Case study

A PHASTAR statistician was tasked to assist with the design of a modular protocol to assess safety, tolerability and pharmacokinetics in a novel oncology compound in an acute indication. The first module consisted of a dose escalation portion with the aim of defining the maximum tolerated dose (MTD). An ongoing first-in-human study in a more robust population was using the Continual Reassessment Method (CRM) to determine the MTD, but due to expected early adverse reactions in the more acute population, each dose cycle would need to involve intra-patient titration. With the titration proposed, it was not possible to fit a dose toxicity curve since the actual dose received within each dose level would be variable between patients. For this reason, CRM was rejected, and a keyboard, or modified toxicity probability interval (mTPI-2) design (Guo et al 2017) was proposed by PHASTAR, which would allow the dose levels to follow an ordinal framework and provide better operating characteristics and specificity than a traditional 3+3 or Rolling-six design.

Starting dose

At the time of protocol design, the tolerability of the proposed starting dose in the more robust population had not been established, but the client was keen not to delay the project, or deny patients a potentially effective dose by starting too low in the second-in-human study. PHASTAR suggested to reduce these risks by incorporating a de-escalation to a dose level -1 into the study design. In effect, the starting dose would be the second lowest dose.

Design details

A recommended dose level for each subsequent cohort of patients was made by a Scientific Review Committee, guided by pre-defined decision rules. The decision rules were based on a mTPI-2 design with a 30% (± 5%) target dose-limiting toxicity (DLT) rate for MTD. Depending on the number of patients treated at a dose level, and the number of DLTs observed in those patients, the possible decisions were: escalate ‘E’, stay at current dose ‘S’, de-escalate ‘D’, or de-escalate and do not use current dose again due to unacceptable toxicity ‘DNT’. A dose level was considered as having unacceptable toxicity (with no additional patients to be enrolled at that dose level) if there was an estimated > 95% probability (P) of exceeding the target DLT rate of 30% (i.e., P[DLT > 30% data] > 95%) with at least 3 patients treated at that dose level. If a ‘stay’ decision was made, additional patients could be enrolled (in cohorts of 3) up to a maximum of 12 DLT-evaluable patients for a given dose level and a maximum of 30 DLT-evaluable patients.

Choice of sample size

The sample size was determined based on a review of operating characteristics provided by PHASTAR through simulations of the proposed design, and design alternatives. Potential risks of overdose, misspecification of the MTD, and time taken to complete this part of the study were presented to the study team who were then able to make an informed decision on the optimum cohort size and maximum number of patients to be treated at each level.

Learn more about our services
We are experts in study design, statistical analysis, data science, data capture and reporting for clinical trials