American Heart Journal
Volume 146, Issue 5 , Pages 754-756, November 2003

New insight into clinical risk scores for patients with acute coronary syndromes

  • David A Morrow, MD, MPH

      Affiliations

    • TIMI Study Group and Cardiovascular Division, Department of Medicine, Brigham and Women's Hospital, Boston, Mass, USA
    • Corresponding Author InformationReprint requests: David A. Morrow, MD, MPH, Cardiovascular Division, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115, USA.

Article Outline

 

See related article on page 811.

In line with the pivotal role of risk stratification in the care of patients with non-ST–elevation acute coronary syndromes (NSTE ACS),1 clinical investigators have devoted substantial effort toward developing tools that enhance the clinician's quantitative assessment of risk. As a result, multiple clinical risk scores have been described and shown to deliver reasonable to very good prognostic performance.2, 3, 4, 5, 6 Although somewhat varied in their practicality, several of these clinical tools are expressed as simple integer scores that may be used conveniently at the bedside in caring for patients with NSTE ACS.5, 6 The evaluation of the Platelet Glycoprotein IIb/IIIa in Unstable Angina: Receptor Suppression Using Integrilin Therapy (PURSUIT) risk score by Brilakis et al7 in this issue of the Journal highlights critical issues to be considered by practitioners as they contemplate use of clinical risk scores in their practice, provides valuable insight into the pathophysiologic basis for the observed risk relationships, and points toward important clinical and research applications for this and other risk scores.

Back to Article Outline

Critical appraisal of clinical risk scores 

Weintraub recently reviewed the key elements involved in evaluating the overall performance of any prognostic model, and clinical risk scores in particular.8 In brief, the clinician should have an understanding of the prognostic discriminatory capacity of the risk score, usually expressed as a C-statistic, which corresponds to the area under a receiver operating characteristic curve.9 Equally important is the degree to which the event rates predicted using the risk score correspond to the observed rates within the population of interest (termed calibration). Both properties should be evaluated not only in the study in which the tool was developed but also, more importantly, in one or more external populations that serve to validate the model. As stressed by Brilakis, the characteristics of the validation set bear immediately on our ability to assess whether the performance of a risk model developed in a selected population, such as from a clinical trial, will be preserved in the general population encountered in routine practice.7

The validation of the PURSUIT risk score5 in a relatively unselected population with acute myocardial infarction treated at the Mayo Clinic is an important step toward supporting the performance of this score in a community-based population. Specifically, this study indicates that the discriminatory capacity of the PURSUIT risk score is maintained in a population reflective of day-to-day practice in Minnesota (C-statistic 0.78). The agreement of the predicted and observed event rates is less robust with an approximately 2-fold higher mortality in the validation cohort (scores 9–12 and 13–16) compared with the predictions from PURSUIT. This discrepancy points toward a limitation of the analysis. To optimally test the discrimination and calibration of the PURSUIT score, it should be modeled based on the actual predictions from PURSUIT (ie, in statistical terms, the same intercept and parameter estimates). Otherwise, the consequence is to develop a new set of predicted event rates for the same score. As these data for the simplified PURSUIT score were not available to the authors, they refit the score to their study population, and thus perhaps overstated the calibration. Nevertheless, with the data available to them, Brilakis et al have provided an informative assessment of the validity of generalizing use of the PURSUIT risk score.

Back to Article Outline

Pathobiologic insights into risk score performance—a window to therapy 

The report from Brilakis also advances our understanding of the pathophysiology underlying the prognostic performance of the PURSUIT score.7 Their finding that the risk score was strongly associated with the probability of underlying left ventricular systolic dysfunction and the severity of coronary artery disease takes the clinician one step closer to the mechanistic link between the risk score values and the probability of death. It is not surprising that a score with heavy emphasis on variables that reflect the hemodynamic consequences of the acute event (heart rate, blood pressure, rales on exam) correlates well with ejection fraction. Nevertheless, few studies have attempted to elucidate the links between specific clinical risk scores and the events that they predict.10 Such efforts are likely to be helpful in directing clinicians and researchers toward interventions that may modify the risk associated with higher risk scores. For example, as the authors note, the PURSUIT risk score appears to be particularly valuable for identifying patients with a high probability of impaired left ventricular function and/or 3-vessel or left main coronary artery disease who may be appropriate candidates for angiography and revascularization.

Other risk scores developed to assess the risk of recurrent ischemic events as well as mortality may derive from different underlying pathobiology. For example, the Thrombolysis in Myocardial Infarction (TIMI) risk score for unstable angina/NSTE myocardial infarction,6 which was optimized to predict death, myocardial infarction, and recurrent ischemia, is associated with the burden of coronary thrombus and impaired epicardial flow, as well as the angiographic extent of coronary artery disease.11 These findings may explain, at least in part, the particular benefit of enoxaparin (vs unfractionated heparin)6 and intravenous glycoprotein IIb/IIIa inhibition12 among patients presenting with NSTE ACS and higher TIMI risk scores. In contrast, other prediction tools that heavily weight medical comorbidities such as renal failure, malignancy, functional status, and evidence of multiorgan failure capture additional prognostic information (at the cost of greater complexity)13 and reflect contributors that are less likely to be modified by potent antithrombin and antiplatelet therapy or an early invasive management strategy.

Back to Article Outline

Clinical and research applications 

Ultimately the clinical value of such simple tools for risk stratification stems from the extent to which they are useful for directing treatment.14 Today's practitioner is faced with an expanding menu of therapeutic options, concurrent with an escalating need to manage the costs of health care delivery. Current guidelines stress the importance of risk assessment in guiding the appropriate selection of patients for whom invasive or aggressive medical interventions are likely to be beneficial.1 Clinical risk scores may be used to identify higher risk patients for whom potent antithrombotic agents and early invasive management are particularly advantageous.6, 12, 15 Effective risk assessment also guides triage to alternative levels of hospital care, assists in making decisions regarding transfer to tertiary facilities, and helps in the implementation of clinical pathways that direct patient care and utilization of resources. Few studies have directly evaluated the application of clinical risk scores in such “critical pathways.”16, 17 Future work dedicated to this purpose should aid in facilitating an evidence-based approach toward enhancing the care of patients with ACS.

In addition to direct employment for patient care, risk scores are useful for a variety of applications in clinical research. In designing clinical trials, simple risk scores may be used to guide enrollment of a population with a high event rate, and thus permit testing for an expected relative treatment effect with a smaller sample size for the trial. Such scores may also be used as a tool for analysis and interpretation of clinical research. For example, the TIMI risk score for ST-elevation myocardial infarction has been used for analyses stratified by risk at presentation, such as risk-adjusted assessment of resource utilization18 and regional variation in outcomes.19 In their report, Brilakis et al demonstrate the potential underutilization of coronary angiography in high-risk patients identified using the PURSUIT risk score.7 Clinical scores may also be applied as a mechanism for comparing the effects of a similar intervention across different clinical trials,20 or for risk-adjusted evaluation of outcomes across hospitals or providers.21 Selection of the appropriate risk model to employ may be based upon the end point of interest (eg, mortality alone vs death and recurrent ischemic events), the objective of the analysis, and the information available. A more complex score that captures information regarding other major comorbidities may be optimal for retrospective comparison of mortality rates across institutions.4, 21 In contrast, a more practical score is preferable as a basis for entry into a clinical trial or for evaluation of patients with ACS at the time of presentation.

Back to Article Outline

Conclusions 

Simple integer clinical risk scores can deliver strong prognostic performance in clinical trials and “real world” populations. The evidence base for integrating such scores into clinical decision making continues to grow. Future research exploring the mechanistic relationships between individual scores and outcome is likely to further enhance their integration into clinical care.

Back to Article Outline

References 

  1. Braunwald E, Antman EM, Beasley JW, et al.  ACC/AHA 2002 guideline update for the management of patients with unstable angina and non-ST-segment elevation myocardial infarction—summary article (a report of the ACC/AHA task force on practice guidelines (Committee on the Management of Patients With Unstable Angina)). J Am Coll Cardiol. 2002;40:1366–1374
  2. Braunwald E, Mark D, Jones R, et al. Unstable angina: diagnosis and management. Clinical practice guideline number 10. AHCPR publication no. 94-0602. Rockville (Md): Agency for Health Care Policy and Research and the NHLBI, Public Health Service, US Department of Health and Human Services; 1994:154
  3. Scirica BM, Cannon CP, McCabe CH, et al.  Prognosis in the thrombolysis in myocardial ischemia III registry according to the Braunwald unstable angina pectoris classification. Am J Cardiol. 2002;90:821–826
  4. Jacobs DR, Kroenke C, Crow R, et al.  PREDICT (A simple risk score for clinical severity and long-term prognosis after hospitalization for acute myocardial infarction or unstable angina: the Minnesota heart survey). Circulation. 1999;100:599–607
  5. Boersma E, Pieper KS, Steyerberg EW, et al.  Predictors of outcome in patients with acute coronary syndromes without persistent ST-segment elevation (results from an international trial of 9461 patients. The PURSUIT Investigators). Circulation. 2000;101:2557–2567
  6. Antman EM, Cohen M, Bernink PJ, et al.  The TIMI risk score for unstable angina/non-ST elevation MI (A method for prognostication and therapeutic decision making). JAMA. 2000;284:835–842
  7. Brilakis ES, Wright RS, Kopecky SL, et al. Association of the PURSUIT risk score with predischarge ejection fraction, angiographic severity of coronary artery disease, and mortality in a nonselected, community-based population with non-ST-elevation acute myocardial infarction. Am Heart J 2003;146:811–8
  8. Weintraub WS. Prediction scores after myocardial infarction (value, limitations, and future directions). Circulation. 2002;106:2292–2293
  9. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982;143:29–36
  10. Mathew V, Farkouh M, Grill DE, et al.  Clinical risk stratification correlates with the angiographic extent of coronary artery disease in unstable angina. J Am Coll Cardiol. 2001;37:2053–2058
  11. Morrow DA, Sabatine MS, Cannon CP, et al.  Higher TIMI Risk Score is associated with coronary thrombus and abnormal flow in non-ST elevation acute coronary syndromes (observations from PRISM-PLUS). [abstract] Circulation. 2001;104:II-648
  12. Morrow DA, Antman EM, Snapinn SM, et al.  An integrated clinical approach to predicting the benefit of tirofiban in non-ST eleva-tion acute coronary syndromes. Application of the TIMI Risk Score for UA/NSTEMI in PRISM-PLUS. Eur Heart J. 2002;23:223–229
  13. Singh M, Reeder GS, Jacobsen SJ, et al.  Scores for post-myocardial infarction risk stratification in the community. Circulation. 2002;106:2309–2314
  14. Cannon CP. Evidence-based risk stratification to target therapies in acute coronary syndromes. Circulation. 2002;106:1588–1591
  15. Cannon CP, Weintraub WS, Demopoulos LA, et al.  Comparison of early invasive and conservative strategies in patients with unstable coronary syndromes treated with the glycoprotein IIb/IIIa inhibitor tirofiban. N Engl J Med. 2001;344:1879–1887
  16. Cannon CP, O'Gara PT. Critical pathway for unstable angina. In:  Cannon CP,  O'Gara PT editor. Critical pathways in cardiology. New York: Lippincott Williams & Wilkins; 2001;p. 131–141
  17. Reilly BM, Evans AT, Schaider JJ, et al.  Impact of a clinical decision rule on hospital triage of patients with suspected acute cardiac ischemia in the emergency department. JAMA. 2002;288:342–350
  18. Morrow DA, Antman EM, Charlesworth A, et al.  TIMI Risk Score for ST-elevation myocardial infarction (a convenient, bedside, clinical score for risk assessment at presentation. An InTIME II substudy). Circulation. 2000;102:2031–2037
  19. Giugliano RP, Llevadot J, Wilcox RG, et al.  Geographic variation in patient and hospital characteristics, management, and clinical outcomes in ST-elevation myocardial infarction treated with fibrinolysis (results from InTIME-II). Eur Heart J. 2001;22:1702–1715
  20. Sabatine MS, Morrow DA, Giugliano RP, et al.  Implications of upstream GPIIB/IIIA inhibition and stenting in the invasive management of UA/NSTEMI (a comparison of TIMI IIIB and TACTICS-TIMI 18). [abstract] Circulation. 2001;104:II-549
  21. Krumholz HM, Chen J, Wang Y, et al.  Comparing AMI mortality among hospitals in patients 65 years of age and older (evaluating methods of risk adjustment). Circulation. 1999;99:2986–2992

PII: S0002-8703(03)00460-5

doi:10.1016/S0002-8703(03)00460-5

American Heart Journal
Volume 146, Issue 5 , Pages 754-756, November 2003

Article Tools