Development and external validation study of a melanoma risk prediction model incorporating clinically assessed naevi and solar lentigines

Abstract

Background Melanoma risk prediction models could be useful for matching preventive interventions to patients' risk.Objectives To develop and validate a model for incident first-primary cutaneous melanoma using clinically assessed risk factors. Methods We used unconditional logistic regression with backward selection from the Australian Melanoma Family Study (461 cases and 329 controls) in which age, sex and city of recruitment were kept in each step, and we externally validated it using the Leeds Melanoma Case-Control Study (960 cases and 513 controls). Candidate predictors included clinically assessed whole-body naevi and solar lentigines, and self-assessed pigmentation phenotype, sun exposure, family history and history of keratinocyte cancer. We evaluated the predictive strength and discrimination of the model risk factors using odds per age-and sex-adjusted SD (OPERA) and the area under curve (AUC), and calibration using the Hosmer-Lemeshow test. Results The final model included the number of naevi ≥ 2 mm in diameter on the whole body, solar lentigines on the upper back (a six-level scale), hair colour at age 18 years and personal history of keratinocyte cancer. Naevi was the strongest risk factor; the OPERA was 3Á51 [95% confidence interval (CI) 2Á71-4Á54] in the Australian study and 2Á56 (95% CI 2Á23-2Á95) in the Leeds study. The AUC was 0Á79 (95% CI 0Á76-0Á83) in the Australian study and 0Á73 (95% CI 0Á70-0Á75) in the Leeds study. The Hosmer-Lemeshow test P-value was 0Á30 in the Australian study and < 0Á001 in the Leeds study. Conclusions This model had good discrimination and could be used by clinicians to stratify patients by melanoma risk for the targeting of preventive interventions.

Reporting Quality of Studies Developing and Validating Melanoma Prediction Models: An Assessment Based on the TRIPOD Statement

“…We included 42 studies in our TRIPOD rating. Forty studies [ 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 , 40 , 41 , 42 , 43 , 44 , 45 , 46 , 47 , 48 , 49 , 50 , 51 , 52 , 53 , 54 , 55 , 56 , 57 , 58 , 59 ] were adopted from the systematic review about risk prediction models for melanoma published in 2020, whereas the remaining two studies [ 60 , 61 ] arose from the updated literature search. Details on the information extracted from the studies are summarized in Table A1 .…”

Risk Prediction Models for Melanoma: A Systematic Review on the Heterogeneity in Model Development and Validation

“…Altogether, 24 of the 26 studies 39,40] from the two systematic reviews [13,15] were included in our analysis. Eight further studies were identified via forward snowballing [46,[49][50][51][52][53][54][55], while eight studies were found in PubMed [45,[56][57][58][59][60][61][62]. Therefore, we included 40 studies in our analysis (Figure 1).…”

Developing risk prediction models for melanoma: balancing better predictive value with ease of clinical implementation

“…Vuong et al . found strong discrimination in the initial test of the model [area under the curve (AUC) = 0·79] as well as in the external validation (AUC = 0·73) . The authors suggest that clinical assessment of whole‐body naevi and solar lentigines improved discrimination between individuals with and without melanoma as naevi were the most predictive risk factor.…”

“…In this issue of the BJD, Vuong et al . describe a prediction model they developed for incident first‐primary cutaneous melanoma using the Australian Melanoma Family Study.…”

“…Other studies, however, found better discrimination with clinically collected naevi data . Prospective studies evaluating highly predictive models using clinical and/or self‐reported data and ways to optimize their adoption in clinical practice are critical for implementing models such as the one developed by Vuong et al . to improve melanoma outcomes.…”