CRITICAL LIMITS TO DEFINE A LAB ADVERSE EVENT DURING PHASE-I STUDIES - A STUDY IN 1134 SUBJECTS

Objective: The first goal of phase I drug development is the determina tion of maximal tolerated dose, which must be established by case-by-c ase analysis, sometimes using a laboratory adverse event. Since no acc urate rule defining lab adverse events, has been validated yet, we pro pose a new ''combined method'' based on combination of two thresholds: inclusion values and magnitude of variation. Using this combined meth od, the label ''lab adverse event'' is applied if any lab value exceed s the inclusion threshold and is associated with a variation from base line exceeding the variation threshold defined from reference change l imit. Thus, this study aimed to test this combined method on a large h ealthy volunteer population, studied in 19 phase I centres worldwide, and on five lab parameters: alanine amino transferase, aspartate amino transferase, alkaline phosphatases, creatinine and polymorphonuclear leukocytes. Methods: The inclusion threshold from each center was used . Reference change limits were defined from volunteers previously incl uded in comparable studies and were expressed as absolute values: incr eases of 10 IU.l(-1) for alanine amino transferase or aspartate amino transferase, 15 IU.l(-1) for alkaline phosphatases, 15 mu mol.l(-1) fo r creatinine and a 0.34 10(9).l(-1) decrease for polymorphonuclear leu kocytes. Comparison between the ''combined method'' and a normal range method was made using positive predictive value and a ratio between r elevant and irrelevant results. This application was implemented in al l young healthy volunteers (1134) included in 38 phase I studies spons ored by Rhone Poulenc Rorer from 1991 to 1993. Results: Seventy seven subjects (6.7%) were indicated in final study reports as having a lab adverse event (reference group). Of 179 subjects with lab abnormalitie s defined by the normal range method, 77 belonged to the reference gro up, inducing a poor 0.43 positive predictive value. Of ninety subjects with lab adverse events defined by the ''combined method'', seventy-f ive belonged to the reference group, inducing a two-fold higher 0.83 p ositive predictive value. The combined method produced a high ratio of relevant/irrelevant results (5 = 75/15) compared with the low ratio ( 0.76 = 77/102) achieved using the normal range method. Conclusion: Thi s new ''combined method'', leading to a better definition of lab adver se event, seems an accurate and useful tool for routine case-by-case a nalysis within phase I drug development studies.