PHARMACEUTICAL STATISTICS

Pharmaceut. Statist. 2009; 8: 84–86

Published online in Wiley InterScience

(www.interscience.wiley.com) DOI: 10.1002/pst.374 LITERATURE REVIEWS

Literature Review October–December

2008

Kevin Carroll1,�,y, S. Krishna Padmanabhan2 and Andrew Stone3

1AstraZeneca Pharmaceuticals, CMOs Office, Alderley Park, Macclesfield, UK2Wyeth Research and Development, Collegeville, PA, USA3AstraZeneca Pharmaceuticals, Clinical Development, Alderley Park, Macclesfield, UK

INTRODUCTION

This review covers the following journals receivedduring the period from October to December2008:� Applied Statistics, volume 57, part 5.� Biometrical Journal, volume 50, issues 5 and 6.� Biometrics, volume 64, issue 4.� Biometrika, volume 95, issue 4.� Biostatistics, volume 9, part 4.� Clinical Trials, volume 5, parts 4, 5 and 6.� Contemporary Clinical Trials, volume 29, part 6.� Drug Information Journal, volume 42, parts 5

and 6.� Journal of Biopharmaceutical Statistics,

volume 18, parts 5 and 6.� Journal of the Royal Statistical Society, Series

A, volume 171, part 4.� Statistics in Medicine, volume 27, parts 21–29.� Statistical Methods in Medical Research,

volume 17, parts 5 and 6.

SELECTED HIGHLIGHTS FROMTHE LITERATURE

Regulation

This paper reviews China’s recently amended andreleased Provisions for Drug Registration.

� Zhao N et al. On the amendments of China’sprovisions for drug registration. Drug Informa-tion Journal 2008; 42:467–475.

Interim analyses, flexible designs and data mon-

itoring committees

Inverse dose–response estimation refers to theinference of an effective dose of some agent thatgives a desired probability of response. Theauthors consider inverse dose response for twoagents, using an adaptive direction-samplingalgorithm.� Hu et al. Bayesian estimation of inverse dose

response. Biometrics 2008; 64(4):1223–1230.

Detecting genetic markers with biologicallyrelevant effects remains a challenge due to multipletesting. In this paper, the authors develop two-stage designs for screening genetic markers whenthe cost of measurements is high.

� Moerkerke B et al. Optimal screeningfor promising genes in 2-stage designs.Biostatistics 2008; 9(4):700–714.

A simple and adaptive two-dimensional dosefinding is proposed. This design can accommodateany type of single-agent methodology in a two-dimensional dose finding setting (combinationtherapy).� Yuan Y et al. Sequential continual reassessment

method for two-dimensional dose finding.Statistics in Medicine 2008; 27(27):5664–5678.yE-mail: kevin.carroll2@astrazeneca.com

*Correspondence to: Kevin Carroll, AstraZeneca Pharma-ceuticals, CMOs Office, Alderley Park, Macclesfield, SK104TG, UK.

Copyright r 2009 John Wiley & Sons, Ltd.

Maintaining the integrity of a clinical trial whenan interim analysis is performed is clearly essential.Some simple, practical guidelines for the interimreview of clinical trial data are offered in thefollowing:

� King DW, Jokinen JD. Some guidelines forinterim data review of clinical trial data. DrugInformation Journal 2008; 42:449–453.

Data analysis

This paper delves into a common situation in drugdevelopment where it is often of interest to exploresuperiority of an agent in a non-inferiority trial(post hoc). The authors provide a closed-testprocedure for this endeavor.

� Logan BR et al. Superiority inferences onindividual endpoints following non-inferioritytesting in clinical trials. Biometrical Journal2008; 50(5):693–703.

A very pertinent issue for oncology trials usingthe Simon’s two-stage design is described in thefollowing paper. The authors consider someapproaches in handling situations where devia-tions or interruptions from the original Simon’stwo-stage design occur because of slower thanexpected patient recruitment.

� Wu Y et al. Approaches to handling data whena phase II trial deviates from the pre-specifiedSimon’s two-stage design. Statistics inMedicine 2008; 27(29):6190–6208.

In the previous review we highlighted a paper,by Moser and McCann, which reformulated thehazard ratio in a way to make it more meaningfulto clinicians. The following paper points outmodifications necessary in the presence of censor-ing and should be read in conjunction; the paperstogether providing a useful adjunctive descriptionof the hazard ratio.

� Buyse M. Reformulating the hazard ratio toenhance communication with clinical investi-gators. Clinical Trials 2008; 6:641–642.

We often take many of our conventions forgranted and it can be useful sometimes to go backto first principles. The following article takes athought provoking look at one such area – one-sided or two-sided testing.

� Freedman LS. An analysis of the controversyover classical one-sided tests. Clinical Trials2008; 6:635–640.

The effect of stratification on the power of manytests has been studied. In the following article theauthors provide practical guidance, on the basis ofa simulation study, of when to stratify analysesusing the Mann–Whitney test.

� Qu Y, Zhao YD, Rahardja D. WilcoxonMann–Whitney test: stratify or not?Journal of Biopharmaceutical Statistics 2008;18(6):1103–1111.

When making between trials, indirect compar-isons, as is essential, say, when an estimate ofeffectiveness versus placebo is desired, but, due topractical and ethical constraints, a new drug canonly be trialled against an active control, can beproblematic. A review of some of the key issues isprovided in the following:

� Julious SA, Wang S-J. How biased are indirectcomparisons, particularly when comparisonsare made over time in controlled trials? DrugInformation Journal 2008; 42:625–633.

Surrogate endpoint validation

Volume 17, issue 5 of Statistical Methods in MedicalResearch is devoted to matters concerning surrogateendpoint validation. A series of authors analysedthe same datasets using different methodologies,which contrasts nicely the varying approaches tosurrogate endpoint validation. The whole issue isworth a read with the following of most interest:

� Buyse M et al. Individual- and trial-levelsurrogacy in colorectal cancer. StatisticalMethods in Medical Research 2008; 17:467–475.

Copyright r 2009 John Wiley & Sons, Ltd. Pharmaceut. Statist. 2009; 8: 84–86DOI: 10.1002/pst

Literature Review 85

� Green EM et al. Surrogate endpoint valida-tion: statistical elegance versus clinical rele-vance. Statistical Methods in Medical Research2008; 17:477–486.

� Hughes MD. Practical issues arising in anexploratory analysis evaluating progression-free survival as a surrogate endpoint foroverall survival in advanced colorectal cancer.Statistical Methods in Medical Research 2008;17:487–495.

Design

This paper contains a broad overview of methodsfor MTD finding in oncology studies, followed byinteresting discussions and a rejoinder.

� Gerke O et al. Optimal phase I dose-escalationtrial designs in oncology – a simulation study.Statistics in Medicine 2008; 27(26):5329–5344.

Miscellaneous (publication bias)

Publication bias is a major and intractableproblem in meta-analysis. The authors introducea procedure to make inferences in this contextbased on the assumption that the probability thata paper is published may depend in someunspecified way on the P-value being claimed bythat study.

� Copas J et al. A robust P-value for treatmenteffect in meta-analysis with publication bias.Statistics in Medicine 2008; 27(21):4267–4278.

Additionally, we would like to highlight thevolume 27, issue 30 of Statistics in Medicine. Itcontains papers from the 28th Annual Conferenceof the International Society for Clinical Biostatis-tics. Many will be of interest to pharmaceuticalstatisticians.

Copyright r 2009 John Wiley & Sons, Ltd. Pharmaceut. Statist. 2009; 8: 84–86DOI: 10.1002/pst

86 Literature Review