Curriculum in Cardiology Clinical trials in acute myocardial infarction: Should we adjust for baseline characteristics? Ewout W. Steyerberg, PhIha Patrick M.M. Bossuyt, PhD,b and Kerry L. L.ee, PhIF Rot&r&m andAn#ste&znr, The Netherlands, and Durham, NC Background CI’ mica1 trials concerning acute myocardial infarction often evaluate short-term death. Several baseline characteristics are predictors of death, most notably age. Adjustment for one or more predictors in a multivariable analysis may be considered to correct the estimate of the treatment effect for any imbalance that by chance may have occurred between the randomized groups. Moreover, adjustment results in a stratified estimate of the effect of treatment. Methods and Results The effects of adjustment (correction for imbalance and stratification) were studied with logis tic regression analysis in the Global Use of Strategies to Open Occluded Coronary Arteries (GUSTO)-1 trial. The primary end point was 30day death, which occurred in 6.3% of 10,348 patients randomly assigned to tissue plasminogen activator and 7.3% of 20,162 patients randomly assigned to streptokinase thrombolytic therapy. This is equivalent to an unadjusted odds ratio of 0.853. No significant imbalance had occurred for any of 17 baseline characteristics considered, including well-known demographic, presenting, and history characteristics. Adjusted for age, the odds ratio was 0.829, which is an 18% increase in estimated effect on the logistic scale. When adjusted for 17 characteristics, the odds ratio was 0.820, an increase of 25%. The increase in effect estimate was largely explained by the stratification effect and only partly by imbalance of predictors. Conclusions Adjustment for predictive baseline characteristics, even when largely balanced, may lead to clearly different estimates of the treatment effect on mortality rates. Adjustment for important predictors such as age is recommended in clinical trials studying patients with acute myocardial infarction. (Am Heart J 2000;139:745-5 1.) See related Editorial on page 761. Randomized clinical trials provide the most reliable evidence of effectiveness of treatments in acute myocar- dial infarction (MI). Trials have established the beneficial effects and relative safety of several thrombolytic agents’ (streptokinase,2,3 tissue plasminogen activator [TPA]*) and adjunctive medical therapy5 (padrenergic antago nists,” angiotensinconverting enzyme inhibitors).7-9 Through similar studies, other adjunctive therapies have been shown to be of no more than limited value (addi- tion of heparin to aspirin, nitrates, calcium-channel blockers, some antiarrhythmic drugs, magnesium).5~8 From othe Center for Clinical Decision Sciences, Deportment of Public Health, Erasmus University, Rotterdom; bthe Deportment of Clinical Epidemiology and Biostotistics, AC(F demic Medical Center, University of Amsterdam; ond <the Deportment of Community ond Family Medicine, Duke University Medical Center, Durham. Supported in port by o grant horn the Netherlands Organization for Scientilic Research (NWO, 59.5.156). Guest Editor for this mowscript was David 1. &Mets, PhD, University of Wisconsin Medical School, Madison. Submitted December 28, 1998; accepted September 20, 1999. Reprint requests: &out W. Steyerberg, PhD, Center for Clinical Decision Sciences, Ee209 I, Deportment of Public Health, Erasmus University, PO Box 1738, 3000 DR Rotterdam, The Netherlands. E-mail: steyerberg~mgz.lgg.eur.nl Copyright 0 2000 by Mosby, Inc. 0002.8703/2000/$12.00+0 4/I/104240 dor: IO. I067/mhi.2000. I04240 Short-term death (within 28 to 42 days from randomiza- tion) is the main end point in most larger trials.*~~sl* Sometimes a composite measure is used, for example, consisting of the combination of death and severe left ventricular dysfunction,7J3 or death and reinfarctiont4 The main value of randomization is that treatment groups are on average comparable in terms of known and unknown patient characteristics. Selection on indi- cation is excluded, a feature that cannot be guaranteed in nonrandomized comparative studies. However, base- line differences in prognosis between treatment groups may well occur, even in relatively large trials. Such differ- ences between treatment groups are from pure chance when treatment allocation was truly random. In 5% of the cases, baseline characteristics will show “significant” imbalance between randomized groups (P < .05>. P val- ues of statistical tests of imbalance cannot be interpreted as indicating whether randomization workedts-17 and may only serve a descriptive purpose. Furthermore, absence of significant differences does not guarantee comparability of prognosis because the impact of imbalance also depends on the prognostic strength of the characteristics considered; a small difference in a very powerful predictor can make treatment groups clearly different in prognosis. Hence the similarity between randomized groups should be judged by an appraisal of the prognostic strength of the baseline characteristics and the magnitude of any imbalance.18