Preoperative C-reactive Protein Predicts Long-term Mortality and Hospital Length of Stay after Primary, Nonemergent Coronary Artery Bypass Grafting

• ❖ Preoperative C reactive protein (CRP) concentrations more than 10 mg/L are predictive of mortality after cardiac surgery in patients with ongoing ischemia, but whether lower values predict mortality in stable patients is unclear

• ❖ In more than 900 patients without ongoing ischemia undergoing coronary artery bypass grafting only, CRP concentrations as low as 3 mg/l were associated with increased 5-yr mortality and increased hospital length of stay

C-REACTIVE protein (CRP) is a well-recognized marker and a possible mediator of systemic inflammation. Low (< 1 mg/l), moderate (1-3 mg/l), and high (> 3 mg/l) CRP risk categories are currently used to predict cardiovascular events in nonsurgical populations.1These validated CRP risk categories have been shown to predict future cardiovascular events in apparently healthy individuals, as well as recurrent myocardial infarction, and death in patients with acute coronary syndromes.2–5Previous studies in cardiac surgical patients suggest that increased preoperative CRP levels are associated with increased short- and long-term morbidity and mortality.6–8However, these studies have, for the most part, included higher-acuity cardiac surgical patients with profoundly increased preoperative CRP levels undergoing combined valve or coronary artery bypass graft (CABG) surgery. The significance of lower preoperative CRP values for predicting postoperative mortality and morbidity in a lower-acuity cardiac patient population has not been defined. Therefore, the aim of this study was to examine the value of preoperative CRP levels for predicting long-term, all-cause mortality and hospital length of stay (HLOS) in a relatively homogeneous patient sample population undergoing primary, nonemergent CABG-only surgery with cardiopulmonary bypass.

Between August 2001 and June 2007, 1,444 patients aged 20 to 89 yr scheduled for primary CABG-only surgery with cardiopulmonary bypass at the Brigham and Women's Hospital, Boston, Massachusetts, and the Texas Heart Institute, St Luke's Episcopal Hospital, Houston, Texas, were enrolled under the auspices of the parent study known as CABG Genomics.∥CABG Genomics is an institutional review board-approved, prospective longitudinal study examining the association between genetic variation and adverse outcomes after CABG surgery. Exclusion criteria for enrollment into the parent study were preoperative hematocrit less than 25% and a history of leukocyte-rich blood product transfusion ≤ 30 days before surgery. After informed consent, preoperative demographic characteristics, environmental risk factors and comorbidities, operative surgical and anesthetic management, and in-hospital events were recorded for all patients. After hospital discharge, the annual incidence of all-cause death was discovered through mailed questionnaires or examination of the Social Security Index for those patients lost to follow-up.

To address the specific aim of this study, we further restricted our analysis to lower-acuity CABG-only patients by excluding patients with increased preoperative cardiac troponin I (cTnI) levels (> 0.03 μg/l, n = 421), evidence of a preoperative infection defined as fever, the need for preoperative antibiotics, or positive blood cultures 7 days before surgery (n = 40); patients having concomitant valve surgery (n = 41), emergent surgery (n = 1), off-pump surgery (n = 32), or reoperative cardiac surgery (n = 1); and patients with missing preoperative CRP values (n = 8).

CRP and cTnI levels were measured from whole blood samples drawn immediately before surgery and on postoperative days (PODs) 1-5. After collection of whole blood, aliquoted citrated plasma samples were stored in liquid-phase nitrogen. CRP and cTnI were calculated using a sandwich immunoassay on Triage® platform (Biosite, Inc., San Diego, CA) by a single blinded core facility. CRP concentrations were calculated using a set of spiked plasma samples as the calibration reference. The analytical sensitivity of the CRP assay was 0.33 mg/l, with a precision of more than 17% the measurable range of the assay (0.33-300 mg/l).

Patients were categorized into the following CRP categories: less than 1, 1-3, 3-10, and more than 10 mg/l based on preoperative CRP levels. The primary outcomes were all-cause mortality occurring up to 7 yr after surgery, and HLOS defined as the number of days spent in the hospital after the operation, with the operative day and last day of hospitalization counted as whole days.

Separate Cox proportional hazards regression models were used to estimate adjusted hazards ratios (HR) for the association between preoperative CRP risk categories and all-cause mortality and HLOS. A separate list of clinical risk factors for all-cause postoperative mortality and HLOS was initially generated through a theory-driven selection from more than 1,500 data points used to phenotype patients enrolled in CABG Genomics. We subsequently built separate proportional hazards models using a data-driven stepwise regression approach by including variables with P ≤ 0.20 on univariate analysis. Demographic variables, including age, gender, race, and institution, were forced into the final models. Because HMG (3-hydroxy-3-methyl-glutaryl)-CoA reductase inhibitor (statin) therapy has been shown to decrease cardiovascular events and mortality in cardiac surgical patients,9,10preoperative statin therapy was also forced into the final models. In addition, increased POD1 cTnI, which has been associated with increased postoperative mortality,11was forced into the final models (see figure legends for a final list of risk factors for each model). The proportional hazards assumption was tested by including interaction terms between preoperative CRP levels and the logarithm of follow-up time in regression models.12The interaction terms were not significant, suggesting that the assumption of proportionality was valid.12Mortality was censored at 7 yr, and HLOS was censored at 30 days. Patients who died before hospital discharge were censored at 30 days or at day of death for the HLOS analysis. Kaplan-Meier survival curves were generated for each preoperative CRP risk category.

Medians with interquartile range are used to describe nonnormally distributed data; otherwise, means and SD are presented. Wilcoxon rank-sum and Fisher's exact tests were used to compare continuous and categorical variables, respectively. A Kruskal-Wallis test was used to compare continuous variables against categorical variables with more than 2 levels. HR ± 95% confidence intervals (CI) was used to describe the effect of preoperative CRP risk categories on outcomes. P ≤ 0.05 was considered statistically significant. JMP 7.0.2 and SAS 9.1.3 (SAS Institute, Cary, NC) were used for statistical analysis.

After exclusions, 916 patients were analyzed: 365 patients in the less than 1 mg/l group, 315 patients in the 1-3 mg/l group, 138 patients in the 3-10 mg/l group, and 98 patients in the more than 10 mg/l group. Baseline demographic and clinical characteristics for each CRP category are listed in table 1. Patients in the higher CRP categories were more likely to be women, current smokers, have a higher body mass index, have a higher incidence of self-reported history of myocardial infarction less than 2 weeks before surgery, and have longer cardiopulmonary bypass times. These patients were less likely to have hypercholesterolemia or to be receiving preoperative aspirin.

Preoperative cTnI levels were less than 0.03 μg/l in all patients analyzed. POD1 cTnI levels were significantly increased in each CRP category compared with respective preoperative cTnI levels (median cTnI levels [interquartile range]: < 1 mg/l group, 1.17 [0.60-2.36]μg/l; 1-3 mg/l group, 1.02 [0.58-2.17]μg/l; 3-10 mg/l group, 1.20 [0.65-2.08]μg/l; and > 10 mg/l group, 1.14 [0.67-2.76]μg/l; P < 0.0001). However, POD1 cTnI levels did not differ significantly between CRP categories (P = 0.36). Perioperative CRP levels stratified by CRP category are shown in figure 1.

Eighty-seven patients (9.5%) died during a mean follow-up period of 4.8 ± 1.5 yr. There was a significant decrease in survival with increasing preoperative CRP category (P = 0.002; fig. 2). The unadjusted HRs (95% CI) for patients in the 1-3, 3-10, and more than 10 mg/l preoperative CRP groups were 1.86 (1.01-3.53), P = 0.05; 2.89 (1.49-5.70), P = 0.002, and 3.34 (1.61-6.86), P = 0.002, respectively, compared with the less than 1 mg/l CRP category as the reference (fig. 3). After adjusting for clinically relevant demographic variables and perioperative risk factors, including preoperative statin use and POD1 cTnI levels, the HRs (95% CI) for patients in the 1-3, 3-10, and more than 10 mg/l preoperative CRP groups were 1.43 (0.75-2.83), P = 0.28; 2.50 (1.22-5.16), P = 0.01; and 2.66 (1.21-5.80), P = 0.02, respectively, compared with the less than 1 mg/l CRP category as the reference. POD1 cTnI levels (P = 0.20) and preoperative statin therapy (P = 0.10) were not independently associated with all-cause mortality after surgery.

Nine hundred four patients (98.2%) were discharged from the hospital within 30 days of surgery. Three patients died before hospital discharge. Patients with preoperative CRP levels in the less than 1 and 1-3 mg/l range stayed a median of 7 days in the hospital, whereas patients with preoperative CRP levels in the 3-10 and more than 10 mg/l range stayed a median of 8 days (P < 0.001). Although the clinical relevance of one additional POD in the hospital may be debatable, the costs incurred nonetheless significant. The unadjusted HRs (95% CI) for patients in the 1-3, 3-10, and more than 10 mg/l preoperative CRP groups were 0.95 (0.82-1.11), P = 0.55; 1.35 (1.11-1.65), P = 0.0.003; and 1.26 (1.01-1.58), P = 0.04, respectively, compared with the less than 1 mg/l CRP category as the reference (fig. 4). The adjusted HRs (95% CI) for patients in the 1-3, 3-10, and more than 10 mg/l preoperative CRP groups were 0.95 (0.81-1.11), P = 0.20; 1.32 (1.07-1.63), P < 0.001; and 1.27 (1.02-1.62), P = 0.001, respectively, compared with the less than 1 mg/l CRP category. After adjustment, POD1 cTnI levels were independently associated with extended HLOS (HR [95% CI]: 2.70 [1.38-5.88], P = 0.003).

Preoperative CRP levels more than 10 mg/l have been previously associated with increased long-term mortality and extended HLOS in patients undergoing primary CABG-only surgery with cardiopulmonary bypass. We now extend this previous work by demonstrating for the first time that preoperative CRP levels ranging as low as 3-10 mg/l are also independent predictors of increased long-term mortality and extended HLOS, even in patients without evidence of ongoing preoperative myocardial injury. These important findings may allow for more objective risk stratification of patients who present for uncomplicated surgical coronary revascularization.

Increased CRP levels have emerged as a strong independent predictor of cardiovascular disease in apparently healthy individuals.1,13In patients with acute coronary syndromes, CRP predictors increased morbidity and mortality independent of increased troponin levels.14,15Several studies have reported increased short- and long-term morbidity and mortality in cardiac surgical patients with increased preoperative CRP levels. Palmerini et al.  8showed decreased survival at 9 months of follow-up after CABG surgery in relatively high-risk patients with unprotected left main coronary artery disease and preoperative CRP levels ≥ 12.2 mg/l. Cappabianca et al.  6showed increased in-hospital and 3-yr all-cause mortality in a heterogeneous patient population with preoperative CRP levels ≥ 5 mg/l undergoing cardiac surgery that included, but was not limited to, CABG surgery. Kangasniemi et al.  7showed significantly increased 12-yr mortality in patients with preoperative CRP levels ≥ 10 mg/l undergoing isolated on-pump CABG surgery. However, these authors were unable to comment specifically on the effects of preoperative CRP levels less than 10 mg/l because of limitations of the CRP assay used. Although these studies add to the growing body of evidence suggesting that increased preoperative CRP levels predict increased postoperative mortality, they are not generalizable to lower-acuity cardiac surgical patients with lower-range preoperative CRP levels less than 10 mg/l undergoing CABG-only surgery. To address this gap in knowledge, we not only corroborate the findings of other by demonstrating an association between preoperative CRP levels more than 10 mg/l and increased postoperative mortality but also show that preoperative CRP levels as low as 3 mg/l are associated with decreased survival and extended HLOS in lower-acuity cardiac surgical patients free of ongoing myocardial ischemia or infarction at the time of nonemergent surgical revascularization.

We did not observe an effect of preoperative statin therapy on long-term mortality after CABG surgery. Although no study to date has examined the effects of preoperative statin therapy on long-term survival after CABG surgery, there are a number of studies that suggest a beneficial association with improved survival up to 30 days postoperatively.9,10,16,17In this study, more than 75% of the patients were treated preoperatively with a statin. Because of the small number of patients not on preoperative statin therapy and the relatively low mortality rate in this study, we may have been underpowered to detect a difference in long-term survival. Furthermore, in light of emerging evidence suggesting beneficial effects on short- and long-term cardiovascular morbidity and mortality in both surgical and nonsurgical cohorts,16,18any inference of discontinuing perioperative statin therapy should not be made based on our findings.

Emerging evidence from the nonsurgical population suggests that even slight increase in circulating CRP levels correlate with poorer cardiovascular outcomes. Similarly, our findings in the cardiac surgical population suggest that preoperative CRP levels well below levels considered increased in previous studies also carries a significant risk of increased long-term mortality and HLOS. However, because we only enrolled patients scheduled to undergo CABG surgery, we are unable to directly compare the predictive value of CRP between surgical and nonsurgical cohorts with regards to outcome. Despite minimal variability of basal CRP levels over long periods of time,19,20it is conceivable that surgical coronary revascularization might alter a patients' risk profile for death by altering postoperative baseline CRP levels. These questions can only be addressed through a prospectively designed study whereby CRP levels are measured at the time of long-term follow-up.

Our findings do not support or refute the important question of whether or not CRP is a proinflammatory mediator of postoperative mortality, or merely a marker. A number of studies implicate CRP as an important mediator in the generation of atheromatous coronary plaque, including uptake of low-density lipoprotein by macrophages,21triggering increased expression of endothelial cell surface adhesion molecules,22and activating complement system proteins.23However, more recent studies have brought into question the biologic role of CRP as a proinflammatory pattern recognition molecule by suggesting that the in vitro  effects of CRP noted above can be ascribed to bacterial by-product or azide contamination of commercially available CRP preparations not subsequently seen when highly purified CRP is used.24,25Moreover, although there is strong evidence that putative CRP gene single nucleotide polymorphisms influence systemic CRP levels, an independent association between these CRP single nucleotide polymorphisms and adverse cardiovascular events has not been definitively established.26–28Confirmation of a direct causal relationship between CRP and cardiovascular outcomes will require further investigation.

In this study, we show that increased preoperative CRP levels predict long-term, all-cause mortality and extended HLOS in a cohort of lower-acuity CABG-only surgical patients. These findings may allow for more objective risk stratification of patients who present for nonemergent, surgical coronary revascularization. Further delineation of the proinflammatory mechanisms of CRP will establish CRP as a potentially modifiable risk predictor in both surgical and nonsurgical patients.

The authors acknowledge the outstanding contributory efforts of James Gosnell, R.N. (Research Coordinator, Department of Anesthesiology, Perioperative and Pain Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts); Kujtim Bodinaku, M.D., Jai Madan, M.D., M.P.H., and Svetlana Gorbatov, M.P.H. (Research Assistants, Department of Anesthesiology, Perioperative and Pain Medicine, Brigham and Women's Hospital, Harvard Medical School); Juliette Dean, R.N., James Chen, R.N., and Jacques Estephan, R.N. (Research Assistants, Department of Anesthesia, Texas Heart Institute, St. Luke's Episcopal Hospital, Baylor College of Medicine, Houston, Texas); and Isabella Canderlaria, B.S. (Research Coordinator, Department of Anesthesia, Texas Heart Institute, St. Luke's Episcopal Hospital, Baylor College of Medicine).