Related Experiment Video
Updated: Apr 28, 2026
A Melanoma Patient-Derived Xenograft Model
Published on: May 20, 2019
A model for lentigo maligna recurrence using melanocyte count as a predictive marker based upon logistic regression
Mark Gorman1, Muhammad A A Khan2, Paul C D Johnson3
1Canniesburn Plastic Surgery Unit, Glasgow, UK.
Background:
The pre-malignant skin lesion lentigo maligna (LM) presents a particular challenge. Pathologists demonstrate poor diagnostic concordance and often struggle to assess whether excision margins are truly negative. This can lead to equivocal histology reports and a lack of clear guidance with which surgeons may rationalise their surgical management plans. Based upon the biological principle that tumour burden increases the chance of recurrence, we propose a shift in diagnostic paradigm, using melanocyte count (MC) at an excision margin to predict LM recurrence.
Methods:
This retrospective study reviewed all cases of LM from a regional UK melanoma centre (1996-2011), to include 167 excisions, from 99 patients. Pathology slides were assessed for MC (blinded) at the most affected margin. Seven secondary markers of neoplasia were additionally evaluated. Logistic regression analysis was used to model the relationship between MC and recurrence.
Results:
MC is a strong predictor of LM recurrence (p < 0.0001). A regression curve predicts risk for individual MCs, which may also be divided into three risk strata; low (0-11% [MC 0-20]), intermediate (15-89% [MC 21-30]), and high risk (92-100% [MC ≥ 31]). MC misclassified 0.6% of cases in the low and high risk groups compared with 21% for pathologists, who were also equivocal for 18% of excisions. MC's inter-rater concordance was high (>0.9). The secondary factors were all independently associated with recurrence, but failed to improve predictive ability supplementary to MC.
Conclusions:
MC confidently predicts LM recurrence and is more accurate and reliable, whilst also reducing the uncertainty of current pathology assessment. Risk estimates for any given MC can be easily charted using the regression curve graph, where confidence interval and risk group boundaries demonstrate the degree of certainty associated with any given prediction. This change in approach is congruent with tumour behaviour. A recurrence 'tipping point' corresponds to the sharp risk increase across the intermediate group's narrow band of MC.
