Author + information
- Mark P. Tschanz, DO, MACM∗ (, )
- Elexis C. McBee, DO, MPH,
- Travis E. Harrell, MD and
- Bryan M. Spalding, MD
- ↵∗Department of Internal Medicine, Naval Medical Center San Diego, 34800 Bob Wilson Drive, San Diego, California 92134
We appreciate the thoughtful study design by Pareek et al. (1) in their recent study of low-dose thiazide and thiazide-like diuretic agents. We agree with the authors that conclusive evidence in the form of an outcome study will be critical to finally put to rest the questions concerning the efficacy of these medications, despite the many publications over the past several years that have supported the use of chlorthalidone over hydrochlorothiazide.
Unfortunately, we are concerned that the significant enrollment difficulties and protocol violations challenge the internal validity of this study and may have led to an overstatement of the study findings. While designing the study to utilize an intention-to-treat analysis (ITT) is noteworthy, the enrollment of only 25% of the patients estimated for a power of 80% and the study completion rate (barely above 50%) limit the utility of this type of analysis. An argument could be made that the ITT analysis is more likely to make a type 2 error (find no difference between treatment groups); however, this assumes strict protocol adherence for the majority of participants. As described by Hernán and Hernández-Díaz (2), adherence problems in a randomized controlled trial often lead to untestable assumptions about treatment groups. While a “per protocol” or “as treated” analysis could be considered for this study, careful adjustment for confounders would be essential. The data required for this adjustment are often unavailable, because subjects may be lost to follow-up. As the number of participants completing a trial on protocol decreases, the trial design becomes more observational and less of a randomized controlled trial, making adjustment for variables critical (3).
While the authors note the limited number of participants as a limitation to the study, we would suggest a greater limitation is the number of protocol violations, which limits the utility of the ITT analysis.
Please note: The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Pareek A.K.,
- Messerli F.H.,
- Chandurkar N.B.,
- et al.
- Hernán M.A.,
- Hernández-Díaz S.