The four-variable probability model developed by Schuchter and colleagues [1] for estimating prognosis of primary melanoma represents a clinically useful approach to a difficult problem. The model's accuracy can be increased considerably, I believe, by using it in a slightly different manner. I suggest that a probability of survival in a subgroup be used to estimate prognosis only if the entire CI and not just the point probability lie outside the intermediate probability range of 0.3 to 0.7. This would enable us to accurately estimate prognosis in 95% of 315 patients in 14 of the 32 subgroups [2]. For the other 18 subgroups, in which the CI lies entirely or partly in the intermediate range, I suggest that a new five-variable probability model be created that incorporates information on one of the two other highly predictive variables (mitotic rate or lymphocyte infiltration). With this model, 36 instead of these 18 subgroups will be formed. It is quite possible that the CI may completely move out from the intermediate range in some or all of these 36 subgroups; this allows us to more accurately estimate prognosis in patients in these subgroups. It would be of interest to know whether Schuchter and associates have data on mitotic rate and lymphocyte infiltration and whether a model that incorporates these variables does actually improve accuracy. If it does, then mention of mitotic rate or lymphocyte infiltration in pathology reports on melanoma should become standard, as is mention of tumor thickness.