Classification of naevus distribution patterns in the Brisbane naevus morphology study

Abstract

There are multiple risk factors which contribute to cutaneous melanoma, including but not limited to, ultraviolet (UV) radiation exposure, the amount of freckling, skin and hair colour, skin phototype, and personal and familial melanoma history. One of the single strongest risk factors for cutaneous melanoma is a high naevus count. In the clinical setting, thorough total body examinations are hindered by the practical limitations of counting high numbers of naevi. To overcome this problem, multiple prediction models which estimate naevus count and cutaneous melanoma risk have been proposed. However, the progression and standardisation of these models has been hampered by the lack of independent validation, variation in methods and incorporated variables, and the absence of consensus as to what constitutes ‘at risk’. Consequently, rapidly and accurately identifying these ‘at risk’ individuals in the clinical setting remains difficult. Individuals with multiple naevi with distinct naevus distribution patterns have been observed clinically. The concept of naevus distribution pattern has not been thoroughly described in the literature nor have these ‘patterns’ been formally characterised. Through this study, we aim to determine if clinically distinct naevus distribution patterns, in at risk individuals, can be recognised, characterised and classified. Furthermore, whether a stratification model can be developed for the future classification of individuals based on naevus distribution pattern. 2D and 3D whole body images were captured from 1225 high risk individuals (with personal and/or familial melanoma history) selected from the Brisbane Naevus Morphology Study (BNMS) using the FotoFinder imaging system (FotoFinder Systems GmbH, Germany) or the Vectra WB360 imaging system (Canfield Scientific Inc, USA). To ensure the accurate identification of true naevi, only naevi ≥ 5 mm were included in naevus counts. Naevi ≥ 5 mm were counted on the head and neck, back, chest and abdomen, upper limbs and lower limbs. Naevus distribution clusters were derived using mclust and k–means cluster analysis based on anatomical regional naevus counts ≥ 5 mm. Naevus counts were adjusted for both age and sex. Individual pigmentation phenotypes and genotypes were also assessed. Three distinct clusters of naevus distribution pattern were identified using mclust analysis in the BNMS population. Cluster 1 (N=427) contained participants with a high total body naevus count (mean 49 ± 34 naevi ≥ 5 mm) with naevi predominantly centralised to the trunk and either the upper or lower limbs. Cluster 2 (N=552) constituted participants with a lower total body naevus count (mean 10 ± 6 naevi ≥ 5 mm) and lower relative naevus counts at each anatomical site in comparison to Cluster 1. Naevi in this cluster were more likely to be located to the trunk (back, chest and abdomen) and extremities. Cluster 3 (N=247) contained participants with the lowest naevus counts (4 ± 5 naevi ≥ 5 mm) from all three groups. In this cluster the rare naevi present were more likely to be located to the back. We have also proposed an early classification model that would enable a rapid assessment of an individual in the clinical setting, by simply performing visual overview of the number of naevi and pattern of naevus distribution across each anatomical site to stratify the patient into a given cluster. In order to further characterise these naevus distribution patterns defined within each cluster, known phenotypic and genotypic melanoma risk factors were also investigated in this population. Hair and eye colour, as well as the number of naevi were positively associated with cluster. Specifically, individuals in Cluster 1 were more likely to have light or red coloured hair and were less likely to have dark coloured eyes, compared to the other two cohorts. Cluster 3 contained a higher proportion of dark haired and dark eye coloured individuals. Correlation to naevus distribution clusters was also observed with some SNPs in pigmentation genes which have been associated with naevus count/ or pigmentation. An example of some of these genes include MC1R, MTAP, OCA2, MITF and PLA2G6. Consistently, SNPs in pigmentation genes previously associated with naevus count and/or pigmentation i.e. MC1R, MTAP, OCA2, MITF and PLA2G6 showed correlation to naevus distribution clusters. A method which classifies individuals into distinct clinical groups based on pattern of naevus distribution across anatomical sites may offer an alternate method for clinicians for rapid assessment, stratification and identification of individuals at risk in the clinical setting. Here we have described three distinct clusters on the basis of pattern of naevus distribution and proposed a novel, rapid, stepwise early classification model for easy clinical assessment by doctors of all levels of expertise. This suggests that such clustering analysis, while on the basis of naevus count and distribution, may also be a proxy for additional risk factors. Future research will aim to replicate and validate these findings in diverse populations to confirm these distinctive groups.