The approaching quantitative revolution in diagnostic anatomic pathology.

In 1989 Clark et al, 1 in a landmark article, used multivari-ate analysis and logistic regression to generate a model predicting survival in stage I primary malignant melanoma. They examined 23 attributes and found that, in addition to tumor thickness, 5, including mitotic rate, presence of tumor-infiltrating lymphocytes, anatomic site of the lesion, sex of the patient, and histologic evidence of regression of the tumor, were independent predictors of 8-year survival. Among variables found not to be independent predictors in their model were presence of plasma cells, ulceration, angiogenesis, and vascular invasion. Previous attempts to apply multivariate analysis in predicting prognosis of melanomas had been less successful (reviewed by Vollmer 2). Before the article by Clark et al, 1 survival prediction in patients with primary stage I melanoma was based, almost exclusively, on tumor thickness. Their article has been followed by a number of others (examples include articles by Barnhill et al 3 and Spatz et al 4); taken together, these studies have begun to enable us to base survival prediction of patients with this complex disease on factors in addition to tumor thickness. The article by Vollmer 5 in this issue of the Journal also applies mathematical modeling to provide a useful answer to a challenging problem in anatomic pathology in general and melanocytic lesions in particular: how to distinguish between Spitz nevi (spindle and epithelioid cell nevi, " juvenile melanomas ") and melanomas. It is by several yardsticks also a landmark paper: (1) It provides a new basis for a decision, in individual cases, based on MIB-1 staining properties, as to whether the lesion is a melanoma or a Spitz nevus. (2) It combines data from several studies to give a single distribution for Spitz nevi and another one for melanomas, thus allowing data from several studies to be integrated as if from a single study. (3) It provides a model for combining data from several studies. (4) It begins to integrate and assess data on immunohisto-chemical markers—the products of the " brown revolution " in diagnostic anatomic pathology—so that these data can be analyzed in a quantitative way to be of use in diagnosing individual lesions. Many of these articles refer to their methods or present their hypotheses and/or their methods in the precise, eloquent language of formal or at least semiformal mathematics. While this is a necessity for these papers to maintain the rigor needed to …