Response-adaptive randomization for survival trials: the parametric approach

Doubly adaptive biased coin design to improve Bayesian clinical trials with time-to-event endpoints

Clinical trialists often face the challenge of balancing scientific questions with other design features, such as improving efficiency, minimizing exposure to inferior treatments, and simultaneously comparing multiple treatments. While Bayesian response adaptive randomization (RAR) is a popular and effective method for achieving these objectives, it is known to have large variability and a lack of explicit theoretical results, making its use in clinical trials a subject of concern. It is desirable to propose a design that targets the same allocation proportion as Bayesian RAR and achieves the above objectives but addresses the concerns over Bayesian RAR. We propose the frequentist doubly adaptive biased coin designs (DBCD) targeting ethical allocation proportions from the Bayesian framework to satisfy different objectives in clinical trials with time-to-event endpoints. We derive the theoretical properties of the proposed adaptive randomization design and show through comprehensive numerical simulations that it can achieve ethical objectives without sacrificing efficiency. Our combined theoretical and numerical results offer a strong foundation for the practical use of RAR in real clinical trials.

Bayesian phase II adaptive randomization by jointly modeling efficacy and toxicity as time-to-event outcomes

The main goals of Phase II trials are to identify the therapeutic efficacy of new treatments and continue monitoring all the possible adverse effects. In Phase II trials, it is important to develop an adaptive randomization (AR) procedure that takes into account both the efficacy and toxicity. In most existing articles, toxicity is modeled as a binary endpoint through an unobservable random effect (frailty) to link the efficacy and toxicity. However, this approach does not capture toxicity profiles that evolve over time. In this article, we propose a new Bayesian adaptive randomization (BAR) procedure using the covariate-adjusted efficacy-toxicity ratio (ETR) index, where efficacy and toxicity are jointly modelled as time-to-event (TTE) outcomes. Furthermore, we also propose early stopping rules for toxicity and futility such that inferior treatments can be dropped at earlier time of trial. Simulation results show that compared to the BAR procedures based solely on the efficacy and that based on TTE efficacy and binary toxicity outcomes, the proposed BAR procedure can better identify the difference in treatment toxicity such that it can assign more patients to the superior treatment arm under some scenarios.

Response-adaptive randomization in clinical trials: from myths to practical considerations

Response-Adaptive Randomization (RAR) is part of a wider class of data-dependent sampling algorithms, for which clinical trials are typically used as a motivating application. In that context, patient allocation to treatments is determined by randomization probabilities that change based on the accrued response data in order to achieve experimental goals. RAR has received abundant theoretical attention from the biostatistical literature since the 1930's and has been the subject of numerous debates. In the last decade, it has received renewed consideration from the applied and methodological communities, driven by well-known practical examples and its widespread use in machine learning. Papers on the subject present different views on its usefulness, and these are not easy to reconcile. This work aims to address this gap by providing a unified, broad and fresh review of methodological and practical issues to consider when debating the use of RAR in clinical trials.

Comparison of treatments with ordinal responses in trials with sequential monitoring and response-adaptive randomization

In clinical trials, comparisons of treatments with ordinal responses are frequently conducted using the proportional odds model. However, the use of this model necessitates the adoption of the proportional odds assumption, which may not be appropriate. In particular, when responses are skewed, the use of the proportional odds model may result in a markedly inflated type I error rate. The latent Weibull distribution has recently been proposed to remedy this problem, and it has been demonstrated to be superior to the proportional odds model, especially when response-adaptive randomization is incorporated. However, there are several drawbacks associated with the latent Weibull model and the previously suggested response-adaptive treatment randomization scheme. In this paper, we propose the modified latent Weibull model to address these issues. Based on the modified latent Weibull model, the original response-adaptive design was also revised. In addition, the group sequential monitoring mechanism was included to enable interim analyses to be performed to determine, during a trial, whether a specific treatment is significantly more effective than another. If so, this will enable the trial to be terminated at a much earlier stage than a trial based on a fixed sample size. We performed a simulation study that clearly demonstrated the merits of our proposed framework. Furthermore, we redesigned a clinical study to further illustrate the advantages of our response-adaptive approach.

Evaluating response-adaptive randomization procedures for recurrent events and terminal event data using a composite endpoint

Recurrent event and terminal event data commonly arise in clinical and observational studies. To evaluate the efficacy of a treatment effect for both types of events, a composite endpoint has been used as a possible assessment, particularly when faced with high costs and a longer follow-up study. To model recurrent event processes complicated by the existence of a terminal event, joint frailty modeling has been typically employed. In this study, the objective was to develop some target-driven response adaptive randomization strategies using a composite endpoint based on joint frailty modeling. We first implemented a balanced randomized design and then investigated the response adaptive randomization. The former is intuitively first adopted while the latter is expected to be desirable and ethical in terms of allocating more subjects to the more effective treatment. The results show that the proposed procedures using a composite endpoint are capable of reducing the number of trial participants who receive inferior treatment while simultaneously reaching a desired optimal target as compared to a balanced randomized design. The R shiny application for calculating the sample size and allocation probabilities is also available. Finally, two clinical trials were used as pilot datasets to introduce the proposed procedures.

Response-adaptive treatment allocation for clinical studies with recurrent event and terminal event data

In long-term clinical studies, recurrent event data are frequently collected to contrast the efficacy of two different treatments. However, the recurrent event process can be stopped by a terminal event, such as death. For analyzing recurrent event and terminal event data, joint frailty modeling has recently received considerable attention because it makes it possible to study the joint evolution over time of both recurrent and terminal event processes and gives consistent and efficient parameters. For a two-arm clinical trial design based on these data sets, there has been limited research on investigating the balanced design, let alone adaptive treatment allocation. Although equal sample size allocation obtained for both treatments is intuitively first adopted in a trial design, if one treatment is expected to be superior, it may be desirable to allocate more subjects to the effective treatment. In this article, we calculate the required sample size based on restricted randomization and then propose a target response-adaptive randomization procedure for recurrent and terminal event outcomes based on the joint frailty model. A randomization procedure, the doubly adaptive biased coin design that targets some optimal allocations, is implemented. The proposed adaptive treatment allocation schemes have been shown to be capable of reducing the number of trial participants who receive inferior treatment while simultaneously reaching an optimal target, as well as retaining a comparable test power as compared to a restricted randomization design. Finally, two clinical studies, the COAPT trial and the A-HeFT trial, are used to illustrate the advantages of adopting the proposed procedure.

An optimal multiarmed response adaptive design for survival outcome with independent censoring

Compromising ethics and precision in the context of a multiarmed clinical trial, an optimal order adjusted response adaptive design is proposed for survival outcomes subject to independent random censoring. The operating characteristics of the proposed design and the follow-up inference are studied both theoretically as well as empirically and are compared with those of the competitors. Applicability of the developed design is further illustrated through redesigning a real clinical trial with survival responses.

Optimal and ethical designs for hypothesis testing in multi-arm exponential trials

Multi-arm clinical trials are complex experiments which involve several objectives. The demand for unequal allocations in a multi-treatment context is growing and adaptive designs are being increasingly used in several areas of medical research. For uncensored and censored exponential responses, we propose a constrained optimization approach in order to derive the design maximizing the power of the multivariate test of homogeneity, under a suitable ethical constraint. In the absence of censoring, we obtain a very simple closed-form solution that dominates the balanced design in terms of power and ethics. Our suggestion can also accommodate delayed responses and staggered entries, and can be implemented via response adaptive rules. While other targets proposed in the literature could present an unethical behavior, the suggested optimal allocation is frequently unbalanced by assigning more patients to the best treatment, both in the absence and presence of censoring. We evaluate the operating characteristics of our proposal theoretically and by simulations, also redesigning a real lung cancer trial, showing that the constrained optimal target guarantees very good performances in terms of ethical demands, power and estimation precision. Therefore, it is a valid and useful tool in designing clinical trials, especially oncological trials and clinical experiments for grave and novel infectious diseases, where the ethical concern is of primary importance.

Response-adaptive treatment allocation for survival trials with clustered right-censored data

A comparison of 2 treatments with survival outcomes in a clinical study may require treatment randomization on clusters of multiple units with correlated responses. For example, for patients with otitis media in both ears, a specific treatment is normally given to a single patient, and hence, the 2 ears constitute a cluster. Statistical procedures are available for comparison of treatment efficacies. The conventional approach for treatment allocation is the adoption of a balanced design, in which half of the patients are assigned to each treatment arm. However, considering the increasing acceptability of responsive-adaptive designs in recent years because of their desirable features, we have developed a response-adaptive treatment allocation scheme for survival trials with clustered data. The proposed treatment allocation scheme is superior to the balanced design in that it allows more patients to receive the better treatment. At the same time, the test power for comparing treatment efficacies using our treatment allocation scheme remains highly competitive. The advantage of the proposed randomization procedure is supported by a simulation study and the redesign of a clinical study.

Adaptive Optimal Designs for Dose-Finding Studies with Time-to-Event Outcomes

We consider optimal design problems for dose-finding studies with censored Weibull time-to-event outcomes. Locally D-optimal designs are investigated for a quadratic dose-response model for log-transformed data subject to right censoring. Two-stage adaptive D-optimal designs using maximum likelihood estimation (MLE) model updating are explored through simulation for a range of different dose-response scenarios and different amounts of censoring in the model. The adaptive optimal designs are found to be nearly as efficient as the locally D-optimal designs. A popular equal allocation design can be highly inefficient when the amount of censored data is high and when the Weibull model hazard is increasing. The issues of sample size planning/early stopping for an adaptive trial are investigated as well. The adaptive D-optimal design with early stopping can potentially reduce study size while achieving similar estimation precision as the fixed allocation design.

An oncology clinical trial design with randomization adaptive to both short- and long-term responses

In oncology clinical trials, both short-term response and long-term survival are important. We propose an urn-based adaptive randomization design to incorporate both of these two outcomes. While short-term response can update the randomization probability quickly to benefit the trial participants, long-term survival outcome can also change the randomization to favor the treatment arm with definitive therapeutic benefit. Using generalized Friedman's urn, we derive an explicit formula for the limiting distribution of the number of subjects assigned to each arm. With prior or hypothetical knowledge on treatment effects, this formula can be used to guide the selection of parameters for the proposed design to achieve desirable patient number ratios between different treatment arms, and thus optimize the operating characteristics of the trial design. Simulation studies show that the proposed design successfully assign more patients to the treatment arms with either better short-term tumor response or long-term survival outcome or both.

Bayesian Adaptive Randomization and Trial Monitoring with Predictive Probability for Time-to-event Endpoint

There has been much development in Bayesian adaptive designs in clinical trials. In the Bayesian paradigm, the posterior predictive distribution characterizes the future possible outcomes given the currently observed data. Based on the interim time-to-event data, we develop a new phase II trial design by combining the strength of both Bayesian adaptive randomization and the predictive probability. By comparing the mean survival times between patients assigned to two treatment arms, more patients are assigned to the better treatment on the basis of adaptive randomization. We continuously monitor the trial using the predictive probability for early termination in the case of superiority or futility. We conduct extensive simulation studies to examine the operating characteristics of four designs: the proposed predictive probability adaptive randomization design, the predictive probability equal randomization design, the posterior probability adaptive randomization design, and the group sequential design. Adaptive randomization designs using predictive probability and posterior probability yield a longer overall median survival time than the group sequential design, but at the cost of a slightly larger sample size. The average sample size using the predictive probability method is generally smaller than that of the posterior probability design.

Comparing three regularization methods to avoid extreme allocation probability in response-adaptive randomization

We examine three variations of the regularization methods for response-adaptive randomization (RAR) and compare their operating characteristics. A power transformation (PT) is applied to refine the randomization probability. The clip method is used to bound the randomization probability within specified limits. A burn-in period of equal randomization (ER) can be added before adaptive randomization (AR). For each method, more patients are assigned to the superior arm and overall response rate increase as the scheme approximates simple AR, while statistical power increases as it approximates ER. We evaluate the performance of the three methods by varying the tuning parameter to control the extent of AR to achieve the same statistical power. When there is no early stopping rule, PT method generally performed the best in yielding higher proportion to the superior arm and higher overall response rate, but with larger variability. The burn-in method showed smallest variability compared with the clip method and the PT method. With the efficacy early stopping rule, all three methods performed more similarly. The PT and clip methods are better than the burn-in method in achieving higher proportion randomized to the superior arm and higher overall response rate but burn-in method required fewer patients in the trial. By carefully choosing the method and the tuning parameter, RAR methods can be tailored to strike a balance between achieving the desired statistical power and enhancing the overall response rate.

Implementing Optimal Allocation in Clinical Trials with Multiple Endpoints

Modern clinical trials are often complex, with multiple competing objectives and multiple endpoints. Such trials should be both ethical and efficient. In this paper, we overcome the obstacles introduced by the large number of unknown parameters and the possible correlations between the multiple endpoints. We obtain the optimal allocation proportions for the following two optimization problems: (1) maximizing the power of the test of homogeneity with a fixed sample size, and (2) minimizing the expected weighted number of failures with a fixed power. Further, we implement these optimal allocations through response-adaptive randomization procedures. Our theoretical results provide the foundation for the implementation and further investigation of the procedure, and our numerical studies demonstrate its ability to achieve diverse objectives.

A Bayesian Hybrid Adaptive Randomisation Design for Clinical Trials with Survival Outcomes

BACKGROUND

Response-adaptive randomisation designs have been proposed to improve the efficiency of phase III randomised clinical trials and improve the outcomes of the clinical trial population. In the setting of failure time outcomes, Zhang and Rosenberger (2007) developed a response-adaptive randomisation approach that targets an optimal allocation, based on a fixed sample size.

OBJECTIVES

The aim of this research is to propose a response-adaptive randomisation procedure for survival trials with an interim monitoring plan, based on the following optimal criterion: for fixed variance of the estimated log hazard ratio, what allocation minimizes the expected hazard of failure? We demonstrate the utility of the design by redesigning a clinical trial on multiple myeloma.

METHODS

To handle continuous monitoring of data, we propose a Bayesian response-adaptive randomisation procedure, where the log hazard ratio is the effect measure of interest. Combining the prior with the normal likelihood, the mean posterior estimate of the log hazard ratio allows derivation of the optimal target allocation. We perform a simulation study to assess and compare the performance of this proposed Bayesian hybrid adaptive design to those of fixed, sequential or adaptive - either frequentist or fully Bayesian - designs. Non informative normal priors of the log hazard ratio were used, as well as mixture of enthusiastic and skeptical priors. Stopping rules based on the posterior distribution of the log hazard ratio were computed. The method is then illustrated by redesigning a phase III randomised clinical trial of chemotherapy in patients with multiple myeloma, with mixture of normal priors elicited from experts.

RESULTS

As expected, there was a reduction in the proportion of observed deaths in the adaptive vs. non-adaptive designs; this reduction was maximized using a Bayes mixture prior, with no clear-cut improvement by using a fully Bayesian procedure. The use of stopping rules allows a slight decrease in the observed proportion of deaths under the alternate hypothesis compared with the adaptive designs with no stopping rules.

CONCLUSIONS

Such Bayesian hybrid adaptive survival trials may be promising alternatives to traditional designs, reducing the duration of survival trials, as well as optimizing the ethical concerns for patients enrolled in the trial.

A surrogate-primary replacement algorithm for response-adaptive randomization in stroke clinical trials

Response-adaptive randomization (RAR) offers clinical investigators benefit by modifying the treatment allocation probabilities to optimize the ethical, operational, or statistical performance of the trial. Delayed primary outcomes and their effect on RAR have been studied in the literature; however, the incorporation of surrogate outcomes has not been fully addressed. We explore the benefits and limitations of surrogate outcome utilization in RAR in the context of acute stroke clinical trials. We propose a novel surrogate-primary (S-P) replacement algorithm where a patient's surrogate outcome is used in the RAR algorithm only until their primary outcome becomes available to replace it. Computer simulations investigate the effect of both the delay in obtaining the primary outcome and the underlying surrogate and primary outcome distributional discrepancies on complete randomization, standard RAR and the S-P replacement algorithm methods. Results show that when the primary outcome is delayed, the S-P replacement algorithm reduces the variability of the treatment allocation probabilities and achieves stabilization sooner. Additionally, the S-P replacement algorithm benefit proved to be robust in that it preserved power and reduced the expected number of failures across a variety of scenarios.

Outcome-adaptive randomization for a delayed outcome with a short-term predictor: imputation-based designs

Delay in the outcome variable is challenging for outcome-adaptive randomization, as it creates a lag between the number of subjects accrued and the information known at the time of the analysis. Motivated by a real-life pediatric ulcerative colitis trial, we consider a case where a short-term predictor is available for the delayed outcome. When a short-term predictor is not considered, studies have shown that the asymptotic properties of many outcome-adaptive randomization designs are little affected unless the lag is unreasonably large relative to the accrual process. These theoretical results assumed independent identical delays, however, whereas delays in the presence of a short-term predictor may only be conditionally homogeneous. We consider delayed outcomes as missing and propose mitigating the delay effect by imputing them. We apply this approach to the doubly adaptive biased coin design (DBCD) for motivating pediatric ulcerative colitis trial. We provide theoretical results that if the delays, although non-homogeneous, are reasonably short relative to the accrual process similarly as in the iid delay case, the lag is also asymptotically ignorable in the sense that a standard DBCD that utilizes only observed outcomes attains target allocation ratios in the limit. Empirical studies, however, indicate that imputation-based DBCDs performed more reliably in finite samples with smaller root mean square errors. The empirical studies assumed a common clinical setting where a delayed outcome is positively correlated with a short-term predictor similarly between treatment arm groups. We varied the strength of the correlation and considered fast and slow accrual settings.

Efficient and ethical response-adaptive randomization designs for multi-arm clinical trials with Weibull time-to-event outcomes

We consider a design problem for a clinical trial with multiple treatment arms and time-to-event primary outcomes that are modeled using the Weibull family of distributions. The D-optimal design for the most precise estimation of model parameters is derived, along with compound optimal allocation designs that provide targeted efficiencies for various estimation problems and ethical considerations. The proposed optimal allocation designs are studied theoretically and are implemented using response-adaptive randomization for a clinical trial with censored Weibull outcomes. We compare the merits of our multiple-objective response-adaptive designs with traditional randomization designs and show that our designs are more flexible, realistic, generally more ethical, and frequently provide higher efficiencies for estimating different sets of parameters.

Near efficient target allocations in response-adaptive randomization

Traditionally optimal target allocation proportions for response-adaptive designs are derived by completely ignoring the actual adaptive randomization procedure. Considering efficiency of the allocation designs, we derive near efficient target proportions to balance between individual and collective ethics. Performance of the derived allocation targets are assessed numerically for binary, normal and exponential responses. Generalization for multiple treatments is also addressed.

Estimation of treatment effect under non-proportional hazards and conditionally independent censoring

In clinical trials with time-to-event outcomes, it is common to estimate the marginal hazard ratio from the proportional hazards model, even when the proportional hazards assumption is not valid. This is unavoidable from the perspective that the estimator must be specified a priori if probability statements about treatment effect estimates are desired. Marginal hazard ratio estimates under non-proportional hazards are still useful, as they can be considered to be average treatment effect estimates over the support of the data. However, as many have shown, under non-proportional hazard, the 'usual' unweighted marginal hazard ratio estimate is a function of the censoring distribution, which is not normally considered to be scientifically relevant when describing the treatment effect. In addition, in many practical settings, the censoring distribution is only conditionally independent (e.g., differing across treatment arms), which further complicates the interpretation. In this paper, we investigate an estimator of the hazard ratio that removes the influence of censoring and propose a consistent robust variance estimator. We compare the coverage probability of the estimator to both the usual Cox model estimator and an estimator proposed by Xu and O'Quigley (2000) when censoring is independent of the covariate. The new estimator should be used for inference that does not depend on the censoring distribution. It is particularly relevant to adaptive clinical trials where, by design, censoring distributions differ across treatment arms.

Covariate-adjusted response-adaptive designs for longitudinal treatment responses: PEMF trial revisited

Response-adaptive designs have become popular for allocation of the entering patients among two or more competing treatments in a phase III clinical trial. Although there are a lot of designs for binary treatment responses, the number of designs involving covariates is very small. Sometimes the patients give repeated responses. The only available response-adaptive allocation design for repeated binary responses is the urn design by Biswas and Dewanji [Biswas A and Dewanji AA. Randomized longitudinal play-the-winner design for repeated binary data. ANZJS 2004; 46: 675–684; Biswas A and Dewanji A. Inference for a RPW-type clinical trial with repeated monitoring for the treatment of rheumatoid arthritis. Biometr J 2004; 46: 769–779.], although it does not take care of the covariates of the patients in the allocation design. In this article, a covariate-adjusted response-adaptive randomisation procedure is developed using the log-odds ratio within the Bayesian framework for longitudinal binary responses. The small sample performance of the proposed allocation procedure is assessed through a simulation study. The proposed procedure is illustrated using some real data set.

Using short-term response information to facilitate adaptive randomization for survival clinical trials

Increased survival is a common goal of cancer clinical trials. Owing to the long periods of observation and follow-up to assess patient survival outcome, it is difficult to use outcome-adaptive randomization in these trials. In practice, often information about a short-term response is quickly available during or shortly after treatment, and this short-term response is a good predictor for long-term survival. For example, complete remission of leukemia can be achieved and measured after a few cycles of treatment. It is a short-term response that is desirable for prolonging survival. We propose a new design for survival trials when such short-term response information is available. We use the short-term information to 'speed up' the adaptation of the randomization procedure. We establish a connection between the short-term response and the long-term survival through a Bayesian model, first by using prior clinical information, and then by dynamically updating the model according to information accumulated in the ongoing trial. Interim monitoring and final decision making are based upon inference on the primary outcome of survival. The new design uses fewer patients, and can more effectively assign patients to the better treatment arms. We demonstrate these properties through simulation studies.