European Heart Journal Advance Access originally published online on January 24, 2007
European Heart Journal 2007 28(3):381; doi:10.1093/eurheartj/ehl480
Prospective randomized study comparing amiodarone vs. amiodarone plus losartan vs. amiodarone plus perindopril for the prevention of atrial fibrillation recurrence in patients with lone paroxysmal atrial fibrillation
Weill Medical College of Cornell University
Department of Internal Medicine
Lincoln Hospital
234, East 149th Street
Bronx
New York 10451
USA
Fax: +1 718 579 4836
E-mail address: bvkanna{at}aol.com
Department of Internal Medicine
Lincoln Hospital
New York
USA
In reference to the interesting study by Yin et al.,1 we would like to raise the following comments regarding randomization and effect size.
The study design follows a non-concealed allocation sequence by serially placing the eligible patients into each of the three groups, and the assigned patients to each of the treatment groups were followed in an open-labelled fashion. Non-concealment may not lead to bias because the allocation in the study is unrelated to any patient baseline characteristics. However, the open treatment allocation may shape the investigator's decision to recruit patient into a particular group, resulting in treatment groups that are not comparable.2 In addition, lack of random concealed allocation makes the differences between treatment groups non-representative of the differences between random samples from a single population. Therefore, the investigators should not mistakenly equate this allocation method to randomization, a process in which the allocation is random and concealed.3
While calculating sample size in clinical studies, the expected event rate in the control group is overestimated usually because of inclusion of stable subjects with limited comorbidity,4 as noted in this study (50% predicted event rate vs. 41% observed event rate among controls). Further, the investigators also overestimated the effect size reduction, i.e. the risk reduction between the control and two treatment groups (60% expected vs. 54 and 41% observed). Even if the expected event rate for the control group remains constant, sample size requirement will increase by the inverse square of the effect size reduction. Overestimation of the effect size reduction could lead to diminished power and limit internal validity of study findings.5
References
- Yin Y, Dalal D, Liu Z, Wu J, Liu D, Lan Y, Dai Y, Su L, Ling Z, She Q, Luo K, Woo K, Dong J. (2006) Prospective randomized study comparing amiodarone vs. amiodarone plus losartan vs. amiodarone plus perindopril for the prevention of atrial fibrillation recurrence in patients with lone paroxysmal atrial fibrillation. Eur Heart J doi:10.1093/eurheartj/ehl135. Published online on July 6.
- Altman DG. (1991) Randomization. BMJ 302:1481–1482.
[Free Full Text] - Altman DG and Bland JM. (1999) Treatment allocation in controlled trials: why randomize? BMJ 318:1209.
[Free Full Text] - Kirby A, Gebski V, Keech AC. (2002) Determining the sample size in a clinical trial. Med J Aust 177:256–257.[Web of Science][Medline]
- Schulz KF and Grimes DA. (2005) Sample size calculations in randomized trials: mandatory and mystical. Lancet 365:1348–1353.[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||