North West Hub for Trials Methodology Research

which include meta-analysis of time-to-event data [3-7], longitudinal data [1] and individual patient data [3-5,8]; outcome reporting and publication bias [10-17]; and multiple treatment comparisons [2]. Building on existing work and in partnership with the Cochrane Collaboration, future research projects will focus on using findings of systematic reviews to inform the design, conduct and analysis of new clinical trials. One area of particular interest is exploiting covariate data from IPD reviews to improve design and interpretation of clinical trials which can be difficult in heterogeneous populations such as epilepsy. See key publications by our group.

Statistical methods for the proper treatment of data of this form, merging information from the two sources, are currently both under-developed and under-used in medical research. The MRC funded JoineR project (“Statistical methodology for the analysis of longitudinal data”) aims to develop new statistical methodologies for the analysis of complex longitudinal data structures, implement the methods in corresponding user-friendly software in the R language and disseminate the results to the medical research community. Standard methods for joint modelling of longitudinal and survival data allow for one event with a single mode of failure and an assumption of independent censoring. Williamson et al (2008) [9] have recently described and applied an extension to joint modelling to allow for competing risks of a single event with multiple possible causes. Further projects will build on existing work in this area.

Methods for improving trial management and allowing more complete data analyses, such as adjusting for non-compliance in clinical trials, will be explored, primarily through research projects conducted in the MSc in Clinical Research Administration programme.

It can occur when treatment schedules are longitudinal, resulting in the possibility of incomplete treatment or drop out, or when toxicity or side-effects of treatment result in treatment switches. Such treatment schedules are common in chronic diseases such as epilepsy and cancer. A project is underway to develop and apply methods of adjustment for non-compliance in the analysis of randomised controlled trials with survival outcomes in order to estimate treatment efficacy.

We have developed methods for predicting the impact of non-adherence on treatment effectiveness [Clin Pharmacol Ther 2003;74:1-8 ; Br J Clin Pharmacol 2008;65:871-8] and cost-effectiveness [Value in Health 2007;10:498-509]. Research in these areas has led to a good-practice checklist for the design and analysis of prospective studies that assess adherence [Clin Ther 2009;31:421-35] and a framework for the economic assessment of adherence-enhancing interventions [Pharmacoeconomics 2007; 25(8):621-35].

Research is being focussed on assessing the use, within trials, of patient-reported resource-use instruments (e.g. questionnaires, diaries etc where the patients report GP visits, medication usage, inpatient visits, outpatient visits, home visits, time off work, travelling expenses, medical procedures). This is a two-part process: a systematic review of HTA-funded trials [Value in Health 2010; in press] followed by the compilation of a compendium of resource-use questionnaires, with a critical assessment of their validity. It follows from earlier work on methods for the estimation of drug costs in economic evaluations [Pharmacoeconomics 2009;27:635-43].

For example, the International League Against Epilepsy has recommended retention time, defined as the time to withdrawal of the randomised drug or addition of another, be one of the primary endpoints for clinical trials of anti-epileptic drugs (AEDs). Patients may decide to withdraw from such treatment because of unacceptable adverse effects (UAE) or switch to an alternative AED because of inadequate seizure control (ISC). The reduction in side effects may be at the expense of a reduction in seizure control but it would be hoped that the latter would be within the defined margin of equivalence. Overall analysis of retention time may miss such differential effects of AEDs on the reasons for withdrawal, which may differ in terms of their relative importance for patients. An analysis of one event type which censors those patients who suffer a different event type, at the time they experience that other event, may be misleading because such an analysis assumes that the competing risks are independent or equivalently that such censoring is non-informative [1].

In another example, in trials designed to delay or avoid irradiation among children with malignant brain tumor, irradiation after disease progression is an important event. However, patients who have disease progression may decline radiotherapy (RT), or those without disease progression may opt for elective RT because of physician or parental decision. Current methods for quantifying the need for RT in such trials is based on the RT-free survival rate obtained via the Kaplan-Meier method of survival analysis. It does not evaluate how each competing event contributes to the delay or advancement of irradiation. To accurately describe the cumulative need for RT in such instances, it is crucial to account for these distinct events and to evaluate how each contributes to the delay or advancement of irradiation via a competing risks analysis [2].

There is considerable debate regarding the choice of test for treatment difference in a randomized clinical trial in the presence of competing risks. Two common methods for testing treatment effects in competing risks situations are the logrank test, corresponding to a test of equality of cause-specific hazards, and Gray’s test based on the cumulative incidence approach. The logrank test censors other event types at the time they occur and assumes the competing risks are independent. The approach based on cumulative incidence makes no such assumption. Our paper [3] investigated these approaches to testing of treatment effects in the presence of competing events.  This paper provides simulation results for the log-rank test comparing cause-specific hazard rates and Gray's test comparing cause-specific cumulative incidence curves. Competing risks settings were considered where all competing events are of interest, events may be negatively correlated, and the degree of correlation may differ in the two treatment groups. In settings where there are treatment effects in opposite directions for the two event types, e.g. the situation in the SANAD trial of AEDs, Gray's test has greater power to detect treatment differences than log-rank analysis.

[1] Williamson PR, Tudur Smith C, Josemir W Sander and Marson AG. Importance of competing risks in the analysis of anti-epileptic drug failure. Trials 2007, 8:12  doi: 10.1186/1745-6215-8-12

[2] Tai B-C, Grundy RG and Machin D. On the importance of accounting for competing risks in pediatric cancer trials designed to delay or avoid radiotherapy:I. Basic concepts and first analyses. Int J Radiation Oncology Biol Phys; Vol. 76, No. 5, pp. 1493–1499, 2010 doi:10.1016/j.ijrobp.2009.03.035

[3] Williamson PR, Kolamunnage-Dona R, Tudur Smith C: The influence of competing-risks setting on the choice of hypothesis test for treatment effect. Biostatistics; 2007 Oct;8(4):689-94