Choice of starting dose for molecular targeted agents (MTA) evaluated in phase I cancer clinical trials (P1T)

Abstract

3586 Background: One tenth of the LD10 (dose lethal to 10% of mice) has been conventionally used to identify a safe human starting dose (SD) in P1T with cytotoxic agents. There is no consensus on which preclinical models and parameters would best predict for a safe SD for MTA. Methods: We reviewed a total of 61 MTA evaluated in 86 P1T for which preclinical data were available. Our objectives were to evaluate which animal models and parameters were used to establish a safe human SD, to compare the SD with the maximum administered dose (MAD), and to calculate the number of dose levels required to achieve MAD. We also analyzed if higher SD would shorten dose escalation while guaranteeing safety. Results: Of the 86 P1T reviewed, the SD was based on data from rodent (33), non- rodent (30), healthy human volunteers (10), other ongoing P1T (5) and not specified (8). For the 44 agents in which animal data were available, 25 followed a modified Fibonacci dose escalation design and 19 used alternative designs. To assure comparability, we only analyzed the 25 agents which utilized the modified Fibonacci design in 29 P1T. Among these 29 P1T, 17 used rodent parameters to determine SD (13 used LD10, 4 used toxic dose low [TDL]), while 12 used non-rodent parameters (6 used no observed adverse event level [NOAEL], 4 used LD10, 2 used TDL). The median ratio to MAD to SD for all 25 agents was 11.4 (1- 45): 11 and 12 for agents with SD based on rodent and non-rodent data respectively. The median number of dose levels to reach MAD from SD was 5.3 (1–9). The table below illustrates the safety and number of dose levels to reach MAD, with varying SD (1x, 2x and 3x SD). Conclusions: Unlike cytotoxic agents, the derivation of SD for MTA is based on different practices using a variety of preclinical parameters. No significant differences are observed in the number of dose levels to reach MAD using SD based on rodent versus non-rodent data. Non-rodent data appear to better predict for a safe human SD than rodent data in P1T of MTA. Starting dose No of dose levels to reach MAD Rodents median (range) No of dose levels to reach MAD Non-rodents median (range) No of unsafe* trials Rodents (n=17) vs non-rodents (n=12) No of unsafe agents* Rodents (n=14) vs non-rodents (n=11) 1x 5.2 (1–9) 5.3 (2–9) 4 vs. 0 4 vs. 0 2x 3.2 (1–6) 3.6 (1–6) 4 vs. 1 4 vs. 1 3x 2.4 (1–5) 2.4 (1–5) 8 vs. 3 7 vs. 3 * Unsafe trial or agent is arbitrarily defined if SD = MAD No significant financial relationships to disclose.