- Friday, 28 June 2013 10:29
ECLAT study has a control group, Yates correction would be completely inappropriate
Response by Dr Farsalinos to Stanton Glantz
Prof Glantz, a well-known prohibitionist has made a comment (uploaded on PLoS One) about the study evaluating the efficiency of e-cigarettes in smoking substitution. Dr Farsalinos has uploaded his answer, accessible here.
The study evaluated the efficiency of electronic cigarette use in a randomized, double blind clinical trial. Two nicotine concentrations were used, and a control group which received non-nicotine electronic cigarette was also included. The control group is not in reality a placebo group, because they received a product which mimics the act of smoking and provides visual and sensory stimulation; thus, it handles several aspects of the addiction to smoking, but does not provide nicotine. Including such a control group leads to underestimation of the true effects on the treatment group. Moreover, no psychological support was provided, which is a key issue and may explain much (if not all) of the efficacy of other smoking-cessation methods like NRTs .
It is extremely important that the authors used intention-to-treat analysis. The failure rate was significant, something that is common in studies of smoking cessation. The fact that all participants were not willing to quit further raises the expected lost-to-follow up rate, as is the use of an old and inefficient electronic cigarette device (the study started in 2009, technology has significantly evolved since then). Treating all these participants as failures in the final analysis was the correct thing to do, but is also another reason for underestimating the true effects of electronic cigarettes.
Yates correction for continuity  is a method to correct for continuity in chi-squared statistics. However, it tends to overcorrect the p-value. Moreover, it is chiefly used when at least one cell of the table has an expected count smaller than 5, and this is not the case in this study. It has also been suggested that this correction is unnecessary even when the sample size is low . In fact, it has been supported that Yates correction “produces a test that is so conservative as to be almost useless” , and several other studies have found such a correction inappropriate [5,6].
Considering the use of an old and inefficient electronic cigarette device, the inherent lack of placebo available for electronic cigarette studies and the inclusion of unmotivated smokers (also supported by their overall good health status, which would further decrease their incentive to quit), it is extremely difficult to discard the results of this study.
1. Moore D, Aveyard P, Connock M, Wang D, Fry-Smith A, Barton P. Effectiveness and safety of nicotine replacement therapy assisted reduction to stop smoking: systematic review and meta-analysis. BMJ. 2009;338:b1024.
2. Yates, F. Contingency table involving small numbers and the χ2test. Journal of the Royal Statistical Society 1934;1: 217–235.
3. Sokal RR, Rohlf FJ.Biometry: The Principles and Practice of Statistics in Biological Research.Oxford: W.H. Freeman 1981,ISBN 0-7167-1254-7.
4. Grizzle JE. Continuity correction in the χ2 test for 2 × 2 tables. The American Statistician 1967;21:28–32.
5. Maxwell EA. Analysis of contingency tables and further reasons for not using Yates correction in 2 × 2 tables. Canadian Journal of Statistics 1976; 4:277–290.
6. Conover WJ. Some reasons for not using the Yates continuity correction on 2×2 contingency tables. Journal of the American Statistical Association 1974;69:374–376.