Baseline clinical factors accurately predicted 30-day mortality after myocardial infarction in patients considered for thrombolysis
ACP J Club. 1995 Sept-Oct;123:48. doi:10.7326/ACPJC-1995-123-2-048
Lee KL, Woodlief LH, Topol EJ, et al., for the GUSTO-I Investigators. Predictors of 30-day mortality in the era of reperfusion for acute myocardial infarction: results from an international trial of 41 021 patients. Circulation. 1995 Mar 15;91:1659-68.
To determine baseline risk factors that predicted 30-day mortality after myocardial infarction (MI) in patients considered for thrombolytic therapy.
Observational post hoc multivariate analysis of patients in the Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries (GUSTO-I) trial.
1081 hospitals in 15 countries.
41 021 patients with acute MI and ST-segment elevation. Patients with standard contraindications to thrombolysis were ineligible. Follow-up was > 99%.
Assessment of prognostic factors
Baseline data were collected: age; sex; race; Killip class; country; smoking status; family history of coronary disease; hypertension; diabetes; hypercholesterolemia; and preadmission events including previous MI, angina, bypass surgery, or angioplasty; location of infarct; or cerebrovascular disease. Thrombolytic strategy, as randomly assigned, was included in the analysis. No other interventions were included.
Main outcome measure
30-day all-cause mortality.
The strongest independent risk factors for 30-day all-cause mortality were older age, lower systolic blood pressure, higher Killip class, higher heart rate, and anterior location of infarction. These 5 characteristics contained most of the baseline prognostic information. The adjusted odds ratio (OR) for the oldest compared with the youngest quartile of patients was 3.88 (95% CI 3.52 to 4.28). For patients in Killip classes III and IV compared with those in class I, adjusted ORs were 4.37 (CI 3.34 to 5.71) and 7.86 (CI 5.88 to 10.49), respectively. Significant but modest effects were seen with previous MI, the interaction of age with Killip class, height, time to treatment, diabetes, weight, smoking, choice of thrombolytic strategy, previous bypass surgery, hypertension, and previous cerebrovascular disease. Sex was a borderline predictor (P = 0.043). Internal validation of the model showed excellent predictive results.
Risk for 30-day mortality was accurately predicted by the interplay of several baseline characteristics in patients with myocardial infarction.
Sources of funding: Bayer; CIBA-Corning; Genentech; ICI Pharmaceuticals; Sanofi Pharmaceuticals.
For article reprint: Dr. K.L. Lee, Biometry Division, Community and Family Medicine, Box 3363, Duke University Medical Center, Durham, NC 27710, USA. FAX 919-286-2947.
Why try to predict risk for death after MI? First, because an understanding of prognosis is almost always helpful to physicians, patients, and families. Second, some therapies may be preferentially directed to high-risk patients if concerns exist about side effects or costs (e.g., tissue plasminogen activator, the price of which is 10-fold higher than that of streptokinase). Third, risk prediction is valuable for researchers designing and interpreting clinical studies and for institutions seeking to benchmark clinical outcomes for quality improvement.
Any risk-prediction algorithm ideally should be accurate, widely applicable, and easy to use. Optimizing these features simultaneously is not easy. For example, an algorithm designed for wide application may either lose accuracy as it is applied to diverse subgroups of patients or require numerous variables that make it less user-friendly.
How does the algorithm devised by Lee and colleagues measure up? Their model shows outstanding predictive accuracy for a remarkable range of patients treated with thrombolytic drugs. Although the model was not tested in an independent population, rigorous methods were used for internal cross-validation. As to applicability, their analysis draws on patients enrolled in the GUSTO trial (1), with the inevitable selection bias that this entails. Short-term mortality in the GUSTO trial, for example, was about half that seen among patients hospitalized with MI outside of trials (2). The generalizability of the model must be established for patients presenting > 6 hours after symptom onset. The overall algorithm itself is too complex for bedside use. A simple scoring system or nomogram, however, is now being prepared by the same team (Lee KL. Personal communication).
Lee and colleagues found that sex was only a weak predictor of 30-day mortality in women (P = 0.043) when other clinical factors were taken into account. Intriguingly, they also showed that taller, heavier persons fared better after MI. Because sex obviously correlates to some extent with body size, inclusion of body size variables may mask a biological disadvantage incurred by women.
The GUSTO findings support the conclusions of Vaccarino and colleagues, who systematically reviewed the literature on sex differences in prognosis after MI. They found insufficient similarities in study designs and data to allow a standard meta-analysis. Therefore, their review uses vote-counting; studies are weighted equally regardless of sample size or the magnitude of sex differences found. Vaccarino and colleagues do, however, convincingly show that the adverse short-term prognosis for women after MI occurs largely because women are older, on average, when they have an MI. Our recent analysis of population-wide data for Ontario provides corroborating evidence. In 1991, the crude (vs age- and sex-adjusted) in-hospital death rates for men and women with MI were 14.4% (15.8%) and 21.9% (17.5%), respectively. The 34% relative difference decreased to 10% with age adjustment (2).
Vaccarino and colleagues also highlight the favorable long-term prognosis for women. Beyond 1 year from the index MI, women have a lower age-adjusted mortality than men because their usual survival advantage over same-aged men re-emerges.
Taken together, these studies offer reassurance that little of the sex difference in mortality after MI is attributable to sex differences in MI treatment that have been reported in several countries. More research is needed to understand better whether (and why) women have even a small short-term mortality disadvantage after MI, and to determine whether sex differences in treatment have adverse effects on quality of life for women.
C. David Naylor, MD
University of TorontoToronto, Ontario, Canada