Yin, Jun

SummaryThe landscape of oncology drug development has recently changed with the emergence of molecularly targeted agents and immunotherapies. These new therapeutic agents appear more likely to induce multiple low or moderate grade toxicities rather than dose limiting toxicities. Various model-based dose finding designs and toxicity severity scoring systems have been proposed to account for toxicity grades, but they are difficult to implement because of the use of complicated dose–toxicity models and the requirement to refit the model at each decision of dose escalation and de-escalation. We propose a generalized Bayesian optimal interval design, gBOIN, that accommodates various existing toxicity grade scoring systems under a unified framework. As a model-assisted design, gBOIN derives its optimal decision rule on the basis of the exponential family of distributions but is carried out in a simple way as the algorithm-based design: its decision of dose escalation and de-escalation involves only a simple comparison of the sample mean of the end point with two prespecified dose escalation and de-escalation boundaries. No model fitting is needed. We show that gBOIN has the desirable finite property of coherence and a large sample property of consistency. Numerical studies show that gBOIN yields good performance that is comparable with or superior to that of some existing, more complicated model-based designs. A Web application for implementing gBOIN is freely available from http://www.trialdesign.org.

Background: Phase I trials are designed to determine the safety, tolerability, and recommended phase 2 dose of therapeutic agents for subsequent testing. The dose-finding paradigm has thus traditionally focused on identifying the maximum tolerable dose of an agent or combination therapy under the assumption that there is a non-decreasing relationship between dose–toxicity and dose–efficacy. The dose is typically determined based on the probability of severe toxicity observed during the first treatment cycle. A novel endpoint, the total toxicity profile, was previously developed to account for the multiple toxicity types and grades experienced in the first cycle. More recently, this was extended to a repeated measures design based on the total toxicity profile to account for longitudinal toxicities over multiple treatment cycles in the absence of within-patient correlation. Methods: In this work, we propose to extend the design in the presence of within-patient correlation. Furthermore, we provide a framework to detect a toxicity time trend (toxicity increasing, decreasing, or stable) over multiple treatment cycles. We utilize a linear mixed model in the Bayesian framework, with the addition of Bayesian risk functions for decision-making in dose assignment. Results: The performance of this design was evaluated using simulation studies and real data from a phase I trial. We demonstrated that using available toxicity data from all cycles of treatment improves the accuracy of maximum tolerated dose identification and allows for the detection of a time trend. The performance is consistent regardless of the strength of the within-patient correlation. In addition, the use of a quasi-continuous total toxicity profile score significantly increased the power to detect time trends compared to when binary data only were used. Conclusion: The increased interest in molecularly targeted agents and immunotherapies in oncology necessitates innovative phase I study designs. Our proposed framework provides a tool to tackle some of the challenges presented by these novel agents, specifically through the ability to understand patterns of toxicity over time, which is important in the cases of cumulative or late toxicities.

675 Background: Recent retrospective analysis of CALGB/SWOG 80405 showed that left-sided colon cancers (LCC) respond differently to biological therapy compared with right-sided tumors. However, differences between rectal cancers (RC) and LCC remain undefined. Herein, we report our exploration of differences between these two groups. Methods: Individual patient data from 4182 patients (pts) with metastatic colorectal cancers, enrolled onto 8 first-line randomized trials, were pooled. Only pts with tumor locations that were clearly defined as LCC (splenic flexure to sigmoid) or RC were included in this analysis. Differences in pt characteristics and disease factors according to LCC vs. RC were identified. The prognostic effect of primary tumor location on OS and PFS was quantified via multivariable Cox proportional hazards modeling stratified by treatment arm within each study and adjusting for age, sex, performance status, and prior surgery. Results: In total, 2,479 (59%) pts with LCC and 1,703 (41%) pts with RC were identified. Pts with RC, compared with LCC, were more likely to be male (68% vs. 62%, p < 0.001), have lung metastases (mets) (56% vs. 37%, p < 0.001), and have 2 or more metastatic sites (64% vs. 60%, p < 0.02), whereas pts with LCC were more likely to have liver mets (84% vs. 76%, p < 0.001). RC had a greater frequency of KRAS mutations (41% vs. 37%, p = 0.04) than LCC but there were no differences in the frequency of BRAF mutations (5% vs. 4%, p = 0.2). In multivariable analysis, no differences in OS or PFS were observed between pts with LCC vs. RC. While risk of death did not differ by primary tumor location, across all pts with LCC or RC, those with liver mets had a 17% increased risk of death compared to those with lung mets (HR = 1.17, p = 0.03) after adjusting for effects of other variables. Forthcoming prognostic analysis of LCC vs. RC within grouped backbone treatments (e.g., FOLFOX and FOLFIRI) is underway. Conclusions: Tumors arising in the rectum may carry clinical and molecular features that are distinct from LCC. Further investigations are warranted to determine whether RC should be treated with the same chemo backbone and biological therapy as LCC.

Phase I oncology trials are designed to identify a safe dose with an acceptable toxicity profile. The dose is typically determined based on the probability of severe toxicity observed during the first treatment cycle, although patients continue to receive treatment for multiple cycles. In addition, the toxicity data from multiple types and grades are typically summarized into a single binary outcome of dose‐limiting toxicity. A novel endpoint, the total toxicity profile, was previously developed to account for the multiple toxicity types and grades. In this work, we propose to account for longitudinal repeated measures of total toxicity profile over multiple treatment cycles, accounting for cumulative toxicity during dosing‐finding. A linear mixed model was utilized in the Bayesian framework, with addition of Bayesian risk functions for decision‐making in dose assignment. The performance of this design is evaluated using simulation studies and compared with the previously proposed quasi‐likelihood continual reassessment method (QLCRM) design. Twelve clinical scenarios incorporating four different locations of maximum tolerated dose and three different time trends (decreasing, increasing, and no effect) were investigated. The proposed repeated measures design was comparable with the QLCRM when only cycle 1 data were utilized in dose‐finding; however, it demonstrated an improvement over the QLCRM when data from multiple cycles were used across all scenarios. Copyright © 2016 John Wiley & Sons, Ltd.