Critique of Medical Research Council Trial of Treatment of Hypertension in Older Adults, BMJ 304:405-412, 1992

Minor Problems

  1. Continuous variables were not used effectively as covariates.
  2. Descriptive statistics did not include quantiles for continuous variables.
  3. Long descriptions of numeric results (e.g., the withdrawals from randomized treatment) would be better presented in charts.
  4. When the authors did allow for varying follow-up time and take into account the actual times to events, they summarized the results using events per person year of follow-up, without mentioning that this kind of summary assumes an exponential (constant attrition) survival model to work efficiently.

Moderate Problems

  1. The statistical methods used in this paper are exceptionally poorly described.
  2. For covariate adjustment and assessment of interactions, the authors have used an inappropriate model, the binary logistic model. This model ignores the fact that follow-up time varies by subject and that events occurring early need to be distinguished from those occurring late.
  3. The study was single-blind; double-blind would have been better.

Severe Problems

  1. This trial was undertaken because an earlier trial did not include subjects older than 64 years. It was a mistake to exclude older subjects from earlier trials; the answer would have already been known, and nothing "magic" happens at age 65.
  2. This is a 3-treatment comparison (diuretic vs. beta-blocker vs. placebo). It is very informative to compare more than two treatments. In a very unusual move, the study designers wanted to have separate placebos to match both active drugs. Apparently they were concerned about pairs of subjects, one on beta-blocker and one on diuretic, comparing notes. The designers then decided to allocate fully 1/2 of all subjects to a placebo in a 3-treatment comparison. This can be inefficient, and raises ethical problems because of the next point.
  3. A placebo-controlled trial launched in 1982 may not have raised ethical eyebrows back then. But the fact that the trial had no sequential monitoring for possible early termination for efficacy or safety is difficult to understand. Statistical techniques for early termination were available in the mid 1970s and were used for the NHLBI Lipids Research Clinicals Program Primary Prevential Trial which started randomizing patients around 1973. The diuretic reduced the event rate for cardiovascular events by a factor of 0.69 over placebo. Having subjects on placebo who had systolic blood pressure between 160 and 209 mmHg for up to seven years is problematic. The diuretic effect would have been significant much earlier.

FrankHarrell

Topic revision: r2 - 21 Nov 2004, FrankHarrell
 

This site is powered by FoswikiCopyright © 2013-2017 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Vanderbilt Biostatistics Wiki? Send feedback