Date

Book (and chapters)/Journal

Presented/ led by

Jan - Mar, 2013

Fairclough, DL. (2010). Design and Analysis of Quality of Life Studies in Clinical Trials. 2nd edition, CRC Press.

The book has been structured in a way that allows one to read each chapter independently. We find the arguments easy to follow and accessible especially for beginners. As such, the materials are not too technical for statisticians. We found the SAS and R codes most useful.

Ch. 2 Study Design and Protocol Development

Tom Morris

Ch. 3 Models for Longitudinal Studies I

Nick Parsons

Ch. 4 Models for Longitudinal Studies II

Nick Parsons

Ch. 5 Moderation and Mediation

Siew Wan Hee

Ch. 14 Composite Endpoints and Summary Measures

Ric Crossman

Ch. 6 Characterization of Missing Data

Helen Parsons

Ch. 7 Analysis of Studies with Missing Data

Peter Kimani

Ch. 8 Simple Imputation

Tom Hamborg

Ch. 9 Multiple Imputation

Joshua Pink

Ch. 10 Pattern Mixture and Other Mixture Models

Joshua Pink

Ch. 11 Random Effects Dependent Dropout

Tom Palmer

Ch. 12 Selection Models

Chris Knox

Ch. 15 Quality Adjusted Life-Years (QALYs) and Q-TWiST

Kamran Khan

Roberts, C. and Roberts, SA. (2005). Design and analysis of clinical trials with clustering effects due to treatment. Clinical Trials, 2: 152-162.

The paper discusses the design and analysis of clinical trial where there is a clustering effect inherent in the mechanism of treatments rather than only due to the randomisation mechanism. The authors identified two scenarios where this might occur, namely, group therapy (interventions delivered in a group setting) and therapist-delivered trials. The statistical design of such clinical trials is more complex than the design of conventional cluster randomised trials where the distribution of cluster sizes and intracluster correlation coefficient (ICC) in each arm can be assumed to be equal. In group therapy and therapist trials, the ICC within and between arms may differ and the mean cluster size is also likely to differ between arms. The paper presents and compares statistical models (random intercept, random coefficient and heteroscedastic) that may be adopted in such scenarios.

We felt that this paper provides a good, broad overview of models one should consider when analysing such trials. There were several typographical errors in the paper, in particular in the tables, which impede understanding. We felt that it would have been very helpful if the authors had provided the R and MLWin syntax that they have used in their analyses and simulation examples. Finally, it would have been beneficial if the authors had presented an example where usage of the random effects models would make a difference to the outcome of the analysis. This may be helpful in persuading clinical collaborators to adopt these models. However, we also noted that two analyses, one accounting for clustering and one that does not account for clustering (OLS analysis), yielding similar result, is a regular occurrence in real-life trials that we have been involved in.

Borenstein, M., Hedges, L. V., Higgins, J. P. T., and Rothstein, H. R. (2009). Introduction to meta-analysis. Wiley.

The book is an excellent introductory book for both statisticians and clinicians who have never done meta-analysis before. The examples were adequate and the insights and discussion of the principles, do’s and don’ts and why’s were presented in simple terms. The step-by-step calculations were presented clearly and logically that will encourage clinicians and students doing them on their own.

However, it was felt that inclusion of some examples on hazard ratio or time-to-event would be beneficial as these are common topics encountered in medical research. In part a bit more technical background would be appreciated. The selection of Excel for the software examples seemed like an odd choice as this software is very uncommon for meta-analysis in practice.

We would recommend readers who are planning to do meta-analysis and are completely unfamiliar with the topic to start exploring chapters in Part 9 first as these chapters present and discuss the question or objective of the intended meta-analysis, whether it is suitable to do one and what should be reported if an analysis is performed and how should they be reported. These are helpful issues as one should not jump into doing a meta-analysis only to realise that it was not necessary or it has been inadequate or not answering the question of the study!

 Part 2: Effect size and precision (chapters 3 to 9)

Part 3: Fixed-effect versus random-effects models (chapters 10 to 14)

 Part 4A: Heterogeneity (chapters 15 to 18)

 Part 4B: Heterogeneity (chapters 19 to 21)

 Part 5: Complex data structyres (chapters 22 to 26)

 Part 6: Other issues (chapters 27 to 30)

 Part 7: Issues related to effect size (chapters 31 to 34)

 Part 8: Further methods (chapters 35 to 38)

 Part 9: Meta-analysis in context (chapters 39 to 43)