Yuhong Yuan, M. Thabane, L. Thabane
Dec 20, 2006
To the Editor: The study of aminotransferase levels by Dr Watkins and colleagues found a high incidence of alanine aminotransferase (ALT) elevations in healthy participants who were randomized to receive medications containing acetaminophen (paracetamol); this was not found in participants randomized to placebo. We have a number of concerns with the methods used in this study. First, the statistical methods did not include a description of sample size determination or definition of the minimal clinically important difference that the study was designed to detect. Second, we are concerned about allocation concealment throughout the study. For example, blinding of patients may not have been effective, since the drugs were commercial products; 3 groups used all tablets but 2 groups used both tablets and caplets. There is no description of blinding of data analysts. The analysis was not performed according to intention to treat. All of these have the potential to introduce bias. Third, regarding results, level of significance was apparently not adjusted for multiple comparisons, which could inflate type I error. The group mean ALT values shown in Figure 2 in the article included data from participants who were taking or who had stopped medications, which does not reflect the ALT time curve after hepatic cell injury in acetaminophen users. The aspartate aminotransferase (AST) elevations followed a similar time course but were smaller (Figure 3), raising a question about the clinical significance. Repeating a test increases the chance of detecting a positive result, which could explain the 38% ALT elevation rate in the placebo group, which was tested around 14 times. Finally, consumption of alcohol by the participants was not considered. We believe that all of these issues suggest caution in interpreting the results of this study.