Histologic features predictive of basal cell carcinoma recurrence: results of a multivariate analysis.

Previous univariate analysis of 30 recurrent and 74 non-recurrent basal cell carcinomas identified 6 histologic parameters predictive of recurrence: distance to the closest resection margin, growth pattern, shape of cell groups, contour of invading edge, degree of peripheral palisading, and nuclear pleomorphism. Re-analysis of the data by multivariate analysis to select the few most important independent prognostic variables identified two parameters in each of two final models: resection margin distance and growth pattern (Model 1), and resection margin distance and shape of cell groups (Model 2). Based on these variables, a logistic regression equation could be derived for each model to calculate the predicted probability of recurrence or non-recurrence.