Does Preoperative Hemoglobin Independently Predict Short-Term Outcomes After Coronary Artery Bypass Graft Surgery?

Transcription

1 CARDIOTHORACIC ANESTHESIOLOGY: The Annals of Thoracic Surgery CME Program is located online at To take the CME activity related to this article, you must have either an STS member or an individual non-member subscription to the journal. Does Preoperative Hemoglobin Independently Predict Short-Term Outcomes After Coronary Artery Bypass Graft Surgery? Melanie L. Bell, PhD, Gary K. Grunwald, PhD, Janet H. Baltz, MSN, Gerald O. McDonald, MD, Missy R. Bell, MBA, MSHA, Frederick L. Grover, MD, and A. Laurie W. Shroyer, PhD Department of Preventive and Social Medicine, University of Otago, Dunedin, New Zealand; Division of Cardiac Research, Eastern Colorado Health Care System Department of Veterans Affairs Medical Center, Departments of Biostatistics and Informatics, Surgery, and Medicine, University of Colorado Denver, Colorado; Department of Veterans Affairs, Office of Patient Care Services, Washington, DC; and Department of Veterans Affairs Medical Center, Northport, New York Background. The role of preoperative hemoglobin in predicting short-term coronary artery bypass graft (CABG) surgery outcomes has not been well established. This study investigated the association between preoperative hemoglobin level with 30-day operative mortality and perioperative complications. Methods. For the period from 1997 to 2005, the Department of Veterans Affairs Continuous Improvement in Cardiac Surgery Program database was used to extract 36,658 CABG-only deidentified records for patients undergoing cardiopulmonary bypass. Univariate and multivariate statistical models were used to predict both mortality and morbidity outcomes for varying levels of hemoglobin. Results. Unadjusted odds of 30-day operative mortality for patients with preoperative hemoglobin level less than 10 g/dl were 2.37 times higher than for patients with hemoglobin levels of 10 g/dl or greater (95% confidence interval: 1.84 to 3.05; p < ). Multivariable effect upon 30-day operative mortality decreased considerably (odds ratio 1.29, 95% confidence interval: 0.99 to 1.68; p 0.064). Conclusions. As several risk factors may occur concurrently with low hemoglobin, preoperative hemoglobin level less than 10 g/dl was not a strong independent predictor of 30-day operative mortality or perioperative morbidity in multivariate models for on-pump CABGonly patients. Hemoglobin and serum creatinine were correlated, with creatinine exhibiting the stronger association with mortality. These findings suggest that moderate or severe anemia may be intertwined with other risk factors, such as serum creatinine or congestive heart failure. For a CABG patient subgroup presenting with a complex clinical preoperative profile, therefore, riskmodel approaches may be inherently limited in separately identifying the predictors of anemic patients outcomes. (Ann Thorac Surg 2008;86: ) 2008 by The Society of Thoracic Surgeons Coronary artery bypass graft surgery (CABG) is one of the most frequently performed major operations in the United States. Although many large databases of CABG procedural outcomes and risk factors have been analyzed, questions remain regarding the role of preoperative hemoglobin in predicting short-term outcomes after CABG. Conceptually, anemic patients may be at higher risk for adverse cardiac surgery events, as the oxygen-carrying capacity of the blood through hemoglobin is crucial for adequate function of organs and tissues. Carson and Accepted for publication July 28, Address correspondence to Dr Shroyer, Research and Development Office, Northport Veterans Affairs Medical Center, 79 Middleville Rd (151), Bldg 62, Room 125, Northport, NY 11768; Annie. Shroyer@va.gov. associates [1] found nontransfused CABG patients with preoperative hemoglobin level less than 11.0 g/dl had particularly poor 30-day outcomes. The role of hemoglobin as a predictor of CABG mortality and morbidity was examined by Zindrou and coworkers [2] in a study of 2,059 patients undergoing CABG between 1993 and They developed univariate and multivariable logistic models for in-hospital ( 28 days) death. Patients with low preoperative hemoglobin had an overall adjusted odds ratio (OR) of 3.2 (p 0.016). An earlier study by Magovern and associates [3] found a much smaller adjusted OR of 1.8. Weightman and associates [4] found hemoglobin to be a nonsignificant predictor, with OR 1.2. In multivariate models, Kulier and associates [5] found significant associations of low pre by The Society of Thoracic Surgeons /08/$34.00 Published by Elsevier Inc doi: /j.athoracsur

2 1416 BELL ET AL Ann Thorac Surg PREOPERATIVE HEMOGLOBIN AND OUTCOMES AFTER CABG 2008;86: operative hemoglobin with postoperative noncardiac complications (adjusted OR 1.95 for hemoglobin 10 g/dl versus 14 g/dl). The strongest association found was with renal complications. Associations found for cardiac complications were not significant. To address the previous conflicting literature findings, the goal for this Department of Veterans Affairs (VA) National Cardiovascular Care Improvement Program (NCCIP) analysis was to examine the association of preoperative hemoglobin with short-term outcomes after CABG-only procedures. Using similar multivariable modeling applied to the VA Continuous Improvement in Cardiac Surgery Program (CICSP) database, the VA NCCIP team sought to identify the univariate and multivariate associations of hemoglobin with 30-day operative mortality, and hemoglobin with postoperative complications to improve future CICSP risk-modeling endeavors. Additionally, the prevalence of risk factors and incidence of short-term outcomes are presented across the range of preoperative hemoglobin levels. Table 1. Patient Characteristics (n 36,658) by Preoperative Hemoglobin Classification a Preoperative Hemoglobin Range (g/dl) WHO Anemia Ranges Severe Moderate Normal Missing Overall Number of patients 245 (0.7%) 764 (2.1%) 1,796 (4.9%) 3,338 (9.1%) 5,638 (15.4%) 24,558 (67.0%) 319 (0.9%) 36,658 Noncardiac risk factors Male Age (years) 66.0 (9.5) 66.7 (9.4) 67.8 (9.2) 67.3 (9.3) 65.7 (9.2) 62.6 (9.1) 62.0 (9.5) 63.8 (9.3) Body surface area (m 2 ) 2.00 (0.20) 1.99 (0.20) 1.98 (0.21) 1.99 (0.21) 2.01 (0.20) 2.04 (0.20) 2.02 (0.18) 2.03 (0.20) COPD Peripheral vascular disease Cerebral vascular disease Serum creatinine mg/dl Serum creatinine (mg/dl) 1.54 (1.24) 1.77 (1.64) 1.58 (1.27) 1.41 (0.98) 1.25 (0.64) 1.13 (0.43) 1.39 (1.33) 1.2 (0.68) Diabetes mellitus Current smoker Partially or totally dependent functional status Cardiac risk factors Prior heart surgery Prior myocardial infarction Prior percutaneous coronary intervention Preoperative intra-aortic balloon pump Preoperative intravenous nitroglycerin CCS anginal class III or IV NYHA functional class III or IV Urgent or emergent surgical priority Preoperative ST-segment depression on electrocardiogram Left ventricular ejection fraction 35% Left main stenosis 50% Three or more coronary arteries with stenosis 50% a For binary variables the percentage with characteristic within each hemoglobin classification is given. For continuous variables, values shown are mean (SD). The final column gives the overall percentage or mean (SD) of patients with the characteristic. CCS Canadian Cardiovascular Society; COPD chronic obstructive pulmonary disease; NYHA New York Heart Association; WHO World Health Organization.

3 Ann Thorac Surg BELL ET AL 2008;86: PREOPERATIVE HEMOGLOBIN AND OUTCOMES AFTER CABG Patients and Methods Patient Population Since 1987, the VA CICSP has collected risk and outcome data on all veterans undergoing cardiac surgery at 44 VA cardiac surgical centers [6, 7]. This CICSP study extracted deidentified records for all CABG-only operations performed with cardiopulmonary bypass for the period from October 1, 1997, to September 30, As part of the VA PL quality improvement endeavors to enhance risk-modeling approaches, this study was approved by local Institutional Review Boards with waiver of informed consent, and waiver of Health Insurance Portability and Accountability Act for authorization received (COMIRB and NVAMC ID: 00298) During 1997 to 2005, there were 44,353 VA CABG-only procedures performed. Of these, 37 records (0.08%) were omitted owing to missing outcome information (operative mortality or major complications), and 7,658 (17.3%) more were omitted because of cardiopulmonary bypass time 0 (off-pump procedures), leaving 36,658 CABGonly operations for analysis in this study. Outcomes The two short-term outcomes evaluated for this study were 30-day operative mortality and 30-day perioperative morbidity. The CICSP 30-day operative mortality is defined as any death occurring during the index hospitalization or within 30 days after surgery, and occasionally includes the rare death after discharge that occurs Fig 1. (A C) Prevalence and incidence of risk factors by hemoglobin level. Risk factors that are nearly constant (body surface area and sex) are not shown. Patients with missing hemoglobin measurements are not shown.

4 1418 BELL ET AL Ann Thorac Surg PREOPERATIVE HEMOGLOBIN AND OUTCOMES AFTER CABG 2008;86: Table 2. Postoperative Outcomes by Preoperative Hemoglobin Classification a WHO Anemia Ranges Missing Preoperative Hemoglobin Range (g/dl) Severe Moderate Normal Number (%) of Patients Number of patients 319 (0.9%) 245 (0.7%) 764 (2.1%) 1,796 (4.9%) 3,338 (9.1%) 5,638 (15.4%) 24,558 (67.0%) 36,658 Operative death ,143 (3.1) Any complication b ,230 (11.5) Endocarditis (0.010) Renal failure requiring (1.2) dialysis Mediastinitis (1.2) Ventilator for ,464 (6.7) hours Reoperation for (2.5) bleeding Reoperation requiring (0.25) bypass Coma (0.46) Stroke (1.7) Cardiac arrest (2.2) a Values given are percentages unless otherwise noted. The final column shows the number and percentage of patients with the complication. the nine specified complications. WHO World Health Organization. b At least one of beyond 30 days after surgery, if a direct result of a cardiac procedural complication. All deaths are verified by matching with the national death registry for all veterans [8]. The CICSP 30-day perioperative morbidity is defined as new occurrence of any of nine major morbidities after CABG, including endocarditis, renal failure requiring dialysis, mediastinitis, reoperation for bleeding, mechanical ventilator used postoperatively for longer than 48 hours, Fig 2. Postoperative morbidity by hemoglobin levels. Complications that have an average incidence less than 1% (endocarditis, reoperation requiring bypass, and coma) are not shown. Patients with missing hemoglobin measurements are not shown. repeat cardiac surgery, stroke, coma for longer than 24 hours, or cardiac arrest requiring cardiopulmonary resuscitation within 30 days after surgery. Preoperative Hemoglobin Assessment Since October 1, 1997, preoperative hemoglobin (the most recent laboratory assessment within 2 weeks before cardiac surgery) was gathered routinely in CICSP. Hemoglobin was analyzed three different ways: (1) as a continuous variable; 2) as a World Health Organization (WHO) categorical variable (hemoglobin 10 g/dl is severe anemia, 10 to 12 g/dl is moderate anemia, and 12 g/dl is normal ) [9]; and (3) as a binary variable (recording hemoglobin 10 g/dl as abnormal using Zindrou s threshold [2]). Unfortunately, the CICSP gathers data only for preoperative hemoglobin status. No CICSP information is available regarding blood products administered preoperatively, intraoperatively, or postoperatively. During or after cardiopulmonary bypass, no CICSP information was available on hemoglobin values (eg, lowest hemoglobin during bypass). Although not a perfect normal distribution, preoperative hemoglobin was roughly symmetric, with a mean of 13.6 g/dl and standard deviation of 1.7 g/dl. Preoperative hemoglobin was missing for 319 CICSP records (0.9%). Analyses were completed by both omitting these records and imputing the missing hemoglobin values to the median (13.7 g/dl), and no substantive differences were observed. Given the similarity of results, only results of analyses with imputation are presented. CICSP Risk Variables In addition to hemoglobin, 22 other preoperative risk variables (10 demographic and 12 cardiac-related data

5 Ann Thorac Surg BELL ET AL 2008;86: PREOPERATIVE HEMOGLOBIN AND OUTCOMES AFTER CABG Table 3. Univariate Associations of Preoperative Hemoglobin (Binary, Continuous, and Categorized) With Operative Mortality and Morbidity OR 95% CI p Value Operative mortality, 30-day Hgb 10 g/dl 2.37 (1.84, 3.05) Per one g/dl decrease in 1.21 (1.17, 1.25) Hgb (1.82, 4.95) (2.19, 3.94) (1.96, 3.01) (1.64, 2.35) (1.34, 1.84) Reference Morbidity a Hgb 10 g/dl 1.87 (1.59, 2.19) Per one g/dl decrease in 1.15 (1.12, 1.17) Hgb (1.08, 2.21) (1.96, 2.80) (1.72, 2.22) (1.44, 1.77) (1.18, 1.41) Reference a Morbidity associated with postoperative complications: endocarditis, renal failure requiring dialysis, mediastinitis, reoperation for bleeding, mechanical ventilator used postoperatively for longer than 48 hours, repeat cardiac surgery, stroke, coma for longer than 24 hours, or cardiac arrest requiring cardiopulmonary resuscitation within 30 days after surgery. CI confidence interval; Hgb hemoglobin; OR odds ratio factors were not uniformly captured throughout the study period. As the best CICSP surrogate metrics, previous heart surgery was used in place of redo CABG operation and surgical priority (urgent or emergent versus elective) was used to approximate cardiogenic shock. To find the best models for 30-day operative mortality, hemoglobin was analyzed both as a continuous and categorical variable, and backward selection was used with the risk factors shown in Table 1. Creatinine was also considered both continuously and as a binary variable, using greater than 1.7 mg/dl to indicate high creatinine. The interaction between hemoglobin and creatinine was also tested. To explore the possibility of nonlinear relationships, the method of fractional polynomials [12, 13] for hemoglobin was used. Postoperative morbidity was analyzed similarly. Since timing of deaths was known, 30-day operative survival was also modeled using Cox proportional hazards models, both univariately for hemoglobin alone and multivariately including the risk variables shown in Table 1. Post-hoc 2 tests of trend were performed to assess risk variables and complications, and their relation to hemoglobin levels. Results Study Population Table 1 shows patient risk factors by preoperative hemoglobin values. Using WHO criteria, 2.8% had severe anemia, 14.0% had moderate anemia, and 82.4% had normal preoperative hemoglobin levels. Figure 1 shows elements) were considered risk-model eligible (Table 1). No risk variables were missing more than 5% of values. To avoid possible bias from excluding records with missing values, these missing values were imputed to the median for continuous variables or the most common category for categorical variables, using the standard CICSP modeling approach. Statistical Analyses Statistical analyses were performed using SAS software (SAS Institute, Cary, North Carolina) [10] and Stata (StataCorp, College Station, Texas) [11] with statistical significance of p set at Risk variables and outcomes (including specific complications) are reported for the prespecified WHO-based hemoglobin ranges (Table 1). Logistic regression was used to assess associations of preoperative hemoglobin with 30-day operative mortality and 30-day perioperative morbidity. Univariate models containing only preoperative hemoglobin were fit, as well as multivariate models including the risk variables shown in Table 1, and hemoglobin, which was forced to remain in the model when selection procedures were performed in multivariate models. For comparison with the results of Zindrou and coworkers [2], the same risk factors were used with the exception of thyroid disease and hypertension, as these CICSP risk Table 4. Multivariate Model for 30-Day Operative Mortality With Hemoglobin Less Than 10 g/dl Variable OR 95% CI p Value Hemoglobin 10 g/dl 1.29 (0.99, 1.68) Age (years) (1.036, 1.050) COPD 1.30 (1.14, 1.48) Peripheral vascular disease 1.49 (1.31, 1.69) Cerebral vascular disease 1.33 (1.16, 1.52) Serum creatinine 1.7 mg/dl 1.67 (1.42, 1.98) Partially or totally dependent 1.22 (1.05, 1.42) functional status Prior heart surgery 2.55 (2.12, 3.05) Prior myocardial infarction 1.43 (1.25, 1.64) Preoperative intra-aortic 1.91 (1.57, 2.31) balloon pump Preoperative intravenous 1.31 (1.11, 1.55) nitroglycerin NYHA functional class III or 1.46 (1.27, 1.67) IV Preoperative ST-segment 1.34 (1.16, 1.55) depression on ECG Left ventricular ejection fraction 35% 1.30 (1.11, 1.53) Hosmer-Lemeshow goodness of fit: 2 8.1, p 0.4; C-index: CI confidence interval; COPD chronic obstructive pulmonary disease; ECG electrocardiogram; NYHA New York Heart Association; OR odds ratio.

6 1420 BELL ET AL Ann Thorac Surg PREOPERATIVE HEMOGLOBIN AND OUTCOMES AFTER CABG 2008;86: Table 5. Multivariate Model for 30-Day Postoperative Morbidity With Hemoglobin Less Than 10 g/dl Indicator Variable OR 95% CI p Value Hemoglobin 10 g/dl 1.20 (1.02, 1.43) Age (years) (1.022, 1.031) Male 1.65 (1.23, 2.21) Body surface area (m 2 ) 1.36 (1.14, 1.62) COPD 1.25 (1.16, 1.35) Peripheral vascular disease 1.26 (1.17, 1.35) Cerebral vascular disease 1.28 (1.19, 1.39) Serum creatinine 1.7 mg/dl 1.69 (1.53, 1.87) Diabetes mellitus 1.19 (1.11, 1.27) Current smoker 1.22 (1.13, 1.32) Partially or totally dependent 1.19 (1.09, 1.30) functional status Prior heart surgery 1.56 (1.37, 1.77) Preoperative intra-aortic 1.92 (1.70, 2.16) balloon pump Preoperative intravenous 1.26 (1.14, 1.39) nitroglycerin NYHA functional class III or 1.44 (1.34, 1.56) IV Urgent or emergent surgical 1.15 (1.05, 1.25) priority Preoperative ST-segment 1.10 (1.01, 1.20) depression on ECG Left ventricular ejection 1.26 (1.14, 1.38) fraction 35% Three or more coronary arteries with stenosis 50% 1.15 (1.06, 1.25) Hosmer-Lemeshow goodness of fit: 2 6.3, p 0.6; C-index: CI confidence interval; COPD chronic obstructive pulmonary disease; ECG electrocardiogram; NYHA New York Heart Association; OR odds ratio. most of the risk factors by hemoglobin level, except for sex and body surface area, as they were nearly constant. Interestingly, the 245 patients (0.7%) with hemoglobin less than 9 g/dl had better risk profiles compared with the 764 patients with hemoglobin 9 to 9.9 g/dl. The risk variables that showed greater prevalence in that group were angina functional class of III or IV and prior heart surgery. Aside from these severely anemic patients (and those with missing hemoglobin values), a dose-response relationship appears with hemoglobin, with highest prevalence of most risk factors in the 9 to 9.9 g/dl range, or in a few cases, the 10 to 10.9 g/dl range. Post-hoc 2 tests of trend yielded p values from to less than for decreasing risk prevalence (except for prior heart surgery and current smoking) with increasing hemoglobin level. Table 2 and Figure 2 show short-term postoperative outcomes by WHO hemoglobin ranges. The 30-day operative mortality rate was 3.11%, and 11.54% of patients had at least one of the nine major complications reported. Both 30-day operative mortality and 30-day perioperative morbidity decreased with increasing hemoglobin range (p values for trend ). Incidence of most (5 of 9) complications was lower in the most severe anemia group as compared with the moderately severe anemia group. A statistically significant dose-response relationship for patients with hemoglobin level of 9 g/dl or greater was found for all complications (p values for trend ranged from 0.02 to ) except for endocarditis. Among the hemoglobin less than 9.0 g/dl group, however, there is a striking degree of renal failure requiring dialysis, which is 2.8 times higher than that of the normal hemoglobin group. Associations of Hemoglobin With Short-Term Clinical Outcomes Univariately, low preoperative hemoglobin is strongly associated with 30-day operative mortality and 30-day perioperative morbidity, as shown in Table 3. The unadjusted odds of operative mortality for patients with preoperative hemoglobin less than 10 g/dl were 2.4 times as great as those for patients with hemoglobin of 10 g/dl or greater (95% confidence interval [CI]: 1.84 to 3.05; p ). When hemoglobin was used as a continuous variable, each 1 g/dl decrease in hemoglobin was associated with increased odds of mortality by a factor of 1.2 (95% CI: 1.17 to 1.25; p ). Similarly, the association of preoperative hemoglobin less than 10 g/dl with morbidity was also statistically significant with OR 1.87 (95% CI: 1.59 to 2.19; p ). Linearity of the univariate log odds with hemoglobin for both mortality and morbidity was examined using fractional polynomials, and no striking deviations were found, indicating the modeling approach used was appropriate. After performing a stepwise, backward-selected multivariate logistic regression starting with the other 22 preoperative clinical risk factors shown in Table 1, the univariate relationships of hemoglobin with outcomes decreased considerably, with the OR for hemoglobin less than 10.0 g/dl decreasing to 1.29 (95% CI: 0.99 to 1.68; p Fig 3. Survival curves for severe, moderate, and normal hemoglobin (Hgb) groups. (CABG coronary artery bypass graft surgery.)

7 Ann Thorac Surg BELL ET AL 2008;86: PREOPERATIVE HEMOGLOBIN AND OUTCOMES AFTER CABG 0.064) for operative mortality, and 1.20 (95% CI: 1.02 to 1.43; p 0.033) for morbidity, as shown in Tables 4 and 5. When considering hemoglobin as a continuous variable, the OR for 30-day operative mortality decreased to 1.03 for a 1 g/dl decrease in hemoglobin (95% CI: 0.99 to 1.06; p 0.085), and for 30-day perioperative morbidity, the OR was 1.02 (95% CI: 1.00 to 1.04; p 0.043). Kaplan-Meier survival curves for hemoglobin groups throughout the 30-day postoperative period (Fig 3) show increasing risk of mortality as hemoglobin level decreases. A Cox proportional hazards model using (continuous) hemoglobin as the sole covariate yielded a hazard ratio of 1.20 per each g/dl decrease in hemoglobin (95% CI: 1.16 to 1.25; p ). However, when the other risk factors were included and a stepwise, backward selection was performed, preoperative hemoglobin no longer was statistically significant with a hazard ratio of 1.03 per g/dl (95% CI: 0.99 to 1.07; p 0.16). Potential Link Between Hemoglobin and Creatinine As hemoglobin and creatinine were correlated univariately (r 0.2), the relationship between these two variables with operative mortality was examined more closely using interaction models with all combinations of hemoglobin as a binary ( 10 g/dl) and continuous variable, and creatinine as binary ( 1.7 mg/dl), as well as the continuous variable log transformed. Each of the models with only creatinine and hemoglobin showed a significant interaction. For example, using binary creatinine and continuous hemoglobin, the OR for an increase in hemoglobin was 0.84 per g/dl for subjects with normal creatinine but was 0.92 per g/dl for subjects with high creatinine (p for interaction 0.047), implying that higher hemoglobin for someone with high creatinine has less effect on odds of death. However, the difference in these hemoglobin odds ratios is not large, and when all other baseline risk factors were entered and a stepwise, backward selection performed, the interaction became nonsignificant. Comment In this large VA database study of 36,658 CABG-only records for patients operated on with cardiopulmonary bypass, preoperative hemoglobin less than 10 g/dl was a strong univariate predictor of short-term outcomes (unadjusted OR for mortality 2.37, p ); however, inclusion of statistically significant risk factors decreased this effect considerably (adjusted OR 1.29, p 0.064). Previous findings regarding preoperative hemoglobin and postoperative outcomes have been in conflict, even regarding multivariate adjusted associations. Zindrou and associates [2] found a significant adjusted odds ratio for low preoperative hemoglobin and mortality of 3.2 (p 0.016). Others found hemoglobin not to be significant, or found smaller odds ratios such as 1.8 for preoperative anemia [3, 4]. Kulier and associates [5] found significant adjusted associations of low preoperative hemoglobin for noncardiac, but not for cardiac complications Discrepant findings may be due in part to methodologic differences. Zindrou and coworkers [2] defined low hemoglobin as 10 g/dl or less; Magovern and associates [3] defined anemia as 12.5 g/dl or less for males, 11.0 g/dl or less for females, or need for preoperative transfusion; and Weightman and associates [4] analyzed hemoglobin as a continuous variable. Each study included different variables for statistical adjustment, and some studies differed in their availability or treatment of preoperative transfusion information. These differences could explain some of the variation in previous results. There are multiple acute and chronic reasons a patient might be anemic preoperatively, including acute or chronic blood loss, decreased red blood cell production (common in renal disease, cancer, or malnutrition), poor red blood cell maturation, or increased red blood cell destruction [15, 16]. A sudden decline in hemoglobin (eg, from acute bleeding) could result in worsened angina, precipitating a need for cardiac surgery. Al Falluji and coworkers [14] reported that in patients with acute myocardial infarction, anemia did not affect 1-year mortality, but anemic patients were more likely to undergo CABG. The impact on any given patient varies with the patient s general condition and underlying pathology. Young, healthy patients tolerate acutely low hemoglobin levels if volume losses are replaced [16, 17]. Carson and associates [18] found that hemoglobin levels from 8.0 to 10.0 g/dl can usually be tolerated without a change in outcome. In a study of outcomes of surgical patients who received no transfusions due to religious reasons, however, Carson and associates [1] noted that the risk of mortality increased for all patients with a preoperative hemoglobin of less than 11.0 g/dl and continued to increase as hemoglobin levels decreased, with a 30-day mortality rate of 33% in patients with hemoglobin levels less than 6.0 g/dl. This association was stronger for patients with cardiovascular disease where, in patients with hemoglobin less than 11 g/dl, the death rate was 4.3 times higher than for patients without cardiac disease. Cardiovascular patients may be particularly sensitive to anemia, having limited ability to compensate through increased heart rate and stroke volume [14, 16]. Shander and associates [17] concluded that because cardiovascular patients undergoing surgery appear to be less tolerant of anemia, the trigger hemoglobin for transfusion should be higher (approximately 10.0 g/dl) compared with those without cardiovascular disease. Failure of CABG patients to compensate in the face of severe anemia may result in tissue hypoxia, cellular failure, and eventual organ dysfunction and failure. Such deaths would occur over the postoperative course, as noted in Figure 3. In this study, patients with greater cardiac and respiratory comorbidities were found more often in the subgroups with below-normal hemoglobin levels. In the multivariate analysis, this diminished the relative impact of preoperative hemoglobin and the dose-response relationship between hemoglobin less than 9 g/dl and most risk factors. Patients in the moderate-to-severe anemia range have more chronic obstructive pulmonary disease, which Carson and colleagues [1, 18] have shown in-

8 1422 BELL ET AL Ann Thorac Surg PREOPERATIVE HEMOGLOBIN AND OUTCOMES AFTER CABG 2008;86: creased relative risk of mortality and morbidity from 4.3 with anemia and cardiovascular disease to 4.8 with anemia and cardiopulmonary disease. Clinical symptoms of ischemia also appear more prevalent, and this group had the highest intra-aortic balloon pump use. Anemia is frequently seen with congestive heart failure and has been associated with greater disease severity, greater left ventricular mass index, and higher mortality rates [19, 20]. Tanner and colleagues [21] found significantly more patients with anemia in New York Heart Association class III and IV compared with class I and II. Lower hemoglobin levels are also associated with higher plasma brain natriuretic peptide levels, a marker for congestive heart failure [22, 23]. Hence, anemia and cardiac function may be intertwined, each having an effect on the other in a potentially spiraling fashion [23]. Preoperative anemia may possibly be a surrogate for other risk factors. Hemoglobin and creatinine appear to be correlated. When creatinine is not included in the multivariable model, hemoglobin becomes statistically significant (OR 1.41, 95% CI: 1.08 to 1.84; p ). The p value decreased even more when a continuous value for hemoglobin was used. Further, there was some evidence of interaction between creatinine and hemoglobin, with hemoglobin possibly acting as a surrogate metric for renal dysfunction or failure. It appears that a triad of anemia, renal insufficiency, and cardiac disease may coexist as a cardiorenal anemia syndrome [24]. Literature findings support that anemia can be both a cause and an effect. The interrelationships of clinical conditions may lead to a vicious cycle, as each condition may impact the others. Interestingly, whereas patients in the low hemoglobin group ( 9 g/dl) exhibited generally better risk profiles, including creatinine levels (similar to the moderate anemia group), these patients had the highest rate of renal failure complications requiring dialysis, although the association was not strong. When hemoglobin (or anemia) and creatinine are tested within the same model, generally creatinine appears to be more significant than hemoglobin, with larger effects [4, 25, 26]. To evaluate the effect of high creatinine, a sensitivity analysis was performed after eliminating all CICSP records with creatinine greater than 2.0. The sensitivity analysis found results similar to the first analysis, albeit slightly attenuated, adding credence to the impact of hemoglobin in the risk model and to the concept of synergy of hemoglobin with cardiac and renal disease. Although the sample size and data quality are VA CICSP strengths, potential limitations include the primarily (99%) male veteran population exhibiting complex, multiple chronic comorbidities. While comprising only 1.0% of the CICSP population, women are disproportionately represented within the anemic subgroups. Females and patients with a smaller body surface area may become more hemodiluted, possibly resulting in negative outcomes [27]. After the start of the surgical procedure, CICSP data were not captured to assess intraoperative or perioperative hemoglobin levels and to monitor blood product use. In summary, preoperative hemoglobin less than 10 g/dl appears univariately as an independent risk factor for mortality in CABG-only surgical patients. In multivariate analyses, however, preoperative hemoglobin is not significant as an independent risk factor for mortality and is only mildly significant for morbidity. Hemoglobin and creatinine are correlated, with creatinine exhibiting the stronger relationship to mortality. These findings suggest that moderate or severe anemia may possibly serve as a surrogate for renal insufficiency. Given these findings, hemoglobin appears to be intertwined with other risk factors (such as severity of cardiac disease or creatinine). As all risk models have inherent limitations, risk models should be used with caution for predicting outcomes for very complex high risk patients (such as the subgroup of patients presenting with anemia, congestive heart failure, and renal insufficiency). As a prudent quality of care endeavor, this study warrants monitoring an anemic post-cabg patient s status more closely for potential renal problems. Funding for this study was initially provided by the Department of Veterans Affairs Health Services Research and Development Grant IHY (Dr Shroyer, Principal Investigator), with ongoing support from the Office of Patient Care Services, VA Central Office, Washington, DC. This project was supported, in part, by the Offices of Research and Development Offices at the Northport and the Eastern Colorado Health Care System Denver Veterans Affairs Medical Centers. References 1. Carson JL, Duff A, Poses RM, et al. Effect of anemia and cardiovascular disease on surgical mortality and morbidity. Lancet 1996;348: Zindrou D, Taylor K, Bagger J. Pre-operative haemoglobin concentration and mortality rate after coronary artery bypass surgery. Lancet 2002;359: Magovern JA, Sakert T, Magovern GJ, et al. A model that predicts morbidity and mortality after coronary artery bypass graft surgery. J Am Coll Cardiol 1996;28: Weightman WM, Gibbs NM, Sheminant MR, Whitford EG, Mahon BD, Newman MA. Drug therapy before coronary artery surgery: nitrates are independent predictors of mortality and beta-adrenergic blockers predict survival. Anesth Analg 1999;88: Kulier A, Levin J, Moser R, et al. Impact of preoperative anemia on outcome in patients undergoing coronary artery bypass graft surgery. Circulation 2007;116: Grover FL, Johnson R, Shroyer LW, Marshall G, Hammermeister KE. The Veterans Affairs Continuous Improvement in Cardiac Surgery study. Ann Thorac Surg 1994;58: Grover FL, Shroyer ALW, Hammermeister KE. Calculating risk and outcome: the Veterans Affairs database. Ann Thorac Surg 1996;62(Suppl): Davis CL, Pierce JR, Henderson W, et al. Assessment of the reliability of data collected for the Department of Veterans Affairs National Surgical Quality Improvement Program. J Am Coll Surg 2007;204: World Health Organization. Nutritional anaemias: report of a WHO scientific group. World Health Organ Tech Rep Ser 1968;405: SAS system for Windows, version 9. Cary, NC: SAS Institute, 2006.

9 Ann Thorac Surg BELL ET AL 2008;86: PREOPERATIVE HEMOGLOBIN AND OUTCOMES AFTER CABG 11. Stata statistical software, release 9. College Station, TX: StataCorp LP, Royston P, Altman DG. Regression using fractional polynomials of continuous covariates: parsimonious parametric modelling (with discussion). Appl Statist 1994;43: Royston P, Sauerbrei W. A new approach to modelling interactions between treatment and continuous covariates in clinical trials by using fractional polynomials. Statist Med 2006;23: Al Falluji N, Lawrence-Nelson J, Kostis JB, Lacy CR, Ranjan R, Wilson AC, and the Myocardial Infarction Data Acquisition System (MIDAS #8) Study Group. Effect of anemia on 1-year mortality in patients with acute myocardial infarction. Am Heart J 2002;144: Greenburg AG. Pathophysiology of anemia. Am J Med 1996;101(Suppl): Armas-Loughran B, Kalra R, Carson JL. Evaluation and management of anemia and bleeding disorders in surgical patients. Med Clin North Am 2003;87: Shander A, Knight K, Thurer R, Adamson J, Spence R. Prevalence and outcomes of anemia in surgery: a systematic review of the literature. Am J Med 2003;116(Suppl): Carson JL, Duff A, Berlin JA, et al. Perioperative blood transfusion and post-operative mortality. JAMA 1998;279: Anand I, McMurray JJ, Whitmore J, et al. Anemia and its relationship to clinical outcome in heart failure. Circulation 2004;110: Foley RN. Anaemia and the heart: what s new in 2003? Nephrol Dial Transplant 2003;18(Suppl 8): Tanner H, Moschovitis G, Kuster GM, et al. The prevalence of anemia in chronic heart failure. Int J Cardiol 2002;86: Tsuji H, Nishino N, Kimura Y, et al. Haemoglobin level influences plasma brain natriuretic peptide concentration. Acta Cardiol 2004;59: Okonko DO, Anker SD. Anemia in chronic heart failure: pathogenetic mechanisms. J Card Fail 2004;10(Suppl): Silverberg D, Wexler D, Blum M, Wollman Y, Iaina A. The cardio-renal anaemia syndrome: does it exist? Nephrol Dial Transplant 2003;18(Suppl 8): Kurki TS, Jarvinen O, Kataja MJ, Laurikka J, Tarkka M. Performance of three pre-operative risk indices: CABDEAL, Euro- SCORE and Cleveland models in a prospective coronary bypass database. Eur J Cardiothorac Surg 2002;21: Bridgewater B, Neve H, Moat N, Hooper T, Jones M. Predicting operative risk for coronary artery surgery in the United Kingdom: a comparison of various risk prediction algorithms. Heart 1998;79: DeFoe GR, Ross CS, Olmstead EM, et al. Lowest hematocrit on bypass and adverse outcomes associated with coronary artery bypass grafting. Ann Thorac Surg 2001;71: