A review of dose-limiting events in phase I trials: antimetabolites show unpredictable relationships between dose and toxicity

Introduction: In a sample of NCIC CTG phase I trials we noted that studies of antimetabolites were frequently confounded by the occurrence of dose-lim iting toxicities (DLT) at doses well below those ultimately recommended (re commended dose, RD) for further study, necessitating frequent expansion of dose levels and usually a change to more conservative dose escalation. This slowed development, exposed patients to ineffective doses of drugs, and ra ises concerns about the safety of current trial designs which include a sin gle patient per dose level. Conversely, some patients treated at the RD may be receiving inadequate doses of anticancer drugs. To determine if this wa s a general phenomenon, we undertook a review of the results of a large gro up of phase I trials of cytotoxic agents. Methods: Starting dose (SD), numb er of dose levels, dose at first DLT (D-DLT), maximum tolerated doses (MTD, dose at which DLT is seen in two or more patients) and RD were extracted f rom the NCI-Canada phase I trial database, and from a literature survey of phase I studies published between 1985 and 1999. Combination phase I and ph ase Ib studies were excluded. Results: The review included 33 trials with a ntimetabolites and 60 with other cytotoxic agents. The median ratio D-DLT/M TD was 0.33 for antimetabolites and 0.75 for other cytotoxic agents (P < 0. 01). Similarly, the median ratio D-DLT/RD was 0.43 for antimetabolites and 1 for other cytotoxic agents (P < 0.01). The median number of dose levels t ested was nine for antimetabolites and six for other cytotoxic agents. Disc ussion: Statistically significant differences in the ratios D-DLT/MTD and D -DLT/RD between antimetabolites and other cytotoxic compounds were noted, c onfirming our initial observations that unpredictable DLT occurs earlier an d at lower dose levels in phase I clinical trials of antimetabolites than w ould be expected as compared to other classes of cytotoxic agents. Toxicity thus appears to be incompletely predicted by dose alone for antimetabolite s. DLT may occur in certain patients at doses well below the RD. Current ph ase I design may not be ideal for development of these compounds, and shoul d focus on pharmacodynamic endpoints in addition to traditional pharmacokin etic and clinical endpoints.