Primary Research Areas

• Statistical methods for dose-finding studies
• Modelling pharmacokinetic and pharmacodynamic data
• Proof-of-concept studies
• Adaptive designs for clinical trials
• Ordered categorical data
• Survival data
• Meta-analysis of clinical trials
• Statistical methods in pharmacogenetics

Statistical methods for dose-finding studies

John Whitehead, Anne Whitehead, Helene Thygesen, Gareth Ridall

Our work concerns first-in-man studies of new drugs, both in the setting of oncology where subjects are patients and responses are toxicities, and more generally where subjects are healthy volunteers and responses are pharmacokinetic and pharmacodynamic assessments. Research has centred on the development of Bayesian decision procedures, and on phase I/II designs in which both adverse events and (surrogate) measures of benefit are assessed. Also of interest is the identification of the optimal doses for more than one drug in combination therapies.

Some recent publications:

• Zhou, Y., Whitehead, J. and Korhonen, P. (2008). Implementation of a Bayesian design in a dose-escalation study of an experimental agent in healthy volunteers. Biometrics 64, 299-308.
• Whitehead, J., Zhou, Y., Mander, A., Ritchie, S., Sabin, A. and Wright, A. (2006). An evaluation of Bayesian designs for dose-escalation studies in healthy volunteers. Statistics in Medicine 25, 433-445.
• Whitehead, J., Zhou, Y., Patterson, S., Webber, D. and Francis, S., (2001). Easy-to-implement Bayesian methods for dose-escalation studies in healthy volunteers. Biostatistics 2, 47-61.
• Whitehead, A., Whitehead, J., Todd, S., Zhou, Y. and Smith, M.K. (2007). Fitting models for the joint action of two drugs using SAS®. Online: Pharmaceutical Statistics. DOI: 10.1002/pst.312.

Modelling pharmacokinetic and pharmacodynamic data

Thomas Jaki, John Whitehead and Helene Thygesen

This is a new area of interest for MPS, and one on which a new professional development course is being prepared. Most work to date has concerned the estimation of the area under the concentration versus time curve for different sampling schemes. We are also interested in problems of scaling drug effects from animals to man.

Some recent publications:

• Wolfsegger M. J., Jaki T. (2009) Assessing systemic drug exposure in repeated dose toxicity studies in the case of complete and incomplete sampling. Biometrical Journal. In press.
• Jaki T., Wolfsegger M. J. (2009) A theoretical framework for estimation of AUCs in complete and incomplete sampling designs. Statistics in Biopharmaceutical Research. Vol 1(No 2), 176-184.
• Jaki T., Wolfsegger M. J., Ploner M. (2009) Confidence Intervals for Ratios of AUCs in the Case of Serial Sampling: A Comparison of Seven Methods. Pharmaceutical Statistics. Vol 8(No 1), 12-24.
• Wolfsegger, M. J., Jaki, T. (2005). Estimation of AUC from 0 to infinity in serial sacrifice designs. Journal of Pharmacokinetics and Pharmacodynamics 32, 757-766.

Proof-of-concept studies

John Whitehead, Thomas Jaki

We are interested in both Bayesian and frequentist approaches to the determination of sample size for phase II studies. Such studies may be conducted to make a go/no go decision for a single treatment, or to select one or more treatments or doses for further study. It is often desirable to include an interim analysis within the design, although less common to need more than one. Futility analyses are very sensible at this early stage of clinical evaluation. Often the decision to proceed has to be taken on the basis of an early endpoint that will not be suitable for later phase III studies: the sample size for such a trial should be related to the eventual treatment effect desired in terms of the definitive long-term outcome.

Some recent publications:

• Whitehead, J. and Valdés-Márquez, E. (2008). Bayesian sample size for exploratory clinical trials incorporating historical data. Statistics in Medicine 27, 2307-2327.
• Stallard, N., Whitehead, J. and Cleall, S. (2005). Decision-making in a phase II clinical trial: a new approach combining Bayesian and frequentist concepts. Pharmaceutical Statistics 4, 119-128.

Adaptive designs for clinical trials

John Whitehead, Anne Whitehead, Thomas Hamborg

There is substantial experience within MPS of developing and implementing group sequential designs for clinical trials, and we are currently involved in the conduct of interim analyses for several trials that we have previously designed. We have also worked on methods for sample size reviews and their implementation. Current research includes designs with a single interim analysis conducted to detect a treatment by factor interaction. The factor might concern the presence of a biomarker, which might be genetic. The choices following the interim analysis are to continue the trial with all subjects, restrict recruitment to those who are bioassay positive, or stop the study due to futility. The overall type I error and a form of power are specified.

Some recent publications:

• Whitehead, J. (1997). The Design and Analysis of Sequential Clinical Trials. Revised Second Edition, Chichester: Wiley.
• Sooriyarachchi, M. R., Whitehead, J., Matsushita, T., Bolland, K. and Whitehead, A. (2003). Incorporating data received after a sequential trial has stopped into the final analysis: implementation and comparison of methods. Biometrics 59, 701-709.
• Whitehead, J., Whitehead, A., Todd, S., Bolland, K. and Sooriyarachchi, M. R. (2001). Mid-trial reviews for sequential clinical trials. Statistics in Medicine 20, 165-176.

Ordered categorical data

Anne Whitehead, John Whitehead, Ting-Li Su

One strand of our current research concerns the correlation between the score statistics for comparing two treatments when these are based on different ordinal assessment scales. For example, in stroke trials the assessment scales might be the Barthel index, the modified Rankin score and the NIH stroke scale. One approach is to use generalised estimating equations, but we are seeking an alternative that is more robust when prognostic factors with many levels (such as treatment centre) are to be fitted. Another interest is repeated assessments of patients on an ordinal scale, with the primary response being the last of them. During interim analyses, some patients will have only short-term assessments. We are building on our work with binary data, and developing methods which allow the inclusion of these patients, without the need to assume any longitudinal model.

Some recent publications:

• Whitehead, A., Sooriyarachchi, M. R., Whitehead, J. and Bolland, K. (2008). Incorporating intermediate binary responses into interim analyses of clinical trials: a comparison of four methods. Statistics in Medicine 27, 1646-1666.
• Sooriyarachchi, M. R., Whitehead, J., Whitehead, A., and Bolland, K. (2006). The sequential analysis of repeated binary responses. Statistics in Medicine 25, 2196-2214.
• Dark, R., Bolland, K. and Whitehead, J. (2003). Statistical methods for ordered categorical data based on a constrained odds model. Biometrical Journal 45, 453-470.

Survival data

John Whitehead, Zakiyah Zain

As for ordinal data, the correlation between different score statistics for comparing the same two treatments is of interest. In the context of survival data, the score statistics will be logrank statistics or their counterparts from a Cox’s proportional hazards regression model adjusting for prognostic factors. The different score statistics will correspond to different waiting times, such as time to death and time to disease progression, or time to loss of sight in the right eye and time to loss of sight in the left eye. Correlations are needed to enable bivariate or multivariate analyses to be made or global tests to be performed. One possible approach is to use the Wei-Lin-Weissfeld method, but that is difficult to understand and does not reduce to a simple logrank analysis in the case of a single endpoint.

Some recent publications:

• Branson, M. and Whitehead, J. (2002). Estimating a treatment effect in survival studies in which patients switch treatment. Statistics in Medicine 21, 2449-2463.
• Whitehead, J. (2001). Predicting the duration of sequential survival studies. Drug Information Journal 35, 1387-1400.

Meta-analysis of clinical trials

Anne Whitehead

MPS has a long-term interest in developing methodology for meta-analysis, particularly individual patient data approaches, and in the conduct of such analyses. Recent work has included meta-analyses of studies in which repeated observations have been made on subjects at different time points across studies. The problems arise when only summary data from published papers are available, and involve the handling of correlation between time points and of missing observations. The properties of cumulative meta-analyses, and the problems of inflating type I error are also of interest.

Some recent publications:

• Whitehead, A. (2002). Meta-analysis of Controlled Clinical Trials. Wiley: Chichester.
• Whitehead, A., Perdomo, C., Pratt, R.D., Birks, J., Wilcock, G.K. and Grimley Evans, J. (2004). Donepezil for the symptomatic treatment of patients with mild to moderate Alzheimer’s disease: a meta-analysis of individual patient data from randomised controlled trials. International Journal of Geriatric Psychiatry 19, 624-633.
• Oliver, D., Connelly, J.B., Victor, C.R., Shaw, F.E., Whitehead, A., Genc, Y., Vanolli, A., Martin, F.C. and Gosney, M.A. (2006). Strategies to prevent falls and fractures in hospitals and care homes and effect of cognitive impairment: systematic review and meta-analyses. (2006). Online: Brit Med Journal, DOI: 10.1136/bmj.39049.706493.55.

Statistical methods in pharmacogenetics

John Whitehead, Helene Thygesen, Ting-Li Su, Thomas Jaki, Matt Sperrin

We are interested in the relationship between the human genome and the risk of adverse drug reactions. We devised a sequential procedure to identify such an association governed by a limit on type I error which was valid even though the genome itself comprises thousands or hundreds of thousands of separate but possibly correlated SNPs. Work is now centred on what to do following the discovery (or suspicion) of such a relationship. In particular, can an exclusion criterion be devised so that subjects whose genome suggests an excessive risk of an adverse drug reaction do not enter clinical studies, and the risk of subsequent ADRs is reduced. A Bayesian formulation is being used and the repeated review of the exclusion rule is being considered.

Some recent publications:

• Kelly, P., Zhou, Y., Whitehead, J. and Bowman, C. (2008). Sequential testing for a gene-drug interaction in a genomewide analysis. Statistics in Medicine 27, 2022-2034.
• Kelly, P., Stallard, N., Zhou, Y., Whitehead, J. and Bowman, C. (2006). Sequential genome-wide association studies for monitoring adverse events in the clinical evaluation of new drugs. Statistics in Medicine 25, 3081-3092.