The use of low-dose aspirin for the primary prevention of cardiovascular disease (CVD) morbidity and mortality continues to be controversial, particularly for older adults. Recently published, robust randomized controlled trials have revealed less cardiovascular benefit from aspirin for primary prevention compared with previous trials; additionally, an increased risk of major bleeding events has been notably more prevalent in older adults.^1-5 These trials have suggested that preventative aspirin use in older adults confers less benefit than other therapies for decreasing atherosclerotic CVD (ASCVD) risk, including blood pressure (BP) control, cholesterol management, and tobacco cessation.^1,6

A recent meta-analysis indicated a composite cardiovascular risk reduction in patients aged 53 to 74 years taking aspirin vs no aspirin; however, this benefit was offset with an even greater increased risk of major bleeding.⁷ This trend was consistent regardless of stratification by 10-year ASCVD risk or presence of diabetes mellitus (DM) diagnosis.^7,8 Additionally, the recently published Aspirin in Reducing Events in the Elderly (ASPREE) trial studied the impacts of aspirin use in healthy adults aged ≥ 70 years and aged ≥ 65 years among Black and Hispanic adults.⁴ The study concluded that the risk of major bleeding with aspirin use was even higher vs the potential cardiovascular benefit in older adults.⁴

With this emerging evidence, guidelines have been updated to represent the need for risk vs benefit considerations regarding aspirin use for primary prevention in older adults.^1,9,10 The most recent guideline update from the American College of Cardiology and American Heart Association (ACC/AHA) recommends against the routine use of aspirin in patients aged > 70 years or those with bleeding risk factors.¹ The guideline recommends considering aspirin use for patients ages 40 to 70 years only after a patient-specific risk vs benefit discussion.¹ Furthermore, the 2020 American Diabetes Association guideline recommends considering aspirin use for primary prevention in adults with DM between ages 50 and 70 only after a risk vs benefit discussion of patient-specific bleeding risk factors and ASCVD risk-enhancing factors.¹⁰

Despite the demonstrated risks for bleeding with the routine use of aspirin, studies indicate that aspirin continues to be used commonly among older adults, often when unnecessary. In the 2017 National Health Interview Survey, about 23% of adults aged > 40 years in the United States without CVD used aspirin daily, and 23% of these did so without recommendation from a health care professional.¹¹ Furthermore, nearly half of adults ages ≥ 70 years and nearly one-quarter of adults with a history of peptic ulcer disease used aspirin daily.¹¹ Although the most recent guidelines from the ACC/AHA do not recommend a 10-year ASCVD risk threshold for therapy, one study illustrated that 12% of older adult patients were inappropriately prescribed aspirin for primary prevention despite a 10-year ASCVD risk of < 6%.^1,12 These studies highlight the large proportion of individuals, particularly older adults, who may be inappropriately taking aspirin for primary prevention.

Deprescribing Program

Deprescribing potentially inappropriate medications (PIMs) is particularly important in the older adult population, as these individuals experience a high risk of adverse effects (AEs), polypharmacy, cognitive decline, and falls related to medication use.^6,13-17 Evidence suggests that mortality outcomes are improved with the implementation of targeted deprescribing efforts based on patient-specific factors.¹⁸ Additionally, deprescribing unnecessary medications may improve adherence to other essential medications and reduce financial burdens.¹⁹ Pharmacists play a crucial role among health care professionals in the implementation of deprescribing practices, and studies have shown that physicians are highly accepting of pharmacists’ deprescribing recommendations.^13,20-22

Despite the evidence for the benefits of deprescribing, limited data are available regarding the impact and feasibility of a targeted aspirin deprescribing approach by nonphysician practitioners.²³ The objective of this study was to implement and evaluate the success of a pharmacist-driven aspirin deprescribing protocol for older adults in a primary care setting.

This aspirin deprescribing protocol was developed by ambulatory care clinical pharmacist or clinical pharmacist practitioners (CPPs), at the William S. Middleton Memorial Veterans Hospital in Madison, Wisconsin. Within the US Department of Veterans Affairs (VA) health care system, CPPs work under a broad scope of practice with the ability to independently prescribe and monitor medications. The protocol was reviewed by physician stakeholders in both primary care and cardiology and a list was generated, including patients from 2 primary care panels aged ≥ 70 years with aspirin on their medication list, either as a prescription or over-the-counter medication, using the VA Information System Technology and Architecture. A CPP or supervised pharmacy intern identified patients from this list who were appropriate for risk/benefit discussions regarding the discontinuation of aspirin. Patients were excluded from the intervention if they had a history of clinical ASCVD, including myocardial infarction (MI), stable or unstable angina, coronary artery disease (CAD), coronary or other arterial revascularization, cerebrovascular accident (CVA), transient ischemic accident (TIA), or peripheral artery disease (PAD), or another documented indication for aspirin use, including pain, flushing (with niacin use), venous thromboembolism prophylaxis, valvular heart disease, or acute or recurrent pericarditis.

After identifying eligible patients, a CPP or pharmacy intern contacted patients by telephone, following a script to guide conversation. All patients were screened for potential appropriate aspirin indications, particularly any history of MI, CAD, CVA, TIA, PAD, or other clinical ASCVD. The patient was asked about their rationale for taking aspirin and patient-specific ASCVD risk-enhancing factors and bleeding risk factors and educated them on lifestyle modalities to reduce ASCVD risk, using the script as a guide. ASCVD risk-enhancing factors included family history of premature MI, inability to achieve BP goal, DM with the inability to achieve blood glucose or hemoglobin A_1c goal, tobacco use, or inadequate statin therapy. Bleeding risk factors included a history of gastrointestinal bleed or peptic ulcer disease, concurrent use of medications that increase bleeding risk, chronic kidney disease, or thrombocytopenia.

Through shared decision making with careful consideration of these factors, we reached a conclusion with each patient to either continue or to deprescribe aspirin. Each discussion was documented in the electronic health record (EHR) using a standard documentation template (eAppendix, available at doi:10.12788/fp.0320). The patient’s medication list also was updated to reflect changes in aspirin use. For patients who declined deprescribing, the CPP or pharmacy intern asked the patient for their primary reason for preferring to continue aspirin, which was subsequently categorized as one of the following: no prior concerns with bleeding, concerns about a future cardiovascular event, wishing to discuss further with their primary care practitioner (PCP), or identifying an appropriate use for aspirin not evident through record review. For the patients who wished to further discuss the issue with their PCP before deprescribing, the patient’s PCP was notified of this preference by a record alert to the note documenting the encounter, and the patient was also encouraged to follow up about this issue. A voicemail was left if the patient did not answer requesting a call back, and a second attempt was made within 2 weeks.

Data Collected

We collected data to assess the proportion of patients for whom aspirin for primary prevention was discontinued. For patients who declined deprescribing, we documented the rationale for continuing aspirin. Additionally, the feasibility of implementation was assessed, including pharmacist time spent on each record review and intervention. Descriptive statistics were generated to evaluate baseline characteristics and intervention outcomes. The time to completion of these tasks was summarized with descriptive statistics.

We reviewed 459 patient records, and 110 were determined eligible for risk/benefit discussions.

Results of the time-study analysis for each intervention indicated that a pharmacy intern or pharmacist spent about 2 minutes reviewing the record of each patient to determine eligibility for risk/benefit discussions. The 110 patients identified as eligible were 24% of the 459 records reviewed. An average (range) of 12 (6-20) minutes was spent on the telephone call plus documentation for each patient contacted. Additionally, we estimated that CPPs and pharmacy interns spent an approximate combined 12 hours in the development and review of materials for this program, including the protocol, script, and documentation templates. This also included about 1 hour to identify appropriate parameters for, and generate, the eligible patient list.

Discussion

The implementation of a pharmacist-driven aspirin deprescribing protocol for older adults in a primary care setting led to the discontinuation of inappropriate aspirin use in nearly half of older adults contacted. Furthermore, opportunities were identified to update medication lists to reflect previously self-discontinued aspirin for older adults. Just over one-quarter of those contacted declined to discontinue or reduce their aspirin dose. It is hypothesized that with these targeted deprescribing interventions, overall risk reduction for bleeding and polypharmacy will be observed for older adults.¹

In addition to deprescribing aspirin, CPPs used shared decision making to initiate risk/benefit discussions and to educate on targeted lifestyle modifications to lower ASCVD risk. While not all patients agreed to discontinue aspirin, all were provided education that may empower them to engage in future discussions with PCPs regarding appropriate aspirin use. Previous pharmacist-led deprescribing initiatives for proton pump inhibitors and other PIMs have indicated that a large percentage of patients who opt to further discuss a deprescribing concern with their PCPs ultimately resulted in deprescribing outcomes.^24,25 Additionally, a recent trial examining pharmacist-led deprescribing of 4 common PIMs in older adults compared the impact of pharmacists leading educational interventions directly to patients with pharmacists making deprescribing recommendations to physicians. Deprescribing was more successful when patients were involved in the decision-making process.²⁶

Limitations

Although this quality improvement initiative resulted in the deprescribing of inappropriate aspirin for many older adults, a limitation is the small sample size within a single institution. The population of male veterans also may limit generalizability to nonmale and nonveteran older adults. As the protocol was initiated within a limited number of primary care teams initially, future implementation into additional primary care teams will increase the number of older adults impacted by risk/benefit discussions regarding aspirin use. This work may not be generalizable to other health care systems. Many patients within the VA receive both their primary and specialty care within the system, which facilitates communication and collaboration between primary and specialty practitioners. The protocol may require workflow adjustments for patients receiving care within multiple systems. Additionally, although the deprescribing protocol was created in collaboration with physicians, CPPs within the VA work under a broad scope of practice that includes independent medication prescribing, deprescribing, and monitoring. This may be a consideration when implementing similar protocols at other sites, as collaborative practice agreements may need to be in place.

Future Directions

The time required to complete these interventions was generally feasible, though this intervention would require some workflow alteration to be incorporated routinely into a CPP’s schedule. The telephone calls were completed as isolated interventions and were not incorporated into existing scheduled primary care appointments. In the future, the aspirin deprescribing protocol could be incorporated into existing pharmacist-led primary care appointments. Based on the outcomes of this study, CPPs are leading an initiative to develop an aspirin deprescribing clinical reminder tool, which may be quickly inserted into a progress note within the EHR and may be incorporated into any primary care visit led by a CPP or PCP.

Conclusions

This study demonstrates that a pharmacist-led aspirin deprescribing protocol in the ambulatory care pharmacy setting was successful in the discontinuation of unnecessary aspirin use in older adults. The protocol also provided opportunities for education on ASCVD risk reduction in all older adults reached. These findings highlight the role of pharmacists in deprescribing PIMs for older adults and identifying opportunities to further streamline risk/benefit discussions on aspirin deprescribing potential within primary care visits.

The use of low-dose aspirin for the primary prevention of cardiovascular disease (CVD) morbidity and mortality continues to be controversial, particularly for older adults. Recently published, robust randomized controlled trials have revealed less cardiovascular benefit from aspirin for primary prevention compared with previous trials; additionally, an increased risk of major bleeding events has been notably more prevalent in older adults.^1-5 These trials have suggested that preventative aspirin use in older adults confers less benefit than other therapies for decreasing atherosclerotic CVD (ASCVD) risk, including blood pressure (BP) control, cholesterol management, and tobacco cessation.^1,6

A recent meta-analysis indicated a composite cardiovascular risk reduction in patients aged 53 to 74 years taking aspirin vs no aspirin; however, this benefit was offset with an even greater increased risk of major bleeding.⁷ This trend was consistent regardless of stratification by 10-year ASCVD risk or presence of diabetes mellitus (DM) diagnosis.^7,8 Additionally, the recently published Aspirin in Reducing Events in the Elderly (ASPREE) trial studied the impacts of aspirin use in healthy adults aged ≥ 70 years and aged ≥ 65 years among Black and Hispanic adults.⁴ The study concluded that the risk of major bleeding with aspirin use was even higher vs the potential cardiovascular benefit in older adults.⁴

With this emerging evidence, guidelines have been updated to represent the need for risk vs benefit considerations regarding aspirin use for primary prevention in older adults.^1,9,10 The most recent guideline update from the American College of Cardiology and American Heart Association (ACC/AHA) recommends against the routine use of aspirin in patients aged > 70 years or those with bleeding risk factors.¹ The guideline recommends considering aspirin use for patients ages 40 to 70 years only after a patient-specific risk vs benefit discussion.¹ Furthermore, the 2020 American Diabetes Association guideline recommends considering aspirin use for primary prevention in adults with DM between ages 50 and 70 only after a risk vs benefit discussion of patient-specific bleeding risk factors and ASCVD risk-enhancing factors.¹⁰

Despite the demonstrated risks for bleeding with the routine use of aspirin, studies indicate that aspirin continues to be used commonly among older adults, often when unnecessary. In the 2017 National Health Interview Survey, about 23% of adults aged > 40 years in the United States without CVD used aspirin daily, and 23% of these did so without recommendation from a health care professional.¹¹ Furthermore, nearly half of adults ages ≥ 70 years and nearly one-quarter of adults with a history of peptic ulcer disease used aspirin daily.¹¹ Although the most recent guidelines from the ACC/AHA do not recommend a 10-year ASCVD risk threshold for therapy, one study illustrated that 12% of older adult patients were inappropriately prescribed aspirin for primary prevention despite a 10-year ASCVD risk of < 6%.^1,12 These studies highlight the large proportion of individuals, particularly older adults, who may be inappropriately taking aspirin for primary prevention.

Deprescribing Program

Deprescribing potentially inappropriate medications (PIMs) is particularly important in the older adult population, as these individuals experience a high risk of adverse effects (AEs), polypharmacy, cognitive decline, and falls related to medication use.^6,13-17 Evidence suggests that mortality outcomes are improved with the implementation of targeted deprescribing efforts based on patient-specific factors.¹⁸ Additionally, deprescribing unnecessary medications may improve adherence to other essential medications and reduce financial burdens.¹⁹ Pharmacists play a crucial role among health care professionals in the implementation of deprescribing practices, and studies have shown that physicians are highly accepting of pharmacists’ deprescribing recommendations.^13,20-22

Despite the evidence for the benefits of deprescribing, limited data are available regarding the impact and feasibility of a targeted aspirin deprescribing approach by nonphysician practitioners.²³ The objective of this study was to implement and evaluate the success of a pharmacist-driven aspirin deprescribing protocol for older adults in a primary care setting.

This aspirin deprescribing protocol was developed by ambulatory care clinical pharmacist or clinical pharmacist practitioners (CPPs), at the William S. Middleton Memorial Veterans Hospital in Madison, Wisconsin. Within the US Department of Veterans Affairs (VA) health care system, CPPs work under a broad scope of practice with the ability to independently prescribe and monitor medications. The protocol was reviewed by physician stakeholders in both primary care and cardiology and a list was generated, including patients from 2 primary care panels aged ≥ 70 years with aspirin on their medication list, either as a prescription or over-the-counter medication, using the VA Information System Technology and Architecture. A CPP or supervised pharmacy intern identified patients from this list who were appropriate for risk/benefit discussions regarding the discontinuation of aspirin. Patients were excluded from the intervention if they had a history of clinical ASCVD, including myocardial infarction (MI), stable or unstable angina, coronary artery disease (CAD), coronary or other arterial revascularization, cerebrovascular accident (CVA), transient ischemic accident (TIA), or peripheral artery disease (PAD), or another documented indication for aspirin use, including pain, flushing (with niacin use), venous thromboembolism prophylaxis, valvular heart disease, or acute or recurrent pericarditis.

After identifying eligible patients, a CPP or pharmacy intern contacted patients by telephone, following a script to guide conversation. All patients were screened for potential appropriate aspirin indications, particularly any history of MI, CAD, CVA, TIA, PAD, or other clinical ASCVD. The patient was asked about their rationale for taking aspirin and patient-specific ASCVD risk-enhancing factors and bleeding risk factors and educated them on lifestyle modalities to reduce ASCVD risk, using the script as a guide. ASCVD risk-enhancing factors included family history of premature MI, inability to achieve BP goal, DM with the inability to achieve blood glucose or hemoglobin A_1c goal, tobacco use, or inadequate statin therapy. Bleeding risk factors included a history of gastrointestinal bleed or peptic ulcer disease, concurrent use of medications that increase bleeding risk, chronic kidney disease, or thrombocytopenia.

Through shared decision making with careful consideration of these factors, we reached a conclusion with each patient to either continue or to deprescribe aspirin. Each discussion was documented in the electronic health record (EHR) using a standard documentation template (eAppendix, available at doi:10.12788/fp.0320). The patient’s medication list also was updated to reflect changes in aspirin use. For patients who declined deprescribing, the CPP or pharmacy intern asked the patient for their primary reason for preferring to continue aspirin, which was subsequently categorized as one of the following: no prior concerns with bleeding, concerns about a future cardiovascular event, wishing to discuss further with their primary care practitioner (PCP), or identifying an appropriate use for aspirin not evident through record review. For the patients who wished to further discuss the issue with their PCP before deprescribing, the patient’s PCP was notified of this preference by a record alert to the note documenting the encounter, and the patient was also encouraged to follow up about this issue. A voicemail was left if the patient did not answer requesting a call back, and a second attempt was made within 2 weeks.

Data Collected

We collected data to assess the proportion of patients for whom aspirin for primary prevention was discontinued. For patients who declined deprescribing, we documented the rationale for continuing aspirin. Additionally, the feasibility of implementation was assessed, including pharmacist time spent on each record review and intervention. Descriptive statistics were generated to evaluate baseline characteristics and intervention outcomes. The time to completion of these tasks was summarized with descriptive statistics.

We reviewed 459 patient records, and 110 were determined eligible for risk/benefit discussions.

Results of the time-study analysis for each intervention indicated that a pharmacy intern or pharmacist spent about 2 minutes reviewing the record of each patient to determine eligibility for risk/benefit discussions. The 110 patients identified as eligible were 24% of the 459 records reviewed. An average (range) of 12 (6-20) minutes was spent on the telephone call plus documentation for each patient contacted. Additionally, we estimated that CPPs and pharmacy interns spent an approximate combined 12 hours in the development and review of materials for this program, including the protocol, script, and documentation templates. This also included about 1 hour to identify appropriate parameters for, and generate, the eligible patient list.

Discussion

The implementation of a pharmacist-driven aspirin deprescribing protocol for older adults in a primary care setting led to the discontinuation of inappropriate aspirin use in nearly half of older adults contacted. Furthermore, opportunities were identified to update medication lists to reflect previously self-discontinued aspirin for older adults. Just over one-quarter of those contacted declined to discontinue or reduce their aspirin dose. It is hypothesized that with these targeted deprescribing interventions, overall risk reduction for bleeding and polypharmacy will be observed for older adults.¹

In addition to deprescribing aspirin, CPPs used shared decision making to initiate risk/benefit discussions and to educate on targeted lifestyle modifications to lower ASCVD risk. While not all patients agreed to discontinue aspirin, all were provided education that may empower them to engage in future discussions with PCPs regarding appropriate aspirin use. Previous pharmacist-led deprescribing initiatives for proton pump inhibitors and other PIMs have indicated that a large percentage of patients who opt to further discuss a deprescribing concern with their PCPs ultimately resulted in deprescribing outcomes.^24,25 Additionally, a recent trial examining pharmacist-led deprescribing of 4 common PIMs in older adults compared the impact of pharmacists leading educational interventions directly to patients with pharmacists making deprescribing recommendations to physicians. Deprescribing was more successful when patients were involved in the decision-making process.²⁶

Limitations

Although this quality improvement initiative resulted in the deprescribing of inappropriate aspirin for many older adults, a limitation is the small sample size within a single institution. The population of male veterans also may limit generalizability to nonmale and nonveteran older adults. As the protocol was initiated within a limited number of primary care teams initially, future implementation into additional primary care teams will increase the number of older adults impacted by risk/benefit discussions regarding aspirin use. This work may not be generalizable to other health care systems. Many patients within the VA receive both their primary and specialty care within the system, which facilitates communication and collaboration between primary and specialty practitioners. The protocol may require workflow adjustments for patients receiving care within multiple systems. Additionally, although the deprescribing protocol was created in collaboration with physicians, CPPs within the VA work under a broad scope of practice that includes independent medication prescribing, deprescribing, and monitoring. This may be a consideration when implementing similar protocols at other sites, as collaborative practice agreements may need to be in place.

Future Directions

The time required to complete these interventions was generally feasible, though this intervention would require some workflow alteration to be incorporated routinely into a CPP’s schedule. The telephone calls were completed as isolated interventions and were not incorporated into existing scheduled primary care appointments. In the future, the aspirin deprescribing protocol could be incorporated into existing pharmacist-led primary care appointments. Based on the outcomes of this study, CPPs are leading an initiative to develop an aspirin deprescribing clinical reminder tool, which may be quickly inserted into a progress note within the EHR and may be incorporated into any primary care visit led by a CPP or PCP.

Conclusions

This study demonstrates that a pharmacist-led aspirin deprescribing protocol in the ambulatory care pharmacy setting was successful in the discontinuation of unnecessary aspirin use in older adults. The protocol also provided opportunities for education on ASCVD risk reduction in all older adults reached. These findings highlight the role of pharmacists in deprescribing PIMs for older adults and identifying opportunities to further streamline risk/benefit discussions on aspirin deprescribing potential within primary care visits.

The use of low-dose aspirin for the primary prevention of cardiovascular disease (CVD) morbidity and mortality continues to be controversial, particularly for older adults. Recently published, robust randomized controlled trials have revealed less cardiovascular benefit from aspirin for primary prevention compared with previous trials; additionally, an increased risk of major bleeding events has been notably more prevalent in older adults.^1-5 These trials have suggested that preventative aspirin use in older adults confers less benefit than other therapies for decreasing atherosclerotic CVD (ASCVD) risk, including blood pressure (BP) control, cholesterol management, and tobacco cessation.^1,6

A recent meta-analysis indicated a composite cardiovascular risk reduction in patients aged 53 to 74 years taking aspirin vs no aspirin; however, this benefit was offset with an even greater increased risk of major bleeding.⁷ This trend was consistent regardless of stratification by 10-year ASCVD risk or presence of diabetes mellitus (DM) diagnosis.^7,8 Additionally, the recently published Aspirin in Reducing Events in the Elderly (ASPREE) trial studied the impacts of aspirin use in healthy adults aged ≥ 70 years and aged ≥ 65 years among Black and Hispanic adults.⁴ The study concluded that the risk of major bleeding with aspirin use was even higher vs the potential cardiovascular benefit in older adults.⁴

With this emerging evidence, guidelines have been updated to represent the need for risk vs benefit considerations regarding aspirin use for primary prevention in older adults.^1,9,10 The most recent guideline update from the American College of Cardiology and American Heart Association (ACC/AHA) recommends against the routine use of aspirin in patients aged > 70 years or those with bleeding risk factors.¹ The guideline recommends considering aspirin use for patients ages 40 to 70 years only after a patient-specific risk vs benefit discussion.¹ Furthermore, the 2020 American Diabetes Association guideline recommends considering aspirin use for primary prevention in adults with DM between ages 50 and 70 only after a risk vs benefit discussion of patient-specific bleeding risk factors and ASCVD risk-enhancing factors.¹⁰

Despite the demonstrated risks for bleeding with the routine use of aspirin, studies indicate that aspirin continues to be used commonly among older adults, often when unnecessary. In the 2017 National Health Interview Survey, about 23% of adults aged > 40 years in the United States without CVD used aspirin daily, and 23% of these did so without recommendation from a health care professional.¹¹ Furthermore, nearly half of adults ages ≥ 70 years and nearly one-quarter of adults with a history of peptic ulcer disease used aspirin daily.¹¹ Although the most recent guidelines from the ACC/AHA do not recommend a 10-year ASCVD risk threshold for therapy, one study illustrated that 12% of older adult patients were inappropriately prescribed aspirin for primary prevention despite a 10-year ASCVD risk of < 6%.^1,12 These studies highlight the large proportion of individuals, particularly older adults, who may be inappropriately taking aspirin for primary prevention.

Deprescribing Program

Deprescribing potentially inappropriate medications (PIMs) is particularly important in the older adult population, as these individuals experience a high risk of adverse effects (AEs), polypharmacy, cognitive decline, and falls related to medication use.^6,13-17 Evidence suggests that mortality outcomes are improved with the implementation of targeted deprescribing efforts based on patient-specific factors.¹⁸ Additionally, deprescribing unnecessary medications may improve adherence to other essential medications and reduce financial burdens.¹⁹ Pharmacists play a crucial role among health care professionals in the implementation of deprescribing practices, and studies have shown that physicians are highly accepting of pharmacists’ deprescribing recommendations.^13,20-22

Despite the evidence for the benefits of deprescribing, limited data are available regarding the impact and feasibility of a targeted aspirin deprescribing approach by nonphysician practitioners.²³ The objective of this study was to implement and evaluate the success of a pharmacist-driven aspirin deprescribing protocol for older adults in a primary care setting.

This aspirin deprescribing protocol was developed by ambulatory care clinical pharmacist or clinical pharmacist practitioners (CPPs), at the William S. Middleton Memorial Veterans Hospital in Madison, Wisconsin. Within the US Department of Veterans Affairs (VA) health care system, CPPs work under a broad scope of practice with the ability to independently prescribe and monitor medications. The protocol was reviewed by physician stakeholders in both primary care and cardiology and a list was generated, including patients from 2 primary care panels aged ≥ 70 years with aspirin on their medication list, either as a prescription or over-the-counter medication, using the VA Information System Technology and Architecture. A CPP or supervised pharmacy intern identified patients from this list who were appropriate for risk/benefit discussions regarding the discontinuation of aspirin. Patients were excluded from the intervention if they had a history of clinical ASCVD, including myocardial infarction (MI), stable or unstable angina, coronary artery disease (CAD), coronary or other arterial revascularization, cerebrovascular accident (CVA), transient ischemic accident (TIA), or peripheral artery disease (PAD), or another documented indication for aspirin use, including pain, flushing (with niacin use), venous thromboembolism prophylaxis, valvular heart disease, or acute or recurrent pericarditis.

After identifying eligible patients, a CPP or pharmacy intern contacted patients by telephone, following a script to guide conversation. All patients were screened for potential appropriate aspirin indications, particularly any history of MI, CAD, CVA, TIA, PAD, or other clinical ASCVD. The patient was asked about their rationale for taking aspirin and patient-specific ASCVD risk-enhancing factors and bleeding risk factors and educated them on lifestyle modalities to reduce ASCVD risk, using the script as a guide. ASCVD risk-enhancing factors included family history of premature MI, inability to achieve BP goal, DM with the inability to achieve blood glucose or hemoglobin A_1c goal, tobacco use, or inadequate statin therapy. Bleeding risk factors included a history of gastrointestinal bleed or peptic ulcer disease, concurrent use of medications that increase bleeding risk, chronic kidney disease, or thrombocytopenia.

Through shared decision making with careful consideration of these factors, we reached a conclusion with each patient to either continue or to deprescribe aspirin. Each discussion was documented in the electronic health record (EHR) using a standard documentation template (eAppendix, available at doi:10.12788/fp.0320). The patient’s medication list also was updated to reflect changes in aspirin use. For patients who declined deprescribing, the CPP or pharmacy intern asked the patient for their primary reason for preferring to continue aspirin, which was subsequently categorized as one of the following: no prior concerns with bleeding, concerns about a future cardiovascular event, wishing to discuss further with their primary care practitioner (PCP), or identifying an appropriate use for aspirin not evident through record review. For the patients who wished to further discuss the issue with their PCP before deprescribing, the patient’s PCP was notified of this preference by a record alert to the note documenting the encounter, and the patient was also encouraged to follow up about this issue. A voicemail was left if the patient did not answer requesting a call back, and a second attempt was made within 2 weeks.

Data Collected

We collected data to assess the proportion of patients for whom aspirin for primary prevention was discontinued. For patients who declined deprescribing, we documented the rationale for continuing aspirin. Additionally, the feasibility of implementation was assessed, including pharmacist time spent on each record review and intervention. Descriptive statistics were generated to evaluate baseline characteristics and intervention outcomes. The time to completion of these tasks was summarized with descriptive statistics.

We reviewed 459 patient records, and 110 were determined eligible for risk/benefit discussions.

Results of the time-study analysis for each intervention indicated that a pharmacy intern or pharmacist spent about 2 minutes reviewing the record of each patient to determine eligibility for risk/benefit discussions. The 110 patients identified as eligible were 24% of the 459 records reviewed. An average (range) of 12 (6-20) minutes was spent on the telephone call plus documentation for each patient contacted. Additionally, we estimated that CPPs and pharmacy interns spent an approximate combined 12 hours in the development and review of materials for this program, including the protocol, script, and documentation templates. This also included about 1 hour to identify appropriate parameters for, and generate, the eligible patient list.

Discussion

The implementation of a pharmacist-driven aspirin deprescribing protocol for older adults in a primary care setting led to the discontinuation of inappropriate aspirin use in nearly half of older adults contacted. Furthermore, opportunities were identified to update medication lists to reflect previously self-discontinued aspirin for older adults. Just over one-quarter of those contacted declined to discontinue or reduce their aspirin dose. It is hypothesized that with these targeted deprescribing interventions, overall risk reduction for bleeding and polypharmacy will be observed for older adults.¹

In addition to deprescribing aspirin, CPPs used shared decision making to initiate risk/benefit discussions and to educate on targeted lifestyle modifications to lower ASCVD risk. While not all patients agreed to discontinue aspirin, all were provided education that may empower them to engage in future discussions with PCPs regarding appropriate aspirin use. Previous pharmacist-led deprescribing initiatives for proton pump inhibitors and other PIMs have indicated that a large percentage of patients who opt to further discuss a deprescribing concern with their PCPs ultimately resulted in deprescribing outcomes.^24,25 Additionally, a recent trial examining pharmacist-led deprescribing of 4 common PIMs in older adults compared the impact of pharmacists leading educational interventions directly to patients with pharmacists making deprescribing recommendations to physicians. Deprescribing was more successful when patients were involved in the decision-making process.²⁶

Limitations

Although this quality improvement initiative resulted in the deprescribing of inappropriate aspirin for many older adults, a limitation is the small sample size within a single institution. The population of male veterans also may limit generalizability to nonmale and nonveteran older adults. As the protocol was initiated within a limited number of primary care teams initially, future implementation into additional primary care teams will increase the number of older adults impacted by risk/benefit discussions regarding aspirin use. This work may not be generalizable to other health care systems. Many patients within the VA receive both their primary and specialty care within the system, which facilitates communication and collaboration between primary and specialty practitioners. The protocol may require workflow adjustments for patients receiving care within multiple systems. Additionally, although the deprescribing protocol was created in collaboration with physicians, CPPs within the VA work under a broad scope of practice that includes independent medication prescribing, deprescribing, and monitoring. This may be a consideration when implementing similar protocols at other sites, as collaborative practice agreements may need to be in place.

Future Directions

The time required to complete these interventions was generally feasible, though this intervention would require some workflow alteration to be incorporated routinely into a CPP’s schedule. The telephone calls were completed as isolated interventions and were not incorporated into existing scheduled primary care appointments. In the future, the aspirin deprescribing protocol could be incorporated into existing pharmacist-led primary care appointments. Based on the outcomes of this study, CPPs are leading an initiative to develop an aspirin deprescribing clinical reminder tool, which may be quickly inserted into a progress note within the EHR and may be incorporated into any primary care visit led by a CPP or PCP.

Conclusions

This study demonstrates that a pharmacist-led aspirin deprescribing protocol in the ambulatory care pharmacy setting was successful in the discontinuation of unnecessary aspirin use in older adults. The protocol also provided opportunities for education on ASCVD risk reduction in all older adults reached. These findings highlight the role of pharmacists in deprescribing PIMs for older adults and identifying opportunities to further streamline risk/benefit discussions on aspirin deprescribing potential within primary care visits.

In 2019, diabetes mellitus (DM) was the seventh leading cause of death in the United States, and currently, about 11% of the American population has a DM diagnosis.¹ Most have a diagnosis of type 2 diabetes (T2DM), which has a strong genetic predisposition, and the risk of developing T2DM increases with age, obesity, and lack of physical activity.^1,2 Nearly one-quarter of veterans have a diagnosis of DM, and DM is the leading cause of comorbidities, such as blindness, end-stage renal disease, and amputation for patients receiving care from the Veterans Health Administration (VHA).² The elevated incidence of DM in the veteran population is attributed to a variety of factors, including exposure to herbicides, such as Agent Orange, advanced age, increased risk of obesity, and limited access to high-quality food.³

After diagnosis, both the American Diabetes Association (ADA) and the American Association of Clinical Endocrinologists and American College of Endocrinology (AACE/ACE) emphasize the appropriate use of lifestyle management and pharmacologic therapy for DM care. The use of pharmacologic agents (oral medications, insulin, or noninsulin injectables) is often determined by efficacy, cost, potential adverse effects (AEs), and patient factors and comorbidities.^4,5

The initial recommendation for pharmacologic treatment for T2DM differs slightly between expert guidelines. The ADA and AACE/ACE recommend any of the following as initial monotherapy, listed in order to represent a hierarchy of usage: metformin, glucagon-like peptide-1 receptor agonists (GLP-1 RAs), sodium-glucose cotransporter 2 (SGLT-2) inhibitors, or dipeptidyl peptidase-4 (DPP-4) inhibitors, with the first 3 agents carrying the strongest recommendations.^4,5 For patients with established atherosclerotic cardiovascular disease (CVD), chronic kidney disease, or heart failure, it is recommended to start a long-acting GLP-1 RA or SGLT-2 inhibitor. For patients with T2DM and hemoglobin A_1c (HbA_1c) between 7.5% and 9.0% at diagnosis, the AACE/ACE recommend initiation of dual therapy using metformin alongside another first-line agent and recommend the addition of another antidiabetic agent if glycemic goals are not met after regular follow-up. AACE/ACE recommend the consideration of insulin therapy in symptomatic patients with HbA_1c > 9.0%.⁵ In contrast, the ADA recommends metformin as first-line therapy for all patients with T2DM and recommends dual therapy using metformin and another preferred agent (selection based on comorbidities) when HbA_1c is 1.5% to 2% above target. The ADA recommends the consideration of insulin with HbA_1c > 10% or with evidence of ongoing catabolism or symptoms of hyperglycemia.⁴ There are several reasons why insulin may be initiated prior to GLP-1 RAs, including profound hyperglycemia at time of diagnosis or implementation of insulin agents prior to commercial availability of GLP-1 RA.

GLP-1 RAs are analogs of the hormone incretin, which increases glucose-dependent insulin secretion, decreases postprandial glucagon secretion, increases satiety, and slows gastric emptying.^6,7 When used in combination with noninsulin agents, GLP-1 RAs have demonstrated HbA_1c reductions of 0.5% to 1.5%.⁸ The use of GLP-1 RAs with basal insulin also has been studied extensively.^6,8-10 When the combination of GLP-1 RAs and basal insulin was compared with basal/bolus insulin regimens, the use of the GLP-1 RAs resulted in lower HbA_1c levels and lower incidence of hypoglycemia.^6,9 Data have demonstrated the complementary mechanisms of using basal insulin and GLP 1 RAs in decreasing HbA_1c levels, insulin requirements, and weight compared with using basal insulin monotherapy and basal/bolus combinations.^6,9-13 Moreover, 3 GLP-1 RA medications currently on the market (liraglutide, dulaglutide, and semaglutide) have displayed cardiovascular and renal benefits, further supporting the use of these medications.^2,5

Despite these benefits, GLP-1 RAs may have bothersome AEs and are associated with a high cost.⁶ In addition, some studies have found that as the length of therapy increases, the positive effects of these agents may diminish.^9,11 In one study, which looked at the impact of the addition of exenatide to patients taking basal or basal/bolus insulin regimens, mean changes in weight were −2.4 kg at 0 to 6 months, −4.3 kg at 6 to 12 months, −6.2 kg at 12 to 18 months, and −5.5 kg at 18 to 27 months. After 18 months, an increase in weight was observed, but the increase remained lower than baseline.¹¹ Another study, conducted over 12 months, found no significant decrease in weight or total daily dose (TDD) of insulin when exenatide or liraglutide were added to various insulin regimens (basal or basal/bolus).¹³ To date, minimal published data exist regarding the addition of newer GLP-1 RAs and the long-term use of these agents beyond 12 months in patients taking basal/bolus insulin regimens. The primary goal of this study was to evaluate the effect of adding GLP-1 RAs to basal/bolus insulin regimens over a 24-month period.

Methods

This study was a retrospective, electronic health record review of all patients on basal and bolus insulin regimens who received additional therapy with a GLP-1 RA at Veteran Health Indiana in Indianapolis from September 1, 2015, to June 30, 2019. Patients meeting inclusion criteria served as their own control. The primary outcome was change in HbA_1c at 3, 6, 12, 18, and 24 months after initiation of the GLP-1 RA. Secondary outcomes included change in weight and TDD of insulin at 3, 6, 12, 18, and 24 months after the initiation of the GLP-1 RAs and incidence of patient-reported or laboratory-confirmed hypoglycemia and other AEs.

Patients were included if they were aged ≥ 18 years with a diagnosis of T2DM, had concomitant prescriptions for both a basal insulin (glargine, detemir, or NPH) and a bolus insulin (aspart, lispro, or regular) before receiving add-on therapy with a GLP-1 RA (exenatide, liraglutide, albiglutide, lixisenatide, dulaglutide, or semaglutide) from September 1, 2015, to June 30, 2019, and had baseline and subsequent HbA_1c measurements available in the electronic health record. Patients were excluded if they had a diagnosis of type 1 DM (T1DM), were followed by an outside clinician for DM care, or if the GLP-1 RA was discontinued before subsequent HbA_1c measurement. The study protocol was approved by the Research and Development Office of Veteran Health Indiana, and the project was deemed exempt from review by the Indiana University Institutional Review Board due to the retrospective nature of the study.

Data analysis was performed using Excel. Change from baseline for each interval was computed, and 1 sample t tests (2-tailed) compared change from baseline to no change. Due to the disparity in the number of patients with data available at each of the time intervals, a mean plot was presented for each group of patients within each interval, allowing mean changes in individual groups to be observed over time.

Results

One hundred twenty-three subjects met inclusion criteria; 16 patients were excluded due to GLP-1 RA discontinuation before follow-up measurement of HbA_1c; 14 were excluded due to patients being managed by a clinician outside of the facility; 1 patient was excluded for lack of documentation regarding baseline and subsequent insulin doses. Ninety-two patient charts were reviewed. Participants had a mean age of 64 years, 95% were male, and 89% were White. Mean baseline HbA_1c was 9.2%, mean body mass index was 38.9, and the mean TDD of insulin was 184 units.

Since some patients switched between GLP-1 RAs throughout the study and there was variation in timing of laboratory and clinic follow-up,

Discussion

Adding a GLP-1 RA to basal/bolus insulin regimens was associated with a statistically significant decrease in HbA_1c at each time point through 18 months. The greatest improvement in glycemic control from baseline was seen at 3 months, with improvements in HbA_1c diminishing at each subsequent period. The study also demonstrated a significant decrease in weight at each time point through 18 months. The greatest decrease in weight was observed at both 6 and 12 months. Statistically significant decreases in TDD were observed at 3, 6, and 12 months. Insulin changes after 12 months were not found to be statistically significant.

Few studies have previously evaluated the use of GLP-1 RAs in patients with T2DM who are already taking basal/bolus insulin regimens. Gyorffy and colleagues reported significant improvements in glycemic control at 3 and 6 months in a sample of 54 patients taking basal/bolus insulin when liraglutide or exenatide was added, although statistical significance was not found at the final 12-month time point.¹³ That study also found a significant decrease in weight at 6 months; however there was not a significant reduction in weight at both 3 and 12 months of GLP-1 RA therapy. There was not a significant decrease in TDD at any of the collected time points. Nonetheless, Gyorffy and colleagues concluded that reduction in TDD leveled off after 12 months, which is consistent with this study’s findings. The small size of the study may have limited the ability to detect statistical significance; however, this study was conducted in a population that was racially diverse and included a higher proportion of women, though average age was similar.¹³

Yoon and colleagues reported weight loss through 18 months, then saw weight increase, though weights did remain lower than baseline. The study also showed no significant change in TDD of insulin after 12 months of concomitant exenatide and insulin therapy.¹¹ Although these results mirror the outcomes observed in this study, Yoon and colleagues did not differentiate results between basal and basal/bolus insulin groups.¹¹ Seino and colleagues observed no significant change in weight after 36 weeks of GLP-1 RA therapy in Japanese patients when used with basal and basal/bolus insulin regimens. Despite the consideration that the population in the study was not overweight (mean body mass index was 25.6), the results of these studies support the idea that effects of GLP-1 RAs on weight and TDD may diminish over time.¹⁴

Within the VHA, GLP-1 RAs are nonformulary medications. Patients must meet certain criteria in order to be approved for these agents, which may include diagnosis of CVD, renal disease, or failure to reach glycemic control with the use of oral agents or insulin. Therefore, participants of this study represent a particular subset of VHA patients, many of whom may have been selected for consideration due to long-standing or uncontrolled T2DM and failure of previous therapies. The baseline demographics support this idea, given poor glycemic control at baseline and high insulin requirements. Once approved for GLP-1 RA therapy, semaglutide is currently the preferred agent within the VHA, with other agents available for select considerations. It should be noted that albiglutide, which was the primary agent selected for some of the patients included in this study, was removed from the market in 2017 for economic considerations.¹⁵ In the case for these patients, a conversion to a formulary-preferred GLP-1 RA was made.

Most of the patients included in this study (70%) were maintained on metformin from baseline throughout the study period. Fifty-seven percent of patients were taking TDD of insulin > 150 units. Considering the significant cost of concentrated insulins, the addition of GLP-1 RAs to standard insulin may prove to be beneficial from a cost standpoint. Additional research in this area may be warranted to establish more data regarding this potential benefit of GLP-1 RAs as add-on therapy.

Many adverse drug reactions were reported at different periods; however, most of these were associated with the gastrointestinal system, which is consistent with current literature, drug labeling, and the mechanism of action.¹⁶ Hypoglycemia occurred in about one-third of the participants; however, it should be noted that alone, GLP-1 RAs are not associated with a high risk of hypoglycemia. Previous studies have found that GLP-1 RA monotherapy is associated with hypoglycemia in 1.6% to 12.6% of patients.^17,18 More likely, the combination of basal/bolus insulin and the GLP-1 RA’s effect on increasing insulin sensitivity through weight loss, improving glucose-dependent insulin secretion, or by decreasing appetite and therefore decreasing carbohydrate intake contributed to the hypoglycemia prevalence.

Limitations and Strengths

Limitations of this study include a small patient population and a gradual reduction in available data as time periods progressed, making even smaller sample sizes for subsequent time periods. A majority of participants were older, males and White race. This could have limited the determination of statistical significance and applicability of the results to other patient populations. Another potential limitation was the retrospective nature of the study design, which may have limited reporting of hypoglycemia and other AEs based on the documentation of the clinician.

Strengths included the study duration and the diversity of GLP-1 RAs used by participants, as the impact of many of these agents has not yet been assessed in the literature. In addition, the retrospective nature of the study allows for a more realistic representation of patient adherence, education, and motivation, which are likely different from those of patients included in prospective clinical trials.

There are no clear guidelines dictating the optimal duration of concomitant GLP-1 RA and insulin therapy; however, our study suggests that there may be continued benefits past short-term use. Also our study suggests that patients with T2DM treated with basal/bolus insulin regimens may glean additional benefit from adding GLP-1 RAs; however, further randomized, controlled studies are warranted, particularly in poorly controlled patients requiring even more aggressive treatment regimens, such as concentrated insulins.

Conclusions

In our study, adding GLP-1 RA to basal/bolus insulin was associated with a significant decrease in HbA_1c from baseline through 18 months. An overall decrease in weight and TDD of insulin was observed through 24 months, but the change in weight was not significant past 18 months, and the change in insulin requirement was not significant past 12 months. Hypoglycemia was observed in almost one-third of patients, and gastrointestinal symptoms were the most common AE observed as a result of adding GLP-1 RAs. More studies are needed to better evaluate the durability and cost benefit of GLP-1 RAs, especially in patients with high insulin requirements.

Acknowledgments

This material is the result of work supported with resources and facilities at Veteran Health Indiana in Indianapolis. Study data were collected and managed using REDCap electronic data capture tools hosted at Veteran Health Indiana. The authors also acknowledge George Eckert for his assistance with data analysis.

In 2019, diabetes mellitus (DM) was the seventh leading cause of death in the United States, and currently, about 11% of the American population has a DM diagnosis.¹ Most have a diagnosis of type 2 diabetes (T2DM), which has a strong genetic predisposition, and the risk of developing T2DM increases with age, obesity, and lack of physical activity.^1,2 Nearly one-quarter of veterans have a diagnosis of DM, and DM is the leading cause of comorbidities, such as blindness, end-stage renal disease, and amputation for patients receiving care from the Veterans Health Administration (VHA).² The elevated incidence of DM in the veteran population is attributed to a variety of factors, including exposure to herbicides, such as Agent Orange, advanced age, increased risk of obesity, and limited access to high-quality food.³

After diagnosis, both the American Diabetes Association (ADA) and the American Association of Clinical Endocrinologists and American College of Endocrinology (AACE/ACE) emphasize the appropriate use of lifestyle management and pharmacologic therapy for DM care. The use of pharmacologic agents (oral medications, insulin, or noninsulin injectables) is often determined by efficacy, cost, potential adverse effects (AEs), and patient factors and comorbidities.^4,5

The initial recommendation for pharmacologic treatment for T2DM differs slightly between expert guidelines. The ADA and AACE/ACE recommend any of the following as initial monotherapy, listed in order to represent a hierarchy of usage: metformin, glucagon-like peptide-1 receptor agonists (GLP-1 RAs), sodium-glucose cotransporter 2 (SGLT-2) inhibitors, or dipeptidyl peptidase-4 (DPP-4) inhibitors, with the first 3 agents carrying the strongest recommendations.^4,5 For patients with established atherosclerotic cardiovascular disease (CVD), chronic kidney disease, or heart failure, it is recommended to start a long-acting GLP-1 RA or SGLT-2 inhibitor. For patients with T2DM and hemoglobin A_1c (HbA_1c) between 7.5% and 9.0% at diagnosis, the AACE/ACE recommend initiation of dual therapy using metformin alongside another first-line agent and recommend the addition of another antidiabetic agent if glycemic goals are not met after regular follow-up. AACE/ACE recommend the consideration of insulin therapy in symptomatic patients with HbA_1c > 9.0%.⁵ In contrast, the ADA recommends metformin as first-line therapy for all patients with T2DM and recommends dual therapy using metformin and another preferred agent (selection based on comorbidities) when HbA_1c is 1.5% to 2% above target. The ADA recommends the consideration of insulin with HbA_1c > 10% or with evidence of ongoing catabolism or symptoms of hyperglycemia.⁴ There are several reasons why insulin may be initiated prior to GLP-1 RAs, including profound hyperglycemia at time of diagnosis or implementation of insulin agents prior to commercial availability of GLP-1 RA.

GLP-1 RAs are analogs of the hormone incretin, which increases glucose-dependent insulin secretion, decreases postprandial glucagon secretion, increases satiety, and slows gastric emptying.^6,7 When used in combination with noninsulin agents, GLP-1 RAs have demonstrated HbA_1c reductions of 0.5% to 1.5%.⁸ The use of GLP-1 RAs with basal insulin also has been studied extensively.^6,8-10 When the combination of GLP-1 RAs and basal insulin was compared with basal/bolus insulin regimens, the use of the GLP-1 RAs resulted in lower HbA_1c levels and lower incidence of hypoglycemia.^6,9 Data have demonstrated the complementary mechanisms of using basal insulin and GLP 1 RAs in decreasing HbA_1c levels, insulin requirements, and weight compared with using basal insulin monotherapy and basal/bolus combinations.^6,9-13 Moreover, 3 GLP-1 RA medications currently on the market (liraglutide, dulaglutide, and semaglutide) have displayed cardiovascular and renal benefits, further supporting the use of these medications.^2,5

Despite these benefits, GLP-1 RAs may have bothersome AEs and are associated with a high cost.⁶ In addition, some studies have found that as the length of therapy increases, the positive effects of these agents may diminish.^9,11 In one study, which looked at the impact of the addition of exenatide to patients taking basal or basal/bolus insulin regimens, mean changes in weight were −2.4 kg at 0 to 6 months, −4.3 kg at 6 to 12 months, −6.2 kg at 12 to 18 months, and −5.5 kg at 18 to 27 months. After 18 months, an increase in weight was observed, but the increase remained lower than baseline.¹¹ Another study, conducted over 12 months, found no significant decrease in weight or total daily dose (TDD) of insulin when exenatide or liraglutide were added to various insulin regimens (basal or basal/bolus).¹³ To date, minimal published data exist regarding the addition of newer GLP-1 RAs and the long-term use of these agents beyond 12 months in patients taking basal/bolus insulin regimens. The primary goal of this study was to evaluate the effect of adding GLP-1 RAs to basal/bolus insulin regimens over a 24-month period.

Methods

This study was a retrospective, electronic health record review of all patients on basal and bolus insulin regimens who received additional therapy with a GLP-1 RA at Veteran Health Indiana in Indianapolis from September 1, 2015, to June 30, 2019. Patients meeting inclusion criteria served as their own control. The primary outcome was change in HbA_1c at 3, 6, 12, 18, and 24 months after initiation of the GLP-1 RA. Secondary outcomes included change in weight and TDD of insulin at 3, 6, 12, 18, and 24 months after the initiation of the GLP-1 RAs and incidence of patient-reported or laboratory-confirmed hypoglycemia and other AEs.

Patients were included if they were aged ≥ 18 years with a diagnosis of T2DM, had concomitant prescriptions for both a basal insulin (glargine, detemir, or NPH) and a bolus insulin (aspart, lispro, or regular) before receiving add-on therapy with a GLP-1 RA (exenatide, liraglutide, albiglutide, lixisenatide, dulaglutide, or semaglutide) from September 1, 2015, to June 30, 2019, and had baseline and subsequent HbA_1c measurements available in the electronic health record. Patients were excluded if they had a diagnosis of type 1 DM (T1DM), were followed by an outside clinician for DM care, or if the GLP-1 RA was discontinued before subsequent HbA_1c measurement. The study protocol was approved by the Research and Development Office of Veteran Health Indiana, and the project was deemed exempt from review by the Indiana University Institutional Review Board due to the retrospective nature of the study.

Data analysis was performed using Excel. Change from baseline for each interval was computed, and 1 sample t tests (2-tailed) compared change from baseline to no change. Due to the disparity in the number of patients with data available at each of the time intervals, a mean plot was presented for each group of patients within each interval, allowing mean changes in individual groups to be observed over time.

Results

One hundred twenty-three subjects met inclusion criteria; 16 patients were excluded due to GLP-1 RA discontinuation before follow-up measurement of HbA_1c; 14 were excluded due to patients being managed by a clinician outside of the facility; 1 patient was excluded for lack of documentation regarding baseline and subsequent insulin doses. Ninety-two patient charts were reviewed. Participants had a mean age of 64 years, 95% were male, and 89% were White. Mean baseline HbA_1c was 9.2%, mean body mass index was 38.9, and the mean TDD of insulin was 184 units.

Since some patients switched between GLP-1 RAs throughout the study and there was variation in timing of laboratory and clinic follow-up,

Discussion

Adding a GLP-1 RA to basal/bolus insulin regimens was associated with a statistically significant decrease in HbA_1c at each time point through 18 months. The greatest improvement in glycemic control from baseline was seen at 3 months, with improvements in HbA_1c diminishing at each subsequent period. The study also demonstrated a significant decrease in weight at each time point through 18 months. The greatest decrease in weight was observed at both 6 and 12 months. Statistically significant decreases in TDD were observed at 3, 6, and 12 months. Insulin changes after 12 months were not found to be statistically significant.

Few studies have previously evaluated the use of GLP-1 RAs in patients with T2DM who are already taking basal/bolus insulin regimens. Gyorffy and colleagues reported significant improvements in glycemic control at 3 and 6 months in a sample of 54 patients taking basal/bolus insulin when liraglutide or exenatide was added, although statistical significance was not found at the final 12-month time point.¹³ That study also found a significant decrease in weight at 6 months; however there was not a significant reduction in weight at both 3 and 12 months of GLP-1 RA therapy. There was not a significant decrease in TDD at any of the collected time points. Nonetheless, Gyorffy and colleagues concluded that reduction in TDD leveled off after 12 months, which is consistent with this study’s findings. The small size of the study may have limited the ability to detect statistical significance; however, this study was conducted in a population that was racially diverse and included a higher proportion of women, though average age was similar.¹³

Yoon and colleagues reported weight loss through 18 months, then saw weight increase, though weights did remain lower than baseline. The study also showed no significant change in TDD of insulin after 12 months of concomitant exenatide and insulin therapy.¹¹ Although these results mirror the outcomes observed in this study, Yoon and colleagues did not differentiate results between basal and basal/bolus insulin groups.¹¹ Seino and colleagues observed no significant change in weight after 36 weeks of GLP-1 RA therapy in Japanese patients when used with basal and basal/bolus insulin regimens. Despite the consideration that the population in the study was not overweight (mean body mass index was 25.6), the results of these studies support the idea that effects of GLP-1 RAs on weight and TDD may diminish over time.¹⁴

Within the VHA, GLP-1 RAs are nonformulary medications. Patients must meet certain criteria in order to be approved for these agents, which may include diagnosis of CVD, renal disease, or failure to reach glycemic control with the use of oral agents or insulin. Therefore, participants of this study represent a particular subset of VHA patients, many of whom may have been selected for consideration due to long-standing or uncontrolled T2DM and failure of previous therapies. The baseline demographics support this idea, given poor glycemic control at baseline and high insulin requirements. Once approved for GLP-1 RA therapy, semaglutide is currently the preferred agent within the VHA, with other agents available for select considerations. It should be noted that albiglutide, which was the primary agent selected for some of the patients included in this study, was removed from the market in 2017 for economic considerations.¹⁵ In the case for these patients, a conversion to a formulary-preferred GLP-1 RA was made.

Most of the patients included in this study (70%) were maintained on metformin from baseline throughout the study period. Fifty-seven percent of patients were taking TDD of insulin > 150 units. Considering the significant cost of concentrated insulins, the addition of GLP-1 RAs to standard insulin may prove to be beneficial from a cost standpoint. Additional research in this area may be warranted to establish more data regarding this potential benefit of GLP-1 RAs as add-on therapy.

Many adverse drug reactions were reported at different periods; however, most of these were associated with the gastrointestinal system, which is consistent with current literature, drug labeling, and the mechanism of action.¹⁶ Hypoglycemia occurred in about one-third of the participants; however, it should be noted that alone, GLP-1 RAs are not associated with a high risk of hypoglycemia. Previous studies have found that GLP-1 RA monotherapy is associated with hypoglycemia in 1.6% to 12.6% of patients.^17,18 More likely, the combination of basal/bolus insulin and the GLP-1 RA’s effect on increasing insulin sensitivity through weight loss, improving glucose-dependent insulin secretion, or by decreasing appetite and therefore decreasing carbohydrate intake contributed to the hypoglycemia prevalence.

Limitations and Strengths

Limitations of this study include a small patient population and a gradual reduction in available data as time periods progressed, making even smaller sample sizes for subsequent time periods. A majority of participants were older, males and White race. This could have limited the determination of statistical significance and applicability of the results to other patient populations. Another potential limitation was the retrospective nature of the study design, which may have limited reporting of hypoglycemia and other AEs based on the documentation of the clinician.

Strengths included the study duration and the diversity of GLP-1 RAs used by participants, as the impact of many of these agents has not yet been assessed in the literature. In addition, the retrospective nature of the study allows for a more realistic representation of patient adherence, education, and motivation, which are likely different from those of patients included in prospective clinical trials.

There are no clear guidelines dictating the optimal duration of concomitant GLP-1 RA and insulin therapy; however, our study suggests that there may be continued benefits past short-term use. Also our study suggests that patients with T2DM treated with basal/bolus insulin regimens may glean additional benefit from adding GLP-1 RAs; however, further randomized, controlled studies are warranted, particularly in poorly controlled patients requiring even more aggressive treatment regimens, such as concentrated insulins.

Conclusions

In our study, adding GLP-1 RA to basal/bolus insulin was associated with a significant decrease in HbA_1c from baseline through 18 months. An overall decrease in weight and TDD of insulin was observed through 24 months, but the change in weight was not significant past 18 months, and the change in insulin requirement was not significant past 12 months. Hypoglycemia was observed in almost one-third of patients, and gastrointestinal symptoms were the most common AE observed as a result of adding GLP-1 RAs. More studies are needed to better evaluate the durability and cost benefit of GLP-1 RAs, especially in patients with high insulin requirements.

Acknowledgments

This material is the result of work supported with resources and facilities at Veteran Health Indiana in Indianapolis. Study data were collected and managed using REDCap electronic data capture tools hosted at Veteran Health Indiana. The authors also acknowledge George Eckert for his assistance with data analysis.

In 2019, diabetes mellitus (DM) was the seventh leading cause of death in the United States, and currently, about 11% of the American population has a DM diagnosis.¹ Most have a diagnosis of type 2 diabetes (T2DM), which has a strong genetic predisposition, and the risk of developing T2DM increases with age, obesity, and lack of physical activity.^1,2 Nearly one-quarter of veterans have a diagnosis of DM, and DM is the leading cause of comorbidities, such as blindness, end-stage renal disease, and amputation for patients receiving care from the Veterans Health Administration (VHA).² The elevated incidence of DM in the veteran population is attributed to a variety of factors, including exposure to herbicides, such as Agent Orange, advanced age, increased risk of obesity, and limited access to high-quality food.³

After diagnosis, both the American Diabetes Association (ADA) and the American Association of Clinical Endocrinologists and American College of Endocrinology (AACE/ACE) emphasize the appropriate use of lifestyle management and pharmacologic therapy for DM care. The use of pharmacologic agents (oral medications, insulin, or noninsulin injectables) is often determined by efficacy, cost, potential adverse effects (AEs), and patient factors and comorbidities.^4,5

The initial recommendation for pharmacologic treatment for T2DM differs slightly between expert guidelines. The ADA and AACE/ACE recommend any of the following as initial monotherapy, listed in order to represent a hierarchy of usage: metformin, glucagon-like peptide-1 receptor agonists (GLP-1 RAs), sodium-glucose cotransporter 2 (SGLT-2) inhibitors, or dipeptidyl peptidase-4 (DPP-4) inhibitors, with the first 3 agents carrying the strongest recommendations.^4,5 For patients with established atherosclerotic cardiovascular disease (CVD), chronic kidney disease, or heart failure, it is recommended to start a long-acting GLP-1 RA or SGLT-2 inhibitor. For patients with T2DM and hemoglobin A_1c (HbA_1c) between 7.5% and 9.0% at diagnosis, the AACE/ACE recommend initiation of dual therapy using metformin alongside another first-line agent and recommend the addition of another antidiabetic agent if glycemic goals are not met after regular follow-up. AACE/ACE recommend the consideration of insulin therapy in symptomatic patients with HbA_1c > 9.0%.⁵ In contrast, the ADA recommends metformin as first-line therapy for all patients with T2DM and recommends dual therapy using metformin and another preferred agent (selection based on comorbidities) when HbA_1c is 1.5% to 2% above target. The ADA recommends the consideration of insulin with HbA_1c > 10% or with evidence of ongoing catabolism or symptoms of hyperglycemia.⁴ There are several reasons why insulin may be initiated prior to GLP-1 RAs, including profound hyperglycemia at time of diagnosis or implementation of insulin agents prior to commercial availability of GLP-1 RA.

GLP-1 RAs are analogs of the hormone incretin, which increases glucose-dependent insulin secretion, decreases postprandial glucagon secretion, increases satiety, and slows gastric emptying.^6,7 When used in combination with noninsulin agents, GLP-1 RAs have demonstrated HbA_1c reductions of 0.5% to 1.5%.⁸ The use of GLP-1 RAs with basal insulin also has been studied extensively.^6,8-10 When the combination of GLP-1 RAs and basal insulin was compared with basal/bolus insulin regimens, the use of the GLP-1 RAs resulted in lower HbA_1c levels and lower incidence of hypoglycemia.^6,9 Data have demonstrated the complementary mechanisms of using basal insulin and GLP 1 RAs in decreasing HbA_1c levels, insulin requirements, and weight compared with using basal insulin monotherapy and basal/bolus combinations.^6,9-13 Moreover, 3 GLP-1 RA medications currently on the market (liraglutide, dulaglutide, and semaglutide) have displayed cardiovascular and renal benefits, further supporting the use of these medications.^2,5

Despite these benefits, GLP-1 RAs may have bothersome AEs and are associated with a high cost.⁶ In addition, some studies have found that as the length of therapy increases, the positive effects of these agents may diminish.^9,11 In one study, which looked at the impact of the addition of exenatide to patients taking basal or basal/bolus insulin regimens, mean changes in weight were −2.4 kg at 0 to 6 months, −4.3 kg at 6 to 12 months, −6.2 kg at 12 to 18 months, and −5.5 kg at 18 to 27 months. After 18 months, an increase in weight was observed, but the increase remained lower than baseline.¹¹ Another study, conducted over 12 months, found no significant decrease in weight or total daily dose (TDD) of insulin when exenatide or liraglutide were added to various insulin regimens (basal or basal/bolus).¹³ To date, minimal published data exist regarding the addition of newer GLP-1 RAs and the long-term use of these agents beyond 12 months in patients taking basal/bolus insulin regimens. The primary goal of this study was to evaluate the effect of adding GLP-1 RAs to basal/bolus insulin regimens over a 24-month period.

Methods

This study was a retrospective, electronic health record review of all patients on basal and bolus insulin regimens who received additional therapy with a GLP-1 RA at Veteran Health Indiana in Indianapolis from September 1, 2015, to June 30, 2019. Patients meeting inclusion criteria served as their own control. The primary outcome was change in HbA_1c at 3, 6, 12, 18, and 24 months after initiation of the GLP-1 RA. Secondary outcomes included change in weight and TDD of insulin at 3, 6, 12, 18, and 24 months after the initiation of the GLP-1 RAs and incidence of patient-reported or laboratory-confirmed hypoglycemia and other AEs.

Patients were included if they were aged ≥ 18 years with a diagnosis of T2DM, had concomitant prescriptions for both a basal insulin (glargine, detemir, or NPH) and a bolus insulin (aspart, lispro, or regular) before receiving add-on therapy with a GLP-1 RA (exenatide, liraglutide, albiglutide, lixisenatide, dulaglutide, or semaglutide) from September 1, 2015, to June 30, 2019, and had baseline and subsequent HbA_1c measurements available in the electronic health record. Patients were excluded if they had a diagnosis of type 1 DM (T1DM), were followed by an outside clinician for DM care, or if the GLP-1 RA was discontinued before subsequent HbA_1c measurement. The study protocol was approved by the Research and Development Office of Veteran Health Indiana, and the project was deemed exempt from review by the Indiana University Institutional Review Board due to the retrospective nature of the study.

Data analysis was performed using Excel. Change from baseline for each interval was computed, and 1 sample t tests (2-tailed) compared change from baseline to no change. Due to the disparity in the number of patients with data available at each of the time intervals, a mean plot was presented for each group of patients within each interval, allowing mean changes in individual groups to be observed over time.

Results

One hundred twenty-three subjects met inclusion criteria; 16 patients were excluded due to GLP-1 RA discontinuation before follow-up measurement of HbA_1c; 14 were excluded due to patients being managed by a clinician outside of the facility; 1 patient was excluded for lack of documentation regarding baseline and subsequent insulin doses. Ninety-two patient charts were reviewed. Participants had a mean age of 64 years, 95% were male, and 89% were White. Mean baseline HbA_1c was 9.2%, mean body mass index was 38.9, and the mean TDD of insulin was 184 units.

Since some patients switched between GLP-1 RAs throughout the study and there was variation in timing of laboratory and clinic follow-up,

Discussion

Adding a GLP-1 RA to basal/bolus insulin regimens was associated with a statistically significant decrease in HbA_1c at each time point through 18 months. The greatest improvement in glycemic control from baseline was seen at 3 months, with improvements in HbA_1c diminishing at each subsequent period. The study also demonstrated a significant decrease in weight at each time point through 18 months. The greatest decrease in weight was observed at both 6 and 12 months. Statistically significant decreases in TDD were observed at 3, 6, and 12 months. Insulin changes after 12 months were not found to be statistically significant.

Few studies have previously evaluated the use of GLP-1 RAs in patients with T2DM who are already taking basal/bolus insulin regimens. Gyorffy and colleagues reported significant improvements in glycemic control at 3 and 6 months in a sample of 54 patients taking basal/bolus insulin when liraglutide or exenatide was added, although statistical significance was not found at the final 12-month time point.¹³ That study also found a significant decrease in weight at 6 months; however there was not a significant reduction in weight at both 3 and 12 months of GLP-1 RA therapy. There was not a significant decrease in TDD at any of the collected time points. Nonetheless, Gyorffy and colleagues concluded that reduction in TDD leveled off after 12 months, which is consistent with this study’s findings. The small size of the study may have limited the ability to detect statistical significance; however, this study was conducted in a population that was racially diverse and included a higher proportion of women, though average age was similar.¹³

Yoon and colleagues reported weight loss through 18 months, then saw weight increase, though weights did remain lower than baseline. The study also showed no significant change in TDD of insulin after 12 months of concomitant exenatide and insulin therapy.¹¹ Although these results mirror the outcomes observed in this study, Yoon and colleagues did not differentiate results between basal and basal/bolus insulin groups.¹¹ Seino and colleagues observed no significant change in weight after 36 weeks of GLP-1 RA therapy in Japanese patients when used with basal and basal/bolus insulin regimens. Despite the consideration that the population in the study was not overweight (mean body mass index was 25.6), the results of these studies support the idea that effects of GLP-1 RAs on weight and TDD may diminish over time.¹⁴

Within the VHA, GLP-1 RAs are nonformulary medications. Patients must meet certain criteria in order to be approved for these agents, which may include diagnosis of CVD, renal disease, or failure to reach glycemic control with the use of oral agents or insulin. Therefore, participants of this study represent a particular subset of VHA patients, many of whom may have been selected for consideration due to long-standing or uncontrolled T2DM and failure of previous therapies. The baseline demographics support this idea, given poor glycemic control at baseline and high insulin requirements. Once approved for GLP-1 RA therapy, semaglutide is currently the preferred agent within the VHA, with other agents available for select considerations. It should be noted that albiglutide, which was the primary agent selected for some of the patients included in this study, was removed from the market in 2017 for economic considerations.¹⁵ In the case for these patients, a conversion to a formulary-preferred GLP-1 RA was made.

Most of the patients included in this study (70%) were maintained on metformin from baseline throughout the study period. Fifty-seven percent of patients were taking TDD of insulin > 150 units. Considering the significant cost of concentrated insulins, the addition of GLP-1 RAs to standard insulin may prove to be beneficial from a cost standpoint. Additional research in this area may be warranted to establish more data regarding this potential benefit of GLP-1 RAs as add-on therapy.

Many adverse drug reactions were reported at different periods; however, most of these were associated with the gastrointestinal system, which is consistent with current literature, drug labeling, and the mechanism of action.¹⁶ Hypoglycemia occurred in about one-third of the participants; however, it should be noted that alone, GLP-1 RAs are not associated with a high risk of hypoglycemia. Previous studies have found that GLP-1 RA monotherapy is associated with hypoglycemia in 1.6% to 12.6% of patients.^17,18 More likely, the combination of basal/bolus insulin and the GLP-1 RA’s effect on increasing insulin sensitivity through weight loss, improving glucose-dependent insulin secretion, or by decreasing appetite and therefore decreasing carbohydrate intake contributed to the hypoglycemia prevalence.

Limitations and Strengths

Limitations of this study include a small patient population and a gradual reduction in available data as time periods progressed, making even smaller sample sizes for subsequent time periods. A majority of participants were older, males and White race. This could have limited the determination of statistical significance and applicability of the results to other patient populations. Another potential limitation was the retrospective nature of the study design, which may have limited reporting of hypoglycemia and other AEs based on the documentation of the clinician.

Strengths included the study duration and the diversity of GLP-1 RAs used by participants, as the impact of many of these agents has not yet been assessed in the literature. In addition, the retrospective nature of the study allows for a more realistic representation of patient adherence, education, and motivation, which are likely different from those of patients included in prospective clinical trials.

There are no clear guidelines dictating the optimal duration of concomitant GLP-1 RA and insulin therapy; however, our study suggests that there may be continued benefits past short-term use. Also our study suggests that patients with T2DM treated with basal/bolus insulin regimens may glean additional benefit from adding GLP-1 RAs; however, further randomized, controlled studies are warranted, particularly in poorly controlled patients requiring even more aggressive treatment regimens, such as concentrated insulins.

Conclusions

In our study, adding GLP-1 RA to basal/bolus insulin was associated with a significant decrease in HbA_1c from baseline through 18 months. An overall decrease in weight and TDD of insulin was observed through 24 months, but the change in weight was not significant past 18 months, and the change in insulin requirement was not significant past 12 months. Hypoglycemia was observed in almost one-third of patients, and gastrointestinal symptoms were the most common AE observed as a result of adding GLP-1 RAs. More studies are needed to better evaluate the durability and cost benefit of GLP-1 RAs, especially in patients with high insulin requirements.

Acknowledgments

This material is the result of work supported with resources and facilities at Veteran Health Indiana in Indianapolis. Study data were collected and managed using REDCap electronic data capture tools hosted at Veteran Health Indiana. The authors also acknowledge George Eckert for his assistance with data analysis.

Perioperative hyperglycemia, defined as blood glucose levels ≥ 180 mg/dL in the immediate pre- and postoperative period, is associated with increased postoperative morbidity, including infections, preoperative interventions, and in-hospital mortality.^1-3 Despite being identified as a barrier to optimal perioperative glycemic control, limited evidence is available on patient or health care practitioner (HCP) adherence to preoperative insulin protocols.^4-6

Background

Despite mounting evidence of the advantages of maintaining perioperative glucose levels between 80 and 180 mg/dL, available guidelines vary in their recommendations for long-acting basal insulin dosing.^7-10 The Society of Ambulatory Anesthesia suggests using 100% of the prescribed evening dosage of long-acting basal insulin dose on the night before surgery in patients without a history of nocturnal or morning hypoglycemia (category 2A evidence).⁹ However, the revised 2016 United Kingdom National Health Service consensus guideline recommends using 80% to 100% of the prescribed evening dosage of long-acting basal insulin dose on the night before surgery.⁷ The 2022 American Diabetes Association references an observational study of patients with type 2 DM (T2DM) treated with evening-only, long-acting glargine insulin, indicating that the optimal basal insulin dose on the evening before surgery is about 75% of the outpatient dose.^5,10 However, in a randomized, prospective open trial of patients with DM treated with evening-only long-acting basal insulin, no significant difference was noted in the target day of surgery (DOS) glucose levels among different dosing strategies on the evening before surgery.⁶ Presently, the optimal dose of long-acting insulin analogs on the evening before surgery is unknown.

Additionally, little is known about the other factors that influence perioperative glycemic control. Several barriers to optimal perioperative care of patients with DM have been identified, including lack of prioritization by HCPs, lack of knowledge about current evidence-based recommendations, and lack of patient information and involvement.⁴ To determine the effect of patient adherence to preoperative medication instructions on postoperative outcome, a cross-sectional study assessed surgical patients admitted to the postanesthetic care unit (PACU) and found that only 70% of patients with insulin-treated DM took their medications preoperatively. Additionally, 23% of nonadherent patients who omitted their medications either did not understand or forgot preoperative medication management instructions. Preoperative DM medication omission was associated with higher rates of hyperglycemia in the PACU (23.8% vs 3.6%; P = .02).¹¹ Importantly, to our knowledge, the extent of HCP adherence to DM management protocols and the subsequent effect on DOS hyperglycemia has not been examined until now.For patients with DM treated with an evening dose of long-acting basal insulin (ie, either once-daily long-acting basal insulin in the evening or twice-daily long-acting basal insulin, both morning and evening) presenting for elective noncardiac surgery, our aim was to decrease the rate of DOS hyperglycemia from 29% (our baseline) to 15% by intensifying the dose of insulin on the evening before surgery without increasing the rate of hypoglycemia. We also sought to determine the rates of HCP adherence to our insulin protocols as well as patients’ self-reported adherence to HCP instructions over the course of this quality improvement (QI) initiative.

Quality Improvement Program

Our surgical department consists of 11 surgical subspecialties that performed approximately 4400 noncardiac surgeries in 2019. All patients undergoing elective surgery are evaluated in the preoperative clinic, which is staffed by an anesthesiology professional (attending and resident physicians, nurse practitioners, and physician assistants) and internal medicine attending physicians. At the preoperative visit, each patient is evaluated by anesthesiology; medically complex patients may also be referred to an internal medicine professional for further risk stratification and optimization before surgery.

At the preoperative clinic visit, HCPs prepare written patient instructions for the preoperative management of medications, including glucose-lowering medications, based on a DM management protocol that was implemented in 2016 for the preoperative management of insulin, noninsulin injectable agents, and oral hyperglycemic agents. According to this protocol, patients with DM treated with evening long-acting basal insulin (eg, glargine insulin) are instructed to take 50% of their usual evening dose the evening before surgery. A preoperative clinic nurse reviews the final preoperative medication instructions with the patient at the end of the clinic visit. Patients are also instructed to avoid oral intake other than water and necessary medications after midnight before surgery regardless of the time of surgery. On the DOS, the patient’s blood glucose level is measured on arrival to the presurgical area.

Our QI initiative focused only on the dose of self-administered, long-acting basal insulin on the evening before surgery. The effect of the morning of surgery long-acting insulin dose on the DOS glucose levels largely depends on the timing of surgery, which is variable; therefore, we did not target this dose for our initiative. Patients receiving intermediate-acting neutral protamine Hagedorn (NPH) insulin were excluded because our protocol does not recommend a dose reduction for NPH insulin on the evening before surgery.

We developed a comprehensive driver diagram to help elucidate the different factors contributing to DOS hyperglycemia and to guide specific QI interventions.¹² Some of the identified contributors to DOS hyperglycemia, such as the length of preoperative fasting and timing of surgery, are unpredictable and were deemed difficult to address preoperatively. Other contributors to DOS hyperglycemia, such as outpatient DM management, often require interventions over several months, which is well beyond the time usually allotted for preoperative evaluation and optimization. On the other hand, immediate preoperative insulin dosing directly affects DOS glycemic control; therefore, improvement of the preoperative insulin management protocol to optimize the dosage on the evening before surgery was considered to be an achievable QI goal with the potential for decreasing the rate of DOS hyperglycemia in patients presenting for elective noncardiac surgery.

We used the Model for Understanding Success in Quality (MUSIQ) as a framework to identify key contextual factors that may affect the success of our QI project.¹³ Limited resource availability and difficulty with dissemination of protocol changes in the preoperative clinic were determined to be potential barriers to the successful implementation of our QI initiative. Nonetheless, senior leadership support, microsystem QI culture, QI team skills, and physician involvement supported the implementation. The revised Standards for Quality Improvement Reporting Excellence (SQUIRE 2.0) guidelines were followed for this study.¹⁴

Interventions

With stakeholder input from anesthesiology, internal medicine, endocrinology, and nursing, we designed an intervention to iteratively change the HCP protocol instructions for long-acting insulin dosing on the evening before surgery. In phase 1 of the study (October 1, 2018, to March 11, 2019), we obtained baseline data on the rates of DOS hyperglycemia (blood glucose ≥ 180 mg/dL) and hypoglycemia (blood glucose < 80 mg/dL), as well as patient and HCP adherence rates to our existing preoperative DM protocol. For phase 2 (March 12, 2019, to July 22, 2019), the preoperative DM management protocol was changed to increase the dose of long-acting basal insulin on the evening before surgery for patients with hemoglobin A_1c (HbA_1c) levels > 8% from 50% of the usual outpatient dose to 100%. Finally, in phase 3 (July 23, 2019, to March 12, 2020), the protocol was changed to increase the dose of long-acting basal insulin on the evening before surgery for patients with HbA_1c levels ≤ 8% from 50% of the usual outpatient dose to 75% while sustaining the phase 2 change. Preoperative HCPs were informed of the protocol changes in person and were provided with electronic and hard copies of each new protocol.

Protocol

We used a prospective cohort design of 424 consecutive patients with DM who presented for preoperative evaluation for elective noncardiac surgery between October 1, 2018, and March 12, 2020. For the subset of 195 patients treated with an evening dose of long-acting basal insulin, we examined the effect of intensification of this preoperative basal insulin dose on DOS hyperglycemia and hypoglycemia, HCP adherence to iterative changes of the protocol, and patient adherence to HCP instructions on preoperative medication dosing. The QI project was concluded when elective surgeries were paused due to the COVID-19 pandemic.

We created a standardized preoperative data collection form that included information on the most recent HbA_1c, time, dose, and type of patient-administered insulin on the evening before surgery, and DOS blood glucose level. A preoperative clinic nurse completed the standardized preoperative data collection form. The HCP’s preoperative medication instructions and the preoperative data collection forms were gathered for review and data analysis.

The primary outcome was DOS hyperglycemia (blood glucose levels ≥ 180 mg/dL). We monitored the rate of DOS hypoglycemia (blood glucose levels < 80 mg/dL) as a balancing measure to ensure safety with long-acting basal insulin intensification. Although hypoglycemia is defined as a blood glucose level < 70 mg/dL, a target glucose range of 80 mg/dL to 180 mg/dL is recommended during the perioperative period.⁸ Therefore, we chose a more conservative definition of hypoglycemia (blood glucose levels < 80 mg/dL) to adhere to the recommended perioperative glucose target range.

Process measures included HCP adherence to each protocol change, which was assessed by comparing written preoperative patient instructions to the current protocol. Similarly, patient adherence to HCP-recommended long-acting basal insulin dosing was assessed by comparing written preoperative patient instructions to the patient’s self-reported time and dose of long-acting basal insulin on the evening before surgery. For any discrepancy between the HCP instructions and protocol or HCP-recommended dose and patient self-reported dose of long-acting basal insulin, a detailed chart review was performed to determine the etiology.

Statistical Analysis

We used the statistical process p-control chart to assess the effect of iterative changes to the preoperative long-acting basal insulin protocol on DOS hyperglycemia. The proportion defective (rate of DOS hyperglycemia) was plotted against time to determine whether the observed variations in the rate of DOS hyperglycemia over time were attributable to random common causes or special causes because of our intervention. The lower control limit (LCL) and upper control limit (UCL) define the limits of expected outcome measures in a stable process prior to introducing changes and were set at 3 SDs from the mean to balance the likelihood of type I (false-positive) and type II (false-negative) errors. Because of the variable interval sample sizes, we used the CRITBINOM function of Microsoft Excel to calculate the exact UCL satisfying the 3 SD limits of 0.99865.¹⁵ The Shewhart rules (outliers, runs or shifts, trends, sawtooth) were used to analyze the p-control chart to identify special cause signals resulting from our interventions.¹⁶ We used the statistical process t-control chart to record the time (days) between the few occurrences of DOS hypoglycemia because cases of hypoglycemia were rare.

Ethical Consideration

The Human Research Protection Program, Associate Chief of Staff for Research and Development, and Quality, Safety, and Values department reviewed this project in accordance with the Veterans Health Administration Program Guide 1200.21 and determined that it was a nonresearch operations activity; thus, approval by an institutional review board was not needed. The authors declare no competing interests.

Patient Outcomes

We prospectively followed 424 consecutive patients with DM undergoing elective noncardiac surgery from the time of the preoperative clinic evaluation until DOS; 195 patients were on evening

A subgroup analysis of DOS glucose levels in insulin-treated patients with preoperative HbA_1c levels > 8% did not demonstrate a change in the rate of

Discussion

The management and prevention of immediate perioperative hyperglycemia and glycemic variability have attracted attention as evidence has mounted for their association with postoperative morbidity and mortality.^1,2,17 Available guidelines for preventing DOS hyperglycemia vary in their recommendations for preoperative insulin management.^7-10 Notably, concerns about iatrogenic hypoglycemia often hinder efforts to lower rates of DOS hyperglycemia.⁴ We successfully implemented an iterative intensification protocol for preoperative long-acting basal insulin doses on the evening before surgery but did not observe a lower rate of hyperglycemia. Importantly, we also did not observe a higher rate of hypoglycemia on the DOS, as observed in a previous study.⁵

The observational study by Demma and colleagues found that patients receiving 75% of their evening, long-acting basal insulin dose were significantly more likely to achieve target blood glucose levels of 100 to 180 mg/dL than patients receiving no insulin at all (78% vs 0%; P = .001). However, no significant difference was noted when this group was compared with patients receiving 50% of their evening, long-acting basal insulin doses (78% vs 70%; P = .56). This is more clinically pertinent as it is generally accepted that the evening, long-acting insulin dose should not be entirely withheld on the evening before surgery.⁵

These findings are consistent with our observation that the rate of DOS hyperglycemia did not decrease with intensification of the evening, long-acting insulin dose from 50% to 100% of the prescribed dose in patients with HbA_1c levels > 8% (phase 2) and 50% to 75% of the prescribed dose in patients with HbA_1c levels ≤ 8% (phase 3). In the study by Demma and colleagues, few patients presented with preoperative hypoglycemia (2.7%) but all had received 100% of their evening, long-acting basal insulin dose, suggesting a significant increase in the rate of hypoglycemia compared with patients receiving lower doses of insulin (P = .01).⁵ However, long-term DM control as assessed by HbA_1c level was available for < 10% of the patients, making it difficult to evaluate the effect of overall DM control on the results.⁵ In our study, preoperative HbA_1c levels were available for 99.5% of the patients and only those with HbA_1c levels > 8% received 100% of their evening, long-acting insulin dose on the evening before surgery. Notably, we did not observe a higher rate of hypoglycemia in this patient population, indicating that preoperative insulin dose intensification is safe for this subgroup.

Although HCP adherence to perioperative DM management protocols has been identified as a predominant barrier to the delivery of optimal perioperative DM care, prior studies of various preoperative insulin protocols to reduce perioperative hyperglycemia have not reported HCP adherence to their insulin protocols or its effect on DOS hyperglycemia.^4-6 Additionally, patient adherence to HCP instructions is a key factor identified in our driver diagram that may influence DOS hyperglycemia, a hypothesis that is supported by a prior cross-sectional study showing an increased rate of hyperglycemia in the PACU with omission of preoperative DM medication.¹¹ In our study, patient adherence to preoperative medication management instructions was higher than reported previously and remained consistently high regardless of protocol changes, which may explain why patient adherence did not affect the rate of DOS hyperglycemia.

Although not part of our study protocol, our preoperative HCPs routinely prepare written patient instructions for the preoperative management of medications for all patients, which likely explains higher patient adherence to instructions in our study than seen in the previous study where written instructions were only encouraged.¹¹ However, HCP adherence to the protocol decreased after our phase 2 changes and was associated with a transient increase in DOS hyperglycemia rates. The DOS hyperglycemia rates returned to baseline levels with ongoing QI efforts and education to improve HCP adherence to protocol.

Limitations

Our QI initiative had several limitations. Nearly all patients were male veterans with T2DM, and most were older (range, 50-89 years). This limits the generalizability to women, younger patients, and people with type 1 DM. Additionally, our data collection relied on completion and collection of the preoperative form by different HCPs, allowing for sampling bias if some patients with DM undergoing elective noncardiac surgery were missed. Furthermore, although we could verify HCP adherence to the preoperative DM management protocols by reviewing their written instructions, we relied on patients’ self-reported adherence to the preoperative instructions. Finally, we did not evaluate postoperative blood glucose levels because the effect of intraoperative factors such as fluid, insulin, and glucocorticoid administration on postoperative glucose levels are variable. To the best of our knowledge, no other major systematic changes occurred in the preoperative care of patients with DM during the study period.

Conclusions

The findings of our QI initiative suggest that HCP adherence to preoperative DM management protocols may be a key contributor to DOS hyperglycemia and that ensuring HCP adherence may be as important as preoperative insulin dose adjustments. To our knowledge, this is the first study to report rates of HCP adherence to preoperative DM management protocols and its effect on DOS hyperglycemia. We will focus future QI efforts on optimizing HCP adherence to preoperative DM management protocols at our institution.

Acknowledgments

We thank our endocrinology expert, Dr. Kristina Utzschneider, for her guidance in designing this improvement project and our academic research coach, Dr. Helene Starks, for her help in editing the manuscript.

Perioperative hyperglycemia, defined as blood glucose levels ≥ 180 mg/dL in the immediate pre- and postoperative period, is associated with increased postoperative morbidity, including infections, preoperative interventions, and in-hospital mortality.^1-3 Despite being identified as a barrier to optimal perioperative glycemic control, limited evidence is available on patient or health care practitioner (HCP) adherence to preoperative insulin protocols.^4-6

Background

Despite mounting evidence of the advantages of maintaining perioperative glucose levels between 80 and 180 mg/dL, available guidelines vary in their recommendations for long-acting basal insulin dosing.^7-10 The Society of Ambulatory Anesthesia suggests using 100% of the prescribed evening dosage of long-acting basal insulin dose on the night before surgery in patients without a history of nocturnal or morning hypoglycemia (category 2A evidence).⁹ However, the revised 2016 United Kingdom National Health Service consensus guideline recommends using 80% to 100% of the prescribed evening dosage of long-acting basal insulin dose on the night before surgery.⁷ The 2022 American Diabetes Association references an observational study of patients with type 2 DM (T2DM) treated with evening-only, long-acting glargine insulin, indicating that the optimal basal insulin dose on the evening before surgery is about 75% of the outpatient dose.^5,10 However, in a randomized, prospective open trial of patients with DM treated with evening-only long-acting basal insulin, no significant difference was noted in the target day of surgery (DOS) glucose levels among different dosing strategies on the evening before surgery.⁶ Presently, the optimal dose of long-acting insulin analogs on the evening before surgery is unknown.

Additionally, little is known about the other factors that influence perioperative glycemic control. Several barriers to optimal perioperative care of patients with DM have been identified, including lack of prioritization by HCPs, lack of knowledge about current evidence-based recommendations, and lack of patient information and involvement.⁴ To determine the effect of patient adherence to preoperative medication instructions on postoperative outcome, a cross-sectional study assessed surgical patients admitted to the postanesthetic care unit (PACU) and found that only 70% of patients with insulin-treated DM took their medications preoperatively. Additionally, 23% of nonadherent patients who omitted their medications either did not understand or forgot preoperative medication management instructions. Preoperative DM medication omission was associated with higher rates of hyperglycemia in the PACU (23.8% vs 3.6%; P = .02).¹¹ Importantly, to our knowledge, the extent of HCP adherence to DM management protocols and the subsequent effect on DOS hyperglycemia has not been examined until now.For patients with DM treated with an evening dose of long-acting basal insulin (ie, either once-daily long-acting basal insulin in the evening or twice-daily long-acting basal insulin, both morning and evening) presenting for elective noncardiac surgery, our aim was to decrease the rate of DOS hyperglycemia from 29% (our baseline) to 15% by intensifying the dose of insulin on the evening before surgery without increasing the rate of hypoglycemia. We also sought to determine the rates of HCP adherence to our insulin protocols as well as patients’ self-reported adherence to HCP instructions over the course of this quality improvement (QI) initiative.

Quality Improvement Program

Our surgical department consists of 11 surgical subspecialties that performed approximately 4400 noncardiac surgeries in 2019. All patients undergoing elective surgery are evaluated in the preoperative clinic, which is staffed by an anesthesiology professional (attending and resident physicians, nurse practitioners, and physician assistants) and internal medicine attending physicians. At the preoperative visit, each patient is evaluated by anesthesiology; medically complex patients may also be referred to an internal medicine professional for further risk stratification and optimization before surgery.

At the preoperative clinic visit, HCPs prepare written patient instructions for the preoperative management of medications, including glucose-lowering medications, based on a DM management protocol that was implemented in 2016 for the preoperative management of insulin, noninsulin injectable agents, and oral hyperglycemic agents. According to this protocol, patients with DM treated with evening long-acting basal insulin (eg, glargine insulin) are instructed to take 50% of their usual evening dose the evening before surgery. A preoperative clinic nurse reviews the final preoperative medication instructions with the patient at the end of the clinic visit. Patients are also instructed to avoid oral intake other than water and necessary medications after midnight before surgery regardless of the time of surgery. On the DOS, the patient’s blood glucose level is measured on arrival to the presurgical area.

Our QI initiative focused only on the dose of self-administered, long-acting basal insulin on the evening before surgery. The effect of the morning of surgery long-acting insulin dose on the DOS glucose levels largely depends on the timing of surgery, which is variable; therefore, we did not target this dose for our initiative. Patients receiving intermediate-acting neutral protamine Hagedorn (NPH) insulin were excluded because our protocol does not recommend a dose reduction for NPH insulin on the evening before surgery.

We developed a comprehensive driver diagram to help elucidate the different factors contributing to DOS hyperglycemia and to guide specific QI interventions.¹² Some of the identified contributors to DOS hyperglycemia, such as the length of preoperative fasting and timing of surgery, are unpredictable and were deemed difficult to address preoperatively. Other contributors to DOS hyperglycemia, such as outpatient DM management, often require interventions over several months, which is well beyond the time usually allotted for preoperative evaluation and optimization. On the other hand, immediate preoperative insulin dosing directly affects DOS glycemic control; therefore, improvement of the preoperative insulin management protocol to optimize the dosage on the evening before surgery was considered to be an achievable QI goal with the potential for decreasing the rate of DOS hyperglycemia in patients presenting for elective noncardiac surgery.

We used the Model for Understanding Success in Quality (MUSIQ) as a framework to identify key contextual factors that may affect the success of our QI project.¹³ Limited resource availability and difficulty with dissemination of protocol changes in the preoperative clinic were determined to be potential barriers to the successful implementation of our QI initiative. Nonetheless, senior leadership support, microsystem QI culture, QI team skills, and physician involvement supported the implementation. The revised Standards for Quality Improvement Reporting Excellence (SQUIRE 2.0) guidelines were followed for this study.¹⁴

Interventions

With stakeholder input from anesthesiology, internal medicine, endocrinology, and nursing, we designed an intervention to iteratively change the HCP protocol instructions for long-acting insulin dosing on the evening before surgery. In phase 1 of the study (October 1, 2018, to March 11, 2019), we obtained baseline data on the rates of DOS hyperglycemia (blood glucose ≥ 180 mg/dL) and hypoglycemia (blood glucose < 80 mg/dL), as well as patient and HCP adherence rates to our existing preoperative DM protocol. For phase 2 (March 12, 2019, to July 22, 2019), the preoperative DM management protocol was changed to increase the dose of long-acting basal insulin on the evening before surgery for patients with hemoglobin A_1c (HbA_1c) levels > 8% from 50% of the usual outpatient dose to 100%. Finally, in phase 3 (July 23, 2019, to March 12, 2020), the protocol was changed to increase the dose of long-acting basal insulin on the evening before surgery for patients with HbA_1c levels ≤ 8% from 50% of the usual outpatient dose to 75% while sustaining the phase 2 change. Preoperative HCPs were informed of the protocol changes in person and were provided with electronic and hard copies of each new protocol.

Protocol

We used a prospective cohort design of 424 consecutive patients with DM who presented for preoperative evaluation for elective noncardiac surgery between October 1, 2018, and March 12, 2020. For the subset of 195 patients treated with an evening dose of long-acting basal insulin, we examined the effect of intensification of this preoperative basal insulin dose on DOS hyperglycemia and hypoglycemia, HCP adherence to iterative changes of the protocol, and patient adherence to HCP instructions on preoperative medication dosing. The QI project was concluded when elective surgeries were paused due to the COVID-19 pandemic.

We created a standardized preoperative data collection form that included information on the most recent HbA_1c, time, dose, and type of patient-administered insulin on the evening before surgery, and DOS blood glucose level. A preoperative clinic nurse completed the standardized preoperative data collection form. The HCP’s preoperative medication instructions and the preoperative data collection forms were gathered for review and data analysis.

The primary outcome was DOS hyperglycemia (blood glucose levels ≥ 180 mg/dL). We monitored the rate of DOS hypoglycemia (blood glucose levels < 80 mg/dL) as a balancing measure to ensure safety with long-acting basal insulin intensification. Although hypoglycemia is defined as a blood glucose level < 70 mg/dL, a target glucose range of 80 mg/dL to 180 mg/dL is recommended during the perioperative period.⁸ Therefore, we chose a more conservative definition of hypoglycemia (blood glucose levels < 80 mg/dL) to adhere to the recommended perioperative glucose target range.

Process measures included HCP adherence to each protocol change, which was assessed by comparing written preoperative patient instructions to the current protocol. Similarly, patient adherence to HCP-recommended long-acting basal insulin dosing was assessed by comparing written preoperative patient instructions to the patient’s self-reported time and dose of long-acting basal insulin on the evening before surgery. For any discrepancy between the HCP instructions and protocol or HCP-recommended dose and patient self-reported dose of long-acting basal insulin, a detailed chart review was performed to determine the etiology.

Statistical Analysis

We used the statistical process p-control chart to assess the effect of iterative changes to the preoperative long-acting basal insulin protocol on DOS hyperglycemia. The proportion defective (rate of DOS hyperglycemia) was plotted against time to determine whether the observed variations in the rate of DOS hyperglycemia over time were attributable to random common causes or special causes because of our intervention. The lower control limit (LCL) and upper control limit (UCL) define the limits of expected outcome measures in a stable process prior to introducing changes and were set at 3 SDs from the mean to balance the likelihood of type I (false-positive) and type II (false-negative) errors. Because of the variable interval sample sizes, we used the CRITBINOM function of Microsoft Excel to calculate the exact UCL satisfying the 3 SD limits of 0.99865.¹⁵ The Shewhart rules (outliers, runs or shifts, trends, sawtooth) were used to analyze the p-control chart to identify special cause signals resulting from our interventions.¹⁶ We used the statistical process t-control chart to record the time (days) between the few occurrences of DOS hypoglycemia because cases of hypoglycemia were rare.

Ethical Consideration

The Human Research Protection Program, Associate Chief of Staff for Research and Development, and Quality, Safety, and Values department reviewed this project in accordance with the Veterans Health Administration Program Guide 1200.21 and determined that it was a nonresearch operations activity; thus, approval by an institutional review board was not needed. The authors declare no competing interests.

Patient Outcomes

We prospectively followed 424 consecutive patients with DM undergoing elective noncardiac surgery from the time of the preoperative clinic evaluation until DOS; 195 patients were on evening

A subgroup analysis of DOS glucose levels in insulin-treated patients with preoperative HbA_1c levels > 8% did not demonstrate a change in the rate of

Discussion

The management and prevention of immediate perioperative hyperglycemia and glycemic variability have attracted attention as evidence has mounted for their association with postoperative morbidity and mortality.^1,2,17 Available guidelines for preventing DOS hyperglycemia vary in their recommendations for preoperative insulin management.^7-10 Notably, concerns about iatrogenic hypoglycemia often hinder efforts to lower rates of DOS hyperglycemia.⁴ We successfully implemented an iterative intensification protocol for preoperative long-acting basal insulin doses on the evening before surgery but did not observe a lower rate of hyperglycemia. Importantly, we also did not observe a higher rate of hypoglycemia on the DOS, as observed in a previous study.⁵

The observational study by Demma and colleagues found that patients receiving 75% of their evening, long-acting basal insulin dose were significantly more likely to achieve target blood glucose levels of 100 to 180 mg/dL than patients receiving no insulin at all (78% vs 0%; P = .001). However, no significant difference was noted when this group was compared with patients receiving 50% of their evening, long-acting basal insulin doses (78% vs 70%; P = .56). This is more clinically pertinent as it is generally accepted that the evening, long-acting insulin dose should not be entirely withheld on the evening before surgery.⁵

These findings are consistent with our observation that the rate of DOS hyperglycemia did not decrease with intensification of the evening, long-acting insulin dose from 50% to 100% of the prescribed dose in patients with HbA_1c levels > 8% (phase 2) and 50% to 75% of the prescribed dose in patients with HbA_1c levels ≤ 8% (phase 3). In the study by Demma and colleagues, few patients presented with preoperative hypoglycemia (2.7%) but all had received 100% of their evening, long-acting basal insulin dose, suggesting a significant increase in the rate of hypoglycemia compared with patients receiving lower doses of insulin (P = .01).⁵ However, long-term DM control as assessed by HbA_1c level was available for < 10% of the patients, making it difficult to evaluate the effect of overall DM control on the results.⁵ In our study, preoperative HbA_1c levels were available for 99.5% of the patients and only those with HbA_1c levels > 8% received 100% of their evening, long-acting insulin dose on the evening before surgery. Notably, we did not observe a higher rate of hypoglycemia in this patient population, indicating that preoperative insulin dose intensification is safe for this subgroup.

Although HCP adherence to perioperative DM management protocols has been identified as a predominant barrier to the delivery of optimal perioperative DM care, prior studies of various preoperative insulin protocols to reduce perioperative hyperglycemia have not reported HCP adherence to their insulin protocols or its effect on DOS hyperglycemia.^4-6 Additionally, patient adherence to HCP instructions is a key factor identified in our driver diagram that may influence DOS hyperglycemia, a hypothesis that is supported by a prior cross-sectional study showing an increased rate of hyperglycemia in the PACU with omission of preoperative DM medication.¹¹ In our study, patient adherence to preoperative medication management instructions was higher than reported previously and remained consistently high regardless of protocol changes, which may explain why patient adherence did not affect the rate of DOS hyperglycemia.

Although not part of our study protocol, our preoperative HCPs routinely prepare written patient instructions for the preoperative management of medications for all patients, which likely explains higher patient adherence to instructions in our study than seen in the previous study where written instructions were only encouraged.¹¹ However, HCP adherence to the protocol decreased after our phase 2 changes and was associated with a transient increase in DOS hyperglycemia rates. The DOS hyperglycemia rates returned to baseline levels with ongoing QI efforts and education to improve HCP adherence to protocol.

Limitations

Our QI initiative had several limitations. Nearly all patients were male veterans with T2DM, and most were older (range, 50-89 years). This limits the generalizability to women, younger patients, and people with type 1 DM. Additionally, our data collection relied on completion and collection of the preoperative form by different HCPs, allowing for sampling bias if some patients with DM undergoing elective noncardiac surgery were missed. Furthermore, although we could verify HCP adherence to the preoperative DM management protocols by reviewing their written instructions, we relied on patients’ self-reported adherence to the preoperative instructions. Finally, we did not evaluate postoperative blood glucose levels because the effect of intraoperative factors such as fluid, insulin, and glucocorticoid administration on postoperative glucose levels are variable. To the best of our knowledge, no other major systematic changes occurred in the preoperative care of patients with DM during the study period.

Conclusions

The findings of our QI initiative suggest that HCP adherence to preoperative DM management protocols may be a key contributor to DOS hyperglycemia and that ensuring HCP adherence may be as important as preoperative insulin dose adjustments. To our knowledge, this is the first study to report rates of HCP adherence to preoperative DM management protocols and its effect on DOS hyperglycemia. We will focus future QI efforts on optimizing HCP adherence to preoperative DM management protocols at our institution.

Acknowledgments

We thank our endocrinology expert, Dr. Kristina Utzschneider, for her guidance in designing this improvement project and our academic research coach, Dr. Helene Starks, for her help in editing the manuscript.

Perioperative hyperglycemia, defined as blood glucose levels ≥ 180 mg/dL in the immediate pre- and postoperative period, is associated with increased postoperative morbidity, including infections, preoperative interventions, and in-hospital mortality.^1-3 Despite being identified as a barrier to optimal perioperative glycemic control, limited evidence is available on patient or health care practitioner (HCP) adherence to preoperative insulin protocols.^4-6

Background

Despite mounting evidence of the advantages of maintaining perioperative glucose levels between 80 and 180 mg/dL, available guidelines vary in their recommendations for long-acting basal insulin dosing.^7-10 The Society of Ambulatory Anesthesia suggests using 100% of the prescribed evening dosage of long-acting basal insulin dose on the night before surgery in patients without a history of nocturnal or morning hypoglycemia (category 2A evidence).⁹ However, the revised 2016 United Kingdom National Health Service consensus guideline recommends using 80% to 100% of the prescribed evening dosage of long-acting basal insulin dose on the night before surgery.⁷ The 2022 American Diabetes Association references an observational study of patients with type 2 DM (T2DM) treated with evening-only, long-acting glargine insulin, indicating that the optimal basal insulin dose on the evening before surgery is about 75% of the outpatient dose.^5,10 However, in a randomized, prospective open trial of patients with DM treated with evening-only long-acting basal insulin, no significant difference was noted in the target day of surgery (DOS) glucose levels among different dosing strategies on the evening before surgery.⁶ Presently, the optimal dose of long-acting insulin analogs on the evening before surgery is unknown.

Additionally, little is known about the other factors that influence perioperative glycemic control. Several barriers to optimal perioperative care of patients with DM have been identified, including lack of prioritization by HCPs, lack of knowledge about current evidence-based recommendations, and lack of patient information and involvement.⁴ To determine the effect of patient adherence to preoperative medication instructions on postoperative outcome, a cross-sectional study assessed surgical patients admitted to the postanesthetic care unit (PACU) and found that only 70% of patients with insulin-treated DM took their medications preoperatively. Additionally, 23% of nonadherent patients who omitted their medications either did not understand or forgot preoperative medication management instructions. Preoperative DM medication omission was associated with higher rates of hyperglycemia in the PACU (23.8% vs 3.6%; P = .02).¹¹ Importantly, to our knowledge, the extent of HCP adherence to DM management protocols and the subsequent effect on DOS hyperglycemia has not been examined until now.For patients with DM treated with an evening dose of long-acting basal insulin (ie, either once-daily long-acting basal insulin in the evening or twice-daily long-acting basal insulin, both morning and evening) presenting for elective noncardiac surgery, our aim was to decrease the rate of DOS hyperglycemia from 29% (our baseline) to 15% by intensifying the dose of insulin on the evening before surgery without increasing the rate of hypoglycemia. We also sought to determine the rates of HCP adherence to our insulin protocols as well as patients’ self-reported adherence to HCP instructions over the course of this quality improvement (QI) initiative.

Quality Improvement Program

Our surgical department consists of 11 surgical subspecialties that performed approximately 4400 noncardiac surgeries in 2019. All patients undergoing elective surgery are evaluated in the preoperative clinic, which is staffed by an anesthesiology professional (attending and resident physicians, nurse practitioners, and physician assistants) and internal medicine attending physicians. At the preoperative visit, each patient is evaluated by anesthesiology; medically complex patients may also be referred to an internal medicine professional for further risk stratification and optimization before surgery.

At the preoperative clinic visit, HCPs prepare written patient instructions for the preoperative management of medications, including glucose-lowering medications, based on a DM management protocol that was implemented in 2016 for the preoperative management of insulin, noninsulin injectable agents, and oral hyperglycemic agents. According to this protocol, patients with DM treated with evening long-acting basal insulin (eg, glargine insulin) are instructed to take 50% of their usual evening dose the evening before surgery. A preoperative clinic nurse reviews the final preoperative medication instructions with the patient at the end of the clinic visit. Patients are also instructed to avoid oral intake other than water and necessary medications after midnight before surgery regardless of the time of surgery. On the DOS, the patient’s blood glucose level is measured on arrival to the presurgical area.

Our QI initiative focused only on the dose of self-administered, long-acting basal insulin on the evening before surgery. The effect of the morning of surgery long-acting insulin dose on the DOS glucose levels largely depends on the timing of surgery, which is variable; therefore, we did not target this dose for our initiative. Patients receiving intermediate-acting neutral protamine Hagedorn (NPH) insulin were excluded because our protocol does not recommend a dose reduction for NPH insulin on the evening before surgery.

We developed a comprehensive driver diagram to help elucidate the different factors contributing to DOS hyperglycemia and to guide specific QI interventions.¹² Some of the identified contributors to DOS hyperglycemia, such as the length of preoperative fasting and timing of surgery, are unpredictable and were deemed difficult to address preoperatively. Other contributors to DOS hyperglycemia, such as outpatient DM management, often require interventions over several months, which is well beyond the time usually allotted for preoperative evaluation and optimization. On the other hand, immediate preoperative insulin dosing directly affects DOS glycemic control; therefore, improvement of the preoperative insulin management protocol to optimize the dosage on the evening before surgery was considered to be an achievable QI goal with the potential for decreasing the rate of DOS hyperglycemia in patients presenting for elective noncardiac surgery.

We used the Model for Understanding Success in Quality (MUSIQ) as a framework to identify key contextual factors that may affect the success of our QI project.¹³ Limited resource availability and difficulty with dissemination of protocol changes in the preoperative clinic were determined to be potential barriers to the successful implementation of our QI initiative. Nonetheless, senior leadership support, microsystem QI culture, QI team skills, and physician involvement supported the implementation. The revised Standards for Quality Improvement Reporting Excellence (SQUIRE 2.0) guidelines were followed for this study.¹⁴

Interventions

With stakeholder input from anesthesiology, internal medicine, endocrinology, and nursing, we designed an intervention to iteratively change the HCP protocol instructions for long-acting insulin dosing on the evening before surgery. In phase 1 of the study (October 1, 2018, to March 11, 2019), we obtained baseline data on the rates of DOS hyperglycemia (blood glucose ≥ 180 mg/dL) and hypoglycemia (blood glucose < 80 mg/dL), as well as patient and HCP adherence rates to our existing preoperative DM protocol. For phase 2 (March 12, 2019, to July 22, 2019), the preoperative DM management protocol was changed to increase the dose of long-acting basal insulin on the evening before surgery for patients with hemoglobin A_1c (HbA_1c) levels > 8% from 50% of the usual outpatient dose to 100%. Finally, in phase 3 (July 23, 2019, to March 12, 2020), the protocol was changed to increase the dose of long-acting basal insulin on the evening before surgery for patients with HbA_1c levels ≤ 8% from 50% of the usual outpatient dose to 75% while sustaining the phase 2 change. Preoperative HCPs were informed of the protocol changes in person and were provided with electronic and hard copies of each new protocol.

Protocol

We used a prospective cohort design of 424 consecutive patients with DM who presented for preoperative evaluation for elective noncardiac surgery between October 1, 2018, and March 12, 2020. For the subset of 195 patients treated with an evening dose of long-acting basal insulin, we examined the effect of intensification of this preoperative basal insulin dose on DOS hyperglycemia and hypoglycemia, HCP adherence to iterative changes of the protocol, and patient adherence to HCP instructions on preoperative medication dosing. The QI project was concluded when elective surgeries were paused due to the COVID-19 pandemic.

We created a standardized preoperative data collection form that included information on the most recent HbA_1c, time, dose, and type of patient-administered insulin on the evening before surgery, and DOS blood glucose level. A preoperative clinic nurse completed the standardized preoperative data collection form. The HCP’s preoperative medication instructions and the preoperative data collection forms were gathered for review and data analysis.

The primary outcome was DOS hyperglycemia (blood glucose levels ≥ 180 mg/dL). We monitored the rate of DOS hypoglycemia (blood glucose levels < 80 mg/dL) as a balancing measure to ensure safety with long-acting basal insulin intensification. Although hypoglycemia is defined as a blood glucose level < 70 mg/dL, a target glucose range of 80 mg/dL to 180 mg/dL is recommended during the perioperative period.⁸ Therefore, we chose a more conservative definition of hypoglycemia (blood glucose levels < 80 mg/dL) to adhere to the recommended perioperative glucose target range.

Process measures included HCP adherence to each protocol change, which was assessed by comparing written preoperative patient instructions to the current protocol. Similarly, patient adherence to HCP-recommended long-acting basal insulin dosing was assessed by comparing written preoperative patient instructions to the patient’s self-reported time and dose of long-acting basal insulin on the evening before surgery. For any discrepancy between the HCP instructions and protocol or HCP-recommended dose and patient self-reported dose of long-acting basal insulin, a detailed chart review was performed to determine the etiology.

Statistical Analysis

We used the statistical process p-control chart to assess the effect of iterative changes to the preoperative long-acting basal insulin protocol on DOS hyperglycemia. The proportion defective (rate of DOS hyperglycemia) was plotted against time to determine whether the observed variations in the rate of DOS hyperglycemia over time were attributable to random common causes or special causes because of our intervention. The lower control limit (LCL) and upper control limit (UCL) define the limits of expected outcome measures in a stable process prior to introducing changes and were set at 3 SDs from the mean to balance the likelihood of type I (false-positive) and type II (false-negative) errors. Because of the variable interval sample sizes, we used the CRITBINOM function of Microsoft Excel to calculate the exact UCL satisfying the 3 SD limits of 0.99865.¹⁵ The Shewhart rules (outliers, runs or shifts, trends, sawtooth) were used to analyze the p-control chart to identify special cause signals resulting from our interventions.¹⁶ We used the statistical process t-control chart to record the time (days) between the few occurrences of DOS hypoglycemia because cases of hypoglycemia were rare.

Ethical Consideration

The Human Research Protection Program, Associate Chief of Staff for Research and Development, and Quality, Safety, and Values department reviewed this project in accordance with the Veterans Health Administration Program Guide 1200.21 and determined that it was a nonresearch operations activity; thus, approval by an institutional review board was not needed. The authors declare no competing interests.

Patient Outcomes

We prospectively followed 424 consecutive patients with DM undergoing elective noncardiac surgery from the time of the preoperative clinic evaluation until DOS; 195 patients were on evening

A subgroup analysis of DOS glucose levels in insulin-treated patients with preoperative HbA_1c levels > 8% did not demonstrate a change in the rate of

Discussion

The management and prevention of immediate perioperative hyperglycemia and glycemic variability have attracted attention as evidence has mounted for their association with postoperative morbidity and mortality.^1,2,17 Available guidelines for preventing DOS hyperglycemia vary in their recommendations for preoperative insulin management.^7-10 Notably, concerns about iatrogenic hypoglycemia often hinder efforts to lower rates of DOS hyperglycemia.⁴ We successfully implemented an iterative intensification protocol for preoperative long-acting basal insulin doses on the evening before surgery but did not observe a lower rate of hyperglycemia. Importantly, we also did not observe a higher rate of hypoglycemia on the DOS, as observed in a previous study.⁵

The observational study by Demma and colleagues found that patients receiving 75% of their evening, long-acting basal insulin dose were significantly more likely to achieve target blood glucose levels of 100 to 180 mg/dL than patients receiving no insulin at all (78% vs 0%; P = .001). However, no significant difference was noted when this group was compared with patients receiving 50% of their evening, long-acting basal insulin doses (78% vs 70%; P = .56). This is more clinically pertinent as it is generally accepted that the evening, long-acting insulin dose should not be entirely withheld on the evening before surgery.⁵

These findings are consistent with our observation that the rate of DOS hyperglycemia did not decrease with intensification of the evening, long-acting insulin dose from 50% to 100% of the prescribed dose in patients with HbA_1c levels > 8% (phase 2) and 50% to 75% of the prescribed dose in patients with HbA_1c levels ≤ 8% (phase 3). In the study by Demma and colleagues, few patients presented with preoperative hypoglycemia (2.7%) but all had received 100% of their evening, long-acting basal insulin dose, suggesting a significant increase in the rate of hypoglycemia compared with patients receiving lower doses of insulin (P = .01).⁵ However, long-term DM control as assessed by HbA_1c level was available for < 10% of the patients, making it difficult to evaluate the effect of overall DM control on the results.⁵ In our study, preoperative HbA_1c levels were available for 99.5% of the patients and only those with HbA_1c levels > 8% received 100% of their evening, long-acting insulin dose on the evening before surgery. Notably, we did not observe a higher rate of hypoglycemia in this patient population, indicating that preoperative insulin dose intensification is safe for this subgroup.

Although HCP adherence to perioperative DM management protocols has been identified as a predominant barrier to the delivery of optimal perioperative DM care, prior studies of various preoperative insulin protocols to reduce perioperative hyperglycemia have not reported HCP adherence to their insulin protocols or its effect on DOS hyperglycemia.^4-6 Additionally, patient adherence to HCP instructions is a key factor identified in our driver diagram that may influence DOS hyperglycemia, a hypothesis that is supported by a prior cross-sectional study showing an increased rate of hyperglycemia in the PACU with omission of preoperative DM medication.¹¹ In our study, patient adherence to preoperative medication management instructions was higher than reported previously and remained consistently high regardless of protocol changes, which may explain why patient adherence did not affect the rate of DOS hyperglycemia.

Although not part of our study protocol, our preoperative HCPs routinely prepare written patient instructions for the preoperative management of medications for all patients, which likely explains higher patient adherence to instructions in our study than seen in the previous study where written instructions were only encouraged.¹¹ However, HCP adherence to the protocol decreased after our phase 2 changes and was associated with a transient increase in DOS hyperglycemia rates. The DOS hyperglycemia rates returned to baseline levels with ongoing QI efforts and education to improve HCP adherence to protocol.

Limitations

Our QI initiative had several limitations. Nearly all patients were male veterans with T2DM, and most were older (range, 50-89 years). This limits the generalizability to women, younger patients, and people with type 1 DM. Additionally, our data collection relied on completion and collection of the preoperative form by different HCPs, allowing for sampling bias if some patients with DM undergoing elective noncardiac surgery were missed. Furthermore, although we could verify HCP adherence to the preoperative DM management protocols by reviewing their written instructions, we relied on patients’ self-reported adherence to the preoperative instructions. Finally, we did not evaluate postoperative blood glucose levels because the effect of intraoperative factors such as fluid, insulin, and glucocorticoid administration on postoperative glucose levels are variable. To the best of our knowledge, no other major systematic changes occurred in the preoperative care of patients with DM during the study period.

Conclusions

The findings of our QI initiative suggest that HCP adherence to preoperative DM management protocols may be a key contributor to DOS hyperglycemia and that ensuring HCP adherence may be as important as preoperative insulin dose adjustments. To our knowledge, this is the first study to report rates of HCP adherence to preoperative DM management protocols and its effect on DOS hyperglycemia. We will focus future QI efforts on optimizing HCP adherence to preoperative DM management protocols at our institution.

Acknowledgments

We thank our endocrinology expert, Dr. Kristina Utzschneider, for her guidance in designing this improvement project and our academic research coach, Dr. Helene Starks, for her help in editing the manuscript.

Prolonged intensive care unit (ICU) stays, variably defined as > 48 h to > 14 days, are a known complication of cardiac surgery.^1-8 Prolonged stays are associated with higher resource utilization and higher mortality.^2,3,9-12 Although there are several cardiac surgery risk models that can be used preoperatively to identify patients at risk for prolonged ICU stay, factors that influence outcomes for patients who experience prolonged ICU stays are poorly understood.^2,13-19 Little information is available to inform discussions between health care practitioners (HCPs) and patients throughout a prolonged ICU stay, especially those ≥ 7 days.

As cardiac surgical complexity, patient age, and preexisting comorbidities have increased over time, so has the need to provide patients and HCPs with data to inform decision making, enhance prognostication, and set realistic expectations at varying time intervals during prolonged ICU stay. The purpose of this study was to evaluate short- and long-term outcomes in cardiac surgery patients after prolonged ICU stays at relevant time intervals (7, 14, 21, and 28 days) and to determine factors that may predict a patient’s outcome after a prolonged ICU stay.

Methods

The University of Michigan Health System Institutional Review Board approved this study and waived informed consent. We merged the University of Michigan Medical Center Society of Thoracic Surgeons (STS) database, which is updated periodically with late mortality, with elements of the electronic health record (EHR). Adult patients were included if they had cardiac surgery at the University of Michigan between January 2, 2001, and December 31, 2011. Late mortality was updated through December 1, 2014. Data are presented as frequency (%), mean (SD), and median (IQR) as appropriate. Bivariate comparisons between survivors and nonsurvivors were done with χ² or Fisher exact test for categorical data, Student t test for continuous normally distributed data, and Wilcoxon rank sum test for continuous not normally distributed data. To determine factors associated with operative mortality (death within 30 days of surgery or hospital discharge, whichever occurred later), we used logistic regression with forward selection. All available factors were initially entered in the models.

Separate logistic models were created based on all data available at days 7, 14, 21, and 28. Final models consisted of factors with statistically significant P values (< .05) and adjusted odds ratios (AORs) with 95% CIs that excluded 1. To determine factors associated with late mortality, we used a Cox proportional hazard model, which used data available at discharge and STS complications. As these complications did not include their timing, they could only be used in models created at discharge and not for days 7, 14, 21, and 28 models. Final models consisted of factors with P values < .05 and 95% CIs of the AORs or the hazard ratios (HRs) that excluded 1. As the EHR did not start recording data until January 2, 2004, and its capture of data remained incomplete for several years, rather than imputing these missing data or excluding these patients, we chose to create an extra categorical level for each factor to represent missing data. For continuous factors with missing data, we first converted the continuous data to terciles and the missing data became the fourth level.^20,21

The discrimination of the logistic models were determined by the c-statistic and for the Cox proportional hazards model with the Harrell concordance index (C index). Time trends were assessed with the Cochran-Armitage trend test. P < .05 was deemed statistically significant. Statistics were calculated with SPSS versions 21-23 or SAS 9.4.

Results

Of 8309 admissions to the ICU after cardiac surgery, 1174 (14%) had ICU stays ≥ 7 days, 386 (5%) ≥ 14 days, 201 (2%) ≥ 21 days, and 80 (0.9%) ≥ 28 days. The prolonged ICU study population was mostly male, White race, with a mean (SD) age of 62 (14) years. Patients had a variety of comorbidities, most notably 61% had hypertension and half had heart failure. Valve surgery (55%) was the most common procedure (n = 651). Twenty-nine percent required > 1 procedure (eAppendix 1).

The operative mortality

Using multivariable logistic regression to adjust for factors associated with mortality, we found that receiving mechanical ventilation on the day of analysis was associated with increased operative mortality with AOR increasing from 3.35 (95% CI, 2.82-3.98) for

Discussion

We found that operative mortality increased the longer the patient stayed in the ICU, ranging from 11% for ≥ 7 days to 35% for ≥ 28 days. We further found that in ICU survivors, median (IQR) survival was 10.7 (0.7) years. While previous studies have evaluated prolonged ICU stays, they have been limited by studying limited subpopulations, such as patients who are dependent on dialysis or octogenarians, or used a single cutoff to define prolonged ICU stays, variably defined from > 48 hours to > 14 days.^2-7,9-12,22 Our study is similar to others that used ≥ 2 cutoffs.^1,8 However, our study was novel by providing 4 cutoffs to improve temporal prediction of hospital outcomes. Unlike a study by Ryan and colleagues, which found no increase in mortality with longer stay (43.5% for ≥ 14 days and 45% for ≥ 28 days), our study findings are similar to those of Yu and colleagues (11.1% mortality for prolonged ICU stays of 1 to 2 weeks, 26.6% for 2 to 4 weeks, and 31% for > 4 weeks) and others (8%, 3 to 14 days; 40%, >14 days; 10%, 1 to 2 weeks; 25.7% > 2 weeks) in finding a progressively increased hospital mortality with longer ICU stays.^1,4,5,8 These differences may be related to different ICU populations or to improvements in care since Ryan and colleagues study was conducted.

Fewer studies have evaluated factors associated with mortality in cardiac surgery prolonged ICU stay patients. Our study is similar to other studies that evaluated risk factors by finding associations between a variety of comorbidities and process of care associated with both operative and long-term mortality; however, comparison between these studies is limited by the varying factors analyzed.^{1,3,5,6,8,9,11} We found that mechanical ventilation on days 7, 14, 21, and 28 was strongly associated with operative mortality, similar to noncardiac surgery patients and cardiac surgery patients.^6,23,24 While we found several processes of care, such as catecholamine use and transfusions to be associated with mortality, which is similar to other studies, notably, we did not find an association between renal replacement therapy and mortality.^1,25 While there is an association between renal replacement therapy and mortality in ICU patients, its status in cardiac surgery patients with prolonged ICU stays is less clear.²⁶ While Ryan and colleagues found an association between renal replacement therapy and hospital mortality in patients staying ≥ 14 days, they did not find it in patients staying ≥.

28 days.¹ Other studies of prolonged ICU stays for cardiac surgery patients have also failed to find an association between renal replacement therapy and mortality.^5,6,9 Importantly, practice that expedites liberation from mechanical ventilation, such as fast tracking, daily spontaneous breathing trials, extubation to noninvasive respiratory support, and pulmonary rehabilitation may all have potential to limit mechanical ventilation duration and improve hospital survival and deserve further study.^27-29Median (IQR) survival in hospital survivors was 10.7 (0.7) years, which is generally better than previously reported, but similar to that reported by Silberman and colleagues.^{2,4,6,8,11,12} Differences between these studies may relate to different patient populations within the cardiac surgery ICUs, definitions of prolonged ICU stays, or eras of care. Further study is needed to clarify these discrepancies. We found that cardiac transplantation and obesity were associated with the least risk of dying, while smoking, lung disease, and postoperative pneumonia were independently associated with increased hazard of dying. The obesity paradox, where obesity is protective, has been previously observed in cardiac surgery patients.³⁰

Strengths and Limitations

There are several limitations of this study. This is a single center study, and our patient population and processes of care may differ from other centers, limiting its generalizability. Notably, we do fewer coronary bypass operations and more aortic reconstructions and ventricular assist device insertions than do many other centers. Second, we did not have laboratory values for about one-third of patients (preceded EHR implementation). However, we were able to compensate for this by binning values and including missing data as an extra bin.^20,21

The main strength of this study is that we were able to combine disparate records to assess a large number of potential factors associated with both operative and long-term mortality. This produced models that had good to very good discrimination. By producing models at 7, 14, 21, and 28 days to predict operative mortality and a model at discharge, it may help to provide objective data to facilitate conversations with patients and their families. However, further studies to externally validate these models should be conducted.

Conclusions

We found that longer prolonged ICU stays are associated with both operative and late mortality. Receiving mechanical ventilation on days 7, 14, 21, or 28 was strongly associated with operative mortality.

Prolonged intensive care unit (ICU) stays, variably defined as > 48 h to > 14 days, are a known complication of cardiac surgery.^1-8 Prolonged stays are associated with higher resource utilization and higher mortality.^2,3,9-12 Although there are several cardiac surgery risk models that can be used preoperatively to identify patients at risk for prolonged ICU stay, factors that influence outcomes for patients who experience prolonged ICU stays are poorly understood.^2,13-19 Little information is available to inform discussions between health care practitioners (HCPs) and patients throughout a prolonged ICU stay, especially those ≥ 7 days.

As cardiac surgical complexity, patient age, and preexisting comorbidities have increased over time, so has the need to provide patients and HCPs with data to inform decision making, enhance prognostication, and set realistic expectations at varying time intervals during prolonged ICU stay. The purpose of this study was to evaluate short- and long-term outcomes in cardiac surgery patients after prolonged ICU stays at relevant time intervals (7, 14, 21, and 28 days) and to determine factors that may predict a patient’s outcome after a prolonged ICU stay.

Methods

The University of Michigan Health System Institutional Review Board approved this study and waived informed consent. We merged the University of Michigan Medical Center Society of Thoracic Surgeons (STS) database, which is updated periodically with late mortality, with elements of the electronic health record (EHR). Adult patients were included if they had cardiac surgery at the University of Michigan between January 2, 2001, and December 31, 2011. Late mortality was updated through December 1, 2014. Data are presented as frequency (%), mean (SD), and median (IQR) as appropriate. Bivariate comparisons between survivors and nonsurvivors were done with χ² or Fisher exact test for categorical data, Student t test for continuous normally distributed data, and Wilcoxon rank sum test for continuous not normally distributed data. To determine factors associated with operative mortality (death within 30 days of surgery or hospital discharge, whichever occurred later), we used logistic regression with forward selection. All available factors were initially entered in the models.

Separate logistic models were created based on all data available at days 7, 14, 21, and 28. Final models consisted of factors with statistically significant P values (< .05) and adjusted odds ratios (AORs) with 95% CIs that excluded 1. To determine factors associated with late mortality, we used a Cox proportional hazard model, which used data available at discharge and STS complications. As these complications did not include their timing, they could only be used in models created at discharge and not for days 7, 14, 21, and 28 models. Final models consisted of factors with P values < .05 and 95% CIs of the AORs or the hazard ratios (HRs) that excluded 1. As the EHR did not start recording data until January 2, 2004, and its capture of data remained incomplete for several years, rather than imputing these missing data or excluding these patients, we chose to create an extra categorical level for each factor to represent missing data. For continuous factors with missing data, we first converted the continuous data to terciles and the missing data became the fourth level.^20,21

The discrimination of the logistic models were determined by the c-statistic and for the Cox proportional hazards model with the Harrell concordance index (C index). Time trends were assessed with the Cochran-Armitage trend test. P < .05 was deemed statistically significant. Statistics were calculated with SPSS versions 21-23 or SAS 9.4.

Results

Of 8309 admissions to the ICU after cardiac surgery, 1174 (14%) had ICU stays ≥ 7 days, 386 (5%) ≥ 14 days, 201 (2%) ≥ 21 days, and 80 (0.9%) ≥ 28 days. The prolonged ICU study population was mostly male, White race, with a mean (SD) age of 62 (14) years. Patients had a variety of comorbidities, most notably 61% had hypertension and half had heart failure. Valve surgery (55%) was the most common procedure (n = 651). Twenty-nine percent required > 1 procedure (eAppendix 1).

The operative mortality

Using multivariable logistic regression to adjust for factors associated with mortality, we found that receiving mechanical ventilation on the day of analysis was associated with increased operative mortality with AOR increasing from 3.35 (95% CI, 2.82-3.98) for

Discussion

We found that operative mortality increased the longer the patient stayed in the ICU, ranging from 11% for ≥ 7 days to 35% for ≥ 28 days. We further found that in ICU survivors, median (IQR) survival was 10.7 (0.7) years. While previous studies have evaluated prolonged ICU stays, they have been limited by studying limited subpopulations, such as patients who are dependent on dialysis or octogenarians, or used a single cutoff to define prolonged ICU stays, variably defined from > 48 hours to > 14 days.^2-7,9-12,22 Our study is similar to others that used ≥ 2 cutoffs.^1,8 However, our study was novel by providing 4 cutoffs to improve temporal prediction of hospital outcomes. Unlike a study by Ryan and colleagues, which found no increase in mortality with longer stay (43.5% for ≥ 14 days and 45% for ≥ 28 days), our study findings are similar to those of Yu and colleagues (11.1% mortality for prolonged ICU stays of 1 to 2 weeks, 26.6% for 2 to 4 weeks, and 31% for > 4 weeks) and others (8%, 3 to 14 days; 40%, >14 days; 10%, 1 to 2 weeks; 25.7% > 2 weeks) in finding a progressively increased hospital mortality with longer ICU stays.^1,4,5,8 These differences may be related to different ICU populations or to improvements in care since Ryan and colleagues study was conducted.

Fewer studies have evaluated factors associated with mortality in cardiac surgery prolonged ICU stay patients. Our study is similar to other studies that evaluated risk factors by finding associations between a variety of comorbidities and process of care associated with both operative and long-term mortality; however, comparison between these studies is limited by the varying factors analyzed.^{1,3,5,6,8,9,11} We found that mechanical ventilation on days 7, 14, 21, and 28 was strongly associated with operative mortality, similar to noncardiac surgery patients and cardiac surgery patients.^6,23,24 While we found several processes of care, such as catecholamine use and transfusions to be associated with mortality, which is similar to other studies, notably, we did not find an association between renal replacement therapy and mortality.^1,25 While there is an association between renal replacement therapy and mortality in ICU patients, its status in cardiac surgery patients with prolonged ICU stays is less clear.²⁶ While Ryan and colleagues found an association between renal replacement therapy and hospital mortality in patients staying ≥ 14 days, they did not find it in patients staying ≥.

28 days.¹ Other studies of prolonged ICU stays for cardiac surgery patients have also failed to find an association between renal replacement therapy and mortality.^5,6,9 Importantly, practice that expedites liberation from mechanical ventilation, such as fast tracking, daily spontaneous breathing trials, extubation to noninvasive respiratory support, and pulmonary rehabilitation may all have potential to limit mechanical ventilation duration and improve hospital survival and deserve further study.^27-29Median (IQR) survival in hospital survivors was 10.7 (0.7) years, which is generally better than previously reported, but similar to that reported by Silberman and colleagues.^{2,4,6,8,11,12} Differences between these studies may relate to different patient populations within the cardiac surgery ICUs, definitions of prolonged ICU stays, or eras of care. Further study is needed to clarify these discrepancies. We found that cardiac transplantation and obesity were associated with the least risk of dying, while smoking, lung disease, and postoperative pneumonia were independently associated with increased hazard of dying. The obesity paradox, where obesity is protective, has been previously observed in cardiac surgery patients.³⁰

Strengths and Limitations

There are several limitations of this study. This is a single center study, and our patient population and processes of care may differ from other centers, limiting its generalizability. Notably, we do fewer coronary bypass operations and more aortic reconstructions and ventricular assist device insertions than do many other centers. Second, we did not have laboratory values for about one-third of patients (preceded EHR implementation). However, we were able to compensate for this by binning values and including missing data as an extra bin.^20,21

The main strength of this study is that we were able to combine disparate records to assess a large number of potential factors associated with both operative and long-term mortality. This produced models that had good to very good discrimination. By producing models at 7, 14, 21, and 28 days to predict operative mortality and a model at discharge, it may help to provide objective data to facilitate conversations with patients and their families. However, further studies to externally validate these models should be conducted.

Conclusions

We found that longer prolonged ICU stays are associated with both operative and late mortality. Receiving mechanical ventilation on days 7, 14, 21, or 28 was strongly associated with operative mortality.

Prolonged intensive care unit (ICU) stays, variably defined as > 48 h to > 14 days, are a known complication of cardiac surgery.^1-8 Prolonged stays are associated with higher resource utilization and higher mortality.^2,3,9-12 Although there are several cardiac surgery risk models that can be used preoperatively to identify patients at risk for prolonged ICU stay, factors that influence outcomes for patients who experience prolonged ICU stays are poorly understood.^2,13-19 Little information is available to inform discussions between health care practitioners (HCPs) and patients throughout a prolonged ICU stay, especially those ≥ 7 days.

As cardiac surgical complexity, patient age, and preexisting comorbidities have increased over time, so has the need to provide patients and HCPs with data to inform decision making, enhance prognostication, and set realistic expectations at varying time intervals during prolonged ICU stay. The purpose of this study was to evaluate short- and long-term outcomes in cardiac surgery patients after prolonged ICU stays at relevant time intervals (7, 14, 21, and 28 days) and to determine factors that may predict a patient’s outcome after a prolonged ICU stay.

Methods

The University of Michigan Health System Institutional Review Board approved this study and waived informed consent. We merged the University of Michigan Medical Center Society of Thoracic Surgeons (STS) database, which is updated periodically with late mortality, with elements of the electronic health record (EHR). Adult patients were included if they had cardiac surgery at the University of Michigan between January 2, 2001, and December 31, 2011. Late mortality was updated through December 1, 2014. Data are presented as frequency (%), mean (SD), and median (IQR) as appropriate. Bivariate comparisons between survivors and nonsurvivors were done with χ² or Fisher exact test for categorical data, Student t test for continuous normally distributed data, and Wilcoxon rank sum test for continuous not normally distributed data. To determine factors associated with operative mortality (death within 30 days of surgery or hospital discharge, whichever occurred later), we used logistic regression with forward selection. All available factors were initially entered in the models.

Separate logistic models were created based on all data available at days 7, 14, 21, and 28. Final models consisted of factors with statistically significant P values (< .05) and adjusted odds ratios (AORs) with 95% CIs that excluded 1. To determine factors associated with late mortality, we used a Cox proportional hazard model, which used data available at discharge and STS complications. As these complications did not include their timing, they could only be used in models created at discharge and not for days 7, 14, 21, and 28 models. Final models consisted of factors with P values < .05 and 95% CIs of the AORs or the hazard ratios (HRs) that excluded 1. As the EHR did not start recording data until January 2, 2004, and its capture of data remained incomplete for several years, rather than imputing these missing data or excluding these patients, we chose to create an extra categorical level for each factor to represent missing data. For continuous factors with missing data, we first converted the continuous data to terciles and the missing data became the fourth level.^20,21

The discrimination of the logistic models were determined by the c-statistic and for the Cox proportional hazards model with the Harrell concordance index (C index). Time trends were assessed with the Cochran-Armitage trend test. P < .05 was deemed statistically significant. Statistics were calculated with SPSS versions 21-23 or SAS 9.4.

Results

Of 8309 admissions to the ICU after cardiac surgery, 1174 (14%) had ICU stays ≥ 7 days, 386 (5%) ≥ 14 days, 201 (2%) ≥ 21 days, and 80 (0.9%) ≥ 28 days. The prolonged ICU study population was mostly male, White race, with a mean (SD) age of 62 (14) years. Patients had a variety of comorbidities, most notably 61% had hypertension and half had heart failure. Valve surgery (55%) was the most common procedure (n = 651). Twenty-nine percent required > 1 procedure (eAppendix 1).

The operative mortality

Using multivariable logistic regression to adjust for factors associated with mortality, we found that receiving mechanical ventilation on the day of analysis was associated with increased operative mortality with AOR increasing from 3.35 (95% CI, 2.82-3.98) for

Discussion

We found that operative mortality increased the longer the patient stayed in the ICU, ranging from 11% for ≥ 7 days to 35% for ≥ 28 days. We further found that in ICU survivors, median (IQR) survival was 10.7 (0.7) years. While previous studies have evaluated prolonged ICU stays, they have been limited by studying limited subpopulations, such as patients who are dependent on dialysis or octogenarians, or used a single cutoff to define prolonged ICU stays, variably defined from > 48 hours to > 14 days.^2-7,9-12,22 Our study is similar to others that used ≥ 2 cutoffs.^1,8 However, our study was novel by providing 4 cutoffs to improve temporal prediction of hospital outcomes. Unlike a study by Ryan and colleagues, which found no increase in mortality with longer stay (43.5% for ≥ 14 days and 45% for ≥ 28 days), our study findings are similar to those of Yu and colleagues (11.1% mortality for prolonged ICU stays of 1 to 2 weeks, 26.6% for 2 to 4 weeks, and 31% for > 4 weeks) and others (8%, 3 to 14 days; 40%, >14 days; 10%, 1 to 2 weeks; 25.7% > 2 weeks) in finding a progressively increased hospital mortality with longer ICU stays.^1,4,5,8 These differences may be related to different ICU populations or to improvements in care since Ryan and colleagues study was conducted.

Fewer studies have evaluated factors associated with mortality in cardiac surgery prolonged ICU stay patients. Our study is similar to other studies that evaluated risk factors by finding associations between a variety of comorbidities and process of care associated with both operative and long-term mortality; however, comparison between these studies is limited by the varying factors analyzed.^{1,3,5,6,8,9,11} We found that mechanical ventilation on days 7, 14, 21, and 28 was strongly associated with operative mortality, similar to noncardiac surgery patients and cardiac surgery patients.^6,23,24 While we found several processes of care, such as catecholamine use and transfusions to be associated with mortality, which is similar to other studies, notably, we did not find an association between renal replacement therapy and mortality.^1,25 While there is an association between renal replacement therapy and mortality in ICU patients, its status in cardiac surgery patients with prolonged ICU stays is less clear.²⁶ While Ryan and colleagues found an association between renal replacement therapy and hospital mortality in patients staying ≥ 14 days, they did not find it in patients staying ≥.

28 days.¹ Other studies of prolonged ICU stays for cardiac surgery patients have also failed to find an association between renal replacement therapy and mortality.^5,6,9 Importantly, practice that expedites liberation from mechanical ventilation, such as fast tracking, daily spontaneous breathing trials, extubation to noninvasive respiratory support, and pulmonary rehabilitation may all have potential to limit mechanical ventilation duration and improve hospital survival and deserve further study.^27-29Median (IQR) survival in hospital survivors was 10.7 (0.7) years, which is generally better than previously reported, but similar to that reported by Silberman and colleagues.^{2,4,6,8,11,12} Differences between these studies may relate to different patient populations within the cardiac surgery ICUs, definitions of prolonged ICU stays, or eras of care. Further study is needed to clarify these discrepancies. We found that cardiac transplantation and obesity were associated with the least risk of dying, while smoking, lung disease, and postoperative pneumonia were independently associated with increased hazard of dying. The obesity paradox, where obesity is protective, has been previously observed in cardiac surgery patients.³⁰

Strengths and Limitations

There are several limitations of this study. This is a single center study, and our patient population and processes of care may differ from other centers, limiting its generalizability. Notably, we do fewer coronary bypass operations and more aortic reconstructions and ventricular assist device insertions than do many other centers. Second, we did not have laboratory values for about one-third of patients (preceded EHR implementation). However, we were able to compensate for this by binning values and including missing data as an extra bin.^20,21

The main strength of this study is that we were able to combine disparate records to assess a large number of potential factors associated with both operative and long-term mortality. This produced models that had good to very good discrimination. By producing models at 7, 14, 21, and 28 days to predict operative mortality and a model at discharge, it may help to provide objective data to facilitate conversations with patients and their families. However, further studies to externally validate these models should be conducted.

Conclusions

We found that longer prolonged ICU stays are associated with both operative and late mortality. Receiving mechanical ventilation on days 7, 14, 21, or 28 was strongly associated with operative mortality.

PHILADELPHIA – Women with spondyloarthritis (SpA) who are desiring pregnancy may want to consider decreasing use or discontinuing use (with supervision) of nonsteroidal anti-inflammatory drugs before conception, new data suggest.

Researchers have found a connection between NSAID use and age and a significantly longer time to conception among women with spondyloarthritis. Sabrina Hamroun, MMed, with the rheumatology department at the University Hospital Cochin, Paris, presented the findings during a press conference at the annual meeting of the American College of Rheumatology.

Subfertility group took an average of 16 months to get pregnant

Subfertility was observed in 40 (45.4%) of the women, with an average time to conception of 16.1 months. A woman was considered subfertile if her time to conception was more than 12 months or if she did not become pregnant.

The average preconception Bath Ankylosing Spondylitis Disease Activity Index score was 2.9 (+/- 2.1), the authors noted. The average age of the participants was 32 years.

Twenty-three patients were treated with NSAIDs, eight with corticosteroids, 12 with conventional synthetic disease-modifying antirheumatic drugs, and 61 with biologics.

Researchers adjusted for factors including age, body mass index, disease duration and severity, smoking, form of SpA (axial, peripheral, or both), and medication in the preconception period.

They found significant associations between longer time to conception and age (hazard ratio, 1.22; 95% confidence interval, 1.08-1.40; P < .001), and a much higher hazard ratio with the use of NSAIDs during preconception (HR, 3.01; 95% CI, 2.15-3.85; P = .01).

Some data unavailable

Ms. Hamroun acknowledged that no data were available on the frequency of sexual intercourse or quality of life, factors that could affect time to conception. Women were asked when they discontinued contraceptive use and actively began trying to become pregnant.

She stated that information on the dose of NSAIDs used by the patients was incomplete, noting, “We were therefore unable to adjust the results of our statistical analyses on the dose used by patients.”

Additionally, because the study participants were patients at tertiary centers in France and had more severe disease, the results may not be generalizable to all women of childbearing age. Patients with less severe SpA are often managed in outpatient settings in France, she said.

When asked about alternatives to NSAIDs, Ms. Hamroun said that anti–tumor necrosis factor agents with low placental passage may be a good alternative “if a woman with long-standing difficulties to conceive needs a regular use of NSAIDs to control disease activity, in the absence of any other cause of subfertility.”

The patient’s age must also be considered, she noted.

“A therapeutic switch may be favored in a woman over 35 years of age, for example, whose fertility is already impaired by age,” Ms. Hamroun said.

As for the mechanism that might explain the effects of NSAIDs on conception, Ms. Hamroun said that prostaglandins are essential to ovulation and embryo implantation and explained that NSAIDs may work against ovulation and result in poor implantation (miscarriage) by blocking prostaglandins.

She pointed out that her results are in line with the ACR’s recommendation to discontinue NSAID use during the preconception period in women with SpA who are having difficulty conceiving.
 

Control before conception is important

Sinead Maguire, MD, a clinical and research fellow in the Spondylitis Program at Toronto (Ont.) Western Hospital who was not part of the study, said the study highlights the importance of optimizing disease control before conception.

“There are a number of things rheumatologists can do to support our SpA patients when they are trying to conceive,” she told this news organization. “One of the most important issues to address is ensuring their SpA is in remission and continues to remain so. For that reason, if a woman is requiring regular NSAIDs for symptom control, the results of this study might encourage me to consider a biologic agent sooner to ensure remission.”

She urged women who want to become pregnant to discuss medications with their rheumatologist before trying to conceive.

“It is very exciting to see studies such as this so that rheumatologists can provide answers to our patients’ questions with evidence-based advice,” she said.

Ms. Hamroun and several coauthors had no disclosures. Other coauthors disclosed relationships with companies including Merck/MSD, Novartis, Janssen, AbbVie/Abbott, Amgen, AstraZeneca, Biogen, Bristol-Myers Squibb, Galapagos, Eli Lilly, Novartis, and/or UCB. Dr. Maguire reports no relevant financial relationships.

A version of this article first appeared on Medscape.com.

PHILADELPHIA – Women with spondyloarthritis (SpA) who are desiring pregnancy may want to consider decreasing use or discontinuing use (with supervision) of nonsteroidal anti-inflammatory drugs before conception, new data suggest.

Researchers have found a connection between NSAID use and age and a significantly longer time to conception among women with spondyloarthritis. Sabrina Hamroun, MMed, with the rheumatology department at the University Hospital Cochin, Paris, presented the findings during a press conference at the annual meeting of the American College of Rheumatology.

Subfertility group took an average of 16 months to get pregnant

Subfertility was observed in 40 (45.4%) of the women, with an average time to conception of 16.1 months. A woman was considered subfertile if her time to conception was more than 12 months or if she did not become pregnant.

The average preconception Bath Ankylosing Spondylitis Disease Activity Index score was 2.9 (+/- 2.1), the authors noted. The average age of the participants was 32 years.

Twenty-three patients were treated with NSAIDs, eight with corticosteroids, 12 with conventional synthetic disease-modifying antirheumatic drugs, and 61 with biologics.

Researchers adjusted for factors including age, body mass index, disease duration and severity, smoking, form of SpA (axial, peripheral, or both), and medication in the preconception period.

They found significant associations between longer time to conception and age (hazard ratio, 1.22; 95% confidence interval, 1.08-1.40; P < .001), and a much higher hazard ratio with the use of NSAIDs during preconception (HR, 3.01; 95% CI, 2.15-3.85; P = .01).

Some data unavailable

Ms. Hamroun acknowledged that no data were available on the frequency of sexual intercourse or quality of life, factors that could affect time to conception. Women were asked when they discontinued contraceptive use and actively began trying to become pregnant.

She stated that information on the dose of NSAIDs used by the patients was incomplete, noting, “We were therefore unable to adjust the results of our statistical analyses on the dose used by patients.”

Additionally, because the study participants were patients at tertiary centers in France and had more severe disease, the results may not be generalizable to all women of childbearing age. Patients with less severe SpA are often managed in outpatient settings in France, she said.

When asked about alternatives to NSAIDs, Ms. Hamroun said that anti–tumor necrosis factor agents with low placental passage may be a good alternative “if a woman with long-standing difficulties to conceive needs a regular use of NSAIDs to control disease activity, in the absence of any other cause of subfertility.”

The patient’s age must also be considered, she noted.

“A therapeutic switch may be favored in a woman over 35 years of age, for example, whose fertility is already impaired by age,” Ms. Hamroun said.

As for the mechanism that might explain the effects of NSAIDs on conception, Ms. Hamroun said that prostaglandins are essential to ovulation and embryo implantation and explained that NSAIDs may work against ovulation and result in poor implantation (miscarriage) by blocking prostaglandins.

She pointed out that her results are in line with the ACR’s recommendation to discontinue NSAID use during the preconception period in women with SpA who are having difficulty conceiving.
 

Control before conception is important

Sinead Maguire, MD, a clinical and research fellow in the Spondylitis Program at Toronto (Ont.) Western Hospital who was not part of the study, said the study highlights the importance of optimizing disease control before conception.

“There are a number of things rheumatologists can do to support our SpA patients when they are trying to conceive,” she told this news organization. “One of the most important issues to address is ensuring their SpA is in remission and continues to remain so. For that reason, if a woman is requiring regular NSAIDs for symptom control, the results of this study might encourage me to consider a biologic agent sooner to ensure remission.”

She urged women who want to become pregnant to discuss medications with their rheumatologist before trying to conceive.

“It is very exciting to see studies such as this so that rheumatologists can provide answers to our patients’ questions with evidence-based advice,” she said.

Ms. Hamroun and several coauthors had no disclosures. Other coauthors disclosed relationships with companies including Merck/MSD, Novartis, Janssen, AbbVie/Abbott, Amgen, AstraZeneca, Biogen, Bristol-Myers Squibb, Galapagos, Eli Lilly, Novartis, and/or UCB. Dr. Maguire reports no relevant financial relationships.

A version of this article first appeared on Medscape.com.

PHILADELPHIA – Women with spondyloarthritis (SpA) who are desiring pregnancy may want to consider decreasing use or discontinuing use (with supervision) of nonsteroidal anti-inflammatory drugs before conception, new data suggest.

Researchers have found a connection between NSAID use and age and a significantly longer time to conception among women with spondyloarthritis. Sabrina Hamroun, MMed, with the rheumatology department at the University Hospital Cochin, Paris, presented the findings during a press conference at the annual meeting of the American College of Rheumatology.

Subfertility group took an average of 16 months to get pregnant

Subfertility was observed in 40 (45.4%) of the women, with an average time to conception of 16.1 months. A woman was considered subfertile if her time to conception was more than 12 months or if she did not become pregnant.

The average preconception Bath Ankylosing Spondylitis Disease Activity Index score was 2.9 (+/- 2.1), the authors noted. The average age of the participants was 32 years.

Twenty-three patients were treated with NSAIDs, eight with corticosteroids, 12 with conventional synthetic disease-modifying antirheumatic drugs, and 61 with biologics.

Researchers adjusted for factors including age, body mass index, disease duration and severity, smoking, form of SpA (axial, peripheral, or both), and medication in the preconception period.

They found significant associations between longer time to conception and age (hazard ratio, 1.22; 95% confidence interval, 1.08-1.40; P < .001), and a much higher hazard ratio with the use of NSAIDs during preconception (HR, 3.01; 95% CI, 2.15-3.85; P = .01).

Some data unavailable

Ms. Hamroun acknowledged that no data were available on the frequency of sexual intercourse or quality of life, factors that could affect time to conception. Women were asked when they discontinued contraceptive use and actively began trying to become pregnant.

She stated that information on the dose of NSAIDs used by the patients was incomplete, noting, “We were therefore unable to adjust the results of our statistical analyses on the dose used by patients.”

Additionally, because the study participants were patients at tertiary centers in France and had more severe disease, the results may not be generalizable to all women of childbearing age. Patients with less severe SpA are often managed in outpatient settings in France, she said.

When asked about alternatives to NSAIDs, Ms. Hamroun said that anti–tumor necrosis factor agents with low placental passage may be a good alternative “if a woman with long-standing difficulties to conceive needs a regular use of NSAIDs to control disease activity, in the absence of any other cause of subfertility.”

The patient’s age must also be considered, she noted.

“A therapeutic switch may be favored in a woman over 35 years of age, for example, whose fertility is already impaired by age,” Ms. Hamroun said.

As for the mechanism that might explain the effects of NSAIDs on conception, Ms. Hamroun said that prostaglandins are essential to ovulation and embryo implantation and explained that NSAIDs may work against ovulation and result in poor implantation (miscarriage) by blocking prostaglandins.

She pointed out that her results are in line with the ACR’s recommendation to discontinue NSAID use during the preconception period in women with SpA who are having difficulty conceiving.
 

Control before conception is important

Sinead Maguire, MD, a clinical and research fellow in the Spondylitis Program at Toronto (Ont.) Western Hospital who was not part of the study, said the study highlights the importance of optimizing disease control before conception.

“There are a number of things rheumatologists can do to support our SpA patients when they are trying to conceive,” she told this news organization. “One of the most important issues to address is ensuring their SpA is in remission and continues to remain so. For that reason, if a woman is requiring regular NSAIDs for symptom control, the results of this study might encourage me to consider a biologic agent sooner to ensure remission.”

She urged women who want to become pregnant to discuss medications with their rheumatologist before trying to conceive.

“It is very exciting to see studies such as this so that rheumatologists can provide answers to our patients’ questions with evidence-based advice,” she said.

Ms. Hamroun and several coauthors had no disclosures. Other coauthors disclosed relationships with companies including Merck/MSD, Novartis, Janssen, AbbVie/Abbott, Amgen, AstraZeneca, Biogen, Bristol-Myers Squibb, Galapagos, Eli Lilly, Novartis, and/or UCB. Dr. Maguire reports no relevant financial relationships.

A version of this article first appeared on Medscape.com.

PHILADELPHIA – Combining the immunomodulatory agent mycophenolate with the antifibrotic pirfenidone led to more rapid improvement and showed a trend to be more effective than mycophenolate mofetil alone for treating the signs and symptoms of scleroderma-related interstitial lung disease, but the combination therapy came with an increase in side effects, according to results from the Scleroderma Lung Study III.

Dinesh Khanna, MBBS, MSc, of the University of Michigan, Ann Arbor, presented the results at the annual meeting of the American College of Rheumatology. He noted some problems with the study – namely its small size, enrolling only 51 patients, about one-third of its original goal. But he also said it showed a potential signal for efficacy and that the study itself could serve as a “template” for future studies of combination mycophenolate mofetil (MMF) plus pirfenidone therapy for scleroderma-related interstitial lung disease (SSc-ILD).

“The pirfenidone patients had quite a bit more GI side effects and photosensitivity, and those are known side effects,” Dr. Khanna said in an interview. “So the combination therapy had more side effects but trends to higher efficacy.”

The design of SLS-III, a phase 2 clinical trial, was a challenge, Dr. Khanna explained. The goal was to enroll 150 SSc-ILD patients who hadn’t had any previous treatment for their disease. Finding those patients proved difficult. “In fact, if you look at the recent history, 70% of the patients with early diffuse scleroderma are on MMF,” he said in his presentation. Compounding low study enrollment was the intervening COVID-19 pandemic, he added.
 

Testing a faster-acting combination

Nonetheless, the trial managed to enroll 27 patients in the combination therapy group and 24 in the MMF-plus-placebo group and compared their outcomes over 18 months. Study dosing was 1,500 mg MMF twice daily and pirfenidone 801 mg three times daily, titrated to the tolerable dose.

Despite the study’s being underpowered, Dr. Khanna said, it still reported some notable outcomes that merit further investigation. “I think what was intriguing in the study was the long-term benefit in the patient-reported outcomes and the structural changes,” he said in the interview.

Among those notable outcomes was a clinically significant change in forced vital capacity (FVC) percentage for the combination vs. the placebo groups: 2.24% vs. 2.09%. He also noted that the combination group saw a somewhat more robust improvement in FVC at six months: 2.59% (± 0.98%) vs. 0.92% (± 1.1%) in the placebo group.

The combination group showed greater improvements in high-resolution computed tomography-evaluated lung involvement and lung fibrosis and patient-reported outcomes, including a statistically significant 3.67-point greater improvement in PROMIS-29 physical function score (4.42 vs. 0.75).

The patients on combination therapy had higher rates of serious adverse events (SAEs), and seven discontinued one or both study drugs early, all in the combined arm. Four combination therapy patients had six SAEs, compared to two placebo patients with three SAEs. In the combination group, SAEs included chest pain, herpes zoster ophthalmicus, nodular basal cell cancer, marginal zone B cell lymphoma, renal crisis, and dyspnea. SAEs in the placebo group were colitis, COVID-19 and hypoxic respiratory failure.

Study design challenges

Nonetheless, Dr. Khanna said the SLS-III data are consistent with the SLS-II findings, with mean improvements in FVC of 2.24% and 2.1%, respectively.

“The next study may be able to replicate what we tried to do, keeping in mind that there are really no MMF-naive patients who are walking around,” Dr. Khanna said. “So the challenge is about the feasibility of recruiting within a trial vs. trying to show a statistical difference between the drug and placebo.”

This study could serve as a foundation for future studies of MMF in patients with SSc-ILD, Robert Spiera, MD, of the Hospital for Special Surgery in New York, said in an interview. “There are lessons to be learned both from the study but also from prior studies looking at MMF use in the background in patients treated with other drugs in clinical trials,” he said.

Dr. Spiera noted that the study had other challenges besides the difficulty in recruiting patients who hadn’t been on MMF therapy. “A great challenge is that the benefit with regard to the impact on the lungs from MMF seems most prominent in the first 6 months to a year to even 2 years that somebody is on the drug,” he said.

The other challenge with this study is that a large proportion of patients had limited systemic disease and relatively lower levels of skin disease compared with other studies of patients on MMF, Dr. Spiera said.

“The optimal treatment of scleroderma-associated lung disease remains a very important and not-adequately met need,” he said. “Particularly, we’re looking for drugs that are tolerable in a patient population that are very prone to GI side effects in general. This study and others have taught us a lot about trial design, and I think more globally this will allow us to move this field forward.”

Dr. Khanna disclosed relationships with Actelion, Boehringer Ingelheim, Bristol-Myers Squibb, CSL Behring, Horizon Therapeutics USA, Janssen Global Services, Prometheus Biosciences, Mitsubishi Tanabe Pharma Corp., Genentech/Roche, Theraly, and Pfizer. Genentech provided funding for the study and pirfenidone and placebo drugs at no cost.

Dr. Spiera disclosed relationships with GlaxoSmithKline, Boehringer-Ingelheim, Corbus Pharmaceutical, InflaRx, AbbVie/Abbott, Sanofi, Novartis, Chemocentryx, Roche and Vera.

PHILADELPHIA – Combining the immunomodulatory agent mycophenolate with the antifibrotic pirfenidone led to more rapid improvement and showed a trend to be more effective than mycophenolate mofetil alone for treating the signs and symptoms of scleroderma-related interstitial lung disease, but the combination therapy came with an increase in side effects, according to results from the Scleroderma Lung Study III.

Dinesh Khanna, MBBS, MSc, of the University of Michigan, Ann Arbor, presented the results at the annual meeting of the American College of Rheumatology. He noted some problems with the study – namely its small size, enrolling only 51 patients, about one-third of its original goal. But he also said it showed a potential signal for efficacy and that the study itself could serve as a “template” for future studies of combination mycophenolate mofetil (MMF) plus pirfenidone therapy for scleroderma-related interstitial lung disease (SSc-ILD).

“The pirfenidone patients had quite a bit more GI side effects and photosensitivity, and those are known side effects,” Dr. Khanna said in an interview. “So the combination therapy had more side effects but trends to higher efficacy.”

The design of SLS-III, a phase 2 clinical trial, was a challenge, Dr. Khanna explained. The goal was to enroll 150 SSc-ILD patients who hadn’t had any previous treatment for their disease. Finding those patients proved difficult. “In fact, if you look at the recent history, 70% of the patients with early diffuse scleroderma are on MMF,” he said in his presentation. Compounding low study enrollment was the intervening COVID-19 pandemic, he added.
 

Testing a faster-acting combination

Nonetheless, the trial managed to enroll 27 patients in the combination therapy group and 24 in the MMF-plus-placebo group and compared their outcomes over 18 months. Study dosing was 1,500 mg MMF twice daily and pirfenidone 801 mg three times daily, titrated to the tolerable dose.

Despite the study’s being underpowered, Dr. Khanna said, it still reported some notable outcomes that merit further investigation. “I think what was intriguing in the study was the long-term benefit in the patient-reported outcomes and the structural changes,” he said in the interview.

Among those notable outcomes was a clinically significant change in forced vital capacity (FVC) percentage for the combination vs. the placebo groups: 2.24% vs. 2.09%. He also noted that the combination group saw a somewhat more robust improvement in FVC at six months: 2.59% (± 0.98%) vs. 0.92% (± 1.1%) in the placebo group.

The combination group showed greater improvements in high-resolution computed tomography-evaluated lung involvement and lung fibrosis and patient-reported outcomes, including a statistically significant 3.67-point greater improvement in PROMIS-29 physical function score (4.42 vs. 0.75).

The patients on combination therapy had higher rates of serious adverse events (SAEs), and seven discontinued one or both study drugs early, all in the combined arm. Four combination therapy patients had six SAEs, compared to two placebo patients with three SAEs. In the combination group, SAEs included chest pain, herpes zoster ophthalmicus, nodular basal cell cancer, marginal zone B cell lymphoma, renal crisis, and dyspnea. SAEs in the placebo group were colitis, COVID-19 and hypoxic respiratory failure.

Study design challenges

Nonetheless, Dr. Khanna said the SLS-III data are consistent with the SLS-II findings, with mean improvements in FVC of 2.24% and 2.1%, respectively.

“The next study may be able to replicate what we tried to do, keeping in mind that there are really no MMF-naive patients who are walking around,” Dr. Khanna said. “So the challenge is about the feasibility of recruiting within a trial vs. trying to show a statistical difference between the drug and placebo.”

This study could serve as a foundation for future studies of MMF in patients with SSc-ILD, Robert Spiera, MD, of the Hospital for Special Surgery in New York, said in an interview. “There are lessons to be learned both from the study but also from prior studies looking at MMF use in the background in patients treated with other drugs in clinical trials,” he said.

Dr. Spiera noted that the study had other challenges besides the difficulty in recruiting patients who hadn’t been on MMF therapy. “A great challenge is that the benefit with regard to the impact on the lungs from MMF seems most prominent in the first 6 months to a year to even 2 years that somebody is on the drug,” he said.

The other challenge with this study is that a large proportion of patients had limited systemic disease and relatively lower levels of skin disease compared with other studies of patients on MMF, Dr. Spiera said.

“The optimal treatment of scleroderma-associated lung disease remains a very important and not-adequately met need,” he said. “Particularly, we’re looking for drugs that are tolerable in a patient population that are very prone to GI side effects in general. This study and others have taught us a lot about trial design, and I think more globally this will allow us to move this field forward.”

Dr. Khanna disclosed relationships with Actelion, Boehringer Ingelheim, Bristol-Myers Squibb, CSL Behring, Horizon Therapeutics USA, Janssen Global Services, Prometheus Biosciences, Mitsubishi Tanabe Pharma Corp., Genentech/Roche, Theraly, and Pfizer. Genentech provided funding for the study and pirfenidone and placebo drugs at no cost.

Dr. Spiera disclosed relationships with GlaxoSmithKline, Boehringer-Ingelheim, Corbus Pharmaceutical, InflaRx, AbbVie/Abbott, Sanofi, Novartis, Chemocentryx, Roche and Vera.

PHILADELPHIA – Combining the immunomodulatory agent mycophenolate with the antifibrotic pirfenidone led to more rapid improvement and showed a trend to be more effective than mycophenolate mofetil alone for treating the signs and symptoms of scleroderma-related interstitial lung disease, but the combination therapy came with an increase in side effects, according to results from the Scleroderma Lung Study III.

Dinesh Khanna, MBBS, MSc, of the University of Michigan, Ann Arbor, presented the results at the annual meeting of the American College of Rheumatology. He noted some problems with the study – namely its small size, enrolling only 51 patients, about one-third of its original goal. But he also said it showed a potential signal for efficacy and that the study itself could serve as a “template” for future studies of combination mycophenolate mofetil (MMF) plus pirfenidone therapy for scleroderma-related interstitial lung disease (SSc-ILD).

“The pirfenidone patients had quite a bit more GI side effects and photosensitivity, and those are known side effects,” Dr. Khanna said in an interview. “So the combination therapy had more side effects but trends to higher efficacy.”

The design of SLS-III, a phase 2 clinical trial, was a challenge, Dr. Khanna explained. The goal was to enroll 150 SSc-ILD patients who hadn’t had any previous treatment for their disease. Finding those patients proved difficult. “In fact, if you look at the recent history, 70% of the patients with early diffuse scleroderma are on MMF,” he said in his presentation. Compounding low study enrollment was the intervening COVID-19 pandemic, he added.
 

Testing a faster-acting combination

Nonetheless, the trial managed to enroll 27 patients in the combination therapy group and 24 in the MMF-plus-placebo group and compared their outcomes over 18 months. Study dosing was 1,500 mg MMF twice daily and pirfenidone 801 mg three times daily, titrated to the tolerable dose.

Despite the study’s being underpowered, Dr. Khanna said, it still reported some notable outcomes that merit further investigation. “I think what was intriguing in the study was the long-term benefit in the patient-reported outcomes and the structural changes,” he said in the interview.

Among those notable outcomes was a clinically significant change in forced vital capacity (FVC) percentage for the combination vs. the placebo groups: 2.24% vs. 2.09%. He also noted that the combination group saw a somewhat more robust improvement in FVC at six months: 2.59% (± 0.98%) vs. 0.92% (± 1.1%) in the placebo group.

The combination group showed greater improvements in high-resolution computed tomography-evaluated lung involvement and lung fibrosis and patient-reported outcomes, including a statistically significant 3.67-point greater improvement in PROMIS-29 physical function score (4.42 vs. 0.75).

The patients on combination therapy had higher rates of serious adverse events (SAEs), and seven discontinued one or both study drugs early, all in the combined arm. Four combination therapy patients had six SAEs, compared to two placebo patients with three SAEs. In the combination group, SAEs included chest pain, herpes zoster ophthalmicus, nodular basal cell cancer, marginal zone B cell lymphoma, renal crisis, and dyspnea. SAEs in the placebo group were colitis, COVID-19 and hypoxic respiratory failure.

Study design challenges

Nonetheless, Dr. Khanna said the SLS-III data are consistent with the SLS-II findings, with mean improvements in FVC of 2.24% and 2.1%, respectively.

“The next study may be able to replicate what we tried to do, keeping in mind that there are really no MMF-naive patients who are walking around,” Dr. Khanna said. “So the challenge is about the feasibility of recruiting within a trial vs. trying to show a statistical difference between the drug and placebo.”

This study could serve as a foundation for future studies of MMF in patients with SSc-ILD, Robert Spiera, MD, of the Hospital for Special Surgery in New York, said in an interview. “There are lessons to be learned both from the study but also from prior studies looking at MMF use in the background in patients treated with other drugs in clinical trials,” he said.

Dr. Spiera noted that the study had other challenges besides the difficulty in recruiting patients who hadn’t been on MMF therapy. “A great challenge is that the benefit with regard to the impact on the lungs from MMF seems most prominent in the first 6 months to a year to even 2 years that somebody is on the drug,” he said.

The other challenge with this study is that a large proportion of patients had limited systemic disease and relatively lower levels of skin disease compared with other studies of patients on MMF, Dr. Spiera said.

“The optimal treatment of scleroderma-associated lung disease remains a very important and not-adequately met need,” he said. “Particularly, we’re looking for drugs that are tolerable in a patient population that are very prone to GI side effects in general. This study and others have taught us a lot about trial design, and I think more globally this will allow us to move this field forward.”

Dr. Khanna disclosed relationships with Actelion, Boehringer Ingelheim, Bristol-Myers Squibb, CSL Behring, Horizon Therapeutics USA, Janssen Global Services, Prometheus Biosciences, Mitsubishi Tanabe Pharma Corp., Genentech/Roche, Theraly, and Pfizer. Genentech provided funding for the study and pirfenidone and placebo drugs at no cost.

Dr. Spiera disclosed relationships with GlaxoSmithKline, Boehringer-Ingelheim, Corbus Pharmaceutical, InflaRx, AbbVie/Abbott, Sanofi, Novartis, Chemocentryx, Roche and Vera.

In my last column, I discussed blogging as a great way to capture the attention of anyone interested in your practice, especially prospective patients. If you are already blogging – or would like to consider a less crowded and competitive activity – podcasting might be the answer. At this writing (November 2022), more than 600 million blogs are online, compared with about 2 million podcasts, and relatively few of them are run by physicians. With podcasts, you have a better chance of standing out in a crowded online world.

Starting a podcast is not difficult, but there are several steps you need to go through before launching one.

As with blogging, start by outlining a long-range plan. Your general topic will probably be your specialty, but you will need to narrow your focus to a few specific subjects, such as the problems you see most often, or a subspecialty that you concentrate on. You can always expand your topic later, as you get more popular. Choose a name for your podcast, and purchase a domain name that accurately describes it.

You will also need to choose a hosting service. Numerous inexpensive hosting platforms are available, and a simple Google search will find them for you. Many of them provide free learning materials, helpful creative tools, and customer support to get you through the confusing technical aspects. They can also help you choose a music introduction (to add a bit of polish), and help you piece together your audio segments. Buzzsprout, RSS.com, and Podbean get good reviews on many sites. (As always, I have no financial interest in any company or service mentioned herein.)

Hosting services can assist you in creating a template – a framework that you can reuse each time you record an episode – containing your intro and exit music, tracks for your conversations, etc. This will make your podcasts instantly recognizable each time your listeners tune in.

Many podcasting experts recommend recruiting a co-host. This can be an associate within your practice, a friend who practices elsewhere, or perhaps a resident in an academic setting. You will be able to spread the workload of creating, editing, and promoting. Plus, it is much easier to generate interesting content when two people are having a conversation, rather than one person lecturing from a prepared script. You might also consider having multiple co-hosts, either to expand episodes into group discussions, or to take turns working with you in covering different subjects.

How long you make your podcast is entirely up to you. Some consultants recommend specific time frames, such as 5 minutes (because that’s an average attention span), or 28 minutes (because that’s the average driving commute time). There are short podcasts and long ones; whatever works for you is fine, as long as you don’t drift off the topic. Furthermore, no one says they must all be the same length; when you are finished talking, you are done. And no one says you must stick with one subject throughout. Combining several short segments might hold more listeners’ interest and will make it easier to share small clips on social media.

Content guidelines are similar to those for blogs. Give people content that will be of interest or benefit to them. Talk about subjects – medical and otherwise – that are relevant to your practice or are prominent in the news.

As with blogs, try to avoid polarizing political discussions, and while it’s fine to discuss treatments and procedures that you offer, aggressive solicitation tends to make viewers look elsewhere. Keep any medical advice in general terms; don’t portray any specific patients as examples.

When your podcast is ready, your hosting platform will show you how to submit it to iTunes, and how to submit your podcast RSS feed to other podcast directories. As you upload new episodes, your host will automatically update your RSS feed, so that any directory you are listed on will receive the new episode.

Once you are uploaded, you can use your host’s social sharing tools to spread the word. As with blogs, use social media, such as your practice’s Facebook page, to push podcast updates into patients’ feeds and track relevant Twitter hashtags to find online communities that might be interested in your subject matter. You should also find your episode embed code (which your host will have) and place it in a prominent place on your website so patients can listen directly from there.

Transcriptions are another excellent promotional tool. Search engines will “read” your podcasts and list them in searches. Some podcast hosts will do transcribing for a fee, but there are independent transcription services as well.

Dr. Eastern practices dermatology and dermatologic surgery in Belleville, N.J. He is the author of numerous articles and textbook chapters, and is a longtime monthly columnist for Dermatology News. Write to him at [email protected].

In my last column, I discussed blogging as a great way to capture the attention of anyone interested in your practice, especially prospective patients. If you are already blogging – or would like to consider a less crowded and competitive activity – podcasting might be the answer. At this writing (November 2022), more than 600 million blogs are online, compared with about 2 million podcasts, and relatively few of them are run by physicians. With podcasts, you have a better chance of standing out in a crowded online world.

Starting a podcast is not difficult, but there are several steps you need to go through before launching one.

As with blogging, start by outlining a long-range plan. Your general topic will probably be your specialty, but you will need to narrow your focus to a few specific subjects, such as the problems you see most often, or a subspecialty that you concentrate on. You can always expand your topic later, as you get more popular. Choose a name for your podcast, and purchase a domain name that accurately describes it.

You will also need to choose a hosting service. Numerous inexpensive hosting platforms are available, and a simple Google search will find them for you. Many of them provide free learning materials, helpful creative tools, and customer support to get you through the confusing technical aspects. They can also help you choose a music introduction (to add a bit of polish), and help you piece together your audio segments. Buzzsprout, RSS.com, and Podbean get good reviews on many sites. (As always, I have no financial interest in any company or service mentioned herein.)

Hosting services can assist you in creating a template – a framework that you can reuse each time you record an episode – containing your intro and exit music, tracks for your conversations, etc. This will make your podcasts instantly recognizable each time your listeners tune in.

Many podcasting experts recommend recruiting a co-host. This can be an associate within your practice, a friend who practices elsewhere, or perhaps a resident in an academic setting. You will be able to spread the workload of creating, editing, and promoting. Plus, it is much easier to generate interesting content when two people are having a conversation, rather than one person lecturing from a prepared script. You might also consider having multiple co-hosts, either to expand episodes into group discussions, or to take turns working with you in covering different subjects.

How long you make your podcast is entirely up to you. Some consultants recommend specific time frames, such as 5 minutes (because that’s an average attention span), or 28 minutes (because that’s the average driving commute time). There are short podcasts and long ones; whatever works for you is fine, as long as you don’t drift off the topic. Furthermore, no one says they must all be the same length; when you are finished talking, you are done. And no one says you must stick with one subject throughout. Combining several short segments might hold more listeners’ interest and will make it easier to share small clips on social media.

Content guidelines are similar to those for blogs. Give people content that will be of interest or benefit to them. Talk about subjects – medical and otherwise – that are relevant to your practice or are prominent in the news.

As with blogs, try to avoid polarizing political discussions, and while it’s fine to discuss treatments and procedures that you offer, aggressive solicitation tends to make viewers look elsewhere. Keep any medical advice in general terms; don’t portray any specific patients as examples.

When your podcast is ready, your hosting platform will show you how to submit it to iTunes, and how to submit your podcast RSS feed to other podcast directories. As you upload new episodes, your host will automatically update your RSS feed, so that any directory you are listed on will receive the new episode.

Once you are uploaded, you can use your host’s social sharing tools to spread the word. As with blogs, use social media, such as your practice’s Facebook page, to push podcast updates into patients’ feeds and track relevant Twitter hashtags to find online communities that might be interested in your subject matter. You should also find your episode embed code (which your host will have) and place it in a prominent place on your website so patients can listen directly from there.

Transcriptions are another excellent promotional tool. Search engines will “read” your podcasts and list them in searches. Some podcast hosts will do transcribing for a fee, but there are independent transcription services as well.

Dr. Eastern practices dermatology and dermatologic surgery in Belleville, N.J. He is the author of numerous articles and textbook chapters, and is a longtime monthly columnist for Dermatology News. Write to him at [email protected].

In my last column, I discussed blogging as a great way to capture the attention of anyone interested in your practice, especially prospective patients. If you are already blogging – or would like to consider a less crowded and competitive activity – podcasting might be the answer. At this writing (November 2022), more than 600 million blogs are online, compared with about 2 million podcasts, and relatively few of them are run by physicians. With podcasts, you have a better chance of standing out in a crowded online world.

Starting a podcast is not difficult, but there are several steps you need to go through before launching one.

As with blogging, start by outlining a long-range plan. Your general topic will probably be your specialty, but you will need to narrow your focus to a few specific subjects, such as the problems you see most often, or a subspecialty that you concentrate on. You can always expand your topic later, as you get more popular. Choose a name for your podcast, and purchase a domain name that accurately describes it.

You will also need to choose a hosting service. Numerous inexpensive hosting platforms are available, and a simple Google search will find them for you. Many of them provide free learning materials, helpful creative tools, and customer support to get you through the confusing technical aspects. They can also help you choose a music introduction (to add a bit of polish), and help you piece together your audio segments. Buzzsprout, RSS.com, and Podbean get good reviews on many sites. (As always, I have no financial interest in any company or service mentioned herein.)

Hosting services can assist you in creating a template – a framework that you can reuse each time you record an episode – containing your intro and exit music, tracks for your conversations, etc. This will make your podcasts instantly recognizable each time your listeners tune in.

Many podcasting experts recommend recruiting a co-host. This can be an associate within your practice, a friend who practices elsewhere, or perhaps a resident in an academic setting. You will be able to spread the workload of creating, editing, and promoting. Plus, it is much easier to generate interesting content when two people are having a conversation, rather than one person lecturing from a prepared script. You might also consider having multiple co-hosts, either to expand episodes into group discussions, or to take turns working with you in covering different subjects.

How long you make your podcast is entirely up to you. Some consultants recommend specific time frames, such as 5 minutes (because that’s an average attention span), or 28 minutes (because that’s the average driving commute time). There are short podcasts and long ones; whatever works for you is fine, as long as you don’t drift off the topic. Furthermore, no one says they must all be the same length; when you are finished talking, you are done. And no one says you must stick with one subject throughout. Combining several short segments might hold more listeners’ interest and will make it easier to share small clips on social media.

Content guidelines are similar to those for blogs. Give people content that will be of interest or benefit to them. Talk about subjects – medical and otherwise – that are relevant to your practice or are prominent in the news.

As with blogs, try to avoid polarizing political discussions, and while it’s fine to discuss treatments and procedures that you offer, aggressive solicitation tends to make viewers look elsewhere. Keep any medical advice in general terms; don’t portray any specific patients as examples.

When your podcast is ready, your hosting platform will show you how to submit it to iTunes, and how to submit your podcast RSS feed to other podcast directories. As you upload new episodes, your host will automatically update your RSS feed, so that any directory you are listed on will receive the new episode.

Once you are uploaded, you can use your host’s social sharing tools to spread the word. As with blogs, use social media, such as your practice’s Facebook page, to push podcast updates into patients’ feeds and track relevant Twitter hashtags to find online communities that might be interested in your subject matter. You should also find your episode embed code (which your host will have) and place it in a prominent place on your website so patients can listen directly from there.

Transcriptions are another excellent promotional tool. Search engines will “read” your podcasts and list them in searches. Some podcast hosts will do transcribing for a fee, but there are independent transcription services as well.

Dr. Eastern practices dermatology and dermatologic surgery in Belleville, N.J. He is the author of numerous articles and textbook chapters, and is a longtime monthly columnist for Dermatology News. Write to him at [email protected].

PHILADELPHIA – Aggressive therapy using conventional synthetic disease-modifying antirheumatic drugs (DMARDs) in combination with biologic agents early, soon after a child is diagnosed with polyarticular juvenile idiopathic arthritis (pJIA), enabled more patients to achieve clinical remission and longer times in inactive disease than more conventional therapeutic approaches, 3-year results of prospective, observational study demonstrated.

The results of The Childhood Arthritis and Rheumatology Research Alliance STOP-JIA study, which Yukiko Kimura, MD, presented at the annual meeting of the American College of Rheumatology, showed early combination therapy had benefits, compared with other treatment strategies that were more evident at 3 years than at 1 year of study.

“The STOP-JIA study showed that, after 3 years, patients who started a biologic early on in combination with methotrexate spent more time in inactive disease and achieved clinical remission more often when compared to those started on traditional step-up therapy,” Dr. Kimura, chief of pediatric rheumatology at Hackensack (N.J.) Meridian Health and professor of pediatrics at the Hackensack Meridian School of Medicine, said at a press conference. “This study shows that the treatment of poly-JIA patients receive initially very early on in their disease matters even 3 years after that treatment was started.”

The study compared three CARRA consensus treatment plans (CTP) for untreated pediatric pJIA patients: step-up (SU) – starting conventional synthetic DMARD therapy and adding a biologic if needed after 3 or more months; early-combination (EC) therapy – starting synthetic and biologic DMARDs together; and biologic first (BF) therapy – starting biologic DMARD monotherapy.

Dr. Kimura explained the rationale for the study. “Since biologic treatments were introduced more than 20 years ago, the prognosis for JIA significantly improved. These very effective medicines often work wonders, quickly reducing pain and inflammation in joint disease activity,” she said in the press conference. “What is not known, however, is when is the best time to start these very effective treatments.”

The most common approach is to start with a synthetic DMARD, typically methotrexate, and wait before starting a biologic, Dr. Kimura said.

“But even though methotrexate can work very well by itself, it does not work for every patient, and we don’t know whether waiting months for it to work and then starting a biologic might potentially lessen their effectiveness,” Dr. Kimura added. “We don’t know if there’s a window of opportunity that’s lost while waiting to see whether methotrexate will work.”

The study originally enrolled 400 patients, 297 of whom completed the 3-year visit – 190 in SU, 76 in EC and 31 in BF. At 12 months, the study found no statistically significant difference in clinically inactive disease (CID) between the groups, Dr. Kimura said.

Even at the 3-year visit, the percentage of patients in CID off glucocorticoids and clinical Juvenile Arthritis Disease Activity Score based on 10 joints inactive disease (cJADAS 10 ID) did not differ among the three groups, Dr. Kimura said in presenting the results. “But,” she added, “greater proportions of early-combination CTP group were able to achieve clinical remissions and spend more time with inactive disease in both CID and cJADAS 10.”

A closer look at the outcomes showed some separation between early-combination therapy and the other two treatment plans. The incidence of clinical remission (at any time point over 36 months) was 67.1% in the EC group vs. 49.1% and 47.3%, respectively, in the BF and SU groups, Dr. Kimura said. “The difference between the early-combination and step-up groups was highly significant [P = .007],” she added.

EC also had an edge in the percentage of time patients spent in CID (over 36 months): 39.2% versus 32% and 27.4%, respectively, in the BF and SU groups (P = .006 for EV vs. SU), as well as cJADAS 10 ID (50.6% in EC group vs. 42.8% and 37.5%, respectively in the BF and SU groups; P = .005 for EC vs. SU).

Dr. Kimura said that the STOP JIA trial will continue with longer-term analysis and ongoing monitoring of study patients through the CARRA registry. “These longer-term analyses and readouts will be important because even though the results at 12 months didn’t seem as definitive, it seems the longer we go, the more impact we see of the treatments that were started early on in this disease.”

The findings from this study are “significantly important,” Nina T. Washington, MD, MPH, a pediatric rheumatologist at the University of New Mexico Hospital, Albuquerque, and the Mary Bridge Children’s Hospital in Tacoma, Wash., said in an interview. “At least for the past decade we’ve really been advocating towards earlier and aggressive therapy, and that’s what this study shows: the sooner you can treat this disease, the sooner you can attack those joints that are inflamed, the better outcome you give the patient.”

The study also confirms that pediatric rheumatologists are not overtreating patients with pJIA, she added.

“In a sense we’re actually treating and preventing and if you have a child that has arthritis, it’s okay to treat that child,” Dr. Washington said. “For me that’s the most reassuring thing: that I’m not necessarily going overboard. If I have a child with polyarticular JIA and they have multiple inflamed joints and I have the evidence as they’re sitting in front of me, and I treat them. I’m going to give them the best outcome.”

The Patient Centered Outcomes Research Institute provided study funding. Dr. Kimura is chair of the CARRA JIA disease research committee and cochair of the CARRA Registry and Research Oversight Committee. She disclosed a financial relationship with Genentech. Dr. Washington has no relevant relationships to disclose.
 

PHILADELPHIA – Aggressive therapy using conventional synthetic disease-modifying antirheumatic drugs (DMARDs) in combination with biologic agents early, soon after a child is diagnosed with polyarticular juvenile idiopathic arthritis (pJIA), enabled more patients to achieve clinical remission and longer times in inactive disease than more conventional therapeutic approaches, 3-year results of prospective, observational study demonstrated.

The results of The Childhood Arthritis and Rheumatology Research Alliance STOP-JIA study, which Yukiko Kimura, MD, presented at the annual meeting of the American College of Rheumatology, showed early combination therapy had benefits, compared with other treatment strategies that were more evident at 3 years than at 1 year of study.

“The STOP-JIA study showed that, after 3 years, patients who started a biologic early on in combination with methotrexate spent more time in inactive disease and achieved clinical remission more often when compared to those started on traditional step-up therapy,” Dr. Kimura, chief of pediatric rheumatology at Hackensack (N.J.) Meridian Health and professor of pediatrics at the Hackensack Meridian School of Medicine, said at a press conference. “This study shows that the treatment of poly-JIA patients receive initially very early on in their disease matters even 3 years after that treatment was started.”

The study compared three CARRA consensus treatment plans (CTP) for untreated pediatric pJIA patients: step-up (SU) – starting conventional synthetic DMARD therapy and adding a biologic if needed after 3 or more months; early-combination (EC) therapy – starting synthetic and biologic DMARDs together; and biologic first (BF) therapy – starting biologic DMARD monotherapy.

Dr. Kimura explained the rationale for the study. “Since biologic treatments were introduced more than 20 years ago, the prognosis for JIA significantly improved. These very effective medicines often work wonders, quickly reducing pain and inflammation in joint disease activity,” she said in the press conference. “What is not known, however, is when is the best time to start these very effective treatments.”

The most common approach is to start with a synthetic DMARD, typically methotrexate, and wait before starting a biologic, Dr. Kimura said.

“But even though methotrexate can work very well by itself, it does not work for every patient, and we don’t know whether waiting months for it to work and then starting a biologic might potentially lessen their effectiveness,” Dr. Kimura added. “We don’t know if there’s a window of opportunity that’s lost while waiting to see whether methotrexate will work.”

The study originally enrolled 400 patients, 297 of whom completed the 3-year visit – 190 in SU, 76 in EC and 31 in BF. At 12 months, the study found no statistically significant difference in clinically inactive disease (CID) between the groups, Dr. Kimura said.

Even at the 3-year visit, the percentage of patients in CID off glucocorticoids and clinical Juvenile Arthritis Disease Activity Score based on 10 joints inactive disease (cJADAS 10 ID) did not differ among the three groups, Dr. Kimura said in presenting the results. “But,” she added, “greater proportions of early-combination CTP group were able to achieve clinical remissions and spend more time with inactive disease in both CID and cJADAS 10.”

A closer look at the outcomes showed some separation between early-combination therapy and the other two treatment plans. The incidence of clinical remission (at any time point over 36 months) was 67.1% in the EC group vs. 49.1% and 47.3%, respectively, in the BF and SU groups, Dr. Kimura said. “The difference between the early-combination and step-up groups was highly significant [P = .007],” she added.

EC also had an edge in the percentage of time patients spent in CID (over 36 months): 39.2% versus 32% and 27.4%, respectively, in the BF and SU groups (P = .006 for EV vs. SU), as well as cJADAS 10 ID (50.6% in EC group vs. 42.8% and 37.5%, respectively in the BF and SU groups; P = .005 for EC vs. SU).

Dr. Kimura said that the STOP JIA trial will continue with longer-term analysis and ongoing monitoring of study patients through the CARRA registry. “These longer-term analyses and readouts will be important because even though the results at 12 months didn’t seem as definitive, it seems the longer we go, the more impact we see of the treatments that were started early on in this disease.”

The findings from this study are “significantly important,” Nina T. Washington, MD, MPH, a pediatric rheumatologist at the University of New Mexico Hospital, Albuquerque, and the Mary Bridge Children’s Hospital in Tacoma, Wash., said in an interview. “At least for the past decade we’ve really been advocating towards earlier and aggressive therapy, and that’s what this study shows: the sooner you can treat this disease, the sooner you can attack those joints that are inflamed, the better outcome you give the patient.”

The study also confirms that pediatric rheumatologists are not overtreating patients with pJIA, she added.

“In a sense we’re actually treating and preventing and if you have a child that has arthritis, it’s okay to treat that child,” Dr. Washington said. “For me that’s the most reassuring thing: that I’m not necessarily going overboard. If I have a child with polyarticular JIA and they have multiple inflamed joints and I have the evidence as they’re sitting in front of me, and I treat them. I’m going to give them the best outcome.”

The Patient Centered Outcomes Research Institute provided study funding. Dr. Kimura is chair of the CARRA JIA disease research committee and cochair of the CARRA Registry and Research Oversight Committee. She disclosed a financial relationship with Genentech. Dr. Washington has no relevant relationships to disclose.
 

PHILADELPHIA – Aggressive therapy using conventional synthetic disease-modifying antirheumatic drugs (DMARDs) in combination with biologic agents early, soon after a child is diagnosed with polyarticular juvenile idiopathic arthritis (pJIA), enabled more patients to achieve clinical remission and longer times in inactive disease than more conventional therapeutic approaches, 3-year results of prospective, observational study demonstrated.

The results of The Childhood Arthritis and Rheumatology Research Alliance STOP-JIA study, which Yukiko Kimura, MD, presented at the annual meeting of the American College of Rheumatology, showed early combination therapy had benefits, compared with other treatment strategies that were more evident at 3 years than at 1 year of study.

“The STOP-JIA study showed that, after 3 years, patients who started a biologic early on in combination with methotrexate spent more time in inactive disease and achieved clinical remission more often when compared to those started on traditional step-up therapy,” Dr. Kimura, chief of pediatric rheumatology at Hackensack (N.J.) Meridian Health and professor of pediatrics at the Hackensack Meridian School of Medicine, said at a press conference. “This study shows that the treatment of poly-JIA patients receive initially very early on in their disease matters even 3 years after that treatment was started.”

The study compared three CARRA consensus treatment plans (CTP) for untreated pediatric pJIA patients: step-up (SU) – starting conventional synthetic DMARD therapy and adding a biologic if needed after 3 or more months; early-combination (EC) therapy – starting synthetic and biologic DMARDs together; and biologic first (BF) therapy – starting biologic DMARD monotherapy.

Dr. Kimura explained the rationale for the study. “Since biologic treatments were introduced more than 20 years ago, the prognosis for JIA significantly improved. These very effective medicines often work wonders, quickly reducing pain and inflammation in joint disease activity,” she said in the press conference. “What is not known, however, is when is the best time to start these very effective treatments.”

The most common approach is to start with a synthetic DMARD, typically methotrexate, and wait before starting a biologic, Dr. Kimura said.

“But even though methotrexate can work very well by itself, it does not work for every patient, and we don’t know whether waiting months for it to work and then starting a biologic might potentially lessen their effectiveness,” Dr. Kimura added. “We don’t know if there’s a window of opportunity that’s lost while waiting to see whether methotrexate will work.”

The study originally enrolled 400 patients, 297 of whom completed the 3-year visit – 190 in SU, 76 in EC and 31 in BF. At 12 months, the study found no statistically significant difference in clinically inactive disease (CID) between the groups, Dr. Kimura said.

Even at the 3-year visit, the percentage of patients in CID off glucocorticoids and clinical Juvenile Arthritis Disease Activity Score based on 10 joints inactive disease (cJADAS 10 ID) did not differ among the three groups, Dr. Kimura said in presenting the results. “But,” she added, “greater proportions of early-combination CTP group were able to achieve clinical remissions and spend more time with inactive disease in both CID and cJADAS 10.”

A closer look at the outcomes showed some separation between early-combination therapy and the other two treatment plans. The incidence of clinical remission (at any time point over 36 months) was 67.1% in the EC group vs. 49.1% and 47.3%, respectively, in the BF and SU groups, Dr. Kimura said. “The difference between the early-combination and step-up groups was highly significant [P = .007],” she added.

EC also had an edge in the percentage of time patients spent in CID (over 36 months): 39.2% versus 32% and 27.4%, respectively, in the BF and SU groups (P = .006 for EV vs. SU), as well as cJADAS 10 ID (50.6% in EC group vs. 42.8% and 37.5%, respectively in the BF and SU groups; P = .005 for EC vs. SU).

Dr. Kimura said that the STOP JIA trial will continue with longer-term analysis and ongoing monitoring of study patients through the CARRA registry. “These longer-term analyses and readouts will be important because even though the results at 12 months didn’t seem as definitive, it seems the longer we go, the more impact we see of the treatments that were started early on in this disease.”

The findings from this study are “significantly important,” Nina T. Washington, MD, MPH, a pediatric rheumatologist at the University of New Mexico Hospital, Albuquerque, and the Mary Bridge Children’s Hospital in Tacoma, Wash., said in an interview. “At least for the past decade we’ve really been advocating towards earlier and aggressive therapy, and that’s what this study shows: the sooner you can treat this disease, the sooner you can attack those joints that are inflamed, the better outcome you give the patient.”

The study also confirms that pediatric rheumatologists are not overtreating patients with pJIA, she added.

“In a sense we’re actually treating and preventing and if you have a child that has arthritis, it’s okay to treat that child,” Dr. Washington said. “For me that’s the most reassuring thing: that I’m not necessarily going overboard. If I have a child with polyarticular JIA and they have multiple inflamed joints and I have the evidence as they’re sitting in front of me, and I treat them. I’m going to give them the best outcome.”

The Patient Centered Outcomes Research Institute provided study funding. Dr. Kimura is chair of the CARRA JIA disease research committee and cochair of the CARRA Registry and Research Oversight Committee. She disclosed a financial relationship with Genentech. Dr. Washington has no relevant relationships to disclose.
 

This transcript has been edited for clarity.

Welcome to Impact Factor, your weekly dose of commentary on a new medical study. I’m Dr. F. Perry Wilson of the Yale School of Medicine.

On March 26, 2022, Hawaii became the last state in the United States to lift its indoor mask mandate. By the time the current school year started, there were essentially no public school mask mandates either.

Whether you viewed the mask as an emblem of stalwart defiance against a rampaging virus, or a scarlet letter emblematic of the overreaches of public policy, you probably aren’t seeing them much anymore.

And yet, the debate about masks still rages. Who was right, who was wrong? Who trusted science, and what does the science even say? If we brought our country into marriage counseling, would we be told it is time to move on?  To look forward, not backward? To plan for our bright future together?

Perhaps. But this question isn’t really moot just because masks have largely disappeared in the United States. Variants may emerge that lead to more infection waves – and other pandemics may occur in the future. And so I think it is important to discuss a study that, with quite rigorous analysis, attempts to answer the following question: Did masking in schools lower students’ and teachers’ risk of COVID?

We are talking about this study, appearing in the New England Journal of Medicine. The short version goes like this.

Researchers had access to two important sources of data. One – an accounting of all the teachers and students (more than 300,000 of them) in 79 public, noncharter school districts in Eastern Massachusetts who tested positive for COVID every week. Two – the date that each of those school districts lifted their mask mandates or (in the case of two districts) didn’t.

Right away, I’m sure you’re thinking of potential issues. Districts that kept masks even when the statewide ban was lifted are likely quite a bit different from districts that dropped masks right away. You’re right, of course – hold on to that thought; we’ll get there.

But first – the big question – would districts that kept their masks on longer do better when it comes to the rate of COVID infection?

When everyone was masking, COVID case rates were pretty similar. Statewide mandates are lifted in late February – and most school districts remove their mandates within a few weeks – the black line are the two districts (Boston and Chelsea) where mask mandates remained in place.

As time marched on, the case rates in the various districts spread out – with districts that kept masks on longer doing better than those that took them off, and districts that kept masks on the whole time doing best of all.

Prior to the mask mandate lifting, you see very similar COVID rates in districts that would eventually remove the mandate and those that would not, with a bit of noise around the initial Omicron wave which saw just a huge amount of people get infected.

And then, after the mandate was lifted, separation. Districts that held on to masks longer had lower rates of COVID infection.

In all, over the 15-weeks of the study, there were roughly 12,000 extra cases of COVID in the mask-free school districts, which corresponds to about 35% of the total COVID burden during that time. And, yes, kids do well with COVID – on average. But 12,000 extra cases is enough to translate into a significant number of important clinical outcomes – think hospitalizations and post-COVID syndromes. And of course, maybe most importantly, missed school days. Positive kids were not allowed in class no matter what district they were in.

Okay – I promised we’d address confounders. This was not a cluster-randomized trial, where some school districts had their mandates removed based on the vicissitudes of a virtual coin flip, as much as many of us would have been interested to see that. The decision to remove masks was up to the various school boards – and they had a lot of pressure on them from many different directions. But all we need to worry about is whether any of those things that pressure a school board to keep masks on would ALSO lead to fewer COVID cases. That’s how confounders work, and how you can get false results in a study like this.

And yes – districts that kept the masks on longer were different than those who took them right off. But check out how they were different.

The districts that kept masks on longer had more low-income students. More Black and Latino students. More students per classroom. These are all risk factors that increase the risk of COVID infection. In other words, the confounding here goes in the opposite direction of the results. If anything, these factors should make you more certain that masking works.

The authors also adjusted for other factors – the community transmission of COVID-19, vaccination rates, school district sizes, and so on. No major change in the results.

One concern I addressed to Dr. Ellie Murray, the biostatistician on the study – could districts that removed masks simply have been testing more to compensate, leading to increased capturing of cases?

If anything, the schools that kept masks on were testing more than the schools that took them off – again that would tend to imply that the results are even stronger than what was reported.

Is this a perfect study? Of course not – it’s one study, it’s from one state. And the relatively large effects from keeping masks on for one or 2 weeks require us to really embrace the concept of exponential growth of infections, but, if COVID has taught us anything, it is that small changes in initial conditions can have pretty big effects.

My daughter, who goes to a public school here in Connecticut, unmasked, was home with COVID this past week. She’s fine. But you know what? She missed a week of school. I worked from home to be with her – though I didn’t test positive. And that is a real cost to both of us that I think we need to consider when we consider the value of masks. Yes, they’re annoying – but if they keep kids in school, might they be worth it? Perhaps not for now, as cases aren’t surging. But in the future, be it a particularly concerning variant, or a whole new pandemic, we should not discount the simple, cheap, and apparently beneficial act of wearing masks to decrease transmission.

Dr. Perry Wilson is an associate professor of medicine and director of the Clinical and Translational Research Accelerator at Yale University, New Haven, Conn. He disclosed no relevant conflicts of interest.

A version of this article first appeared on Medscape.com.

This transcript has been edited for clarity.

Welcome to Impact Factor, your weekly dose of commentary on a new medical study. I’m Dr. F. Perry Wilson of the Yale School of Medicine.

On March 26, 2022, Hawaii became the last state in the United States to lift its indoor mask mandate. By the time the current school year started, there were essentially no public school mask mandates either.

Whether you viewed the mask as an emblem of stalwart defiance against a rampaging virus, or a scarlet letter emblematic of the overreaches of public policy, you probably aren’t seeing them much anymore.

And yet, the debate about masks still rages. Who was right, who was wrong? Who trusted science, and what does the science even say? If we brought our country into marriage counseling, would we be told it is time to move on?  To look forward, not backward? To plan for our bright future together?

Perhaps. But this question isn’t really moot just because masks have largely disappeared in the United States. Variants may emerge that lead to more infection waves – and other pandemics may occur in the future. And so I think it is important to discuss a study that, with quite rigorous analysis, attempts to answer the following question: Did masking in schools lower students’ and teachers’ risk of COVID?

We are talking about this study, appearing in the New England Journal of Medicine. The short version goes like this.

Researchers had access to two important sources of data. One – an accounting of all the teachers and students (more than 300,000 of them) in 79 public, noncharter school districts in Eastern Massachusetts who tested positive for COVID every week. Two – the date that each of those school districts lifted their mask mandates or (in the case of two districts) didn’t.

Right away, I’m sure you’re thinking of potential issues. Districts that kept masks even when the statewide ban was lifted are likely quite a bit different from districts that dropped masks right away. You’re right, of course – hold on to that thought; we’ll get there.

But first – the big question – would districts that kept their masks on longer do better when it comes to the rate of COVID infection?

When everyone was masking, COVID case rates were pretty similar. Statewide mandates are lifted in late February – and most school districts remove their mandates within a few weeks – the black line are the two districts (Boston and Chelsea) where mask mandates remained in place.

As time marched on, the case rates in the various districts spread out – with districts that kept masks on longer doing better than those that took them off, and districts that kept masks on the whole time doing best of all.

Prior to the mask mandate lifting, you see very similar COVID rates in districts that would eventually remove the mandate and those that would not, with a bit of noise around the initial Omicron wave which saw just a huge amount of people get infected.

And then, after the mandate was lifted, separation. Districts that held on to masks longer had lower rates of COVID infection.

In all, over the 15-weeks of the study, there were roughly 12,000 extra cases of COVID in the mask-free school districts, which corresponds to about 35% of the total COVID burden during that time. And, yes, kids do well with COVID – on average. But 12,000 extra cases is enough to translate into a significant number of important clinical outcomes – think hospitalizations and post-COVID syndromes. And of course, maybe most importantly, missed school days. Positive kids were not allowed in class no matter what district they were in.

Okay – I promised we’d address confounders. This was not a cluster-randomized trial, where some school districts had their mandates removed based on the vicissitudes of a virtual coin flip, as much as many of us would have been interested to see that. The decision to remove masks was up to the various school boards – and they had a lot of pressure on them from many different directions. But all we need to worry about is whether any of those things that pressure a school board to keep masks on would ALSO lead to fewer COVID cases. That’s how confounders work, and how you can get false results in a study like this.

And yes – districts that kept the masks on longer were different than those who took them right off. But check out how they were different.

The districts that kept masks on longer had more low-income students. More Black and Latino students. More students per classroom. These are all risk factors that increase the risk of COVID infection. In other words, the confounding here goes in the opposite direction of the results. If anything, these factors should make you more certain that masking works.

The authors also adjusted for other factors – the community transmission of COVID-19, vaccination rates, school district sizes, and so on. No major change in the results.

One concern I addressed to Dr. Ellie Murray, the biostatistician on the study – could districts that removed masks simply have been testing more to compensate, leading to increased capturing of cases?

If anything, the schools that kept masks on were testing more than the schools that took them off – again that would tend to imply that the results are even stronger than what was reported.

Is this a perfect study? Of course not – it’s one study, it’s from one state. And the relatively large effects from keeping masks on for one or 2 weeks require us to really embrace the concept of exponential growth of infections, but, if COVID has taught us anything, it is that small changes in initial conditions can have pretty big effects.

My daughter, who goes to a public school here in Connecticut, unmasked, was home with COVID this past week. She’s fine. But you know what? She missed a week of school. I worked from home to be with her – though I didn’t test positive. And that is a real cost to both of us that I think we need to consider when we consider the value of masks. Yes, they’re annoying – but if they keep kids in school, might they be worth it? Perhaps not for now, as cases aren’t surging. But in the future, be it a particularly concerning variant, or a whole new pandemic, we should not discount the simple, cheap, and apparently beneficial act of wearing masks to decrease transmission.

Dr. Perry Wilson is an associate professor of medicine and director of the Clinical and Translational Research Accelerator at Yale University, New Haven, Conn. He disclosed no relevant conflicts of interest.

A version of this article first appeared on Medscape.com.

This transcript has been edited for clarity.

Welcome to Impact Factor, your weekly dose of commentary on a new medical study. I’m Dr. F. Perry Wilson of the Yale School of Medicine.

On March 26, 2022, Hawaii became the last state in the United States to lift its indoor mask mandate. By the time the current school year started, there were essentially no public school mask mandates either.

Whether you viewed the mask as an emblem of stalwart defiance against a rampaging virus, or a scarlet letter emblematic of the overreaches of public policy, you probably aren’t seeing them much anymore.

And yet, the debate about masks still rages. Who was right, who was wrong? Who trusted science, and what does the science even say? If we brought our country into marriage counseling, would we be told it is time to move on?  To look forward, not backward? To plan for our bright future together?

Perhaps. But this question isn’t really moot just because masks have largely disappeared in the United States. Variants may emerge that lead to more infection waves – and other pandemics may occur in the future. And so I think it is important to discuss a study that, with quite rigorous analysis, attempts to answer the following question: Did masking in schools lower students’ and teachers’ risk of COVID?

We are talking about this study, appearing in the New England Journal of Medicine. The short version goes like this.

Researchers had access to two important sources of data. One – an accounting of all the teachers and students (more than 300,000 of them) in 79 public, noncharter school districts in Eastern Massachusetts who tested positive for COVID every week. Two – the date that each of those school districts lifted their mask mandates or (in the case of two districts) didn’t.

Right away, I’m sure you’re thinking of potential issues. Districts that kept masks even when the statewide ban was lifted are likely quite a bit different from districts that dropped masks right away. You’re right, of course – hold on to that thought; we’ll get there.

But first – the big question – would districts that kept their masks on longer do better when it comes to the rate of COVID infection?

When everyone was masking, COVID case rates were pretty similar. Statewide mandates are lifted in late February – and most school districts remove their mandates within a few weeks – the black line are the two districts (Boston and Chelsea) where mask mandates remained in place.

As time marched on, the case rates in the various districts spread out – with districts that kept masks on longer doing better than those that took them off, and districts that kept masks on the whole time doing best of all.

Prior to the mask mandate lifting, you see very similar COVID rates in districts that would eventually remove the mandate and those that would not, with a bit of noise around the initial Omicron wave which saw just a huge amount of people get infected.

And then, after the mandate was lifted, separation. Districts that held on to masks longer had lower rates of COVID infection.

In all, over the 15-weeks of the study, there were roughly 12,000 extra cases of COVID in the mask-free school districts, which corresponds to about 35% of the total COVID burden during that time. And, yes, kids do well with COVID – on average. But 12,000 extra cases is enough to translate into a significant number of important clinical outcomes – think hospitalizations and post-COVID syndromes. And of course, maybe most importantly, missed school days. Positive kids were not allowed in class no matter what district they were in.

Okay – I promised we’d address confounders. This was not a cluster-randomized trial, where some school districts had their mandates removed based on the vicissitudes of a virtual coin flip, as much as many of us would have been interested to see that. The decision to remove masks was up to the various school boards – and they had a lot of pressure on them from many different directions. But all we need to worry about is whether any of those things that pressure a school board to keep masks on would ALSO lead to fewer COVID cases. That’s how confounders work, and how you can get false results in a study like this.

And yes – districts that kept the masks on longer were different than those who took them right off. But check out how they were different.

The districts that kept masks on longer had more low-income students. More Black and Latino students. More students per classroom. These are all risk factors that increase the risk of COVID infection. In other words, the confounding here goes in the opposite direction of the results. If anything, these factors should make you more certain that masking works.

The authors also adjusted for other factors – the community transmission of COVID-19, vaccination rates, school district sizes, and so on. No major change in the results.

One concern I addressed to Dr. Ellie Murray, the biostatistician on the study – could districts that removed masks simply have been testing more to compensate, leading to increased capturing of cases?

If anything, the schools that kept masks on were testing more than the schools that took them off – again that would tend to imply that the results are even stronger than what was reported.

Is this a perfect study? Of course not – it’s one study, it’s from one state. And the relatively large effects from keeping masks on for one or 2 weeks require us to really embrace the concept of exponential growth of infections, but, if COVID has taught us anything, it is that small changes in initial conditions can have pretty big effects.

My daughter, who goes to a public school here in Connecticut, unmasked, was home with COVID this past week. She’s fine. But you know what? She missed a week of school. I worked from home to be with her – though I didn’t test positive. And that is a real cost to both of us that I think we need to consider when we consider the value of masks. Yes, they’re annoying – but if they keep kids in school, might they be worth it? Perhaps not for now, as cases aren’t surging. But in the future, be it a particularly concerning variant, or a whole new pandemic, we should not discount the simple, cheap, and apparently beneficial act of wearing masks to decrease transmission.

Dr. Perry Wilson is an associate professor of medicine and director of the Clinical and Translational Research Accelerator at Yale University, New Haven, Conn. He disclosed no relevant conflicts of interest.

A version of this article first appeared on Medscape.com.