Assessment of Cardiac Risk and the Cardiology Consultation

Reperfusion of an Ischemic Myocardium

Surgical interventions requiring interruption of blood flow to the heart must, out of necessity, be followed by restoration of perfusion. Numerous experimental studies have provided compelling evidence that reperfusion, although essential for tissue or organ survival, or both, is not without risk because of the extension of cell damage as a result of reperfusion itself. Myocardial ischemia of limited duration (< 20 minutes), followed by reperfusion, are accompanied by functional recovery without evidence of structural injury or biochemical evidence of tissue injury.^8,9

Paradoxically, reperfusion of cardiac tissue, which has been subjected to an extended period of ischemia, results in a phenomenon known as myocardial reperfusion injury.^10–12 Thus, a paradox exists in that tissue viability can be maintained only if reperfusion is instituted within a reasonable time period, but only at the risk for extending the injury beyond that caused by the ischemic insult itself. This is supported by the observation that ventricular fibrillation was prominent when the regionally ischemic canine heart was subjected to reperfusion.¹³ Jennings et al¹⁴ reported adverse structural and electrophysiologic changes associated with reperfusion of the ischemic canine heart, and Hearse¹⁵ introduced the concept of an oxygen paradox in noting cardiac muscle enzyme release and alterations in ultrastructure when isolated hearts were reoxygenated after a period of hypoxic perfusion.

Myocardial reperfusion injury is defined as the death of myocytes, alive at the time of reperfusion, as a direct result of one or more events initiated by reperfusion. Myocardial cell damage results from the restoration of blood flow to the previously ischemic heart, thereby extending the region of irreversible injury beyond that caused by the ischemic insult alone. The cellular damage that results from reperfusion can be reversible or irreversible, depending on the length of the ischemic insult. If reperfusion is initiated within 20 minutes after the onset of ischemia, the resulting myocardial injury is reversible and is characterized functionally by depressed myocardial contractility, which eventually recovers completely. Myocardial tissue necrosis is not detectable in the previously ischemic region, although functional impairment of contractility may persist for a variable period, a phenomenon known as myocardial stunning. Initiating reperfusion after a duration of ischemia of longer than 20 minutes, however, results in irreversible myocardial injury or cellular necrosis. The extent of tissue necrosis that develops during reperfusion is directly related to the duration of the ischemic event. Tissue necrosis originates in the subendocardial regions of the ischemic myocardium and extends to the subepicardial regions of the area at risk, often referred to as the wavefront phenomenon. The cell death that occurs during reperfusion can be characterized microscopically by explosive swelling, which includes disruption of the tissue lattice, contraction bands, mitochondrial swelling, and calcium phosphate deposits within mitochondria.¹³

The magnitude of reperfusion injury is directly related to the magnitude of the ischemic injury that precedes it. In its most severe form, it manifests in a “no-reflow” phenomenon. In cardiac surgery, prevention of myocardial injury after the release of the AXC, including the prevention of no reflow, is directly dependent on the adequacy of myocardial protection during the period of aortic clamping. The combination of ischemic and reperfusion injury is probably the most frequent and serious type of injury that leads to poor outcomes in cardiac surgery today (see Chapters 2, 3, 6, 12 to 14, 18, and 28).

Basic science investigations (in mouse, human, and porcine hearts) have implicated acidosis as a primary trigger of apoptosis. Acidosis, reoxygenation, and reperfusion, but not hypoxia (or ischemia) alone, are strong stimuli for programmed cell death, as well as the demonstration that cardiac apoptosis can lead to heart failure.^16,17 This suggests that apoptotic changes might be triggered in the course of a cardiac operation, thus effecting an injurious cascade of adverse clinical events that manifest late in the postoperative course.

Based on the previous discussion, it is clear that a significant portion of perioperative cardiac morbidity is related primarily to intraoperative factors. However, preoperative risk factors may influence ischemia/reperfusion injury.

Adverse Systemic Effects of Cardiopulmonary Bypass

In addition to the effects of disruption and restoration of myocardial blood flow, cardiac morbidity may result from many of the components used to perform cardiovascular operations, which lead to systemic insults that result from CPB circuit-induced contact activation. Inflammation in cardiac surgical patients is produced by complex humoral and cellular interactions, including activation, generation, or expression of thrombin, complement, cytokines, neutrophils, adhesion molecules, mast cells, and multiple inflammatory mediators.¹⁸ Because of the redundancy of the inflammatory cascades, profound amplification occurs to produce multiorgan system dysfunction that can manifest as coagulopathy, respiratory failure, myocardial dysfunction, renal insufficiency, and neurocognitive defects. Coagulation and inflammation also are linked closely through networks of both humoral and cellular components, including proteases of the clotting and fibrinolytic cascades, as well as tissue factor. Vascular endothelial cells mediate inflammation and the cross-talk between coagulation and inflammation. Surgery alone activates specific hemostatic responses, activation of immune mechanisms, and inflammatory responses mediated by the release of various cytokines and chemokines (see Chapters 8 and 28 to 31). This complex inflammatory reaction can lead to death from nonischemic causes and suggests that preoperative risk factors may not predict morbidity. The ability to risk-adjust populations is critical to study interventions that may influence these responses to CPB.

Assessment of Cardiac Function

Cardiac contractile dysfunction is the most prominent feature of myocardial injury, despite the fact that there are virtually no perfect measures of postoperative cardiac function.

The need for inotropic support, thermodilution cardiac output (CO) measurements, and transesophageal echocardiography (TEE) may represent practical intraoperative options for cardiac contractility evaluation. The need for inotropic support and CO measurements are not reliable measures because they depend on loading conditions and practitioner variability. Failure to wean from CPB, in the absence of systemic factors such as hyperkalemia and acidosis, is the best evidence of intraoperative myocardial injury or cardiac dysfunction; but it also may be multifactorial and, therefore, a less robust outcome measure.

RWMAs follow the onset of ischemia in 10 to 15 seconds. Echocardiography can, therefore, be a sensitive and rapid monitor for cardiac ischemia/injury.²¹ If the RWMA is irreversible, this indicates irreversible myocardial necrosis (see Chapters 11 through 14). The importance of TEE assessment of cardiac function is further enhanced by its value as a predictor of long-term survival.²² In patients undergoing CABG, a postoperative decrease in left ventricular ejection fraction (LVEF) compared with preoperative baseline predicts decreased long-term survival.²³

The use of TEE is complicated because myocardial stunning (postischemic transient ventricular dysfunction) is a common cause of new postoperative RWMAs, which are transient. However, the appearance of a new ventricular RWMA in the postoperative period, whether caused by irreversible AMI or by reversible myocardial stunning, is an indication of some form of inadequate myocardial protection during the intraoperative period and, therefore, of interest for the assessment of new interventions. Echocardiographic and Doppler systems also have the limitation of being sensitive to alterations in loading conditions, similar to the need for inotropic support and CO determinations.²⁴ The interpretation of TEE images is also operator dependent.²⁵ In addition, there are nonischemic causes of RWMAs, such as conduction abnormalities, ventricular pacing, and myocarditis, which confound the use of this outcome measure for the assessment of ischemic morbidity.

Electrocardiography Monitoring

The presence of new persistent Q waves of at least 0.03-second duration, broadening of preexisting Q waves, or new QS deflections on the postoperative ECG have been considered evidence of perioperative AMI.²⁶ However, new Q waves also may be caused by unmasking of an old MI and therefore not indicative of a new AMI. Crescenzi et al²⁷ demonstrated that the association of a new Q wave and high levels of biomarkers was strongly associated with postoperative cardiac events, whereas the isolated appearance of a new Q wave had no impact on the postoperative cardiac outcome. In addition, new Q waves may actually disappear over time.²⁸ Signs of non–Q-wave MI, such as ST-T wave changes, are even less reliable signs of AMI after cardiac surgery in the absence of biochemical evidence. ST-segment changes are even less specific for perioperative MI because they can be caused by changes in body position, hypothermia, transient conduction abnormalities, and electrolyte imbalances (see Chapter 15).

Serum Biochemical Markers to Detect Myocardial Injury

Serum biomarkers have become the primary means of assessing the presence and extent of AMI after cardiac surgery. Serum biomarkers that are indicative of myocardial damage include the following (with post-insult peak time given in parentheses): myoglobin (4 hours), total CK (16 hours), CK-MB isoenzyme (24 hours), troponins I and T (24 hours), and lactate dehydrogenase (LDH) (76 hours). The CK-MB isoenzyme has been used most widely, but studies have suggested that troponin I is the most sensitive and specific in depicting myocardial ischemia and infarction.^29–34

With respect to CK-MB, the definition of an optimal cutoff has been defined best by the correlation of multiples of the upper limit of normal (ULN) for the laboratory and medium- and long-term outcomes. For example, Klatte et al³⁵ reported on the implications of CK-MB in 2918 high-risk CABG patients enrolled in a clinical trial of an anti-ischemic agent. The unadjusted 6-month mortality rates were 3.4%, 5.8%, 7.8%, and 20.2% for patients with a postoperative peak CK-MB ratio (peak CK-MB value/ULN for laboratory test) of less than 5, ≥5 to <10, ≥10 to < 20, and ≥20 ULN, respectively.³⁵ The relation remained statistically significant after adjustment for ejection fraction (EF), congestive heart failure (CHF), cerebrovascular disease, peripheral vascular disease, cardiac arrhythmias, and the method of cardioplegia delivery. In the Arterial Revascularization Therapies Study (ARTS), 496 patients with multivessel CAD undergoing CABG were evaluated by CK-MB testing and followed after surgery at 30 days and 1 year.³⁶ Patients with increased cardiac enzyme levels after CABG were at increased risk for both death and repeat AMI within the first 30 days. CK-MB increase also was independently related to late adverse outcome.

Studies suggest that postcardiac surgery monitoring of troponins can be used to assess myocardial injury and risk stratification. Increased cardiac-specific troponin I or T in patients after CABG has been associated with a cardiac cause of death and with major postoperative complications within 2 years after CABG.^37,38 The ACC/ESC definition includes biomarkers but does not include specific criteria for diagnosing post-CABG AMI using cardiac biomarkers.¹⁹

There are a few new biomarkers of perioperative cardiac injury or ischemia under development. Brain natriuretic peptide (BNP) could be detected in the early stages of ischemia and decreases shortly after ischemic insult, allowing better detection of reinjury.³⁹ BNP concentrations after CABG in the patients who had cardiac events within 2 years were significantly greater than those in the patients free of cardiac events.⁴⁰ Soluble CD40 ligand (sCD40L) is another early biomarker of myocardial ischemia,⁴¹ and CPB causes an increase in the concentration of plasma sCD40L. A corresponding decrease in platelet CD40L suggests that this prothrombotic and proinflammatory protein was derived primarily from platelets and may contribute to the thrombotic and inflammatory complications associated with CPB.⁴² Future research will be required to determine how these biomarkers will be used to assess outcome after cardiac surgery.

Variability in Diagnosis of Perioperative Myocardial Infarction

The variability in diagnosing perioperative AMI has been studied by Jain and colleagues,⁴³ who evaluated data from 566 patients at 20 clinical sites, collected as part of a clinical trial. The occurrence of AMI by Q-wave, CK-MB, or autopsy criteria was determined. Of the 25% of patients who met the Q-wave, CK-MB, or autopsy criteria for AMI, 19% had increased CK-MB concentrations, as well as ECG changes. Q-wave and CK-MB or autopsy criteria for AMI were met by 4% of patients. Multicenter data collection showed a substantial variation in the incidence of AMI and an overall incidence rate of up to 25%. The definition of perioperative AMI was highly variable depending on the definitions used.

Clinicians are still in search for a “gold standard” approach to diagnose perioperative AMI. Perioperative myocardial necrosis/injury ranges from mild to severe and can have ischemic and nonischemic origin in patients undergoing cardiac surgery. Perioperative ECG changes, including Q-waves, and new RWMAs on ECGs are less reliable than in the nonperioperative arena. Currently, troponin I or T is the best indicator of myocardial damage after cardiac surgery. The level of enzymes correlates with the extension of the injury, but there is no universal cutoff point defining perioperative MI.

Predictors of Postoperative Morbidity and Mortality

Clinical and angiographic predictors of operative mortality were initially defined from the Coronary Artery Surgery Study (CASS).^47,48 A total of 6630 patients underwent isolated CABG between 1975 and 1978. Women had a significantly greater mortality rate than men; mortality increased with advancing age in men, but this was not a significant factor in women. Increasing severity of angina, manifestations of heart failure, and number and extent of coronary artery stenoses all correlated with greater mortality, whereas EF was not a predictor. Urgency of surgery was a strong predictor of outcome, with those patients requiring emergency surgery in the presence of a 90% left main coronary artery stenosis sustaining a 40% mortality rate.

A risk-scoring scheme for cardiac surgery (CABG and valve) was introduced by Paiement et al³ at the Montreal Heart Institute in 1983. Eight risk factors were identified: (1) poor left ventricular (LV) function, (2) CHF, (3) unstable angina or recent (within 6 weeks) MI, (4) age greater than 65 years, (5) severe obesity (body mass index > 30 kg/m²), (6) reoperation, (7) emergency surgery, and (8) other significant or uncontrolled systemic disturbances. Three classifications were identified: patients with none of these factors (normal), those presenting with one risk factor (increased risk), and those with more than one factor (high risk). In a study of 500 consecutive cardiac surgical patients, it was found that operative mortality increased with increasing risk (confirming their scoring system).

One of the most commonly used scoring systems for CABG was developed by Parsonnet and colleagues (Table 1-1).⁴⁹ Fourteen risk factors were identified for in-hospital or 30-day mortality after univariate regression analysis of 3500 consecutive operations. An additive model was constructed and prospectively evaluated in 1332 cardiac procedures. Five categories of risk were identified with increasing mortality rates, complication rates, and length of stay at the Newark Beth Israel Medical Center. The Parsonnet Index frequently is used as a benchmark for comparison among institutions. However, the Parsonnet model was created earlier than the other models and may not be representative of the current practice of CABG. During the period after publication of the Parsonnet model, numerous technical advances now in routine use have diminished CABG mortality rates.

TABLE 1-1 Components of the Additive Model

BP, blood pressure; CABG, coronary artery bypass graft; HD, heart disease; Hemo, hemodialysis; IABP, intra-aortic balloon pump; PA, pulmonary artery; PD, peritoneal dialysis; PTCA, percutaneous transluminal coronary angioplasty.

* On the actual worksheet, these risk factors require justification.

† Values were predictive of increased risk for operative mortality in univariate analysis.

From Parsonnet V, Dean D, Bernstein A: A method of uniform stratification of risk for evaluating the results of surgery in acquired adult heart disease. Circulation 79:I3, 1989, by permission.

Bernstein and Parsonnet⁵⁰ simplified the risk-adjusted scoring system in 2000 to provide a handy tool in preoperative discussions with patients and their families, and for preoperative risk stratification calculation. The authors developed a logistic regression model in which 47 potential risk factors were considered, and a method requiring only simple addition and graphic interpretation was designed for relatively easily approximating the estimated risk. The final estimates provided by the simplified model correlated well with the observed mortality (Figure 1-2).

Figure 1-2 Preoperative Risk-Estimation Worksheet.

(From Bernstein AD, Parsonnet V: Bedside estimation of risk as an aid for decision-making in cardiac surgery. Ann Thorac Surg 69:823, 2000, by permission from the Society of Thoracic Surgeons.)

O’Connor et al⁵¹ used data collected from 3055 patients undergoing isolated CABG at five clinical centers between 1987 and 1989 to develop a multivariate numerical score. A regression model was developed in a training set and subsequently validated in a test set. Independent predictors of in-hospital mortality included patient age, body surface area, comorbidity score, prior CABG, EF, LV end-diastolic pressure, and priority of surgery. The validated multivariate prediction rule was robust in predicting the in-hospital mortality for an individual patient, and the authors proposed that it could be used to contrast observed and expected mortality rates for an institution or a particular clinician.

Higgins et al⁵² developed a Clinical Severity Score for CABG at The Cleveland Clinic. A multivariate logistic regression model to predict perioperative risk was developed in 5051 patients undergoing CABG between 1986 and 1988, and subsequently validated in a cohort of 4069 patients. Independent predictors of in-hospital and 30-day mortality were emergency procedure, preoperative serum creatinine level of greater than 168 μmol/L, severe LV dysfunction, preoperative hematocrit of less than 34%, increasing age, chronic pulmonary disease, prior vascular surgery, reoperation, and mitral valve insufficiency. Predictors of morbidity (AMI and use of the intra-aortic balloon pump [IABP], mechanical ventilation for ≥3 days, neurologic deficit, oliguric or anuric renal failure, or serious infection) included diabetes mellitus, body weight of 65 kg or less, aortic stenosis, and cerebrovascular disease. Each independent predictor was assigned a weight or score, with increasing mortality and morbidity associated with an increasing total score.

The New York State model of Hannan et al⁵³ collected data over the years of 1989 through 1992 with 57,187 patients in a study with 14 variables. It was validated in 30 institutions. The mortality definition was “in-hospital.” The crude mortality rate was 3.1%; the receiver operating characteristic (ROC) curve was 0.7, with the Hosmer-Lemeshow (H-L) statistic less than 0.005. Observed mortality was 3.7%, and the expected mortality rate was 2.8%. They included only isolated CABG operations.

The Society of Thoracic Surgeons (STS) national database represents the most robust source of data for calculating risk-adjusted scoring systems. Established in 1989, the database has grown to include 892 participating hospitals in 2008. This provider-supported database allows participants to benchmark their risk-adjusted results against regional and national standards. This National Adult Cardiac Surgery Database (STS NCD) has become one of the largest in the world. New patient data are brought into the STS database on an annual and now semiannual basis. These new data have been analyzed, modeled, and tested using a variety of statistical algorithms. Since 1990, when more complete data collection was achieved, risk stratification models were developed for both CABG and valve replacement surgery. Models developed in 1995 and 1996 were shown to have good predictive value (Table 1-2; Figure 1-3).^54,55 In 1999, the STS analyzed the database for valve replacement with and without CABG to determine trends in risk stratification. Between 1986 and 1995, 86,580 patients were analyzed. The model evaluated the influence of 51 preoperative variables on operative mortality by univariate and multivariate analyses for the overall population and for each subset. After the significant risk factors were determined by univariate analysis, a standard logistic regression analysis was performed using the training-set population to develop a formal model. The test-set population then was used to determine the validity of the model. The preoperative risk factors associated with greatest operative mortality rates were salvage status, renal failure (dialysis dependent and nondialysis dependent), emergent status, multiple reoperations, and New York Heart Association class IV. The multivariate logistic regression analysis identified 30 independent preoperative risk factors among the 6 valvular models, isolated or in combination with CABG. The addition of CABG increased the mortality rate significantly for all age groups and for all subset models.⁵⁶

TABLE 1-2 Risk Model Results

From Shroyer AL, Plomondon ME, Grover FL, et al: The 1996 coronary artery bypass risk model: The Society of Thoracic Surgeons Adult Cardiac National Database. Ann Thorac Surg 67:1205, 1999, by permission of Society of Thoracic Surgeons.

Figure 1-3 A, Ordered risk deciles with equal number of records per group. After the predicted risk for each patient in the test set was determined, the patient records were arranged sequentially in order of predicted risk. The population was divided into 10 groups of equal size. The predicted mortality rate was compared with the actual mortality for each of the 10 groups. Dashed lines represent range of predicted mortality for a group of patients; bars represent actual mortality for a group of patients. B, Ordered risk deciles with equal number of deaths per group. After the predicted risk for each patient in the test set was determined, the patient records were arranged sequentially in order of predicted risk. The population was divided into 10 groups with equal numbers of deaths in each group. The predicted mortality was compared with the actual mortality for each of the 10 groups. Dashed lines represent range of predicted mortality for a group of patients; bars represent actual mortality for a group of patients. C, Ordered risk categories in clinically relevant groupings. After the predicted risk for each patient in the test set was determined, the patient records were arranged sequentially in order of predicted risk. The population was divided into seven clinically relevant risk categories. The predicted mortality was compared with the actual mortality for each of the seven groups. Dashed lines represent range of predicted mortality for a group of patients; bars represent actual mortality for a group of patients.

(A–C, From Shroyer AL, Plomondon ME, Grover FL, et al: The 1996 coronary artery bypass risk model: The Society of Thoracic Surgeons Adult Cardiac National Database. Ann Thorac Surg 67:1205, 1999, by permission of the Society of Thoracic Surgeons.)

There are currently three general STS risk models: CABG, valve (aortic or mitral), and valve plus CABG. These apply to seven specific, precisely defined procedures: the CABG model refers to an isolated CABG; the valve model includes isolated aortic or mitral valve replacement and mitral valve repair; and the valve and CABG model includes aortic valve replacement and CABG, mitral valve replacement and CABG, and mitral valve repair and CABG. Besides operative mortality, these models were developed for eight additional end points: reoperation, permanent stroke, renal failure, deep sternal wound infection, prolonged (> 24 hours) ventilation, major morbidity, and operative death, and finally short (< 6 days) and long (> 14 days) postoperative length of stay.^57–59 These models are updated periodically, every few years, and calibrated annually to provide an immediate and accurate tool for regional and national benchmarking, and have been proposed for public reporting. The calibration of the risk factors is based on the observed/expected (O/E) ratio, and calibration factors are updated quarterly. The expected mortality (E) is calibrated to obtain the national E/O ratio.

Tu et al⁶⁰ collected data from 13,098 patients undergoing cardiac surgery between 1991 and 1993 at all nine adult cardiac surgery institutions in Ontario, Canada. Six variables (age, sex, LV function, type of surgery, urgency of surgery, and repeat operation) predicted in-hospital mortality, ICU stay, and postoperative stay in days after cardiac surgery. Subsequently, the Working Group Panel on the Collaborative CABG Database Project categorized 44 clinical variables into 7 core, 13 level 1, and 24 level 2 variables, to reflect their relative importance in determining short-term mortality after CABG. Using data from 5517 patients undergoing isolated CABG at 9 institutions in Ontario in 1993, a series of models were developed. The incorporation of additional variables beyond the original six added little to the prediction of in-hospital mortality.

Spivack et al⁶¹ collected data during 1991 and 1992 and included 513 patients with 15 variables, validated only in their institution. They used only an isolated CABG population, and the outcomes measured were mortality and morbidity. The morbidity definition was ventilator time and ICU days. Both prolonged mechanical ventilation and death were rare events (8.3% and 2.0%, respectively). The combination of reduced LVEF and the presence of selected preexisting comorbid conditions (clinical CHF, angina, current smoking, diabetes) served as modest risk factors for prolonged mechanical ventilation; their absence strongly predicted an uncomplicated postoperative respiratory course.

The European System for Cardiac Operative Risk Evaluation (EuroSCORE) for cardiac operative risk evaluation was constructed from an analysis of 19,030 patients undergoing a diverse group of cardiac surgical procedures from 128 centers across Europe (Tables 1-3 and 1-4).^62,63 The following risk factors were associated with increased mortality: age, female sex, serum creatinine, extracardiac arteriopathy, chronic airway disease, severe neurologic dysfunction, previous cardiac surgery, recent MI, LVEF, chronic CHF, pulmonary hypertension, active endocarditis, unstable angina, procedure urgency, critical preoperative condition, ventricular septal rupture, noncoronary surgery, and thoracic aortic surgery.

TABLE 1-3 Risk Factors, Definitions, and Weights (Score)

CABG, coronary artery bypass graft surgery; LVEF, left ventricular ejection fraction.

From Nashef SA, Roques F, Michel P, et al: European system for cardiac operative risk evaluation (EuroSCORE). Eur J Cardiothorac Surg 16:9, 1999.

EuroSCORE provided a unique opportunity to assess the true risk of cardiac surgery in the absence of any identifiable risk factors. For the purposes of this analysis, baseline mortality figures were calculated in patients in whom no preoperative risk factors could be identified (including risk factors that were not found to have a significant impact in this study, such as diabetes and hypertension). When all such patients were excluded, it was gratifying to note the extremely low current mortality for cardiac surgery in Europe: 0% for atrial septal defect repair, 0.4% for CABG, and barely more than 1% for single valve repair or replacement.

During the 2000s, this additive EuroSCORE has been used widely and validated across different centers in Europe and across the world, making it a primary tool for risk stratification in cardiac surgery.^64–75 Although its accuracy has been well established for CABG and isolated valve procedures, its predictive ability in combined CABG and valve procedures has been less well studied. Karthik et al⁶⁶ showed that, in patients undergoing combined procedures, the additive EuroSCORE significantly underpredicted the risk compared with the observed mortality. In this subset, they determined that the logistic EuroSCORE is a better and more accurate method of risk assessment.

Dupuis et al⁷⁶ attempted to simplify the approach to risk of cardiac surgical procedures in a manner similar to the original American Society of Anesthesiologists (ASA) physical status classification. They developed a score that uses a simple continuous categorization, using five classes plus an emergency status (Table 1-5). The Cardiac Anesthesia Risk Evaluation (CARE) score model collected data from 1996 to 1999 and included 3548 patients to predict both in-hospital mortality and a diverse group of major morbidities. It combined clinical judgment and the recognition of three risk factors previously identified by multifactorial risk indices: comorbid conditions categorized as controlled or uncontrolled, the surgical complexity, and the urgency of the procedure. The CARE score demonstrated similar or superior predictive characteristics compared with the more complex indices.

TABLE 1-5 Cardiac Anesthesia Risk Evaluation Score

* Examples: controlled hypertension, diabetes mellitus, peripheral vascular disease, chronic obstructive pulmonary disease, controlled systemic diseases, others as judged by clinicians.

† Examples: unstable angina treated with intravenous heparin or nitroglycerin, preoperative intra-aortic balloon pump, heart failure with pulmonary or peripheral edema, uncontrolled hypertension, renal insufficiency (creatinine level > 140 μmol/L, debilitating systemic diseases, others as judged by clinicians).

‡ Examples: reoperation, combined valve and coronary artery surgery, multiple valve surgery, left ventricular aneurysmectomy, repair of ventricular septal defect after myocardial infarction, coronary artery bypass of diffuse or heavily calcified vessels, others as judged by clinicians.

From Dupuis JY, Wang F, Nathan H, et al: The cardiac anesthesia risk evaluation score: A clinically useful predictor of mortality and morbidity after cardiac surgery. Anesthesiology 94:194, 2001, by permission.

Nowicki et al⁷⁷ used data on 8943 cardiac valve surgery patients aged 30 years and older from eight northern New England medical centers from 1991 through 2001 to develop a model to predict in-hospital mortality. In the multivariate analysis, 11 variables in the aortic model (older age, lower body surface area, prior cardiac operation, increased creatinine, prior stroke, NYHA class IV, CHF, atrial fibrillation, acuity, year of surgery, and concomitant CABG) and 10 variables in the mitral model (female sex, older age, diabetes, CAD, prior cerebrovascular accident, increased creatinine, NYHA class IV, CHF, acuity, and valve replacement) remained independent predictors of the outcome. They developed a look-up table for mortality rate based on a simple scoring system.

Hannan and colleagues⁷⁸ also evaluated predictors of mortality after valve surgery but used data from 14,190 patients from New York State. A total of 18 independent risk factors were identified in the 6 models of differing combinations of valve and CABG. Shock and dialysis-dependent renal failure were among the most significant risk factors in all models. The risk factors and odds ratios are shown in Tables 1-6, 1-7, and 1-8. They also studied which risk factors are associated with early readmission (within 30 days) after CABG. Of 16,325 total patients, 2111 (12.9%) were readmitted within 30 days for reasons related to CABG. Eleven risk factors were found to be independently associated with greater readmission rates: older age, female sex, African American race, greater body surface area, previous AMI within 1 week, and six comorbidities. After controlling for these preoperative patient-level risk factors, two provider characteristics (annual surgeon CABG volume < 100 and hospital risk-adjusted mortality rate in the highest decile) and two postoperative factors (discharge to nursing home or rehabilitation/acute care facility and length of stay during index CABG admission of ≥5 days) also were related to greater readmission rates. The development of several excellent risk models for cardiac valve surgery provides a powerful new tool to improve patient care, select procedures, counsel patients, and compare outcomes (see Chapter 19).⁷⁹

Consistency Among Risk Indices

Many different variables have been found to be associated with the increased risk during cardiac surgery, but only a few variables consistently have been found to be major risk factors across multiple and very diverse study settings. Age, female sex, LV function, body habitus, reoperation, type of surgery, and urgency of surgery were some variables consistently present in most of the models (Box 1-3).

Although a variety of investigators have found different comorbid diseases to be significant risk factors, no diseases have been shown to be consistent risk factors, with the possible exception of renal dysfunction and diabetes. These two comorbidities have been shown to be important risk factors in a majority of the studies (Box 1-4).

Applicability of Risk Indices to a Given Population

It is critical to understand how these indices were created to understand how best to apply a given risk index to a specific patient or population. Specifically, the application of these risk models must be done with caution and after careful study for any specific population. One issue is that the profile of patients undergoing cardiac surgery is constantly changing, and patients who previously would not have been considered for surgery (and thus not included in the development data set) are now undergoing surgery. Therefore, the models require continuous updating and revision. In addition, cardiac surgery itself is changing with the increasing use of off-pump and less invasive procedures, which may change the nature of the influence of preexisting conditions.

One critical factor in the choice of model to use for a given practice is to understand the clinical goals used in the original development process. In addition, despite extensive research and widespread use of risk models in cardiac surgery, there are methodologic problems. The extent of the details in the reports varies greatly. Different conclusions can be reached depending on the risk model used. Processes critical to the development of risk models are shown in Figure 1-4.

Figure 1-4 Risk model development.

(From Omar RZ, Ambler G, Royston P, et al: Cardiac surgery risk modeling for mortality: A review of current practice and suggestions for improvement. Ann Thorac Surg 77:2232, 2004, by permission of the Society of Thoracic Surgeons.)

The underlying assumption in the development of any risk index is that specific factors (disease history, physical findings, laboratory data, nature of surgery) cannot be modified with respect to their influence on outcome; that is, the perioperative period is essentially a black box. If a specific factor is left untreated, it could lead to major morbidity or mortality. For example, the urgency of the planned surgical procedure and baseline comorbidities cannot be changed. However, the models themselves depend on the appropriate selection of baseline variables or risk factors to study, and their prevalence in the population of interest is critical for them to affect outcome. For example, referral patterns to a given institution may result in an absence of certain patient populations and, therefore, the risk factor would not appear in the model. Also, the use of multivariate logistic regression may eliminate biologically important risk factors, which are not present in sufficient numbers to achieve statistical significance.

In developing a risk index, it is also important to validate the model and to benchmark it against other known means of assessing risks. It is important to determine whether the index predicts morbidity, mortality, or both. Typically, a model’s performance is first evaluated on the developmental data, evaluating its goodness of fit. Alternatively, the original data can be split and the model can be built on half of the data and validated on the other half. This reduces the total number of patients and outcomes available to create the model. This method is best suited to situations in which data on tens of thousands of patients are available. This internal validation does not provide the practitioner with information on the generalizability of the model. External validation on a large, completely independent test dataset is the best approach to satisfying this requirement.

In addition to validation, calibration refers to a model’s ability to predict mortality accurately. Numerous tests can be applied, the most common being the H-L test. If the P value from an H-L test is greater than 0.05, the current practice of the developers is to claim that the model predicts mortality accurately.

Discrimination is the ability of a model to distinguish patients who die from those who survive. The area under the ROC is the common method of assessing this facet of the model. In brief, the test is determined by evaluating all possible pairs of patients, determining whether the predicted probability of death should ideally be greater for the patient who died than for the one who survived. The ROC area is the percentage of pairs for which this is true. The current practice in cardiac surgery is to conclude that a model discriminates well if the ROC area is greater than 0.7. If predictions are used to identify surgical centers or surgeons with unexpectedly high or low rates, achieving a high ROC area alone is not adequate, but good calibration is also critical. A poorly calibrated model may cause large numbers of institutions or surgeons to reveal excessively high or low rates of mortality, when, in fact, the fault lies with the model, not the clinical performance. If predictions are used to stratify patients by disease severity to compare treatments or to decide on patient management, both calibration and discrimination aspects are important.

A key problem in the development of cardiac surgery risk stratification models is the evolving practice of surgery. This includes new procedures, or variations on older procedures, which may affect perioperative risk and not be accounted for in the data used to develop the model. Despite these limitations, calibrated and validated risk model remains the most objective tool currently available. Clinicians need to understand the specific model, its strengths and weaknesses, to appropriately apply the model in academic research, patient counseling, benchmarking, and management of resources.

Renal Dysfunction

Renal dysfunction has been shown to be an important risk factor for surgical mortality in patients undergoing cardiac surgery.^80–82 However, the spectrum of what constitutes renal dysfunction is broad, with some models defining it as increased creatinine levels and others defining it as dialysis dependency.

The Northern New England Cardiovascular Study Group reported a 12.2% in-hospital mortality rate after CABG in patients on chronic dialysis versus a 3.0% mortality rate in patients not on dialysis.⁸³ However, the incidence of dialysis dependency in the cardiac surgical population is sufficiently low (e.g., 0.5% in New York State) so that it may not enter into many of the models developed.

Acute kidney injury (AKI) after cardiac surgery carries significant morbidity and mortality. Patients who experienced development of severe renal dysfunction (defined as glomerular filtration rate [GFR] < 30 mL/min) after CABG had an almost 10% mortality rate compared with 1% mortality in those with normal renal function.⁸⁴ Poor outcome associated with perioperative AKI has led to development of predictive models of AKI to identify patients at risk. One of the recent models predicts need for renal replacement therapy (RRT) after cardiac surgery. Wijeysundera et al⁸⁵ retrospectively studied a cohort of 20,131 cardiac surgery patients at 2 hospitals in Ontario, Canada. Multivariate predictors of RRT were preoperative estimated GFR, diabetes mellitus requiring medication, LVEF, previous cardiac surgery, procedure, urgency of surgery, and preoperative IABP. An estimated GFR less than or equal to 30 mL/min was assigned 2 points; other components were assigned 1 point each: estimated GFR of 31 to 60 mL/min, diabetes mellitus, EF less than or equal to 40%, previous cardiac surgery, procedure other than CABG, IABP, and nonelective case. Among the 53% of patients with low risk scores (≤1), the risk for RRT was 0.4%; by comparison, this risk was 10% among the 6% of patients with high-risk scores (≥4). Another group developed a robust prediction rule to assist clinicians in identifying patients with normal, or near-normal, preoperative renal function who are at high risk for development of severe renal insufficiency.⁸⁶ In a multivariate model, the preoperative patient characteristics most strongly associated with postoperative severe renal insufficiency included age, sex, white blood cell count > 12,000, prior CABG, CHF, peripheral vascular disease, diabetes, hypertension, and preoperative IABP.

A major issue with respect to the development of indices to predict perioperative renal failure is that the pathophysiology of perioperative AKI includes inflammatory, nephrotoxic, and hemodynamic insults. This multifactorial nature of AKI might be one of the reasons that a limited single-strategy approach has not been successful.⁸⁷ Contrast agents used for angiography before cardiac surgery represent one of the modifiable nephrotoxic factors perioperatively. Delaying cardiac surgery beyond 24 hours after the exposure and minimizing the contrast agent load can decrease the incidence of AKI in elective cardiac surgery cases.⁸⁸

Uniformity of AKI definition (Risk of renal dysfunction, Injury to the kidney, Failure of kidney function, Loss of kidney function, and End-stage kidney disease; RIFLE) improved risk stratification models and utilization of early biomarkers of AKI hopefully will provide tools to design clinical trials addressing this important issue.^89,90

Diabetes

The association between diabetes and mortality with cardiac surgery has been inconsistent, with some studies supporting the association, whereas other studies do not.^91–98 Several recent trials have evaluated outcome between CABG and PCI in patients with diabetes. In the CARDia (Coronary Artery Revascularization in Diabetes) trial,⁹⁹ a total of 510 patients with diabetes with multivessel or complex single-vessel CAD from 24 centers were randomized to PCI plus stenting (and routine abciximab) or CABG. At 1 year of follow-up, the composite rate of death, MI, and stroke was 10.5% in the CABG group and 13.0% in the PCI group (hazard ratio [HR]: 1.25; 95% CI: 0.75 to 2.09; P = 0.39), all-cause mortality rates were 3.2% and 3.2%, and the rates of death, MI, stroke, or repeat revascularization were 11.3% and 19.3% (HR: 1.77; 95% CI: 1.11 to 2.82; P = 0.02), respectively. The Bypass Angioplasty Revascularization Investigation 2 Diabetes (BARI 2D) trial randomized 2368 patients with both type 2 diabetes and heart disease to undergo either prompt revascularization with intensive medical therapy or intensive medical therapy alone, and to undergo either insulin-sensitization or insulin-provision therapy.¹⁰⁰ In patients with more extensive CAD, similar to those enrolled in the CABG stratum, prompt CABG, in the absence of contraindications, intensive medical therapy, and an insulin sensitization strategy appears to be a preferred therapeutic strategy to reduce the incidence of MI.¹⁰¹

Acute Coronary Syndrome

Patients with a recent episode of non–ST-segment elevation acute coronary syndrome before CABG have greater rates of operative morbidity and mortality than do patients with stable coronary syndromes.¹⁰² However, a recent report of the American College of Cardiology Foundation, in collaboration with numerous other societies, has published appropriateness for coronary revascularization.¹⁰³ There are numerous Class A recommendations for revascularization and, therefore, many patients may come to the operating room directly after coronary angiography and potentially after attempted stent placement with antiplatelet agents. There is evidence to suggest that delaying CABG for 3 to 7 days in patients after ST-elevation myocardial infarction (STEMI) or non–ST-elevation myocardial infarction (NSTEMI) is beneficial in selected stable patients with contraindications to PCI. In addition, patients with a hemodynamically significant right ventricular MI should be allowed to recover the injured ventricle.¹⁰⁴

Nonexercise (Pharmacologic) Stress Testing

Pharmacologic stress testing has been advocated for patients in whom exercise tolerance is limited, both by comorbid diseases and by symptomatic peripheral vascular disease. Often, these patients may not stress themselves sufficiently during daily life to provoke symptoms of myocardial ischemia or CHF. Pharmacologic stress testing techniques either increase myocardial oxygen demand (dobutamine)¹¹¹ or produce coronary vasodilatation leading to coronary flow redistribution (dipyridamole/adenosine).¹¹² Echocardiographic or nuclear scintigraphic imaging (SPECT) are used in conjunction with the pharmacologic therapy to perform myocardial perfusion imaging for risk stratification and myocardial viability assessment (Box 1-6) (see Chapters 2, 3, 6, 11 to 15, and 18).