|
|
A Validated Prediction Model for all Forms of Acute Coronary Syndrome: Estimating the Risk of 6-Month Post Discharge Death in an International RegistryEagle KA, Lim MJ, Dabbous OH, et al. J Am Med Assoc 2004; 291:2727-33Reviewer: KW Tim Park, MD
Background: The acute coronary syndrome (ACS) is comprised of conditions ranging from unstable angina to both non-ST-segment elevation and ST-segment elevation myocardial infarction (NSTEMI and STEMI). In order to help with clinical decision making, prediction models on outcome after ACS have been developed. Previous risk-prediction models for outcome after ACS were developed from large randomized clinical trials and focused on in-hospital mortality, and their generalizability to risk prediction in the average clinical setting has been questioned.1-6 The current study sought to develop a prediction model of all-cause 6-month mortality after all types of ACS in patients similar to those encountered in routine clinical practice. Methods: The Global Registry of Acute Coronary Events (GRACE) is a multinational registry of patients admitted with ACS at 94 hospitals in 14 countries. To be entered in the registry, the patient had to be ≥18-years-old, admitted with a presumptive diagnosis of ACS, and had one of the following: ECG changes consistent with ACS, serial increases in serum cardiac markers, and/or documentation of coronary artery disease (CAD). The qualifying ACS must not have been precipitated by significant noncardiovascular comorbidity such as acute anemia or hyperthyroidism. At discharge, all cases were categorized as unstable angina, NSTEMI, or STEMI. At ~ 6 months post discharge, patients were followed up to ascertain their vital status. The primary endpoint was all-cause mortality within 6 months of discharge. Using patients' baseline demographics, comorbidities, symptoms and signs at presentation, in-hospital treatments and procedures, a prediction model was developed in all patients enrolled in GRACE between April 1, 1999 and March 31, 2002 (n=17,142), using stepwise Cox proportional hazards regression. Then, the model was tested in a validation cohort of consecutive patients enrolled in GRACE between April 1, 2002 and December 31, 2003 (n=7,638). Results: The overall follow-up rate in the development cohort was 87.5% (n=15,007) for death. The development cohort of patients was 65 ñ 13 in age and 67% men. Fifty-eight percent had hypertension and prior or current smoking history, 24% had diabetes mellitus, 32% had a history of MI, 46% had hyperlipidemia, and 10% had a history of congestive heart failure (CHF). Forty-three percent were on aspirin, 25% on angiotensin converting enzyme inhibitors, 31% on á-blockers, 25% on oral nitrates, 23% on statins, and 21% on calcium channel blockers. By stepwise Cox proportional hazards regression, nine predictors for the 6-month mortality were identified: age per 10-year increase above 40, history of MI, history of CHF, increase in heart rate, decrease in systolic pressure, elevated serum creatinine, elevated cardiac enzymes, ST-segment depression, and no in-hospital percutaneous coronary intervention (PCI). Based on the model's variable coefficients, these prediction variables were given weighted scores. When the model was then applied to the validation cohort, the model performed well in all forms of ACS with a c statistic of at least 0.70. Discussion and Comments: As the authors of the study point out, previous prediction models developed from large clinical trials are robust for the specific endpoints in the population in which they were developed. Many ACS trials, however, used combined endpoints that included need for revascularization and this variable is influenced so much by local practice style and availability of a cardiac catheterization laboratory. This has limited the generalizability of the previous models. Another limitation of previous models has been the arbitrary distinction between NSTEMI and STEMI, even though it has been shown that in-hospital mortality is similar whether the ST segment deviation is elevation or depression.7 The current model developed from GRACE utilized an unambiguous end-point of all-cause 6-month mortality and is equally robust for all forms of ACS, including unstable angina, NSTEMI, and STEMI. Furthermore, patients enrolled in the registry are consecutive patients admitted with ACS and are population-based, rather than study protocol-based. Therefore, this model may be more generalizable than previous models. While the model is quite generalizable, the model does not account for all possible predictors - e.g., the authors did not consider socioeconomic factors. Another notable factor about the current model is that it is applicable to patients who are discharged alive after being admitted with ACS. It may thus be applicable to patients presenting for major noncardiac surgery after a prior admission with ACS. Current ACC/AHA guidelines on preoperative cardiac evaluation consider all forms of ACS within a month of the surgery as a major clinical predictor, whereas history of stable angina or MI > 1 month is considered an intermediate clinical predictor.8 The current study suggests that as far as mortality is concerned, one may need to make distinctions among different forms of ACS with whose history a patient may be coming to a major noncardiac surgery. For example, a 55-year-old without prior history of MI or CHF, presents with unstable angina, but is well managed with systolic pressure > 140 and resting heart rate < 70 bpm. His serum creatinine is 1.0 and he has no cardiac enzyme elevations. His symptoms are resolved with PCI. He would get a score of 54 on this model, with an associated 6-month predicted mortality of < 1%. On the other hand, a 75-year-old with prior history of CHF rules in for an MI with ST depression and enzyme elevation. His systolic pressure drops to 90's and his resting heart rate was in 100's before being controlled with β-blockade. He also has a serum creatinine of 2 mg/dl. His score on this model would be 186, with an associated 6-month predicted mortality of about 30%. The implications for perioperative mortality and cardiac morbidity for these two hypothetical patients, though both have ACS and a major clinical predictor, cannot be the same. This study calls for a study to develop a perioperative post-ACS prediction model that takes into account different forms of ACS. References:
Table of Contents:
© Society of Cardiovascular Anesthesiologists Questions or comments? Please send email to webmaster@scahq.org |