Vascular

Editor’s Note: The views expressed in this article are solely those of the authors and do not reflect the official policy or position of the Department of State or the United States Government. This version of the article was peer-reviewed.

Venous thromboembolism (VTE) associated with travel has emerged as an important public health concern over the past decade. Numerous epidemiologic and case control studies have reported air travel as a risk factor for the development of VTE and have attempted to determine who is at risk and which precautions need to be taken to prevent this potentially fatal event.^1–7 Often referred to as “traveler’s thrombosis” or “flight-related deep vein thrombosis,” VTE can also develop after long trips by automobile, bus, or train.^8,9 Although the absolute risk is very low, this threat appears to be about three times higher in travelers and increases with longer trips.³

See related patient information material

This article focuses on defining VTE and recognizing its clinical features, as well as providing recommendations and guidelines to prevent, diagnose, and treat this complication in people who travel.

WHAT IS VENOUS THROMBOEMBOLISM?

Deep vein thrombosis and pulmonary embolism represent different manifestations of the same clinical entity, ie, VTE. VTE is a common, lethal disease that affects hospitalized and nonhospitalized patients, frequently recurs, is often overlooked, may be asymptomatic, and may result in long-term complications that include pulmonary hypertension and the postthrombotic syndrome.

Deep vein thrombosis of the upper extremities is generally related to an indwelling venous catheter or a central line being used for long-term administration of antibiotics, chemotherapy, or nutrition. A condition known as Paget-Schroetter syndrome or “effort thrombosis” may be seen in younger or athletic people who have a history of strenuous or unusual arm exercise.

RISK FACTORS FOR VTE

Common inherited risk factors include:

• Factor V Leiden mutation
• Prothrombin gene mutation G20210A
• Hyperhomocysteinemia
• Deficiency of the natural anticoagulant proteins C, S, or antithrombin
• Elevated levels of factor VIII (may be inherited or acquired).

Acquired risk factors include:

• Older age
• Immobilization or stasis (such as sitting for long periods of time while traveling)
• Surgery (most notably orthopedic procedures including hip and knee replacement and repair of a hip fracture)
• Trauma
• Stroke
• Acute medical illness (including congestive heart failure, chronic obstructive pulmonary disease, pneumonia)
• The antiphospholipid syndrome (consisting of a lupus anticoagulant, anticardiolipin antibodies, or both)
• Pregnancy and the postpartum state
• Use of oral contraceptives or hormone replacement therapy
• Cancer (including the myeloproliferative disorders) and certain chemotherapeutic agents
• Obesity (a body mass index > 30 kg/m², see www.nhlbisupport.com/bmi/)
• Inflammatory bowel disease
• Previous VTE
• A central venous catheter or pacemaker
• Nephrotic syndrome.

In addition, emerging risk factors more recently recognized include male sex, persistence of elevated factor VIII levels, and the continued presence of an elevated D-dimer level or deep vein thrombosis on duplex ultrasonography once anticoagulation treatment is completed. There is also evidence of an association between VTE and risk factors for atherosclerotic arterial disease such as smoking, hypertension, hyperlipidemia, and diabetes.

CLINICAL MANIFESTATIONS OF VTE

Patients with deep vein thrombosis may complain of pain, swelling, or both in the leg or arm. Physical examination may reveal increased warmth, tenderness, erythema, edema, or dilated (collateral) veins, most notable on the upper thigh or calf (for deep vein thrombosis in the lower extremity) or the chest wall (for upper-extremity deep vein thrombosis). The examiner may also observe a tender, palpable cord, which represents a superficial vein thrombosis involving the great and small saphenous veins (Figure 1). In extreme situations, the limb may be cyanotic or gangrenous.

DIAGNOSIS OF VTE

Clinical examination alone is generally insufficient to confirm a diagnosis of deep vein thrombosis or pulmonary embolism. Venous duplex ultrasonography is the most dependable investigation for deep vein thrombosis, but other tests include D-dimer and imaging studies such as computed tomographic venography or magnetic resonance venography of the lower extremities. A more invasive approach is venography; formerly considered the gold standard, it is now generally used only when the diagnosis is in doubt after noninvasive testing. The diagnosis of acute pulmonary embolism is best made by spiral computed tomography.

Other studies that may prove helpful include a ventilation-perfusion lung scan for patients who cannot undergo computed tomography due to a contrast allergy or renal insufficiency. Pulmonary angiography, while the gold standard, is less commonly used today, given the specificity and sensitivity of computed tomography.

Echocardiography at the bedside may be useful for patients too sick to move, although the study may not be diagnostic unless thrombi are seen in the heart or pulmonary arteries.

TREATMENT OF VTE

For acute deep venous thrombosis

Acute deep vein thrombosis is now treated on an outpatient basis under most circumstances.

Unfractionated heparin is given intravenously for patients who need to be hospitalized, or subcutaneously in full dose for inpatient or outpatient treatment.

Low-molecular-weight heparins are available in subcutaneous preparations and can be given on an outpatient basis.

Fondaparinux (Arixtra), a factor Xa inhibitor, can also be given subcutaneously on an outpatient basis. Equivalent products are available outside the United States.

Warfarin (Coumadin), an oral vitamin K inhibitor, is the agent of choice for long-term management of deep vein thrombosis.

Other oral agents are available outside the United States.

For pulmonary embolism

Outpatient treatment of pulmonary embolism is not yet advised: an initial hospitalization is necessary. The same anticoagulants used for deep vein thrombosis are also used for acute pulmonary embolism.

Empiric treatment in underdeveloped countries

VTE may be an even greater concern on an outbound trip to a remote area, where medical care capabilities may be less than ideal and diagnostic and treatment options may be limited.

If there is a high pretest probability of acute VTE (Table 2, Table 3) and no diagnostic methods are available, empiric treatment with any of the parenteral anticoagulant agents listed in Table 4 is an option until the diagnosis can be confirmed. Caveats:

• Care must be taken to be certain there is not a strong contraindication to the use of anticoagulation, such as bleeding or a drug allergy.
• Neither unfractionated heparin nor any of the low-molecular-weight heparins should be given to a patient who has a history of heparin-induced thrombocytopenia.
• In patients who have chronic kidney disease (creatinine clearance less than 30 mL/minute), the dosage of low-molecular-weight heparins must be adjusted and factor Xa inhibitors avoided. Both of these types of anticoagulants should be avoided in patients on hemodialysis.

More aggressive therapy

Under select circumstances a more aggressive approach to the treatment of VTE may be necessary. These options are usually indicated for a patient with a massive deep vein thrombosis of a lower extremity and for certain patients with an upper extremity deep vein thrombosis. Treatments include catheter-directed thrombolytic therapy and endovenous or surgical thrombectomy.

Thrombolytic therapy is recommended for a patient with an acute pulmonary embolism who is clinically unstable (systolic blood pressure lower than 90 mm Hg), if there is no contraindication to its use (bleeding risk or recent stroke or surgery). Thrombolytic therapy is also an option for those at low risk of bleeding with an acute pulmonary embolism who have signs and symptoms of right heart failure proven by echocardiography.

Surgical pulmonary embolectomy for acute massive pulmonary embolism and mechanical thrombectomy for extensive deep vein thrombosis are generally available only at highly sophisticated tertiary care centers.

An inferior vena cava filter is advised in patients with acute deep vein thrombosis or pulmonary embolism who cannot be fully anticoagulated, to prevent the clot from migrating from the lower extremities to the lungs. These filters are available as either permanent or temporary implants. Some temporary versions can remain in place for up to 150 days after insertion.

PREVENTION OF VTE

Prevention is the standard of care for all patients admitted to the hospital and in select individuals as outpatients who are at high risk of VTE.

Mechanical compression (graduated compression stockings, intermittent pneumatic compression devices) has proven effective in reducing the incidence of deep vein thrombosis and pulmonary embolism postoperatively in patients who cannot take anticoagulants. One study has demonstrated that compression stockings may also be effective in preventing VTE during travel.¹²

ABSOLUTE RISK IS LOW

Over the past decade, special attention has been paid to travel as a risk factor for developing VTE.¹³ Traveler’s thrombosis has become an important public health concern. Numerous publications and epidemiologic studies have targeted air travel in an attempt to determine who is at risk and what precautions are necessary to prevent this complication.^1–7,9

The incidence of VTE following air travel is reported to be 3.2 per 1,000 person-years.⁴ While this incidence is relatively low, it is still 3.2 times higher than in the healthy population that is not flying.

The more serious complication of VTE, ie, acute pulmonary embolism, occurs less often. In three studies, the reported incidence ranged from 1.65 per million patients in flights longer than 8 hours to a high of 4.8 per million patients in flights longer than 12 hours or distances exceeding 10,000 km (6,200 miles).^5,14,15 For the 400 passengers on the average long-haul flight of 12 hours, there is at most a 0.2% chance that somebody on the plane will have a symptomatic VTE).

RISK FACTORS IN LONG-DISTANCE TRAVELERS

The risk of traveler’s thrombosis has recently attracted the attention of passengers and the airline industry. Airlines are now openly discussing the risk and providing reminders such as exercises that should be undertaken in-flight (see the patient information page that accompanies this article). Some airlines are recommending that all patients consult their doctor to assess their personal risk of deep vein thrombosis before flying.

The most common risk factors for VTE in travelers are well established and are additive (Table 1). The extent of the additive risk, however, is not entirely clear.

What is clear is that when VTE occurs it is a life-altering and life-threatening event. If it occurs on an outbound trip, the local resources and capabilities available at the destination may not be adequate for optimal treatment. If a traveler experiences a VTE event on an outbound trip, an emergency return trip to the continental United States or a regional center of expertise may be required. There is an additive risk with this subsequent travel event if the patient is not given immediate treatment first (Table 4). Hence, treatment prior to evacuation should be strongly considered.

The traveler must also be aware that VTE can be recognized up to 2 months after a long-haul flight, though it is especially a concern within the first 2 weeks after travel.^2,4,16,17

RECOMMENDATIONS FOR LONG-DISTANCE AIR TRAVELERS

Each person should be evaluated on a case-by-case basis for his or her need for VTE prophylaxis. Medical guidelines for airline passengers have been published by the Aerospace Medical Association and the American College of Chest Physicians (ACCP).^18,19 In general, travelers should:

• Exercise the legs by flexing and extending the ankles at regular intervals while seated (see the patient information material that accompanies this article) and frequently contracting the calf muscles.
• Walk about the cabin periodically, 5 minutes for every hour on longer-duration flights (over 4 hours) and when flight conditions permit.
• Drink adequate amounts of water and fruit juices to maintain good hydration.¹⁷
• Avoid alcohol and caffeinated beverages, which are dehydrating.
• Be careful about eating too much during the flight.
• Request an aisle seat if you are at risk
• Do not place baggage underneath the seat in front of you, because that reduces the ability to move the legs.
• Do not sleep in a cramped position, and avoid the use of any type of sleep aid.
• Avoid wearing constrictive clothing around the lower extremities or waist.

We recommend that all airplane passengers take the steps listed above to reduce venous stasis and avoid dehydration, even though these measures have not been proven effective in clinical trials.¹⁹

The ACCP further advises that decisions about pharmacologic prophylaxis of VTE for airplane passengers at high risk should be made on an individual basis, considering that there are potential adverse effects of prophylaxis and that these may outweigh the benefits. For long-distance travelers with additional risk factors for VTE, we suggest the following:

• Use of properly fitted, below-the-knee graduated compression stockings providing 15 to 30 mm Hg of pressure at the ankle (particularly when large varicosities or leg edema is present)
• For people at very high risk, a single prophylactic dose of a low-molecular-weight heparin or a factor Xa inhibitor injected just before departure (Table 5)
• Aspirin is not recommended as it is not effective for the prevention of VTE.²⁰

SUMMARY FOR THE AIR TRAVELER

All travelers on long flights should perform standard VTE prophylaxis exercises (see the patient information pages accompanying this article). Although VTE is uncommon, people with additional risk factors who travel frequently either on multiple flights in a short period of time or on very long flights should be evaluated on a case-by-case basis for a more aggressive approach to prevention (compression support hose or prophylactic administration of a low-molecular-weight heparin or a factor Xa inhibitor).

Should a VTE event occur during travel, the patient should seek medical care immediately. The standard evaluation of a patient with a suspected VTE should include an estimation of the pretest probability of disease (Table 2, Table 3), followed by duplex ultrasonography of the upper or lower extremity to detect a deep vein thrombosis. If symptoms dictate, then spiral computed tomography, ventilation-perfusion lung scan, or pulmonary angiography (where available) should be ordered to diagnose acute pulmonary embolism. A positive D-dimer blood test alone is not diagnostic and may not be available in more remote locations. A negative D-dimer test result is most helpful to exclude VTE.

Standard therapy for VTE is immediate treatment with one of the anticoagulants listed in Table 4, unless the patient has a contraindication to treatment, such as bleeding or allergy. Immediate evacuation is recommended if the patient has a life-threatening pulmonary embolism, defined as hemodynamic instability (hypotension with a blood pressure under 90 mm Hg systolic or signs of right heart failure) that cannot be treated at a local facility. An air ambulance should be used to transport these patients. If the patient has an iliofemoral deep vein thrombosis, it is also advisable that he or she be considered for evacuation if severe symptoms are present, such as pain, swelling, or cyanosis. Unless contraindicated, all patients should be given either full-dose intravenous or full-dose subcutaneous heparin or subcutaneous injection of a readily available low-molecular-weight heparin preparations or factor Xa inhibitor at once.²¹

Editor’s Note: The views expressed in this article are solely those of the authors and do not reflect the official policy or position of the Department of State or the United States Government. This version of the article was peer-reviewed.

Venous thromboembolism (VTE) associated with travel has emerged as an important public health concern over the past decade. Numerous epidemiologic and case control studies have reported air travel as a risk factor for the development of VTE and have attempted to determine who is at risk and which precautions need to be taken to prevent this potentially fatal event.^1–7 Often referred to as “traveler’s thrombosis” or “flight-related deep vein thrombosis,” VTE can also develop after long trips by automobile, bus, or train.^8,9 Although the absolute risk is very low, this threat appears to be about three times higher in travelers and increases with longer trips.³

See related patient information material

This article focuses on defining VTE and recognizing its clinical features, as well as providing recommendations and guidelines to prevent, diagnose, and treat this complication in people who travel.

WHAT IS VENOUS THROMBOEMBOLISM?

Deep vein thrombosis and pulmonary embolism represent different manifestations of the same clinical entity, ie, VTE. VTE is a common, lethal disease that affects hospitalized and nonhospitalized patients, frequently recurs, is often overlooked, may be asymptomatic, and may result in long-term complications that include pulmonary hypertension and the postthrombotic syndrome.

Deep vein thrombosis of the upper extremities is generally related to an indwelling venous catheter or a central line being used for long-term administration of antibiotics, chemotherapy, or nutrition. A condition known as Paget-Schroetter syndrome or “effort thrombosis” may be seen in younger or athletic people who have a history of strenuous or unusual arm exercise.

RISK FACTORS FOR VTE

Common inherited risk factors include:

• Factor V Leiden mutation
• Prothrombin gene mutation G20210A
• Hyperhomocysteinemia
• Deficiency of the natural anticoagulant proteins C, S, or antithrombin
• Elevated levels of factor VIII (may be inherited or acquired).

Acquired risk factors include:

• Older age
• Immobilization or stasis (such as sitting for long periods of time while traveling)
• Surgery (most notably orthopedic procedures including hip and knee replacement and repair of a hip fracture)
• Trauma
• Stroke
• Acute medical illness (including congestive heart failure, chronic obstructive pulmonary disease, pneumonia)
• The antiphospholipid syndrome (consisting of a lupus anticoagulant, anticardiolipin antibodies, or both)
• Pregnancy and the postpartum state
• Use of oral contraceptives or hormone replacement therapy
• Cancer (including the myeloproliferative disorders) and certain chemotherapeutic agents
• Obesity (a body mass index > 30 kg/m², see www.nhlbisupport.com/bmi/)
• Inflammatory bowel disease
• Previous VTE
• A central venous catheter or pacemaker
• Nephrotic syndrome.

In addition, emerging risk factors more recently recognized include male sex, persistence of elevated factor VIII levels, and the continued presence of an elevated D-dimer level or deep vein thrombosis on duplex ultrasonography once anticoagulation treatment is completed. There is also evidence of an association between VTE and risk factors for atherosclerotic arterial disease such as smoking, hypertension, hyperlipidemia, and diabetes.

CLINICAL MANIFESTATIONS OF VTE

Patients with deep vein thrombosis may complain of pain, swelling, or both in the leg or arm. Physical examination may reveal increased warmth, tenderness, erythema, edema, or dilated (collateral) veins, most notable on the upper thigh or calf (for deep vein thrombosis in the lower extremity) or the chest wall (for upper-extremity deep vein thrombosis). The examiner may also observe a tender, palpable cord, which represents a superficial vein thrombosis involving the great and small saphenous veins (Figure 1). In extreme situations, the limb may be cyanotic or gangrenous.

DIAGNOSIS OF VTE

Clinical examination alone is generally insufficient to confirm a diagnosis of deep vein thrombosis or pulmonary embolism. Venous duplex ultrasonography is the most dependable investigation for deep vein thrombosis, but other tests include D-dimer and imaging studies such as computed tomographic venography or magnetic resonance venography of the lower extremities. A more invasive approach is venography; formerly considered the gold standard, it is now generally used only when the diagnosis is in doubt after noninvasive testing. The diagnosis of acute pulmonary embolism is best made by spiral computed tomography.

Other studies that may prove helpful include a ventilation-perfusion lung scan for patients who cannot undergo computed tomography due to a contrast allergy or renal insufficiency. Pulmonary angiography, while the gold standard, is less commonly used today, given the specificity and sensitivity of computed tomography.

Echocardiography at the bedside may be useful for patients too sick to move, although the study may not be diagnostic unless thrombi are seen in the heart or pulmonary arteries.

TREATMENT OF VTE

For acute deep venous thrombosis

Acute deep vein thrombosis is now treated on an outpatient basis under most circumstances.

Unfractionated heparin is given intravenously for patients who need to be hospitalized, or subcutaneously in full dose for inpatient or outpatient treatment.

Low-molecular-weight heparins are available in subcutaneous preparations and can be given on an outpatient basis.

Fondaparinux (Arixtra), a factor Xa inhibitor, can also be given subcutaneously on an outpatient basis. Equivalent products are available outside the United States.

Warfarin (Coumadin), an oral vitamin K inhibitor, is the agent of choice for long-term management of deep vein thrombosis.

Other oral agents are available outside the United States.

For pulmonary embolism

Outpatient treatment of pulmonary embolism is not yet advised: an initial hospitalization is necessary. The same anticoagulants used for deep vein thrombosis are also used for acute pulmonary embolism.

Empiric treatment in underdeveloped countries

VTE may be an even greater concern on an outbound trip to a remote area, where medical care capabilities may be less than ideal and diagnostic and treatment options may be limited.

If there is a high pretest probability of acute VTE (Table 2, Table 3) and no diagnostic methods are available, empiric treatment with any of the parenteral anticoagulant agents listed in Table 4 is an option until the diagnosis can be confirmed. Caveats:

• Care must be taken to be certain there is not a strong contraindication to the use of anticoagulation, such as bleeding or a drug allergy.
• Neither unfractionated heparin nor any of the low-molecular-weight heparins should be given to a patient who has a history of heparin-induced thrombocytopenia.
• In patients who have chronic kidney disease (creatinine clearance less than 30 mL/minute), the dosage of low-molecular-weight heparins must be adjusted and factor Xa inhibitors avoided. Both of these types of anticoagulants should be avoided in patients on hemodialysis.

More aggressive therapy

Under select circumstances a more aggressive approach to the treatment of VTE may be necessary. These options are usually indicated for a patient with a massive deep vein thrombosis of a lower extremity and for certain patients with an upper extremity deep vein thrombosis. Treatments include catheter-directed thrombolytic therapy and endovenous or surgical thrombectomy.

Thrombolytic therapy is recommended for a patient with an acute pulmonary embolism who is clinically unstable (systolic blood pressure lower than 90 mm Hg), if there is no contraindication to its use (bleeding risk or recent stroke or surgery). Thrombolytic therapy is also an option for those at low risk of bleeding with an acute pulmonary embolism who have signs and symptoms of right heart failure proven by echocardiography.

Surgical pulmonary embolectomy for acute massive pulmonary embolism and mechanical thrombectomy for extensive deep vein thrombosis are generally available only at highly sophisticated tertiary care centers.

An inferior vena cava filter is advised in patients with acute deep vein thrombosis or pulmonary embolism who cannot be fully anticoagulated, to prevent the clot from migrating from the lower extremities to the lungs. These filters are available as either permanent or temporary implants. Some temporary versions can remain in place for up to 150 days after insertion.

PREVENTION OF VTE

Prevention is the standard of care for all patients admitted to the hospital and in select individuals as outpatients who are at high risk of VTE.

Mechanical compression (graduated compression stockings, intermittent pneumatic compression devices) has proven effective in reducing the incidence of deep vein thrombosis and pulmonary embolism postoperatively in patients who cannot take anticoagulants. One study has demonstrated that compression stockings may also be effective in preventing VTE during travel.¹²

ABSOLUTE RISK IS LOW

Over the past decade, special attention has been paid to travel as a risk factor for developing VTE.¹³ Traveler’s thrombosis has become an important public health concern. Numerous publications and epidemiologic studies have targeted air travel in an attempt to determine who is at risk and what precautions are necessary to prevent this complication.^1–7,9

The incidence of VTE following air travel is reported to be 3.2 per 1,000 person-years.⁴ While this incidence is relatively low, it is still 3.2 times higher than in the healthy population that is not flying.

The more serious complication of VTE, ie, acute pulmonary embolism, occurs less often. In three studies, the reported incidence ranged from 1.65 per million patients in flights longer than 8 hours to a high of 4.8 per million patients in flights longer than 12 hours or distances exceeding 10,000 km (6,200 miles).^5,14,15 For the 400 passengers on the average long-haul flight of 12 hours, there is at most a 0.2% chance that somebody on the plane will have a symptomatic VTE).

RISK FACTORS IN LONG-DISTANCE TRAVELERS

The risk of traveler’s thrombosis has recently attracted the attention of passengers and the airline industry. Airlines are now openly discussing the risk and providing reminders such as exercises that should be undertaken in-flight (see the patient information page that accompanies this article). Some airlines are recommending that all patients consult their doctor to assess their personal risk of deep vein thrombosis before flying.

The most common risk factors for VTE in travelers are well established and are additive (Table 1). The extent of the additive risk, however, is not entirely clear.

What is clear is that when VTE occurs it is a life-altering and life-threatening event. If it occurs on an outbound trip, the local resources and capabilities available at the destination may not be adequate for optimal treatment. If a traveler experiences a VTE event on an outbound trip, an emergency return trip to the continental United States or a regional center of expertise may be required. There is an additive risk with this subsequent travel event if the patient is not given immediate treatment first (Table 4). Hence, treatment prior to evacuation should be strongly considered.

The traveler must also be aware that VTE can be recognized up to 2 months after a long-haul flight, though it is especially a concern within the first 2 weeks after travel.^2,4,16,17

RECOMMENDATIONS FOR LONG-DISTANCE AIR TRAVELERS

Each person should be evaluated on a case-by-case basis for his or her need for VTE prophylaxis. Medical guidelines for airline passengers have been published by the Aerospace Medical Association and the American College of Chest Physicians (ACCP).^18,19 In general, travelers should:

• Exercise the legs by flexing and extending the ankles at regular intervals while seated (see the patient information material that accompanies this article) and frequently contracting the calf muscles.
• Walk about the cabin periodically, 5 minutes for every hour on longer-duration flights (over 4 hours) and when flight conditions permit.
• Drink adequate amounts of water and fruit juices to maintain good hydration.¹⁷
• Avoid alcohol and caffeinated beverages, which are dehydrating.
• Be careful about eating too much during the flight.
• Request an aisle seat if you are at risk
• Do not place baggage underneath the seat in front of you, because that reduces the ability to move the legs.
• Do not sleep in a cramped position, and avoid the use of any type of sleep aid.
• Avoid wearing constrictive clothing around the lower extremities or waist.

We recommend that all airplane passengers take the steps listed above to reduce venous stasis and avoid dehydration, even though these measures have not been proven effective in clinical trials.¹⁹

The ACCP further advises that decisions about pharmacologic prophylaxis of VTE for airplane passengers at high risk should be made on an individual basis, considering that there are potential adverse effects of prophylaxis and that these may outweigh the benefits. For long-distance travelers with additional risk factors for VTE, we suggest the following:

• Use of properly fitted, below-the-knee graduated compression stockings providing 15 to 30 mm Hg of pressure at the ankle (particularly when large varicosities or leg edema is present)
• For people at very high risk, a single prophylactic dose of a low-molecular-weight heparin or a factor Xa inhibitor injected just before departure (Table 5)
• Aspirin is not recommended as it is not effective for the prevention of VTE.²⁰

SUMMARY FOR THE AIR TRAVELER

All travelers on long flights should perform standard VTE prophylaxis exercises (see the patient information pages accompanying this article). Although VTE is uncommon, people with additional risk factors who travel frequently either on multiple flights in a short period of time or on very long flights should be evaluated on a case-by-case basis for a more aggressive approach to prevention (compression support hose or prophylactic administration of a low-molecular-weight heparin or a factor Xa inhibitor).

Should a VTE event occur during travel, the patient should seek medical care immediately. The standard evaluation of a patient with a suspected VTE should include an estimation of the pretest probability of disease (Table 2, Table 3), followed by duplex ultrasonography of the upper or lower extremity to detect a deep vein thrombosis. If symptoms dictate, then spiral computed tomography, ventilation-perfusion lung scan, or pulmonary angiography (where available) should be ordered to diagnose acute pulmonary embolism. A positive D-dimer blood test alone is not diagnostic and may not be available in more remote locations. A negative D-dimer test result is most helpful to exclude VTE.

Standard therapy for VTE is immediate treatment with one of the anticoagulants listed in Table 4, unless the patient has a contraindication to treatment, such as bleeding or allergy. Immediate evacuation is recommended if the patient has a life-threatening pulmonary embolism, defined as hemodynamic instability (hypotension with a blood pressure under 90 mm Hg systolic or signs of right heart failure) that cannot be treated at a local facility. An air ambulance should be used to transport these patients. If the patient has an iliofemoral deep vein thrombosis, it is also advisable that he or she be considered for evacuation if severe symptoms are present, such as pain, swelling, or cyanosis. Unless contraindicated, all patients should be given either full-dose intravenous or full-dose subcutaneous heparin or subcutaneous injection of a readily available low-molecular-weight heparin preparations or factor Xa inhibitor at once.²¹

Editor’s Note: The views expressed in this article are solely those of the authors and do not reflect the official policy or position of the Department of State or the United States Government. This version of the article was peer-reviewed.

Venous thromboembolism (VTE) associated with travel has emerged as an important public health concern over the past decade. Numerous epidemiologic and case control studies have reported air travel as a risk factor for the development of VTE and have attempted to determine who is at risk and which precautions need to be taken to prevent this potentially fatal event.^1–7 Often referred to as “traveler’s thrombosis” or “flight-related deep vein thrombosis,” VTE can also develop after long trips by automobile, bus, or train.^8,9 Although the absolute risk is very low, this threat appears to be about three times higher in travelers and increases with longer trips.³

See related patient information material

This article focuses on defining VTE and recognizing its clinical features, as well as providing recommendations and guidelines to prevent, diagnose, and treat this complication in people who travel.

WHAT IS VENOUS THROMBOEMBOLISM?

Deep vein thrombosis and pulmonary embolism represent different manifestations of the same clinical entity, ie, VTE. VTE is a common, lethal disease that affects hospitalized and nonhospitalized patients, frequently recurs, is often overlooked, may be asymptomatic, and may result in long-term complications that include pulmonary hypertension and the postthrombotic syndrome.

Deep vein thrombosis of the upper extremities is generally related to an indwelling venous catheter or a central line being used for long-term administration of antibiotics, chemotherapy, or nutrition. A condition known as Paget-Schroetter syndrome or “effort thrombosis” may be seen in younger or athletic people who have a history of strenuous or unusual arm exercise.

RISK FACTORS FOR VTE

Common inherited risk factors include:

• Factor V Leiden mutation
• Prothrombin gene mutation G20210A
• Hyperhomocysteinemia
• Deficiency of the natural anticoagulant proteins C, S, or antithrombin
• Elevated levels of factor VIII (may be inherited or acquired).

Acquired risk factors include:

• Older age
• Immobilization or stasis (such as sitting for long periods of time while traveling)
• Surgery (most notably orthopedic procedures including hip and knee replacement and repair of a hip fracture)
• Trauma
• Stroke
• Acute medical illness (including congestive heart failure, chronic obstructive pulmonary disease, pneumonia)
• The antiphospholipid syndrome (consisting of a lupus anticoagulant, anticardiolipin antibodies, or both)
• Pregnancy and the postpartum state
• Use of oral contraceptives or hormone replacement therapy
• Cancer (including the myeloproliferative disorders) and certain chemotherapeutic agents
• Obesity (a body mass index > 30 kg/m², see www.nhlbisupport.com/bmi/)
• Inflammatory bowel disease
• Previous VTE
• A central venous catheter or pacemaker
• Nephrotic syndrome.

In addition, emerging risk factors more recently recognized include male sex, persistence of elevated factor VIII levels, and the continued presence of an elevated D-dimer level or deep vein thrombosis on duplex ultrasonography once anticoagulation treatment is completed. There is also evidence of an association between VTE and risk factors for atherosclerotic arterial disease such as smoking, hypertension, hyperlipidemia, and diabetes.

CLINICAL MANIFESTATIONS OF VTE

Patients with deep vein thrombosis may complain of pain, swelling, or both in the leg or arm. Physical examination may reveal increased warmth, tenderness, erythema, edema, or dilated (collateral) veins, most notable on the upper thigh or calf (for deep vein thrombosis in the lower extremity) or the chest wall (for upper-extremity deep vein thrombosis). The examiner may also observe a tender, palpable cord, which represents a superficial vein thrombosis involving the great and small saphenous veins (Figure 1). In extreme situations, the limb may be cyanotic or gangrenous.

DIAGNOSIS OF VTE

Clinical examination alone is generally insufficient to confirm a diagnosis of deep vein thrombosis or pulmonary embolism. Venous duplex ultrasonography is the most dependable investigation for deep vein thrombosis, but other tests include D-dimer and imaging studies such as computed tomographic venography or magnetic resonance venography of the lower extremities. A more invasive approach is venography; formerly considered the gold standard, it is now generally used only when the diagnosis is in doubt after noninvasive testing. The diagnosis of acute pulmonary embolism is best made by spiral computed tomography.

Other studies that may prove helpful include a ventilation-perfusion lung scan for patients who cannot undergo computed tomography due to a contrast allergy or renal insufficiency. Pulmonary angiography, while the gold standard, is less commonly used today, given the specificity and sensitivity of computed tomography.

Echocardiography at the bedside may be useful for patients too sick to move, although the study may not be diagnostic unless thrombi are seen in the heart or pulmonary arteries.

TREATMENT OF VTE

For acute deep venous thrombosis

Acute deep vein thrombosis is now treated on an outpatient basis under most circumstances.

Unfractionated heparin is given intravenously for patients who need to be hospitalized, or subcutaneously in full dose for inpatient or outpatient treatment.

Low-molecular-weight heparins are available in subcutaneous preparations and can be given on an outpatient basis.

Fondaparinux (Arixtra), a factor Xa inhibitor, can also be given subcutaneously on an outpatient basis. Equivalent products are available outside the United States.

Warfarin (Coumadin), an oral vitamin K inhibitor, is the agent of choice for long-term management of deep vein thrombosis.

Other oral agents are available outside the United States.

For pulmonary embolism

Outpatient treatment of pulmonary embolism is not yet advised: an initial hospitalization is necessary. The same anticoagulants used for deep vein thrombosis are also used for acute pulmonary embolism.

Empiric treatment in underdeveloped countries

VTE may be an even greater concern on an outbound trip to a remote area, where medical care capabilities may be less than ideal and diagnostic and treatment options may be limited.

If there is a high pretest probability of acute VTE (Table 2, Table 3) and no diagnostic methods are available, empiric treatment with any of the parenteral anticoagulant agents listed in Table 4 is an option until the diagnosis can be confirmed. Caveats:

• Care must be taken to be certain there is not a strong contraindication to the use of anticoagulation, such as bleeding or a drug allergy.
• Neither unfractionated heparin nor any of the low-molecular-weight heparins should be given to a patient who has a history of heparin-induced thrombocytopenia.
• In patients who have chronic kidney disease (creatinine clearance less than 30 mL/minute), the dosage of low-molecular-weight heparins must be adjusted and factor Xa inhibitors avoided. Both of these types of anticoagulants should be avoided in patients on hemodialysis.

More aggressive therapy

Under select circumstances a more aggressive approach to the treatment of VTE may be necessary. These options are usually indicated for a patient with a massive deep vein thrombosis of a lower extremity and for certain patients with an upper extremity deep vein thrombosis. Treatments include catheter-directed thrombolytic therapy and endovenous or surgical thrombectomy.

Thrombolytic therapy is recommended for a patient with an acute pulmonary embolism who is clinically unstable (systolic blood pressure lower than 90 mm Hg), if there is no contraindication to its use (bleeding risk or recent stroke or surgery). Thrombolytic therapy is also an option for those at low risk of bleeding with an acute pulmonary embolism who have signs and symptoms of right heart failure proven by echocardiography.

Surgical pulmonary embolectomy for acute massive pulmonary embolism and mechanical thrombectomy for extensive deep vein thrombosis are generally available only at highly sophisticated tertiary care centers.

An inferior vena cava filter is advised in patients with acute deep vein thrombosis or pulmonary embolism who cannot be fully anticoagulated, to prevent the clot from migrating from the lower extremities to the lungs. These filters are available as either permanent or temporary implants. Some temporary versions can remain in place for up to 150 days after insertion.

PREVENTION OF VTE

Prevention is the standard of care for all patients admitted to the hospital and in select individuals as outpatients who are at high risk of VTE.

Mechanical compression (graduated compression stockings, intermittent pneumatic compression devices) has proven effective in reducing the incidence of deep vein thrombosis and pulmonary embolism postoperatively in patients who cannot take anticoagulants. One study has demonstrated that compression stockings may also be effective in preventing VTE during travel.¹²

ABSOLUTE RISK IS LOW

Over the past decade, special attention has been paid to travel as a risk factor for developing VTE.¹³ Traveler’s thrombosis has become an important public health concern. Numerous publications and epidemiologic studies have targeted air travel in an attempt to determine who is at risk and what precautions are necessary to prevent this complication.^1–7,9

The incidence of VTE following air travel is reported to be 3.2 per 1,000 person-years.⁴ While this incidence is relatively low, it is still 3.2 times higher than in the healthy population that is not flying.

The more serious complication of VTE, ie, acute pulmonary embolism, occurs less often. In three studies, the reported incidence ranged from 1.65 per million patients in flights longer than 8 hours to a high of 4.8 per million patients in flights longer than 12 hours or distances exceeding 10,000 km (6,200 miles).^5,14,15 For the 400 passengers on the average long-haul flight of 12 hours, there is at most a 0.2% chance that somebody on the plane will have a symptomatic VTE).

RISK FACTORS IN LONG-DISTANCE TRAVELERS

The risk of traveler’s thrombosis has recently attracted the attention of passengers and the airline industry. Airlines are now openly discussing the risk and providing reminders such as exercises that should be undertaken in-flight (see the patient information page that accompanies this article). Some airlines are recommending that all patients consult their doctor to assess their personal risk of deep vein thrombosis before flying.

The most common risk factors for VTE in travelers are well established and are additive (Table 1). The extent of the additive risk, however, is not entirely clear.

What is clear is that when VTE occurs it is a life-altering and life-threatening event. If it occurs on an outbound trip, the local resources and capabilities available at the destination may not be adequate for optimal treatment. If a traveler experiences a VTE event on an outbound trip, an emergency return trip to the continental United States or a regional center of expertise may be required. There is an additive risk with this subsequent travel event if the patient is not given immediate treatment first (Table 4). Hence, treatment prior to evacuation should be strongly considered.

The traveler must also be aware that VTE can be recognized up to 2 months after a long-haul flight, though it is especially a concern within the first 2 weeks after travel.^2,4,16,17

RECOMMENDATIONS FOR LONG-DISTANCE AIR TRAVELERS

Each person should be evaluated on a case-by-case basis for his or her need for VTE prophylaxis. Medical guidelines for airline passengers have been published by the Aerospace Medical Association and the American College of Chest Physicians (ACCP).^18,19 In general, travelers should:

• Exercise the legs by flexing and extending the ankles at regular intervals while seated (see the patient information material that accompanies this article) and frequently contracting the calf muscles.
• Walk about the cabin periodically, 5 minutes for every hour on longer-duration flights (over 4 hours) and when flight conditions permit.
• Drink adequate amounts of water and fruit juices to maintain good hydration.¹⁷
• Avoid alcohol and caffeinated beverages, which are dehydrating.
• Be careful about eating too much during the flight.
• Request an aisle seat if you are at risk
• Do not place baggage underneath the seat in front of you, because that reduces the ability to move the legs.
• Do not sleep in a cramped position, and avoid the use of any type of sleep aid.
• Avoid wearing constrictive clothing around the lower extremities or waist.

We recommend that all airplane passengers take the steps listed above to reduce venous stasis and avoid dehydration, even though these measures have not been proven effective in clinical trials.¹⁹

The ACCP further advises that decisions about pharmacologic prophylaxis of VTE for airplane passengers at high risk should be made on an individual basis, considering that there are potential adverse effects of prophylaxis and that these may outweigh the benefits. For long-distance travelers with additional risk factors for VTE, we suggest the following:

• Use of properly fitted, below-the-knee graduated compression stockings providing 15 to 30 mm Hg of pressure at the ankle (particularly when large varicosities or leg edema is present)
• For people at very high risk, a single prophylactic dose of a low-molecular-weight heparin or a factor Xa inhibitor injected just before departure (Table 5)
• Aspirin is not recommended as it is not effective for the prevention of VTE.²⁰

SUMMARY FOR THE AIR TRAVELER

All travelers on long flights should perform standard VTE prophylaxis exercises (see the patient information pages accompanying this article). Although VTE is uncommon, people with additional risk factors who travel frequently either on multiple flights in a short period of time or on very long flights should be evaluated on a case-by-case basis for a more aggressive approach to prevention (compression support hose or prophylactic administration of a low-molecular-weight heparin or a factor Xa inhibitor).

Should a VTE event occur during travel, the patient should seek medical care immediately. The standard evaluation of a patient with a suspected VTE should include an estimation of the pretest probability of disease (Table 2, Table 3), followed by duplex ultrasonography of the upper or lower extremity to detect a deep vein thrombosis. If symptoms dictate, then spiral computed tomography, ventilation-perfusion lung scan, or pulmonary angiography (where available) should be ordered to diagnose acute pulmonary embolism. A positive D-dimer blood test alone is not diagnostic and may not be available in more remote locations. A negative D-dimer test result is most helpful to exclude VTE.

Standard therapy for VTE is immediate treatment with one of the anticoagulants listed in Table 4, unless the patient has a contraindication to treatment, such as bleeding or allergy. Immediate evacuation is recommended if the patient has a life-threatening pulmonary embolism, defined as hemodynamic instability (hypotension with a blood pressure under 90 mm Hg systolic or signs of right heart failure) that cannot be treated at a local facility. An air ambulance should be used to transport these patients. If the patient has an iliofemoral deep vein thrombosis, it is also advisable that he or she be considered for evacuation if severe symptoms are present, such as pain, swelling, or cyanosis. Unless contraindicated, all patients should be given either full-dose intravenous or full-dose subcutaneous heparin or subcutaneous injection of a readily available low-molecular-weight heparin preparations or factor Xa inhibitor at once.²¹

Patients often ask primary care physicians and cardiologists about the measurement of biomarkers for cardiovascular disease and about the efficacy of preventive measures.

Although studies have shown that elevated homocysteine is a risk factor for cardiovascular and peripheral arterial disease^1–3 and that supplementation with folic acid, vitamin B₆, and vitamin B₁₂ lowers homocysteine levels,^4,5 it is unclear whether such supplementation prevents cardiovascular events. As a result, there is no consensus about whose homocysteine levels should be measured and who, if anyone, should receive homocysteine-lowering therapies.

The aim of this paper is to examine whether the evidence is sufficient to recommend homocysteine testing to guide the prevention and treatment of cardiovascular disease, or to recommend using folic acid, vitamin B₆, and vitamin B₁₂ for primary or secondary prevention of cardiovascular disease.

HISTORY OF HOMOCYSTEINE AS A RISK MARKER

Homocysteine is an amino acid formed from the metabolism of methionine, an essential amino acid derived from dietary protein. Although homocysteine was first isolated by Butz and du Vigneaud in 1932,⁶ it was not until 1964 that Gibson et al⁷ reported that patients with homocystinuria (more about this below) had vascular anomalies and arterial thrombosis. In 1969, McCully⁸ made the connection between elevated homocysteine levels and the risk of atherosclerosis.

Several possible mechanisms for the association between homocysteine and atherosclerosis have been demonstrated in experimental models. These include stimulation of smooth muscle growth, reduction in endothelial cell growth, impaired endothelial cell relaxation, decreased synthesis of high-density lipoprotein, promotion of autoimmune response, and accumulation of inflammatory monocytes in atherosclerotic plaques.^3,9,10

In view of these findings, researchers have been evaluating whether homocysteine-lowering therapies decrease the risk of cardiovascular disease.

CAUSES OF ELEVATED PLASMA HOMOCYSTEINE

Vitamin deficiencies. Vitamin B₆ (pyridoxine), vitamin B₁₂ (cyanocobalamin), and folic acid are cofactors required for homocysteine metabolism, and deficiency in any or all of these leads to disruption of the relevant metabolic pathways.

Renal impairment. A low glomerular filtration rate has also been correlated with an elevated plasma homocysteine concentration. This makes sense, since the kidneys perform up to 70% of the clearance of homocysteine, although a cause-and-effect relationship is unclear.¹³

Inborn errors of homocysteine metabolism.Homocystinuria, ie, an abnormal elevation of homocysteine in the urine, is caused by several autosomal recessive disorders. People with these genetic variations have extremely high homocysteine levels.

A deficiency in the enzyme cystathionine beta-synthase is quite rare (the incidence in newborn babies has been found to be 1 in 344,000 worldwide and 1 in 65,000 in Ireland and Australia¹⁴), but leads to homocysteine levels greater than 100 μmol/L and often causes cardiovascular disease by the age of 30 years.¹⁵

A deficiency in the enzyme methylene tetrahydrofolate reductase (MTHFR) is a more common cause of mildly to moderately elevated plasma homocysteine levels.¹⁶ The MTHFR deficiency involves a variation at position 677 in the MTHFR gene in which cytosine is replaced by thymidine (thus called C677T or 677C>T).¹⁷ Ten percent of the population are homozygous for this variant (TT), 43% are heterozygous (CT), and 47% are unaffected (CC). Heterozygotes have slightly higher homocysteine levels than unaffected people, while people with the TT genotype have approximately 20% higher homocysteine levels.¹⁷

ELEVATED HOMOCYSTEINE IS COMMON

In a study of a population in Norway from 1992 to 1993, 8.5% had mild elevations in homocysteine (plasma levels 15–29.99 μmol/L), 0.8% had moderate elevations (30–99.99 μmol/L), and 0.02% had severe elevations (≥ 100 μmol/L).^13,18 The prevalence of hyperhomocysteinemia in the United States is probably much lower, given that supplementation of white flour and cereal grains with folic acid has been mandatory since 1998, but this is not well described in the literature.

HOW GREAT IS THE RISK?

Studies over the past 10 to 20 years have shown that elevated homocysteine is a marker of risk of cardiovascular disease. The association was first noted in patients with cystathionine beta-synthase deficiency, who tend to have premature cardiovascular disease.

However, studies of patients with MTHFR 677C>T have yielded mixed results. Although several meta-analyses found up to a 42% higher rate of ischemic heart disease and stroke in patients homozygous for MTHFR 677C>T (the TT genotype) than in those with the CC genotype,^17,19,20 two other large meta-analyses did not find an association between this variant and vascular risk.^21,22

Nonetheless, in a meta-analysis of the association between homocysteine and cardiovascular disease, Wald et al¹⁷ found that for every 5-μmol/L increase in serum homocysteine concentration, the risk of ischemic heart disease increased 20% to 30%.

TRIALS OF HOMOCYSTEINE-LOWERING THERAPY HAVE HAD MIXED RESULTS

Primary prevention of cardiovascular disease

Given the finding that treatment with folic acid lowers homocysteine—initially noted in patients with homocystinuria—researchers hypothesized that treatment with folic acid, vitamin B₆, and vitamin B₁₂ would decrease the risk of cardiovascular disease.

Thus, in its recent evaluation of novel risk markers of cardiovascular disease, the United States Preventive Services Task Force ^28,29 does not recommend measuring the plasma homocysteine level in the evaluation of either low-risk or intermediate-risk populations, finding no evidence that it adds any useful information in predicting major coronary events beyond what one could get from calculating the Framingham Risk Score. The task force also found no evidence that treating people who have elevated homocysteine levels decreases their risk of subsequent cardiovascular events.

In addition, a recent Cochrane Database review of eight randomized controlled trials in patients at low risk did not find a lower risk of myocardial infarction (fatal or nonfatal), stroke, or death from any cause in patients receiving B-complex vitamins.³⁰

Secondary prevention of cardiovascular disease

Bazzano et al,⁴⁷ in a meta-analysis published in 2006, evaluated 12 randomized controlled trials of folic acid supplementation in patients with known cardiovascular disease and did not find that treated patients had better cardiovascular outcomes. The mean homocysteine level was elevated (> 15 μmol/L) at baseline in only 4 of the 12 trials. However, in 1 of these 4 trials, there was no difference in outcomes comparing those with and without elevated homocysteine.³¹

Albert et al⁴ more recently evaluated the effect of a combination pill containing folic acid, vitamin B₆, and vitamin B₁₂ on cardiovascular events in women at high risk, ie, those with a history of cardiovascular disease or having three or more coronary risk factors. Treatment did not decrease the rate of the composite outcome of cardiovascular disease mortality, stroke, myocardial infarction, or coronary revascularization, although the homocysteine level decreased by a mean of 30% in the treated group. However, only 27.7% of the participants had an elevated homocysteine level. One might not expect patients to benefit from such treatment if they had normal homocysteine levels to begin with.

Ebbing et al,⁵ in a trial published in 2008, investigated the effect of folic acid, vitamin B₁₂, and vitamin B₆ supplements on the risks of death from any cause and of cardiovascular events in patients undergoing coronary angiography. Outcomes were no better in the treatment group than in the control group, despite a mean decrease in homocysteine level of 19%. However, over 90% of the participants had a normal homocysteine level.

Mager et al,³² in a study published in 2009, looked specifically at whether patients with coronary artery disease and elevated homocysteine levels (> 15 μmol/L) would benefit from folate-based vitamin therapy. In this subset, the incidence of death from any cause was lower in the treated group than in the control group (4% vs 32%, P < .001), an association that was not present in patients with normal homocysteine levels.

The SEARCH trial (Study of the Effectiveness of Additional Reductions in Cholesterol and Homocysteine),³³ recently published, was a double-blind, randomized controlled trial of vitamin B₁₂ and folic acid treatment in 12,064 patients who had survived a myocardial infarction. Although those who received the vitamin therapy had a 28% reduction in homocysteine level, no clinical benefit was demonstrated. Of note, 66% of the patients had a homocysteine level lower than 14 μmol/L at baseline.

Restenosis after angioplasty

Results are also mixed regarding whether folic acid supplements modify the risk of restenosis after coronary angioplasty.

Namazi et al⁴⁸ evaluated the effect of folic acid supplementation on in-stent restenosis in 200 patients and found no difference between the treatment and placebo groups in the rates of either restenosis or target-vessel revascularization.

Schnyder et al⁴⁹ evaluated the effect of folic acid, vitamin B₆, and vitamin B₁₂ treatment on the rate of coronary restenosis (in cases of balloon angioplasty) or in-stent restenosis (if a stent was used). Patients receiving treatment had lower rates of restenosis or instent restenosis (40% vs 48%, P = .01) and of need for target-vessel revascularization (11% vs 22%, P = .047). The mean homocysteine level was not elevated in this study either, and the researchers did not analyze the outcomes according to whether patients had high or normal homocysteine levels.

Lange et al³⁵ also evaluated the effect of folic acid, vitamin B₆, and vitamin B₁₂ treatment on coronary in-stent restenosis. Paradoxically, the rate was higher with treatment in the overall group (mean homocysteine level 12.2 μmol/L), leading to a higher incidence of target-vessel revascularization. Patients who had a baseline elevation in homocysteine level had a nonsignificant trend toward a lower rate of in-stent restenosis.

The evidence is also mixed for using folic acid and other B vitamins to prevent cerebrovascular disease and peripheral vascular disease. Although a 2007 meta-analysis found that folic acid supplementation decreased the risk of a first stroke by 18% (P = .045),⁵⁰ a later meta-analysis contradicts this finding.⁵¹

A 2009 meta-analysis found that patients with peripheral arterial disease had higher homocysteine levels than controls, but it did not find any benefit from supplementation, owing to heterogeneity of the clinical end points used.⁵² Indeed, a 2009 Cochrane Database Systematic Review found that there were no adequate trials of the treatment of patients with peripheral vascular disease who have elevated plasma homocysteine.⁵³

However, immediately after the Cochrane review was published, Khandanpour et al³⁶ published the results of a trial of the effect of folic acid and 5-methyltetrahydrofolate (an active form of folic acid) supplementation on the ankle-brachial pressure index and the pulse-wave velocity in patients with peripheral arterial disease. These measures improved with 16 weeks of treatment. For the ankle-brachial pressure index, the P value was less than .01 for folic acid and .009 for 5-methyltetrahydrofolate; for the pulse-wave velocity, the P value was .051 for folic acid and .011 for 5-methyltetrahydrofolate.

Kidney disease

One could postulate that patients with end-stage renal disease or chronic kidney disease might benefit the most from folic acid supplementation, given the correlation of elevations in homocysteine levels with decline in glomerular filtration rate.

However, only one study found a lower rate of cardiovascular events with folic acid supplementation in dialysis patients, and the difference was not statistically significant (25% vs 36%, P < .08).³¹ Further, several studies found no benefit of folic acid supplementation in patients with chronic kidney disease.^11,12,37

FUTURE DIRECTIONS AND RECOMMENDATIONS

Many experts have suggested that the existing evidence indicates that the homocysteine-lowering therapies folic acid, vitamin B₆, and vitamin B₁₂ do not lower the risk of cardiovascular disease.^38,54–59 Indeed, the American Heart Association guidelines for cardiovascular disease prevention in women do not recommend folic acid supplementation to prevent cardiovascular disease.⁶⁰ (Recommendations for men are the same as for women.) However, most of the clinical trials have not selected and treated patients with elevated homocysteine levels, but have instead included all patients regardless of homocysteine level.

At least two large ongoing trials are currently evaluating B-vitamin therapy for secondary prevention, but neither trial is looking specifically at patients with elevated homocysteine levels.^61,62

Thus, instead of concluding that no patients could benefit from homocysteine-lowering treatment, future studies need to clarify:

• Whether patients with elevated homocysteine would benefit from such treatment
• At what level it would be appropriate to start treatment
• The appropriate target homocysteine level with treatment.

Particularly given the recent finding that folic acid supplementation may increase cancer risk,⁶³ these questions need closer scrutiny.

Patients often ask primary care physicians and cardiologists about the measurement of biomarkers for cardiovascular disease and about the efficacy of preventive measures.

Although studies have shown that elevated homocysteine is a risk factor for cardiovascular and peripheral arterial disease^1–3 and that supplementation with folic acid, vitamin B₆, and vitamin B₁₂ lowers homocysteine levels,^4,5 it is unclear whether such supplementation prevents cardiovascular events. As a result, there is no consensus about whose homocysteine levels should be measured and who, if anyone, should receive homocysteine-lowering therapies.

The aim of this paper is to examine whether the evidence is sufficient to recommend homocysteine testing to guide the prevention and treatment of cardiovascular disease, or to recommend using folic acid, vitamin B₆, and vitamin B₁₂ for primary or secondary prevention of cardiovascular disease.

HISTORY OF HOMOCYSTEINE AS A RISK MARKER

Homocysteine is an amino acid formed from the metabolism of methionine, an essential amino acid derived from dietary protein. Although homocysteine was first isolated by Butz and du Vigneaud in 1932,⁶ it was not until 1964 that Gibson et al⁷ reported that patients with homocystinuria (more about this below) had vascular anomalies and arterial thrombosis. In 1969, McCully⁸ made the connection between elevated homocysteine levels and the risk of atherosclerosis.

Several possible mechanisms for the association between homocysteine and atherosclerosis have been demonstrated in experimental models. These include stimulation of smooth muscle growth, reduction in endothelial cell growth, impaired endothelial cell relaxation, decreased synthesis of high-density lipoprotein, promotion of autoimmune response, and accumulation of inflammatory monocytes in atherosclerotic plaques.^3,9,10

In view of these findings, researchers have been evaluating whether homocysteine-lowering therapies decrease the risk of cardiovascular disease.

CAUSES OF ELEVATED PLASMA HOMOCYSTEINE

Vitamin deficiencies. Vitamin B₆ (pyridoxine), vitamin B₁₂ (cyanocobalamin), and folic acid are cofactors required for homocysteine metabolism, and deficiency in any or all of these leads to disruption of the relevant metabolic pathways.

Renal impairment. A low glomerular filtration rate has also been correlated with an elevated plasma homocysteine concentration. This makes sense, since the kidneys perform up to 70% of the clearance of homocysteine, although a cause-and-effect relationship is unclear.¹³

Inborn errors of homocysteine metabolism.Homocystinuria, ie, an abnormal elevation of homocysteine in the urine, is caused by several autosomal recessive disorders. People with these genetic variations have extremely high homocysteine levels.

A deficiency in the enzyme cystathionine beta-synthase is quite rare (the incidence in newborn babies has been found to be 1 in 344,000 worldwide and 1 in 65,000 in Ireland and Australia¹⁴), but leads to homocysteine levels greater than 100 μmol/L and often causes cardiovascular disease by the age of 30 years.¹⁵

A deficiency in the enzyme methylene tetrahydrofolate reductase (MTHFR) is a more common cause of mildly to moderately elevated plasma homocysteine levels.¹⁶ The MTHFR deficiency involves a variation at position 677 in the MTHFR gene in which cytosine is replaced by thymidine (thus called C677T or 677C>T).¹⁷ Ten percent of the population are homozygous for this variant (TT), 43% are heterozygous (CT), and 47% are unaffected (CC). Heterozygotes have slightly higher homocysteine levels than unaffected people, while people with the TT genotype have approximately 20% higher homocysteine levels.¹⁷

ELEVATED HOMOCYSTEINE IS COMMON

In a study of a population in Norway from 1992 to 1993, 8.5% had mild elevations in homocysteine (plasma levels 15–29.99 μmol/L), 0.8% had moderate elevations (30–99.99 μmol/L), and 0.02% had severe elevations (≥ 100 μmol/L).^13,18 The prevalence of hyperhomocysteinemia in the United States is probably much lower, given that supplementation of white flour and cereal grains with folic acid has been mandatory since 1998, but this is not well described in the literature.

HOW GREAT IS THE RISK?

Studies over the past 10 to 20 years have shown that elevated homocysteine is a marker of risk of cardiovascular disease. The association was first noted in patients with cystathionine beta-synthase deficiency, who tend to have premature cardiovascular disease.

However, studies of patients with MTHFR 677C>T have yielded mixed results. Although several meta-analyses found up to a 42% higher rate of ischemic heart disease and stroke in patients homozygous for MTHFR 677C>T (the TT genotype) than in those with the CC genotype,^17,19,20 two other large meta-analyses did not find an association between this variant and vascular risk.^21,22

Nonetheless, in a meta-analysis of the association between homocysteine and cardiovascular disease, Wald et al¹⁷ found that for every 5-μmol/L increase in serum homocysteine concentration, the risk of ischemic heart disease increased 20% to 30%.

TRIALS OF HOMOCYSTEINE-LOWERING THERAPY HAVE HAD MIXED RESULTS

Primary prevention of cardiovascular disease

Given the finding that treatment with folic acid lowers homocysteine—initially noted in patients with homocystinuria—researchers hypothesized that treatment with folic acid, vitamin B₆, and vitamin B₁₂ would decrease the risk of cardiovascular disease.

Thus, in its recent evaluation of novel risk markers of cardiovascular disease, the United States Preventive Services Task Force ^28,29 does not recommend measuring the plasma homocysteine level in the evaluation of either low-risk or intermediate-risk populations, finding no evidence that it adds any useful information in predicting major coronary events beyond what one could get from calculating the Framingham Risk Score. The task force also found no evidence that treating people who have elevated homocysteine levels decreases their risk of subsequent cardiovascular events.

In addition, a recent Cochrane Database review of eight randomized controlled trials in patients at low risk did not find a lower risk of myocardial infarction (fatal or nonfatal), stroke, or death from any cause in patients receiving B-complex vitamins.³⁰

Secondary prevention of cardiovascular disease

Bazzano et al,⁴⁷ in a meta-analysis published in 2006, evaluated 12 randomized controlled trials of folic acid supplementation in patients with known cardiovascular disease and did not find that treated patients had better cardiovascular outcomes. The mean homocysteine level was elevated (> 15 μmol/L) at baseline in only 4 of the 12 trials. However, in 1 of these 4 trials, there was no difference in outcomes comparing those with and without elevated homocysteine.³¹

Albert et al⁴ more recently evaluated the effect of a combination pill containing folic acid, vitamin B₆, and vitamin B₁₂ on cardiovascular events in women at high risk, ie, those with a history of cardiovascular disease or having three or more coronary risk factors. Treatment did not decrease the rate of the composite outcome of cardiovascular disease mortality, stroke, myocardial infarction, or coronary revascularization, although the homocysteine level decreased by a mean of 30% in the treated group. However, only 27.7% of the participants had an elevated homocysteine level. One might not expect patients to benefit from such treatment if they had normal homocysteine levels to begin with.

Ebbing et al,⁵ in a trial published in 2008, investigated the effect of folic acid, vitamin B₁₂, and vitamin B₆ supplements on the risks of death from any cause and of cardiovascular events in patients undergoing coronary angiography. Outcomes were no better in the treatment group than in the control group, despite a mean decrease in homocysteine level of 19%. However, over 90% of the participants had a normal homocysteine level.

Mager et al,³² in a study published in 2009, looked specifically at whether patients with coronary artery disease and elevated homocysteine levels (> 15 μmol/L) would benefit from folate-based vitamin therapy. In this subset, the incidence of death from any cause was lower in the treated group than in the control group (4% vs 32%, P < .001), an association that was not present in patients with normal homocysteine levels.

The SEARCH trial (Study of the Effectiveness of Additional Reductions in Cholesterol and Homocysteine),³³ recently published, was a double-blind, randomized controlled trial of vitamin B₁₂ and folic acid treatment in 12,064 patients who had survived a myocardial infarction. Although those who received the vitamin therapy had a 28% reduction in homocysteine level, no clinical benefit was demonstrated. Of note, 66% of the patients had a homocysteine level lower than 14 μmol/L at baseline.

Restenosis after angioplasty

Results are also mixed regarding whether folic acid supplements modify the risk of restenosis after coronary angioplasty.

Namazi et al⁴⁸ evaluated the effect of folic acid supplementation on in-stent restenosis in 200 patients and found no difference between the treatment and placebo groups in the rates of either restenosis or target-vessel revascularization.

Schnyder et al⁴⁹ evaluated the effect of folic acid, vitamin B₆, and vitamin B₁₂ treatment on the rate of coronary restenosis (in cases of balloon angioplasty) or in-stent restenosis (if a stent was used). Patients receiving treatment had lower rates of restenosis or instent restenosis (40% vs 48%, P = .01) and of need for target-vessel revascularization (11% vs 22%, P = .047). The mean homocysteine level was not elevated in this study either, and the researchers did not analyze the outcomes according to whether patients had high or normal homocysteine levels.

Lange et al³⁵ also evaluated the effect of folic acid, vitamin B₆, and vitamin B₁₂ treatment on coronary in-stent restenosis. Paradoxically, the rate was higher with treatment in the overall group (mean homocysteine level 12.2 μmol/L), leading to a higher incidence of target-vessel revascularization. Patients who had a baseline elevation in homocysteine level had a nonsignificant trend toward a lower rate of in-stent restenosis.

The evidence is also mixed for using folic acid and other B vitamins to prevent cerebrovascular disease and peripheral vascular disease. Although a 2007 meta-analysis found that folic acid supplementation decreased the risk of a first stroke by 18% (P = .045),⁵⁰ a later meta-analysis contradicts this finding.⁵¹

A 2009 meta-analysis found that patients with peripheral arterial disease had higher homocysteine levels than controls, but it did not find any benefit from supplementation, owing to heterogeneity of the clinical end points used.⁵² Indeed, a 2009 Cochrane Database Systematic Review found that there were no adequate trials of the treatment of patients with peripheral vascular disease who have elevated plasma homocysteine.⁵³

However, immediately after the Cochrane review was published, Khandanpour et al³⁶ published the results of a trial of the effect of folic acid and 5-methyltetrahydrofolate (an active form of folic acid) supplementation on the ankle-brachial pressure index and the pulse-wave velocity in patients with peripheral arterial disease. These measures improved with 16 weeks of treatment. For the ankle-brachial pressure index, the P value was less than .01 for folic acid and .009 for 5-methyltetrahydrofolate; for the pulse-wave velocity, the P value was .051 for folic acid and .011 for 5-methyltetrahydrofolate.

Kidney disease

One could postulate that patients with end-stage renal disease or chronic kidney disease might benefit the most from folic acid supplementation, given the correlation of elevations in homocysteine levels with decline in glomerular filtration rate.

However, only one study found a lower rate of cardiovascular events with folic acid supplementation in dialysis patients, and the difference was not statistically significant (25% vs 36%, P < .08).³¹ Further, several studies found no benefit of folic acid supplementation in patients with chronic kidney disease.^11,12,37

FUTURE DIRECTIONS AND RECOMMENDATIONS

Many experts have suggested that the existing evidence indicates that the homocysteine-lowering therapies folic acid, vitamin B₆, and vitamin B₁₂ do not lower the risk of cardiovascular disease.^38,54–59 Indeed, the American Heart Association guidelines for cardiovascular disease prevention in women do not recommend folic acid supplementation to prevent cardiovascular disease.⁶⁰ (Recommendations for men are the same as for women.) However, most of the clinical trials have not selected and treated patients with elevated homocysteine levels, but have instead included all patients regardless of homocysteine level.

At least two large ongoing trials are currently evaluating B-vitamin therapy for secondary prevention, but neither trial is looking specifically at patients with elevated homocysteine levels.^61,62

Thus, instead of concluding that no patients could benefit from homocysteine-lowering treatment, future studies need to clarify:

• Whether patients with elevated homocysteine would benefit from such treatment
• At what level it would be appropriate to start treatment
• The appropriate target homocysteine level with treatment.

Particularly given the recent finding that folic acid supplementation may increase cancer risk,⁶³ these questions need closer scrutiny.

Patients often ask primary care physicians and cardiologists about the measurement of biomarkers for cardiovascular disease and about the efficacy of preventive measures.

Although studies have shown that elevated homocysteine is a risk factor for cardiovascular and peripheral arterial disease^1–3 and that supplementation with folic acid, vitamin B₆, and vitamin B₁₂ lowers homocysteine levels,^4,5 it is unclear whether such supplementation prevents cardiovascular events. As a result, there is no consensus about whose homocysteine levels should be measured and who, if anyone, should receive homocysteine-lowering therapies.

The aim of this paper is to examine whether the evidence is sufficient to recommend homocysteine testing to guide the prevention and treatment of cardiovascular disease, or to recommend using folic acid, vitamin B₆, and vitamin B₁₂ for primary or secondary prevention of cardiovascular disease.

HISTORY OF HOMOCYSTEINE AS A RISK MARKER

Homocysteine is an amino acid formed from the metabolism of methionine, an essential amino acid derived from dietary protein. Although homocysteine was first isolated by Butz and du Vigneaud in 1932,⁶ it was not until 1964 that Gibson et al⁷ reported that patients with homocystinuria (more about this below) had vascular anomalies and arterial thrombosis. In 1969, McCully⁸ made the connection between elevated homocysteine levels and the risk of atherosclerosis.

Several possible mechanisms for the association between homocysteine and atherosclerosis have been demonstrated in experimental models. These include stimulation of smooth muscle growth, reduction in endothelial cell growth, impaired endothelial cell relaxation, decreased synthesis of high-density lipoprotein, promotion of autoimmune response, and accumulation of inflammatory monocytes in atherosclerotic plaques.^3,9,10

In view of these findings, researchers have been evaluating whether homocysteine-lowering therapies decrease the risk of cardiovascular disease.

CAUSES OF ELEVATED PLASMA HOMOCYSTEINE

Vitamin deficiencies. Vitamin B₆ (pyridoxine), vitamin B₁₂ (cyanocobalamin), and folic acid are cofactors required for homocysteine metabolism, and deficiency in any or all of these leads to disruption of the relevant metabolic pathways.

Renal impairment. A low glomerular filtration rate has also been correlated with an elevated plasma homocysteine concentration. This makes sense, since the kidneys perform up to 70% of the clearance of homocysteine, although a cause-and-effect relationship is unclear.¹³

Inborn errors of homocysteine metabolism.Homocystinuria, ie, an abnormal elevation of homocysteine in the urine, is caused by several autosomal recessive disorders. People with these genetic variations have extremely high homocysteine levels.

A deficiency in the enzyme cystathionine beta-synthase is quite rare (the incidence in newborn babies has been found to be 1 in 344,000 worldwide and 1 in 65,000 in Ireland and Australia¹⁴), but leads to homocysteine levels greater than 100 μmol/L and often causes cardiovascular disease by the age of 30 years.¹⁵

A deficiency in the enzyme methylene tetrahydrofolate reductase (MTHFR) is a more common cause of mildly to moderately elevated plasma homocysteine levels.¹⁶ The MTHFR deficiency involves a variation at position 677 in the MTHFR gene in which cytosine is replaced by thymidine (thus called C677T or 677C>T).¹⁷ Ten percent of the population are homozygous for this variant (TT), 43% are heterozygous (CT), and 47% are unaffected (CC). Heterozygotes have slightly higher homocysteine levels than unaffected people, while people with the TT genotype have approximately 20% higher homocysteine levels.¹⁷

ELEVATED HOMOCYSTEINE IS COMMON

In a study of a population in Norway from 1992 to 1993, 8.5% had mild elevations in homocysteine (plasma levels 15–29.99 μmol/L), 0.8% had moderate elevations (30–99.99 μmol/L), and 0.02% had severe elevations (≥ 100 μmol/L).^13,18 The prevalence of hyperhomocysteinemia in the United States is probably much lower, given that supplementation of white flour and cereal grains with folic acid has been mandatory since 1998, but this is not well described in the literature.

HOW GREAT IS THE RISK?

Studies over the past 10 to 20 years have shown that elevated homocysteine is a marker of risk of cardiovascular disease. The association was first noted in patients with cystathionine beta-synthase deficiency, who tend to have premature cardiovascular disease.

However, studies of patients with MTHFR 677C>T have yielded mixed results. Although several meta-analyses found up to a 42% higher rate of ischemic heart disease and stroke in patients homozygous for MTHFR 677C>T (the TT genotype) than in those with the CC genotype,^17,19,20 two other large meta-analyses did not find an association between this variant and vascular risk.^21,22

Nonetheless, in a meta-analysis of the association between homocysteine and cardiovascular disease, Wald et al¹⁷ found that for every 5-μmol/L increase in serum homocysteine concentration, the risk of ischemic heart disease increased 20% to 30%.

TRIALS OF HOMOCYSTEINE-LOWERING THERAPY HAVE HAD MIXED RESULTS

Primary prevention of cardiovascular disease

Given the finding that treatment with folic acid lowers homocysteine—initially noted in patients with homocystinuria—researchers hypothesized that treatment with folic acid, vitamin B₆, and vitamin B₁₂ would decrease the risk of cardiovascular disease.

Thus, in its recent evaluation of novel risk markers of cardiovascular disease, the United States Preventive Services Task Force ^28,29 does not recommend measuring the plasma homocysteine level in the evaluation of either low-risk or intermediate-risk populations, finding no evidence that it adds any useful information in predicting major coronary events beyond what one could get from calculating the Framingham Risk Score. The task force also found no evidence that treating people who have elevated homocysteine levels decreases their risk of subsequent cardiovascular events.

In addition, a recent Cochrane Database review of eight randomized controlled trials in patients at low risk did not find a lower risk of myocardial infarction (fatal or nonfatal), stroke, or death from any cause in patients receiving B-complex vitamins.³⁰

Secondary prevention of cardiovascular disease

Bazzano et al,⁴⁷ in a meta-analysis published in 2006, evaluated 12 randomized controlled trials of folic acid supplementation in patients with known cardiovascular disease and did not find that treated patients had better cardiovascular outcomes. The mean homocysteine level was elevated (> 15 μmol/L) at baseline in only 4 of the 12 trials. However, in 1 of these 4 trials, there was no difference in outcomes comparing those with and without elevated homocysteine.³¹

Albert et al⁴ more recently evaluated the effect of a combination pill containing folic acid, vitamin B₆, and vitamin B₁₂ on cardiovascular events in women at high risk, ie, those with a history of cardiovascular disease or having three or more coronary risk factors. Treatment did not decrease the rate of the composite outcome of cardiovascular disease mortality, stroke, myocardial infarction, or coronary revascularization, although the homocysteine level decreased by a mean of 30% in the treated group. However, only 27.7% of the participants had an elevated homocysteine level. One might not expect patients to benefit from such treatment if they had normal homocysteine levels to begin with.

Ebbing et al,⁵ in a trial published in 2008, investigated the effect of folic acid, vitamin B₁₂, and vitamin B₆ supplements on the risks of death from any cause and of cardiovascular events in patients undergoing coronary angiography. Outcomes were no better in the treatment group than in the control group, despite a mean decrease in homocysteine level of 19%. However, over 90% of the participants had a normal homocysteine level.

Mager et al,³² in a study published in 2009, looked specifically at whether patients with coronary artery disease and elevated homocysteine levels (> 15 μmol/L) would benefit from folate-based vitamin therapy. In this subset, the incidence of death from any cause was lower in the treated group than in the control group (4% vs 32%, P < .001), an association that was not present in patients with normal homocysteine levels.

The SEARCH trial (Study of the Effectiveness of Additional Reductions in Cholesterol and Homocysteine),³³ recently published, was a double-blind, randomized controlled trial of vitamin B₁₂ and folic acid treatment in 12,064 patients who had survived a myocardial infarction. Although those who received the vitamin therapy had a 28% reduction in homocysteine level, no clinical benefit was demonstrated. Of note, 66% of the patients had a homocysteine level lower than 14 μmol/L at baseline.

Restenosis after angioplasty

Results are also mixed regarding whether folic acid supplements modify the risk of restenosis after coronary angioplasty.

Namazi et al⁴⁸ evaluated the effect of folic acid supplementation on in-stent restenosis in 200 patients and found no difference between the treatment and placebo groups in the rates of either restenosis or target-vessel revascularization.

Schnyder et al⁴⁹ evaluated the effect of folic acid, vitamin B₆, and vitamin B₁₂ treatment on the rate of coronary restenosis (in cases of balloon angioplasty) or in-stent restenosis (if a stent was used). Patients receiving treatment had lower rates of restenosis or instent restenosis (40% vs 48%, P = .01) and of need for target-vessel revascularization (11% vs 22%, P = .047). The mean homocysteine level was not elevated in this study either, and the researchers did not analyze the outcomes according to whether patients had high or normal homocysteine levels.

Lange et al³⁵ also evaluated the effect of folic acid, vitamin B₆, and vitamin B₁₂ treatment on coronary in-stent restenosis. Paradoxically, the rate was higher with treatment in the overall group (mean homocysteine level 12.2 μmol/L), leading to a higher incidence of target-vessel revascularization. Patients who had a baseline elevation in homocysteine level had a nonsignificant trend toward a lower rate of in-stent restenosis.

The evidence is also mixed for using folic acid and other B vitamins to prevent cerebrovascular disease and peripheral vascular disease. Although a 2007 meta-analysis found that folic acid supplementation decreased the risk of a first stroke by 18% (P = .045),⁵⁰ a later meta-analysis contradicts this finding.⁵¹

A 2009 meta-analysis found that patients with peripheral arterial disease had higher homocysteine levels than controls, but it did not find any benefit from supplementation, owing to heterogeneity of the clinical end points used.⁵² Indeed, a 2009 Cochrane Database Systematic Review found that there were no adequate trials of the treatment of patients with peripheral vascular disease who have elevated plasma homocysteine.⁵³

However, immediately after the Cochrane review was published, Khandanpour et al³⁶ published the results of a trial of the effect of folic acid and 5-methyltetrahydrofolate (an active form of folic acid) supplementation on the ankle-brachial pressure index and the pulse-wave velocity in patients with peripheral arterial disease. These measures improved with 16 weeks of treatment. For the ankle-brachial pressure index, the P value was less than .01 for folic acid and .009 for 5-methyltetrahydrofolate; for the pulse-wave velocity, the P value was .051 for folic acid and .011 for 5-methyltetrahydrofolate.

Kidney disease

One could postulate that patients with end-stage renal disease or chronic kidney disease might benefit the most from folic acid supplementation, given the correlation of elevations in homocysteine levels with decline in glomerular filtration rate.

However, only one study found a lower rate of cardiovascular events with folic acid supplementation in dialysis patients, and the difference was not statistically significant (25% vs 36%, P < .08).³¹ Further, several studies found no benefit of folic acid supplementation in patients with chronic kidney disease.^11,12,37

FUTURE DIRECTIONS AND RECOMMENDATIONS

Many experts have suggested that the existing evidence indicates that the homocysteine-lowering therapies folic acid, vitamin B₆, and vitamin B₁₂ do not lower the risk of cardiovascular disease.^38,54–59 Indeed, the American Heart Association guidelines for cardiovascular disease prevention in women do not recommend folic acid supplementation to prevent cardiovascular disease.⁶⁰ (Recommendations for men are the same as for women.) However, most of the clinical trials have not selected and treated patients with elevated homocysteine levels, but have instead included all patients regardless of homocysteine level.

At least two large ongoing trials are currently evaluating B-vitamin therapy for secondary prevention, but neither trial is looking specifically at patients with elevated homocysteine levels.^61,62

Thus, instead of concluding that no patients could benefit from homocysteine-lowering treatment, future studies need to clarify:

• Whether patients with elevated homocysteine would benefit from such treatment
• At what level it would be appropriate to start treatment
• The appropriate target homocysteine level with treatment.

Particularly given the recent finding that folic acid supplementation may increase cancer risk,⁶³ these questions need closer scrutiny.

For patients with carotid artery stenosis, percutaneous intervention with stenting is as good as surgery (carotid endarterectomy). This was the major finding of the recently completed Carotid Revascularization Endarterectomy Versus Stenting Trial (CREST)¹—with some qualifications.

CREST is the latest in a series of clinical trials of treatment of carotid stenosis that have generated reams of numbers and much debate. The topic of surgery vs percutaneous intervention is a moving target, as techniques evolve and improve. We believe the CREST results are valuable and should help inform decisions about treatment in the “real world.”

In this article, we offer a critical review of CREST, with a careful evaluation of its methods, results, and conclusions.

AN EVOLVING FIELD

Despite improvements in diagnosis and management, stroke remains one of the leading causes of morbidity and death in the United States, with an annual incidence of 780,000 cases and 270,000 deaths.^2,3

Randomized trials of stenting have had mixed results, leading the Centers for Medicare and Medicaid Services (CMS) to adopt strict reimbursement policies. Currently, CMS reimburses for stenting only in symptomatic cases with at least 50% carotid artery stenosis. It also reimburses for stenting in asymptomatic cases in patients at high risk with 80% or greater stenosis, but only if the patients are enrolled in ongoing clinical trials or registries.

CREST compared stenting with endarterectomy and provided important insights into each approach.¹

BEFORE CREST

Endarterectomy is superior to medical therapy for symptomatic stenosis

First described in 1953, carotid endarterectomy became the most widely used invasive treatment for significant carotid stenosis.⁹ Several studies have described patient subsets that benefit from this procedure.

NASCET (the North American Symptomatic Carotid Endarterectomy Trial)¹⁰ assigned 2,226 patients with symptomatic stenosis (transient ischemic attack or stroke within the past 180 days) to medical management or endarterectomy.

Surgery was associated with a 65% lower rate of ipsilateral cerebral events in patients with 70% or greater stenosis.¹⁰ Surgery was also found to be superior in patients with moderate disease (50% to 69% stenosis), but the difference only approached statistical significance. In patients with stenosis of less than 50%, the outcomes were similar with endarterectomy and medical management.¹¹

ECST (the European Carotid Surgery Trial)¹² included a similar population of 3,024 patients. Those with high-grade disease (stenosis ≥ 80%) had significantly better outcomes with endarterectomy, but in those with stenosis less than 70%, surgery was no better than drug therapy.

Comment. NASCET and ECST taught us that endarterectomy is clearly superior to medical therapy in patients with severe symptomatic carotid disease. However, both trials excluded patients at high surgical risk, eg, those with severe coronary artery disease, kidney disease, or heart failure. Additionally, medical management was not aggressive by today’s standards in terms of control of blood pressure and hyperlipidemia, and this could have skewed the results in favor of carotid endarterectomy.

The case for carotid endarterectomy for asymptomatic stenosis

Endarterectomy has also been compared with drug therapy for asymp tomatic carotid artery stenosis in several trials.^13–15

ACAS (the Asymptomatic Carotid Atherosclerosis Study)¹⁵ assigned 1,662 patients who had no symptoms and had at least 60% carotid artery stenosis to endarterectomy or to medical management, and found a relative risk reduction of 53% in favor of surgery.¹⁵

The Veterans Affairs Cooperative Study Group¹⁴ corroborated these results in 444 patients with asymptomatic stenosis of greater than 50%. Endarterectomy was associated with a 61% lower risk of transient ischemic attack, transient monocular blindness, or stroke compared with medical therapy. However, there was no statistically significant difference in rates of stroke or death at 30 days.¹⁴

ACST (the Asymptomatic Carotid Surgery Trial),¹³ the largest study to compare carotid endarterectomy with drug therapy for asymptomatic stenosis, randomized 3,120 patients to surgery or drug therapy. The net 5-year risk of stroke was 6.4% with endarterectomy vs 11.8% with drug therapy (P < .0001). The rate of fatal stroke was also lower with endarterectomy: 2.1% vs 4.2% (P = .006).¹³

Comment. The results of these and other studies of endarterectomy vs medical therapy may not be applicable to current practice, since medical therapy has evolved and the risks with current drug therapy are likely much lower than seen in these trials, some of which began 2 decades ago. Another problem with interpreting these trials is that they excluded surgically “high-risk” patients, which limits the generalizability of the findings to this particular patient population.

The American Heart Association and the American Stroke Association have, on the basis of these trials, recommended carotid endarterectomy in patients with^7,8,16:

• Ipsilateral, symptomatic carotid artery stenosis of 70% to 99% (class I, level of evidence A)
• Symptomatic stenosis of 50% to 69%, depending on patient-specific factors such as age, sex, and comorbidities
• High-grade asymptomatic carotid stenosis, if the patients are carefully selected and the surgery is performed by surgeons with procedural morbidity and mortality rates of less than 3% (class I, level of evidence A).

While carotid endarterectomy is proven to be more efficacious than medical management in certain patient subsets, studies favoring surgery over medical therapy have been criticized because they excluded patients with significant comorbidities. In addition, surgery has been associated with significant cardiovascular events, wound complications, and cranial nerve damage, and it requires general anesthesia in most cases.^12,17–19 These and other factors spurred the development of less invasive, percutaneous approaches for patients with substantial comorbidities.

So far, several trials have investigated carotid angioplasty with or without stents and with or without devices to capture distal emboli. This interest set the stage for CREST.^20,21

Initial attempts at angioplasty without distal protection were not very successful. A meta-analysis of nonrandomized trials that included 714 patients from the initial 13 studies of angioplasty (with or without stenting) and 6,970 patients from 20 studies of carotid endarterectomy found angioplasty to be possibly associated with higher rates of stroke within 30 days of the procedure.²⁰

With improvements in technology, routine use of embolic protection devices, more experience, and better selection of patients, the outcome of carotid stenting has improved. In fact, a meta-analysis comparing stenting without an embolic protection device (26 trials with 2,357 patients) vs stenting with an embolic protection device (11 trials with 839 patients) showed that embolic protection led to significantly better outcomes with fewer strokes—outcomes arguably similar to those of carotid endarterectomy.²¹

SAPPHIRE (the Stenting and Angioplasty With Protection in Patients at High Risk for Endarterectomy trial)²² was the only completed US trial until CREST that compared carotid artery stenting with distal protection against surgery. It included 334 high-risk patients with either symptomatic stenosis of 50% or greater or asymptomatic stenosis of 80% or greater.

The results suggested that the outcomes with stenting with embolic protection were in fact similar to those of endarterectomy, with possibly fewer complications.²³ The benefit persisted up to 2 years.²²

The US Food and Drug Administration (FDA), on the basis of these data, approved the use of stenting with distal protection for high-risk patients, and the CMS reimburses for symptomatic stenosis of 50% or greater and for asymptomatic stenosis of 80% or greater as long as the patient is enrolled in a registry.

SPACE (the Stent-Protected Angioplasty Versus Carotid Endarterectomy in Symptomatic Patients trial),²⁴ conducted in Germany, included 1,214 patients with symptomatic stenosis of at least 50%. Results were similar in terms of the combined primary end point of stroke or death at 30 days. However, the results were not similar enough to prove that stenting is not inferior to surgery, according to preset study criteria.

EVA-3S (the Endarterectomy Versus Stenting in Patients With Symptomatic Severe Carotid Stenosis trial),²⁵ in France, evaluated 527 patients with symptomatic carotid disease (stenosis ≥ 60%), but was terminated early due to significantly higher rates of death or stroke at 30 days in the stenting group.

Comment. SPACE and EVA-3S have been widely criticized for not mandating the use of an embolic protection device (used in 27% of cases in SPACE and in 91.9% of cases in EVA-3S). Questions were also raised about the experience level of the operators who performed the carotid stenting: up to 39% of the primary operators involved in stent placement were trainees.²⁶ Also, myocardial infarction (MI), an important complication of carotid endarterectomy, was not included in the primary end point.

ICSS (the International Carotid Stenting Study)²⁷ compared stenting with endarterectomy in 1,713 patients with symptomatic carotid stenosis of greater than 50%. The primary end point was the rate of fatal or disabling stroke at 3 years.

An interim safety analysis at 120 days of follow-up showed the primary end point had occurred in 4.0% of stenting cases vs 3.2% of endarterectomy cases, a difference that was not statistically significant (hazard ratio [HR] 1.28, 95% confidence interval [CI] 0.77–2.11). However, the risk of any stroke was higher with stenting, with a rate of 7.7% vs 4.1% in the surgical group—a statistically significant difference (HR 1.92, 95% CI 1.27–2.89).

In a substudy of ICSS,²⁸ the investigators corroborated these findings, using magnetic resonance imaging to evaluate for new ischemic brain lesions periprocedurally. They found more new ischemic brain lesions in patients who underwent stenting than in patients who underwent surgery—a statistically significant finding.

Comment. ICSS had limitations: eg, it included only patients with symptoms, and the training for the stenting procedure was not standardized. Furthermore, the use of embolic protection devices was not mandated in stenting procedures.

Because of the controversial and incongruous findings of the above trials, there has been much anticipation for further large, appropriately conducted, randomized controlled trials such as CREST.

CREST STUDY DESIGN

CREST was a prospective, multicenter randomized controlled trial with blinded end point adjudication. Assignment to stenting or surgery occurred in a one-to-one fashion, and patients were stratified by medical center and symptomatic status.

Conducted at 108 sites in the United States and nine sites in Canada, CREST was supported by a grant from the National Institutes of Health and by the manufacturer of the catheter and stent delivery and embolic protection systems. The manufacturer’s representative held a nonvoting position on the executive committee and reviewed the manuscript of the results before submission.

CREST included patients with or without symptoms

CREST was initially designed to compare carotid artery stenting vs carotid endarterectomy in patients with symptoms, but enrollment was later extended to patients without symptoms.

Patients with symptoms were included if they had stenosis of at least 50% on angiography, at least 70% on ultrasonography, or at least 70% on computed tomographic angiography or magnetic resonance angiography if stenosis on ultrasonography was 50% to 69%. Carotid artery stenosis was considered symptomatic if the patient had a transient ischemic attack, amaurosis fugax, or minor disabling stroke in the hemisphere supplied by the target vessel within 180 days of randomization.

Patients without symptoms were eligible if they had at least 60% stenosis on angiography, at least 70% stenosis on ultrasonography, or at least 80% stenosis on computed tomographic angiography or magnetic resonance angiography if the stenosis was 50% to 69% on ultrasonography.

Other eligibility criteria included favorable anatomy and clinical stability for both stenting and surgical procedures.

Exclusion criteria were evolving stroke, history of major stroke, chronic or paroxysmal atrial fibrillation on anticoagulation therapy, MI within the previous 30 days, and unstable angina.

Patients undergoing stenting received aspirin and clopidogrel (Plavix) before and up to 30 days after the procedure. Continuation of antiplatelet therapy was recommended beyond 1 month.

Patients undergoing endarterectomy received aspirin before surgery and continued to receive aspirin for at least 1 year.

Alternatives to aspirin in both groups were ticlopidine (Ticlid), clopidogrel, or aspirin with extended-release dipyridamole (Aggrenox).

End points: Stroke, MI, death

The primary end point was a composite of periprocedural clinical stroke (any type), MI, or death, and of ipsilateral stroke up to 4 years after the procedure. Secondary analyses were also planned for evaluation of treatment modification by age, symptom status, and sex.

Stroke was defined as any acute neurologic ischemic event lasting at least 24 hours with focal signs and symptoms.

Two separate definitions were applied to distinguish major stroke from nonmajor stroke. Major stroke was defined as a National Institutes of Health Stroke Scale (NIHSS) score greater than 9 or records suggesting that the event was a disabling stroke if admitted to another facility. Nonmajor stroke included an event that did not fit these criteria. The stroke review process was initiated with a significant neurologic event, a positive transient ischemia attack or stroke questionnaire, or a two-point or greater increase in the NIHSS score.

MI was defined as a combination of an elevation of cardiac enzymes to at least twice the laboratory upper limit of normal, as well as clinical signs suggesting MI or electrocardiographic evidence of ischemia.²⁹

Stroke was adjudicated by two independent neurologists, and MI was adjudicated by two independent cardiologists blinded to treatment group assignment.

The Rankin scale, the transient ischemic attack and stroke questionnaire, and the Medical Outcomes Survey were also used to assess for disability and quality of life in long-term follow-up.

Intention-to-treat analysis

Intention-to-treat survival analysis was used along with time-to-event statistical modeling with adjustment for major baseline covariates. Differences in outcomes were assessed, and a noninferiority analysis was performed. Kaplan-Meier estimates were constructed of the proportion of patients remaining free of the composite end point at 30 days, 6 months, 1 year, and annually thereafter, and of the associated confidence intervals. The hazard ratios between groups were estimated after adjustment for important covariates.

Most patients enrolled were available for analysis

From December 2000 to July 2008, 2,522 patients were enrolled; 1,271 were assigned to stenting, and 1,251 were assigned to surgery. After randomization, 2.8% of the patients assigned to stenting withdrew consent, 5.7% underwent surgery, and 2.6% were lost to follow-up. Of those assigned to surgery, 5.1% withdrew consent, 1.0% underwent stenting, and 3.8% were lost to follow-up.

A ‘conventional-risk’ patient population

The trial sought to include a “conventional-risk” patient population to make the study more applicable to real-world practice. The mean age was 69 years in both groups. Of the 2,522 patients enrolled:

• 35% were women
• 47% had asymptomatic carotid disease
• 86% had carotid stenosis of 70% or greater
• 86% had hypertension
• 30% had diabetes mellitus
• 83% had hyperlipidemia
• 26% were current smokers
• 42% had a history of cardiovascular disease
• 21% had undergone coronary artery bypass grafting surgery.

The only statistically significant difference in measured baseline variables between the two treatment groups was a slightly higher rate of dyslipidemia in the group undergoing surgery.

The interventionalists and surgeons were highly experienced

Operators performing stenting underwent a lead-in phase of training, with close supervision and scrutiny before eligibility. Of patients undergoing stenting, 96.1% also received an embolic protection device. Antiplatelet therapy was continued in 99% of the patients.

The surgeons performing endarterectomy were experienced and had documented low complication rates. General anesthesia was used in 90% of surgical patients. Shunts were used during surgery in 57%, and patches were used in 62%. After endarterectomy, 91% of the patients received antiplatelet therapy.

CREST STUDY RESULTS: STENTING WAS AS GOOD AS SURGERY

Periprocedural outcomes

• Stroke, MI, or death: 5.2% with stenting vs 4.5% with surgery, HR 1.18, 95% CI 0.82–1.68, P = .38
• Stroke: 4.1% vs 2.3%, HR 1.79, 95% CI 1.14–2.82, P = .01
• 
  Major ipsilateral stroke: 0.9% vs 0.3%, HR 2.67, 95% CI 0.85–8.40, P = .09.
• MI: 1.1% vs 2.3%, HR 0.50, 95% CI 0.26–0.94, P = .03
• Cranial nerve palsy: 0.3% vs 4.8%, HR 0.07, 95% CI 0.02–0.18, P < .0001 (Table 2).

Outcomes at 4 years

• 

  Brott TG, et al; CREST Investigators. Stenting versus endarterectomy for treatment of carotid-artery stenosis. N Engl J Med 2010; 363:11–23. Copyright 2010, Massachusetts Medical Society. All rights reserved.

  The primary end point (periprocedural stroke, MI, or death, or ipsilateral stroke within 4 years after the procedure): 7.2% with stenting vs 6.8% with surgery, HR 1.11, 95% CI 0.81–1.51, P = .51. A Kaplan-Meier analysis showed similar findings with statistically similar outcomes (Figure 2).
• Ipsilateral stroke: 2.0% vs 2.4%, HR 0.94, 95% CI 0.50–1.76, P = .85.

The primary outcome was analyzed for interactions of baseline variables, and no effect was detected for symptomatic status or sex. There was a suggestion of an interaction with age, with older patients (over age 70) benefiting more from endarterectomy.

Quality-of-life indices showed that both major and minor strokes were likely to produce long-term physical limitations, with minor stroke associated with worse mental and physical health at 1 year. The effect of periprocedural MI on long-term physical and mental health was less certain. The increased incidence of cranial nerve palsy noted with endarterectomy has been found before and has had no effect on quality of life.

WHAT DO THE CREST FINDINGS MEAN?

CREST is the largest trial to date to compare stenting and surgery. It is an important addition to the literature, not only because of its size, but also because it focused on a real-world patient population. For this reason, its results are more applicable to patients seen in primary care clinics, ie, with peripheral vascular disease, coronary artery disease, diabetes mellitus, hypertension, and smoking.

As noted, previous studies of endarterectomy had strict inclusion and exclusion criteria, which selected against patients at high surgical risk. Therefore, the CREST findings are of greater relevance when comparing stenting and endarterectomy.

Periprocedural and long-term neurologic outcomes

CREST showed similar findings for the composite end point of periprocedural stroke, death, or MI (ie, within 30 days of the procedure) and long-term stroke, establishing similar outcomes in patients undergoing stenting and surgery.

However, an analysis of the individual components of the composite end point showed significant differences between the two treatments. The risk of ipsilateral periprocedural stroke was higher with stenting; these events were defined as nonmajor by NIHSS criteria. The risk of contralateral stroke was similar and low with each treatment.

While the increased risk of periprocedural ipsilateral stroke was not synonymous with an increased risk of major stroke, post hoc analysis showed that any stroke was associated with decreased physical and mental health at 1 year. Therefore, patients who had even a minor stroke did worse from a physical and mental standpoint, a finding that argues for the superiority of surgery in selected patients at risk of periprocedural stroke.

If periprocedural stroke is excluded, the risk of long-term ipsilateral stroke was similar for each treatment, and extremely low (2% for stenting, 2.4% for surgery). Despite this, given the importance of periprocedural minor and major stroke, better predictive models are needed to identify patients at risk of procedural neurologic events. These prediction models will allow better patient selection.

The CREST data and medical therapy

The rates of stroke in this trial were similar to those observed with current medical treatment (approximately 1% per year), especially for patients with asymptomatic disease. Such findings introduce fresh controversy in the necessity of performing either procedure for this patient subset and may lead to further studies evaluating current medical therapy vs intervention.

Periprocedural myocardial infarction

Vascular surgery has long been associated with high cardiovascular risk, especially an increased risk of periprocedural MI.³⁰ Findings from CREST provide further evidence of the risk of MI with endarterectomy in a real-world patient population. Given the evidence of a strong correlation between periprocedural cardiac enzyme elevations and adverse outcomes, the increased incidence of periprocedural MI is worrisome.³¹ As with risk assessment for periprocedural stroke, better predictive models are needed for patients at risk of cardiovascular events during endarterectomy.

Procedural complications

Carotid endarterectomy entails incisions in the neck with disruption of tissue planes, as opposed to catheter entry site wounds with stenting. The more invasive nature of endarterectomy thus carries a higher risk of wound complications. In fact, in the NASCET trial, the risk of wound complications was 9.3%.^10,19 In CREST, surgery carried a higher risk of wound complications compared with stenting (42 vs 0 cases), although stenting involved more periprocedural transfusions, presumably due to retroperitoneal bleeding in four patients.

Use of general anesthesia is also associated with adverse outcomes.^17,18 In CREST, 90% of endarterectomy procedures required general anesthesia, whereas none of the stenting procedures required this.

Cranial nerve palsy is an often overlooked but real complication after these procedures. Cranial nerve palsies can lead to vocal, swallowing, and sensory problems that can have a transient or permanent impact on quality of life. In CREST, as in EVA-3S, SAPPHIRE, and ICSS, this risk was substantially higher with surgery,^23,25,27 although the long-term consequences of these palsies were not found to affect quality of life at 1 year of follow-up.

HOW CREST FINDINGS COMPARE WITH PREVIOUS STUDIES

Patients in CREST enjoyed overall better outcomes than in previous studies. In earlier trials of surgery vs medical therapy, the rates of adverse outcomes were higher than in CREST. In NASCET, the risk of ipsilateral stroke was 9% with surgery, with 2.5% being fatal or disabling strokes.¹⁰ In the ECST, rates of major stroke or death with endarterectomy were 7.0% within 30 days of surgery and 37.0% at a mean follow-up of 6.1 years.¹²

In earlier studies of surgery vs stenting, outcomes at 30 days were also substantially worse than those in CREST. In the EVA-3S trial, the 30-day incidence of stroke or death was 3.9% after surgery and 9.6% after stenting. These findings were similar at 6 months in EVA-3S, with a 6.1% rate of adverse events after surgery and 11.7% after stenting.²⁵ In the SAPPHIRE trial, the cumulative incidence of stroke and death at 1 year was 21.4% for surgery and 13.6% for stenting.²³

Overall, the CREST results show better outcomes than in previous trials. This may be due to improvements in technical aspects of the interventions and to more aggressive drug therapy. Also, because of the high number of patients enrolled in CREST, surgeons and interventionalists were required to meet eligibility criteria, which could have contributed to the improved outcomes.³²

CREST was also unique in that stenting was done with an embolic protection device whenever possible, and this also likely had an impact on outcomes.

CREST vs ICSS

CREST and ICSS, published within a few months of each other, seem to have arrived at entirely different conclusions. As both studies are well-designed randomized controlled trials, these distinct results have yielded much controversy. However, closer scrutiny sheds light as to why the results may be different.

While ICSS focused only on patients with symptoms, CREST also included those without symptoms. The difference in patient populations is itself enough to account for the different outcomes.

Also, the interim analysis of ICSS was at 120 days, which makes periprocedural events a more dominant factor in outcomes, whereas these events likely do not last into the long term, as was the case in CREST. Analysis of the ICSS data at a later follow-up date may show results more similar to those of CREST.

The design of ICSS was also different than CREST. In ICSS, the use of an embolic protection device in stenting was not mandated, and the study lacked a lead-in phase of intensive training for those performing stenting. Furthermore, MI was adjudicated only when clinically recognized, which is different than the more rigorous method used in CREST.

Yet despite these differences, CREST and ICSS shed light on a controversial area of carotid stenosis management, and both studies boasted low rates of periprocedural complications. Clinicians should keep in mind the inclusion criteria and the technical specificities of these trials in order to explain to patients the risks and benefits of stenting and surgery, and to arrive at a decision together.

Limitations

The results of CREST should also be reviewed carefully due to a number of limitations. The study began in 2000 with symptomatic patients only, and began enrolling asymptomatic patients in 2005, so that the methodology of the study was changed midway. However, the investigators performed a subgroup analysis to distinguish between outcomes of the symptomatic and the asymptomatic groups and found no statistical interaction for the primary end point based on symptom status.

Despite careful patient selection, many of the predictors of adverse outcomes with stenting, such as lesion length, level of calcification, and lesion location, were not accounted for in the earlier days of enrollment. This may have had an impact on the incidence of stroke in patients enrolled in the early years of the trial. We await the analysis of predictors of perioperative stroke from CREST.

TAKE-HOME POINTS AND FUTURE DIRECTIONS

The CREST findings show that outcomes with stenting are similar to those with surgery in both the short term and the long term, and that the choice of management should be individualized. Each patient’s risk of MI and stroke should be considered based on a variety of factors, including the severity of coronary artery disease, the length of the carotid lesion, the level of calcification, the location of the lesion, and aortic atheroma. The treatment should be selected after also taking into account the patient’s preference and the available expertise, and only after a comprehensive discussion with the patient.

For patients with carotid artery stenosis, percutaneous intervention with stenting is as good as surgery (carotid endarterectomy). This was the major finding of the recently completed Carotid Revascularization Endarterectomy Versus Stenting Trial (CREST)¹—with some qualifications.

CREST is the latest in a series of clinical trials of treatment of carotid stenosis that have generated reams of numbers and much debate. The topic of surgery vs percutaneous intervention is a moving target, as techniques evolve and improve. We believe the CREST results are valuable and should help inform decisions about treatment in the “real world.”

In this article, we offer a critical review of CREST, with a careful evaluation of its methods, results, and conclusions.

AN EVOLVING FIELD

Despite improvements in diagnosis and management, stroke remains one of the leading causes of morbidity and death in the United States, with an annual incidence of 780,000 cases and 270,000 deaths.^2,3

Randomized trials of stenting have had mixed results, leading the Centers for Medicare and Medicaid Services (CMS) to adopt strict reimbursement policies. Currently, CMS reimburses for stenting only in symptomatic cases with at least 50% carotid artery stenosis. It also reimburses for stenting in asymptomatic cases in patients at high risk with 80% or greater stenosis, but only if the patients are enrolled in ongoing clinical trials or registries.

CREST compared stenting with endarterectomy and provided important insights into each approach.¹

BEFORE CREST

Endarterectomy is superior to medical therapy for symptomatic stenosis

First described in 1953, carotid endarterectomy became the most widely used invasive treatment for significant carotid stenosis.⁹ Several studies have described patient subsets that benefit from this procedure.

NASCET (the North American Symptomatic Carotid Endarterectomy Trial)¹⁰ assigned 2,226 patients with symptomatic stenosis (transient ischemic attack or stroke within the past 180 days) to medical management or endarterectomy.

Surgery was associated with a 65% lower rate of ipsilateral cerebral events in patients with 70% or greater stenosis.¹⁰ Surgery was also found to be superior in patients with moderate disease (50% to 69% stenosis), but the difference only approached statistical significance. In patients with stenosis of less than 50%, the outcomes were similar with endarterectomy and medical management.¹¹

ECST (the European Carotid Surgery Trial)¹² included a similar population of 3,024 patients. Those with high-grade disease (stenosis ≥ 80%) had significantly better outcomes with endarterectomy, but in those with stenosis less than 70%, surgery was no better than drug therapy.

Comment. NASCET and ECST taught us that endarterectomy is clearly superior to medical therapy in patients with severe symptomatic carotid disease. However, both trials excluded patients at high surgical risk, eg, those with severe coronary artery disease, kidney disease, or heart failure. Additionally, medical management was not aggressive by today’s standards in terms of control of blood pressure and hyperlipidemia, and this could have skewed the results in favor of carotid endarterectomy.

The case for carotid endarterectomy for asymptomatic stenosis

Endarterectomy has also been compared with drug therapy for asymp tomatic carotid artery stenosis in several trials.^13–15

ACAS (the Asymptomatic Carotid Atherosclerosis Study)¹⁵ assigned 1,662 patients who had no symptoms and had at least 60% carotid artery stenosis to endarterectomy or to medical management, and found a relative risk reduction of 53% in favor of surgery.¹⁵

The Veterans Affairs Cooperative Study Group¹⁴ corroborated these results in 444 patients with asymptomatic stenosis of greater than 50%. Endarterectomy was associated with a 61% lower risk of transient ischemic attack, transient monocular blindness, or stroke compared with medical therapy. However, there was no statistically significant difference in rates of stroke or death at 30 days.¹⁴

ACST (the Asymptomatic Carotid Surgery Trial),¹³ the largest study to compare carotid endarterectomy with drug therapy for asymptomatic stenosis, randomized 3,120 patients to surgery or drug therapy. The net 5-year risk of stroke was 6.4% with endarterectomy vs 11.8% with drug therapy (P < .0001). The rate of fatal stroke was also lower with endarterectomy: 2.1% vs 4.2% (P = .006).¹³

Comment. The results of these and other studies of endarterectomy vs medical therapy may not be applicable to current practice, since medical therapy has evolved and the risks with current drug therapy are likely much lower than seen in these trials, some of which began 2 decades ago. Another problem with interpreting these trials is that they excluded surgically “high-risk” patients, which limits the generalizability of the findings to this particular patient population.

The American Heart Association and the American Stroke Association have, on the basis of these trials, recommended carotid endarterectomy in patients with^7,8,16:

• Ipsilateral, symptomatic carotid artery stenosis of 70% to 99% (class I, level of evidence A)
• Symptomatic stenosis of 50% to 69%, depending on patient-specific factors such as age, sex, and comorbidities
• High-grade asymptomatic carotid stenosis, if the patients are carefully selected and the surgery is performed by surgeons with procedural morbidity and mortality rates of less than 3% (class I, level of evidence A).

While carotid endarterectomy is proven to be more efficacious than medical management in certain patient subsets, studies favoring surgery over medical therapy have been criticized because they excluded patients with significant comorbidities. In addition, surgery has been associated with significant cardiovascular events, wound complications, and cranial nerve damage, and it requires general anesthesia in most cases.^12,17–19 These and other factors spurred the development of less invasive, percutaneous approaches for patients with substantial comorbidities.

So far, several trials have investigated carotid angioplasty with or without stents and with or without devices to capture distal emboli. This interest set the stage for CREST.^20,21

Initial attempts at angioplasty without distal protection were not very successful. A meta-analysis of nonrandomized trials that included 714 patients from the initial 13 studies of angioplasty (with or without stenting) and 6,970 patients from 20 studies of carotid endarterectomy found angioplasty to be possibly associated with higher rates of stroke within 30 days of the procedure.²⁰

With improvements in technology, routine use of embolic protection devices, more experience, and better selection of patients, the outcome of carotid stenting has improved. In fact, a meta-analysis comparing stenting without an embolic protection device (26 trials with 2,357 patients) vs stenting with an embolic protection device (11 trials with 839 patients) showed that embolic protection led to significantly better outcomes with fewer strokes—outcomes arguably similar to those of carotid endarterectomy.²¹

SAPPHIRE (the Stenting and Angioplasty With Protection in Patients at High Risk for Endarterectomy trial)²² was the only completed US trial until CREST that compared carotid artery stenting with distal protection against surgery. It included 334 high-risk patients with either symptomatic stenosis of 50% or greater or asymptomatic stenosis of 80% or greater.

The results suggested that the outcomes with stenting with embolic protection were in fact similar to those of endarterectomy, with possibly fewer complications.²³ The benefit persisted up to 2 years.²²

The US Food and Drug Administration (FDA), on the basis of these data, approved the use of stenting with distal protection for high-risk patients, and the CMS reimburses for symptomatic stenosis of 50% or greater and for asymptomatic stenosis of 80% or greater as long as the patient is enrolled in a registry.

SPACE (the Stent-Protected Angioplasty Versus Carotid Endarterectomy in Symptomatic Patients trial),²⁴ conducted in Germany, included 1,214 patients with symptomatic stenosis of at least 50%. Results were similar in terms of the combined primary end point of stroke or death at 30 days. However, the results were not similar enough to prove that stenting is not inferior to surgery, according to preset study criteria.

EVA-3S (the Endarterectomy Versus Stenting in Patients With Symptomatic Severe Carotid Stenosis trial),²⁵ in France, evaluated 527 patients with symptomatic carotid disease (stenosis ≥ 60%), but was terminated early due to significantly higher rates of death or stroke at 30 days in the stenting group.

Comment. SPACE and EVA-3S have been widely criticized for not mandating the use of an embolic protection device (used in 27% of cases in SPACE and in 91.9% of cases in EVA-3S). Questions were also raised about the experience level of the operators who performed the carotid stenting: up to 39% of the primary operators involved in stent placement were trainees.²⁶ Also, myocardial infarction (MI), an important complication of carotid endarterectomy, was not included in the primary end point.

ICSS (the International Carotid Stenting Study)²⁷ compared stenting with endarterectomy in 1,713 patients with symptomatic carotid stenosis of greater than 50%. The primary end point was the rate of fatal or disabling stroke at 3 years.

An interim safety analysis at 120 days of follow-up showed the primary end point had occurred in 4.0% of stenting cases vs 3.2% of endarterectomy cases, a difference that was not statistically significant (hazard ratio [HR] 1.28, 95% confidence interval [CI] 0.77–2.11). However, the risk of any stroke was higher with stenting, with a rate of 7.7% vs 4.1% in the surgical group—a statistically significant difference (HR 1.92, 95% CI 1.27–2.89).

In a substudy of ICSS,²⁸ the investigators corroborated these findings, using magnetic resonance imaging to evaluate for new ischemic brain lesions periprocedurally. They found more new ischemic brain lesions in patients who underwent stenting than in patients who underwent surgery—a statistically significant finding.

Comment. ICSS had limitations: eg, it included only patients with symptoms, and the training for the stenting procedure was not standardized. Furthermore, the use of embolic protection devices was not mandated in stenting procedures.

Because of the controversial and incongruous findings of the above trials, there has been much anticipation for further large, appropriately conducted, randomized controlled trials such as CREST.

CREST STUDY DESIGN

CREST was a prospective, multicenter randomized controlled trial with blinded end point adjudication. Assignment to stenting or surgery occurred in a one-to-one fashion, and patients were stratified by medical center and symptomatic status.

Conducted at 108 sites in the United States and nine sites in Canada, CREST was supported by a grant from the National Institutes of Health and by the manufacturer of the catheter and stent delivery and embolic protection systems. The manufacturer’s representative held a nonvoting position on the executive committee and reviewed the manuscript of the results before submission.

CREST included patients with or without symptoms

CREST was initially designed to compare carotid artery stenting vs carotid endarterectomy in patients with symptoms, but enrollment was later extended to patients without symptoms.

Patients with symptoms were included if they had stenosis of at least 50% on angiography, at least 70% on ultrasonography, or at least 70% on computed tomographic angiography or magnetic resonance angiography if stenosis on ultrasonography was 50% to 69%. Carotid artery stenosis was considered symptomatic if the patient had a transient ischemic attack, amaurosis fugax, or minor disabling stroke in the hemisphere supplied by the target vessel within 180 days of randomization.

Patients without symptoms were eligible if they had at least 60% stenosis on angiography, at least 70% stenosis on ultrasonography, or at least 80% stenosis on computed tomographic angiography or magnetic resonance angiography if the stenosis was 50% to 69% on ultrasonography.

Other eligibility criteria included favorable anatomy and clinical stability for both stenting and surgical procedures.

Exclusion criteria were evolving stroke, history of major stroke, chronic or paroxysmal atrial fibrillation on anticoagulation therapy, MI within the previous 30 days, and unstable angina.

Patients undergoing stenting received aspirin and clopidogrel (Plavix) before and up to 30 days after the procedure. Continuation of antiplatelet therapy was recommended beyond 1 month.

Patients undergoing endarterectomy received aspirin before surgery and continued to receive aspirin for at least 1 year.

Alternatives to aspirin in both groups were ticlopidine (Ticlid), clopidogrel, or aspirin with extended-release dipyridamole (Aggrenox).

End points: Stroke, MI, death

The primary end point was a composite of periprocedural clinical stroke (any type), MI, or death, and of ipsilateral stroke up to 4 years after the procedure. Secondary analyses were also planned for evaluation of treatment modification by age, symptom status, and sex.

Stroke was defined as any acute neurologic ischemic event lasting at least 24 hours with focal signs and symptoms.

Two separate definitions were applied to distinguish major stroke from nonmajor stroke. Major stroke was defined as a National Institutes of Health Stroke Scale (NIHSS) score greater than 9 or records suggesting that the event was a disabling stroke if admitted to another facility. Nonmajor stroke included an event that did not fit these criteria. The stroke review process was initiated with a significant neurologic event, a positive transient ischemia attack or stroke questionnaire, or a two-point or greater increase in the NIHSS score.

MI was defined as a combination of an elevation of cardiac enzymes to at least twice the laboratory upper limit of normal, as well as clinical signs suggesting MI or electrocardiographic evidence of ischemia.²⁹

Stroke was adjudicated by two independent neurologists, and MI was adjudicated by two independent cardiologists blinded to treatment group assignment.

The Rankin scale, the transient ischemic attack and stroke questionnaire, and the Medical Outcomes Survey were also used to assess for disability and quality of life in long-term follow-up.

Intention-to-treat analysis

Intention-to-treat survival analysis was used along with time-to-event statistical modeling with adjustment for major baseline covariates. Differences in outcomes were assessed, and a noninferiority analysis was performed. Kaplan-Meier estimates were constructed of the proportion of patients remaining free of the composite end point at 30 days, 6 months, 1 year, and annually thereafter, and of the associated confidence intervals. The hazard ratios between groups were estimated after adjustment for important covariates.

Most patients enrolled were available for analysis

From December 2000 to July 2008, 2,522 patients were enrolled; 1,271 were assigned to stenting, and 1,251 were assigned to surgery. After randomization, 2.8% of the patients assigned to stenting withdrew consent, 5.7% underwent surgery, and 2.6% were lost to follow-up. Of those assigned to surgery, 5.1% withdrew consent, 1.0% underwent stenting, and 3.8% were lost to follow-up.

A ‘conventional-risk’ patient population

The trial sought to include a “conventional-risk” patient population to make the study more applicable to real-world practice. The mean age was 69 years in both groups. Of the 2,522 patients enrolled:

• 35% were women
• 47% had asymptomatic carotid disease
• 86% had carotid stenosis of 70% or greater
• 86% had hypertension
• 30% had diabetes mellitus
• 83% had hyperlipidemia
• 26% were current smokers
• 42% had a history of cardiovascular disease
• 21% had undergone coronary artery bypass grafting surgery.

The only statistically significant difference in measured baseline variables between the two treatment groups was a slightly higher rate of dyslipidemia in the group undergoing surgery.

The interventionalists and surgeons were highly experienced

Operators performing stenting underwent a lead-in phase of training, with close supervision and scrutiny before eligibility. Of patients undergoing stenting, 96.1% also received an embolic protection device. Antiplatelet therapy was continued in 99% of the patients.

The surgeons performing endarterectomy were experienced and had documented low complication rates. General anesthesia was used in 90% of surgical patients. Shunts were used during surgery in 57%, and patches were used in 62%. After endarterectomy, 91% of the patients received antiplatelet therapy.

CREST STUDY RESULTS: STENTING WAS AS GOOD AS SURGERY

Periprocedural outcomes

• Stroke, MI, or death: 5.2% with stenting vs 4.5% with surgery, HR 1.18, 95% CI 0.82–1.68, P = .38
• Stroke: 4.1% vs 2.3%, HR 1.79, 95% CI 1.14–2.82, P = .01
• 
  Major ipsilateral stroke: 0.9% vs 0.3%, HR 2.67, 95% CI 0.85–8.40, P = .09.
• MI: 1.1% vs 2.3%, HR 0.50, 95% CI 0.26–0.94, P = .03
• Cranial nerve palsy: 0.3% vs 4.8%, HR 0.07, 95% CI 0.02–0.18, P < .0001 (Table 2).

Outcomes at 4 years

• 

  Brott TG, et al; CREST Investigators. Stenting versus endarterectomy for treatment of carotid-artery stenosis. N Engl J Med 2010; 363:11–23. Copyright 2010, Massachusetts Medical Society. All rights reserved.

  The primary end point (periprocedural stroke, MI, or death, or ipsilateral stroke within 4 years after the procedure): 7.2% with stenting vs 6.8% with surgery, HR 1.11, 95% CI 0.81–1.51, P = .51. A Kaplan-Meier analysis showed similar findings with statistically similar outcomes (Figure 2).
• Ipsilateral stroke: 2.0% vs 2.4%, HR 0.94, 95% CI 0.50–1.76, P = .85.

The primary outcome was analyzed for interactions of baseline variables, and no effect was detected for symptomatic status or sex. There was a suggestion of an interaction with age, with older patients (over age 70) benefiting more from endarterectomy.

Quality-of-life indices showed that both major and minor strokes were likely to produce long-term physical limitations, with minor stroke associated with worse mental and physical health at 1 year. The effect of periprocedural MI on long-term physical and mental health was less certain. The increased incidence of cranial nerve palsy noted with endarterectomy has been found before and has had no effect on quality of life.

WHAT DO THE CREST FINDINGS MEAN?

CREST is the largest trial to date to compare stenting and surgery. It is an important addition to the literature, not only because of its size, but also because it focused on a real-world patient population. For this reason, its results are more applicable to patients seen in primary care clinics, ie, with peripheral vascular disease, coronary artery disease, diabetes mellitus, hypertension, and smoking.

As noted, previous studies of endarterectomy had strict inclusion and exclusion criteria, which selected against patients at high surgical risk. Therefore, the CREST findings are of greater relevance when comparing stenting and endarterectomy.

Periprocedural and long-term neurologic outcomes

CREST showed similar findings for the composite end point of periprocedural stroke, death, or MI (ie, within 30 days of the procedure) and long-term stroke, establishing similar outcomes in patients undergoing stenting and surgery.

However, an analysis of the individual components of the composite end point showed significant differences between the two treatments. The risk of ipsilateral periprocedural stroke was higher with stenting; these events were defined as nonmajor by NIHSS criteria. The risk of contralateral stroke was similar and low with each treatment.

While the increased risk of periprocedural ipsilateral stroke was not synonymous with an increased risk of major stroke, post hoc analysis showed that any stroke was associated with decreased physical and mental health at 1 year. Therefore, patients who had even a minor stroke did worse from a physical and mental standpoint, a finding that argues for the superiority of surgery in selected patients at risk of periprocedural stroke.

If periprocedural stroke is excluded, the risk of long-term ipsilateral stroke was similar for each treatment, and extremely low (2% for stenting, 2.4% for surgery). Despite this, given the importance of periprocedural minor and major stroke, better predictive models are needed to identify patients at risk of procedural neurologic events. These prediction models will allow better patient selection.

The CREST data and medical therapy

The rates of stroke in this trial were similar to those observed with current medical treatment (approximately 1% per year), especially for patients with asymptomatic disease. Such findings introduce fresh controversy in the necessity of performing either procedure for this patient subset and may lead to further studies evaluating current medical therapy vs intervention.

Periprocedural myocardial infarction

Vascular surgery has long been associated with high cardiovascular risk, especially an increased risk of periprocedural MI.³⁰ Findings from CREST provide further evidence of the risk of MI with endarterectomy in a real-world patient population. Given the evidence of a strong correlation between periprocedural cardiac enzyme elevations and adverse outcomes, the increased incidence of periprocedural MI is worrisome.³¹ As with risk assessment for periprocedural stroke, better predictive models are needed for patients at risk of cardiovascular events during endarterectomy.

Procedural complications

Carotid endarterectomy entails incisions in the neck with disruption of tissue planes, as opposed to catheter entry site wounds with stenting. The more invasive nature of endarterectomy thus carries a higher risk of wound complications. In fact, in the NASCET trial, the risk of wound complications was 9.3%.^10,19 In CREST, surgery carried a higher risk of wound complications compared with stenting (42 vs 0 cases), although stenting involved more periprocedural transfusions, presumably due to retroperitoneal bleeding in four patients.

Use of general anesthesia is also associated with adverse outcomes.^17,18 In CREST, 90% of endarterectomy procedures required general anesthesia, whereas none of the stenting procedures required this.

Cranial nerve palsy is an often overlooked but real complication after these procedures. Cranial nerve palsies can lead to vocal, swallowing, and sensory problems that can have a transient or permanent impact on quality of life. In CREST, as in EVA-3S, SAPPHIRE, and ICSS, this risk was substantially higher with surgery,^23,25,27 although the long-term consequences of these palsies were not found to affect quality of life at 1 year of follow-up.

HOW CREST FINDINGS COMPARE WITH PREVIOUS STUDIES

Patients in CREST enjoyed overall better outcomes than in previous studies. In earlier trials of surgery vs medical therapy, the rates of adverse outcomes were higher than in CREST. In NASCET, the risk of ipsilateral stroke was 9% with surgery, with 2.5% being fatal or disabling strokes.¹⁰ In the ECST, rates of major stroke or death with endarterectomy were 7.0% within 30 days of surgery and 37.0% at a mean follow-up of 6.1 years.¹²

In earlier studies of surgery vs stenting, outcomes at 30 days were also substantially worse than those in CREST. In the EVA-3S trial, the 30-day incidence of stroke or death was 3.9% after surgery and 9.6% after stenting. These findings were similar at 6 months in EVA-3S, with a 6.1% rate of adverse events after surgery and 11.7% after stenting.²⁵ In the SAPPHIRE trial, the cumulative incidence of stroke and death at 1 year was 21.4% for surgery and 13.6% for stenting.²³

Overall, the CREST results show better outcomes than in previous trials. This may be due to improvements in technical aspects of the interventions and to more aggressive drug therapy. Also, because of the high number of patients enrolled in CREST, surgeons and interventionalists were required to meet eligibility criteria, which could have contributed to the improved outcomes.³²

CREST was also unique in that stenting was done with an embolic protection device whenever possible, and this also likely had an impact on outcomes.

CREST vs ICSS

CREST and ICSS, published within a few months of each other, seem to have arrived at entirely different conclusions. As both studies are well-designed randomized controlled trials, these distinct results have yielded much controversy. However, closer scrutiny sheds light as to why the results may be different.

While ICSS focused only on patients with symptoms, CREST also included those without symptoms. The difference in patient populations is itself enough to account for the different outcomes.

Also, the interim analysis of ICSS was at 120 days, which makes periprocedural events a more dominant factor in outcomes, whereas these events likely do not last into the long term, as was the case in CREST. Analysis of the ICSS data at a later follow-up date may show results more similar to those of CREST.

The design of ICSS was also different than CREST. In ICSS, the use of an embolic protection device in stenting was not mandated, and the study lacked a lead-in phase of intensive training for those performing stenting. Furthermore, MI was adjudicated only when clinically recognized, which is different than the more rigorous method used in CREST.

Yet despite these differences, CREST and ICSS shed light on a controversial area of carotid stenosis management, and both studies boasted low rates of periprocedural complications. Clinicians should keep in mind the inclusion criteria and the technical specificities of these trials in order to explain to patients the risks and benefits of stenting and surgery, and to arrive at a decision together.

Limitations

The results of CREST should also be reviewed carefully due to a number of limitations. The study began in 2000 with symptomatic patients only, and began enrolling asymptomatic patients in 2005, so that the methodology of the study was changed midway. However, the investigators performed a subgroup analysis to distinguish between outcomes of the symptomatic and the asymptomatic groups and found no statistical interaction for the primary end point based on symptom status.

Despite careful patient selection, many of the predictors of adverse outcomes with stenting, such as lesion length, level of calcification, and lesion location, were not accounted for in the earlier days of enrollment. This may have had an impact on the incidence of stroke in patients enrolled in the early years of the trial. We await the analysis of predictors of perioperative stroke from CREST.

TAKE-HOME POINTS AND FUTURE DIRECTIONS

The CREST findings show that outcomes with stenting are similar to those with surgery in both the short term and the long term, and that the choice of management should be individualized. Each patient’s risk of MI and stroke should be considered based on a variety of factors, including the severity of coronary artery disease, the length of the carotid lesion, the level of calcification, the location of the lesion, and aortic atheroma. The treatment should be selected after also taking into account the patient’s preference and the available expertise, and only after a comprehensive discussion with the patient.

For patients with carotid artery stenosis, percutaneous intervention with stenting is as good as surgery (carotid endarterectomy). This was the major finding of the recently completed Carotid Revascularization Endarterectomy Versus Stenting Trial (CREST)¹—with some qualifications.

CREST is the latest in a series of clinical trials of treatment of carotid stenosis that have generated reams of numbers and much debate. The topic of surgery vs percutaneous intervention is a moving target, as techniques evolve and improve. We believe the CREST results are valuable and should help inform decisions about treatment in the “real world.”

In this article, we offer a critical review of CREST, with a careful evaluation of its methods, results, and conclusions.

AN EVOLVING FIELD

Despite improvements in diagnosis and management, stroke remains one of the leading causes of morbidity and death in the United States, with an annual incidence of 780,000 cases and 270,000 deaths.^2,3

Randomized trials of stenting have had mixed results, leading the Centers for Medicare and Medicaid Services (CMS) to adopt strict reimbursement policies. Currently, CMS reimburses for stenting only in symptomatic cases with at least 50% carotid artery stenosis. It also reimburses for stenting in asymptomatic cases in patients at high risk with 80% or greater stenosis, but only if the patients are enrolled in ongoing clinical trials or registries.

CREST compared stenting with endarterectomy and provided important insights into each approach.¹

BEFORE CREST

Endarterectomy is superior to medical therapy for symptomatic stenosis

First described in 1953, carotid endarterectomy became the most widely used invasive treatment for significant carotid stenosis.⁹ Several studies have described patient subsets that benefit from this procedure.

NASCET (the North American Symptomatic Carotid Endarterectomy Trial)¹⁰ assigned 2,226 patients with symptomatic stenosis (transient ischemic attack or stroke within the past 180 days) to medical management or endarterectomy.

Surgery was associated with a 65% lower rate of ipsilateral cerebral events in patients with 70% or greater stenosis.¹⁰ Surgery was also found to be superior in patients with moderate disease (50% to 69% stenosis), but the difference only approached statistical significance. In patients with stenosis of less than 50%, the outcomes were similar with endarterectomy and medical management.¹¹

ECST (the European Carotid Surgery Trial)¹² included a similar population of 3,024 patients. Those with high-grade disease (stenosis ≥ 80%) had significantly better outcomes with endarterectomy, but in those with stenosis less than 70%, surgery was no better than drug therapy.

Comment. NASCET and ECST taught us that endarterectomy is clearly superior to medical therapy in patients with severe symptomatic carotid disease. However, both trials excluded patients at high surgical risk, eg, those with severe coronary artery disease, kidney disease, or heart failure. Additionally, medical management was not aggressive by today’s standards in terms of control of blood pressure and hyperlipidemia, and this could have skewed the results in favor of carotid endarterectomy.

The case for carotid endarterectomy for asymptomatic stenosis

Endarterectomy has also been compared with drug therapy for asymp tomatic carotid artery stenosis in several trials.^13–15

ACAS (the Asymptomatic Carotid Atherosclerosis Study)¹⁵ assigned 1,662 patients who had no symptoms and had at least 60% carotid artery stenosis to endarterectomy or to medical management, and found a relative risk reduction of 53% in favor of surgery.¹⁵

The Veterans Affairs Cooperative Study Group¹⁴ corroborated these results in 444 patients with asymptomatic stenosis of greater than 50%. Endarterectomy was associated with a 61% lower risk of transient ischemic attack, transient monocular blindness, or stroke compared with medical therapy. However, there was no statistically significant difference in rates of stroke or death at 30 days.¹⁴

ACST (the Asymptomatic Carotid Surgery Trial),¹³ the largest study to compare carotid endarterectomy with drug therapy for asymptomatic stenosis, randomized 3,120 patients to surgery or drug therapy. The net 5-year risk of stroke was 6.4% with endarterectomy vs 11.8% with drug therapy (P < .0001). The rate of fatal stroke was also lower with endarterectomy: 2.1% vs 4.2% (P = .006).¹³

Comment. The results of these and other studies of endarterectomy vs medical therapy may not be applicable to current practice, since medical therapy has evolved and the risks with current drug therapy are likely much lower than seen in these trials, some of which began 2 decades ago. Another problem with interpreting these trials is that they excluded surgically “high-risk” patients, which limits the generalizability of the findings to this particular patient population.

The American Heart Association and the American Stroke Association have, on the basis of these trials, recommended carotid endarterectomy in patients with^7,8,16:

• Ipsilateral, symptomatic carotid artery stenosis of 70% to 99% (class I, level of evidence A)
• Symptomatic stenosis of 50% to 69%, depending on patient-specific factors such as age, sex, and comorbidities
• High-grade asymptomatic carotid stenosis, if the patients are carefully selected and the surgery is performed by surgeons with procedural morbidity and mortality rates of less than 3% (class I, level of evidence A).

While carotid endarterectomy is proven to be more efficacious than medical management in certain patient subsets, studies favoring surgery over medical therapy have been criticized because they excluded patients with significant comorbidities. In addition, surgery has been associated with significant cardiovascular events, wound complications, and cranial nerve damage, and it requires general anesthesia in most cases.^12,17–19 These and other factors spurred the development of less invasive, percutaneous approaches for patients with substantial comorbidities.

So far, several trials have investigated carotid angioplasty with or without stents and with or without devices to capture distal emboli. This interest set the stage for CREST.^20,21

Initial attempts at angioplasty without distal protection were not very successful. A meta-analysis of nonrandomized trials that included 714 patients from the initial 13 studies of angioplasty (with or without stenting) and 6,970 patients from 20 studies of carotid endarterectomy found angioplasty to be possibly associated with higher rates of stroke within 30 days of the procedure.²⁰

With improvements in technology, routine use of embolic protection devices, more experience, and better selection of patients, the outcome of carotid stenting has improved. In fact, a meta-analysis comparing stenting without an embolic protection device (26 trials with 2,357 patients) vs stenting with an embolic protection device (11 trials with 839 patients) showed that embolic protection led to significantly better outcomes with fewer strokes—outcomes arguably similar to those of carotid endarterectomy.²¹

SAPPHIRE (the Stenting and Angioplasty With Protection in Patients at High Risk for Endarterectomy trial)²² was the only completed US trial until CREST that compared carotid artery stenting with distal protection against surgery. It included 334 high-risk patients with either symptomatic stenosis of 50% or greater or asymptomatic stenosis of 80% or greater.

The results suggested that the outcomes with stenting with embolic protection were in fact similar to those of endarterectomy, with possibly fewer complications.²³ The benefit persisted up to 2 years.²²

The US Food and Drug Administration (FDA), on the basis of these data, approved the use of stenting with distal protection for high-risk patients, and the CMS reimburses for symptomatic stenosis of 50% or greater and for asymptomatic stenosis of 80% or greater as long as the patient is enrolled in a registry.

SPACE (the Stent-Protected Angioplasty Versus Carotid Endarterectomy in Symptomatic Patients trial),²⁴ conducted in Germany, included 1,214 patients with symptomatic stenosis of at least 50%. Results were similar in terms of the combined primary end point of stroke or death at 30 days. However, the results were not similar enough to prove that stenting is not inferior to surgery, according to preset study criteria.

EVA-3S (the Endarterectomy Versus Stenting in Patients With Symptomatic Severe Carotid Stenosis trial),²⁵ in France, evaluated 527 patients with symptomatic carotid disease (stenosis ≥ 60%), but was terminated early due to significantly higher rates of death or stroke at 30 days in the stenting group.

Comment. SPACE and EVA-3S have been widely criticized for not mandating the use of an embolic protection device (used in 27% of cases in SPACE and in 91.9% of cases in EVA-3S). Questions were also raised about the experience level of the operators who performed the carotid stenting: up to 39% of the primary operators involved in stent placement were trainees.²⁶ Also, myocardial infarction (MI), an important complication of carotid endarterectomy, was not included in the primary end point.

ICSS (the International Carotid Stenting Study)²⁷ compared stenting with endarterectomy in 1,713 patients with symptomatic carotid stenosis of greater than 50%. The primary end point was the rate of fatal or disabling stroke at 3 years.

An interim safety analysis at 120 days of follow-up showed the primary end point had occurred in 4.0% of stenting cases vs 3.2% of endarterectomy cases, a difference that was not statistically significant (hazard ratio [HR] 1.28, 95% confidence interval [CI] 0.77–2.11). However, the risk of any stroke was higher with stenting, with a rate of 7.7% vs 4.1% in the surgical group—a statistically significant difference (HR 1.92, 95% CI 1.27–2.89).

In a substudy of ICSS,²⁸ the investigators corroborated these findings, using magnetic resonance imaging to evaluate for new ischemic brain lesions periprocedurally. They found more new ischemic brain lesions in patients who underwent stenting than in patients who underwent surgery—a statistically significant finding.

Comment. ICSS had limitations: eg, it included only patients with symptoms, and the training for the stenting procedure was not standardized. Furthermore, the use of embolic protection devices was not mandated in stenting procedures.

Because of the controversial and incongruous findings of the above trials, there has been much anticipation for further large, appropriately conducted, randomized controlled trials such as CREST.

CREST STUDY DESIGN

CREST was a prospective, multicenter randomized controlled trial with blinded end point adjudication. Assignment to stenting or surgery occurred in a one-to-one fashion, and patients were stratified by medical center and symptomatic status.

Conducted at 108 sites in the United States and nine sites in Canada, CREST was supported by a grant from the National Institutes of Health and by the manufacturer of the catheter and stent delivery and embolic protection systems. The manufacturer’s representative held a nonvoting position on the executive committee and reviewed the manuscript of the results before submission.

CREST included patients with or without symptoms

CREST was initially designed to compare carotid artery stenting vs carotid endarterectomy in patients with symptoms, but enrollment was later extended to patients without symptoms.

Patients with symptoms were included if they had stenosis of at least 50% on angiography, at least 70% on ultrasonography, or at least 70% on computed tomographic angiography or magnetic resonance angiography if stenosis on ultrasonography was 50% to 69%. Carotid artery stenosis was considered symptomatic if the patient had a transient ischemic attack, amaurosis fugax, or minor disabling stroke in the hemisphere supplied by the target vessel within 180 days of randomization.

Patients without symptoms were eligible if they had at least 60% stenosis on angiography, at least 70% stenosis on ultrasonography, or at least 80% stenosis on computed tomographic angiography or magnetic resonance angiography if the stenosis was 50% to 69% on ultrasonography.

Other eligibility criteria included favorable anatomy and clinical stability for both stenting and surgical procedures.

Exclusion criteria were evolving stroke, history of major stroke, chronic or paroxysmal atrial fibrillation on anticoagulation therapy, MI within the previous 30 days, and unstable angina.

Patients undergoing stenting received aspirin and clopidogrel (Plavix) before and up to 30 days after the procedure. Continuation of antiplatelet therapy was recommended beyond 1 month.

Patients undergoing endarterectomy received aspirin before surgery and continued to receive aspirin for at least 1 year.

Alternatives to aspirin in both groups were ticlopidine (Ticlid), clopidogrel, or aspirin with extended-release dipyridamole (Aggrenox).

End points: Stroke, MI, death

The primary end point was a composite of periprocedural clinical stroke (any type), MI, or death, and of ipsilateral stroke up to 4 years after the procedure. Secondary analyses were also planned for evaluation of treatment modification by age, symptom status, and sex.

Stroke was defined as any acute neurologic ischemic event lasting at least 24 hours with focal signs and symptoms.

Two separate definitions were applied to distinguish major stroke from nonmajor stroke. Major stroke was defined as a National Institutes of Health Stroke Scale (NIHSS) score greater than 9 or records suggesting that the event was a disabling stroke if admitted to another facility. Nonmajor stroke included an event that did not fit these criteria. The stroke review process was initiated with a significant neurologic event, a positive transient ischemia attack or stroke questionnaire, or a two-point or greater increase in the NIHSS score.

MI was defined as a combination of an elevation of cardiac enzymes to at least twice the laboratory upper limit of normal, as well as clinical signs suggesting MI or electrocardiographic evidence of ischemia.²⁹

Stroke was adjudicated by two independent neurologists, and MI was adjudicated by two independent cardiologists blinded to treatment group assignment.

The Rankin scale, the transient ischemic attack and stroke questionnaire, and the Medical Outcomes Survey were also used to assess for disability and quality of life in long-term follow-up.

Intention-to-treat analysis

Intention-to-treat survival analysis was used along with time-to-event statistical modeling with adjustment for major baseline covariates. Differences in outcomes were assessed, and a noninferiority analysis was performed. Kaplan-Meier estimates were constructed of the proportion of patients remaining free of the composite end point at 30 days, 6 months, 1 year, and annually thereafter, and of the associated confidence intervals. The hazard ratios between groups were estimated after adjustment for important covariates.

Most patients enrolled were available for analysis

From December 2000 to July 2008, 2,522 patients were enrolled; 1,271 were assigned to stenting, and 1,251 were assigned to surgery. After randomization, 2.8% of the patients assigned to stenting withdrew consent, 5.7% underwent surgery, and 2.6% were lost to follow-up. Of those assigned to surgery, 5.1% withdrew consent, 1.0% underwent stenting, and 3.8% were lost to follow-up.

A ‘conventional-risk’ patient population

The trial sought to include a “conventional-risk” patient population to make the study more applicable to real-world practice. The mean age was 69 years in both groups. Of the 2,522 patients enrolled:

• 35% were women
• 47% had asymptomatic carotid disease
• 86% had carotid stenosis of 70% or greater
• 86% had hypertension
• 30% had diabetes mellitus
• 83% had hyperlipidemia
• 26% were current smokers
• 42% had a history of cardiovascular disease
• 21% had undergone coronary artery bypass grafting surgery.

The only statistically significant difference in measured baseline variables between the two treatment groups was a slightly higher rate of dyslipidemia in the group undergoing surgery.

The interventionalists and surgeons were highly experienced

Operators performing stenting underwent a lead-in phase of training, with close supervision and scrutiny before eligibility. Of patients undergoing stenting, 96.1% also received an embolic protection device. Antiplatelet therapy was continued in 99% of the patients.

The surgeons performing endarterectomy were experienced and had documented low complication rates. General anesthesia was used in 90% of surgical patients. Shunts were used during surgery in 57%, and patches were used in 62%. After endarterectomy, 91% of the patients received antiplatelet therapy.

CREST STUDY RESULTS: STENTING WAS AS GOOD AS SURGERY

Periprocedural outcomes

• Stroke, MI, or death: 5.2% with stenting vs 4.5% with surgery, HR 1.18, 95% CI 0.82–1.68, P = .38
• Stroke: 4.1% vs 2.3%, HR 1.79, 95% CI 1.14–2.82, P = .01
• 
  Major ipsilateral stroke: 0.9% vs 0.3%, HR 2.67, 95% CI 0.85–8.40, P = .09.
• MI: 1.1% vs 2.3%, HR 0.50, 95% CI 0.26–0.94, P = .03
• Cranial nerve palsy: 0.3% vs 4.8%, HR 0.07, 95% CI 0.02–0.18, P < .0001 (Table 2).

Outcomes at 4 years

• 

  Brott TG, et al; CREST Investigators. Stenting versus endarterectomy for treatment of carotid-artery stenosis. N Engl J Med 2010; 363:11–23. Copyright 2010, Massachusetts Medical Society. All rights reserved.

  The primary end point (periprocedural stroke, MI, or death, or ipsilateral stroke within 4 years after the procedure): 7.2% with stenting vs 6.8% with surgery, HR 1.11, 95% CI 0.81–1.51, P = .51. A Kaplan-Meier analysis showed similar findings with statistically similar outcomes (Figure 2).
• Ipsilateral stroke: 2.0% vs 2.4%, HR 0.94, 95% CI 0.50–1.76, P = .85.

The primary outcome was analyzed for interactions of baseline variables, and no effect was detected for symptomatic status or sex. There was a suggestion of an interaction with age, with older patients (over age 70) benefiting more from endarterectomy.

Quality-of-life indices showed that both major and minor strokes were likely to produce long-term physical limitations, with minor stroke associated with worse mental and physical health at 1 year. The effect of periprocedural MI on long-term physical and mental health was less certain. The increased incidence of cranial nerve palsy noted with endarterectomy has been found before and has had no effect on quality of life.

WHAT DO THE CREST FINDINGS MEAN?

CREST is the largest trial to date to compare stenting and surgery. It is an important addition to the literature, not only because of its size, but also because it focused on a real-world patient population. For this reason, its results are more applicable to patients seen in primary care clinics, ie, with peripheral vascular disease, coronary artery disease, diabetes mellitus, hypertension, and smoking.

As noted, previous studies of endarterectomy had strict inclusion and exclusion criteria, which selected against patients at high surgical risk. Therefore, the CREST findings are of greater relevance when comparing stenting and endarterectomy.

Periprocedural and long-term neurologic outcomes

CREST showed similar findings for the composite end point of periprocedural stroke, death, or MI (ie, within 30 days of the procedure) and long-term stroke, establishing similar outcomes in patients undergoing stenting and surgery.

However, an analysis of the individual components of the composite end point showed significant differences between the two treatments. The risk of ipsilateral periprocedural stroke was higher with stenting; these events were defined as nonmajor by NIHSS criteria. The risk of contralateral stroke was similar and low with each treatment.

While the increased risk of periprocedural ipsilateral stroke was not synonymous with an increased risk of major stroke, post hoc analysis showed that any stroke was associated with decreased physical and mental health at 1 year. Therefore, patients who had even a minor stroke did worse from a physical and mental standpoint, a finding that argues for the superiority of surgery in selected patients at risk of periprocedural stroke.

If periprocedural stroke is excluded, the risk of long-term ipsilateral stroke was similar for each treatment, and extremely low (2% for stenting, 2.4% for surgery). Despite this, given the importance of periprocedural minor and major stroke, better predictive models are needed to identify patients at risk of procedural neurologic events. These prediction models will allow better patient selection.

The CREST data and medical therapy

The rates of stroke in this trial were similar to those observed with current medical treatment (approximately 1% per year), especially for patients with asymptomatic disease. Such findings introduce fresh controversy in the necessity of performing either procedure for this patient subset and may lead to further studies evaluating current medical therapy vs intervention.

Periprocedural myocardial infarction

Vascular surgery has long been associated with high cardiovascular risk, especially an increased risk of periprocedural MI.³⁰ Findings from CREST provide further evidence of the risk of MI with endarterectomy in a real-world patient population. Given the evidence of a strong correlation between periprocedural cardiac enzyme elevations and adverse outcomes, the increased incidence of periprocedural MI is worrisome.³¹ As with risk assessment for periprocedural stroke, better predictive models are needed for patients at risk of cardiovascular events during endarterectomy.

Procedural complications

Carotid endarterectomy entails incisions in the neck with disruption of tissue planes, as opposed to catheter entry site wounds with stenting. The more invasive nature of endarterectomy thus carries a higher risk of wound complications. In fact, in the NASCET trial, the risk of wound complications was 9.3%.^10,19 In CREST, surgery carried a higher risk of wound complications compared with stenting (42 vs 0 cases), although stenting involved more periprocedural transfusions, presumably due to retroperitoneal bleeding in four patients.

Use of general anesthesia is also associated with adverse outcomes.^17,18 In CREST, 90% of endarterectomy procedures required general anesthesia, whereas none of the stenting procedures required this.

Cranial nerve palsy is an often overlooked but real complication after these procedures. Cranial nerve palsies can lead to vocal, swallowing, and sensory problems that can have a transient or permanent impact on quality of life. In CREST, as in EVA-3S, SAPPHIRE, and ICSS, this risk was substantially higher with surgery,^23,25,27 although the long-term consequences of these palsies were not found to affect quality of life at 1 year of follow-up.

HOW CREST FINDINGS COMPARE WITH PREVIOUS STUDIES

Patients in CREST enjoyed overall better outcomes than in previous studies. In earlier trials of surgery vs medical therapy, the rates of adverse outcomes were higher than in CREST. In NASCET, the risk of ipsilateral stroke was 9% with surgery, with 2.5% being fatal or disabling strokes.¹⁰ In the ECST, rates of major stroke or death with endarterectomy were 7.0% within 30 days of surgery and 37.0% at a mean follow-up of 6.1 years.¹²

In earlier studies of surgery vs stenting, outcomes at 30 days were also substantially worse than those in CREST. In the EVA-3S trial, the 30-day incidence of stroke or death was 3.9% after surgery and 9.6% after stenting. These findings were similar at 6 months in EVA-3S, with a 6.1% rate of adverse events after surgery and 11.7% after stenting.²⁵ In the SAPPHIRE trial, the cumulative incidence of stroke and death at 1 year was 21.4% for surgery and 13.6% for stenting.²³

Overall, the CREST results show better outcomes than in previous trials. This may be due to improvements in technical aspects of the interventions and to more aggressive drug therapy. Also, because of the high number of patients enrolled in CREST, surgeons and interventionalists were required to meet eligibility criteria, which could have contributed to the improved outcomes.³²

CREST was also unique in that stenting was done with an embolic protection device whenever possible, and this also likely had an impact on outcomes.

CREST vs ICSS

CREST and ICSS, published within a few months of each other, seem to have arrived at entirely different conclusions. As both studies are well-designed randomized controlled trials, these distinct results have yielded much controversy. However, closer scrutiny sheds light as to why the results may be different.

While ICSS focused only on patients with symptoms, CREST also included those without symptoms. The difference in patient populations is itself enough to account for the different outcomes.

Also, the interim analysis of ICSS was at 120 days, which makes periprocedural events a more dominant factor in outcomes, whereas these events likely do not last into the long term, as was the case in CREST. Analysis of the ICSS data at a later follow-up date may show results more similar to those of CREST.

The design of ICSS was also different than CREST. In ICSS, the use of an embolic protection device in stenting was not mandated, and the study lacked a lead-in phase of intensive training for those performing stenting. Furthermore, MI was adjudicated only when clinically recognized, which is different than the more rigorous method used in CREST.

Yet despite these differences, CREST and ICSS shed light on a controversial area of carotid stenosis management, and both studies boasted low rates of periprocedural complications. Clinicians should keep in mind the inclusion criteria and the technical specificities of these trials in order to explain to patients the risks and benefits of stenting and surgery, and to arrive at a decision together.

Limitations

The results of CREST should also be reviewed carefully due to a number of limitations. The study began in 2000 with symptomatic patients only, and began enrolling asymptomatic patients in 2005, so that the methodology of the study was changed midway. However, the investigators performed a subgroup analysis to distinguish between outcomes of the symptomatic and the asymptomatic groups and found no statistical interaction for the primary end point based on symptom status.

Despite careful patient selection, many of the predictors of adverse outcomes with stenting, such as lesion length, level of calcification, and lesion location, were not accounted for in the earlier days of enrollment. This may have had an impact on the incidence of stroke in patients enrolled in the early years of the trial. We await the analysis of predictors of perioperative stroke from CREST.

TAKE-HOME POINTS AND FUTURE DIRECTIONS

The CREST findings show that outcomes with stenting are similar to those with surgery in both the short term and the long term, and that the choice of management should be individualized. Each patient’s risk of MI and stroke should be considered based on a variety of factors, including the severity of coronary artery disease, the length of the carotid lesion, the level of calcification, the location of the lesion, and aortic atheroma. The treatment should be selected after also taking into account the patient’s preference and the available expertise, and only after a comprehensive discussion with the patient.

Along with massive pulmonary embolism, the most feared thromboembolic event is the clot that migrates to the brain, resulting in life-altering stroke. We assess this risk in a semiquantitative manner in patients with atrial fibrillation using the CHADS₂ score, hoping to maximize the benefits of anticoagulation while reducing the risks. We recognize that patients at the greatest risk of stroke in this setting are those with a history of a prior stroke. Also, patients bedridden with a recent cerebrovascular accident (CVA) seem to be hypercoagulable, potentially adding risk to recent injury. Thus, we try to start anticoagulation as soon as feasible after the diagnosis of a possible thrombotic event.

But the decision to start or resume anticoagulation is especially agonizing in a patient who has suffered an intracerebral hemorrhage. In this issue of the Journal, Drs. Joshua Goldstein and Steven Greenberg and Dr. Franklin Michota provide a thoughtful discussion of the issues we need to consider in these patients.

While not contributing to the prevention of additional CVAs or other arterial thrombotic events, a modality often underused in the prevention of thrombotic disease is the application (not just the ordering) of compressive leg stockings to bedridden hospitalized patients who cannot, for any reason, be provided pharmacologic anticoagulation therapy. I just completed a stint of hospital consultation, and I was pleased to see the widespread integration of prophylactic anticoagulation therapy, but somewhat dismayed by the number of compressive stockings I watched pumping with vigor, but to no one’s benefit, as they were draped over a bed rail.

As we struggle with complex clinical decisions, we need to also be attentive to the simple and the seemingly mundane: using the foam dispenser at the door, offering the verbal greeting and patient touch at the bedside, and rewrapping the pneumatic stockings that have somehow migrated between mattress and footboard.

Along with massive pulmonary embolism, the most feared thromboembolic event is the clot that migrates to the brain, resulting in life-altering stroke. We assess this risk in a semiquantitative manner in patients with atrial fibrillation using the CHADS₂ score, hoping to maximize the benefits of anticoagulation while reducing the risks. We recognize that patients at the greatest risk of stroke in this setting are those with a history of a prior stroke. Also, patients bedridden with a recent cerebrovascular accident (CVA) seem to be hypercoagulable, potentially adding risk to recent injury. Thus, we try to start anticoagulation as soon as feasible after the diagnosis of a possible thrombotic event.

But the decision to start or resume anticoagulation is especially agonizing in a patient who has suffered an intracerebral hemorrhage. In this issue of the Journal, Drs. Joshua Goldstein and Steven Greenberg and Dr. Franklin Michota provide a thoughtful discussion of the issues we need to consider in these patients.

While not contributing to the prevention of additional CVAs or other arterial thrombotic events, a modality often underused in the prevention of thrombotic disease is the application (not just the ordering) of compressive leg stockings to bedridden hospitalized patients who cannot, for any reason, be provided pharmacologic anticoagulation therapy. I just completed a stint of hospital consultation, and I was pleased to see the widespread integration of prophylactic anticoagulation therapy, but somewhat dismayed by the number of compressive stockings I watched pumping with vigor, but to no one’s benefit, as they were draped over a bed rail.

As we struggle with complex clinical decisions, we need to also be attentive to the simple and the seemingly mundane: using the foam dispenser at the door, offering the verbal greeting and patient touch at the bedside, and rewrapping the pneumatic stockings that have somehow migrated between mattress and footboard.

Along with massive pulmonary embolism, the most feared thromboembolic event is the clot that migrates to the brain, resulting in life-altering stroke. We assess this risk in a semiquantitative manner in patients with atrial fibrillation using the CHADS₂ score, hoping to maximize the benefits of anticoagulation while reducing the risks. We recognize that patients at the greatest risk of stroke in this setting are those with a history of a prior stroke. Also, patients bedridden with a recent cerebrovascular accident (CVA) seem to be hypercoagulable, potentially adding risk to recent injury. Thus, we try to start anticoagulation as soon as feasible after the diagnosis of a possible thrombotic event.

But the decision to start or resume anticoagulation is especially agonizing in a patient who has suffered an intracerebral hemorrhage. In this issue of the Journal, Drs. Joshua Goldstein and Steven Greenberg and Dr. Franklin Michota provide a thoughtful discussion of the issues we need to consider in these patients.

While not contributing to the prevention of additional CVAs or other arterial thrombotic events, a modality often underused in the prevention of thrombotic disease is the application (not just the ordering) of compressive leg stockings to bedridden hospitalized patients who cannot, for any reason, be provided pharmacologic anticoagulation therapy. I just completed a stint of hospital consultation, and I was pleased to see the widespread integration of prophylactic anticoagulation therapy, but somewhat dismayed by the number of compressive stockings I watched pumping with vigor, but to no one’s benefit, as they were draped over a bed rail.

As we struggle with complex clinical decisions, we need to also be attentive to the simple and the seemingly mundane: using the foam dispenser at the door, offering the verbal greeting and patient touch at the bedside, and rewrapping the pneumatic stockings that have somehow migrated between mattress and footboard.

The reality of blood pressure measurement is that human beings do not do it very well. The time has come to delegate this job to machines that can do it better.

Several automatic devices are available. Used in physicians’ offices and in patients’ homes, they can eliminate some types of observer error as well as the “white-coat effect,” ie, the tendency of some patients to have higher blood pressure when medical personnel are present than in their natural environment. Since a difference of only a few millimeters of mercury can affect the physician’s decision to start or to modify treatment, measurements that more accurately reflect a person’s average blood pressure are to be desired.

In the following pages, we review the problems that plague blood pressure measurement by human observers, and we describe the advantages of automatic devices.

SHORTCOMINGS OF OFFICE BLOOD PRESSURE MEASUREMENT

For decades, large surveys have provided invaluable information on the prevalence of hypertension, its relationship to cardiovascular disease, and the benefits of treating it.^1–3 Unfortunately, the percentage of hypertensive patients whose blood pressure is under control has remained low despite our increased knowledge about hypertension’s diagnosis and therapy.⁴

In these surveys, blood pressure was measured by auscultation by human observers using mercury or aneroid sphygmomanometers, and most physicians still use this method in clinical practice. But in spite of multiple guidelines for accurate measurement of blood pressure in the office, the overall accuracy and reproducibility of office blood pressure measurements remain poor.^5–7

The accuracy of blood pressure measurement with aneroid and mercury manometers is affected by a number of observer errors and patient factors.^8,9

Observer errors

Failure to prepare the patient. National guidelines⁵ state that before having their blood pressure taken, patients should be allowed to sit quietly for at least 5 minutes, which often does not happen. Another error is that clinicians rarely discourage patients from smoking cigarettes or drinking coffee in the 30 minutes prior to measurement.

Equipment and layout problems. Equipment should be properly calibrated and validated. ⁵ However, even if the sphygmomanometer is periodically calibrated, too often it is mounted on the wall adjacent to the examination table in the examination room, making it difficult to provide a comfortable seat with back and arm support during the reading. The measurement should be done with the patient sitting in a chair (not on an examination table), with feet on the floor and the arm supported at the level of the heart. If the forearm is not supported in the horizontal position and with the cuff at heart level, the blood pressure and heart rate tend to be higher.¹⁰ Further, the diastolic blood pressure and heart rate may be misleadingly low with the patient supine rather than seated,^11,12 so readings should be taken with the patient sitting.

Miscuffing, ie, the use of a blood pressure cuff that is too large or, more often, too small for the patient’s arm, is a common source of error. The cuff bladder should encircle at least 80% of the arm.⁵ However, some offices do not have a large blood pressure cuff for overweight patients or a pediatric cuff for children or adults with arms of small circumference. It is recommended that a large blood pressure cuff be used routinely in adults, since a smaller cuff gives falsely high readings in people with large upper arms (circumference > 29 cm).^13,14

Digit preference. Many physicians round off the blood pressure to the nearest 5 or 10 mm Hg. This problem may go along with:

Deflating the cuff too rapidly.

Talking to the patient while taking the blood pressure can contribute to higher readings.⁹

Not taking enough readings. Ideally, at the initial visit, blood pressure should be measured in both arms with the patient seated, and another reading should be taken with the patient standing. The arm with the higher pressure should be used for subsequent readings. Physicians should not make any treatment decisions based on blood pressure during an initial clinic visit, and at least two readings should be taken even on subsequent visits. However, owing to time constraints in busy clinical practices, treatment decisions are often based on single readings or on multiple readings on a single visit.

Discrepancies between observers. The blood pressure readings obtained by the nurse or medical assistant may differ significantly from those obtained by the physician. These differences can be large enough to affect treatment decisions,^15,16 and they can be partially corrected by adequate training of all medical personnel who take blood pressure, doctors as well as nurses.

Given that time is tight in busy clinical practices and a trained blood pressure nurse or technician is usually not available, we will probably not see any significant improvement in the accuracy of blood pressure measurement using older technology and current physician practices.

The white-coat effect

Most patients have a higher level of anxiety, and therefore higher blood pressure, in the physician’s office or clinic than in their normal environment (as revealed by ambulatory monitoring or home blood pressure measurements), a phenomenon commonly called the white-coat effect.

Several factors can increase this effect, such as observer-patient interaction during the measurement. The effect tends to be greatest in the initial measurement, but can persist through multiple readings by the doctor or nurse during the same visit.

Whether the white-coat effect is due purely to patient anxiety about an office visit or to a conditioned response has been a point of interest in clinical studies. Regardless, it may result in the misdiagnosis of hypertension or in overestimation of the severity of hypertension and may lead to overly aggressive therapy. Antihypertensive treatment may be unnecessary in the absence of concurrent cardiovascular risk factors.¹⁷

“White-coat hypertension” or “isolated office hypertension” is the condition in which a patient who is not on antihypertensive drug therapy has persistently elevated blood pressure in the clinic or office (> 140/90 mm Hg) but normal daytime ambulatory blood pressure (< 135/85 mm Hg).¹⁸ Since patients may have an elevated reading when seen for a first office visit, at least several visits are required to establish the diagnosis. Multiple studies have suggested that white-coat hypertension may account for 20% to 25% of the hypertensive population, particularly in older patients, mainly women.^19,20

Both white-coat hypertension and the white-coat effect can be avoided by using an automatic and programmable device that can take multiple readings after the clinician leaves the examination room (more about this below).²¹

MEASURING BLOOD PRESSURE OUTSIDE THE OFFICE

Recent studies have reported that the information obtained by 24-hour ambulatory blood pressure monitoring and by self-measurement of blood pressure in the home more accurately reflects the patient’s risk of future cardiovascular events than do conventional blood pressure measurements taken in the physician’s office. ^22–24 Current national guidelines recognize this pattern and the frequent measurement inaccuracies observed in clinical practice, and they are recommending including out-of-office measurements in the diagnosis of hypertension. ^25,26

Ambulatory monitoring provides the most accurate measurement of out-of-office blood pressure. With ambulatory monitoring, the normal mean daytime pressure is considered to be lower than 135/85 mm Hg, in contrast to the 140/90 mm Hg cutoff used in the physician’s office with standard aneroid or mercury devices.

Self-monitoring of blood pressure at home has now become widely available with single-measurement oscillometric devices. (Oscillometric means that these devices measure the blood pressure by sensing the oscillations in pressure in the cuff induced by the pulsation of the brachial artery, as opposed to auscultating the Korotkoff sounds.) Blood pressures lower than 135/85 mm Hg outside the clinician’s office are considered normal with these devices.

However, despite its proven value, ambulatory monitoring is neither widely available nor cost-effective for the long-term management of hypertension. Furthermore, few physicians recommend that patients take their blood pressure at home, although the information obtained can be of significant value in the patient’s long-term management.

AUTOMATED MEASUREMENT IN THE OFFICE

In recent years, several automated oscillometric sphygmomanometers have been developed for measuring blood pressure in the office, and more are on the way. These devices can be programmed to take multiple readings without a clinician observer in the examination room, thus reducing the white-coat response.

Omron (Kyoto, Japan) makes several devices, including the HEM-907 and the HEM-705, that have been used in the clinical setting. ^21,27–29 They can be programmed to take two or three readings at intervals of 1 to 2 minutes, with up to 5 minutes before the first reading. Unfortunately, data were not recorded with the patient alone in the room in many studies of the Omron devices, even though the devices meet national and international standards for accuracy.

The Microlife Watch BP Office (Microlife, Widnau, Switzerland) is currently undergoing development.³⁰

The BpTRU (BpTRU Medical Devices, Coquitlam, BC, Canada) has enjoyed greater clinical acceptance, since it can take up to five blood pressure readings at intervals of 1 to 5 minutes, and calculates the mean of all five readings, taken with the patient resting comfortably in a quiet room without a clinician present.

The accuracy and durability of the device has been well established. Since the BpTRU self-calibrates between every blood pressure measurement, periodic calibration has not been required. The device can be placed on a table, mounted on the wall, or mounted on a cart if used in several locations in the office.

At Cleveland Clinic, several departments are using the BpTRU on a daily basis. Soon, we will be able to transfer data directly from the BpTRU to our electronic medical record system.

Studies of the BpTRU device

To date, most of the studies of automated office blood pressure measurement have used the BpTRU with the recording interval set at 1 to 2 minutes.

Myers³¹ used the BpTRU device in 50 hypertensive patients. The physician took the patient’s blood pressure with a mercury sphygmomanometer while the BpTRU device made the first reading, and then he left the room. The next five readings were taken at 2-minute intervals with the patient alone in the room. The mean initial reading by the machine was 162/85 mm Hg; the reading by the physician was 163/86 mm Hg. The third automatic reading was the lowest (averaging 140/84 mm Hg), and the mean of the five automated readings was 142/80 mm Hg, which was significantly lower than the initial reading obtained by the physician (P < .001).

In another study, Myers et al³² compared the measurements obtained by 24-hour ambulatory monitoring and by the BpTRU device (the mean of five readings obtained at 1-minute or 2-minute intervals) in 309 hypertensive patients. The mean blood pressure with the Bp-TRU was 132/75 mm Hg, which correlated well with the mean awake ambulatory blood pressure (134/77 mm Hg; r = 0.62 for the systolic pressure and 0.72 for the diastolic pressure).

We recently reviewed the records of 278 patients seen in our preventive medicine clinic (D.G. Vidt, MD, unpublished data, November 2009). The group included patients with and without established hypertension, and among the hypertensive group, both treated and untreated individuals. We had initially set the device to take readings at 3-minute intervals following the initial nurse-initiated reading. But in view of the recent data on the Bp-TRU using shorter intervals, we also obtained readings in 51 patients with the device set to record at 2-minute intervals, and then in 72 additional patients at 1-minute intervals. In all three groups, blood pressure had stabilized by the third reading after the clinician had left the room. These observations support those reported by Myers et al.^31,32 Of particular importance is the observation that the white-coat effect dissipates within 2 to 3 minutes after the clinician leaves the room.³³

The shorter measurement intervals can add up in a busy office practice, in which the time relegated to taking blood pressure is often limited.

In fact, waiting 5 minutes between measurements may allow the patient to become too relaxed and the blood pressure to drop too low vis-a-vis the gold standard, ambulatory monitoring. Culleton and colleagues³⁴ compared the blood pressure in 107 hypertensive patients as measured four ways: by the referring physician, by a nurse who was trained to adhere to the protocol of the Canadian Hypertension Education Program, by 24-hour ambulatory monitoring, and by the BpTRU (the mean of five readings obtained at 5-minute intervals). The mean measured values were:

• 150/90 mm Hg by the referring physician
• 139/86 mm Hg by the nurse
• 142/85 mm Hg by ambulatory monitoring
• 132/82 mm Hg by the BpTRU device.

Although the BpTRU reduced the white-coat effect and white-coat hypertension, it underestimated the blood pressure, leading to misclassification of hypertension. Using 140/90 mm Hg as the cutoff for whether the patient was hypertensive and using ambulatory monitoring as the gold standard, the BpTRU misclassified more than half of the patients, agreeing with the classification of hypertensive or not hypertensive by ambulatory monitoring in only 48%. The authors recommended that the BpTRU not be set at 5-minute measurement intervals.³⁴

WHAT ROLE FOR AUTOMATED READINGS IN THE OFFICE?

Although automatic devices, by enabling the physician to leave the room, can eliminate the white-coat effect and white-coat hypertension, physicians must continue to take care to avoid the other potential errors of office blood pressure measurement addressed earlier in this review, for example, by positioning the patient correctly and using a cuff that is large enough. These issues can take on more importance as the clinician leaves the patient alone for brief periods during measurements.

In view of its perennial inaccuracies, some experts have suggested that we abandon routine office measurement of blood pressure.^35,36 In its place, ambulatory monitoring would be used for diagnosis and for periodic follow-up. In addition, patients would regularly take their pressure at home with approved, single-measurement oscillometric devices. Unfortunately, in our health care system, periodic ambulatory monitoring for hypertension management would impose a significant financial burden on patients at this time.³⁷

Of particular importance is the observation that the mean of five readings with the BpTRU device, obtained at 1- or 2-minute intervals, closely approximates the mean awake blood pressure obtained in the same patient with an ambulatory monitor.^32,38 The ability to obtain readings that correlate exceptionally well with mean daytime ambulatory pressure suggests that this device could well reduce the need for ambulatory monitoring, with its associated cost. The ability to negate the white-coat effect with the use of the BpTRU in the office setting also has particular importance, not only for patient office readings, but for the diagnosis and subsequent treatment of hypertension in individual patients.

Most clinical decisions about the treatment of hypertension are still made on the basis of office determinations of blood pressure. Most office practices still rely on the aneroid manometer or, decreasingly, mercury sphygmomanometers. As noted earlier, although auscultatory blood pressure measurement appears to be simple, it is fraught with a host of observer- or patient-induced errors that not only lead to inaccurate diagnoses, but may also result in the mismanagement of hypertension. Even single-measurement oscillometric devices, now used in a minority of clinical practices, are associated with many of the same measurement issues that lead to overestimation of blood pressure.

We believe the time has come for broader use of oscillometric devices in the outpatient setting. While many available oscillometric devices for use in the home could also be used in the physician’s office, they carry the similar disadvantage of providing only a single measurement. The major disadvantage of all single-measurement devices is the continued presence of the clinician during the reading and the associated white-coat effect observed in most patients.

It is highly likely that the next Joint National Committee Report on Hypertension will further emphasize the role of automated blood pressure devices in the outpatient setting.
 

Acknowledgment: The authors wish to acknowledge the contributions of Deborah McCoy, RN, and Maria Eckhouse, RN.

The reality of blood pressure measurement is that human beings do not do it very well. The time has come to delegate this job to machines that can do it better.

Several automatic devices are available. Used in physicians’ offices and in patients’ homes, they can eliminate some types of observer error as well as the “white-coat effect,” ie, the tendency of some patients to have higher blood pressure when medical personnel are present than in their natural environment. Since a difference of only a few millimeters of mercury can affect the physician’s decision to start or to modify treatment, measurements that more accurately reflect a person’s average blood pressure are to be desired.

In the following pages, we review the problems that plague blood pressure measurement by human observers, and we describe the advantages of automatic devices.

SHORTCOMINGS OF OFFICE BLOOD PRESSURE MEASUREMENT

For decades, large surveys have provided invaluable information on the prevalence of hypertension, its relationship to cardiovascular disease, and the benefits of treating it.^1–3 Unfortunately, the percentage of hypertensive patients whose blood pressure is under control has remained low despite our increased knowledge about hypertension’s diagnosis and therapy.⁴

In these surveys, blood pressure was measured by auscultation by human observers using mercury or aneroid sphygmomanometers, and most physicians still use this method in clinical practice. But in spite of multiple guidelines for accurate measurement of blood pressure in the office, the overall accuracy and reproducibility of office blood pressure measurements remain poor.^5–7

The accuracy of blood pressure measurement with aneroid and mercury manometers is affected by a number of observer errors and patient factors.^8,9

Observer errors

Failure to prepare the patient. National guidelines⁵ state that before having their blood pressure taken, patients should be allowed to sit quietly for at least 5 minutes, which often does not happen. Another error is that clinicians rarely discourage patients from smoking cigarettes or drinking coffee in the 30 minutes prior to measurement.

Equipment and layout problems. Equipment should be properly calibrated and validated. ⁵ However, even if the sphygmomanometer is periodically calibrated, too often it is mounted on the wall adjacent to the examination table in the examination room, making it difficult to provide a comfortable seat with back and arm support during the reading. The measurement should be done with the patient sitting in a chair (not on an examination table), with feet on the floor and the arm supported at the level of the heart. If the forearm is not supported in the horizontal position and with the cuff at heart level, the blood pressure and heart rate tend to be higher.¹⁰ Further, the diastolic blood pressure and heart rate may be misleadingly low with the patient supine rather than seated,^11,12 so readings should be taken with the patient sitting.

Miscuffing, ie, the use of a blood pressure cuff that is too large or, more often, too small for the patient’s arm, is a common source of error. The cuff bladder should encircle at least 80% of the arm.⁵ However, some offices do not have a large blood pressure cuff for overweight patients or a pediatric cuff for children or adults with arms of small circumference. It is recommended that a large blood pressure cuff be used routinely in adults, since a smaller cuff gives falsely high readings in people with large upper arms (circumference > 29 cm).^13,14

Digit preference. Many physicians round off the blood pressure to the nearest 5 or 10 mm Hg. This problem may go along with:

Deflating the cuff too rapidly.

Talking to the patient while taking the blood pressure can contribute to higher readings.⁹

Not taking enough readings. Ideally, at the initial visit, blood pressure should be measured in both arms with the patient seated, and another reading should be taken with the patient standing. The arm with the higher pressure should be used for subsequent readings. Physicians should not make any treatment decisions based on blood pressure during an initial clinic visit, and at least two readings should be taken even on subsequent visits. However, owing to time constraints in busy clinical practices, treatment decisions are often based on single readings or on multiple readings on a single visit.

Discrepancies between observers. The blood pressure readings obtained by the nurse or medical assistant may differ significantly from those obtained by the physician. These differences can be large enough to affect treatment decisions,^15,16 and they can be partially corrected by adequate training of all medical personnel who take blood pressure, doctors as well as nurses.

Given that time is tight in busy clinical practices and a trained blood pressure nurse or technician is usually not available, we will probably not see any significant improvement in the accuracy of blood pressure measurement using older technology and current physician practices.

The white-coat effect

Most patients have a higher level of anxiety, and therefore higher blood pressure, in the physician’s office or clinic than in their normal environment (as revealed by ambulatory monitoring or home blood pressure measurements), a phenomenon commonly called the white-coat effect.

Several factors can increase this effect, such as observer-patient interaction during the measurement. The effect tends to be greatest in the initial measurement, but can persist through multiple readings by the doctor or nurse during the same visit.

Whether the white-coat effect is due purely to patient anxiety about an office visit or to a conditioned response has been a point of interest in clinical studies. Regardless, it may result in the misdiagnosis of hypertension or in overestimation of the severity of hypertension and may lead to overly aggressive therapy. Antihypertensive treatment may be unnecessary in the absence of concurrent cardiovascular risk factors.¹⁷

“White-coat hypertension” or “isolated office hypertension” is the condition in which a patient who is not on antihypertensive drug therapy has persistently elevated blood pressure in the clinic or office (> 140/90 mm Hg) but normal daytime ambulatory blood pressure (< 135/85 mm Hg).¹⁸ Since patients may have an elevated reading when seen for a first office visit, at least several visits are required to establish the diagnosis. Multiple studies have suggested that white-coat hypertension may account for 20% to 25% of the hypertensive population, particularly in older patients, mainly women.^19,20

Both white-coat hypertension and the white-coat effect can be avoided by using an automatic and programmable device that can take multiple readings after the clinician leaves the examination room (more about this below).²¹

MEASURING BLOOD PRESSURE OUTSIDE THE OFFICE

Recent studies have reported that the information obtained by 24-hour ambulatory blood pressure monitoring and by self-measurement of blood pressure in the home more accurately reflects the patient’s risk of future cardiovascular events than do conventional blood pressure measurements taken in the physician’s office. ^22–24 Current national guidelines recognize this pattern and the frequent measurement inaccuracies observed in clinical practice, and they are recommending including out-of-office measurements in the diagnosis of hypertension. ^25,26

Ambulatory monitoring provides the most accurate measurement of out-of-office blood pressure. With ambulatory monitoring, the normal mean daytime pressure is considered to be lower than 135/85 mm Hg, in contrast to the 140/90 mm Hg cutoff used in the physician’s office with standard aneroid or mercury devices.

Self-monitoring of blood pressure at home has now become widely available with single-measurement oscillometric devices. (Oscillometric means that these devices measure the blood pressure by sensing the oscillations in pressure in the cuff induced by the pulsation of the brachial artery, as opposed to auscultating the Korotkoff sounds.) Blood pressures lower than 135/85 mm Hg outside the clinician’s office are considered normal with these devices.

However, despite its proven value, ambulatory monitoring is neither widely available nor cost-effective for the long-term management of hypertension. Furthermore, few physicians recommend that patients take their blood pressure at home, although the information obtained can be of significant value in the patient’s long-term management.

AUTOMATED MEASUREMENT IN THE OFFICE

In recent years, several automated oscillometric sphygmomanometers have been developed for measuring blood pressure in the office, and more are on the way. These devices can be programmed to take multiple readings without a clinician observer in the examination room, thus reducing the white-coat response.

Omron (Kyoto, Japan) makes several devices, including the HEM-907 and the HEM-705, that have been used in the clinical setting. ^21,27–29 They can be programmed to take two or three readings at intervals of 1 to 2 minutes, with up to 5 minutes before the first reading. Unfortunately, data were not recorded with the patient alone in the room in many studies of the Omron devices, even though the devices meet national and international standards for accuracy.

The Microlife Watch BP Office (Microlife, Widnau, Switzerland) is currently undergoing development.³⁰

The BpTRU (BpTRU Medical Devices, Coquitlam, BC, Canada) has enjoyed greater clinical acceptance, since it can take up to five blood pressure readings at intervals of 1 to 5 minutes, and calculates the mean of all five readings, taken with the patient resting comfortably in a quiet room without a clinician present.

The accuracy and durability of the device has been well established. Since the BpTRU self-calibrates between every blood pressure measurement, periodic calibration has not been required. The device can be placed on a table, mounted on the wall, or mounted on a cart if used in several locations in the office.

At Cleveland Clinic, several departments are using the BpTRU on a daily basis. Soon, we will be able to transfer data directly from the BpTRU to our electronic medical record system.

Studies of the BpTRU device

To date, most of the studies of automated office blood pressure measurement have used the BpTRU with the recording interval set at 1 to 2 minutes.

Myers³¹ used the BpTRU device in 50 hypertensive patients. The physician took the patient’s blood pressure with a mercury sphygmomanometer while the BpTRU device made the first reading, and then he left the room. The next five readings were taken at 2-minute intervals with the patient alone in the room. The mean initial reading by the machine was 162/85 mm Hg; the reading by the physician was 163/86 mm Hg. The third automatic reading was the lowest (averaging 140/84 mm Hg), and the mean of the five automated readings was 142/80 mm Hg, which was significantly lower than the initial reading obtained by the physician (P < .001).

In another study, Myers et al³² compared the measurements obtained by 24-hour ambulatory monitoring and by the BpTRU device (the mean of five readings obtained at 1-minute or 2-minute intervals) in 309 hypertensive patients. The mean blood pressure with the Bp-TRU was 132/75 mm Hg, which correlated well with the mean awake ambulatory blood pressure (134/77 mm Hg; r = 0.62 for the systolic pressure and 0.72 for the diastolic pressure).

We recently reviewed the records of 278 patients seen in our preventive medicine clinic (D.G. Vidt, MD, unpublished data, November 2009). The group included patients with and without established hypertension, and among the hypertensive group, both treated and untreated individuals. We had initially set the device to take readings at 3-minute intervals following the initial nurse-initiated reading. But in view of the recent data on the Bp-TRU using shorter intervals, we also obtained readings in 51 patients with the device set to record at 2-minute intervals, and then in 72 additional patients at 1-minute intervals. In all three groups, blood pressure had stabilized by the third reading after the clinician had left the room. These observations support those reported by Myers et al.^31,32 Of particular importance is the observation that the white-coat effect dissipates within 2 to 3 minutes after the clinician leaves the room.³³

The shorter measurement intervals can add up in a busy office practice, in which the time relegated to taking blood pressure is often limited.

In fact, waiting 5 minutes between measurements may allow the patient to become too relaxed and the blood pressure to drop too low vis-a-vis the gold standard, ambulatory monitoring. Culleton and colleagues³⁴ compared the blood pressure in 107 hypertensive patients as measured four ways: by the referring physician, by a nurse who was trained to adhere to the protocol of the Canadian Hypertension Education Program, by 24-hour ambulatory monitoring, and by the BpTRU (the mean of five readings obtained at 5-minute intervals). The mean measured values were:

• 150/90 mm Hg by the referring physician
• 139/86 mm Hg by the nurse
• 142/85 mm Hg by ambulatory monitoring
• 132/82 mm Hg by the BpTRU device.

Although the BpTRU reduced the white-coat effect and white-coat hypertension, it underestimated the blood pressure, leading to misclassification of hypertension. Using 140/90 mm Hg as the cutoff for whether the patient was hypertensive and using ambulatory monitoring as the gold standard, the BpTRU misclassified more than half of the patients, agreeing with the classification of hypertensive or not hypertensive by ambulatory monitoring in only 48%. The authors recommended that the BpTRU not be set at 5-minute measurement intervals.³⁴

WHAT ROLE FOR AUTOMATED READINGS IN THE OFFICE?

Although automatic devices, by enabling the physician to leave the room, can eliminate the white-coat effect and white-coat hypertension, physicians must continue to take care to avoid the other potential errors of office blood pressure measurement addressed earlier in this review, for example, by positioning the patient correctly and using a cuff that is large enough. These issues can take on more importance as the clinician leaves the patient alone for brief periods during measurements.

In view of its perennial inaccuracies, some experts have suggested that we abandon routine office measurement of blood pressure.^35,36 In its place, ambulatory monitoring would be used for diagnosis and for periodic follow-up. In addition, patients would regularly take their pressure at home with approved, single-measurement oscillometric devices. Unfortunately, in our health care system, periodic ambulatory monitoring for hypertension management would impose a significant financial burden on patients at this time.³⁷

Of particular importance is the observation that the mean of five readings with the BpTRU device, obtained at 1- or 2-minute intervals, closely approximates the mean awake blood pressure obtained in the same patient with an ambulatory monitor.^32,38 The ability to obtain readings that correlate exceptionally well with mean daytime ambulatory pressure suggests that this device could well reduce the need for ambulatory monitoring, with its associated cost. The ability to negate the white-coat effect with the use of the BpTRU in the office setting also has particular importance, not only for patient office readings, but for the diagnosis and subsequent treatment of hypertension in individual patients.

Most clinical decisions about the treatment of hypertension are still made on the basis of office determinations of blood pressure. Most office practices still rely on the aneroid manometer or, decreasingly, mercury sphygmomanometers. As noted earlier, although auscultatory blood pressure measurement appears to be simple, it is fraught with a host of observer- or patient-induced errors that not only lead to inaccurate diagnoses, but may also result in the mismanagement of hypertension. Even single-measurement oscillometric devices, now used in a minority of clinical practices, are associated with many of the same measurement issues that lead to overestimation of blood pressure.

We believe the time has come for broader use of oscillometric devices in the outpatient setting. While many available oscillometric devices for use in the home could also be used in the physician’s office, they carry the similar disadvantage of providing only a single measurement. The major disadvantage of all single-measurement devices is the continued presence of the clinician during the reading and the associated white-coat effect observed in most patients.

It is highly likely that the next Joint National Committee Report on Hypertension will further emphasize the role of automated blood pressure devices in the outpatient setting.
 

Acknowledgment: The authors wish to acknowledge the contributions of Deborah McCoy, RN, and Maria Eckhouse, RN.

The reality of blood pressure measurement is that human beings do not do it very well. The time has come to delegate this job to machines that can do it better.

Several automatic devices are available. Used in physicians’ offices and in patients’ homes, they can eliminate some types of observer error as well as the “white-coat effect,” ie, the tendency of some patients to have higher blood pressure when medical personnel are present than in their natural environment. Since a difference of only a few millimeters of mercury can affect the physician’s decision to start or to modify treatment, measurements that more accurately reflect a person’s average blood pressure are to be desired.

In the following pages, we review the problems that plague blood pressure measurement by human observers, and we describe the advantages of automatic devices.

SHORTCOMINGS OF OFFICE BLOOD PRESSURE MEASUREMENT

For decades, large surveys have provided invaluable information on the prevalence of hypertension, its relationship to cardiovascular disease, and the benefits of treating it.^1–3 Unfortunately, the percentage of hypertensive patients whose blood pressure is under control has remained low despite our increased knowledge about hypertension’s diagnosis and therapy.⁴

In these surveys, blood pressure was measured by auscultation by human observers using mercury or aneroid sphygmomanometers, and most physicians still use this method in clinical practice. But in spite of multiple guidelines for accurate measurement of blood pressure in the office, the overall accuracy and reproducibility of office blood pressure measurements remain poor.^5–7

The accuracy of blood pressure measurement with aneroid and mercury manometers is affected by a number of observer errors and patient factors.^8,9

Observer errors

Failure to prepare the patient. National guidelines⁵ state that before having their blood pressure taken, patients should be allowed to sit quietly for at least 5 minutes, which often does not happen. Another error is that clinicians rarely discourage patients from smoking cigarettes or drinking coffee in the 30 minutes prior to measurement.

Equipment and layout problems. Equipment should be properly calibrated and validated. ⁵ However, even if the sphygmomanometer is periodically calibrated, too often it is mounted on the wall adjacent to the examination table in the examination room, making it difficult to provide a comfortable seat with back and arm support during the reading. The measurement should be done with the patient sitting in a chair (not on an examination table), with feet on the floor and the arm supported at the level of the heart. If the forearm is not supported in the horizontal position and with the cuff at heart level, the blood pressure and heart rate tend to be higher.¹⁰ Further, the diastolic blood pressure and heart rate may be misleadingly low with the patient supine rather than seated,^11,12 so readings should be taken with the patient sitting.

Miscuffing, ie, the use of a blood pressure cuff that is too large or, more often, too small for the patient’s arm, is a common source of error. The cuff bladder should encircle at least 80% of the arm.⁵ However, some offices do not have a large blood pressure cuff for overweight patients or a pediatric cuff for children or adults with arms of small circumference. It is recommended that a large blood pressure cuff be used routinely in adults, since a smaller cuff gives falsely high readings in people with large upper arms (circumference > 29 cm).^13,14

Digit preference. Many physicians round off the blood pressure to the nearest 5 or 10 mm Hg. This problem may go along with:

Deflating the cuff too rapidly.

Talking to the patient while taking the blood pressure can contribute to higher readings.⁹

Not taking enough readings. Ideally, at the initial visit, blood pressure should be measured in both arms with the patient seated, and another reading should be taken with the patient standing. The arm with the higher pressure should be used for subsequent readings. Physicians should not make any treatment decisions based on blood pressure during an initial clinic visit, and at least two readings should be taken even on subsequent visits. However, owing to time constraints in busy clinical practices, treatment decisions are often based on single readings or on multiple readings on a single visit.

Discrepancies between observers. The blood pressure readings obtained by the nurse or medical assistant may differ significantly from those obtained by the physician. These differences can be large enough to affect treatment decisions,^15,16 and they can be partially corrected by adequate training of all medical personnel who take blood pressure, doctors as well as nurses.

Given that time is tight in busy clinical practices and a trained blood pressure nurse or technician is usually not available, we will probably not see any significant improvement in the accuracy of blood pressure measurement using older technology and current physician practices.

The white-coat effect

Most patients have a higher level of anxiety, and therefore higher blood pressure, in the physician’s office or clinic than in their normal environment (as revealed by ambulatory monitoring or home blood pressure measurements), a phenomenon commonly called the white-coat effect.

Several factors can increase this effect, such as observer-patient interaction during the measurement. The effect tends to be greatest in the initial measurement, but can persist through multiple readings by the doctor or nurse during the same visit.

Whether the white-coat effect is due purely to patient anxiety about an office visit or to a conditioned response has been a point of interest in clinical studies. Regardless, it may result in the misdiagnosis of hypertension or in overestimation of the severity of hypertension and may lead to overly aggressive therapy. Antihypertensive treatment may be unnecessary in the absence of concurrent cardiovascular risk factors.¹⁷

“White-coat hypertension” or “isolated office hypertension” is the condition in which a patient who is not on antihypertensive drug therapy has persistently elevated blood pressure in the clinic or office (> 140/90 mm Hg) but normal daytime ambulatory blood pressure (< 135/85 mm Hg).¹⁸ Since patients may have an elevated reading when seen for a first office visit, at least several visits are required to establish the diagnosis. Multiple studies have suggested that white-coat hypertension may account for 20% to 25% of the hypertensive population, particularly in older patients, mainly women.^19,20

Both white-coat hypertension and the white-coat effect can be avoided by using an automatic and programmable device that can take multiple readings after the clinician leaves the examination room (more about this below).²¹

MEASURING BLOOD PRESSURE OUTSIDE THE OFFICE

Recent studies have reported that the information obtained by 24-hour ambulatory blood pressure monitoring and by self-measurement of blood pressure in the home more accurately reflects the patient’s risk of future cardiovascular events than do conventional blood pressure measurements taken in the physician’s office. ^22–24 Current national guidelines recognize this pattern and the frequent measurement inaccuracies observed in clinical practice, and they are recommending including out-of-office measurements in the diagnosis of hypertension. ^25,26

Ambulatory monitoring provides the most accurate measurement of out-of-office blood pressure. With ambulatory monitoring, the normal mean daytime pressure is considered to be lower than 135/85 mm Hg, in contrast to the 140/90 mm Hg cutoff used in the physician’s office with standard aneroid or mercury devices.

Self-monitoring of blood pressure at home has now become widely available with single-measurement oscillometric devices. (Oscillometric means that these devices measure the blood pressure by sensing the oscillations in pressure in the cuff induced by the pulsation of the brachial artery, as opposed to auscultating the Korotkoff sounds.) Blood pressures lower than 135/85 mm Hg outside the clinician’s office are considered normal with these devices.

However, despite its proven value, ambulatory monitoring is neither widely available nor cost-effective for the long-term management of hypertension. Furthermore, few physicians recommend that patients take their blood pressure at home, although the information obtained can be of significant value in the patient’s long-term management.

AUTOMATED MEASUREMENT IN THE OFFICE

In recent years, several automated oscillometric sphygmomanometers have been developed for measuring blood pressure in the office, and more are on the way. These devices can be programmed to take multiple readings without a clinician observer in the examination room, thus reducing the white-coat response.

Omron (Kyoto, Japan) makes several devices, including the HEM-907 and the HEM-705, that have been used in the clinical setting. ^21,27–29 They can be programmed to take two or three readings at intervals of 1 to 2 minutes, with up to 5 minutes before the first reading. Unfortunately, data were not recorded with the patient alone in the room in many studies of the Omron devices, even though the devices meet national and international standards for accuracy.

The Microlife Watch BP Office (Microlife, Widnau, Switzerland) is currently undergoing development.³⁰

The BpTRU (BpTRU Medical Devices, Coquitlam, BC, Canada) has enjoyed greater clinical acceptance, since it can take up to five blood pressure readings at intervals of 1 to 5 minutes, and calculates the mean of all five readings, taken with the patient resting comfortably in a quiet room without a clinician present.

The accuracy and durability of the device has been well established. Since the BpTRU self-calibrates between every blood pressure measurement, periodic calibration has not been required. The device can be placed on a table, mounted on the wall, or mounted on a cart if used in several locations in the office.

At Cleveland Clinic, several departments are using the BpTRU on a daily basis. Soon, we will be able to transfer data directly from the BpTRU to our electronic medical record system.

Studies of the BpTRU device

To date, most of the studies of automated office blood pressure measurement have used the BpTRU with the recording interval set at 1 to 2 minutes.

Myers³¹ used the BpTRU device in 50 hypertensive patients. The physician took the patient’s blood pressure with a mercury sphygmomanometer while the BpTRU device made the first reading, and then he left the room. The next five readings were taken at 2-minute intervals with the patient alone in the room. The mean initial reading by the machine was 162/85 mm Hg; the reading by the physician was 163/86 mm Hg. The third automatic reading was the lowest (averaging 140/84 mm Hg), and the mean of the five automated readings was 142/80 mm Hg, which was significantly lower than the initial reading obtained by the physician (P < .001).

In another study, Myers et al³² compared the measurements obtained by 24-hour ambulatory monitoring and by the BpTRU device (the mean of five readings obtained at 1-minute or 2-minute intervals) in 309 hypertensive patients. The mean blood pressure with the Bp-TRU was 132/75 mm Hg, which correlated well with the mean awake ambulatory blood pressure (134/77 mm Hg; r = 0.62 for the systolic pressure and 0.72 for the diastolic pressure).

We recently reviewed the records of 278 patients seen in our preventive medicine clinic (D.G. Vidt, MD, unpublished data, November 2009). The group included patients with and without established hypertension, and among the hypertensive group, both treated and untreated individuals. We had initially set the device to take readings at 3-minute intervals following the initial nurse-initiated reading. But in view of the recent data on the Bp-TRU using shorter intervals, we also obtained readings in 51 patients with the device set to record at 2-minute intervals, and then in 72 additional patients at 1-minute intervals. In all three groups, blood pressure had stabilized by the third reading after the clinician had left the room. These observations support those reported by Myers et al.^31,32 Of particular importance is the observation that the white-coat effect dissipates within 2 to 3 minutes after the clinician leaves the room.³³

The shorter measurement intervals can add up in a busy office practice, in which the time relegated to taking blood pressure is often limited.

In fact, waiting 5 minutes between measurements may allow the patient to become too relaxed and the blood pressure to drop too low vis-a-vis the gold standard, ambulatory monitoring. Culleton and colleagues³⁴ compared the blood pressure in 107 hypertensive patients as measured four ways: by the referring physician, by a nurse who was trained to adhere to the protocol of the Canadian Hypertension Education Program, by 24-hour ambulatory monitoring, and by the BpTRU (the mean of five readings obtained at 5-minute intervals). The mean measured values were:

• 150/90 mm Hg by the referring physician
• 139/86 mm Hg by the nurse
• 142/85 mm Hg by ambulatory monitoring
• 132/82 mm Hg by the BpTRU device.

Although the BpTRU reduced the white-coat effect and white-coat hypertension, it underestimated the blood pressure, leading to misclassification of hypertension. Using 140/90 mm Hg as the cutoff for whether the patient was hypertensive and using ambulatory monitoring as the gold standard, the BpTRU misclassified more than half of the patients, agreeing with the classification of hypertensive or not hypertensive by ambulatory monitoring in only 48%. The authors recommended that the BpTRU not be set at 5-minute measurement intervals.³⁴

WHAT ROLE FOR AUTOMATED READINGS IN THE OFFICE?

Although automatic devices, by enabling the physician to leave the room, can eliminate the white-coat effect and white-coat hypertension, physicians must continue to take care to avoid the other potential errors of office blood pressure measurement addressed earlier in this review, for example, by positioning the patient correctly and using a cuff that is large enough. These issues can take on more importance as the clinician leaves the patient alone for brief periods during measurements.

In view of its perennial inaccuracies, some experts have suggested that we abandon routine office measurement of blood pressure.^35,36 In its place, ambulatory monitoring would be used for diagnosis and for periodic follow-up. In addition, patients would regularly take their pressure at home with approved, single-measurement oscillometric devices. Unfortunately, in our health care system, periodic ambulatory monitoring for hypertension management would impose a significant financial burden on patients at this time.³⁷

Of particular importance is the observation that the mean of five readings with the BpTRU device, obtained at 1- or 2-minute intervals, closely approximates the mean awake blood pressure obtained in the same patient with an ambulatory monitor.^32,38 The ability to obtain readings that correlate exceptionally well with mean daytime ambulatory pressure suggests that this device could well reduce the need for ambulatory monitoring, with its associated cost. The ability to negate the white-coat effect with the use of the BpTRU in the office setting also has particular importance, not only for patient office readings, but for the diagnosis and subsequent treatment of hypertension in individual patients.

Most clinical decisions about the treatment of hypertension are still made on the basis of office determinations of blood pressure. Most office practices still rely on the aneroid manometer or, decreasingly, mercury sphygmomanometers. As noted earlier, although auscultatory blood pressure measurement appears to be simple, it is fraught with a host of observer- or patient-induced errors that not only lead to inaccurate diagnoses, but may also result in the mismanagement of hypertension. Even single-measurement oscillometric devices, now used in a minority of clinical practices, are associated with many of the same measurement issues that lead to overestimation of blood pressure.

We believe the time has come for broader use of oscillometric devices in the outpatient setting. While many available oscillometric devices for use in the home could also be used in the physician’s office, they carry the similar disadvantage of providing only a single measurement. The major disadvantage of all single-measurement devices is the continued presence of the clinician during the reading and the associated white-coat effect observed in most patients.

It is highly likely that the next Joint National Committee Report on Hypertension will further emphasize the role of automated blood pressure devices in the outpatient setting.
 

Acknowledgment: The authors wish to acknowledge the contributions of Deborah McCoy, RN, and Maria Eckhouse, RN.