User login
An AI-based skin assessment app may drive up healthcare visits for benign lesions, with unclear benefits for skin cancer detection, a Dutch clinical trial has found.
The trial, of nearly 20,000 patients in one health insurance plan, found that those given free access to the app were no more likely to be diagnosed with skin cancer over 1 year than participants assigned to a control group with no app access. They were, however, more likely to make healthcare visits for benign skin lesions.
The results came as a surprise, lead researcher Marlies Wakkee, MD, PhD, said during a presentation at the European Association of Dermato-Oncology (EADO) Congress 2026, held in Prague, Czech Republic.
“We were a bit flabbergasted,” said Wakkee, of Erasmus MC in Rotterdam, Netherlands. “We were, of course, expecting that those who would use this intervention app would have more skin cancer diagnoses than those who did not.”
She did, however, point to a potential reason for the lack of benefit: A deeper look at the data suggested that participants in the control group might have been particularly motivated to see their doctor for suspicious skin growths.
Can AI Apps Fill a Gap?
Wakkee pointed out that routine skin cancer screening via clinical skin examination is considered infeasible in many countries. Current guidance from the US Preventive Services Task Force says there is insufficient evidence to assess the balance of benefits and harms from widespread screening.
A plethora of AI-based skin assessment apps have entered the market in recent years, Wakkee said, and in theory, they have the potential to aid in earlier skin cancer diagnosis. But, she added, the technology also comes with potential harms, ranging from spurring healthcare visits for benign lesions to missing true cancers.
The current trial focused on the SkinVision app. It relies on a convolutional neural network to analyze images of skin lesions captured by the user’s smartphone and provides risk assessments of low, medium or high; a tele-dermatology team is available for support.
The app has been reimbursed in Netherlands via health insurance companies since 2019, and by 2021, it was available to 2.2 million insurees, with an uptake of about 1%, according to Wakkee.
In a previous study, the researchers used insurance claims data to study 18,960 app users and compare them with 56,880 nonusers. They found that app use was associated with an increased likelihood of being diagnosed with cutaneous malignancies and premalignancies but also benign tumors and nevi.
“So there’s a group in there that just is very worried about their skin,” Wakkee said.
To investigate further, her team conducted the SPOT-study, a randomized controlled trial in which roughly 226,000 adults covered by a Dutch nonprofit health insurance provider were invited to take part.
Of those, just over 19,000 agreed and were randomly assigned to either an intervention group that had free access to the skin app for 12 months or a control group that had no access. They were told that if they had any skin lesions they were worried about, they should visit their general practitioner.
During that period, the study found there was no significant difference in rates of histologically verified melanoma between the intervention and control groups, at 0.26% vs 0.31% — a risk difference of -0.05% (P = .68).
Similarly, the groups showed no difference in rates of any type of skin cancer, including squamous cell and basal cell carcinomas, at 2.66% in the intervention group vs 2.27% in the control group (P = .10). Rates of premalignant lesions were also comparable (6.9% vs 6.3%; P = .23).
The researchers then examined participants’ claims data to look at healthcare visits for benign skin lesions. There, app users did have a significantly higher rate, at 3.9% vs 2.6% (P < .001).
A Case of Inherent Bias?
The lack of benefit for skin cancer detection prompted the researchers to view the data from a different angle. They compared their trial participants with over 200,000 nonresponders from the health insurance plan. And that’s when a difference emerged.
Overall, trial participants were nearly three times more likely to have a skin premalignancy or malignancy diagnosed during that period, at 6.7% vs 2.4% (P < .001).
Wakkee said that because trial participants were told that the study aimed to gauge “the potential impact of this technology” in assessing skin lesions, that might have created an inherent bias. Participants assigned to the control group may have been motivated to have any worrisome skin growth checked out by their general practitioners.
In addition, Wakkee cautioned that the 12-month results are based on a small number of cancer cases, making it difficult to draw firm conclusions about the app’s performance. The trial has a second phase, where both groups were given free access to the app for 12 months, then followed for an additional 24 months.
Longer-term data are needed, Wakkee noted, in part to see whether people’s app usage changes over time.
Future Questions
Audience members at the presentation raised questions about how AI-based apps could be best deployed for skin cancer detection — including whether they might work better in the hands of clinicians rather than patients.
Wakkee said that clinicians would need a more advanced technology than that included in the app used in this trial. But future studies, she said, will look at whether the app can be used in a more targeted way, specifically, as a triage tool for people who are already concerned about something on their skin, to help them decide if they need to visit their doctor.
One presentation attendee wondered whether people given a low-risk result by the app were likely to be reassured or still make an appointment.
Wakkee said her team has begun to dig into that question. In a pilot study, 50 patients who wanted to see their general practitioner for a skin lesion were asked: If you received a low-risk rating on the skin app, would you still visit your doctor?
“Half of them said they would stay at home,” Wakkee said. She added, however, that her team is conducting a follow-up study to see what people actually do.
The trial was supported by SkinVision. The researchers declared having no relevant financial relationships.
This article was previously published by Medscape.
An AI-based skin assessment app may drive up healthcare visits for benign lesions, with unclear benefits for skin cancer detection, a Dutch clinical trial has found.
The trial, of nearly 20,000 patients in one health insurance plan, found that those given free access to the app were no more likely to be diagnosed with skin cancer over 1 year than participants assigned to a control group with no app access. They were, however, more likely to make healthcare visits for benign skin lesions.
The results came as a surprise, lead researcher Marlies Wakkee, MD, PhD, said during a presentation at the European Association of Dermato-Oncology (EADO) Congress 2026, held in Prague, Czech Republic.
“We were a bit flabbergasted,” said Wakkee, of Erasmus MC in Rotterdam, Netherlands. “We were, of course, expecting that those who would use this intervention app would have more skin cancer diagnoses than those who did not.”
She did, however, point to a potential reason for the lack of benefit: A deeper look at the data suggested that participants in the control group might have been particularly motivated to see their doctor for suspicious skin growths.
Can AI Apps Fill a Gap?
Wakkee pointed out that routine skin cancer screening via clinical skin examination is considered infeasible in many countries. Current guidance from the US Preventive Services Task Force says there is insufficient evidence to assess the balance of benefits and harms from widespread screening.
A plethora of AI-based skin assessment apps have entered the market in recent years, Wakkee said, and in theory, they have the potential to aid in earlier skin cancer diagnosis. But, she added, the technology also comes with potential harms, ranging from spurring healthcare visits for benign lesions to missing true cancers.
The current trial focused on the SkinVision app. It relies on a convolutional neural network to analyze images of skin lesions captured by the user’s smartphone and provides risk assessments of low, medium or high; a tele-dermatology team is available for support.
The app has been reimbursed in Netherlands via health insurance companies since 2019, and by 2021, it was available to 2.2 million insurees, with an uptake of about 1%, according to Wakkee.
In a previous study, the researchers used insurance claims data to study 18,960 app users and compare them with 56,880 nonusers. They found that app use was associated with an increased likelihood of being diagnosed with cutaneous malignancies and premalignancies but also benign tumors and nevi.
“So there’s a group in there that just is very worried about their skin,” Wakkee said.
To investigate further, her team conducted the SPOT-study, a randomized controlled trial in which roughly 226,000 adults covered by a Dutch nonprofit health insurance provider were invited to take part.
Of those, just over 19,000 agreed and were randomly assigned to either an intervention group that had free access to the skin app for 12 months or a control group that had no access. They were told that if they had any skin lesions they were worried about, they should visit their general practitioner.
During that period, the study found there was no significant difference in rates of histologically verified melanoma between the intervention and control groups, at 0.26% vs 0.31% — a risk difference of -0.05% (P = .68).
Similarly, the groups showed no difference in rates of any type of skin cancer, including squamous cell and basal cell carcinomas, at 2.66% in the intervention group vs 2.27% in the control group (P = .10). Rates of premalignant lesions were also comparable (6.9% vs 6.3%; P = .23).
The researchers then examined participants’ claims data to look at healthcare visits for benign skin lesions. There, app users did have a significantly higher rate, at 3.9% vs 2.6% (P < .001).
A Case of Inherent Bias?
The lack of benefit for skin cancer detection prompted the researchers to view the data from a different angle. They compared their trial participants with over 200,000 nonresponders from the health insurance plan. And that’s when a difference emerged.
Overall, trial participants were nearly three times more likely to have a skin premalignancy or malignancy diagnosed during that period, at 6.7% vs 2.4% (P < .001).
Wakkee said that because trial participants were told that the study aimed to gauge “the potential impact of this technology” in assessing skin lesions, that might have created an inherent bias. Participants assigned to the control group may have been motivated to have any worrisome skin growth checked out by their general practitioners.
In addition, Wakkee cautioned that the 12-month results are based on a small number of cancer cases, making it difficult to draw firm conclusions about the app’s performance. The trial has a second phase, where both groups were given free access to the app for 12 months, then followed for an additional 24 months.
Longer-term data are needed, Wakkee noted, in part to see whether people’s app usage changes over time.
Future Questions
Audience members at the presentation raised questions about how AI-based apps could be best deployed for skin cancer detection — including whether they might work better in the hands of clinicians rather than patients.
Wakkee said that clinicians would need a more advanced technology than that included in the app used in this trial. But future studies, she said, will look at whether the app can be used in a more targeted way, specifically, as a triage tool for people who are already concerned about something on their skin, to help them decide if they need to visit their doctor.
One presentation attendee wondered whether people given a low-risk result by the app were likely to be reassured or still make an appointment.
Wakkee said her team has begun to dig into that question. In a pilot study, 50 patients who wanted to see their general practitioner for a skin lesion were asked: If you received a low-risk rating on the skin app, would you still visit your doctor?
“Half of them said they would stay at home,” Wakkee said. She added, however, that her team is conducting a follow-up study to see what people actually do.
The trial was supported by SkinVision. The researchers declared having no relevant financial relationships.
This article was previously published by Medscape.
An AI-based skin assessment app may drive up healthcare visits for benign lesions, with unclear benefits for skin cancer detection, a Dutch clinical trial has found.
The trial, of nearly 20,000 patients in one health insurance plan, found that those given free access to the app were no more likely to be diagnosed with skin cancer over 1 year than participants assigned to a control group with no app access. They were, however, more likely to make healthcare visits for benign skin lesions.
The results came as a surprise, lead researcher Marlies Wakkee, MD, PhD, said during a presentation at the European Association of Dermato-Oncology (EADO) Congress 2026, held in Prague, Czech Republic.
“We were a bit flabbergasted,” said Wakkee, of Erasmus MC in Rotterdam, Netherlands. “We were, of course, expecting that those who would use this intervention app would have more skin cancer diagnoses than those who did not.”
She did, however, point to a potential reason for the lack of benefit: A deeper look at the data suggested that participants in the control group might have been particularly motivated to see their doctor for suspicious skin growths.
Can AI Apps Fill a Gap?
Wakkee pointed out that routine skin cancer screening via clinical skin examination is considered infeasible in many countries. Current guidance from the US Preventive Services Task Force says there is insufficient evidence to assess the balance of benefits and harms from widespread screening.
A plethora of AI-based skin assessment apps have entered the market in recent years, Wakkee said, and in theory, they have the potential to aid in earlier skin cancer diagnosis. But, she added, the technology also comes with potential harms, ranging from spurring healthcare visits for benign lesions to missing true cancers.
The current trial focused on the SkinVision app. It relies on a convolutional neural network to analyze images of skin lesions captured by the user’s smartphone and provides risk assessments of low, medium or high; a tele-dermatology team is available for support.
The app has been reimbursed in Netherlands via health insurance companies since 2019, and by 2021, it was available to 2.2 million insurees, with an uptake of about 1%, according to Wakkee.
In a previous study, the researchers used insurance claims data to study 18,960 app users and compare them with 56,880 nonusers. They found that app use was associated with an increased likelihood of being diagnosed with cutaneous malignancies and premalignancies but also benign tumors and nevi.
“So there’s a group in there that just is very worried about their skin,” Wakkee said.
To investigate further, her team conducted the SPOT-study, a randomized controlled trial in which roughly 226,000 adults covered by a Dutch nonprofit health insurance provider were invited to take part.
Of those, just over 19,000 agreed and were randomly assigned to either an intervention group that had free access to the skin app for 12 months or a control group that had no access. They were told that if they had any skin lesions they were worried about, they should visit their general practitioner.
During that period, the study found there was no significant difference in rates of histologically verified melanoma between the intervention and control groups, at 0.26% vs 0.31% — a risk difference of -0.05% (P = .68).
Similarly, the groups showed no difference in rates of any type of skin cancer, including squamous cell and basal cell carcinomas, at 2.66% in the intervention group vs 2.27% in the control group (P = .10). Rates of premalignant lesions were also comparable (6.9% vs 6.3%; P = .23).
The researchers then examined participants’ claims data to look at healthcare visits for benign skin lesions. There, app users did have a significantly higher rate, at 3.9% vs 2.6% (P < .001).
A Case of Inherent Bias?
The lack of benefit for skin cancer detection prompted the researchers to view the data from a different angle. They compared their trial participants with over 200,000 nonresponders from the health insurance plan. And that’s when a difference emerged.
Overall, trial participants were nearly three times more likely to have a skin premalignancy or malignancy diagnosed during that period, at 6.7% vs 2.4% (P < .001).
Wakkee said that because trial participants were told that the study aimed to gauge “the potential impact of this technology” in assessing skin lesions, that might have created an inherent bias. Participants assigned to the control group may have been motivated to have any worrisome skin growth checked out by their general practitioners.
In addition, Wakkee cautioned that the 12-month results are based on a small number of cancer cases, making it difficult to draw firm conclusions about the app’s performance. The trial has a second phase, where both groups were given free access to the app for 12 months, then followed for an additional 24 months.
Longer-term data are needed, Wakkee noted, in part to see whether people’s app usage changes over time.
Future Questions
Audience members at the presentation raised questions about how AI-based apps could be best deployed for skin cancer detection — including whether they might work better in the hands of clinicians rather than patients.
Wakkee said that clinicians would need a more advanced technology than that included in the app used in this trial. But future studies, she said, will look at whether the app can be used in a more targeted way, specifically, as a triage tool for people who are already concerned about something on their skin, to help them decide if they need to visit their doctor.
One presentation attendee wondered whether people given a low-risk result by the app were likely to be reassured or still make an appointment.
Wakkee said her team has begun to dig into that question. In a pilot study, 50 patients who wanted to see their general practitioner for a skin lesion were asked: If you received a low-risk rating on the skin app, would you still visit your doctor?
“Half of them said they would stay at home,” Wakkee said. She added, however, that her team is conducting a follow-up study to see what people actually do.
The trial was supported by SkinVision. The researchers declared having no relevant financial relationships.
This article was previously published by Medscape.