Practical recommendations for implementing a Bayesian adaptive phase I design during a pandemic

Abstract

Background Modern designs for dose-finding studies (e.g., model-based designs such as continual reassessment method) have been shown to substantially improve the ability to determine a suitable dose for efficacy testing when compared to traditional designs such as the 3 + 3 design. However, implementing such designs requires time and specialist knowledge. Methods We present a practical approach to developing a model-based design to help support uptake of these methods; in particular, we lay out how to derive the necessary parameters and who should input, and when, to these decisions. Designing a model-based, dose-finding trial is demonstrated using a treatment within the AGILE platform trial, a phase I/II adaptive design for novel COVID-19 treatments. Results We present discussion of the practical delivery of AGILE, covering what information was found to support principled decision making by the Safety Review Committee, and what could be contained within a statistical analysis plan. We also discuss additional challenges we encountered in the study and discuss more generally what (unplanned) adaptations may be acceptable (or not) in studies using model-based designs. Conclusions This example demonstrates both how to design and deliver an adaptive dose-finding trial in order to support uptake of these methods.

Treating intrusive memories after trauma in healthcare workers: a Bayesian adaptive randomised trial developing an imagery-competing task intervention

“…In contrast to p-values, BFs retain their meaning in situations where data are provided over time, regardless of any sampling decisions, therefore data can be analysed repeatedly as it becomes available, without needing special corrections (see Schönbrodt and colleagues [28]). Therefore, the Bayesian approach allows interim analyses to guide decision-making during RCTs, making them attractive when using adaptive trial designs [38, 39].…”

“…Therefore, the Bayesian approach allows interim analyses to guide decision-making during RCTs, making them attractive when using adaptive trial designs [38,39].…”

Practical implementation of the partial ordering continual reassessment method in a Phase I combination‐schedule dose‐finding trial

“…For example, considering ( 12) and ( 16), it is believed that there is 90% probability that ( 16) is more toxic given that schedule is thought to have a larger effect on toxicity than the single dose of M. In contrast, comparing, ( 14) and (17), it is thought that there is an equal chance of one of these being more toxic than the other given that the total average dose of M1774 is slightly higher for (14) together with the single dose of M1774 but the schedule is less intensive. Similar prior probabilities were assumed to hold for 100 mg of niraparib, for example, regimen (2), ( 6) and ( 4), (7).…”

“…In Table 2, all anti-diagonal moves correspond to potentially unknown ordering between the regimens. For example, comparing (14) and (17) it is unknown which of these is more toxic as the total average amount of M1774 is slightly higher for ( 14), the single dose is higher for ( 14) but the schedule is less intensive. Similarly, comparing the regimens with the total amount of M1774 (eg, (12) vs ( 16) and ( 15) vs (18)) it is unknown a-priori whether the increase in the individual doses of M is more (or less) prominent than the reduction in the schedule.…”

“…Example of the dose-transition pathways including the set of admissible regimens recommended by the model for the first 3 cohorts (each with 3 participants) depending on the number of experienced DLTs (from 0 to 3) and using minimization of the normalized distance criterion If 1 DLT is observed for the first cohort, the regimens ( 14) and ( 17) are in the admissible set of regimens together with lower regimens, and the model recommends de-escalating to regimen (14). If no DLT is observed at this regimen, the model will recommend to escalate back to the starting regimen (18).…”