Robust group sequential designs for trials with survival endpoints and delayed response.

Randomized clinical trials in oncology typically utilize time-to-event endpoints such as progression-free survival or overall survival as their primary efficacy endpoints, and the most commonly used statistical test to analyze these endpoints is the log-rank test. The power of the log-rank test depends on the behavior of the hazard ratio of the treatment arm to the control arm. Under the assumption of proportional hazards, the log-rank test is asymptotically fully efficient. However, this proportionality assumption does not hold true if there is a delayed treatment effect. Cancer immunology has evolved over time and several cancer vaccines are available in the market for treating existing cancers. This includes sipuleucel-T for metastatic hormone-refractory prostate cancer, nivolumab for metastatic melanoma, and pembrolizumab for advanced nonsmall-cell lung cancer. As cancer vaccines require some time to elicit an immune response, a delayed treatment effect is observed, resulting in a violation of the proportional hazards assumption. Thus, the traditional log-rank test may not be optimal for testing immuno-oncology drugs in randomized clinical trials. Moreover, the new immuno-oncology compounds have been shown to be very effective in prolonging overall survival. Therefore, it is desirable to implement a group sequential design with the possibility of early stopping for overwhelming efficacy. In this paper, we investigate the max-combo test, which utilizes the maximum of two weighted log-rank statistics, as a robust alternative to the log-rank test. The new test is implemented for two-stage designs with possible early stopping at the interim analysis time point. Two classes of weights are investigated for the max-combo test: the Fleming and Harrington (1981) Gρ,γ$G^{\rho , \gamma }$ weights and the Magirr and Burman (2019) modest (τ∗)$ (\tau ^{*})$  weights.

Adaptive enrichment trials: What are the benefits?

When planning a Phase III clinical trial, suppose a certain subset of patients is expected to respond particularly well to the new treatment. Adaptive enrichment designs make use of interim data in selecting the target population for the remainder of the trial, either continuing with the full population or restricting recruitment to the subset of patients. We define a multiple testing procedure that maintains strong control of the familywise error rate, while allowing for the adaptive sampling procedure. We derive the Bayes optimal rule for deciding whether or not to restrict recruitment to the subset after the interim analysis and present an efficient algorithm to facilitate simulation-based optimisation, enabling the construction of Bayes optimal rules in a wide variety of problem formulations. We compare adaptive enrichment designs with traditional nonadaptive designs in a broad range of examples and draw clear conclusions about the potential benefits of adaptive enrichment.

A gatekeeping procedure to test a primary and a secondary endpoint in a group sequential design with multiple interim looks.

Glimm et al. (2010) and Tamhane et al. (2010) studied the problem of testing a primary and a secondary endpoint, subject to a gatekeeping constraint, using a group sequential design (GSD) with K=2 looks. In this article, we greatly extend the previous results to multiple (K>2) looks. If the familywise error rate (FWER) is to be controlled at a preassigned α level then it is clear that the primary boundary must be of level α. We show under what conditions one α-level primary boundary is uniformly more powerful than another. Based on this result, we recommend the choice of the O'Brien and Fleming (1979) boundary over the Pocock (1977) boundary for the primary endpoint. For the secondary endpoint the choice of the boundary is more complicated since under certain conditions the secondary boundary can be refined to have a nominal level α'>α, while still controlling the FWER at level α, thus boosting the secondary power. We carry out secondary power comparisons via simulation between different choices of primary-secondary boundary combinations. The methodology is applied to the data from the RALES study (Pitt et al., 1999; Wittes et al., 2001). An R library package gsrsb to implement the proposed methodology is made available on CRAN.

Improved two-stage group sequential procedures for testing a secondary endpoint after the primary endpoint achieves significance.

In two-stage group sequential trials with a primary and a secondary endpoint, the overall type I error rate for the primary endpoint is often controlled by an α-level boundary, such as an O'Brien-Fleming or Pocock boundary. Following a hierarchical testing sequence, the secondary endpoint is tested only if the primary endpoint achieves statistical significance either at an interim analysis or at the final analysis. To control the type I error rate for the secondary endpoint, this is tested using a Bonferroni procedure or any α-level group sequential method. In comparison with marginal testing, there is an overall power loss for the test of the secondary endpoint since a claim of a positive result depends on the significance of the primary endpoint in the hierarchical testing sequence. We propose two group sequential testing procedures with improved secondary power: the improved Bonferroni procedure and the improved Pocock procedure. The proposed procedures use the correlation between the interim and final statistics for the secondary endpoint while applying graphical approaches to transfer the significance level from the primary endpoint to the secondary endpoint. The procedures control the familywise error rate (FWER) strongly by construction and this is confirmed via simulation. We also compare the proposed procedures with other commonly used group sequential procedures in terms of control of the FWER and the power of rejecting the secondary hypothesis. An example is provided to illustrate the procedures.

Comparing oncology clinical programs by use of innovative designs and expected net present value optimization: Which adaptive approach leads to the best result?

Designing an oncology clinical program is more challenging than designing a single study. The standard approaches have been proven to be not very successful during the last decade; the failure rate of Phase 2 and Phase 3 trials in oncology remains high. Improving a development strategy by applying innovative statistical methods is one of the major objectives of a drug development process. The oncology sub-team on Adaptive Program under the Drug Information Association Adaptive Design Scientific Working Group (DIA ADSWG) evaluated hypothetical oncology programs with two competing treatments and published the work in the Therapeutic Innovation and Regulatory Science journal in January 2014. Five oncology development programs based on different Phase 2 designs, including adaptive designs and a standard two parallel arm Phase 3 design were simulated and compared in terms of the probability of clinical program success and expected net present value (eNPV). In this article, we consider eight Phase2/Phase3 development programs based on selected combinations of five Phase 2 study designs and three Phase 3 study designs. We again used the probability of program success and eNPV to compare simulated programs. For the development strategies, we considered that the eNPV showed robust improvement for each successive strategy, with the highest being for a three-arm response adaptive randomization design in Phase 2 and a group sequential design with 5 analyses in Phase 3.

Optimizing the data combination rule for seamless phase II/III clinical trials.

We consider seamless phase II/III clinical trials that compare K treatments with a common control in phase II then test the most promising treatment against control in phase III. The final hypothesis test for the selected treatment can use data from both phases, subject to controlling the familywise type I error rate. We show that the choice of method for conducting the final hypothesis test has a substantial impact on the power to demonstrate that an effective treatment is superior to control. To understand these differences in power, we derive decision rules maximizing power for particular configurations of treatment effects. A rule with such an optimal frequentist property is found as the solution to a multivariate Bayes decision problem. The optimal rules that we derive depend on the assumed configuration of treatment means. However, we are able to identify two decision rules with robust efficiency: a rule using a weighted average of the phase II and phase III data on the selected treatment and control, and a closed testing procedure using an inverse normal combination rule and a Dunnett test for intersection hypotheses. For the first of these rules, we find the optimal division of a given total sample size between phases II and III. We also assess the value of using phase II data in the final analysis and find that for many plausible scenarios, between 50% and 70% of the phase II numbers on the selected treatment and control would need to be added to the phase III sample size in order to achieve the same increase in power. © 2014 The Authors. Statistics in Medicine published by John Wiley & Sons Ltd.

Adaptive sample size modification in clinical trials: start small then ask for more?

We consider sample size re-estimation in a clinical trial, in particular when there is a significant delay before the measurement of patient response. Mehta and Pocock have proposed methods in which sample size is increased when interim results fall in a 'promising zone' where it is deemed worthwhile to increase conditional power by adding more subjects. Our analysis reveals potential pitfalls in applying this approach. Mehta and Pocock use results of Chen, DeMets and Lan to identify when increasing sample size, but applying a conventional level α significance test at the end of the trial does not inflate the type I error rate: we have found the greatest gains in power per additional observation are liable to lie outside the region defined by this method. Mehta and Pocock increase sample size to achieve a particular conditional power, calculated under the current estimate of treatment effect: this leads to high increases in sample size for a small range of interim outcomes, whereas we have found it more efficient to make moderate increases in sample size over a wider range of cases. If the aforementioned pitfalls are avoided, we believe the broad framework proposed by Mehta and Pocock is valuable for clinical trial design. Working in this framework, we propose sample size rules that apply explicitly the principle of adding observations when they are most beneficial. The resulting trial designs are closely related to efficient group sequential tests for a delayed response proposed by Hampson and Jennison.

Model free audit methodology for bias evaluation of tumour progression in oncology.

Many oncology studies incorporate a blinded independent central review (BICR) to make an assessment of the integrity of the primary endpoint, progression free survival. Recently, it has been suggested that, in order to assess the potential for bias amongst investigators, a BICR amongst only a sample of patients could be performed; if evidence of bias is detected, according to a predefined threshold, the BICR is then assessed in all patients, otherwise, it is concluded that the sample was sufficient to rule out meaningful levels of bias. In this paper, we present an approach that adapts a method originally created for defining futility bounds in group sequential designs. The hazard ratio ratio, the ratio of the hazard ratio (HR) for the treatment effect estimated from the BICR to the corresponding HR for the investigator assessments, is used as the metric to define bias. The approach is simple to implement and ensures a high probability that a substantial true bias will be detected. In the absence of bias, there is a high probability of accepting the accuracy of local evaluations based on the sample, in which case an expensive BICR of all patients is avoided. The properties of the approach are demonstrated by retrospective application to a completed Phase III trial in colorectal cancer. The same approach could easily be adapted for other disease settings, and for test statistics other than the hazard ratio.

Type 2 Diabetes risk alleles in Peptidyl-glycine Alpha-amidating Monooxygenase influence GLP-1 levels and response to GLP-1 Receptor Agonists.

Patients with type 2 diabetes vary in their response to currently available therapeutic agents (including GLP-1 receptor agonists) leading to suboptimal glycemic control and increased risk of complications. We show that human carriers of hypomorphic T2D-risk alleles in the gene encoding peptidyl-glycine alpha-amidating monooxygenase (PAM), as well as Pam-knockout mice, display increased resistance to GLP-1 in vivo. Pam inactivation in mice leads to reduced gastric GLP-1R expression and faster gastric emptying: this persists during GLP-1R agonist treatment and is rescued when GLP-1R activity is antagonized, indicating resistance to GLP-1's gastric slowing properties. Meta-analysis of human data from studies examining GLP-1R agonist response (including RCTs) reveals a relative loss of 44% and 20% of glucose lowering (measured by glycated hemoglobin) in individuals with hypomorphic PAM alleles p.S539W and p.D536G treated with GLP-1R agonist. Genetic variation in PAM has effects on incretin signaling that alters response to medication used commonly for treatment of T2D.

An adaptive seamless phase II/III design for oncology trials with subpopulation selection using correlated survival endpoints.

Although the statistical methods enabling efficient adaptive seamless designs are increasingly well established, it is important to continue to use the endpoints and specifications that best suit the therapy area and stage of development concerned when conducting such a trial. Approaches exist that allow adaptive designs to continue seamlessly either in a subpopulation of patients or in the whole population on the basis of data obtained from the first stage of a phase II/III design: our proposed design adds extra flexibility by also allowing the trial to continue in all patients but with both the subgroup and the full population as co-primary populations. Further, methodology is presented which controls the Type-I error rate at less than 2.5% when the phase II and III endpoints are different but correlated time-to-event endpoints. The operating characteristics of the design are described along with a discussion of the practical aspects in an oncology setting.

Processes Underlying Glycemic Deterioration in Type 2 Diabetes: An IMI DIRECT Study.

OBJECTIVE: We investigated the processes underlying glycemic deterioration in type 2 diabetes (T2D). RESEARCH DESIGN AND METHODS: A total of 732 recently diagnosed patients with T2D from the Innovative Medicines Initiative Diabetes Research on Patient Stratification (IMI DIRECT) study were extensively phenotyped over 3 years, including measures of insulin sensitivity (OGIS), ß-cell glucose sensitivity (GS), and insulin clearance (CLIm) from mixed meal tests, liver enzymes, lipid profiles, and baseline regional fat from MRI. The associations between the longitudinal metabolic patterns and HbA1c deterioration, adjusted for changes in BMI and in diabetes medications, were assessed via stepwise multivariable linear and logistic regression. RESULTS: Faster HbA1c progression was independently associated with faster deterioration of OGIS and GS and increasing CLIm; visceral or liver fat, HDL-cholesterol, and triglycerides had further independent, though weaker, roles (R 2 = 0.38). A subgroup of patients with a markedly higher progression rate (fast progressors) was clearly distinguishable considering these variables only (discrimination capacity from area under the receiver operating characteristic = 0.94). The proportion of fast progressors was reduced from 56% to 8-10% in subgroups in which only one trait among OGIS, GS, and CLIm was relatively stable (odds ratios 0.07-0.09). T2D polygenic risk score and baseline pancreatic fat, glucagon-like peptide 1, glucagon, diet, and physical activity did not show an independent role. CONCLUSIONS: Deteriorating insulin sensitivity and ß-cell function, increasing insulin clearance, high visceral or liver fat, and worsening of the lipid profile are the crucial factors mediating glycemic deterioration of patients with T2D in the initial phase of the disease. Stabilization of a single trait among insulin sensitivity, ß-cell function, and insulin clearance may be relevant to prevent progression.

Optimal group-sequential designs for simultaneous testing of superiority and non-inferiority.

Confirmatory clinical trials comparing the efficacy of a new treatment with an active control typically aim at demonstrating either superiority or non-inferiority. In the latter case, the objective is to show that the experimental treatment is not worse than the active control by more than a pre-specified non-inferiority margin. We consider two classes of group-sequential designs that combine the superiority and non-inferiority objectives: non-adaptive designs with fixed group sizes and adaptive designs where future group sizes may be based on the observed treatment effect. For both classes, we derive group-sequential designs meeting error probability constraints that have the lowest possible expected sample size averaged over a set of values of the treatment effect. These optimized designs provide an efficient means of reducing expected sample size under a range of treatment effects, even when the separate objectives of proving superiority and non-inferiority would require quite different fixed sample sizes. We also present error spending versions of group-sequential designs that are easily implementable and can handle unpredictable group sizes or information levels. We find the adaptive choice of group sizes to yield some modest efficiency gains; alternatively, expected sample size may be reduced by adding another interim analysis to a non-adaptive group-sequential design.

TriMaster: randomised double-blind crossover study of a DPP4 inhibitor, SGLT2 inhibitor and thiazolidinedione as second-line or third-line therapy in patients with type 2 diabetes who have suboptimal glycaemic control on metformin treatment with or without a sulfonylurea-a MASTERMIND study protocol.

INTRODUCTION: Pharmaceutical treatment options for patients with type 2 diabetes mellitus (T2DM) have increased to include multiple classes of oral glucose-lowering agents but without accompanying guidance on which of these may most benefit individual patients. Clinicians lack information for treatment intensification after first-line metformin therapy. Stratifying patients by simple clinical characteristics may improve care by targeting treatment options to those in whom they are most effective. This academically designed and run three-way crossover trial aims to test a stratification approach using three standard oral glucose-lowering agents. METHODS AND ANALYSIS: TriMaster is a randomised, double-blind, crossover trial taking place at up to 25 clinical sites across England, Scotland and Wales. 520 patients with T2DM treated with either metformin alone, or metformin and a sulfonylurea who have glycated haemoglobin (HbA1c) >58 mmol/mol will be randomised to receive 16 weeks each of a dipeptidyl peptidase-4 inhibitor, sodium-glucose co-transporter-2 inhibitor and thiazolidinedione in random order. Participants will be assessed at the end of each treatment period, providing clinical and biochemical data, and their experience of side effects. Participant preference will be assessed on completion of all three treatments. The primary endpoint is HbA1c after 4 months of therapy (allowing a range of 12-18 weeks for analysis). Secondary endpoints include participant-reported preference between the three treatments, tolerability and prevalence of side effects. ETHICAL APPROVAL: This study was approved by National Health Service Health Research Authority Research Ethics Committee South Central-Oxford A, study 16/SC/0147. Written informed consent will be obtained from all participants. Results will be submitted to a peer-reviewed journal and presented at relevant scientific meetings. A lay summary of results will be made available to all participants. TRIAL REGISTRATION NUMBERS: 12039221; 2015-002790-38 and NCT02653209.

A multi-stage drop-the-losers design for multi-arm clinical trials.

Multi-arm multi-stage trials can improve the efficiency of the drug development process when multiple new treatments are available for testing. A group-sequential approach can be used in order to design multi-arm multi-stage trials, using an extension to Dunnett's multiple-testing procedure. The actual sample size used in such a trial is a random variable that has high variability. This can cause problems when applying for funding as the cost will also be generally highly variable. This motivates a type of design that provides the efficiency advantages of a group-sequential multi-arm multi-stage design, but has a fixed sample size. One such design is the two-stage drop-the-losers design, in which a number of experimental treatments, and a control treatment, are assessed at a prescheduled interim analysis. The best-performing experimental treatment and the control treatment then continue to a second stage. In this paper, we discuss extending this design to have more than two stages, which is shown to considerably reduce the sample size required. We also compare the resulting sample size requirements to the sample size distribution of analogous group-sequential multi-arm multi-stage designs. The sample size required for a multi-stage drop-the-losers design is usually higher than, but close to, the median sample size of a group-sequential multi-arm multi-stage trial. In many practical scenarios, the disadvantage of a slight loss in average efficiency would be overcome by the huge advantage of a fixed sample size. We assess the impact of delay between recruitment and assessment as well as unknown variance on the drop-the-losers designs.

Costs and Treatment Pathways for Type 2 Diabetes in the UK: A Mastermind Cohort Study.

INTRODUCTION: Medication therapy for type 2 diabetes has become increasingly complex, and there are few reliable data on the current state of clinical practice. We report treatment pathways and associated costs of medication therapy for people with type 2 diabetes in the UK, their variability and changes over time. METHODS: Prescription and biomarker data for 7159 people with type 2 diabetes were extracted from the GoDARTS cohort study, covering the period 1989-2013. Average follow-up was 10 years. Individuals were prescribed on average 2.4 (SD: 1.2) drugs with average annual costs of £241. We calculated summary statistics for first- and second-line therapies. Linear regression models were used to estimate associations between therapy characteristics and baseline patient characteristics. RESULTS: Average time from diagnosis to first prescription was 3 years (SD: 4.0 years). Almost all first-line therapy (98%) was monotherapy, with average annual cost of £83 (SD: £204) for 3.8 (SD: 3.5) years. Second-line therapy was initiated in 73% of all individuals, at an average annual cost of £219 (SD: £305). Therapies involving insulin were markedly more expensive than other common therapies. Baseline HbA1c was unrelated to future therapy costs, but higher average HbA1c levels over time were associated with higher costs. CONCLUSIONS: Medication therapy has undergone substantial changes during the period covered in this study. For example, therapy is initiated earlier and is less expensive than in the past. The data provided in this study will prove useful for future modelling studies, e.g. of stratified treatment approaches.

Adaptive seamless designs: selection and prospective testing of hypotheses.

There is a current trend towards clinical protocols which involve an initial "selection" phase followed by a hypothesis testing phase. The selection phase may involve a choice between competing treatments or different dose levels of a drug, between different target populations, between different endpoints, or between a superiority and a non-inferiority hypothesis. Clearly there can be benefits in elapsed time and economy in organizational effort if both phases can be designed up front as one experiment, with little downtime between phases. Adaptive designs have been proposed as a way to handle these selection/testing problems. They offer flexibility and allow final inferences to depend on data from both phases, while maintaining control of overall false positive rates. We review and critique the methods, give worked examples and discuss the efficiency of adaptive designs relative to more conventional procedures. Where gains are possible using the adaptive approach, a variety of logistical, operational, data handling and other practical difficulties remain to be overcome if adaptive, seamless designs are to be effectively implemented.

Efficient group sequential designs when there are several effect sizes under consideration.

We consider the construction of efficient group sequential designs where the goal is a low expected sample size not only at the null hypothesis and the alternative (taken to be the minimal clinically meaningful effect size), but also at more optimistic anticipated effect sizes. Pre-specified Type I error rate and power requirements can be achieved both by standard group sequential tests and by more recently proposed adaptive procedures. We investigate four nested classes of designs: (A) group sequential tests with equal group sizes and stopping boundaries determined by a monomial error spending function (the 'rho-family'); (B) as A but the initial group size is allowed to be different from the others; (C) group sequential tests with arbitrary group sizes and arbitrary boundaries, fixed in advance; (D) adaptive tests-as C but at each analysis, future group sizes and critical values are updated depending on the current value of the test statistic. By examining the performance of optimal procedures within each class, we conclude that class B provides simple and efficient designs with efficiency close to that of the more complex designs of classes C and D. We provide tables and figures illustrating the performances of optimal designs within each class and defining the optimal procedures of classes A and B.

Meta-analyses and adaptive group sequential designs in the clinical development process.

The clinical development process can be viewed as a succession of trials, possibly overlapping in calendar time. The design of each trial may be influenced by results from previous studies and other currently proceeding trials, as well as by external information. Results from all of these trials must be considered together in order to assess the efficacy and safety of the proposed new treatment. Meta-analysis techniques provide a formal way of combining the information. We examine how such methods can be used in combining results from: (1) a collection of separate studies, (2) a sequence of studies in an organized development program, and (3) stages within a single study using a (possibly adaptive) group sequential design. We present two examples. The first example concerns the combining of results from a Phase IIb trial using several dose levels or treatment arms with those of the Phase III trial comparing the treatment selected in Phase IIb against a control This enables a "seamless transition" from Phase IIb to Phase III. The second example examines the use of combination tests to analyze data from an adaptive group sequential trial.