Elias Meyer, Emily Hammarstrom-Wickens, Jemma Greenin, Fulvio Di Stefano.
2021 PSI Conference Session given by career young statisticians. Second session of the conference which focuses on innovative trial design and contains the following talks – “Decision rules for identifying combination therapies in open-entry, randomized controlled platform trials”, ““What would happen if...?” – Using simulation to determine an optimal sample size re-estimation strategy at the trial design stage”, “Testing… Testing… 1,2,3: Multiple Testing and Combination Testing for Treatment Selection, Adaptive Seamless Designs” and “A comparison of estimation methods adjusting for selection bias in adaptive enrichment designs with time-to-event endpoints”.
Decision rules for identifying combination therapies in open-entry, randomized controlled platform trials - Elias Meyer
The design and conduct of platform trials have become increasingly popular for drug development programs, attracting interest from statisticians, clinicians and regulatory agencies. Many statistical questions related to designing platform trials - such as what is the impact of decision rules, sharing of information across cohorts, and allocation ratios on operating characteristics and error rates – remain unanswered. In many platform trials, the definition of error rates is not straightforward as classical error rate concepts are not applicable. In particular, the strict control of the family-wise Type I error rate may not be applicable in certain settings. For an open-entry, exploratory platform trial design comparing combination therapies to the respective monotherapies and standard-of-care, we define a set of error rates and operating characteristics and then use these as a measure to compare a set of design parameters under a range of simulation assumptions. When setting up the simulations, we aimed for realistic trial trajectories, e.g. in case one compound is found to be superior to standard-of-care, it could become the new standard-of-care in future cohorts. Our results indicate that the method of data sharing, exact specification of decision rules and quality of the biomarker used to make interim decisions all strongly contribute to the operating characteristics of the platform trial. Together with the potential flexibility and complexity of a platform trial, which also impact the achieved operating characteristics, this implies that utmost care needs to be given to evaluation of different assumptions and design parameters at the design stage.
Master protocol and especially platform trials have become increasingly popular in recent years, however designing them is challenging as many different factors influence the trial’s operating characteristics. In this talk, we will shed light on the impact of some assumptions and design parameters, as well as define appropriate operating characteristics.
“What would happen if...?” – Using simulation to determine an optimal sample size re-estimation strategy at the trial design stage - Emily Hammarstrom-Wickens
When proving something mathematically is difficult or limited to specific situations, simulation studies may be an effective solution. In a clinical trial setting, simulations can represent the results of simultaneously performing many real trials, creating an empirical distribution for measures of interest, such as the Type I error. Anticipated parameters, such as the expected treatment effect, and the sensitivity in changes to such parameters can be assessed via simulating plausible scenarios for a clinical trial. Transparent and well-planned simulation studies can be modified to suit particular requirements, making them both a useful and informative tool when designing a clinical trial.
A motivating example is the ROLARR trial (Jayne et al., 2017), which failed to show a statistically significant treatment effect, potentially as a result from the misspecification of the sample size assumptions. A proposed solution to combat this misspecification is the idea of using accumulated data from the study at a specified time point to adjust these initial assumptions and re-estimate the sample size. This is the basis of sample size re-estimation (SSR) methods.
Through the use of simulations, this presentation looks retrospectively at what could have been done at the design stage of the trial by focusing on Bayesian SSR methods.
Testing… Testing… 1,2,3: Multiple Testing and Combination Testing for Treatment Selection, Adaptive Seamless Designs - Jemma Greenin
The cost of drug-development is increasing, as is the timeline of drug discovery. To solve the increasing fear of stagnation in the development of novel compounds many methods are suggested for innovation, including the adaptive designs of clinical trials.
Modifications applied through an adaptive design approach can utilize the data generated during the trial, thus making it more flexible as well as reducing the time factors necessary for completion, hence reducing costs.
My presentation aims to review the methods of an adaptive seamless phase II/III design, combining the objectives of separate Phase IIb and Phase III into a single trial. To successfully execute this, and maintain the type I error, combination testing, multiple testing and treatment selection must all be considered.
Possible methods for combination testing include the inverse chi-squared Fisher and weighted inverse normal and multiple testing methods include Bonferroni, Sime’s or Dunnett. Treatment selection trial designs have been proposed by Bauer and Kohne as well as Thall, Simon and Ellenberg. By using simulations, I will discuss the validity and efficiency of the different combination of these methods available to use.
There is no single approach for designing adaptive studies. The best results rely on clearly defining the objectives of the adaptation combined with a well-chosen design and methodology. The adaptive design proposed by Thall, Simon and Ellenberg was seen to be the least conservative and the most powerful.
The cost of drug-development is increasing, as is the timeline of drug discovery. In this presentation I will describe a possible solution to this problem through the use of adaptive trial designs. Why is it not as simple as just combining phases?
A comparison of estimation methods adjusting for selection bias in adaptive enrichment designs with time-to-event endpoints - Fulvio Di Stefano
Adaptive enrichment designs in clinical trials have been developed to enhance drug developments. They permit, at interim analyses during the trial, to select the sub-population that benefits the most from the treatment. This results in improvements both in terms of resources and ethics, by reducing the number of patients receiving non effective treatments. Because of this selection, the naive maximum likelihood estimation of the treatment effect, commonly used in classical randomized controlled trials, is biased. In the literature, several methods have been proposed to obtain a better estimation of the treatments’ effects in such contexts. To date, most of the works have focused on normally endpoints, and some estimators have been proposed for time-to-event endpoints but they have not all been compared side-by-side. In this work, we conduct an extensive simulation study, inspired by a real case-study in Heart Failure, to compare the maximum-likelihood estimator (MLE) with an unbiased estimator, shrinkage estimators and bias-adjusted estimators for the estimation of the treatment effect with time-to-event data. The performances of the estimators are evaluated in terms of bias, variance and mean squared error. Based on the results, we recommend using the unbiased estimator and the single-iteration bias-adjusted estimator: the former completely eradicates the bias, but is highly variable with respect to a naive estimation; the latter is less biased than the MLE estimator and only slightly more variable.
We analyze six estimation methods that adjust for selection bias in enrichment designs with time-to-event endpoints. We use simulation to assess the properties of the estimators and apply them to a real case study, suggesting two estimators to be used.