In Phase I clinical studies, the maximum tolerated dose has to be determined by a case by case analysis sometimes using a laboratory adverse effect, e.g. an increase in alanine amino transferase (ALT). For this reason a threshold to discriminate between significant or non significant adverse changes in ALT is required particularly in Phase I studies, in order to deal with the very common "close to the limit values". Previous methods (limit of normal range or normal range plus an arbitrary margin) do not solve this problem. The authors propose a new method taking into account the threshold used as inclusion criteria for ALT (R) and the range of spontaneous variations measured under identical Phase I study conditions (V). The (R) and (V) thresholds, respectively, are defined as 50 IU.l-1 and a 50% increase, from baseline. Thus an ALT value is recognized as a "significant adverse experience" if it exceeds 50 IU.l-1 above an increase from baseline exceeding 50% of the baseline value. To highlight the value of the method, it was implemented in a one year period including 8 studies and 134 subjects. The sensitivity, specificity and positive predictive value of various methods were compared. The results showed the following: Six out of 134 subjects had significant adverse changes in ALT (4%); and all these 6 subjects were detected by the proposed new method without error. Eight subjects including two false positives, were detected by an use of the normal range limit, and only 4 were detected using, the 10% margin.(ABSTRACT TRUNCATED AT 250 WORDS)