I have just attended a small meeting on intervention trials in hypertension organized by the International and European Societies of Hypertension, the proceedings of which will be published soon in the Journal of Hypertension. I am delighted to see the continued and warm cooperation of our two journals. More specifically, I am continually impressed with how far we have come since our early Veterans Administration Cooperative Studies organized by Edward D. Freis over 30 years ago. During these years we have seen the established efficacy of antihypertensive agents in reducing arterial pressure and controlling outcomes of hypertensive cardiac and vascular diseases. We have seen a reversal of cardiovascular morbidity and mortality that is associated with marked reductions in the prevalence of hypertensive emergencies and accelerated and malignant hypertension and, of course, dramatic reductions in deaths from stroke and coronary heart disease (CHD).
As we have become well aware, we have been served well by the relatively recent developments in “post hoc analyses” of studies and trials. Nevertheless, cautions were raised at this symposium on the conduct of meta-analysis and the use of drug study databases. One presentation was concerned with the “pros and cons” of meta-analysis, suggesting that perhaps a better term would be “pros and cautions.” However, the subject of these analyses might be more appropriately termed “cautions and cons.” With the recent plethora of publications employing these and other sophisticated epidemiological and biostatistical techniques, another presentation emphasized the need for more rigorous and careful review, analysis, and inclusion of the extant, but peer-reviewed only, published literature. We must keep in mind that by publishing a meta-analysis, we do not necessarily confer truth to that analysis. Inclusion of studies in a meta-analysis requires a meticulous and highly discriminating literature search and review.
We emphasize again that such references must be restricted only to peer-reviewed publications, and editors of major hypertension journals must insist that the literature review not be restricted exclusively to a Medline or Medlars search. Although these indexes are most serviceable, the world literature is not restricted to the 3- to 6-year retrospective data storage that are instantly available on our desktop computers. The selection of a handful of key words and phrases provided by a thoughtful author by no means satisfies the necessity of a more-comprehensive literature review. Too often in submitted manuscripts the author believes that such a search of the extant published data and analysis is satisfied. Clearly, this is an inadequate aspect of the overall research work, and it ill serves the overall needs and purposes of a comprehensive manuscript, the author, and, most of all, the readers. We recently heard one critic of poorly done meta-analyses say that the relationship of “analysis to meta-analysis is analogous to physics to metaphysics.” We all must be aware of the adage of which we were reminded at the symposium: “If garbage is included in any analysis, for sure, garbage will come out.”
Conclusions of Study Findings
Most participants at this symposium, I believe, had been convinced from the analysis of the first 14 multicenter trials of antihypertensive therapy by Collins et al1 that the antihypertensive drugs that had reduced arterial pressure reduced deaths from stroke. Some had been less convinced (at the time) that the deaths from CHD were as dramatically reduced; certainly, they were less than predicted. Few who attended the symposium ever believed that lipid elevations that may have been produced by the diuretics accounted for this “less than predicted significant reduction in CHD.” Since subsequent meta-analysis of subsequent studies employing lower doses of diuretics satisfied the original predictions, a more reasonable conclusion seems that hypokalemia contributing to sudden cardiac death and, hence, deaths from CHD seems more reasonable.2 The obvious lesson from this standing issue is that whatever conclusions are drawn from those trials, readers and other scientists must be extremely cautious in making inferences and conclusions. The consequence of making erroneous conclusions may be extremely costly in falsely justifying new, additional, and confirmatory studies.
There is always a great need for us to be certain of the terminology we use. In the initial 14-study meta-analysis referred to above, the planning groups for each of these studies defined CHD to include deaths from myocardial infarction; angina pectoris without pathologically demonstrable myocardial infarction, lethal congestive heart failure, or pulmonary edema; lethal cardiac arrhythmias; or otherwise unexplainable sudden cardiac deaths. However, it does not appear to be certain that recently designed multicenter trials have made clear-cut definitions of deaths from CHD. Thus, many of the deaths from CHD in the early 14 multicenter trials may have been produced by myocardial infarction or occlusive epicardial coronary artery disease; however, they may also have resulted from hypertensive coronary arteriolar (microvascular) disease or even from antihypertensive treatment (eg, hypokalemia). It would be of immense value to have such outcomes defined to obviate post hoc problems. And, it was of great interest to note that most participants at this symposium were unable to define the current criteria of ongoing studies for CHD.
Another definition problem relates to just what is meant by hypertension. There are major differences in this definition among the many studies in the United States and Europe. In the United States, hypertension is defined and classified (ie, “staged”) according to the heights of systolic and diastolic pressures. Thus, a patient has hypertension if systolic and diastolic pressures exceed 140 and 90 mm Hg, respectively. In Europe, however, even though systolic hypertension is now considered treatable, the existence of hypertension is usually defined by diastolic pressure, with a minimal level of 95 mm Hg (based on the World Health Organization criteria). Thus, studies from the United States and elsewhere must be qualified accordingly because many of our current, ongoing studies involve patients with lower levels of diastolic pressure and with systolic hypertension. Again, when retrospective analyses of studies are pursued, investigators should consider these differences in definition.
It seems clear that we have learned and benefited much from the past intervention trials in hypertension. Although we no doubt will continue to learn much from ongoing and future trials, there is one lesson we may not have learned: We must analyze the data and define our terms, goals, and end points more precisely, or, for certain, new and avoidable controversies will arise.
Collins R, Peto R, MacMahon S, Hebert P, Fiebach NH, Eberlein KA, Godwin J, Qizilbash N, Taylor JO, Hennekens CH. Blood pressure, stroke, and coronary heart disease. Part 2, Short-term reductions in blood pressure: overview of randomised drug trials in their epidemiological context. Lancet. 1990;335:827-838.