Nonmelanoma Skin Cancer

Dermatologists have a generally favorable attitude regarding the use of artificial intelligence (AI) in their practices, but few have actually used it yet, according to the results of a small survey.

Just 9% of the 90 respondents acknowledged that they have used AI in their practices, while 81% said they had not, and 10% weren’t sure or didn’t know. Despite that lack of familiarity, however, “many embrace the potential positive benefits, such as reducing misdiagnoses” and a majority (94.5%) “would use it at least in certain scenarios,” Vishal A. Patel, MD, and associates said in the Journal of Drugs in Dermatology.

Dermatologists aged 40 years and under were more likely to have used AI previously: 15% reported previous experience, compared with 4% of those over age 40 – but the difference in “age did not have a significant effect on perception of AI,” the investigators noted, adding that most of the dermatologists over 40 believe “that AI would be most beneficial and used for detection of malignant skin lesions.”

The survey also asked about ways the respondents would use AI to help their patients. Almost two-thirds of respondents (66%) chose analysis and management of electronic health records “for research purposes to improve patient outcomes,” compared with 56% who chose identifying unknown/screening skin lesions “with a list of differential diagnoses,” 32% who chose telemedicine, and 26% who chose primary surveys of skin, said Dr. Patel, director of cutaneous oncology at the George Washington University Cancer Center in Washington, and coauthors.

The respondents were fairly evenly split when asked about the possible impact of nondermatologists using AI in the near future to detect skin lesions, such as melanomas, on the need for dermatologists. Just over a quarter said that the need for dermatologists will be decreased all (about 4.4%) or some (about 21.1%) of the time, and 24.4% said that the need will be increased, with the largest share (39.9%) of respondents choosing the middle ground: neither increased or decreased, the investigators reported.

The survey form was emailed to 850 members of the Orlando Dermatology, Aesthetic & Surgical Conference listserv, with responses accepted from April 13 to May 14, 2021. The investigators noted that the response rate was low enough to be a limiting factor, making selection bias “by those with a particular interest in the topic” a possibility.

No funding sources for the study were disclosed. Dr. Patel disclosed that he is chief medical officer for Lazarus AI, the other authors had no disclosures listed.

Dermatologists have a generally favorable attitude regarding the use of artificial intelligence (AI) in their practices, but few have actually used it yet, according to the results of a small survey.

Just 9% of the 90 respondents acknowledged that they have used AI in their practices, while 81% said they had not, and 10% weren’t sure or didn’t know. Despite that lack of familiarity, however, “many embrace the potential positive benefits, such as reducing misdiagnoses” and a majority (94.5%) “would use it at least in certain scenarios,” Vishal A. Patel, MD, and associates said in the Journal of Drugs in Dermatology.

Dermatologists aged 40 years and under were more likely to have used AI previously: 15% reported previous experience, compared with 4% of those over age 40 – but the difference in “age did not have a significant effect on perception of AI,” the investigators noted, adding that most of the dermatologists over 40 believe “that AI would be most beneficial and used for detection of malignant skin lesions.”

The survey also asked about ways the respondents would use AI to help their patients. Almost two-thirds of respondents (66%) chose analysis and management of electronic health records “for research purposes to improve patient outcomes,” compared with 56% who chose identifying unknown/screening skin lesions “with a list of differential diagnoses,” 32% who chose telemedicine, and 26% who chose primary surveys of skin, said Dr. Patel, director of cutaneous oncology at the George Washington University Cancer Center in Washington, and coauthors.

The respondents were fairly evenly split when asked about the possible impact of nondermatologists using AI in the near future to detect skin lesions, such as melanomas, on the need for dermatologists. Just over a quarter said that the need for dermatologists will be decreased all (about 4.4%) or some (about 21.1%) of the time, and 24.4% said that the need will be increased, with the largest share (39.9%) of respondents choosing the middle ground: neither increased or decreased, the investigators reported.

The survey form was emailed to 850 members of the Orlando Dermatology, Aesthetic & Surgical Conference listserv, with responses accepted from April 13 to May 14, 2021. The investigators noted that the response rate was low enough to be a limiting factor, making selection bias “by those with a particular interest in the topic” a possibility.

No funding sources for the study were disclosed. Dr. Patel disclosed that he is chief medical officer for Lazarus AI, the other authors had no disclosures listed.

Dermatologists have a generally favorable attitude regarding the use of artificial intelligence (AI) in their practices, but few have actually used it yet, according to the results of a small survey.

Just 9% of the 90 respondents acknowledged that they have used AI in their practices, while 81% said they had not, and 10% weren’t sure or didn’t know. Despite that lack of familiarity, however, “many embrace the potential positive benefits, such as reducing misdiagnoses” and a majority (94.5%) “would use it at least in certain scenarios,” Vishal A. Patel, MD, and associates said in the Journal of Drugs in Dermatology.

Dermatologists aged 40 years and under were more likely to have used AI previously: 15% reported previous experience, compared with 4% of those over age 40 – but the difference in “age did not have a significant effect on perception of AI,” the investigators noted, adding that most of the dermatologists over 40 believe “that AI would be most beneficial and used for detection of malignant skin lesions.”

The survey also asked about ways the respondents would use AI to help their patients. Almost two-thirds of respondents (66%) chose analysis and management of electronic health records “for research purposes to improve patient outcomes,” compared with 56% who chose identifying unknown/screening skin lesions “with a list of differential diagnoses,” 32% who chose telemedicine, and 26% who chose primary surveys of skin, said Dr. Patel, director of cutaneous oncology at the George Washington University Cancer Center in Washington, and coauthors.

The respondents were fairly evenly split when asked about the possible impact of nondermatologists using AI in the near future to detect skin lesions, such as melanomas, on the need for dermatologists. Just over a quarter said that the need for dermatologists will be decreased all (about 4.4%) or some (about 21.1%) of the time, and 24.4% said that the need will be increased, with the largest share (39.9%) of respondents choosing the middle ground: neither increased or decreased, the investigators reported.

The survey form was emailed to 850 members of the Orlando Dermatology, Aesthetic & Surgical Conference listserv, with responses accepted from April 13 to May 14, 2021. The investigators noted that the response rate was low enough to be a limiting factor, making selection bias “by those with a particular interest in the topic” a possibility.

No funding sources for the study were disclosed. Dr. Patel disclosed that he is chief medical officer for Lazarus AI, the other authors had no disclosures listed.

If you worry that artificial intelligence (AI) will one day replace your own clinical acumen as a dermatologist, Vishal A. Patel, MD, advises you to think differently.

“AI is meant to be an enhancement strategy, a support tool to improve our diagnostic abilities,” Dr. Patel, a Mohs surgeon who is director of cutaneous oncology at the George Washington University Cancer Center, Washington, said during the ODAC Dermatology, Aesthetic & Surgical Conference. “Dermatologists should embrace AI and drive how it is utilized – be the captain of the plane (technology) and the passenger (patient). If we’re not in the forefront of the plane, we’re not to be able to dictate which way we are going with this.”

In 2019, a group of German researchers found that AI can improve accuracy and efficiency of specialists in classifying skin cancer based on dermoscopic images. “I really do believe this is going to be the future,” said Dr. Patel, who was not involved with the study. “Current research involves using supervised learning on known outcomes to determine inputs to predict them. In dermatology, think of identifying melanoma from clinical or dermoscopic images or predicting metastasis risk from digitized pathology slides.”

However, there are currently no universal guidelines on how large an AI dataset needs to be to yield accurate results. In the dermatology literature, most AI datasets range between 600 and 14,000 examples, Dr. Patel said, with a large study-specific variation in performance. “Misleading results can result from unanticipated training errors,” he said.

“The AI network may learn its intended task or an unrelated situational cue. For example, you can use great images to predict melanoma, but you may have an unintended poor outcome related to images that have, say, a ruler inside of them clustered within the melanoma diagnoses.” And unbeknown to the system’s developer, “the algorithm picks up that the ruler is predictive of an image being a melanoma and not the pigmented lesion itself.” In other words, the algorithm is only as good as the dataset being used, he said. “This is the key element, to ask what the dataset is that’s training the tool that you may one day use.”
 

Convolutional neural network

In 2017, a seminal study published in Nature showed that for classification of melanoma and epidermal lesions, a type of AI used in image processing known as a convolutional neural network (CNN) was on par with dermatologists and outperformed the average. For epidermal lesions, the network was one standard deviation higher above the average for dermatologists, while for melanocytic lesions, the network was just below one standard deviation above the average of the dermatologists. A CNN “clearly can perform well because it works on a different level than how our brains work,” Dr. Patel said.

In a separate study, a CNN trained to recognize melanoma in dermoscopic images was compared to 58 international dermatologists with varying levels of dermoscopy experience; 29% were “beginners,” with less than 2 years of experience; 19% were “skilled,” with 2-5 years of experience; and 52% were “experts,” with at least 5 years of experience. The analysis consisted of two experiments: In level I, dermatologists classified lesions based on dermoscopy only. In level II, dermatologists were provided dermoscopy, clinical images, and additional clinical information, while the CNN was trained on images only. The researchers found that most dermatologists were outperformed by the CNN. “Physicians of all different levels of training and experience may benefit from assistance by a CNN’s image classification,” they concluded.
 

Gene expression profiling

Another aspect of AI is gene expression profiling (GEP), which Dr. Patel defined as the evaluation of frequency and intensity of genetic activity at once to create a global picture of cellular function. “It’s AI that uses machine learning to evaluate genetic expression to assess lesion behavior,” he explained.

One GEP test on the market is the Pigmented Lesion Assay (PLA) from DermTech, a noninvasive test that looks at the expression of two genes to predict if a lesion is malignant or not. “Based on their validation set, they have shown some impressive numbers,” with sensitivities above 90%, and published registry data that have shown higher sensitivities “and even specificities above 90%,” he said.

“On the surface, it looks like this would be a useful test,” Dr. Patel said. A study published in 2021 looked at the evidence of applying real-world evidence with this test to see if results held up. Based on the authors’ analysis, he noted, “you would need a sensitivity and specificity of 95% to yield a positivity rate of 9.5% for the PLA test, which is what has been reported in real-world use. So, there’s a disconnect somewhere and we are not quite there yet.” That may be a result of the dataset itself not being as uniform between the validation and the training datasets, he continued. Also, the expression of certain genes is different “if you don’t have a clean input variable” of what the test is being used for, he added.

“If you’re not mirroring the dataset, you’re not going to get clean data,” he said. “So, if you’re using this on younger patients or for sun-damaged lesional skin or nonmelanocytic lesions around sun-damaged areas, there are variable expressions that may not be accurately captured by that algorithm. This might help explain the real-world variation that we’re seeing.”

Another GEP test in use is the 31-Gene Expression Profile Test for Melanoma, which evaluates gene expressions in melanoma tumors and what the behavior of that tumor may be. The test has been available for more than a decade “and there is a lot of speculation about its use,” Dr. Patel said. “A recent paper attempted to come up with an algorithm of how to use this, but there’s a lot of concern about the endpoints of what changes in management might result from this test. That is what we need to be thinking about. There’s a lot of back and forth about this.”

In 2020, authors of a consensus statement on prognostic GEP in cutaneous melanoma concluded that before GEP testing is routinely used, the clinical benefit in the management of patients with melanoma should be established through further clinical investigation. Dr. Patel recommended the accompanying editorial on GEP in melanoma, written by Hensin Tsao, MD, PhD, and Warren H. Chan, MS, in JAMA Dermatology.

In Dr. Patel’s opinion, T1a melanomas (0.8 mm, nonulcerated) do not need routine GEP, but the GEP test may be useful in cases that are in the “gray zone,” such as those with T1b or some borderline T2a melanomas (> 0.8 mm, < 1.2mm, nonulcerated, but with high mitosis, etc.); patients with unique coexisting conditions such as pregnancy, and patients who may not tolerate sentinel lymph node biopsy (SLNB) or adjuvant therapy.

Echoing sentiments expressed in the JAMA Dermatology editorial, he advised dermatologists to “remember your training and know the data. GEP predicting survival is not the same as SLNB positive rate. GEP should not replace standard guidelines in T2a and higher melanomas. Nodal sampling remains part of all major guidelines and determines adjuvant therapy.”

He cited the characterization of GEP in the editorial as “a powerful technology” that heralds the age of personalized medicine, but it is not ready for ubiquitous use. Prospective studies and time will lead to highly accurate tools.”

Dr. Patel disclosed that he is chief medical officer for Lazarus AI.

If you worry that artificial intelligence (AI) will one day replace your own clinical acumen as a dermatologist, Vishal A. Patel, MD, advises you to think differently.

“AI is meant to be an enhancement strategy, a support tool to improve our diagnostic abilities,” Dr. Patel, a Mohs surgeon who is director of cutaneous oncology at the George Washington University Cancer Center, Washington, said during the ODAC Dermatology, Aesthetic & Surgical Conference. “Dermatologists should embrace AI and drive how it is utilized – be the captain of the plane (technology) and the passenger (patient). If we’re not in the forefront of the plane, we’re not to be able to dictate which way we are going with this.”

In 2019, a group of German researchers found that AI can improve accuracy and efficiency of specialists in classifying skin cancer based on dermoscopic images. “I really do believe this is going to be the future,” said Dr. Patel, who was not involved with the study. “Current research involves using supervised learning on known outcomes to determine inputs to predict them. In dermatology, think of identifying melanoma from clinical or dermoscopic images or predicting metastasis risk from digitized pathology slides.”

However, there are currently no universal guidelines on how large an AI dataset needs to be to yield accurate results. In the dermatology literature, most AI datasets range between 600 and 14,000 examples, Dr. Patel said, with a large study-specific variation in performance. “Misleading results can result from unanticipated training errors,” he said.

“The AI network may learn its intended task or an unrelated situational cue. For example, you can use great images to predict melanoma, but you may have an unintended poor outcome related to images that have, say, a ruler inside of them clustered within the melanoma diagnoses.” And unbeknown to the system’s developer, “the algorithm picks up that the ruler is predictive of an image being a melanoma and not the pigmented lesion itself.” In other words, the algorithm is only as good as the dataset being used, he said. “This is the key element, to ask what the dataset is that’s training the tool that you may one day use.”
 

Convolutional neural network

In 2017, a seminal study published in Nature showed that for classification of melanoma and epidermal lesions, a type of AI used in image processing known as a convolutional neural network (CNN) was on par with dermatologists and outperformed the average. For epidermal lesions, the network was one standard deviation higher above the average for dermatologists, while for melanocytic lesions, the network was just below one standard deviation above the average of the dermatologists. A CNN “clearly can perform well because it works on a different level than how our brains work,” Dr. Patel said.

In a separate study, a CNN trained to recognize melanoma in dermoscopic images was compared to 58 international dermatologists with varying levels of dermoscopy experience; 29% were “beginners,” with less than 2 years of experience; 19% were “skilled,” with 2-5 years of experience; and 52% were “experts,” with at least 5 years of experience. The analysis consisted of two experiments: In level I, dermatologists classified lesions based on dermoscopy only. In level II, dermatologists were provided dermoscopy, clinical images, and additional clinical information, while the CNN was trained on images only. The researchers found that most dermatologists were outperformed by the CNN. “Physicians of all different levels of training and experience may benefit from assistance by a CNN’s image classification,” they concluded.
 

Gene expression profiling

Another aspect of AI is gene expression profiling (GEP), which Dr. Patel defined as the evaluation of frequency and intensity of genetic activity at once to create a global picture of cellular function. “It’s AI that uses machine learning to evaluate genetic expression to assess lesion behavior,” he explained.

One GEP test on the market is the Pigmented Lesion Assay (PLA) from DermTech, a noninvasive test that looks at the expression of two genes to predict if a lesion is malignant or not. “Based on their validation set, they have shown some impressive numbers,” with sensitivities above 90%, and published registry data that have shown higher sensitivities “and even specificities above 90%,” he said.

“On the surface, it looks like this would be a useful test,” Dr. Patel said. A study published in 2021 looked at the evidence of applying real-world evidence with this test to see if results held up. Based on the authors’ analysis, he noted, “you would need a sensitivity and specificity of 95% to yield a positivity rate of 9.5% for the PLA test, which is what has been reported in real-world use. So, there’s a disconnect somewhere and we are not quite there yet.” That may be a result of the dataset itself not being as uniform between the validation and the training datasets, he continued. Also, the expression of certain genes is different “if you don’t have a clean input variable” of what the test is being used for, he added.

“If you’re not mirroring the dataset, you’re not going to get clean data,” he said. “So, if you’re using this on younger patients or for sun-damaged lesional skin or nonmelanocytic lesions around sun-damaged areas, there are variable expressions that may not be accurately captured by that algorithm. This might help explain the real-world variation that we’re seeing.”

Another GEP test in use is the 31-Gene Expression Profile Test for Melanoma, which evaluates gene expressions in melanoma tumors and what the behavior of that tumor may be. The test has been available for more than a decade “and there is a lot of speculation about its use,” Dr. Patel said. “A recent paper attempted to come up with an algorithm of how to use this, but there’s a lot of concern about the endpoints of what changes in management might result from this test. That is what we need to be thinking about. There’s a lot of back and forth about this.”

In 2020, authors of a consensus statement on prognostic GEP in cutaneous melanoma concluded that before GEP testing is routinely used, the clinical benefit in the management of patients with melanoma should be established through further clinical investigation. Dr. Patel recommended the accompanying editorial on GEP in melanoma, written by Hensin Tsao, MD, PhD, and Warren H. Chan, MS, in JAMA Dermatology.

In Dr. Patel’s opinion, T1a melanomas (0.8 mm, nonulcerated) do not need routine GEP, but the GEP test may be useful in cases that are in the “gray zone,” such as those with T1b or some borderline T2a melanomas (> 0.8 mm, < 1.2mm, nonulcerated, but with high mitosis, etc.); patients with unique coexisting conditions such as pregnancy, and patients who may not tolerate sentinel lymph node biopsy (SLNB) or adjuvant therapy.

Echoing sentiments expressed in the JAMA Dermatology editorial, he advised dermatologists to “remember your training and know the data. GEP predicting survival is not the same as SLNB positive rate. GEP should not replace standard guidelines in T2a and higher melanomas. Nodal sampling remains part of all major guidelines and determines adjuvant therapy.”

He cited the characterization of GEP in the editorial as “a powerful technology” that heralds the age of personalized medicine, but it is not ready for ubiquitous use. Prospective studies and time will lead to highly accurate tools.”

Dr. Patel disclosed that he is chief medical officer for Lazarus AI.

If you worry that artificial intelligence (AI) will one day replace your own clinical acumen as a dermatologist, Vishal A. Patel, MD, advises you to think differently.

“AI is meant to be an enhancement strategy, a support tool to improve our diagnostic abilities,” Dr. Patel, a Mohs surgeon who is director of cutaneous oncology at the George Washington University Cancer Center, Washington, said during the ODAC Dermatology, Aesthetic & Surgical Conference. “Dermatologists should embrace AI and drive how it is utilized – be the captain of the plane (technology) and the passenger (patient). If we’re not in the forefront of the plane, we’re not to be able to dictate which way we are going with this.”

In 2019, a group of German researchers found that AI can improve accuracy and efficiency of specialists in classifying skin cancer based on dermoscopic images. “I really do believe this is going to be the future,” said Dr. Patel, who was not involved with the study. “Current research involves using supervised learning on known outcomes to determine inputs to predict them. In dermatology, think of identifying melanoma from clinical or dermoscopic images or predicting metastasis risk from digitized pathology slides.”

However, there are currently no universal guidelines on how large an AI dataset needs to be to yield accurate results. In the dermatology literature, most AI datasets range between 600 and 14,000 examples, Dr. Patel said, with a large study-specific variation in performance. “Misleading results can result from unanticipated training errors,” he said.

“The AI network may learn its intended task or an unrelated situational cue. For example, you can use great images to predict melanoma, but you may have an unintended poor outcome related to images that have, say, a ruler inside of them clustered within the melanoma diagnoses.” And unbeknown to the system’s developer, “the algorithm picks up that the ruler is predictive of an image being a melanoma and not the pigmented lesion itself.” In other words, the algorithm is only as good as the dataset being used, he said. “This is the key element, to ask what the dataset is that’s training the tool that you may one day use.”
 

Convolutional neural network

In 2017, a seminal study published in Nature showed that for classification of melanoma and epidermal lesions, a type of AI used in image processing known as a convolutional neural network (CNN) was on par with dermatologists and outperformed the average. For epidermal lesions, the network was one standard deviation higher above the average for dermatologists, while for melanocytic lesions, the network was just below one standard deviation above the average of the dermatologists. A CNN “clearly can perform well because it works on a different level than how our brains work,” Dr. Patel said.

In a separate study, a CNN trained to recognize melanoma in dermoscopic images was compared to 58 international dermatologists with varying levels of dermoscopy experience; 29% were “beginners,” with less than 2 years of experience; 19% were “skilled,” with 2-5 years of experience; and 52% were “experts,” with at least 5 years of experience. The analysis consisted of two experiments: In level I, dermatologists classified lesions based on dermoscopy only. In level II, dermatologists were provided dermoscopy, clinical images, and additional clinical information, while the CNN was trained on images only. The researchers found that most dermatologists were outperformed by the CNN. “Physicians of all different levels of training and experience may benefit from assistance by a CNN’s image classification,” they concluded.
 

Gene expression profiling

Another aspect of AI is gene expression profiling (GEP), which Dr. Patel defined as the evaluation of frequency and intensity of genetic activity at once to create a global picture of cellular function. “It’s AI that uses machine learning to evaluate genetic expression to assess lesion behavior,” he explained.

One GEP test on the market is the Pigmented Lesion Assay (PLA) from DermTech, a noninvasive test that looks at the expression of two genes to predict if a lesion is malignant or not. “Based on their validation set, they have shown some impressive numbers,” with sensitivities above 90%, and published registry data that have shown higher sensitivities “and even specificities above 90%,” he said.

“On the surface, it looks like this would be a useful test,” Dr. Patel said. A study published in 2021 looked at the evidence of applying real-world evidence with this test to see if results held up. Based on the authors’ analysis, he noted, “you would need a sensitivity and specificity of 95% to yield a positivity rate of 9.5% for the PLA test, which is what has been reported in real-world use. So, there’s a disconnect somewhere and we are not quite there yet.” That may be a result of the dataset itself not being as uniform between the validation and the training datasets, he continued. Also, the expression of certain genes is different “if you don’t have a clean input variable” of what the test is being used for, he added.

“If you’re not mirroring the dataset, you’re not going to get clean data,” he said. “So, if you’re using this on younger patients or for sun-damaged lesional skin or nonmelanocytic lesions around sun-damaged areas, there are variable expressions that may not be accurately captured by that algorithm. This might help explain the real-world variation that we’re seeing.”

Another GEP test in use is the 31-Gene Expression Profile Test for Melanoma, which evaluates gene expressions in melanoma tumors and what the behavior of that tumor may be. The test has been available for more than a decade “and there is a lot of speculation about its use,” Dr. Patel said. “A recent paper attempted to come up with an algorithm of how to use this, but there’s a lot of concern about the endpoints of what changes in management might result from this test. That is what we need to be thinking about. There’s a lot of back and forth about this.”

In 2020, authors of a consensus statement on prognostic GEP in cutaneous melanoma concluded that before GEP testing is routinely used, the clinical benefit in the management of patients with melanoma should be established through further clinical investigation. Dr. Patel recommended the accompanying editorial on GEP in melanoma, written by Hensin Tsao, MD, PhD, and Warren H. Chan, MS, in JAMA Dermatology.

In Dr. Patel’s opinion, T1a melanomas (0.8 mm, nonulcerated) do not need routine GEP, but the GEP test may be useful in cases that are in the “gray zone,” such as those with T1b or some borderline T2a melanomas (> 0.8 mm, < 1.2mm, nonulcerated, but with high mitosis, etc.); patients with unique coexisting conditions such as pregnancy, and patients who may not tolerate sentinel lymph node biopsy (SLNB) or adjuvant therapy.

Echoing sentiments expressed in the JAMA Dermatology editorial, he advised dermatologists to “remember your training and know the data. GEP predicting survival is not the same as SLNB positive rate. GEP should not replace standard guidelines in T2a and higher melanomas. Nodal sampling remains part of all major guidelines and determines adjuvant therapy.”

He cited the characterization of GEP in the editorial as “a powerful technology” that heralds the age of personalized medicine, but it is not ready for ubiquitous use. Prospective studies and time will lead to highly accurate tools.”

Dr. Patel disclosed that he is chief medical officer for Lazarus AI.

The Diagnosis: Basal Cell Carcinoma

Histopathology was consistent with fungating basal cell carcinoma (BCC). The nodules were comprised of syncytial basaloid cells with high nuclear to cytoplasmic ratios, numerous mitotic figures, fibromyxoid stroma, and peripheral nuclear palisading (Figure). Fortunately, no perineural or lymphovascular invasion was identified, and the margins of the specimen were negative. Despite the high-risk nature of giant BCC, the mass was solitary without notable local invasion, leaving it amendable to surgery. On follow-up, the patient has remained recurrence free, and her hemoglobin level has since stabilized.

Skin cancer is the most common malignancy worldwide, and BCC accounts for more than 80% of nonmelanoma skin cancers in the United States. The incidence is on the rise due to the aging population and increasing cumulative skin exposure.¹ Risk factors include both individual physical characteristics and environmental exposures. Individuals with lighter skin tones, red and blonde hair, and blue and green eyes are at an increased risk.² UV radiation exposure is the most important cause of BCC.³ Chronic immunosuppression and exposure to arsenic, ionizing radiation, and psoralen plus UVA radiation also have been linked to the development of BCC.^4-6 Basal cell carcinomas most commonly arise on sun-exposed areas such as the face, though more than 10% of cases appear on the trunk.⁷ Lesions characteristically remain localized, and growth rate is variable; however, when left untreated, BCCs have the potential to become locally destructive and difficult to treat.

Advanced BCCs are tumors that penetrate deeply into the skin. They often are not amenable to traditional therapy and/ or metastasize. Those that grow to a diameter greater than 5 cm, as in our patient, are known as giant BCCs. Only 0.5% to 1% of BCCs are giant BCCs⁸ ; they typically are more aggressive in nature with higher rates of local recurrence and metastasis. Individuals who develop giant BCCs either have had a delay in access to medical care or a history of BCC that was inadequately managed.^9,10 During the COVID-19 pandemic, patient access to health care was substantially impacted during lockdowns. As in our patient, skin neoplasms and other medical conditions may present in later stages due to medical neglect.^11,12 Metastasis is rare, even in advanced BCCs. A review of the literature from 1984 estimated that the incidence of metastasis of BCCs is 1 in 1000 to 35,000. Metastasis portends a poor prognosis with a median overall survival of 8 to 14 months.¹³ An updated review in 2013 found similar outcomes.¹⁴

The choice of management for BCCs depends on the risk for recurrence as well as individual patient factors. Characteristics such as tumor size, location, histology, whether it is a primary or recurrent lesion, and the presence of chronic skin disease determine the recurrence rate.¹⁵ The management of advanced BCCs often requires a multidisciplinary approach, as these neoplasms may not be amenable to local therapy without causing substantial morbidity. Mohs micrographic surgery is the treatment of choice for BCCs at high risk for recurrence.¹⁶ Standard surgical excision with postoperative margin assessment is acceptable when Mohs micrographic surgery is not available.¹⁷ Radiation therapy is an alternative for patients who are not candidates for surgery.¹⁸

Recently, improved understanding of the molecular pathogenesis of BCCs has led to the development of novel systemic therapies. The Hedgehog signaling pathway has been found to play a critical role in the development of most BCCs.¹⁹ Vismodegib and sonidegib are small-molecule inhibitors of the Hedgehog signaling pathway approved for the treatment of locally advanced and metastatic BCCs that are not amenable to surgery or radiation. Approximately 50% of advanced BCCs respond to these therapies; however, long-term treatment may be limited by intolerable side effects and the development of resistance.²⁰ Basal cell carcinomas that spread to lymph nodes or distant sites are treated with traditional systemic therapy. Historically, conventional cytotoxic chemotherapies, such as platinum-containing regimens, were employed with limited benefit and notable morbidity.²¹

The differential diagnosis for our patient included several other cutaneous neoplasms. Squamous cell carcinoma is the second most common type of skin cancer. Similar to BCC, it can reach a substantial size if left untreated. Risk factors include chronic inflammation, exposure to radiation or chemical carcinogens, burns, human papillomavirus, and other chronic infections. Giant squamous cell carcinomas have high malignant potential and require imaging to assess the extent of invasion and for metastasis. Surgery typically is necessary for both staging and treatment. Adjuvant therapy also may be necessary.^22,23

Internal malignant neoplasms rarely present as cutaneous metastases. Breast cancer, melanoma, and cancers of the upper respiratory tract most frequently metastasize to the skin. Although colorectal cancer (CRC) rarely metastasizes to the skin, it is an important cause of cutaneous metastasis due to its high incidence in the general population. When it does spread to the skin, CRC preferentially affects the abdominal wall. Lesions typically resemble the primary tumor but may appear anaplastic. The occurrence of cutaneous metastasis suggests latestage disease and carries a poor prognosis.²⁴

Merkel cell carcinoma and melanoma are aggressive skin cancers with high mortality rates. The former is rarer but more lethal. Merkel cell carcinomas typically occur in elderly white men on sun-exposed areas of the skin. Tumors present as asymptomatic, rapidly expanding, blue-red, firm nodules. Immunosuppression and UV light exposure are notable risk factors.²⁵ Of the 4 major subtypes of cutaneous melanoma, superficial spreading is the most common, followed by nodular, lentigo maligna, and acral lentiginous.²⁶ Superficial spreading melanoma characteristically presents as an expanding asymmetric macule or thin plaque with irregular borders and variation in size and color (black, brown, or red). Nodular melanoma usually presents as symmetric in shape and color (amelanotic, black, or brown). Early recognition by both the patient and clinician is essential in preventing tumor growth and progression.²⁷

Our patient’s presentation was highly concerning for cutaneous metastasis given her history of CRC. Furthermore, the finding of severe anemia was atypical for skin cancer and more characteristic of the prior malignancy. Imaging revealed a locally confined mass with no evidence of extension, lymph node involvement, or additional lesions. The diagnosis was clinched with histopathologic examination.

The Diagnosis: Basal Cell Carcinoma

Histopathology was consistent with fungating basal cell carcinoma (BCC). The nodules were comprised of syncytial basaloid cells with high nuclear to cytoplasmic ratios, numerous mitotic figures, fibromyxoid stroma, and peripheral nuclear palisading (Figure). Fortunately, no perineural or lymphovascular invasion was identified, and the margins of the specimen were negative. Despite the high-risk nature of giant BCC, the mass was solitary without notable local invasion, leaving it amendable to surgery. On follow-up, the patient has remained recurrence free, and her hemoglobin level has since stabilized.

Skin cancer is the most common malignancy worldwide, and BCC accounts for more than 80% of nonmelanoma skin cancers in the United States. The incidence is on the rise due to the aging population and increasing cumulative skin exposure.¹ Risk factors include both individual physical characteristics and environmental exposures. Individuals with lighter skin tones, red and blonde hair, and blue and green eyes are at an increased risk.² UV radiation exposure is the most important cause of BCC.³ Chronic immunosuppression and exposure to arsenic, ionizing radiation, and psoralen plus UVA radiation also have been linked to the development of BCC.^4-6 Basal cell carcinomas most commonly arise on sun-exposed areas such as the face, though more than 10% of cases appear on the trunk.⁷ Lesions characteristically remain localized, and growth rate is variable; however, when left untreated, BCCs have the potential to become locally destructive and difficult to treat.

Advanced BCCs are tumors that penetrate deeply into the skin. They often are not amenable to traditional therapy and/ or metastasize. Those that grow to a diameter greater than 5 cm, as in our patient, are known as giant BCCs. Only 0.5% to 1% of BCCs are giant BCCs⁸ ; they typically are more aggressive in nature with higher rates of local recurrence and metastasis. Individuals who develop giant BCCs either have had a delay in access to medical care or a history of BCC that was inadequately managed.^9,10 During the COVID-19 pandemic, patient access to health care was substantially impacted during lockdowns. As in our patient, skin neoplasms and other medical conditions may present in later stages due to medical neglect.^11,12 Metastasis is rare, even in advanced BCCs. A review of the literature from 1984 estimated that the incidence of metastasis of BCCs is 1 in 1000 to 35,000. Metastasis portends a poor prognosis with a median overall survival of 8 to 14 months.¹³ An updated review in 2013 found similar outcomes.¹⁴

The choice of management for BCCs depends on the risk for recurrence as well as individual patient factors. Characteristics such as tumor size, location, histology, whether it is a primary or recurrent lesion, and the presence of chronic skin disease determine the recurrence rate.¹⁵ The management of advanced BCCs often requires a multidisciplinary approach, as these neoplasms may not be amenable to local therapy without causing substantial morbidity. Mohs micrographic surgery is the treatment of choice for BCCs at high risk for recurrence.¹⁶ Standard surgical excision with postoperative margin assessment is acceptable when Mohs micrographic surgery is not available.¹⁷ Radiation therapy is an alternative for patients who are not candidates for surgery.¹⁸

Recently, improved understanding of the molecular pathogenesis of BCCs has led to the development of novel systemic therapies. The Hedgehog signaling pathway has been found to play a critical role in the development of most BCCs.¹⁹ Vismodegib and sonidegib are small-molecule inhibitors of the Hedgehog signaling pathway approved for the treatment of locally advanced and metastatic BCCs that are not amenable to surgery or radiation. Approximately 50% of advanced BCCs respond to these therapies; however, long-term treatment may be limited by intolerable side effects and the development of resistance.²⁰ Basal cell carcinomas that spread to lymph nodes or distant sites are treated with traditional systemic therapy. Historically, conventional cytotoxic chemotherapies, such as platinum-containing regimens, were employed with limited benefit and notable morbidity.²¹

The differential diagnosis for our patient included several other cutaneous neoplasms. Squamous cell carcinoma is the second most common type of skin cancer. Similar to BCC, it can reach a substantial size if left untreated. Risk factors include chronic inflammation, exposure to radiation or chemical carcinogens, burns, human papillomavirus, and other chronic infections. Giant squamous cell carcinomas have high malignant potential and require imaging to assess the extent of invasion and for metastasis. Surgery typically is necessary for both staging and treatment. Adjuvant therapy also may be necessary.^22,23

Internal malignant neoplasms rarely present as cutaneous metastases. Breast cancer, melanoma, and cancers of the upper respiratory tract most frequently metastasize to the skin. Although colorectal cancer (CRC) rarely metastasizes to the skin, it is an important cause of cutaneous metastasis due to its high incidence in the general population. When it does spread to the skin, CRC preferentially affects the abdominal wall. Lesions typically resemble the primary tumor but may appear anaplastic. The occurrence of cutaneous metastasis suggests latestage disease and carries a poor prognosis.²⁴

Merkel cell carcinoma and melanoma are aggressive skin cancers with high mortality rates. The former is rarer but more lethal. Merkel cell carcinomas typically occur in elderly white men on sun-exposed areas of the skin. Tumors present as asymptomatic, rapidly expanding, blue-red, firm nodules. Immunosuppression and UV light exposure are notable risk factors.²⁵ Of the 4 major subtypes of cutaneous melanoma, superficial spreading is the most common, followed by nodular, lentigo maligna, and acral lentiginous.²⁶ Superficial spreading melanoma characteristically presents as an expanding asymmetric macule or thin plaque with irregular borders and variation in size and color (black, brown, or red). Nodular melanoma usually presents as symmetric in shape and color (amelanotic, black, or brown). Early recognition by both the patient and clinician is essential in preventing tumor growth and progression.²⁷

Our patient’s presentation was highly concerning for cutaneous metastasis given her history of CRC. Furthermore, the finding of severe anemia was atypical for skin cancer and more characteristic of the prior malignancy. Imaging revealed a locally confined mass with no evidence of extension, lymph node involvement, or additional lesions. The diagnosis was clinched with histopathologic examination.

The Diagnosis: Basal Cell Carcinoma

Histopathology was consistent with fungating basal cell carcinoma (BCC). The nodules were comprised of syncytial basaloid cells with high nuclear to cytoplasmic ratios, numerous mitotic figures, fibromyxoid stroma, and peripheral nuclear palisading (Figure). Fortunately, no perineural or lymphovascular invasion was identified, and the margins of the specimen were negative. Despite the high-risk nature of giant BCC, the mass was solitary without notable local invasion, leaving it amendable to surgery. On follow-up, the patient has remained recurrence free, and her hemoglobin level has since stabilized.

Skin cancer is the most common malignancy worldwide, and BCC accounts for more than 80% of nonmelanoma skin cancers in the United States. The incidence is on the rise due to the aging population and increasing cumulative skin exposure.¹ Risk factors include both individual physical characteristics and environmental exposures. Individuals with lighter skin tones, red and blonde hair, and blue and green eyes are at an increased risk.² UV radiation exposure is the most important cause of BCC.³ Chronic immunosuppression and exposure to arsenic, ionizing radiation, and psoralen plus UVA radiation also have been linked to the development of BCC.^4-6 Basal cell carcinomas most commonly arise on sun-exposed areas such as the face, though more than 10% of cases appear on the trunk.⁷ Lesions characteristically remain localized, and growth rate is variable; however, when left untreated, BCCs have the potential to become locally destructive and difficult to treat.

Advanced BCCs are tumors that penetrate deeply into the skin. They often are not amenable to traditional therapy and/ or metastasize. Those that grow to a diameter greater than 5 cm, as in our patient, are known as giant BCCs. Only 0.5% to 1% of BCCs are giant BCCs⁸ ; they typically are more aggressive in nature with higher rates of local recurrence and metastasis. Individuals who develop giant BCCs either have had a delay in access to medical care or a history of BCC that was inadequately managed.^9,10 During the COVID-19 pandemic, patient access to health care was substantially impacted during lockdowns. As in our patient, skin neoplasms and other medical conditions may present in later stages due to medical neglect.^11,12 Metastasis is rare, even in advanced BCCs. A review of the literature from 1984 estimated that the incidence of metastasis of BCCs is 1 in 1000 to 35,000. Metastasis portends a poor prognosis with a median overall survival of 8 to 14 months.¹³ An updated review in 2013 found similar outcomes.¹⁴

The choice of management for BCCs depends on the risk for recurrence as well as individual patient factors. Characteristics such as tumor size, location, histology, whether it is a primary or recurrent lesion, and the presence of chronic skin disease determine the recurrence rate.¹⁵ The management of advanced BCCs often requires a multidisciplinary approach, as these neoplasms may not be amenable to local therapy without causing substantial morbidity. Mohs micrographic surgery is the treatment of choice for BCCs at high risk for recurrence.¹⁶ Standard surgical excision with postoperative margin assessment is acceptable when Mohs micrographic surgery is not available.¹⁷ Radiation therapy is an alternative for patients who are not candidates for surgery.¹⁸

Recently, improved understanding of the molecular pathogenesis of BCCs has led to the development of novel systemic therapies. The Hedgehog signaling pathway has been found to play a critical role in the development of most BCCs.¹⁹ Vismodegib and sonidegib are small-molecule inhibitors of the Hedgehog signaling pathway approved for the treatment of locally advanced and metastatic BCCs that are not amenable to surgery or radiation. Approximately 50% of advanced BCCs respond to these therapies; however, long-term treatment may be limited by intolerable side effects and the development of resistance.²⁰ Basal cell carcinomas that spread to lymph nodes or distant sites are treated with traditional systemic therapy. Historically, conventional cytotoxic chemotherapies, such as platinum-containing regimens, were employed with limited benefit and notable morbidity.²¹

The differential diagnosis for our patient included several other cutaneous neoplasms. Squamous cell carcinoma is the second most common type of skin cancer. Similar to BCC, it can reach a substantial size if left untreated. Risk factors include chronic inflammation, exposure to radiation or chemical carcinogens, burns, human papillomavirus, and other chronic infections. Giant squamous cell carcinomas have high malignant potential and require imaging to assess the extent of invasion and for metastasis. Surgery typically is necessary for both staging and treatment. Adjuvant therapy also may be necessary.^22,23

Internal malignant neoplasms rarely present as cutaneous metastases. Breast cancer, melanoma, and cancers of the upper respiratory tract most frequently metastasize to the skin. Although colorectal cancer (CRC) rarely metastasizes to the skin, it is an important cause of cutaneous metastasis due to its high incidence in the general population. When it does spread to the skin, CRC preferentially affects the abdominal wall. Lesions typically resemble the primary tumor but may appear anaplastic. The occurrence of cutaneous metastasis suggests latestage disease and carries a poor prognosis.²⁴

Merkel cell carcinoma and melanoma are aggressive skin cancers with high mortality rates. The former is rarer but more lethal. Merkel cell carcinomas typically occur in elderly white men on sun-exposed areas of the skin. Tumors present as asymptomatic, rapidly expanding, blue-red, firm nodules. Immunosuppression and UV light exposure are notable risk factors.²⁵ Of the 4 major subtypes of cutaneous melanoma, superficial spreading is the most common, followed by nodular, lentigo maligna, and acral lentiginous.²⁶ Superficial spreading melanoma characteristically presents as an expanding asymmetric macule or thin plaque with irregular borders and variation in size and color (black, brown, or red). Nodular melanoma usually presents as symmetric in shape and color (amelanotic, black, or brown). Early recognition by both the patient and clinician is essential in preventing tumor growth and progression.²⁷

Our patient’s presentation was highly concerning for cutaneous metastasis given her history of CRC. Furthermore, the finding of severe anemia was atypical for skin cancer and more characteristic of the prior malignancy. Imaging revealed a locally confined mass with no evidence of extension, lymph node involvement, or additional lesions. The diagnosis was clinched with histopathologic examination.

The Diagnosis: Kaposi Sarcoma

A punch biopsy of a lesion on the right side of the back revealed a diffuse, poorly circumscribed, spindle cell neoplasm of the papillary and reticular dermis with associated vascular and pseudovascular spaces distended by erythrocytes (Figure 1). Immunostaining was positive for human herpesvirus 8 (HHV-8)(Figure 2), ETS-related gene, CD31, and CD34 and negative for pan cytokeratin, confirming the diagnosis of Kaposi sarcoma (KS). Bacterial, fungal, and mycobacterial tissue cultures were negative. The patient was tested for HIV and referred to infectious disease and oncology. He subsequently was found to have HIV with a viral load greater than 1 million copies. He was started on antiretroviral therapy and Pneumocystis jirovecii pneumonia prophylaxis. Computed tomography of the chest, abdomen, and pelvis showed bilateral, multifocal, perihilar, flame-shaped consolidations suggestive of KS. The patient later disclosed having an intermittent dry cough of more than a year’s duration with occasional bright red blood per rectum after bowel movements. After workup, the patient was found to have cytomegalovirus esophagitis/gastritis and candidal esophagitis that were treated with valganciclovir and fluconazole, respectively.

Kaposi sarcoma is an angioproliferative, AIDSdefining disease associated with HHV-8. There are 4 types of KS as defined by the populations they affect. AIDS-associated KS occurs in individuals with HIV, as seen in our patient. It often is accompanied by extensive mucocutaneous and visceral lesions, as well as systemic symptoms such as fever, weight loss, and diarrhea.¹ Classic KS is a variant that presents in older men of Mediterranean, Eastern European, and South American descent. Cutaneous lesions typically are distributed on the lower extremities.^2,3 Endemic (African) KS is seen in HIV-negative children and young adults in equatorial Africa. It most commonly affects the lower extremities or lymph nodes and usually follows a more aggressive course.² Lastly, iatrogenic KS is associated with immunosuppressive medications or conditions, such as organ transplantation, chemotherapy, and rheumatologic disorders.^3,4

Kaposi sarcoma commonly presents as violaceous or dark red macules, patches, papules, plaques, and nodules on various parts of the body (Figure 3). Lesions typically begin as macules and progress into plaques or nodules. Our patient presented as a deceptively healthy young man with lesions at various stages of development. In addition to the skin and oral mucosa, the lungs, lymph nodes, and gastrointestinal tract commonly are involved in AIDS-associated KS.⁵ Patients may experience symptoms of internal involvement, including bleeding, hematochezia, odynophagia, or dyspnea.

The differential diagnosis includes conditions that can mimic KS, including bacillary angiomatosis, angioinvasive fungal disease, sarcoid, and other malignancies. A skin biopsy is the gold standard for definitive diagnosis of KS. Histopathology shows a vascular proliferation in the dermis and spindle cell proliferation.⁶ Kaposi sarcoma stains positively for factor VIII–related antigen, CD31, and CD34.² Additionally, staining for HHV-8 gene products, such as latency-associated nuclear antigen 1, is helpful in differentiating KS from other conditions.⁷

In HIV-associated KS, the mainstay of treatment is initiation of highly active antiretroviral therapy. Typically, as the CD4 count rises with treatment, the tumor burden classic KS, effective treatment options include recurrent cryotherapy or intralesional chemotherapeutics, such as vincristine, for localized lesions; for widespread disease, pegylated liposomal doxorubicin or radiation have been found to be effective options. Lastly, for patients with iatrogenic KS, reducing immunosuppressive medications is a reasonable first step in management. If this does not yield adequate improvement, transitioning from calcineurin inhibitors (eg, cyclosporine) to proliferation signal inhibitors (eg, sirolimus) may lead to resolution.⁷

The Diagnosis: Kaposi Sarcoma

A punch biopsy of a lesion on the right side of the back revealed a diffuse, poorly circumscribed, spindle cell neoplasm of the papillary and reticular dermis with associated vascular and pseudovascular spaces distended by erythrocytes (Figure 1). Immunostaining was positive for human herpesvirus 8 (HHV-8)(Figure 2), ETS-related gene, CD31, and CD34 and negative for pan cytokeratin, confirming the diagnosis of Kaposi sarcoma (KS). Bacterial, fungal, and mycobacterial tissue cultures were negative. The patient was tested for HIV and referred to infectious disease and oncology. He subsequently was found to have HIV with a viral load greater than 1 million copies. He was started on antiretroviral therapy and Pneumocystis jirovecii pneumonia prophylaxis. Computed tomography of the chest, abdomen, and pelvis showed bilateral, multifocal, perihilar, flame-shaped consolidations suggestive of KS. The patient later disclosed having an intermittent dry cough of more than a year’s duration with occasional bright red blood per rectum after bowel movements. After workup, the patient was found to have cytomegalovirus esophagitis/gastritis and candidal esophagitis that were treated with valganciclovir and fluconazole, respectively.

Kaposi sarcoma is an angioproliferative, AIDSdefining disease associated with HHV-8. There are 4 types of KS as defined by the populations they affect. AIDS-associated KS occurs in individuals with HIV, as seen in our patient. It often is accompanied by extensive mucocutaneous and visceral lesions, as well as systemic symptoms such as fever, weight loss, and diarrhea.¹ Classic KS is a variant that presents in older men of Mediterranean, Eastern European, and South American descent. Cutaneous lesions typically are distributed on the lower extremities.^2,3 Endemic (African) KS is seen in HIV-negative children and young adults in equatorial Africa. It most commonly affects the lower extremities or lymph nodes and usually follows a more aggressive course.² Lastly, iatrogenic KS is associated with immunosuppressive medications or conditions, such as organ transplantation, chemotherapy, and rheumatologic disorders.^3,4

Kaposi sarcoma commonly presents as violaceous or dark red macules, patches, papules, plaques, and nodules on various parts of the body (Figure 3). Lesions typically begin as macules and progress into plaques or nodules. Our patient presented as a deceptively healthy young man with lesions at various stages of development. In addition to the skin and oral mucosa, the lungs, lymph nodes, and gastrointestinal tract commonly are involved in AIDS-associated KS.⁵ Patients may experience symptoms of internal involvement, including bleeding, hematochezia, odynophagia, or dyspnea.

The differential diagnosis includes conditions that can mimic KS, including bacillary angiomatosis, angioinvasive fungal disease, sarcoid, and other malignancies. A skin biopsy is the gold standard for definitive diagnosis of KS. Histopathology shows a vascular proliferation in the dermis and spindle cell proliferation.⁶ Kaposi sarcoma stains positively for factor VIII–related antigen, CD31, and CD34.² Additionally, staining for HHV-8 gene products, such as latency-associated nuclear antigen 1, is helpful in differentiating KS from other conditions.⁷

In HIV-associated KS, the mainstay of treatment is initiation of highly active antiretroviral therapy. Typically, as the CD4 count rises with treatment, the tumor burden classic KS, effective treatment options include recurrent cryotherapy or intralesional chemotherapeutics, such as vincristine, for localized lesions; for widespread disease, pegylated liposomal doxorubicin or radiation have been found to be effective options. Lastly, for patients with iatrogenic KS, reducing immunosuppressive medications is a reasonable first step in management. If this does not yield adequate improvement, transitioning from calcineurin inhibitors (eg, cyclosporine) to proliferation signal inhibitors (eg, sirolimus) may lead to resolution.⁷

The Diagnosis: Kaposi Sarcoma

A punch biopsy of a lesion on the right side of the back revealed a diffuse, poorly circumscribed, spindle cell neoplasm of the papillary and reticular dermis with associated vascular and pseudovascular spaces distended by erythrocytes (Figure 1). Immunostaining was positive for human herpesvirus 8 (HHV-8)(Figure 2), ETS-related gene, CD31, and CD34 and negative for pan cytokeratin, confirming the diagnosis of Kaposi sarcoma (KS). Bacterial, fungal, and mycobacterial tissue cultures were negative. The patient was tested for HIV and referred to infectious disease and oncology. He subsequently was found to have HIV with a viral load greater than 1 million copies. He was started on antiretroviral therapy and Pneumocystis jirovecii pneumonia prophylaxis. Computed tomography of the chest, abdomen, and pelvis showed bilateral, multifocal, perihilar, flame-shaped consolidations suggestive of KS. The patient later disclosed having an intermittent dry cough of more than a year’s duration with occasional bright red blood per rectum after bowel movements. After workup, the patient was found to have cytomegalovirus esophagitis/gastritis and candidal esophagitis that were treated with valganciclovir and fluconazole, respectively.

Kaposi sarcoma is an angioproliferative, AIDSdefining disease associated with HHV-8. There are 4 types of KS as defined by the populations they affect. AIDS-associated KS occurs in individuals with HIV, as seen in our patient. It often is accompanied by extensive mucocutaneous and visceral lesions, as well as systemic symptoms such as fever, weight loss, and diarrhea.¹ Classic KS is a variant that presents in older men of Mediterranean, Eastern European, and South American descent. Cutaneous lesions typically are distributed on the lower extremities.^2,3 Endemic (African) KS is seen in HIV-negative children and young adults in equatorial Africa. It most commonly affects the lower extremities or lymph nodes and usually follows a more aggressive course.² Lastly, iatrogenic KS is associated with immunosuppressive medications or conditions, such as organ transplantation, chemotherapy, and rheumatologic disorders.^3,4

Kaposi sarcoma commonly presents as violaceous or dark red macules, patches, papules, plaques, and nodules on various parts of the body (Figure 3). Lesions typically begin as macules and progress into plaques or nodules. Our patient presented as a deceptively healthy young man with lesions at various stages of development. In addition to the skin and oral mucosa, the lungs, lymph nodes, and gastrointestinal tract commonly are involved in AIDS-associated KS.⁵ Patients may experience symptoms of internal involvement, including bleeding, hematochezia, odynophagia, or dyspnea.

The differential diagnosis includes conditions that can mimic KS, including bacillary angiomatosis, angioinvasive fungal disease, sarcoid, and other malignancies. A skin biopsy is the gold standard for definitive diagnosis of KS. Histopathology shows a vascular proliferation in the dermis and spindle cell proliferation.⁶ Kaposi sarcoma stains positively for factor VIII–related antigen, CD31, and CD34.² Additionally, staining for HHV-8 gene products, such as latency-associated nuclear antigen 1, is helpful in differentiating KS from other conditions.⁷

In HIV-associated KS, the mainstay of treatment is initiation of highly active antiretroviral therapy. Typically, as the CD4 count rises with treatment, the tumor burden classic KS, effective treatment options include recurrent cryotherapy or intralesional chemotherapeutics, such as vincristine, for localized lesions; for widespread disease, pegylated liposomal doxorubicin or radiation have been found to be effective options. Lastly, for patients with iatrogenic KS, reducing immunosuppressive medications is a reasonable first step in management. If this does not yield adequate improvement, transitioning from calcineurin inhibitors (eg, cyclosporine) to proliferation signal inhibitors (eg, sirolimus) may lead to resolution.⁷

The Diagnosis: Dermatofibrosarcoma Protuberans

Histopathologic examination showed a dermal tumor composed of spindle cells in a storiform arrangement (Figure 1). Immunohistochemistry demonstrated positive CD34 staining of the tumoral cells (Figure 2). Clinical review, histopathologic examination, and immunohistochemistry confirmed a diagnosis of dermatofibrosarcoma protuberans (DFSP). The patient underwent Mohs micrographic surgery (MMS) with clear margins after 3 stages, followed by repair with a rotation flap. No evidence of recurrence was found at 4-year follow-up.

Dermatofibrosarcoma protuberans is a rare low-grade sarcoma of fibroblast origin with an annual incidence of 0.8 to 5 cases per million individuals.¹ It typically presents in patients aged 30 to 50 years on the trunk, scalp, or proximal extremities as an asymptomatic, flesh-colored, erythematous or brown, indurated plaque or nodule.² Due to its variable presentation, these lesions often may be misdiagnosed as lipomas or epidermoid cysts, preventing proper targeted treatment. Therefore, suspicious enlarging indurated nodules require a lower threshold for biopsy.¹

A definitive diagnosis of DFSP is achieved after a biopsy and histopathologic evaluation. Hematoxylin and eosin staining typically shows diffuse infiltration of the dermis and the subcutaneous fat by densely packed, cytologic, relatively uniform, spindle-shaped tumor cells arranged in a characteristic storiform shape. Tumor cells are spread along the septae of the subcutaneous fatty tissue.³ Immunohistochemistry is characterized by positive CD34 and negative factor XIIIa, with rare exceptions.

The differential diagnosis includes lipoma, epidermoid cyst, plexiform fibrohistiocytic tumor, and malignant peripheral nerve sheath tumor.³ Positive CD34 immunostaining, negative S-100 staining, and a storiform pattern of spindle cells can assist in differentiating DFSP from these possible differential diagnoses; lesions of these other entities are characterized by different pathologic findings. Lipomas are composed of fat tissue, epidermoid cysts have epithelial-lined cysts filled with keratin, plexiform fibrohistiocytic tumors have plexiform rays of fibrous tissue extending into fat with negative CD34 staining, and malignant peripheral nerve sheath tumors have fleshy variegated masses involving the peripheral nerve trunks with partial S-100 staining.^4-7 Additional evaluation to confirm DFSP can be accomplished by analysis of tumor samples by fluorescence in situ hybridization or reverse transcriptase–polymerase chain reaction to detect chromosomal translocations and fusion gene transcripts, as chromosomal translocations may be found in more than 90% of cases.³

Early diagnosis of DFSP is beneficial, as it can help prevent recurrence as well as metastasis. Studies have attempted to document the risk for recurrence as well as metastasis based on characteristic features and treatment strategies of DFSP. In a study of 186 patients, 3 had metastatic disease to the lungs, the most common site of metastasis.⁸ These 3 patients had fibrosarcomatous transformation within DFSP, emphasizing the importance of detailing this finding early in the diagnosis, as it was characterized by a higher degree of cellularity, cytologic atypia, mitotic activity, and negative CD34 immunostaining.⁹ In patients with suspected metastasis, lymph node ultrasonography, chest radiography, and computed tomography may be utilized.³

When treating DFSP, the goal is complete removal of the tumor with clear margins. Mohs micrographic surgery, modified MMS, and wide local excision (WLE) with 2- to 4-cm margins are appropriate treatment options, though MMS is the treatment of choice. A study comparing MMS and WLE demonstrated 3% and 30.8% recurrence rates, respectively.⁸ In MMS, complete margin evaluation on microscopy is performed after each stage to ensure negative surgical margins. The presence of positive surgical margins elicits continued resection until the margins are clear.^10,11

Other treatment modalities may be considered for patients with DFSP. Molecular therapy with imatinib, an oral tyrosine kinase inhibitor targeting platelet-derived growth factor–regulated expression, can be utilized for inoperable tumors; however, additional clinical trials are required to ensure efficacy.³ Surgical removal of the possible remaining tumor is still recommended after molecular therapy. Radiotherapy is an additional method of treatment that may be used for inoperable tumors.³

Dermatofibrosarcoma protuberans is a rare lowgrade sarcoma of fibroblast origin that typically does not metastasize but often has notable subclinical extension and recurrence. Differentiating DFSP from other tumors often may be difficult. A protuberant, flesh-colored, slowgrowing, and asymptomatic lesion often may be confused with lipomas or epidermoid cysts; therefore, biopsies with immunohistostaining for suspicious lesions is required.¹² Mohs micrographic surgery has evolved as the treatment of choice for this tumor, though WLE and new targeted molecular therapies still are considered. Proper diagnosis and treatment of DFSP is paramount in preventing future morbidity.

The Diagnosis: Dermatofibrosarcoma Protuberans

Histopathologic examination showed a dermal tumor composed of spindle cells in a storiform arrangement (Figure 1). Immunohistochemistry demonstrated positive CD34 staining of the tumoral cells (Figure 2). Clinical review, histopathologic examination, and immunohistochemistry confirmed a diagnosis of dermatofibrosarcoma protuberans (DFSP). The patient underwent Mohs micrographic surgery (MMS) with clear margins after 3 stages, followed by repair with a rotation flap. No evidence of recurrence was found at 4-year follow-up.

Dermatofibrosarcoma protuberans is a rare low-grade sarcoma of fibroblast origin with an annual incidence of 0.8 to 5 cases per million individuals.¹ It typically presents in patients aged 30 to 50 years on the trunk, scalp, or proximal extremities as an asymptomatic, flesh-colored, erythematous or brown, indurated plaque or nodule.² Due to its variable presentation, these lesions often may be misdiagnosed as lipomas or epidermoid cysts, preventing proper targeted treatment. Therefore, suspicious enlarging indurated nodules require a lower threshold for biopsy.¹

A definitive diagnosis of DFSP is achieved after a biopsy and histopathologic evaluation. Hematoxylin and eosin staining typically shows diffuse infiltration of the dermis and the subcutaneous fat by densely packed, cytologic, relatively uniform, spindle-shaped tumor cells arranged in a characteristic storiform shape. Tumor cells are spread along the septae of the subcutaneous fatty tissue.³ Immunohistochemistry is characterized by positive CD34 and negative factor XIIIa, with rare exceptions.

The differential diagnosis includes lipoma, epidermoid cyst, plexiform fibrohistiocytic tumor, and malignant peripheral nerve sheath tumor.³ Positive CD34 immunostaining, negative S-100 staining, and a storiform pattern of spindle cells can assist in differentiating DFSP from these possible differential diagnoses; lesions of these other entities are characterized by different pathologic findings. Lipomas are composed of fat tissue, epidermoid cysts have epithelial-lined cysts filled with keratin, plexiform fibrohistiocytic tumors have plexiform rays of fibrous tissue extending into fat with negative CD34 staining, and malignant peripheral nerve sheath tumors have fleshy variegated masses involving the peripheral nerve trunks with partial S-100 staining.^4-7 Additional evaluation to confirm DFSP can be accomplished by analysis of tumor samples by fluorescence in situ hybridization or reverse transcriptase–polymerase chain reaction to detect chromosomal translocations and fusion gene transcripts, as chromosomal translocations may be found in more than 90% of cases.³

Early diagnosis of DFSP is beneficial, as it can help prevent recurrence as well as metastasis. Studies have attempted to document the risk for recurrence as well as metastasis based on characteristic features and treatment strategies of DFSP. In a study of 186 patients, 3 had metastatic disease to the lungs, the most common site of metastasis.⁸ These 3 patients had fibrosarcomatous transformation within DFSP, emphasizing the importance of detailing this finding early in the diagnosis, as it was characterized by a higher degree of cellularity, cytologic atypia, mitotic activity, and negative CD34 immunostaining.⁹ In patients with suspected metastasis, lymph node ultrasonography, chest radiography, and computed tomography may be utilized.³

When treating DFSP, the goal is complete removal of the tumor with clear margins. Mohs micrographic surgery, modified MMS, and wide local excision (WLE) with 2- to 4-cm margins are appropriate treatment options, though MMS is the treatment of choice. A study comparing MMS and WLE demonstrated 3% and 30.8% recurrence rates, respectively.⁸ In MMS, complete margin evaluation on microscopy is performed after each stage to ensure negative surgical margins. The presence of positive surgical margins elicits continued resection until the margins are clear.^10,11

Other treatment modalities may be considered for patients with DFSP. Molecular therapy with imatinib, an oral tyrosine kinase inhibitor targeting platelet-derived growth factor–regulated expression, can be utilized for inoperable tumors; however, additional clinical trials are required to ensure efficacy.³ Surgical removal of the possible remaining tumor is still recommended after molecular therapy. Radiotherapy is an additional method of treatment that may be used for inoperable tumors.³

Dermatofibrosarcoma protuberans is a rare lowgrade sarcoma of fibroblast origin that typically does not metastasize but often has notable subclinical extension and recurrence. Differentiating DFSP from other tumors often may be difficult. A protuberant, flesh-colored, slowgrowing, and asymptomatic lesion often may be confused with lipomas or epidermoid cysts; therefore, biopsies with immunohistostaining for suspicious lesions is required.¹² Mohs micrographic surgery has evolved as the treatment of choice for this tumor, though WLE and new targeted molecular therapies still are considered. Proper diagnosis and treatment of DFSP is paramount in preventing future morbidity.

The Diagnosis: Dermatofibrosarcoma Protuberans

Histopathologic examination showed a dermal tumor composed of spindle cells in a storiform arrangement (Figure 1). Immunohistochemistry demonstrated positive CD34 staining of the tumoral cells (Figure 2). Clinical review, histopathologic examination, and immunohistochemistry confirmed a diagnosis of dermatofibrosarcoma protuberans (DFSP). The patient underwent Mohs micrographic surgery (MMS) with clear margins after 3 stages, followed by repair with a rotation flap. No evidence of recurrence was found at 4-year follow-up.

Dermatofibrosarcoma protuberans is a rare low-grade sarcoma of fibroblast origin with an annual incidence of 0.8 to 5 cases per million individuals.¹ It typically presents in patients aged 30 to 50 years on the trunk, scalp, or proximal extremities as an asymptomatic, flesh-colored, erythematous or brown, indurated plaque or nodule.² Due to its variable presentation, these lesions often may be misdiagnosed as lipomas or epidermoid cysts, preventing proper targeted treatment. Therefore, suspicious enlarging indurated nodules require a lower threshold for biopsy.¹

A definitive diagnosis of DFSP is achieved after a biopsy and histopathologic evaluation. Hematoxylin and eosin staining typically shows diffuse infiltration of the dermis and the subcutaneous fat by densely packed, cytologic, relatively uniform, spindle-shaped tumor cells arranged in a characteristic storiform shape. Tumor cells are spread along the septae of the subcutaneous fatty tissue.³ Immunohistochemistry is characterized by positive CD34 and negative factor XIIIa, with rare exceptions.

The differential diagnosis includes lipoma, epidermoid cyst, plexiform fibrohistiocytic tumor, and malignant peripheral nerve sheath tumor.³ Positive CD34 immunostaining, negative S-100 staining, and a storiform pattern of spindle cells can assist in differentiating DFSP from these possible differential diagnoses; lesions of these other entities are characterized by different pathologic findings. Lipomas are composed of fat tissue, epidermoid cysts have epithelial-lined cysts filled with keratin, plexiform fibrohistiocytic tumors have plexiform rays of fibrous tissue extending into fat with negative CD34 staining, and malignant peripheral nerve sheath tumors have fleshy variegated masses involving the peripheral nerve trunks with partial S-100 staining.^4-7 Additional evaluation to confirm DFSP can be accomplished by analysis of tumor samples by fluorescence in situ hybridization or reverse transcriptase–polymerase chain reaction to detect chromosomal translocations and fusion gene transcripts, as chromosomal translocations may be found in more than 90% of cases.³

Early diagnosis of DFSP is beneficial, as it can help prevent recurrence as well as metastasis. Studies have attempted to document the risk for recurrence as well as metastasis based on characteristic features and treatment strategies of DFSP. In a study of 186 patients, 3 had metastatic disease to the lungs, the most common site of metastasis.⁸ These 3 patients had fibrosarcomatous transformation within DFSP, emphasizing the importance of detailing this finding early in the diagnosis, as it was characterized by a higher degree of cellularity, cytologic atypia, mitotic activity, and negative CD34 immunostaining.⁹ In patients with suspected metastasis, lymph node ultrasonography, chest radiography, and computed tomography may be utilized.³

When treating DFSP, the goal is complete removal of the tumor with clear margins. Mohs micrographic surgery, modified MMS, and wide local excision (WLE) with 2- to 4-cm margins are appropriate treatment options, though MMS is the treatment of choice. A study comparing MMS and WLE demonstrated 3% and 30.8% recurrence rates, respectively.⁸ In MMS, complete margin evaluation on microscopy is performed after each stage to ensure negative surgical margins. The presence of positive surgical margins elicits continued resection until the margins are clear.^10,11

Other treatment modalities may be considered for patients with DFSP. Molecular therapy with imatinib, an oral tyrosine kinase inhibitor targeting platelet-derived growth factor–regulated expression, can be utilized for inoperable tumors; however, additional clinical trials are required to ensure efficacy.³ Surgical removal of the possible remaining tumor is still recommended after molecular therapy. Radiotherapy is an additional method of treatment that may be used for inoperable tumors.³

Dermatofibrosarcoma protuberans is a rare lowgrade sarcoma of fibroblast origin that typically does not metastasize but often has notable subclinical extension and recurrence. Differentiating DFSP from other tumors often may be difficult. A protuberant, flesh-colored, slowgrowing, and asymptomatic lesion often may be confused with lipomas or epidermoid cysts; therefore, biopsies with immunohistostaining for suspicious lesions is required.¹² Mohs micrographic surgery has evolved as the treatment of choice for this tumor, though WLE and new targeted molecular therapies still are considered. Proper diagnosis and treatment of DFSP is paramount in preventing future morbidity.

People with cancer are often desperate to know what caused their disease. Was it something they did? Something they could have prevented?

vitanovski/Thinkstock.com

In a recent analysis, experts estimated that about 40% of cancers can be explained by known, often modifiable risk factors. Smoking and obesity represent the primary drivers, though a host of other factors – germline mutations, alcohol, infections, or environmental pollutants like asbestos – contribute to cancer risk as well.

But what about the remaining 60% of cancers?

The study suggests that, although many of these cases likely have an underlying lifestyle or environmental component, experts still do not fully understand their origin story. And a small but significant number may simply be caused by chance.

Here’s what experts suspect those missing causes might be, and why they can be so difficult to confirm.
 

Possibility 1: Known risk factors contribute more than we realize

For certain factors, a straight line can be drawn to cancer.

Take smoking, for instance. Decades of research have helped scientists clearly delineate tobacco’s carcinogenic effects. Researchers have pinpointed a unique set of mutations in the tumors of smokers that can be seen when cells grown in a dish are exposed to the carcinogens present in tobacco.

In addition, experts have been able to collect robust data from epidemiologic studies on smoking prevalence as well as associated cancer risks and deaths, in large part because an individual’s lifetime tobacco exposure is fairly easy to measure.

“The evidence for smoking is incredibly consistent,” Paul Brennan, PhD, a cancer epidemiologist at the World Health Organization’s International Agency for Research on Cancer, said in an interview.

For other known risk factors, such as obesity and air pollution, many more questions than answers remain.

Because of the limitations in how such factors are measured, we are likely downplaying their effects, said Richard Martin, PhD, a professor of clinical epidemiology at the University of Bristol (England).

Take obesity. Excess body weight is associated with an increased risk of at least 13 cancers. Although risk estimates vary by study and cancer type, according to a global snapshot from 2012, being overweight or obese accounted for about 4% of all cancers worldwide – 1% in low-income countries and as high as 8% in high-income countries.

However, Dr. Brennan believes “we have underestimated the effect of obesity [on cancer].”

A key reason, he said, is most studies use body mass index to determine whether someone is overweight or obese, but BMI is a poor measure of body fat. BMI does not differentiate between fat and muscle, which means two people with the same height and weight can have the same BMI, even if one is an athlete who eats lean meats and vegetables while the other lives a sedentary life and consumes large quantities of processed foods and alcohol.

On top of that, studies often only calculate a person’s BMI once, and a single measurement can’t tell you how a person’s weight has fluctuated in recent years or across different stages of their life. However, recent analyses suggest that obesity status over time may be more relevant to cancer risk than one-off measures.

In addition, many studies now suggest that alterations to our gut microbes and high blood insulin level – often seen in people who are overweight or obese – may increase the risk of cancer and speed the growth of tumors.

When these additional factors are considered, the impact of excess body fat may ultimately play a much more significant role in cancer risk. In fact, according to Dr. Brennan, “if we estimate [the effects of obesity] properly, it might at some point become the main cause of cancer.”
 

Possibility 2: Environmental or lifestyle factors remain under the radar

Researchers have linked many substances we consume or are exposed to in our daily lives – air pollution, toxins from industrial waste, and highly processed foods – to cancer. But the extent or contribution of potential carcinogens in our surroundings, particularly those found almost everywhere at low levels, is still largely unknown.

One simple reason is the effects of many of these substances remain difficult to assess. For instance, it is much harder to study the impact of pollutants found in food or water, in which a given population will share similar exposure levels versus tobacco, where it is possible to compare a person who smokes a pack of cigarettes a day with a person who does not smoke.

“If you’ve got exposures that are ubiquitous, it can be difficult to discern their [individual] roles,” Dr. Martin said. “There are many causes that we [likely] don’t really know because everyone has been exposed.”

On the flip side, some carcinogenic substances that people encounter for limited periods might be missed if studies are not performed at the time of exposure.

“What’s in the body at age 40 may not reflect what you were exposed at age 5-10 on the playground or soccer field,” said Graham Colditz, MD, PhD, an epidemiologist and public health expert at Washington University, St. Louis. “The technology keeps changing so we can get better measures of what you’ve got exposure to today, but how that relates to 5, 10, 15 years ago is probably very variable.”

In addition, researchers have found that many carcinogens do not cause specific mutations in a cell’s DNA; rather, studies suggest that most carcinogens lead to cancer-promoting changes in cells, such as inflammation.

“We need to think of how potential carcinogens are causing cancer,” Dr. Brennan said. Instead of provoking mutations, potential carcinogens may use a “whole other kind of pathway.” When, for instance, inflammation becomes chronic, it may spur a cascade of events that ultimately leads to cancer.

Finally, not much is known about what causes cancers in low- and middle-income countries. Most of the research to date has been in high-income countries, such the United States, Australia, and parts of Europe.

“There’s a real lack of robust epidemiological studies in other parts of the world, Latin America, Africa, parts of Asia,” Marc Gunter, PhD, a molecular epidemiologist at the IARC, told this news organization.
 

Possibility 3: Some cancers occur by chance

When it comes to cancer risk, an element of chance may be at play. Cancer can occur in individuals who have very little exposure to known carcinogens or have no family history of cancer.

“We all know there are people who get cancer who eat very healthy diets, are never overweight, and never smoke,” Dr. Gunter said. “Then there are people on the other end of the extreme who don’t get cancer.”

But what fraction of cancers are attributable to chance?

A controversial 2017 study published in Science suggested that, based on the rate of cell turnover in healthy tissues in the lung, pancreas, and other parts of the body, only about one-third of cancers could be linked to environmental or genetic factors. The rest, the authors claimed, occurred because of random mutations that accumulated in a person’s DNA – in other words, bad luck.

That study brought on a flood of criticism from scientists who pointed to serious flaws in the work that led the researchers to significantly overestimate the share of chance-related cancers.

The actual proportion of cancers that occur by chance is much lower, according to Dr. Brennan. “If you look at international comparisons [of cancer rates] and take a conservative estimate, you see that maybe 10% or 15% of cancers are really chance.”

Whether some cancers are caused by bad luck or undiscovered risk factors remains an open question.

But the bottom line is many unknown causes of cancer are likely environmental or lifestyle related, which means that, in theory, they can be altered, even prevented.

“There is always going to be some element of chance, but you can modify your chance, depending on your lifestyle and maybe other factors, which we don’t fully understand yet,” Dr. Gunter said.

The good news is that, when it comes to prevention, there are many ways to modify our behaviors – such as consuming fewer processed meats, going for a daily walk, or getting vaccinated against cancer-causing viruses – to improve our chances of living cancer free. And as scientists better understand more about what causes cancer, possibilities for prevention will only grow.

“There is a constant, slow growth [in knowledge] that is lowering the overall risk of cancer,” Dr. Brennan said. “We’re never going to eliminate cancer, but we will be able to control it as a disease.”

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

People with cancer are often desperate to know what caused their disease. Was it something they did? Something they could have prevented?

vitanovski/Thinkstock.com

In a recent analysis, experts estimated that about 40% of cancers can be explained by known, often modifiable risk factors. Smoking and obesity represent the primary drivers, though a host of other factors – germline mutations, alcohol, infections, or environmental pollutants like asbestos – contribute to cancer risk as well.

But what about the remaining 60% of cancers?

The study suggests that, although many of these cases likely have an underlying lifestyle or environmental component, experts still do not fully understand their origin story. And a small but significant number may simply be caused by chance.

Here’s what experts suspect those missing causes might be, and why they can be so difficult to confirm.
 

Possibility 1: Known risk factors contribute more than we realize

For certain factors, a straight line can be drawn to cancer.

Take smoking, for instance. Decades of research have helped scientists clearly delineate tobacco’s carcinogenic effects. Researchers have pinpointed a unique set of mutations in the tumors of smokers that can be seen when cells grown in a dish are exposed to the carcinogens present in tobacco.

In addition, experts have been able to collect robust data from epidemiologic studies on smoking prevalence as well as associated cancer risks and deaths, in large part because an individual’s lifetime tobacco exposure is fairly easy to measure.

“The evidence for smoking is incredibly consistent,” Paul Brennan, PhD, a cancer epidemiologist at the World Health Organization’s International Agency for Research on Cancer, said in an interview.

For other known risk factors, such as obesity and air pollution, many more questions than answers remain.

Because of the limitations in how such factors are measured, we are likely downplaying their effects, said Richard Martin, PhD, a professor of clinical epidemiology at the University of Bristol (England).

Take obesity. Excess body weight is associated with an increased risk of at least 13 cancers. Although risk estimates vary by study and cancer type, according to a global snapshot from 2012, being overweight or obese accounted for about 4% of all cancers worldwide – 1% in low-income countries and as high as 8% in high-income countries.

However, Dr. Brennan believes “we have underestimated the effect of obesity [on cancer].”

A key reason, he said, is most studies use body mass index to determine whether someone is overweight or obese, but BMI is a poor measure of body fat. BMI does not differentiate between fat and muscle, which means two people with the same height and weight can have the same BMI, even if one is an athlete who eats lean meats and vegetables while the other lives a sedentary life and consumes large quantities of processed foods and alcohol.

On top of that, studies often only calculate a person’s BMI once, and a single measurement can’t tell you how a person’s weight has fluctuated in recent years or across different stages of their life. However, recent analyses suggest that obesity status over time may be more relevant to cancer risk than one-off measures.

In addition, many studies now suggest that alterations to our gut microbes and high blood insulin level – often seen in people who are overweight or obese – may increase the risk of cancer and speed the growth of tumors.

When these additional factors are considered, the impact of excess body fat may ultimately play a much more significant role in cancer risk. In fact, according to Dr. Brennan, “if we estimate [the effects of obesity] properly, it might at some point become the main cause of cancer.”
 

Possibility 2: Environmental or lifestyle factors remain under the radar

Researchers have linked many substances we consume or are exposed to in our daily lives – air pollution, toxins from industrial waste, and highly processed foods – to cancer. But the extent or contribution of potential carcinogens in our surroundings, particularly those found almost everywhere at low levels, is still largely unknown.

One simple reason is the effects of many of these substances remain difficult to assess. For instance, it is much harder to study the impact of pollutants found in food or water, in which a given population will share similar exposure levels versus tobacco, where it is possible to compare a person who smokes a pack of cigarettes a day with a person who does not smoke.

“If you’ve got exposures that are ubiquitous, it can be difficult to discern their [individual] roles,” Dr. Martin said. “There are many causes that we [likely] don’t really know because everyone has been exposed.”

On the flip side, some carcinogenic substances that people encounter for limited periods might be missed if studies are not performed at the time of exposure.

“What’s in the body at age 40 may not reflect what you were exposed at age 5-10 on the playground or soccer field,” said Graham Colditz, MD, PhD, an epidemiologist and public health expert at Washington University, St. Louis. “The technology keeps changing so we can get better measures of what you’ve got exposure to today, but how that relates to 5, 10, 15 years ago is probably very variable.”

In addition, researchers have found that many carcinogens do not cause specific mutations in a cell’s DNA; rather, studies suggest that most carcinogens lead to cancer-promoting changes in cells, such as inflammation.

“We need to think of how potential carcinogens are causing cancer,” Dr. Brennan said. Instead of provoking mutations, potential carcinogens may use a “whole other kind of pathway.” When, for instance, inflammation becomes chronic, it may spur a cascade of events that ultimately leads to cancer.

Finally, not much is known about what causes cancers in low- and middle-income countries. Most of the research to date has been in high-income countries, such the United States, Australia, and parts of Europe.

“There’s a real lack of robust epidemiological studies in other parts of the world, Latin America, Africa, parts of Asia,” Marc Gunter, PhD, a molecular epidemiologist at the IARC, told this news organization.
 

Possibility 3: Some cancers occur by chance

When it comes to cancer risk, an element of chance may be at play. Cancer can occur in individuals who have very little exposure to known carcinogens or have no family history of cancer.

“We all know there are people who get cancer who eat very healthy diets, are never overweight, and never smoke,” Dr. Gunter said. “Then there are people on the other end of the extreme who don’t get cancer.”

But what fraction of cancers are attributable to chance?

A controversial 2017 study published in Science suggested that, based on the rate of cell turnover in healthy tissues in the lung, pancreas, and other parts of the body, only about one-third of cancers could be linked to environmental or genetic factors. The rest, the authors claimed, occurred because of random mutations that accumulated in a person’s DNA – in other words, bad luck.

That study brought on a flood of criticism from scientists who pointed to serious flaws in the work that led the researchers to significantly overestimate the share of chance-related cancers.

The actual proportion of cancers that occur by chance is much lower, according to Dr. Brennan. “If you look at international comparisons [of cancer rates] and take a conservative estimate, you see that maybe 10% or 15% of cancers are really chance.”

Whether some cancers are caused by bad luck or undiscovered risk factors remains an open question.

But the bottom line is many unknown causes of cancer are likely environmental or lifestyle related, which means that, in theory, they can be altered, even prevented.

“There is always going to be some element of chance, but you can modify your chance, depending on your lifestyle and maybe other factors, which we don’t fully understand yet,” Dr. Gunter said.

The good news is that, when it comes to prevention, there are many ways to modify our behaviors – such as consuming fewer processed meats, going for a daily walk, or getting vaccinated against cancer-causing viruses – to improve our chances of living cancer free. And as scientists better understand more about what causes cancer, possibilities for prevention will only grow.

“There is a constant, slow growth [in knowledge] that is lowering the overall risk of cancer,” Dr. Brennan said. “We’re never going to eliminate cancer, but we will be able to control it as a disease.”

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

People with cancer are often desperate to know what caused their disease. Was it something they did? Something they could have prevented?

vitanovski/Thinkstock.com

In a recent analysis, experts estimated that about 40% of cancers can be explained by known, often modifiable risk factors. Smoking and obesity represent the primary drivers, though a host of other factors – germline mutations, alcohol, infections, or environmental pollutants like asbestos – contribute to cancer risk as well.

But what about the remaining 60% of cancers?

The study suggests that, although many of these cases likely have an underlying lifestyle or environmental component, experts still do not fully understand their origin story. And a small but significant number may simply be caused by chance.

Here’s what experts suspect those missing causes might be, and why they can be so difficult to confirm.
 

Possibility 1: Known risk factors contribute more than we realize

For certain factors, a straight line can be drawn to cancer.

Take smoking, for instance. Decades of research have helped scientists clearly delineate tobacco’s carcinogenic effects. Researchers have pinpointed a unique set of mutations in the tumors of smokers that can be seen when cells grown in a dish are exposed to the carcinogens present in tobacco.

In addition, experts have been able to collect robust data from epidemiologic studies on smoking prevalence as well as associated cancer risks and deaths, in large part because an individual’s lifetime tobacco exposure is fairly easy to measure.

“The evidence for smoking is incredibly consistent,” Paul Brennan, PhD, a cancer epidemiologist at the World Health Organization’s International Agency for Research on Cancer, said in an interview.

For other known risk factors, such as obesity and air pollution, many more questions than answers remain.

Because of the limitations in how such factors are measured, we are likely downplaying their effects, said Richard Martin, PhD, a professor of clinical epidemiology at the University of Bristol (England).

Take obesity. Excess body weight is associated with an increased risk of at least 13 cancers. Although risk estimates vary by study and cancer type, according to a global snapshot from 2012, being overweight or obese accounted for about 4% of all cancers worldwide – 1% in low-income countries and as high as 8% in high-income countries.

However, Dr. Brennan believes “we have underestimated the effect of obesity [on cancer].”

A key reason, he said, is most studies use body mass index to determine whether someone is overweight or obese, but BMI is a poor measure of body fat. BMI does not differentiate between fat and muscle, which means two people with the same height and weight can have the same BMI, even if one is an athlete who eats lean meats and vegetables while the other lives a sedentary life and consumes large quantities of processed foods and alcohol.

On top of that, studies often only calculate a person’s BMI once, and a single measurement can’t tell you how a person’s weight has fluctuated in recent years or across different stages of their life. However, recent analyses suggest that obesity status over time may be more relevant to cancer risk than one-off measures.

In addition, many studies now suggest that alterations to our gut microbes and high blood insulin level – often seen in people who are overweight or obese – may increase the risk of cancer and speed the growth of tumors.

When these additional factors are considered, the impact of excess body fat may ultimately play a much more significant role in cancer risk. In fact, according to Dr. Brennan, “if we estimate [the effects of obesity] properly, it might at some point become the main cause of cancer.”
 

Possibility 2: Environmental or lifestyle factors remain under the radar

Researchers have linked many substances we consume or are exposed to in our daily lives – air pollution, toxins from industrial waste, and highly processed foods – to cancer. But the extent or contribution of potential carcinogens in our surroundings, particularly those found almost everywhere at low levels, is still largely unknown.

One simple reason is the effects of many of these substances remain difficult to assess. For instance, it is much harder to study the impact of pollutants found in food or water, in which a given population will share similar exposure levels versus tobacco, where it is possible to compare a person who smokes a pack of cigarettes a day with a person who does not smoke.

“If you’ve got exposures that are ubiquitous, it can be difficult to discern their [individual] roles,” Dr. Martin said. “There are many causes that we [likely] don’t really know because everyone has been exposed.”

On the flip side, some carcinogenic substances that people encounter for limited periods might be missed if studies are not performed at the time of exposure.

“What’s in the body at age 40 may not reflect what you were exposed at age 5-10 on the playground or soccer field,” said Graham Colditz, MD, PhD, an epidemiologist and public health expert at Washington University, St. Louis. “The technology keeps changing so we can get better measures of what you’ve got exposure to today, but how that relates to 5, 10, 15 years ago is probably very variable.”

In addition, researchers have found that many carcinogens do not cause specific mutations in a cell’s DNA; rather, studies suggest that most carcinogens lead to cancer-promoting changes in cells, such as inflammation.

“We need to think of how potential carcinogens are causing cancer,” Dr. Brennan said. Instead of provoking mutations, potential carcinogens may use a “whole other kind of pathway.” When, for instance, inflammation becomes chronic, it may spur a cascade of events that ultimately leads to cancer.

Finally, not much is known about what causes cancers in low- and middle-income countries. Most of the research to date has been in high-income countries, such the United States, Australia, and parts of Europe.

“There’s a real lack of robust epidemiological studies in other parts of the world, Latin America, Africa, parts of Asia,” Marc Gunter, PhD, a molecular epidemiologist at the IARC, told this news organization.
 

Possibility 3: Some cancers occur by chance

When it comes to cancer risk, an element of chance may be at play. Cancer can occur in individuals who have very little exposure to known carcinogens or have no family history of cancer.

“We all know there are people who get cancer who eat very healthy diets, are never overweight, and never smoke,” Dr. Gunter said. “Then there are people on the other end of the extreme who don’t get cancer.”

But what fraction of cancers are attributable to chance?

A controversial 2017 study published in Science suggested that, based on the rate of cell turnover in healthy tissues in the lung, pancreas, and other parts of the body, only about one-third of cancers could be linked to environmental or genetic factors. The rest, the authors claimed, occurred because of random mutations that accumulated in a person’s DNA – in other words, bad luck.

That study brought on a flood of criticism from scientists who pointed to serious flaws in the work that led the researchers to significantly overestimate the share of chance-related cancers.

The actual proportion of cancers that occur by chance is much lower, according to Dr. Brennan. “If you look at international comparisons [of cancer rates] and take a conservative estimate, you see that maybe 10% or 15% of cancers are really chance.”

Whether some cancers are caused by bad luck or undiscovered risk factors remains an open question.

But the bottom line is many unknown causes of cancer are likely environmental or lifestyle related, which means that, in theory, they can be altered, even prevented.

“There is always going to be some element of chance, but you can modify your chance, depending on your lifestyle and maybe other factors, which we don’t fully understand yet,” Dr. Gunter said.

The good news is that, when it comes to prevention, there are many ways to modify our behaviors – such as consuming fewer processed meats, going for a daily walk, or getting vaccinated against cancer-causing viruses – to improve our chances of living cancer free. And as scientists better understand more about what causes cancer, possibilities for prevention will only grow.

“There is a constant, slow growth [in knowledge] that is lowering the overall risk of cancer,” Dr. Brennan said. “We’re never going to eliminate cancer, but we will be able to control it as a disease.”

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

The International Skin Imaging Collaboration (ISIC) Artificial Intelligence Working Group has released the first-ever guidelines for developing artificial intelligence (AI) algorithms used in dermatology.

Christopher Smith

The guidelines, published in JAMA Dermatology on Dec. 1, 2021, contain a broad range of recommendations stakeholders should consider when developing and assessing image-based AI algorithms in dermatology. The recommendations are divided into categories of data, technique, technical assessment, and application. ISIC is “an academia and industry partnership designed to facilitate the application of digital skin imaging to help reduce melanoma mortality,” and is organized into different working groups, including the AI working group, according to its website.

“Our goal with these guidelines was to create higher-quality reporting of dataset and algorithm characteristics for dermatology AI,” first author Roxana Daneshjou, MD, PhD, clinical scholar in dermatology, in the department of dermatology at Stanford (Calif.) University, said in an interview. “We hope these guidelines also aid regulatory bodies around the world when they are assessing algorithms to be used in dermatology.”
 

Recommendations for data

The authors recommended that datasets used by AI algorithms have image descriptions and details on image artifacts. “For photography, these include the type of camera used; whether images were taken under standardized or varying conditions; whether they were taken by professional photographers, laymen, or health care professionals; and image quality,” they wrote. They also recommended that developers include in an image description the type of lighting used and whether the photo contains pen markings, hair, tattoos, injuries, surgical effects, or other “physical perturbations.”

Exchangeable image file format data obtained from the camera, and preprocessing procedures like color normalization and “postprocessing” of images, such as filtering, should also be disclosed. In addition, developers should disclose and justify inclusion of images that have been created by an algorithm within a dataset. Any public images used in the datasets should have references, and privately used images should be made public where possible, the authors said.

The ISIC working group guidelines also provided recommendations for patient-level metadata. Each image should include a patient’s geographical location and medical center they visited as well as their age, sex and gender, ethnicity and/or race, and skin tone. Dr. Daneshjou said this was one area where she and her colleagues found a lack of transparency in AI datasets in algorithms in a recent review. “We found that many AI papers provided sparse details about the images used to train and test their algorithms,” Dr. Daneshjou explained. “For example, only 7 out of 70 papers had any information about the skin tones in the images used for developing and/or testing AI algorithms. Understanding the diversity of images used to train and test algorithms is important because algorithms that are developed on images of predominantly white skin likely won’t work as well on Black and brown skin.”

The guideline authors also asked algorithm developers to describe the limitations of not including patient-level metadata information when it is incomplete or unavailable. In addition, “we ask that algorithm developers comment on potential biases of their algorithms,” Dr. Daneshjou said. “For example, an algorithm based only on telemedicine images may not capture the full range of diseases seen within an in-person clinic.”

When describing their AI algorithm, developers should detail their reasoning for the dataset size and partitions, inclusion and exclusion criteria for images, and use of any external samples for test sets. “Authors should consider any differences between the image characteristics used for algorithm development and those that might be encountered in the real world,” the guidelines stated.

Recommendations for technique

How the images in a dataset are labeled is a unique challenge in developing AI algorithms for dermatology, the authors noted. Developers should use histopathological diagnosis in their labeling, but this can sometimes result in label noise.

“Many of the AI algorithms in dermatology use supervised learning, which requires labeled examples to help the algorithm ‘learn’ features for discriminating between lesions. We found that some papers use consensus labeling – dermatologists providing a label – to label skin cancers; however, the standard for diagnosing skin cancer is using histopathology from a biopsy,” she said. “Dermatologists can biopsy seven to eight suspected melanomas before discovering a true melanoma, so dermatologist labeling of skin cancers is prone to label noise.”

ISIC’s guidelines stated a gold standard of labeling for dermatologic images is one area that still needs future research, but currently, “diagnoses, labels and diagnostic groups used in data repositories as well as public ontologies” such as ICD-11, AnatomyMapper, and SNOMED-CT should be included in dermatologic image datasets.

AI developers should also provide a detailed description of their algorithm, which includes methods, work flows, mathematical formulas as well as the generalizability of the algorithm across more than one dataset.
 

Recommendations for technical assessment

“Another important recommendation is that algorithm developers should provide a way for algorithms to be publicly evaluable by researchers,” Dr. Daneshjou said. “Many dermatology AI algorithms do not share either their data or their algorithm. Algorithm sharing is important for assessing reproducibility and robustness.”

Google’s recently announced AI-powered dermatology assistant tool, for example, “has made claims about its accuracy and ability to diagnose skin disease at a dermatologist level, but there is no way for researchers to independently test these claims,” she said. Other options like Model Dermatology, developed by Seung Seog Han, MD, PhD, of the Dermatology Clinic in Seoul, South Korea, and colleagues, offer an application programming interface “that allows researchers to test the algorithm,” Dr. Daneshjou said. “This kind of openness is key for assessing algorithm robustness.”

Developers should also note in their algorithm explanations how performance markers and benchmarks would translate to proposed clinical application. “In this context,” the use case – the context in which the AI application is being used – “should be clearly described – who are the intended users and under what clinical scenario are they using the algorithm,” the authors wrote.
 

Recommendations for application

The guidelines note that use case for the model should also be described by the AI developers. “Our checklist includes delineating use cases for algorithms and describing what use cases may be within the scope of the algorithm versus which use cases are out of scope,” Dr. Daneshjou said. “For example, an algorithm developed to provide decision support to dermatologists, with a human in the loop, may not be accurate enough to release directly to consumers.”

As the goal of AI algorithms in dermatology is eventual implementation for clinicians and patients, the authors asked developers to consider shortcomings and potential harms of the algorithm during implementation. “Ethical considerations and impact on vulnerable populations should also be considered and discussed,” they wrote. An algorithm “suggesting aesthetic medical treatments may have negative effects given the biased nature of beauty standards,” and “an algorithm that diagnoses basal cell carcinomas but lacks any pigmented basal cell carcinomas, which are more often seen in skin of color, will not perform equitably across populations.”

Prior to implementing an AI algorithm, the ISIC working group recommended developers perform prospective clinical trials for validation. Checklists and guidelines like SPIRIT-AI and CONSORT-AI “provide guidance on how to design clinical trials to test AI algorithms,” Dr. Daneshjou said.

After implementation, “I believe we need additional research in how we monitor algorithms after they are deployed clinically, Dr. Daneshjou said. “Currently there are no [Food and Drug Administration]–approved AI algorithms in dermatology; however, there are several applications that have CE mark in Europe, and there are no mechanisms for postmarket surveillance there.
 

'Timely' recommendations

Commenting on the ISIC working group guidelines, Justin M. Ko, MD, MBA, director and chief of medical dermatology for Stanford Health Care, who was not involved with the work, said that the recommendations are timely and provide “a framework for a ‘common language’ around AI datasets specifically tailored to dermatology.” Dr. Ko, chair of the American Academy of Dermatology’s Ad Hoc Task Force on Augmented Intelligence, noted the work by Dr. Daneshjou and colleagues “is consistent with and builds further details” on the position statement released by the AAD AI task force in 2019.

“As machine-learning capabilities and commercial efforts continue to mature, it becomes increasingly important that we are able to ‘look under the hood,’ and evaluate all the critical factors that influence development of these capabilities,” he said in an interview. “A standard set of reporting guidelines not only allows for transparency in evaluating data and performance of models and algorithms, but also forces the consideration of issues of equity, fairness, mitigation of bias, and clinically meaningful outcomes.”

One concern is the impact of AI algorithms on societal or health systems, he noted, which is brought up in the guidelines. “The last thing we would want is the development of robust AI systems that exacerbate access challenges, or generate patient anxiety/worry, or drive low-value utilization, or adds to care team burden, or create a technological barrier to care, or increases inequity in dermatologic care,” he said.

In developing AI algorithms for dermatology, a “major practical issue” is how performance on paper will translate to real-world use, Dr. Ko explained, and the ISIC guidelines “provide a critical step in empowering clinicians, practices, and our field to shape the advent of the AI and augmented intelligence tools and systems to promote and enhance meaningful clinical outcomes, and augment the core patient-clinician relationship and ensure they are grounded in principles of fairness, equity and transparency.”

This research was funded by awards and grants to individual authors from the Charina Fund, a Google Research Award, Melanoma Research Alliance, National Health and Medical Research Council, National Institutes of Health/National Cancer Institute, National Science Foundation, and the Department of Veterans Affairs. The authors disclosed relationships with governmental entities, pharmaceutical companies, technology startups, medical publishers, charitable trusts, consulting firms, dermatology training companies, providers of medical devices, manufacturers of dermatologic products, and other organizations related to the paper in the form of supplied equipment, having founded a company; receiving grants, patents, or personal fees; holding shares; and medical reporting. Dr. Ko reported that he serves as a clinical advisor for Skin Analytics, and has an ongoing research collaboration with Google.

The International Skin Imaging Collaboration (ISIC) Artificial Intelligence Working Group has released the first-ever guidelines for developing artificial intelligence (AI) algorithms used in dermatology.

Christopher Smith

The guidelines, published in JAMA Dermatology on Dec. 1, 2021, contain a broad range of recommendations stakeholders should consider when developing and assessing image-based AI algorithms in dermatology. The recommendations are divided into categories of data, technique, technical assessment, and application. ISIC is “an academia and industry partnership designed to facilitate the application of digital skin imaging to help reduce melanoma mortality,” and is organized into different working groups, including the AI working group, according to its website.

“Our goal with these guidelines was to create higher-quality reporting of dataset and algorithm characteristics for dermatology AI,” first author Roxana Daneshjou, MD, PhD, clinical scholar in dermatology, in the department of dermatology at Stanford (Calif.) University, said in an interview. “We hope these guidelines also aid regulatory bodies around the world when they are assessing algorithms to be used in dermatology.”
 

Recommendations for data

The authors recommended that datasets used by AI algorithms have image descriptions and details on image artifacts. “For photography, these include the type of camera used; whether images were taken under standardized or varying conditions; whether they were taken by professional photographers, laymen, or health care professionals; and image quality,” they wrote. They also recommended that developers include in an image description the type of lighting used and whether the photo contains pen markings, hair, tattoos, injuries, surgical effects, or other “physical perturbations.”

Exchangeable image file format data obtained from the camera, and preprocessing procedures like color normalization and “postprocessing” of images, such as filtering, should also be disclosed. In addition, developers should disclose and justify inclusion of images that have been created by an algorithm within a dataset. Any public images used in the datasets should have references, and privately used images should be made public where possible, the authors said.

The ISIC working group guidelines also provided recommendations for patient-level metadata. Each image should include a patient’s geographical location and medical center they visited as well as their age, sex and gender, ethnicity and/or race, and skin tone. Dr. Daneshjou said this was one area where she and her colleagues found a lack of transparency in AI datasets in algorithms in a recent review. “We found that many AI papers provided sparse details about the images used to train and test their algorithms,” Dr. Daneshjou explained. “For example, only 7 out of 70 papers had any information about the skin tones in the images used for developing and/or testing AI algorithms. Understanding the diversity of images used to train and test algorithms is important because algorithms that are developed on images of predominantly white skin likely won’t work as well on Black and brown skin.”

The guideline authors also asked algorithm developers to describe the limitations of not including patient-level metadata information when it is incomplete or unavailable. In addition, “we ask that algorithm developers comment on potential biases of their algorithms,” Dr. Daneshjou said. “For example, an algorithm based only on telemedicine images may not capture the full range of diseases seen within an in-person clinic.”

When describing their AI algorithm, developers should detail their reasoning for the dataset size and partitions, inclusion and exclusion criteria for images, and use of any external samples for test sets. “Authors should consider any differences between the image characteristics used for algorithm development and those that might be encountered in the real world,” the guidelines stated.

Recommendations for technique

How the images in a dataset are labeled is a unique challenge in developing AI algorithms for dermatology, the authors noted. Developers should use histopathological diagnosis in their labeling, but this can sometimes result in label noise.

“Many of the AI algorithms in dermatology use supervised learning, which requires labeled examples to help the algorithm ‘learn’ features for discriminating between lesions. We found that some papers use consensus labeling – dermatologists providing a label – to label skin cancers; however, the standard for diagnosing skin cancer is using histopathology from a biopsy,” she said. “Dermatologists can biopsy seven to eight suspected melanomas before discovering a true melanoma, so dermatologist labeling of skin cancers is prone to label noise.”

ISIC’s guidelines stated a gold standard of labeling for dermatologic images is one area that still needs future research, but currently, “diagnoses, labels and diagnostic groups used in data repositories as well as public ontologies” such as ICD-11, AnatomyMapper, and SNOMED-CT should be included in dermatologic image datasets.

AI developers should also provide a detailed description of their algorithm, which includes methods, work flows, mathematical formulas as well as the generalizability of the algorithm across more than one dataset.
 

Recommendations for technical assessment

“Another important recommendation is that algorithm developers should provide a way for algorithms to be publicly evaluable by researchers,” Dr. Daneshjou said. “Many dermatology AI algorithms do not share either their data or their algorithm. Algorithm sharing is important for assessing reproducibility and robustness.”

Google’s recently announced AI-powered dermatology assistant tool, for example, “has made claims about its accuracy and ability to diagnose skin disease at a dermatologist level, but there is no way for researchers to independently test these claims,” she said. Other options like Model Dermatology, developed by Seung Seog Han, MD, PhD, of the Dermatology Clinic in Seoul, South Korea, and colleagues, offer an application programming interface “that allows researchers to test the algorithm,” Dr. Daneshjou said. “This kind of openness is key for assessing algorithm robustness.”

Developers should also note in their algorithm explanations how performance markers and benchmarks would translate to proposed clinical application. “In this context,” the use case – the context in which the AI application is being used – “should be clearly described – who are the intended users and under what clinical scenario are they using the algorithm,” the authors wrote.
 

Recommendations for application

The guidelines note that use case for the model should also be described by the AI developers. “Our checklist includes delineating use cases for algorithms and describing what use cases may be within the scope of the algorithm versus which use cases are out of scope,” Dr. Daneshjou said. “For example, an algorithm developed to provide decision support to dermatologists, with a human in the loop, may not be accurate enough to release directly to consumers.”

As the goal of AI algorithms in dermatology is eventual implementation for clinicians and patients, the authors asked developers to consider shortcomings and potential harms of the algorithm during implementation. “Ethical considerations and impact on vulnerable populations should also be considered and discussed,” they wrote. An algorithm “suggesting aesthetic medical treatments may have negative effects given the biased nature of beauty standards,” and “an algorithm that diagnoses basal cell carcinomas but lacks any pigmented basal cell carcinomas, which are more often seen in skin of color, will not perform equitably across populations.”

Prior to implementing an AI algorithm, the ISIC working group recommended developers perform prospective clinical trials for validation. Checklists and guidelines like SPIRIT-AI and CONSORT-AI “provide guidance on how to design clinical trials to test AI algorithms,” Dr. Daneshjou said.

After implementation, “I believe we need additional research in how we monitor algorithms after they are deployed clinically, Dr. Daneshjou said. “Currently there are no [Food and Drug Administration]–approved AI algorithms in dermatology; however, there are several applications that have CE mark in Europe, and there are no mechanisms for postmarket surveillance there.
 

'Timely' recommendations

Commenting on the ISIC working group guidelines, Justin M. Ko, MD, MBA, director and chief of medical dermatology for Stanford Health Care, who was not involved with the work, said that the recommendations are timely and provide “a framework for a ‘common language’ around AI datasets specifically tailored to dermatology.” Dr. Ko, chair of the American Academy of Dermatology’s Ad Hoc Task Force on Augmented Intelligence, noted the work by Dr. Daneshjou and colleagues “is consistent with and builds further details” on the position statement released by the AAD AI task force in 2019.

“As machine-learning capabilities and commercial efforts continue to mature, it becomes increasingly important that we are able to ‘look under the hood,’ and evaluate all the critical factors that influence development of these capabilities,” he said in an interview. “A standard set of reporting guidelines not only allows for transparency in evaluating data and performance of models and algorithms, but also forces the consideration of issues of equity, fairness, mitigation of bias, and clinically meaningful outcomes.”

One concern is the impact of AI algorithms on societal or health systems, he noted, which is brought up in the guidelines. “The last thing we would want is the development of robust AI systems that exacerbate access challenges, or generate patient anxiety/worry, or drive low-value utilization, or adds to care team burden, or create a technological barrier to care, or increases inequity in dermatologic care,” he said.

In developing AI algorithms for dermatology, a “major practical issue” is how performance on paper will translate to real-world use, Dr. Ko explained, and the ISIC guidelines “provide a critical step in empowering clinicians, practices, and our field to shape the advent of the AI and augmented intelligence tools and systems to promote and enhance meaningful clinical outcomes, and augment the core patient-clinician relationship and ensure they are grounded in principles of fairness, equity and transparency.”

This research was funded by awards and grants to individual authors from the Charina Fund, a Google Research Award, Melanoma Research Alliance, National Health and Medical Research Council, National Institutes of Health/National Cancer Institute, National Science Foundation, and the Department of Veterans Affairs. The authors disclosed relationships with governmental entities, pharmaceutical companies, technology startups, medical publishers, charitable trusts, consulting firms, dermatology training companies, providers of medical devices, manufacturers of dermatologic products, and other organizations related to the paper in the form of supplied equipment, having founded a company; receiving grants, patents, or personal fees; holding shares; and medical reporting. Dr. Ko reported that he serves as a clinical advisor for Skin Analytics, and has an ongoing research collaboration with Google.

The International Skin Imaging Collaboration (ISIC) Artificial Intelligence Working Group has released the first-ever guidelines for developing artificial intelligence (AI) algorithms used in dermatology.

Christopher Smith

The guidelines, published in JAMA Dermatology on Dec. 1, 2021, contain a broad range of recommendations stakeholders should consider when developing and assessing image-based AI algorithms in dermatology. The recommendations are divided into categories of data, technique, technical assessment, and application. ISIC is “an academia and industry partnership designed to facilitate the application of digital skin imaging to help reduce melanoma mortality,” and is organized into different working groups, including the AI working group, according to its website.

“Our goal with these guidelines was to create higher-quality reporting of dataset and algorithm characteristics for dermatology AI,” first author Roxana Daneshjou, MD, PhD, clinical scholar in dermatology, in the department of dermatology at Stanford (Calif.) University, said in an interview. “We hope these guidelines also aid regulatory bodies around the world when they are assessing algorithms to be used in dermatology.”
 

Recommendations for data

The authors recommended that datasets used by AI algorithms have image descriptions and details on image artifacts. “For photography, these include the type of camera used; whether images were taken under standardized or varying conditions; whether they were taken by professional photographers, laymen, or health care professionals; and image quality,” they wrote. They also recommended that developers include in an image description the type of lighting used and whether the photo contains pen markings, hair, tattoos, injuries, surgical effects, or other “physical perturbations.”

Exchangeable image file format data obtained from the camera, and preprocessing procedures like color normalization and “postprocessing” of images, such as filtering, should also be disclosed. In addition, developers should disclose and justify inclusion of images that have been created by an algorithm within a dataset. Any public images used in the datasets should have references, and privately used images should be made public where possible, the authors said.

The ISIC working group guidelines also provided recommendations for patient-level metadata. Each image should include a patient’s geographical location and medical center they visited as well as their age, sex and gender, ethnicity and/or race, and skin tone. Dr. Daneshjou said this was one area where she and her colleagues found a lack of transparency in AI datasets in algorithms in a recent review. “We found that many AI papers provided sparse details about the images used to train and test their algorithms,” Dr. Daneshjou explained. “For example, only 7 out of 70 papers had any information about the skin tones in the images used for developing and/or testing AI algorithms. Understanding the diversity of images used to train and test algorithms is important because algorithms that are developed on images of predominantly white skin likely won’t work as well on Black and brown skin.”

The guideline authors also asked algorithm developers to describe the limitations of not including patient-level metadata information when it is incomplete or unavailable. In addition, “we ask that algorithm developers comment on potential biases of their algorithms,” Dr. Daneshjou said. “For example, an algorithm based only on telemedicine images may not capture the full range of diseases seen within an in-person clinic.”

When describing their AI algorithm, developers should detail their reasoning for the dataset size and partitions, inclusion and exclusion criteria for images, and use of any external samples for test sets. “Authors should consider any differences between the image characteristics used for algorithm development and those that might be encountered in the real world,” the guidelines stated.

Recommendations for technique

How the images in a dataset are labeled is a unique challenge in developing AI algorithms for dermatology, the authors noted. Developers should use histopathological diagnosis in their labeling, but this can sometimes result in label noise.

“Many of the AI algorithms in dermatology use supervised learning, which requires labeled examples to help the algorithm ‘learn’ features for discriminating between lesions. We found that some papers use consensus labeling – dermatologists providing a label – to label skin cancers; however, the standard for diagnosing skin cancer is using histopathology from a biopsy,” she said. “Dermatologists can biopsy seven to eight suspected melanomas before discovering a true melanoma, so dermatologist labeling of skin cancers is prone to label noise.”

ISIC’s guidelines stated a gold standard of labeling for dermatologic images is one area that still needs future research, but currently, “diagnoses, labels and diagnostic groups used in data repositories as well as public ontologies” such as ICD-11, AnatomyMapper, and SNOMED-CT should be included in dermatologic image datasets.

AI developers should also provide a detailed description of their algorithm, which includes methods, work flows, mathematical formulas as well as the generalizability of the algorithm across more than one dataset.
 

Recommendations for technical assessment

“Another important recommendation is that algorithm developers should provide a way for algorithms to be publicly evaluable by researchers,” Dr. Daneshjou said. “Many dermatology AI algorithms do not share either their data or their algorithm. Algorithm sharing is important for assessing reproducibility and robustness.”

Google’s recently announced AI-powered dermatology assistant tool, for example, “has made claims about its accuracy and ability to diagnose skin disease at a dermatologist level, but there is no way for researchers to independently test these claims,” she said. Other options like Model Dermatology, developed by Seung Seog Han, MD, PhD, of the Dermatology Clinic in Seoul, South Korea, and colleagues, offer an application programming interface “that allows researchers to test the algorithm,” Dr. Daneshjou said. “This kind of openness is key for assessing algorithm robustness.”

Developers should also note in their algorithm explanations how performance markers and benchmarks would translate to proposed clinical application. “In this context,” the use case – the context in which the AI application is being used – “should be clearly described – who are the intended users and under what clinical scenario are they using the algorithm,” the authors wrote.
 

Recommendations for application

The guidelines note that use case for the model should also be described by the AI developers. “Our checklist includes delineating use cases for algorithms and describing what use cases may be within the scope of the algorithm versus which use cases are out of scope,” Dr. Daneshjou said. “For example, an algorithm developed to provide decision support to dermatologists, with a human in the loop, may not be accurate enough to release directly to consumers.”

As the goal of AI algorithms in dermatology is eventual implementation for clinicians and patients, the authors asked developers to consider shortcomings and potential harms of the algorithm during implementation. “Ethical considerations and impact on vulnerable populations should also be considered and discussed,” they wrote. An algorithm “suggesting aesthetic medical treatments may have negative effects given the biased nature of beauty standards,” and “an algorithm that diagnoses basal cell carcinomas but lacks any pigmented basal cell carcinomas, which are more often seen in skin of color, will not perform equitably across populations.”

Prior to implementing an AI algorithm, the ISIC working group recommended developers perform prospective clinical trials for validation. Checklists and guidelines like SPIRIT-AI and CONSORT-AI “provide guidance on how to design clinical trials to test AI algorithms,” Dr. Daneshjou said.

After implementation, “I believe we need additional research in how we monitor algorithms after they are deployed clinically, Dr. Daneshjou said. “Currently there are no [Food and Drug Administration]–approved AI algorithms in dermatology; however, there are several applications that have CE mark in Europe, and there are no mechanisms for postmarket surveillance there.
 

'Timely' recommendations

Commenting on the ISIC working group guidelines, Justin M. Ko, MD, MBA, director and chief of medical dermatology for Stanford Health Care, who was not involved with the work, said that the recommendations are timely and provide “a framework for a ‘common language’ around AI datasets specifically tailored to dermatology.” Dr. Ko, chair of the American Academy of Dermatology’s Ad Hoc Task Force on Augmented Intelligence, noted the work by Dr. Daneshjou and colleagues “is consistent with and builds further details” on the position statement released by the AAD AI task force in 2019.

“As machine-learning capabilities and commercial efforts continue to mature, it becomes increasingly important that we are able to ‘look under the hood,’ and evaluate all the critical factors that influence development of these capabilities,” he said in an interview. “A standard set of reporting guidelines not only allows for transparency in evaluating data and performance of models and algorithms, but also forces the consideration of issues of equity, fairness, mitigation of bias, and clinically meaningful outcomes.”

One concern is the impact of AI algorithms on societal or health systems, he noted, which is brought up in the guidelines. “The last thing we would want is the development of robust AI systems that exacerbate access challenges, or generate patient anxiety/worry, or drive low-value utilization, or adds to care team burden, or create a technological barrier to care, or increases inequity in dermatologic care,” he said.

In developing AI algorithms for dermatology, a “major practical issue” is how performance on paper will translate to real-world use, Dr. Ko explained, and the ISIC guidelines “provide a critical step in empowering clinicians, practices, and our field to shape the advent of the AI and augmented intelligence tools and systems to promote and enhance meaningful clinical outcomes, and augment the core patient-clinician relationship and ensure they are grounded in principles of fairness, equity and transparency.”

This research was funded by awards and grants to individual authors from the Charina Fund, a Google Research Award, Melanoma Research Alliance, National Health and Medical Research Council, National Institutes of Health/National Cancer Institute, National Science Foundation, and the Department of Veterans Affairs. The authors disclosed relationships with governmental entities, pharmaceutical companies, technology startups, medical publishers, charitable trusts, consulting firms, dermatology training companies, providers of medical devices, manufacturers of dermatologic products, and other organizations related to the paper in the form of supplied equipment, having founded a company; receiving grants, patents, or personal fees; holding shares; and medical reporting. Dr. Ko reported that he serves as a clinical advisor for Skin Analytics, and has an ongoing research collaboration with Google.

There is an increased incidence of locally staged Merkel cell carcinoma (MCC) among patients who live in rural areas of the United States, compared with those in urban and metropolitan areas, yet overall survival is worse in rural areas.

This paradox was discovered in an analysis of data from the Surveillance, Epidemiology, and End Results (SEER) Program that primary author Bryan T. Carroll, MD, PhD, and colleagues presented during a virtual abstract session at the annual meeting of the American Society for Dermatologic Surgery.

“MCC is a rare and aggressive neoplasm of the skin with high mortality,” said coauthor Emma Larson, MD, a dermatology clinical research fellow at University Hospitals of Cleveland. “Previous studies have demonstrated that MCC survival is lower in low–dermatologist density areas. Associations are difficult to characterize without historical staging data aggregated from large registries. We hypothesized that decreased MCC survival is associated with rural counties.”

The researchers used 18 registries from the November 2019 SEER database to retrospectively evaluate adults who were diagnosed with MCC between 2004 and 2015 as confirmed by positive histology. Study endpoints were SEER historic stage at diagnosis and 5-year survival. MCC cases were stratified by 2013 USDA urban-rural continuum codes, which defines metropolitan counties as those with a population of 1 million or more, urban counties as those with a population of less than 1 million, and rural counties as nonmetropolitan counties not adjacent to a metropolitan area.

A total of 6,291 cases with a mean age of 75 years were included in the final analysis: 3,750 from metro areas, 2,235 from urban areas, and 306 from rural areas. A higher proportion of MCC patients from rural areas were male (69% vs. 62% from metro areas and 64% from urban areas) and white (97% vs. 95% and 96%, respectively). “This may contribute to differences in MCC care,” Dr. Larson said. “However, we also found that there is an increased incidence of locally staged disease in rural areas (51%) than in metro (44%) or urban (45%) areas (P = .02). In addition, fewer lymph node surgeries were performed in rural (50%) and urban (51%) areas than in metro areas (45%; P = .01).”

Overall survival was worse among patients in rural areas (a mean of 34 months), compared with those in urban (a mean of 41 months) and metro areas (a mean of 47 months; P = .02). “This may be due to the fact that rural counties have the higher risk factors for MCC incidence and death, but when we account for the confounders, including sex, age, race, and MCC stage, we still found a difference in overall survival in rural counties, compared to metro and urban counties,” Dr. Larson said.

Dr. Carroll, an associate professor of dermatology at University Hospitals of Cleveland, characterized the finding as “not what you’d expect with a higher incidence of local disease. Therefore, there is the potential for mis-staging in rural counties, where we did see that the interrogation of lymph nodes was done less frequently than in urban centers, which were more aligned with National Comprehensive Cancer Network guidelines during this time period. Still, after correction, rural location is still associated with a higher MCC mortality. There is a need for us to further interrogate what the causes are for this disparity in care between rural and urban centers.”

The other study authors were Dustin DeMeo and Christian Scheufele, MD. The researchers reported having no relevant financial disclosures.

[email protected]

There is an increased incidence of locally staged Merkel cell carcinoma (MCC) among patients who live in rural areas of the United States, compared with those in urban and metropolitan areas, yet overall survival is worse in rural areas.

This paradox was discovered in an analysis of data from the Surveillance, Epidemiology, and End Results (SEER) Program that primary author Bryan T. Carroll, MD, PhD, and colleagues presented during a virtual abstract session at the annual meeting of the American Society for Dermatologic Surgery.

“MCC is a rare and aggressive neoplasm of the skin with high mortality,” said coauthor Emma Larson, MD, a dermatology clinical research fellow at University Hospitals of Cleveland. “Previous studies have demonstrated that MCC survival is lower in low–dermatologist density areas. Associations are difficult to characterize without historical staging data aggregated from large registries. We hypothesized that decreased MCC survival is associated with rural counties.”

The researchers used 18 registries from the November 2019 SEER database to retrospectively evaluate adults who were diagnosed with MCC between 2004 and 2015 as confirmed by positive histology. Study endpoints were SEER historic stage at diagnosis and 5-year survival. MCC cases were stratified by 2013 USDA urban-rural continuum codes, which defines metropolitan counties as those with a population of 1 million or more, urban counties as those with a population of less than 1 million, and rural counties as nonmetropolitan counties not adjacent to a metropolitan area.

A total of 6,291 cases with a mean age of 75 years were included in the final analysis: 3,750 from metro areas, 2,235 from urban areas, and 306 from rural areas. A higher proportion of MCC patients from rural areas were male (69% vs. 62% from metro areas and 64% from urban areas) and white (97% vs. 95% and 96%, respectively). “This may contribute to differences in MCC care,” Dr. Larson said. “However, we also found that there is an increased incidence of locally staged disease in rural areas (51%) than in metro (44%) or urban (45%) areas (P = .02). In addition, fewer lymph node surgeries were performed in rural (50%) and urban (51%) areas than in metro areas (45%; P = .01).”

Overall survival was worse among patients in rural areas (a mean of 34 months), compared with those in urban (a mean of 41 months) and metro areas (a mean of 47 months; P = .02). “This may be due to the fact that rural counties have the higher risk factors for MCC incidence and death, but when we account for the confounders, including sex, age, race, and MCC stage, we still found a difference in overall survival in rural counties, compared to metro and urban counties,” Dr. Larson said.

Dr. Carroll, an associate professor of dermatology at University Hospitals of Cleveland, characterized the finding as “not what you’d expect with a higher incidence of local disease. Therefore, there is the potential for mis-staging in rural counties, where we did see that the interrogation of lymph nodes was done less frequently than in urban centers, which were more aligned with National Comprehensive Cancer Network guidelines during this time period. Still, after correction, rural location is still associated with a higher MCC mortality. There is a need for us to further interrogate what the causes are for this disparity in care between rural and urban centers.”

The other study authors were Dustin DeMeo and Christian Scheufele, MD. The researchers reported having no relevant financial disclosures.

[email protected]

There is an increased incidence of locally staged Merkel cell carcinoma (MCC) among patients who live in rural areas of the United States, compared with those in urban and metropolitan areas, yet overall survival is worse in rural areas.

This paradox was discovered in an analysis of data from the Surveillance, Epidemiology, and End Results (SEER) Program that primary author Bryan T. Carroll, MD, PhD, and colleagues presented during a virtual abstract session at the annual meeting of the American Society for Dermatologic Surgery.

“MCC is a rare and aggressive neoplasm of the skin with high mortality,” said coauthor Emma Larson, MD, a dermatology clinical research fellow at University Hospitals of Cleveland. “Previous studies have demonstrated that MCC survival is lower in low–dermatologist density areas. Associations are difficult to characterize without historical staging data aggregated from large registries. We hypothesized that decreased MCC survival is associated with rural counties.”

The researchers used 18 registries from the November 2019 SEER database to retrospectively evaluate adults who were diagnosed with MCC between 2004 and 2015 as confirmed by positive histology. Study endpoints were SEER historic stage at diagnosis and 5-year survival. MCC cases were stratified by 2013 USDA urban-rural continuum codes, which defines metropolitan counties as those with a population of 1 million or more, urban counties as those with a population of less than 1 million, and rural counties as nonmetropolitan counties not adjacent to a metropolitan area.

A total of 6,291 cases with a mean age of 75 years were included in the final analysis: 3,750 from metro areas, 2,235 from urban areas, and 306 from rural areas. A higher proportion of MCC patients from rural areas were male (69% vs. 62% from metro areas and 64% from urban areas) and white (97% vs. 95% and 96%, respectively). “This may contribute to differences in MCC care,” Dr. Larson said. “However, we also found that there is an increased incidence of locally staged disease in rural areas (51%) than in metro (44%) or urban (45%) areas (P = .02). In addition, fewer lymph node surgeries were performed in rural (50%) and urban (51%) areas than in metro areas (45%; P = .01).”

Overall survival was worse among patients in rural areas (a mean of 34 months), compared with those in urban (a mean of 41 months) and metro areas (a mean of 47 months; P = .02). “This may be due to the fact that rural counties have the higher risk factors for MCC incidence and death, but when we account for the confounders, including sex, age, race, and MCC stage, we still found a difference in overall survival in rural counties, compared to metro and urban counties,” Dr. Larson said.

Dr. Carroll, an associate professor of dermatology at University Hospitals of Cleveland, characterized the finding as “not what you’d expect with a higher incidence of local disease. Therefore, there is the potential for mis-staging in rural counties, where we did see that the interrogation of lymph nodes was done less frequently than in urban centers, which were more aligned with National Comprehensive Cancer Network guidelines during this time period. Still, after correction, rural location is still associated with a higher MCC mortality. There is a need for us to further interrogate what the causes are for this disparity in care between rural and urban centers.”

The other study authors were Dustin DeMeo and Christian Scheufele, MD. The researchers reported having no relevant financial disclosures.

[email protected]

The Diagnosis: Nevoid Basal Cell Carcinoma Syndrome

Nevoid basal cell carcinoma syndrome (NBCCS), also known as Gorlin syndrome, is a rare autosomal-dominant disorder that increases the risk for developing various carcinomas and affects multiple organ systems. Nevoid basal cell carcinoma syndrome is estimated at 1 per 40,000 to 60,000 individuals with no sexual predilection.^1,2 Pathogenesis of NBCCS occurs through molecular alterations in the dormant hedgehog signaling pathway, causing constitutive signaling activity and a loss of function in the tumor suppressor patched 1 gene, PTCH1. As a result, the inhibition of smoothened oncogenes is released, Gli proteins are activated, and the hedgehog signaling pathway is no longer quiescent.² Additional loss of function in the suppressor of fused homolog protein, a negative regulator of the hedgehog pathway, allows for further tumor proliferation. The crucial role these genes play in the hedgehog signaling pathway and their mutation association with NBCCS allows for molecular confirmation in the diagnosis of NBCCS. Allelic losses at the PTCH1 gene site are thought to occur in approximately 70% of NBCCS patients.²

Diagnosis of NBCCS is based on genetic testing to examine pathogenic gene variants, notably in the PTCH1 gene, and identification of characteristic clinical findings.² Diagnosis of NBCCS requires either 2 minor suggestive criteria and 1 major suggestive criterion, 2 major suggestive criteria and 1 minor suggestive criterion, or 1 major suggestive criterion with molecular confirmation. The presence of basal cell carcinomas (BCCs) before 20 years of age or an excessive numbers of BCCs, keratocystic odontogenic tumors (KOTs), palmar or plantar pitting, and first-degree relatives with NBCCS are classified as major suggestive criteria.² Nevoid basal cell carcinoma syndrome patients typically have BCCs that crust, ulcerate, or bleed. Minor suggestive criteria for NBCCS are rib abnormalities, skeletal malformations, macrocephaly, cleft lip or palate, and desmoplastic medulloblastoma.^2-4 Suppressor of fused homolog protein mutations may increase the risk for desmoplastic medulloblastoma in NBCCS patients. Our patient had 4 of the major suggestive criteria, including a history of KOTs, multiple BCCs, first-degree relatives with NBCCS, and palmar or plantar pitting (bottom quiz image), while having 1 minor suggestive criterion of frontal bossing.

Patients with NBCCS have high phenotypic variability, as their skin carcinomas do not have the classic features of pearly surfaces or corkscrew telangiectasia that typically are associated with BCCs.¹ Basal cell carcinomas in NBCCS-affected individuals usually are indistinguishable from sporadic lesions that arise in sun-exposed areas, making NBCCS difficult to diagnose. These sporadic lesions often are misdiagnosed as psoriatic or eczematous lesions, and additional subsequent examination is required. The findings of multiple papules and plaques spanning the body as well as lesions with rolled borders and ulcerated bases, indicative of BCCs, aid dermatologists in distinguishing benign lesions from those of NBCCS.¹

Additional differential diagnoses are required to distinguish NBCCS from other similar inherited skin disorders that are characterized by BCCs. The presence of multiple incidental BCCs early in life remains a histopathologic clue for NBCCS diagnosis, as opposed to Rombo syndrome, in which BCCs often develop in adulthood.^2,4 In addition, although both Bazex syndrome and Muir-Torre syndrome are characterized by the early onset of BCCs, the lack of skeletal abnormalities and palmar and plantar pitting distinguish these entities from NBCCS.^2,4 Furthermore, though psoriasis also can present on the scalp, clinical presentation often includes well-demarcated and symmetric plaques that are erythematous and silvery, all of which were not present in our patient and typically are not seen in NBCCS.⁵

The recommended treatment of NBCCS is vismodegib, a specific oncogene inhibitor. This medication suppresses the hedgehog signaling pathway by inhibiting smoothened oncogenes and downstream target molecules, thereby decreasing tumor proliferation.⁶ In doing so, vismodegib inhibits the development of new BCCs while reducing the burden of present ones. Additionally, vismodegib appears to effectively treat KOTs. If successful, this medication may be able to suppress KOTs in patients with NBCCS and thus facilitate surgery.⁶ Additional hedgehog inhibitors include patidegib, sonidegib, and itraconazole. Patidegib gel 2% currently is in phase 3 clinical trials for evaluation of efficacy and safety in treatment of NBCCS. Sonidegib is approved for the treatment of locally advanced BCCs in the United States and the European Union and for both locally advanced BCCs and metastatic BCCs in Switzerland and Australia.⁷ Further research is needed before recommending antifungal itraconazole for NBCCS clinical use.8 Other medications for localized areas include topical application of 5-fluorouracil and imiquimod.²

The Diagnosis: Nevoid Basal Cell Carcinoma Syndrome

Nevoid basal cell carcinoma syndrome (NBCCS), also known as Gorlin syndrome, is a rare autosomal-dominant disorder that increases the risk for developing various carcinomas and affects multiple organ systems. Nevoid basal cell carcinoma syndrome is estimated at 1 per 40,000 to 60,000 individuals with no sexual predilection.^1,2 Pathogenesis of NBCCS occurs through molecular alterations in the dormant hedgehog signaling pathway, causing constitutive signaling activity and a loss of function in the tumor suppressor patched 1 gene, PTCH1. As a result, the inhibition of smoothened oncogenes is released, Gli proteins are activated, and the hedgehog signaling pathway is no longer quiescent.² Additional loss of function in the suppressor of fused homolog protein, a negative regulator of the hedgehog pathway, allows for further tumor proliferation. The crucial role these genes play in the hedgehog signaling pathway and their mutation association with NBCCS allows for molecular confirmation in the diagnosis of NBCCS. Allelic losses at the PTCH1 gene site are thought to occur in approximately 70% of NBCCS patients.²

Diagnosis of NBCCS is based on genetic testing to examine pathogenic gene variants, notably in the PTCH1 gene, and identification of characteristic clinical findings.² Diagnosis of NBCCS requires either 2 minor suggestive criteria and 1 major suggestive criterion, 2 major suggestive criteria and 1 minor suggestive criterion, or 1 major suggestive criterion with molecular confirmation. The presence of basal cell carcinomas (BCCs) before 20 years of age or an excessive numbers of BCCs, keratocystic odontogenic tumors (KOTs), palmar or plantar pitting, and first-degree relatives with NBCCS are classified as major suggestive criteria.² Nevoid basal cell carcinoma syndrome patients typically have BCCs that crust, ulcerate, or bleed. Minor suggestive criteria for NBCCS are rib abnormalities, skeletal malformations, macrocephaly, cleft lip or palate, and desmoplastic medulloblastoma.^2-4 Suppressor of fused homolog protein mutations may increase the risk for desmoplastic medulloblastoma in NBCCS patients. Our patient had 4 of the major suggestive criteria, including a history of KOTs, multiple BCCs, first-degree relatives with NBCCS, and palmar or plantar pitting (bottom quiz image), while having 1 minor suggestive criterion of frontal bossing.

Patients with NBCCS have high phenotypic variability, as their skin carcinomas do not have the classic features of pearly surfaces or corkscrew telangiectasia that typically are associated with BCCs.¹ Basal cell carcinomas in NBCCS-affected individuals usually are indistinguishable from sporadic lesions that arise in sun-exposed areas, making NBCCS difficult to diagnose. These sporadic lesions often are misdiagnosed as psoriatic or eczematous lesions, and additional subsequent examination is required. The findings of multiple papules and plaques spanning the body as well as lesions with rolled borders and ulcerated bases, indicative of BCCs, aid dermatologists in distinguishing benign lesions from those of NBCCS.¹

Additional differential diagnoses are required to distinguish NBCCS from other similar inherited skin disorders that are characterized by BCCs. The presence of multiple incidental BCCs early in life remains a histopathologic clue for NBCCS diagnosis, as opposed to Rombo syndrome, in which BCCs often develop in adulthood.^2,4 In addition, although both Bazex syndrome and Muir-Torre syndrome are characterized by the early onset of BCCs, the lack of skeletal abnormalities and palmar and plantar pitting distinguish these entities from NBCCS.^2,4 Furthermore, though psoriasis also can present on the scalp, clinical presentation often includes well-demarcated and symmetric plaques that are erythematous and silvery, all of which were not present in our patient and typically are not seen in NBCCS.⁵

The recommended treatment of NBCCS is vismodegib, a specific oncogene inhibitor. This medication suppresses the hedgehog signaling pathway by inhibiting smoothened oncogenes and downstream target molecules, thereby decreasing tumor proliferation.⁶ In doing so, vismodegib inhibits the development of new BCCs while reducing the burden of present ones. Additionally, vismodegib appears to effectively treat KOTs. If successful, this medication may be able to suppress KOTs in patients with NBCCS and thus facilitate surgery.⁶ Additional hedgehog inhibitors include patidegib, sonidegib, and itraconazole. Patidegib gel 2% currently is in phase 3 clinical trials for evaluation of efficacy and safety in treatment of NBCCS. Sonidegib is approved for the treatment of locally advanced BCCs in the United States and the European Union and for both locally advanced BCCs and metastatic BCCs in Switzerland and Australia.⁷ Further research is needed before recommending antifungal itraconazole for NBCCS clinical use.8 Other medications for localized areas include topical application of 5-fluorouracil and imiquimod.²

The Diagnosis: Nevoid Basal Cell Carcinoma Syndrome

Nevoid basal cell carcinoma syndrome (NBCCS), also known as Gorlin syndrome, is a rare autosomal-dominant disorder that increases the risk for developing various carcinomas and affects multiple organ systems. Nevoid basal cell carcinoma syndrome is estimated at 1 per 40,000 to 60,000 individuals with no sexual predilection.^1,2 Pathogenesis of NBCCS occurs through molecular alterations in the dormant hedgehog signaling pathway, causing constitutive signaling activity and a loss of function in the tumor suppressor patched 1 gene, PTCH1. As a result, the inhibition of smoothened oncogenes is released, Gli proteins are activated, and the hedgehog signaling pathway is no longer quiescent.² Additional loss of function in the suppressor of fused homolog protein, a negative regulator of the hedgehog pathway, allows for further tumor proliferation. The crucial role these genes play in the hedgehog signaling pathway and their mutation association with NBCCS allows for molecular confirmation in the diagnosis of NBCCS. Allelic losses at the PTCH1 gene site are thought to occur in approximately 70% of NBCCS patients.²

Diagnosis of NBCCS is based on genetic testing to examine pathogenic gene variants, notably in the PTCH1 gene, and identification of characteristic clinical findings.² Diagnosis of NBCCS requires either 2 minor suggestive criteria and 1 major suggestive criterion, 2 major suggestive criteria and 1 minor suggestive criterion, or 1 major suggestive criterion with molecular confirmation. The presence of basal cell carcinomas (BCCs) before 20 years of age or an excessive numbers of BCCs, keratocystic odontogenic tumors (KOTs), palmar or plantar pitting, and first-degree relatives with NBCCS are classified as major suggestive criteria.² Nevoid basal cell carcinoma syndrome patients typically have BCCs that crust, ulcerate, or bleed. Minor suggestive criteria for NBCCS are rib abnormalities, skeletal malformations, macrocephaly, cleft lip or palate, and desmoplastic medulloblastoma.^2-4 Suppressor of fused homolog protein mutations may increase the risk for desmoplastic medulloblastoma in NBCCS patients. Our patient had 4 of the major suggestive criteria, including a history of KOTs, multiple BCCs, first-degree relatives with NBCCS, and palmar or plantar pitting (bottom quiz image), while having 1 minor suggestive criterion of frontal bossing.

Patients with NBCCS have high phenotypic variability, as their skin carcinomas do not have the classic features of pearly surfaces or corkscrew telangiectasia that typically are associated with BCCs.¹ Basal cell carcinomas in NBCCS-affected individuals usually are indistinguishable from sporadic lesions that arise in sun-exposed areas, making NBCCS difficult to diagnose. These sporadic lesions often are misdiagnosed as psoriatic or eczematous lesions, and additional subsequent examination is required. The findings of multiple papules and plaques spanning the body as well as lesions with rolled borders and ulcerated bases, indicative of BCCs, aid dermatologists in distinguishing benign lesions from those of NBCCS.¹

Additional differential diagnoses are required to distinguish NBCCS from other similar inherited skin disorders that are characterized by BCCs. The presence of multiple incidental BCCs early in life remains a histopathologic clue for NBCCS diagnosis, as opposed to Rombo syndrome, in which BCCs often develop in adulthood.^2,4 In addition, although both Bazex syndrome and Muir-Torre syndrome are characterized by the early onset of BCCs, the lack of skeletal abnormalities and palmar and plantar pitting distinguish these entities from NBCCS.^2,4 Furthermore, though psoriasis also can present on the scalp, clinical presentation often includes well-demarcated and symmetric plaques that are erythematous and silvery, all of which were not present in our patient and typically are not seen in NBCCS.⁵

The recommended treatment of NBCCS is vismodegib, a specific oncogene inhibitor. This medication suppresses the hedgehog signaling pathway by inhibiting smoothened oncogenes and downstream target molecules, thereby decreasing tumor proliferation.⁶ In doing so, vismodegib inhibits the development of new BCCs while reducing the burden of present ones. Additionally, vismodegib appears to effectively treat KOTs. If successful, this medication may be able to suppress KOTs in patients with NBCCS and thus facilitate surgery.⁶ Additional hedgehog inhibitors include patidegib, sonidegib, and itraconazole. Patidegib gel 2% currently is in phase 3 clinical trials for evaluation of efficacy and safety in treatment of NBCCS. Sonidegib is approved for the treatment of locally advanced BCCs in the United States and the European Union and for both locally advanced BCCs and metastatic BCCs in Switzerland and Australia.⁷ Further research is needed before recommending antifungal itraconazole for NBCCS clinical use.8 Other medications for localized areas include topical application of 5-fluorouracil and imiquimod.²

An analysis of open-access skin image datasets available to train machine-learning algorithms to identify skin cancer has revealed that darker skin types are markedly underrepresented in the databases, researchers in the United Kingdom report.

Out of 106,950 skin lesions documented in 21 open-access databases and 17 open-access atlases identified by David Wen, BMBCh, from the University of Oxford (England), and colleagues, 2,436 images contained information on Fitzpatrick skin type. Of these, “only 10 images were from individuals with Fitzpatrick skin type V, and only a single image was from an individual with Fitzpatrick skin type VI,” the researchers said. “The ethnicity of these individuals was either Brazilian or unknown.”

In two datasets containing 1,585 images with ethnicity data, “no images were from individuals with an African, Afro-Caribbean, or South Asian background,” Dr. Wen and colleagues noted. “Coupled with the geographical origins of datasets, there was massive under-representation of skin lesion images from darker-skinned populations.”

The results of their systematic review were presented at the National Cancer Research Institute Festival and published on Nov. 9, 2021, in The Lancet Digital Health. To the best of their knowledge, they wrote, this is “the first systematic review of publicly available skin lesion images comprising predominantly dermoscopic and macroscopic images available through open access datasets and atlases.”

Overall, 11 of 14 datasets (79%) were from North America, Europe, or Oceania among datasets with information on country of origin, the researchers said. Either dermoscopic images or macroscopic photographs were the only types of images available in 19 of 21 (91%) datasets. There was some variation in the clinical information available, with 81,662 images (76.4%) containing information on age, 82,848 images (77.5%) having information on gender, and 79,561 images having information about body site (74.4%).

The researchers explained that these datasets might be of limited use in a real-world setting where the images aren’t representative of the population. Artificial intelligence (AI) programs that train using images of patients with one skin type, for example, can potentially misdiagnose patients of another skin type, they said.

“AI programs hold a lot of potential for diagnosing skin cancer because it can look at pictures and quickly and cost-effectively evaluate any worrying spots on the skin,” Dr. Wen said in a press release from the NCRI Festival. “However, it’s important to know about the images and patients used to develop programs, as these influence which groups of people the programs will be most effective for in real-life settings. Research has shown that programs trained on images taken from people with lighter skin types only might not be as accurate for people with darker skin, and vice versa.”

There was also “limited information on who, how and why the images were taken,” Dr. Wen said in the release. “This has implications for the programs developed from these images, due to uncertainty around how they may perform in different groups of people, especially in those who aren’t well represented in datasets, such as those with darker skin. This can potentially lead to the exclusion or even harm of these groups from AI technologies.”

While there are no current guidelines for developing skin image datasets, quality standards are needed, according to the researchers.

“Ensuring equitable digital health includes building unbiased, representative datasets to ensure that the algorithms that are created benefit people of all backgrounds and skin types,” they concluded in the study.

Neil Steven, MBBS, MA, PhD, FRCP, an NCRI Skin Group member who was not involved with the research, stated in the press release that the results from the study by Dr. Wen and colleagues “raise concerns about the ability of AI to assist in skin cancer diagnosis, especially in a global context.”

“I hope this work will continue and help ensure that the progress we make in using AI in medicine will benefit all patients, recognizing that human skin color is highly diverse,” said Dr. Steven, honorary consultant in medical oncology at University Hospitals Birmingham (England) NHS Foundation Trust.

‘We need more images of everybody’

Dermatologist Adewole Adamson, MD, MPP, assistant professor in the department of internal medicine (division of dermatology) at the University of Texas at Austin, said in an interview that a “major potential downside” of algorithms not trained on diverse datasets is the potential for incorrect diagnoses.

“The harms of algorithms used for diagnostic purposes in the skin can be particularly significant because of the scalability of this technology. A lot of thought needs to be put into how these algorithms are developed and tested,” said Dr. Adamson, who reviewed the manuscript of The Lancet Digital Health study but was not involved with the research.

He referred to the results of a recently published study in JAMA Dermatology, which found that only 10% of studies used to develop or test deep-learning algorithms contained metadata on skin tone. “Furthermore, most datasets are from countries where darker skin types are not represented. [These] algorithms therefore likely underperform on people of darker skin types and thus, users should be wary,” Dr. Adamson said.

A consensus guideline should be developed for public AI algorithms, he said, which should have metadata containing information on sex, race/ethnicity, geographic location, skin type, and part of the body. “This distribution should also be reported in any publication of an algorithm so that users can see if the distribution of the population in the training data mirrors that of the population in which it is intended to be used,” he added.

Adam Friedman, MD, professor and chair of dermatology at George Washington University, Washington, who was not involved with the research, said that, while this issue of underrepresentation has been known in dermatology for some time, the strength of the Lancet study is that it is a large study, with a message of “we need more images of everybody.”

“This is probably the broadest study looking at every possible accessible resource and taking an organized approach,” Dr. Friedman said in an interview. “But I think it also raises some important points about how we think about skin tones and how we refer to them as well with respect to misusing classification schemes that we currently have.”

While using ethnicity data and certain Fitzpatrick skin types as a proxy for darker skin is a limitation of the metadata the study authors had available, it also highlights “a broader problem with respect to lexicon regarding skin tone,” he explained.

“Skin does not have a race, it doesn’t have an ethnicity,” Dr. Friedman said.

A dataset that contains not only different skin tones but how different dermatologic conditions look across skin tones is important. “If you just look at one photo of one skin tone, you missed the fact that clinical presentations can be so polymorphic, especially because of different skin tones,” Dr. Friedman said.

“We need to keep pushing this message to ensure that images keep getting collected. We [need to] ensure that there’s quality control with these images and that we’re disseminating them in a way that everyone has access, both from self-learning, but also to teach others,” said Dr. Friedman, coeditor of a recently introduced dermatology atlas showing skin conditions in different skin tones.

Adamson reports no relevant financial relationships. Dr. Friedman is a coeditor of a dermatology atlas supported by Allergan Aesthetics and SkinBetter Science. This study was funded by NHSX and the Health Foundation. Three authors reported being paid employees of Databiology at the time of the study. The other authors reported no relevant financial relationships.

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

An analysis of open-access skin image datasets available to train machine-learning algorithms to identify skin cancer has revealed that darker skin types are markedly underrepresented in the databases, researchers in the United Kingdom report.

Out of 106,950 skin lesions documented in 21 open-access databases and 17 open-access atlases identified by David Wen, BMBCh, from the University of Oxford (England), and colleagues, 2,436 images contained information on Fitzpatrick skin type. Of these, “only 10 images were from individuals with Fitzpatrick skin type V, and only a single image was from an individual with Fitzpatrick skin type VI,” the researchers said. “The ethnicity of these individuals was either Brazilian or unknown.”

In two datasets containing 1,585 images with ethnicity data, “no images were from individuals with an African, Afro-Caribbean, or South Asian background,” Dr. Wen and colleagues noted. “Coupled with the geographical origins of datasets, there was massive under-representation of skin lesion images from darker-skinned populations.”

The results of their systematic review were presented at the National Cancer Research Institute Festival and published on Nov. 9, 2021, in The Lancet Digital Health. To the best of their knowledge, they wrote, this is “the first systematic review of publicly available skin lesion images comprising predominantly dermoscopic and macroscopic images available through open access datasets and atlases.”

Overall, 11 of 14 datasets (79%) were from North America, Europe, or Oceania among datasets with information on country of origin, the researchers said. Either dermoscopic images or macroscopic photographs were the only types of images available in 19 of 21 (91%) datasets. There was some variation in the clinical information available, with 81,662 images (76.4%) containing information on age, 82,848 images (77.5%) having information on gender, and 79,561 images having information about body site (74.4%).

The researchers explained that these datasets might be of limited use in a real-world setting where the images aren’t representative of the population. Artificial intelligence (AI) programs that train using images of patients with one skin type, for example, can potentially misdiagnose patients of another skin type, they said.

“AI programs hold a lot of potential for diagnosing skin cancer because it can look at pictures and quickly and cost-effectively evaluate any worrying spots on the skin,” Dr. Wen said in a press release from the NCRI Festival. “However, it’s important to know about the images and patients used to develop programs, as these influence which groups of people the programs will be most effective for in real-life settings. Research has shown that programs trained on images taken from people with lighter skin types only might not be as accurate for people with darker skin, and vice versa.”

There was also “limited information on who, how and why the images were taken,” Dr. Wen said in the release. “This has implications for the programs developed from these images, due to uncertainty around how they may perform in different groups of people, especially in those who aren’t well represented in datasets, such as those with darker skin. This can potentially lead to the exclusion or even harm of these groups from AI technologies.”

While there are no current guidelines for developing skin image datasets, quality standards are needed, according to the researchers.

“Ensuring equitable digital health includes building unbiased, representative datasets to ensure that the algorithms that are created benefit people of all backgrounds and skin types,” they concluded in the study.

Neil Steven, MBBS, MA, PhD, FRCP, an NCRI Skin Group member who was not involved with the research, stated in the press release that the results from the study by Dr. Wen and colleagues “raise concerns about the ability of AI to assist in skin cancer diagnosis, especially in a global context.”

“I hope this work will continue and help ensure that the progress we make in using AI in medicine will benefit all patients, recognizing that human skin color is highly diverse,” said Dr. Steven, honorary consultant in medical oncology at University Hospitals Birmingham (England) NHS Foundation Trust.

‘We need more images of everybody’

Dermatologist Adewole Adamson, MD, MPP, assistant professor in the department of internal medicine (division of dermatology) at the University of Texas at Austin, said in an interview that a “major potential downside” of algorithms not trained on diverse datasets is the potential for incorrect diagnoses.

“The harms of algorithms used for diagnostic purposes in the skin can be particularly significant because of the scalability of this technology. A lot of thought needs to be put into how these algorithms are developed and tested,” said Dr. Adamson, who reviewed the manuscript of The Lancet Digital Health study but was not involved with the research.

He referred to the results of a recently published study in JAMA Dermatology, which found that only 10% of studies used to develop or test deep-learning algorithms contained metadata on skin tone. “Furthermore, most datasets are from countries where darker skin types are not represented. [These] algorithms therefore likely underperform on people of darker skin types and thus, users should be wary,” Dr. Adamson said.

A consensus guideline should be developed for public AI algorithms, he said, which should have metadata containing information on sex, race/ethnicity, geographic location, skin type, and part of the body. “This distribution should also be reported in any publication of an algorithm so that users can see if the distribution of the population in the training data mirrors that of the population in which it is intended to be used,” he added.

Adam Friedman, MD, professor and chair of dermatology at George Washington University, Washington, who was not involved with the research, said that, while this issue of underrepresentation has been known in dermatology for some time, the strength of the Lancet study is that it is a large study, with a message of “we need more images of everybody.”

“This is probably the broadest study looking at every possible accessible resource and taking an organized approach,” Dr. Friedman said in an interview. “But I think it also raises some important points about how we think about skin tones and how we refer to them as well with respect to misusing classification schemes that we currently have.”

While using ethnicity data and certain Fitzpatrick skin types as a proxy for darker skin is a limitation of the metadata the study authors had available, it also highlights “a broader problem with respect to lexicon regarding skin tone,” he explained.

“Skin does not have a race, it doesn’t have an ethnicity,” Dr. Friedman said.

A dataset that contains not only different skin tones but how different dermatologic conditions look across skin tones is important. “If you just look at one photo of one skin tone, you missed the fact that clinical presentations can be so polymorphic, especially because of different skin tones,” Dr. Friedman said.

“We need to keep pushing this message to ensure that images keep getting collected. We [need to] ensure that there’s quality control with these images and that we’re disseminating them in a way that everyone has access, both from self-learning, but also to teach others,” said Dr. Friedman, coeditor of a recently introduced dermatology atlas showing skin conditions in different skin tones.

Adamson reports no relevant financial relationships. Dr. Friedman is a coeditor of a dermatology atlas supported by Allergan Aesthetics and SkinBetter Science. This study was funded by NHSX and the Health Foundation. Three authors reported being paid employees of Databiology at the time of the study. The other authors reported no relevant financial relationships.

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

An analysis of open-access skin image datasets available to train machine-learning algorithms to identify skin cancer has revealed that darker skin types are markedly underrepresented in the databases, researchers in the United Kingdom report.

Out of 106,950 skin lesions documented in 21 open-access databases and 17 open-access atlases identified by David Wen, BMBCh, from the University of Oxford (England), and colleagues, 2,436 images contained information on Fitzpatrick skin type. Of these, “only 10 images were from individuals with Fitzpatrick skin type V, and only a single image was from an individual with Fitzpatrick skin type VI,” the researchers said. “The ethnicity of these individuals was either Brazilian or unknown.”

In two datasets containing 1,585 images with ethnicity data, “no images were from individuals with an African, Afro-Caribbean, or South Asian background,” Dr. Wen and colleagues noted. “Coupled with the geographical origins of datasets, there was massive under-representation of skin lesion images from darker-skinned populations.”

The results of their systematic review were presented at the National Cancer Research Institute Festival and published on Nov. 9, 2021, in The Lancet Digital Health. To the best of their knowledge, they wrote, this is “the first systematic review of publicly available skin lesion images comprising predominantly dermoscopic and macroscopic images available through open access datasets and atlases.”

Overall, 11 of 14 datasets (79%) were from North America, Europe, or Oceania among datasets with information on country of origin, the researchers said. Either dermoscopic images or macroscopic photographs were the only types of images available in 19 of 21 (91%) datasets. There was some variation in the clinical information available, with 81,662 images (76.4%) containing information on age, 82,848 images (77.5%) having information on gender, and 79,561 images having information about body site (74.4%).

The researchers explained that these datasets might be of limited use in a real-world setting where the images aren’t representative of the population. Artificial intelligence (AI) programs that train using images of patients with one skin type, for example, can potentially misdiagnose patients of another skin type, they said.

“AI programs hold a lot of potential for diagnosing skin cancer because it can look at pictures and quickly and cost-effectively evaluate any worrying spots on the skin,” Dr. Wen said in a press release from the NCRI Festival. “However, it’s important to know about the images and patients used to develop programs, as these influence which groups of people the programs will be most effective for in real-life settings. Research has shown that programs trained on images taken from people with lighter skin types only might not be as accurate for people with darker skin, and vice versa.”

There was also “limited information on who, how and why the images were taken,” Dr. Wen said in the release. “This has implications for the programs developed from these images, due to uncertainty around how they may perform in different groups of people, especially in those who aren’t well represented in datasets, such as those with darker skin. This can potentially lead to the exclusion or even harm of these groups from AI technologies.”

While there are no current guidelines for developing skin image datasets, quality standards are needed, according to the researchers.

“Ensuring equitable digital health includes building unbiased, representative datasets to ensure that the algorithms that are created benefit people of all backgrounds and skin types,” they concluded in the study.

Neil Steven, MBBS, MA, PhD, FRCP, an NCRI Skin Group member who was not involved with the research, stated in the press release that the results from the study by Dr. Wen and colleagues “raise concerns about the ability of AI to assist in skin cancer diagnosis, especially in a global context.”

“I hope this work will continue and help ensure that the progress we make in using AI in medicine will benefit all patients, recognizing that human skin color is highly diverse,” said Dr. Steven, honorary consultant in medical oncology at University Hospitals Birmingham (England) NHS Foundation Trust.

‘We need more images of everybody’

Dermatologist Adewole Adamson, MD, MPP, assistant professor in the department of internal medicine (division of dermatology) at the University of Texas at Austin, said in an interview that a “major potential downside” of algorithms not trained on diverse datasets is the potential for incorrect diagnoses.

“The harms of algorithms used for diagnostic purposes in the skin can be particularly significant because of the scalability of this technology. A lot of thought needs to be put into how these algorithms are developed and tested,” said Dr. Adamson, who reviewed the manuscript of The Lancet Digital Health study but was not involved with the research.

He referred to the results of a recently published study in JAMA Dermatology, which found that only 10% of studies used to develop or test deep-learning algorithms contained metadata on skin tone. “Furthermore, most datasets are from countries where darker skin types are not represented. [These] algorithms therefore likely underperform on people of darker skin types and thus, users should be wary,” Dr. Adamson said.

A consensus guideline should be developed for public AI algorithms, he said, which should have metadata containing information on sex, race/ethnicity, geographic location, skin type, and part of the body. “This distribution should also be reported in any publication of an algorithm so that users can see if the distribution of the population in the training data mirrors that of the population in which it is intended to be used,” he added.

Adam Friedman, MD, professor and chair of dermatology at George Washington University, Washington, who was not involved with the research, said that, while this issue of underrepresentation has been known in dermatology for some time, the strength of the Lancet study is that it is a large study, with a message of “we need more images of everybody.”

“This is probably the broadest study looking at every possible accessible resource and taking an organized approach,” Dr. Friedman said in an interview. “But I think it also raises some important points about how we think about skin tones and how we refer to them as well with respect to misusing classification schemes that we currently have.”

While using ethnicity data and certain Fitzpatrick skin types as a proxy for darker skin is a limitation of the metadata the study authors had available, it also highlights “a broader problem with respect to lexicon regarding skin tone,” he explained.

“Skin does not have a race, it doesn’t have an ethnicity,” Dr. Friedman said.

A dataset that contains not only different skin tones but how different dermatologic conditions look across skin tones is important. “If you just look at one photo of one skin tone, you missed the fact that clinical presentations can be so polymorphic, especially because of different skin tones,” Dr. Friedman said.

“We need to keep pushing this message to ensure that images keep getting collected. We [need to] ensure that there’s quality control with these images and that we’re disseminating them in a way that everyone has access, both from self-learning, but also to teach others,” said Dr. Friedman, coeditor of a recently introduced dermatology atlas showing skin conditions in different skin tones.

Adamson reports no relevant financial relationships. Dr. Friedman is a coeditor of a dermatology atlas supported by Allergan Aesthetics and SkinBetter Science. This study was funded by NHSX and the Health Foundation. Three authors reported being paid employees of Databiology at the time of the study. The other authors reported no relevant financial relationships.

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