Evidence about preoperative risk assessment: Why aren't there better studies?
ACP J Club. 1993 Sept-Oct;119:A16. doi:10.7326/ACPJC-1993-119-2-A16
Many medical disorders increase the risk for adverse outcomes after surgery but, until recently, rigorous studies that have estimated the risks have been sparse, and controlled trials of interventions to minimize risk are still few. Why have good studies of risk assessment and management been so scarce, and what should we expect from future research?
In 1977, Goldman and colleagues set the stage with a large, well-done, prospective study of cardiac risk in noncardiac surgical patients (1). The result was the “Goldman Cardiac Risk Index” that many of us have taught and used. In 1986, a second major prospective study by Detsky and colleagues took the next important steps of validating the Goldman index, testing a modified cardiac risk index, and incorporating the concepts of pretest probability and likelihood ratios into risk assessment (2). Both indices enhanced the clinical prediction of cardiac risk but both lacked sensitivity: Approximately 40% of serious cardiac complications occurred in patients predicted to have relatively low risk. Although imperfect, these indices are the most objective risk measures available and are widely used in clinical practice for many types of noncardiac operations.
A subsequent editorial by Goldman pointed toward the future when it characterized the low rate of original research and publication in preoperative risk assessment compared with other areas of medicine (3). Now, many more investigators are carefully investigating preoperative risk assessment; many reports have recently been published (4-13).
To understand and use this growing literature effectively for evidence-based clinical decisions, it is important to appreciate the inherent methodologic difficulties of research in perioperative care. First, bad surgical outcomes are infrequent overall. This is good for patients but miserable for researchers because it mandates large, expensive, prospective studies or rigorous retrospective methods to achieve adequate statistical power to detect meaningful differences between groups. (One must be careful not to overvalue small studies only because they are prospective or undervalue well-done retrospective studies.)
Second, surgical selection bias is a problem. Generally only patients who actually have operations have been studied, a subset of all patients who might benefit from surgical procedures. We have little information with which to determine whether the decisions to not consider other patients for surgery were correct.
Third, there have been many problems with definitions of outcomes. One study showed that, depending on the criteria used to define a postoperative myocardial infarction, the incidence of myocardial infarction for a group of 232 patients could have varied from 1% to 11% (13). Most investigators are well informed about this methodologic issue for cardiac complications but, until recently, they have paid less attention to explicit criteria for pulmonary complications. Thus, studies have varied on which pulmonary complications were measured and how those complications were defined, with some not defined at all (14). A fourth issue for the discerning reader to consider is heterogeneity of the preoperative surgical case mix. To the extent that investigators aim for homogeneity and limit variability in types of surgery or patients, results will have greater internal validity but may be less generalizable (external validity). On the other hand, more clinically typical heterogeneity in selection of study participants and operations may limit the ability to predict risk for a particular patient having a specific operation. Thus, we must find better ways to wed the methodologic concerns of researchers with the practical needs of clinicians.
A fifth problem for evidence-based preoperative management is the difficulty of sharing decision making among three specialties: internal medicine, surgery, and anesthesiology. The problems of effective communication, dissemination, and continuing medical education are all compounded by interspecialty barriers. Surgeons and anesthesiologists may not easily accept the advice of internists about, for example, prophylaxis for deep-venous thrombosis because of concern about the increased risk for intraoperative or postoperative bleeding. In these situations, internists are responsible for the quality of their recommendations and the orders they write but may have them overruled, raising perplexing medico-legal issues.
Because of the problems for generating evidence for preoperative risk assessment, where do we stand now? For cardiac risk assessment before noncardiac surgery, we have more well-done research to guide us than for pulmonary operative risk. Studies of cardiac risk are now consistently prospective and large or, if small, carefully address pretest probabilities imposed by the prevalence of coronary artery disease and the nature of the surgical procedure. Nevertheless, the problems of variability in patient groups, selection bias, referral bias, unblinded outcome assessment, and low event rates continue to beleaguer us. As a result, the yields of expensive preoperative cardiac evaluations (for example, stress and thallium testing, cardiac catheterization) are not fully defined but appear to be most useful for patients with moderate to high probability of coronary artery disease (15-18). We also have considerable evidence on which to base preoperative cardiac evaluation for patients having vascular surgery for atherosclerotic vascular disease (8, 9, 19). Even so, in addition to the problems mentioned above, studies remain limited by narrow selection of patient groups and types of surgery evaluated (1-4, 15, 18, 20, 21).
Tests of interventions to prevent cardiac complications remain rudimentary. The benefits of coronary artery bypass procedures to reduce the risk for vascular surgery or other operations are not fully defined (8, 9, 19). Further, the yield of preoperative dipyridamole-thallium stress testing and perioperative Holter monitoring to identify patients who may benefit from prophylactic interventions is unclear (1, 8, 22). We also lack randomized trials of interventions (for example, perioperative blockers or nitrates) to prevent cardiac complications in high-risk patients.
We know that routine preoperative chest roentgenograms have a low yield for pulmonary risk and that spirometry is overused with resultant waste of health care resources (6, 14, 23-25). We have solid evidence from clinical trials that perioperative incentive spirometry can prevent postoperative complications, but we are not so far along as in cardiac risk assessment with large, well-done prospective studies to clarify useful evaluative pathways for assessing pulmonary risk for individual patients having nonthoracic operations (14). Nevertheless, research methods in this area are improving. Recent studies have carefully used explicit criteria for pulmonary complications, systematic standardized preoperative evaluation procedures and postoperative surveillance for complications, and assessment of postoperative complications that was blinded to preoperative status and intraoperative course (6, 7).
Future research should target several other areas. Because most studies have focused on either cardiac or pulmonary complications, we have few data on concurrent rates for both types of complications with current surgical and supportive care (5, 26). In some groups, pulmonary complications occur more frequently (27). Studies of economic evaluations of preoperative testing strategies are increasing and more are needed (24, 25, 28, 29). Also, increasing numbers of elderly patients with several chronic diseases are facing surgery. We need research to develop interventions to enhance postoperative functional recovery in older patients.
In summary, much has been accomplished in research about risk assessment for noncardiac surgery in the past 15 years, but there is much room to grow. Clinicians invested in evidence-based medicine will have the opportunity to see this area mature and to incorporate an evolving scientific discipline into their practice. Internists, anesthesiologists, and surgeons will have a better basis to optimize the care of surgical candidates who have medical problems that increase the risk for complications from surgery.
Valerie A. Lawrence, MD
13. Charlson ME, MacKenzie CR, Ales KL, et al. The post-operative electrocardiogram and creatine kinase: implications for diagnosis of myocardial infarction after non-cardiac surgery. J Clin Epidemiol. 1989;42:25-34.
15. Mangano DT, London MJ, Tubau JF, and the Study of Perioperative Ischemia Research Group. Dipyridamole thallium-201 scintigraphy as a preoperative screening test. A reexamination of its predictive potential. Circulation. 1991;84:493-502.
16. Coley CM, Field TS, Abraham SA, Boucher CA, Eagle KA. Usefulness of dipyridamole-thallium scanning for preoperative evaluation of cardiac risk for nonvascular surgery. Am J Cardiol. 1992;69:1280-5.
17. Malenka DJ. Commentary on “Preoperative cardiac risk assessment with thallium scanning helped identify patients at low risk for cardiac complications - 1992.” ACP J Club. 1992;117:52.
20. Mangano DT, Browner WS, Hollenberg M, et al. Association of perioperative myocardial ischemia with cardiac morbidity and mortality in men undergoing noncardiac surgery. N Engl J Med. 1990;323:1781-8.
22. Paul SD, Kondo NI, Eagle KA. Commentary on “Left ventricular hypertrophy, history of hypertension, diabetes mellitus, coronary artery disease, and digoxin use predicted postoperative myocardial ischemia - 1992.” ACP J Club. 1992;117:88-9.
26. Gerson MF, Hurst JM, Hertzberg VS, et al. Prediction of cardiac and pulmonary complications related to elective abdominal and noncardiac thoracic surgery in geriatric patients. Am J Med. 1990;88:101-7.