Quality-Adjusted Meta-analysis of the Hypertension/Coronary Dilemma

Abstract

Primary investigators of randomized drug trials in hypertension were invited to rate quality of such trials. The intention of the survey was to ask if antihypertensive drug therapy reduces incidence of coronary heart disease (CHD) in hypertensive patients. Response was obtained for 7 of the 11 invited investigators, covering 69% of patients and 75% of CHD cases. Principal component analysis was used to construct a quality score based on answers to 12 questions along visual analog scales. The score correlated well with the answer to a global question of overall quality given by the raters. No systematic tendency toward favoring one's own trial could be demonstrated, therefore, all raters have contributed to the rating. The trials with the highest rated quality to answer the research question were Systolic Hypertension in the Elderly Program (SHEP), Australian National Blood Pressure Study, Medical Research Council, Veterans Administration, and European Working Party of Hypertension in the Elderly. The large Heart Detection and Follow-up Program (HDFP) trial was rated at 11th place among the trials with a score of < 40% of the SHEP. The small trials performed in the 1960s were placed at the bottom of the ranking list. Because SHEP is the only trial without diastolic hypertension, results were given with and without SHEP results. When incorporating the quality score into a meta-analysis of CHD outcome, results were dependent on whether SHEP was included or not. For diastolic hypertension only, the effect of therapy was estimated to be about 8% for all higher quality studies, whereas inclusion of the lower quality HDFP changed it to 14%. When isolated systolic hypertension trial was pooled with the others, no major relation to quality rating was observed. A 14% CHD preventive efficacy was established when pooling the three top quality studies. This stayed unchanged until HDFP at rank 11 was included raising this estimate to 16%. Inclusion of the two latest published trials in the elderly, the Medical Research Council trial of treatment of hypertension in older adults and the Swedish Trial in Old Patients with hypertension, did not change this overall estimate of 16% (standard error = 3.8%). It is concluded that if all randomized drug trials in hypertension had the same treatment efficacy, the estimated CHD prevention would be in the range of 15%. Subgroup analyses revealed no relationship to age, but a difference in efficacy was shown depending on whether the trials were performed in the United States or elsewhere. Also, patients at higher risk levels showed better benefit than lower risk patients.