letter to the editor

1 downloads 0 Views 34KB Size Report
Jul 10, 2008 - oncology: for patients or for industry? Trotta et al. [1] raise interesting points regarding ... 8 | August 2008 letter to the editor. Annals of Oncology ...
letter to the editor

Annals of Oncology

Stopping a trial early in oncology: for patients or for industry? Trotta et al. [1] raise interesting points regarding the size of a clinical trial and the consequences of stopping due to

1512 | letter to the editor

Volume 19 | No. 8 | August 2008

letter to the editor

Annals of Oncology

evidence of efficacy. They list trials in which stopping rules have been applied, including some stopped due to superior efficacy in the experimental arm. The objective of clinical evaluation of a new therapy is to collect valid and reliable information to allow the best decisions to be made for the benefit of current and future patients. As complete information will never be available, a balance has to be struck between collecting information and reaching a conclusion. Statisticians have developed methods to ensure that specified values of type I error (a) and power are respected. R. A. Fisher, working with agricultural data that arrive all together at harvest time, developed fixed-sample experimental designs. These can be implemented even when data arrive sequentially, as in clinical trials. However, sample size determination is not an exact matter as it depends on anticipating parameters such as the variability of the data, the average response rate or the overall survival pattern. When data arrive sequentially, there is an opportunity to use the accumulating evidence to reach an ideal sample size in a way that is impossible in settings such as agriculture. This ideal value will not be fixed in advance, but the procedure for determining it from the emerging data can be. Using such rules does not amount to ‘stopping a trial early’: it is a matter of stopping at precisely the right time. Trotta et al. point out that repeated analyses of accumulating data can lead to statistically significant results occurring only by chance (at a rate > a) and that analyses of sequential trials on the basis of methods for fixed-sample designs can lead to biased overestimates of treatment advantage and overstatement of the strength of the evidence (P values that are too small). They do not mention that proper sequential designs guaranteeing specified values of a and power have existed for many years, nor that methods exist for their analysis avoiding bias and leading to accurate P values [2–4]. Neither do they report whether such methods were used in the trials that they reviewed, although they recognize that ‘the studies analysed were formally well designed’. Certainly regulators are well aware of such methods and tend to insist on their proper use [5]. Trotta et al. indicate that the principal beneficiary from stopping a trial due to efficacy will be the trial sponsor. However, there is also benefit for patients otherwise

Volume 19 | No. 8 | August 2008

randomized to control when it is already evident that the experimental treatment is efficacious, and for those outside the trial otherwise denied the new therapy while patients continue to be recruited. Although Trotta et al. provide evidence that stopping rules are being used, they provide no evidence at all for their misuse. They do not reveal how the sample size for a clinical trial can be fixed so accurately that anything smaller is clearly inadequate. J. Whitehead1* & E. Cobo2 1

Medical & Pharmaceutical Statistics Research Unit, Department of Mathematics and Statistics, Fylde College, Lancaster University, Lancaster, UK, 2Departamento de Estadı´stica e Investigacio´n Operativa, Facultad de Matema´ticas y Estadı´stica, Universidad Polite´cnica de Catalun˜a, Barcelona, Spain (*E-mail: [email protected])

acknowledgements Disclosure of interest: Both authors are members of the steering committee (EC) and the data safety monitoring board (JW) of International Citicoline Trial in Acute Stroke, a sequential trial on a drug for stroke, sponsored by Ferrer International. They also provide consultancy on other sequential clinical trials for the pharmaceutical industry. As members of the Royal Statistical Society, both adhere its code of conduct.

references 1. Trotta F, Apolone G, Garattini S, Tafuri G. Stopping a trial early in oncology: for patients or for industry? Ann Oncol 2008; 19: 1347–1353. 2. Whitehead J. The Design and Analysis of Sequential Clinical Trials, 2nd edition. Chichester: Wiley 1997. 3. Jennison C, Turnbull BW. Group Sequential Methods with Applications to Clinical Trials. London: Chapman and Hall 2000. 4. Todd S. A 25-year review of sequential methodology in clinical studies. Stat Med 2007; 26: 137–252. 5. EMEA. Reflection paper on methodological issues in confirmatory clinical trials with flexible design and analysis plan. Doc. Ref. CHMP/EWP/2459/02. www.emea.europa.eu/pdfs/human/ewp/245902en.pdf (19 June 2008, date last accessed).

doi:10.1093/annonc/mdn410 Published online 10 July 2008

letter to the editor | 1513