Bayesian Dose Escalation Study Design with Consideration of Both Early and Late Onset Toxicity
In phase I oncology trials, dose-limiting toxicity (DLT) is often used as the endpoint during early cycles in dose escalation studies to find the maximum tolerated dose (MTD). Bayesian adaptive designs such as the Escalation with Overdose Control (EWOC; Babb et al., Stat Med 17:1103–1120, 1998) methods have been introduced to protect patients from overdosing and account for variability in toxicity estimates.
Because of the short term nature of the DLT endpoint, these methods may not be feasible in practice when a particular adverse event is likely to occur during longer term exposure. To accommodate a longer assessment time while maintaining escalation timelines, Tighiouart et al. (PLoS One 9(3):e93070, 2014) proposed an EWOC design using a time to event toxicity endpoint.
When both early and late onset toxicity are crucial for dose finding studies, we propose to define DLT with two components, one for immediate toxicity in a binary model, and the other for late onset toxicity in a time to event model. The types of immediate toxicity and late onset toxicity are different. Through simulations, we demonstrated that the proposed dose escalation design can incorporate historical knowledge, protect patients from being assigned to toxic doses, and consider early and late onset toxicity while maintaining the escalation timeline. This design has been implemented in a phase 1 dose escalation study with both short term DLT and long term AESI, which allowed the study team to consider both the short term DLT and long term AESI with an acceptable timeline.
KeywordsDose escalation study Late onset toxicity Early and late onset toxicities Escalation with overdose control Time to event toxicity endpoint Bayesian dose finding
The authors appreciate the helpful discussions with Mourad Tighiouart and Pierre Colin. The authors thank referees and the associate editor for their excellent comments and suggestions.
- EMA. Guideline on the evaluation of anticancer medicinal products in man. EMA/CHMP/205/95 Rev.5 (2017)Google Scholar
- Ma, X., Zheng, W., Lu, Y.: Personalized effective dose selection in dose ranging studies. In: Lin, J., Wang, B., Hu, X., Chen, K., Liu, R. (eds.) Statistical Applications from Clinical Trials and Personalized Medicine to Finance and Business Analytics. Springer, Basel (2016)Google Scholar
- Paoletti, X., Toumeau, C.L., Verweij, J., Siiu, L.L., Seymour, L., Postel-Vinay, S., Collette, L., Rizzo, E., Ivy, P., Olmos, D., Massard, C., Lacombe, D., Kaye, S.B., Soria, J.-C.: Defining dose-liming toxicity for phase 1 trials of molecularly targeted agents: results of a DLT-TARGETT international survey. Eur. J. Cancer. 50(12), 2050–2056 (2014)CrossRefGoogle Scholar
- Sverdlov, O., Gao, L.: Phase I/II dose finding designs with efficacy and safety endpoints. In: O’Quigley, J., Iasonos, A., Bornkamp, B. (eds.) Handbook of Methods for Designing, Monitoring, and Analyzing Dose-Finding Trials. CRC Press, Boca Raton (2017)Google Scholar