Despite accumulating learnings from early phases, several uncertainties remain to be addressed when designing pivotal trials. Adaptive trials can help mitigate uncertainties; however, the trade-offs and their impact differ in the confirmatory setting. Quantifying uncertainties and risks and planning for mitigating adaptations are necessary to maximize the chances of success while maintaining the required scientific rigor of pivotal trials. Quantitative strategies can help inform decisions and optimize choices. (Read Part 1 of this article on early-stage development).
Integrating dose selection into pivotal trials
Increased pressure on the early phases of development sometimes leads to uncertainty of the dose selection for the pivotal trial. To optimize a clinical development program, a seamless Phase 2/3 study can be designed to integrate a dose-finding stage into the pivotal trials.
An operationally seamless approach means that the dose-finding and confirmation stages will be done under the same protocol; however, the population enrolled in the dose-finding stage will not be used for the analysis in the confirmatory stage. This approach can offer operational efficiency and save time, but it does not allow for a reduction in the overall sample size.
Inferentially seamless design means that the dose-finding and confirmation stages will be done under the same protocol, and the population enrolled in dose-finding stage will be included in the final analysis. In other words, the final analysis population includes patients enrolled in the conformity stage as well as patients on placebo and selected dose arms from the dose-finding stage. Such a design allows for a substantial reduction in sample size and shortens the program duration. A solid statistical methodology is required to maintain firm control of the familywise type-1 error rate to withstand regulatory scrutiny.
Mitigating uncertainty about the effect size
Despite the best efforts of early development programs to inform the efficacy assumptions for a pivotal trial, uncertainty about the effect size remains a significant risk for the pivotal program. The remaining uncertainty can be explained by many factors, such as moving from a limited number of sites and countries involved in Phase 2 to a larger Phase 3 study, by evolving the standard of care, and so on.
The most frequently used de-risking strategies include group sequential study design and adaptive design.
Group sequential design allows the possibility of early stopping for futility or efficacy. The study is usually designed for more conservative effect size and one or two (though it can be more) interim analyses with the possibility of early stopping if the effect size is in line with optimistic assumptions. In other words, if the efficacy is better than a conservative assumption, the study can be stopped for efficacy at the interim analysis.
Adaptive design with sample size re-estimation allows for the possibility to stop for futility or efficacy, or re-estimate the sample size at the interim analysis. The study is usually planned/powered for the realistically optimistic effect size. The design is based on the concept of a promising zone to inform decisions to stop for futility, continue the trial, or increase the sample size. For example, if the results are in the promising zone to reach a more conservative effect size, a sample size re-estimation will be proposed, and the trial will continue with the revised sample size.
Both strategies are statistically efficient and accepted by regulators. However, from the investor’s perspective, an adaptive design with sample size re-estimation is usually more appealing because the additional resources are committed only if the results fall into a promising zone. It is important to note that there is no a priori most efficient design: the assumptions on optimistic and conservative effect size, enrollment scenarios, and business trade-offs, such as the probability of success, costs, and duration, need to be taken into consideration for simulations that can guide the selection of the optimal study design.
Primary and key secondary endpoint selection
It is important to note that the same clinical outcome can be analyzed using a different statistical approach, resulting in different risks and operational efficiency. It is critical to clearly define clinical scenarios and challenge your statisticians to come up with the most suitable and efficient statistical approaches.
For example, an “event” endpoint can be analyzed in different ways. It can be managed as a binary endpoint or as a time-to-event. It is important to consider not only the assumptions at the end of the follow-up period, but the evolution during the entire period. The conventional time-to-event methods are based on the proportional hazards assumption. However, in some trials such as acute cardiovascular or immune-oncology trials, crossing or late separation of curves can be expected, which means that the assumption of proportional hazards is violated. Therefore, using the “conventional” methods may lead to loss of power and suboptimal interim decisions.
There are many clinical scenarios when clinical developers are interested in several “events.” For example, an intervention can be expected to reduce recurrent hospitalization without improving overall mortality. In this case, we are interested in the efficacy based on the number of hospitalizations, but we still need to take into account mortality. A statistical approach considering mortality as a competing risk can be helpful in such situations.
A clinical hierarchical composite endpoint is another approach to managing multiple events/endpoints. Conventional composite endpoint approaches have been criticized because they do not take into consideration the clinical importance of the components. For example, in a “usual” composite of mortality, stroke, and hospitalization, an outcome will be determined by the first occurring event, not taking into account that mortality is a clinically more important outcome than hospitalizations. Clinical hierarchical endpoint based on Finkelstein-Schoenfeld methods/win ratio allows for the accounting of clinical priorities of the components in the composite. Importantly, these methods can be implemented in the adaptive designs, including sample size re-estimation designs.
In some situations, there is uncertainty about the primary and key secondary endpoint choices. Different methods, such as Bonferroni correction or Hochberg procedures, can be used to manage multiplicity subject to objectives and assumptions.
Personalized medicine, also known as precision medicine, shifted the direction of drug development from traditional ways to biomarker-based approaches, where the biomarker could be a molecular biomarker, digital biomarker, or AI/ML-derived enrichment algorithm.
Population enrichment helps to identify responders and detect treatment effects with smaller sample sizes. It can also rescue a program if investigational therapy works in a subgroup but not in the whole population of patients. Population enrichment has become increasingly appealing with the rising costs of clinical trials.
An adaptive population enrichment design is a flexible approach that helps identify subpopulations of patients most likely to benefit from a treatment. At the interim analysis, a study can be stopped for futility, continue as planned with the overall population, or enrich the study population by enrolling new patients only in subgroups appearing to benefit from the new therapy. Population enrichment and sample size re-estimation designs have been used across many indications in Phase 3 trials.
Each development program is unique, and no single solution can address all needs. Exploration of several quantitative strategies and trial simulations enable data-driven decisions that can optimize your clinical development strategy.
About Natalia Muehlemann
Natalia Muehlemann is Vice President, Clinical Development, at Cytel. Natalia Muehlemann, MD, MBA, has over 20 years of experience in general management, clinical development, and business development in the life sciences. Dr. Muehlemann joined Cytel in 2020, and prior to Cytel, served as Global Category Head, Acute Care - Oncology - Devices at Nestle Health Sciences. She acts as an Expert Jury member for the European Commission’s Innovation Council. Dr. Muehlemann holds an MD and an MBA (IMD) and professional certifications in statistics and data science.
Read more from Perspectives on Enquiry and Evidence:
Sorry no results please clear the filters and try again