Wednesday 31st October 2012

Adaptive designs in clinical research (half day meeting)

Joint meeting with Clinical Trials Research Unit, University of Leeds

Dr James Wason (MRC Biostatistics Unit, Cambridge) [Presentation Slides]

Design of Multi-Arm Multi-Stage Trials

When there are multiple experimental treatments awaiting testing, multi-arm multi-stage (MAMS) trials provide large gains in efficiency over separate randomised trials of each treatment. They allow a shared control group, dropping of ineffective treatments before the end of the trial and stopping the trial early if sufficient evidence of a treatment being superior to control is found. One type of MAMS design proposed is a generalisation of a group-sequential design to multiple experimental treatments. This class of trial involves specification of futility and stopping boundaries which determine the (random) number of treatments that get through to each stage. I will discuss the choice of these stopping boundaries, and how they can be chosen in an optimal way. A disadvantage of group-sequential MAMS trials is that the sample size used is a random variable, which makes applying for funding and logistical planning difficult. An alternative MAMS design is a drop-the-loser design. In drop-the-loser designs a pre-specified number of experimental treatments are dropped at each stage, and so the sample size used is fixed, which is a big advantage. I shall discuss some recent work on drop-the-loser designs and compare their properties to those of group-sequential MAMS designs.

Dena Cohen (CTRU, University of Leeds)

Incorporating Emerging Therapies into Ongoing Randomised Clinical Trials

My planned research aims to investigate the incorporation of an emerging therapy as a new randomisation arm to be included in a clinical trial that is already open to recruitment.

It may take many years to run a clinical trial from concept to reporting, within a rapidly changing drug development environment. Hence in order for trials to be most useful to inform policy and practice it is advantageous for them to be able to adapt to the changing environment where possible. However, currently there is very little literature regarding statistical methods or considerations when adapting a trial by adding a new treatment midway through recruitment. It is becoming increasingly desirable to researchers, regulators and patients to allow this adaptation to be made within a clinical trial to ensure that the trial remains current, and the overall time and cost for determining optimal therapies is minimised.

The objectives of this research are to identify, assess and investigate statistical methods and design considerations when incorporating an emerging therapy as a new randomisation arm in an ongoing clinical trial. These considerations may include, for example, conservation of error, appropriate use of early control data, and optimal allocation ratios. It is essential to ensure that the research integrity is not compromised when adapting a trial in this way, because otherwise the outcomes may be uncertain, which would be unethical and an unacceptable waste of resources. The planned outcomes will recommend when and how this adaptation is possible and advantageous, comprehensively detailing the statistical considerations.

The focus will be on trials in haematological oncology, although generalisability will be ensured to trials with survival outcomes. Recent advances in the treatment of Chronic Lymphocytic Leukaemia (CLL) have led to multiple new treatments being developed which show considerable early promise. There are a number of early phase trials currently in progress, and it is not appropriate to wait for the results of every promising early phase trial in order to design the next large phase III trial. However with the median progression-free survival in newly diagnosed patients approaching 5 years, a phase III trial is likely to take at least 10 years to be reported. If it was possible to add a new promising treatment to an ongoing trial if it emerged during the recruitment phase, this might facilitate selection of the best of the emerging treatments, reduce time to approval of the optimal therapy, reduce the overall costs of trial conduct compared to running separate trials and help to keep the outcomes current.

This presentation will discuss the research aims, detailing some of the considerations to be investigated.

Dr Patrick Phillips (MRC Clinical Trials Unit, London) [Presentation Slides]

Adaptive Trial Designs for Late-Phase Clinical Trials to Improve the Treatment of Tuberculosis

Tuberculosis (TB) is considered the oldest infectious disease known to man, but continues to be major global health problem in the 21st Century. The World Health Organisation declared TB a global emergency in 1993, and in 2010 there were 8.8 million new cases and over 1 million deaths from TB.

The current standard of care is a 4-drug 6-month regimen that has been consistently shown to have efficacy of 90-95% in randomised controlled trials in a variety of settings. These numbers are not reflected in clinical practice with cure rates unfortunately much lower than those seen in clinical trials. The rising levels of drug resistance and co-infection with HIV also contribute to global morbidity and mortality.

New combination regimens for TB that are shorter, safer and easier to adhere to are urgently needed. A growing number of novel combination regimens of new and repurposed drugs are being studied in phase II and phase III clinical trials. Relapse-free cure is the established endpoint for TB clinical trials and therefore phase III trials following patients for a minimum of 18 months are necessary to confirm the efficacy of a new combination regimen. Development of new regimens to shorten and improve the treatment of drug sensitive or drug resistant TB is a long process with multiple phase I safety and phase II efficacy trials even before a combination enters phase III.

Adaptive and other novel trial designs for phase II and phase III trials are drug development tools that can be used to speed combination development. Novel trial designs, including the Multi-Arm Multi-Stage (MAMS) design, that are being implemented by different research groups in current and planned phase II and phase III clinical trials to improve the treatment of TB will be discussed, including presentation of how these designs are more efficient than the traditional fixed sample size designs.

Helen Marshall (CTRU, University of Leeds) [Presentation Slides]

Incorporation of an Additional Interim Analysis During the Running of a Randomised Clinical Trial Using Group Sequential Design Methodology

AZURE is a multi-centre, randomised trial investigating adjuvant zoledronic acid in 3360 breast cancer patients; disease-free survival (DFS) is the primary endpoint. The trial design incorporated one interim analysis to assess for efficacy when at least half the number of DFS events had occurred. This was carried out in 2008 at a stringent alpha level (two-sided 0.005) to retain an overall two-sided 5% significance level using O`Brien and Fleming`s (1979) alpha spending function; the outcome was no data release. Due to a considerably lower than expected DFS event rate, final analysis is now estimated to occur in 2013; however in 2010 clinical practice had clearly started to change in the absence of confirmatory evidence. The Trial Steering Committee therefore felt a change in the analysis plan was desirable to inform current and future clinical practice in a timely way.

To preserve study integrity and to retain the overall alpha level, it was agreed to conduct an additional interim analysis when at least 75% of the number of DFS events had occurred and to include stopping rules for both efficacy and `lack of benefit` that were seen to be clinically meaningful; these were developed with an independent statistician not having access to the first interim analysis result. Probabilities of declaring a false-positive and false-negative result at the second interim analysis were chosen to be 0.5% (one-sided) and 5% respectively. Using group sequential design methodology (Whitehead, 2010), this corresponded to DFS hazard ratio (HR) stopping boundaries of 0.83 and 0.94: declare efficacy if HR<0.83, declare `lack of benefit` if HR>0.94 or no data release if HR = (0.83, 0.94).

Changes to an analysis plan may be required during the running of a trial. Incorporation of an additional interim analysis using group sequential design methodology is an appropriate solution.

The meeting will be held at University of Leeds, Worsley Building, Level 8, Room X, 2pm until 5pm, no registration prior to the meeting required, all welcome.