PROBLEMS IN DEFINING CUTOFF POINTS OF CONTINUOUS PROGNOSTIC FACTORS -EXAMPLE OF TUMOR THICKNESS IN PRIMARY CUTANEOUS MELANOMA

Continuous prognostic factors are often categorized by defining optimi zed cutoff points. One component of criticism of this approach is the problem of multiple testing that leads to an overestimation of the tru e prognostic impact of the variable. The present study focuses on anot her crucial point by investigating the dependence of optimized cutoff points on the observed distribution of the continuous variable. The co ntinuous variable investigated was the vertical tumor thickness accord ing to Breslow, which is known to be the most important prognostic fac tor in primary melanoma. Eased on the data of 5093 patients, stratifie d random samples were drawn out of six artificially created distributi ons of tumor thickness. For each of these samples, Cox models were cal culated to explore optimized cutoff points for tumor thickness togethe r with other prognostic variables. The optimized cutoff points for tum our thickness varied considerably with the underlying distribution. Ev en in samples from the same distribution, the range of cutoff points w as amazingly broad and, for some of the distributions, covered the who le region of possible values. The results of the present study demonst rate that optimized cutoff points are extremely data dependent and var y notably even if prerequisites are constant. Therefore, if the classi fication of a continuous prognostic factor is necessary, it should not be based on the results of one single study, but on consensus discuss ions including the findings of several investigations. (C) 1997 Elsevi er Science Inc.