Meta-analysis of Hypertension Trials in Diabetic Patien...

Meta-analysis of Hypertension Trials in Diabetic Patients

Meta-analysis of Hypertension Trials in Diabetic Patients
2001
Marco Pahor, MD, Bruce M. Psaty, MD, PHD, Michael H. Alderman, MD and Curt D. Furberg, MD, PHD
Diabetes Care
© 2001 by the American Diabetes Association, Inc.

From the Department of Internal Medicine, Sticht Center on Aging (M.P.), and the Department of Public Health Sciences (C.D.F.), Wake Forest University, Winston-Salem, North Carolina; the Departments of Medicine, Epidemiology, and Health Services (B.M.P.), Cardiovascular Health Research Unit, University of Washington, Seattle, Washington; and the Department of Epidemiology and Social Medicine (M.H.A.), Albert Einstein College of Medicine, Bronx, New York.

Address correspondence to Marco Pahor, MD, Department of Internal Medicine, Sticht Center on Aging, Wake Forest University Baptist Medical Center, Medical Center Boulevard, Winston-Salem, NC 27157. E-mail: mpahor@wfubmc.edu .

Methodological and statistical reasons may explain why the U.K. Prospective Diabetes Study (UKPDS) was considered an outlier when compared with the other reviewed studies. Study design, report, and conduct weaknesses limit the conclusions of the UKPDS (1). Several of these weaknesses have been discussed in published articles (2,3). These weaknesses are as follows: 1) the UKPDS hypertension study was not blind, but open label; 2) the dosing regimens for captopril may not have been adequate (captopril was given twice per day; for the treatment of hypertension captopril should be given three times per day [4]); 3) the trial was not powered to detect differences in drug effects on cardiovascular events; 4) none of the blood pressure goals in the tight blood pressure control group (150/85 mmHg) or in the less tight blood pressure control group (initially 200/105 mmHg and recently 180/105 mmHg) are currently recommended according to U.S. or U.K. guidelines for the treatment of hypertension in patients with or without diabetes (5,6); 5) in the less tight blood pressure control group the first-line agent was furosemide, a diuretic not recommended for treatment of hypertension, according to the current guidelines (5,6); and 6) patients randomized to atenolol were significantly more likely to dropout of the trial than those randomized to captopril (35 vs. 22%, P < 0.0001), which may have created biased results.

These weaknesses may have obscured potentially important differences between captopril and atenolol and limited the possible generalizations of the potential benefits of more intense blood pressure control. The current recommended blood pressure target of <130/85 mmHg in patients with diabetes relies solely on observational data (5,7), not on definitive evidence from randomized controlled trials. It is not known whether a blood pressure target of <130/85 mmHg can be achieved safely in the majority of diabetic patients with hypertension. It remains to be established whether the more favorable outcomes in the tight blood pressure control group of the UKPDS are explained by better blood pressure control, the type of first-line drug being used (furosemide versus captopril or atenolol), chance, or any combination of these factors.

As described in previous reports, the Appropriate Blood Pressure Control in Diabetes (ABCD) study was conducted double-blind (8), whereas the Captopril Prevention Project (CAPPP) (9), the Fosinopril Versus Amlodipine Cardiovascular Events Randomized Trial (FACET) (10), and the UKPDS (1) were all conducted open label. None of the individual trials were adequately powered to assess effects on the risk of major cardiovascular events in diabetic patients with hypertension. The updated number of events in the nisoldipine and enalapril groups of ABCD were 27 and 9 for acute myocardial infarction (MI), 11 and 7 for stroke, 8 and 10 for heart failure, and 18 and 14 for all-cause mortality (M.P., personal communication with Dr. R. Estacio). We have estimated the number of combined cardiovascular events (47 in the nisoldipine and 29 in the enalapril group) by adding the number of new MI events and heart failure events to the previously reported number of combined events. In an analyses of the four trials combined with these new data, the odds ratio for ACE inhibitors versus other treatments (95% CI) was 0.72 (0.54-0.96) for acute MI, 0.87 (0.59-1.27) for stroke, 0.81 (0.64-1.01) for cardiovascular events, and 0.90 (0.68-1.20) for all-cause mortality. The tests for heterogeneity were significant for the outcomes of acute MI and cardiovascular events when the data from the UKPDS were combined with the data from the other three trials (P < 0.001 for both outcomes), but the outcomes were not significant when the UKPDS was excluded from the meta-analysis, suggesting that the results of the UKPDS were different from the results of the other three trials. It cannot be assessed whether such heterogeneity was due to the many weaknesses of the UKPDS, the therapeutic equivalence of captopril and atenolol, or chance. Whether atenolol is equivalent to captopril for the treatment of diabetic patients with hypertension remains a question.

Furthermore, it was appropriate not to combine the UKPDS data with the data of the other trials because of the substantial heterogeneity generated by the inclusion of the UKPDS. Thus, only the data from the ABCD study, the CAPPP, and the FACET were included in the final meta-analytic calculations. When the data from the ABCD study, the CAPPP, and the FACET were combined, the patients randomized to an ACE inhibitor had a significantly lower risk of acute MI, cardiovascular events, and all-cause mortality than those randomized to an alternative treatment (Table 1). There were no such differences for the outcome of stroke. The revised data from the ABCD study did not change the conclusions of our meta-analysis (11).

This article may be viewed in its entirety by subscription or by purchase at the website.