Choice of starting dose and escalation for phase I studies of antitumor agents.

The standard approaches to initial dose selection and dose escalation in phase I trials may be inappropriately conservative. These approaches mandate accrual of large numbers of patients, most of whom are treated at low and potentially ineffective doses. We compared the clinically determined maximum tolerable dose (MTD) with the starting dose of 45 drugs that had undergone phase I studies during the period 1977-1989. We also examined the number of dose-escalation steps required to achieve the MTD in relation to nonhematologic and hematologic dose-limiting toxicity. The median ratio of MTD to starting dose for all drugs was 20 (range, < 1-433) and the median number of dose levels studied to reach the MTD was 8 (range, 0-23). For drugs with nonhematologic dose-limiting toxicity, the median ratio of MTD to starting dose was 30 (range, 3-385) as compared with 12.8 (range, < 1-433) for those with hematologic dose-limiting toxicity (P = 0.023). The median number of dose-escalation steps required to reach the MTD was 9 (range, 2-18) for drugs with nonhematologic dose-limiting toxicity as compared with 5.5 (range, 0-23) for those with hematologic dose-limiting toxicity (P = 0.038). We also examined the response rate for 1,110 patients treated with 21 phase-I-study drugs for which response information was available. Responses were reported for 29 patients (2.6%). Among the 476 patients treated at the 3 highest dose steps, 17 responded (3.6%), which is double the response rate obtained at the lower doses (P = 0.08). It is suggested that although the usual methods for choosing starting doses and dose-escalation schemes for phase I studies are safe, they are overly conservative and delay opportunities for therapeutic benefit in phase I and subsequent phase II trials.