Estensione del Disegno Continual Reassessment Methods (CRM) allo studio di misure ripetute di tossicità in studi Clinici oncologici di fase I .

A modified version of Continual Reassessment Methods for longitudinal data in Phase I cancer Clinical Trials. Background: In Phase I studies, the Maximum Tolerated Dose (MTD) of a new drugs is determined on the basis of dose limiting toxicities (DLT) observed during the first cycles of treatment. Cancer therapies are however administered in repeated courses with a risk of cumulative toxicities that could compromise the development of the drugs. Information about DLT occurred during the entire treatment period need to be included in the dose-response model. To this aim, we propose a phase I design that is a extension to the Modified Continual Reassessment Method (M-CRM) proposed by Goodman et al.. Method: We propose two hierarchical dose-toxicity logistic models for binary correlated responses, with one and two parameters respectively. Data on a combination phase I Italian study are used to evaluate differences in the MTD between the traditional design (3+3) used in the study protocol, M-CRM and CRM with mixed models. For the analysis DLT was defined as any toxicity of grade>2 (CTCAE v.3.1). Results: Enrolled patients were sequentially assigned to four dose levels (2.5;5;7.5;10 mg/day). 3+3 patients were assigned each dose except for the second dose (only 3 patients). The MTD from the study was the higher dose level. The number of toxicities for each dose were 0, 0, 2, 2. Results from the logistic, one parameter, mixed model showed, on the average, an high risk of toxicity respect to fixed effect models (M-CRM) on the first cycle of therapy. The third dose level was identified as MTD (probability p=0.325, IC80%[0.23-0.42]). Conclusions: When a cumulative toxicity effect is supposed, the use of a longitudinal design could be appropriate. To avoid a delay in the study duration a limit must be fixed in advance on the number of cycle on which the dose escalation is based. Operating characteristics of the proposed design could be evaluated through a simulation study.