Glucose and cholesterol elevations during thiazide therapy: intention-to-treat versus actual on-therapy experience

This nonrandomized, parallel-group, prospective clinical series in consecutive patients was conducted to determine if the method of analysis of biochemical data collected in long-term treatment trials can explain the controversy regarding the effects of thiazides on plasma glucose or cholesterol.

Fasting serum glucose, potassium, and cholesterol levels were determined before therapy, and yearly thereafter (4.0 +/- 0.1 years) in 634 consecutive patients treated for more than 1 year with antihypertensive monotherapy.

When the data were analyzed by the intention-to-treat paradigm (ignoring changes in therapy), glucose and cholesterol levels in the thiazide group were significantly higher (compared to baseline or to the other therapy group) only at year 1. When the data were analyzed by actual on-therapy experience (censoring patients when therapy is altered)--a more traditional way to examine side effects--these differences were significant for all 5 years studied. Even if as few as 8% of patients who changed therapies were included in an intention-to-treat analysis, these differences became nonsignificant.

These comparisons not only indicate why a dichotomy might exist between clinical practice (analyzed by actual on-therapy experience) and randomized trials (analyzed by intention-to-treat methods), regarding biochemical changes in glucose and cholesterol during thiazide therapy, but also that, for secondary hypotheses, analyzing data by both methods of assignment is important.