Using biomarkers to allocate patients in a response-adaptive clinical trial.

In this paper, we discuss a response adaptive randomization method, and why it should be used in clinical trials for rare diseases compared to a randomized controlled trial with equal fixed randomization. The developed method uses a patient's biomarkers to alter the allocation probability to each treatment, in order to emphasize the benefit to the trial population. The method starts with an initial burn-in period of a small number of patients, who with equal probability, are allocated to each treatment. We then use a regression method to predict the best outcome of the next patient, using their biomarkers and the information from the previous patients. This estimated best treatment is assigned to the next patient with high probability. A completed clinical trial for the effect of catumaxomab on the survival of cancer patients is used as an example to demonstrate the use of the method and the differences to a controlled trial with equal allocation. Different regression procedures are investigated and compared to a randomized controlled trial, using efficacy and ethical measures.

Loss functions in restricted parameter spaces and their Bayesian applications.

Squared error loss remains the most commonly used loss function for constructing a Bayes estimator of the parameter of interest. However, it can lead to suboptimal solutions when a parameter is defined on a restricted space. It can also be an inappropriate choice in the context when an extreme overestimation and/or underestimation results in severe consequences and a more conservative estimator is preferred. We advocate a class of loss functions for parameters defined on restricted spaces which infinitely penalize boundary decisions like the squared error loss does on the real line. We also recall several properties of loss functions such as symmetry, convexity and invariance. We propose generalizations of the squared error loss function for parameters defined on the positive real line and on an interval. We provide explicit solutions for corresponding Bayes estimators and discuss multivariate extensions. Four well-known Bayesian estimation problems are used to demonstrate inferential benefits the novel Bayes estimators can provide in the context of restricted estimation.

A proposal for a new PhD level curriculum on quantitative methods for drug development.

This paper provides an overview of "Improving Design, Evaluation and Analysis of early drug development Studies" (IDEAS), a European Commission-funded network bringing together leading academic institutions and small- to large-sized pharmaceutical companies to train a cohort of graduate-level medical statisticians. The network is composed of a diverse mix of public and private sector partners spread across Europe, which will host 14 early-stage researchers for 36 months. IDEAS training activities are composed of a well-rounded mixture of specialist methodological components and generic transferable skills. Particular attention is paid to fostering collaborations between researchers and supervisors, which span academia and the private sector. Within this paper, we review existing medical statistics programmes (MSc and PhD) and highlight the training they provide on skills relevant to drug development. Motivated by this review and our experiences with the IDEAS project, we propose a concept for a joint, harmonised European PhD programme to train statisticians in quantitative methods for drug development.

Adaptive clinical trials in tuberculosis: applications, challenges and solutions.

Drug development for tuberculosis (TB) faces numerous practical obstacles, including the need for combination treatment with at least three drugs, reliance on possibly unrepresentative animal models which may not reproduce key features of human disease and the lack of a well-validated surrogate endpoint for stable cure. Pivotal Phase III trials are large, lengthy and expensive, and the funding and capacity to conduct them are limited worldwide. More rational methods for the selection of priority regimens for Phase III are urgently needed to avoid costly late-stage failures. We examine the suitability of adaptive clinical trial designs for drug development in TB, focusing on designs for Phase IIB and III trials, where we believe the biggest gains in efficiency can be made. Key areas that may be addressed by such designs are improvements in the selection of doses and combinations of drugs in early clinical development and in maximising the power of confirmatory trials in multidrug-resistant TB, where patient numbers and complexity pose practical limitations. We encourage trialists and regulators in this area to consider the advantages that may be offered by these designs and their potential to more effectively and rapidly identify better treatment regimens for TB patients worldwide.

Flexible sequential designs for multi-arm clinical trials.

Adaptive designs that are based on group-sequential approaches have the benefit of being efficient as stopping boundaries can be found that lead to good operating characteristics with test decisions based solely on sufficient statistics. The drawback of these so called 'pre-planned adaptive' designs is that unexpected design changes are not possible without impacting the error rates. 'Flexible adaptive designs' on the other hand can cope with a large number of contingencies at the cost of reduced efficiency. In this work, we focus on two different approaches for multi-arm multi-stage trials, which are based on group-sequential ideas, and discuss how these 'pre-planned adaptive designs' can be modified to allow for flexibility. We then show how the added flexibility can be used for treatment selection and sample size reassessment and evaluate the impact on the error rates in a simulation study. The results show that an impressive overall procedure can be found by combining a well chosen pre-planned design with an application of the conditional error principle to allow flexible treatment selection.

Considerations on covariates and endpoints in multi-arm multi-stage clinical trials selecting all promising treatments.

In early stages of drug development, there is often uncertainty about the most promising among a set of different treatments. To ensure the best use of resources in such situations, it is important to decide which, if any, of the treatments should be taken forward for further testing. In later development, it has been shown that evaluating more than one dose increases the chance of success substantially. In this work, we discuss how multi-arm multi-stage trials can be designed such that all promising treatments are kept in the study at the interim analyses. We first investigate the impact of deviating from the planned design and show how confidence intervals can be constructed before we consider the impact of important covariates. We show that under orthogonality, the inclusion of covariates has no effect on familywise error rate control in the strong sense. We further show that the derived methodology can be used to investigate non-normal endpoints.

Simultaneous confidence intervals that are compatible with closed testing in adaptive designs.

We describe a general method for finding a confidence region for a parameter vector that is compatible with the decisions of a two-stage closed test procedure in an adaptive experiment. The closed test procedure is characterized by the fact that rejection or nonrejection of a null hypothesis may depend on the decisions for other hypotheses and the compatible confidence region will, in general, have a complex, nonrectangular shape. We find the smallest cross-product of simultaneous confidence intervals containing the region and provide computational shortcuts for calculating the lower bounds on parameters corresponding to the rejected null hypotheses. We illustrate the method with an adaptive phase II/III clinical trial.

Direct effects testing: a two-stage procedure to test for effect size and variable importance for correlated binary predictors and a binary response.

In applications such as medical statistics and genetics, we encounter situations where a large number of highly correlated predictors explain a response. For example, the response may be a disease indicator and the predictors may be treatment indicators or single nucleotide polymorphisms (SNPs). Constructing a good predictive model in such cases is well studied. Less well understood is how to recover the 'true sparsity pattern', that is finding which predictors have direct effects on the response, and indicating the statistical significance of the results. Restricting attention to binary predictors and response, we study the recovery of the true sparsity pattern using a two-stage method that separates establishing the presence of effects from inferring their exact relationship with the predictors. Simulations and a real data application demonstrate that the method discriminates well between associations and direct effects. Comparisons with lasso-based methods demonstrate favourable performance of the proposed method.

Optimization, refinement and reduction of murine in vivo experiments to assess therapeutic approaches for haemophilia A.

The tail cut bleeding model (CUT) is routinely used in factor VIII-deficient mice to assess pharmacodynamic effects of therapeutic strategies for haemophilia A. Results from this model are highly variable, many modifications to the model are reported and at times the animals' wellbeing may be compromised by recording survival as an endpoint. We therefore investigated if the ferric chloride carotid occlusion model (COM) used for thrombosis research can be applied to enhance data quality and animal welfare in haemophilia A research. Relative dose effects and relative dose variations were calculated for the CUT and COM. The requisite sample sizes were estimated and the importance of survival rates to assess rebleeds during recovery was evaluated by correlating initial blood loss to mortality. Relative dose effects increased with higher doses in both models. The COM was more sensitive at lower doses than the CUT, had up to 82% less variation across doses and clearly showed superior accuracy. Only 5% of the sample size required for the CUT would be needed to establish non-inferiority between a specific therapeutic dose in haemophilia A mice and healthy wild-type animals. A strong statistically significant correlation was found between initial blood loss and mortality within 24 h. Our findings clearly suggest that the COM is a valid tool for assessing haemophilia A treatment in vivo. The highly reproducible data means that significantly fewer animals are required and a more humane endpoint can be used by directly assessing clot stability instead of survival rate.