Diversifying dermatology: Improving skin of colour representation

Current recruitment strategies in dermatology research have segregated those with skin of colour (SOC).¹ In Australia, SOC describes a heterogenous group with non-European or mixed ancestry, in addition to First Nations Australians.² As darker skin differs substantially in biology, structure and function compared with lighter skin,² a diverse participant pool is crucial for understanding human variation and disease pathology. However, dermatology data sets are historically homogenous and tend to be biased towards those with European ancestry.¹ Factors contributing to this phenomenon include physician, investigator, industry and patient-related barriers.¹ An example of a physician-related barrier is the light skin model recognised in medical education,³ which evokes clinical uncertainty when conditions present differently in darker skin, despite the addition of SOC to the Australian dermatology training curriculum in 2017.⁴

Australia has the highest incidence of melanoma per capita worldwide, with 36.6 cases per 100,000 age standardised rate (ASR).⁵ This includes 14.1 cases per 100,000 ASR for Aboriginal and Torres Strait Islanders, and 30.4 cases per 100,000 ASR for non-Indigenous Australians.⁶ It is concerning that these data are not stratified further by ethnicity, as melanoma tends to present at more advanced stages in SOC with poorer prognosis.⁷ Acral lentiginous melanoma is the most common subtype in SOC, yet its atypical presentation in non-sun-exposed areas may contribute to oversight or misdiagnosis.⁷ Without conscious effort to incorporate all skin types in research studies, the unique manifestations, risk factors and treatment for melanoma in SOC will remain unknown, whilst mortality rates in these minority populations continue to rise.^{6, 7}

The ongoing Australian Centre of Excellence in Melanoma Imaging and Diagnosis (ACEMID) cohort study⁸ aims to create a large repository of three-dimensional total body photography to aid the development of artificial intelligence (AI) for early melanoma detection. Participants were recruited across 15 metropolitan and regional Australian sites through referrals, social media, traditional media and previous skin cancer research studies,⁸ without actively targeting those with SOC.

In our subanalysis of two ACEMID sites with completed recruitment in Brisbane, Australia, we identified 4% of participants (Table 1) with SOC (i.e. self-reported non-European ancestry). This proportion differs substantially from the Australian population, since census data report that 30% have non-European ancestry.⁹ However, the most prevalent non-European ancestries among participants with SOC were Chinese (24%), Aboriginal and Torres Strait Islander (24%) and Indian (15%), which are a reasonable reflection of Australia's multicultural society.⁹ Therefore, we can draw preliminary conclusions from this small proportion with SOC, whilst acknowledging that additional measures are required to increase the sample size.

Various classification systems exist to classify skin colour. The Fitzpatrick skin type (FST) defines six categories based on self-reported sunburn and tanning reactions, the Taylor Hyperpigmentation Scale (THS) assesses skin colour against coloured cards, and the Individual Typology Angle (ITA) quantifies pigmentation using a chromameter.¹⁰ Despite the subjectivity of FST, it suits our purpose over THS due to a physiological correlation with ultraviolet radiation and thus melanoma, as well as being more convenient than ITA for a clinical setting.¹⁰ Regarding skin types among participants with SOC, lighter skin (FST I–III) predominated (15%, 39% and 35%, respectively; Table 2), whereas darker skin (FST IV–VI) was poorly represented (9%, 2% and 0%, respectively; Table 2). Though beyond our scope, comparing these proportions to the remainder of the cohort may be worthwhile. Given the apparent skew of the SOC population towards lighter skin tones (FST I–III), untargeted recruitment fails to obtain participants with darker skin (FST IV–VI) and subsequently limits data analysis.

Physician, investigator, industry and patient-related barriers could explain why SOC is under-represented. As dermatological manifestations differ in darker skin, physicians may underestimate disease severity and less frequently recommend clinical trials to their patients. Investigators who focus on pre-existing dermatology research will automatically have bias towards European populations. On an industry level, images of darker skin are seldom integrated into public health campaigns, which is likely to lower perceptions of melanoma risk for patients with SOC. These barriers can be overcome through education on systemic failures and the implementation of codesign, to foster relationships, trust and rapport with elders and leaders of minority communities. Empowering stakeholders such as physicians, investigators and patients with SOC with enriched knowledge and recognition of melanoma in darker skin will improve mortality outcomes and data quality, facilitating the development of AI models with diagnostic accuracy for all skin types. Through active engagement of people with SOC, the absence of diversity in current data sets will improve, ensuring a fairer and more equitable future of dermatological research.

This research was conducted with the support of the Australian Cancer Research Foundation (ACRF), NHMRC Clinical Trials and Cohort Studies Grant (APP2001517) and the Australasian College of Dermatologists 2021 Scientific Research Fund Grant.

HPS is a shareholder of MoleMap NZ Limited and e-derm consult GmbH and undertakes regular teledermatological reporting for both companies. HPS is a Medical Consultant for Canfield Scientific Inc. and a Medical Advisor for First Derm.

HPS is a former EIC of the Australasian Journal of Dermatology and a co-author of this article. To minimise bias, they were excluded from all editorial decision-making related to the acceptance of this article for publication.

KK is an Editorial Board member of the Australasian Journal of Dermatology and a co-author of this article. To minimise bias, they were excluded from all editorial decision-making related to the acceptance of this article for publication.

The ACEMID cohort study was approved by the Metro South Health Human Research Ethics Committee (HREC/2019/QMS/57206) & the University of Queensland Human Research Ethics Committee (2019003077).