Background: The conventional phase I trial design yields an estimate of the maximum tolerated dose (MTD) of a new drug defined from treatment toxic effects observed in a small heterogenous cohort of patients. The MTD is specific for those patients treated in the study and may not be reproducible in another patient series, a limitation known as cohort dependency. Purpose: We conducted an opinion survey of phase I investigators in an attempt to characterize physician attitudes and clinical practices regarding assessment of risk of toxic effects in patients. Methods: A physician opinion survey of scaled multiple choice and open-ended questions was distributed to oncologists (faculty and fellows) at National Cancer Institute (NCI)-funded phase I contract institutions. The target sample was narrowed to these specific institutions because of their experience in conducting various phase I trials. Each NCI-funded phase I contractor received an instruction sheet and 25 surveys and envelopes that were made available to oncologists enrolling patients in phase I trials. Results: Of the 75 surveys distributed, 67 (89%) were returned. Mast respondents agreed that unexplained variability in toxicity exists in phase I trials. However, opinion varied considerably among five participating institutions (two-sided P =.001). Informal scoring of patients for toxicity risk prior to treatment was a common practice (85% overall), although the prevalence of this practice varied somewhat among the institutions (two-sided P = .01). Physicians' opinions were mixed regarding the practice of becoming increasingly selective in enrolling patients as the MTD was approached, with strong disagreement noted among institutions (two-sided P =.001). Given background on a hypothetical phase I trial, respondents were asked to rank the usefulness of 27 patient factors for predicting the risk of dose-limiting leukopenia. Eight factors were perceived as strongly informative: Eastern Cooperative Oncology Group performance status of 2 or worse, recent weight loss of more than 10%, multiple bone marrow metastases, two or more prior chemotherapy regimens, history of treatment toxic effects, and prior treatment with carboplatin, mitomycin, or nitrosoureas. Conclusions: Informal assessment of the risk of toxic effects in a patient was found to be common practice among oncologists who enroll patients on phase I trials, and there was strong agreement on the usefulness of several key patient factors for such assessments. Interestingly, interinstitutional variation in opinions and practices were noted, specifically regarding patient-selection bias. Follow-up studies are required to establish if physician-determined assessments of patient risk are consistent and accurate.
