Surgical risk stratification models, 1 including the

Transcription

1 Troponin T and B-Type Natriuretic Peptide After On-Pump Cardiac Surgery Prognostic Impact on 12-Month Mortality and Major Cardiac Events After Adjustment for Postoperative Complications Giovanna A.L. Lurati Buse, MD, MSc; Daniel Bolliger, MD; Esther Seeberger, RN; Jorge Kasper, MD; Martin Grapow, MD; Michael T. Koller, MD; Manfred D. Seeberger, MD; Miodrag Filipovic, MD Background The independent predictive value of troponin T (TNT) after on-pump cardiac surgery was established in several studies. However, adjustment was limited to preoperative risk factors without considering perioperative complications. Data on the prognostic value of postoperative B-type natriuretic peptide (BNP) are scarce. Our aim was to assess the independent value of TNT and BNP to predict 12-month outcome after cardiac surgery with adjustment for preoperative risk estimates and postoperative complications and to report risk stratification gains when considering the European System for Cardiac Operative Risk Evaluation (EuroSCORE) combined with postoperative biomarkers. Methods and Results This prospective cohort study included consecutive patients undergoing on-pump cardiac surgery between 2007 and We evaluated postoperative TNT and BNP, the EuroSCORE, and postoperative complications as predictors of adverse events using Cox regression. The primary end point was death or major adverse cardiac events within 1 year after surgery. We calculated the net reclassification index of TNT and BNP in addition to the EuroSCORE. We enrolled 1559 patients, of whom 176 (11.3%) experienced an event. The adjusted hazard ratio of TNT >0.8 μg/l was 2.13 (95% confidence interval, ) and of BNP >790 ng/l was 2.44 (95% confidence interval, ). The net reclassification index of the addition of TNT and BNP to the EuroSCORE was (95% confidence interval, ). Conclusions Postoperative TNT and BNP are strong predictors of 1-year events after on-pump cardiac surgery independent of preoperative risk factors and postoperative complications. Updating the preoperative EuroSCORE risk with postoperative TNT and BNP after surgery allows for improved prediction of 1-year death or major adverse cardiac events. (Circulation. 2014;130: ) Key Words: cardiopulmonary bypass natriuretic peptides prognosis troponin T Surgical risk stratification models, 1 including the European System for Cardiac Operative Risk Evaluation (EuroSCORE), 2,3 exclusively consider preoperative and intraoperative risk factors. Although useful for healthcare planning and for preoperative risk/benefit assessments, this view may prove fragmentary postoperatively, because it disregards perioperative myocardial ischemic events 4,5 or postoperative complications 6 8 despite their association with mortality. Therefore, there is a need for tools that allow for early postoperative risk reassessment to guide management decisions (eg, transfer to step-down units and postdischarge follow-up schedules). Editorial see p 932 Clinical Perspective on p 957 Troponin T (TNT) has an established prognostic value for short-term 9 11 and midterm outcomes after cardiac surgery. However, previous studies on the predictive value of troponin 9 11 have focused on preoperative risk factors and have not accounted for postoperative complications (eg, infections or acute kidney failure). In-hospital complications, however, may exert a relevant prognostic change by being associated with both elevated cardiac biomarkers postoperatively and midterm outcome. Several studies have described the association between preoperative natriuretic peptides and outcomes after cardiac surgery. 15,16 Evidence supporting the prognostic value of postoperative concentrations, in contrast, is scarce and has been generated mostly in small studies 17,21,22 that are underpowered Received November 1, 2013; accepted June 16, From the Department for Anesthesia, Surgical Intensive Care, Prehospital Emergency Medicine and Pain Therapy (G.A.L.L.B., D.B., E.S., J.K., M.D.S., M.F.) and Cardiac Surgery Department (M.G.), University Hospital of Basel, Basel, Switzerland; Basel Institute for Clinical Epidemiology and Biostatistics, University of Basel, Basel, Switzerland (M.T.K.); and Anesthesiology Department, Kantonsspital St Gallen, St Gallen, Switzerland (M.F.). The online-only Data Supplement is available with this article at Correspondence to Giovanna A.L. Lurati Buse, MD, MSc, Anesthesiology, University Hospital Basel, CH-4031 Basel, Switzerland. giovanna.lurati@usb.ch 2014 American Heart Association, Inc. Circulation is available at DOI: /CIRCULATIONAHA

2 Lurati Buse et al Biomarkers and Outcome After Cardiac Surgery 949 for multiple adjustment. Moreover, the prognostic gain of a risk stratification on the basis of the EuroSCORE alone versus a risk restratification based on the EuroSCORE with the addition of early postoperative cardiac biomarker concentrations (reclassification) is insufficiently known. The aim of this cohort study was, first, to test the hypothesis that postoperative TNT and B-type natriuretic peptide (BNP) concentrations are predictive of 1-year outcome in adults undergoing on-pump cardiac surgery independent of preoperative risk factors and postoperative complications and, second, to estimate reclassification of extended models that include postoperative TNT and BNP information in addition to the EuroSCORE, 2,3,23 an established and clinically widely used preoperative risk stratification tool. Methods Study Design and Participants We conducted a single-center prospective cohort of consecutive adult patients who underwent on-pump cardiac surgery (ie, including isolated coronary artery bypass graft [CABG], single and multiple valvular procedures, and combined CABG and valvular surgery) at the University Hospital of Basel between January 2007 and January Patients were included independent of preoperative TNT. We excluded patients from the current analysis if they underwent offpump surgery or procedures requiring deep hypothermic circulatory arrest. The study was approved by the local ethics board, and all of the participants provided written informed consent. All of the perioperative and intensive care management, including surgical techniques, choice and dosage of catecholamines for cardiopulmonary bypass weaning and in the intensive care unit, coagulation management, blood transfusion triggers, duration of intensive care unit stay, and investigations for and treatment of postoperative complications were at the discretion of the attending clinicians. Trained research personnel prospectively collected detailed data on perioperative patient treatment, surgical techniques, and in-hospital course and conducted follow-up. In-hospital complications were not adjudicated but were extracted from the hospital charts as diagnosed by the attending physicians with the exception of acute kidney injury classification. 24 Its detection is based on the highest postoperative creatinine concentration measured during hospitalization independent of its mention in the hospital charts. For the categories of risk, injury, and failure, 24 the acute kidney injury classification did not require any minimal duration for creatinine value elevations above the various cutoffs. Cardiac Marker Analyses TNT (Elecsys, Roche Diagnostics, F Hoffmann-La Roche Ltd, Basel, Switzerland) and BNP concentrations (ECLIA, Roche Diagnostics, F Hoffmann-La Roche Ltd) were measured routinely at 6:00 am on the first and second postoperative days in all of the patients. Analyses were performed by the central hospital laboratory, which is subjected to practice-certified regulatory quality controls. The 99 th percentile and 10% coefficient of variation of TNT were <0.01 μg/l and μg/l, respectively. The 95 th percentile and 6.7% coefficient of variation of BNP were 135 ng/l and 180 ng/l, respectively. Clinical End Point Definitions and End Point Ascertainment The primary end point was predefined as the composite of death from any cause or the occurrence of nonfatal major adverse cardiac events (MACEs) within 1 year after surgery. MACEs included myocardial infarction, cardiac arrest, need for subsequent surgical or percutaneous coronary intervention, and congestive heart failure requiring hospitalization. Follow-up data were obtained by a questionnaire mailed to all of the patients 1 year after surgery. Patients or next of kin were asked to report the occurrence and date of any hospitalization within the first year after surgery or the vital status of the patient. If the questionnaire was not returned, research personnel contacted the patient by telephone. If unable to contact a patient, we asked his or her general practitioner to provide the health or vital status information. A trained research nurse blinded to the patient data conducted the interviews. If the questionnaire or the interview indicated any hospitalization, the research team contacted the hospital or the patient s general practitioner to obtain appropriate documentation. On the basis of this information, the occurrence of an end point was determined by 2 independent under assumption of adjudicators blinded to in-hospital TNT and BNP data. Statistical Analysis Continuous data are presented as mean with SD or as median with the first to third quartiles (Q1 Q3) as appropriate. Categoric data are presented as absolute numbers and percentages. Baseline differences between patients with events compared with patients without events were calculated with the Student t, Mann-Whitney U, or Pearson χ 2 tests as appropriate. All of the analyses were based on the higher TNT and BNP values measured at 6:00 am on postoperative days 1 and 2. We explored the association between increasing concentrations of TNT and BNP for the composite outcome by calculating the interval likelihood ratio over 10 percentile groups and plotted it. We calculated the association between cardiac biomarkers and outcome at 1 year in a Cox regression model with forced entry. The dependent variable was the time to the composite end point. We modeled cardiac biomarkers linearly, nonlinearly, and dichotomized. For dichotomization, we defined the optimal cutoff using receiver operating characteristic curves of each biomarker for the composite outcome at 1 year under assumption of equal weight for sensitivity and specificity. We predefined the logistic EuroSCORE 23 (continuous percentage increase), sepsis, sternal infection (without sepsis), respiratory infections (pneumonia, ventilator associated pneumonia, or purulent tracheobronchitis without sepsis), and acute kidney injury classification 24 as additional independent variables. The explanatory variables were either acute events (eg, postoperative sepsis) or apply to a single point in time (logistic EuroSCORE). Therefore, we opted to model them as static variables rather than as time-dependent variables. To avoid model overfitting, we decided to limit the number of independent variables. The selection of postoperative variables to include into the model was guided by the following principles: (1) the concept that early complications affecting the cardiovascular system (eg, postoperative cardiogenic or hemorrhagic shock) are reflected by TNT and BNP concentrations); (2) the prevalence of the risk factor; and (3) the prognostic challenge. The decision not to include postoperative extracorporeal membrane oxygenation therapy (ECMO) dependency exemplifies this approach. First, hemodynamic instability requiring ECMO results in elevated TNT and BNP concentrations (ie, ECMO is not as much a confounder but rather a link of the pathophysiologic chain leading to elevated cardiac biomarkers). Second, the requirement for ECMO was exceedingly rare in our cohort. Third, ECMO dependency is such a strong indicator of poor outcome that clinicians hardly need to rely on cardiac biomarkers to restratify such patients. We assessed the models using the Akaike information criterion, the Harrell C-index, total model χ 2 (Wald test), and adjusted R 2. We assessed calibration of the final model by plotting the observed and the predicted composite events. We evaluated collinearity using the variance inflation factor. We internally validated the model by bootstrapping with 1000 samples. We generated 4 risk stratification strata ( 5%, >5% to 10%, >10% to 15%, and >15%) according to the EuroSCORE, the full model, and a model including the EuroSCORE and early cardiac biomarker information. We then assessed the net reclassification index (NRI) for EuroSCORE versus the full model and for EuroSCORE versus EuroSCORE and TNT plus BNP. We calculated the NRI in patients with events, in those without events, and in the overall NRI. 25 A secondary Cox regression model with the same independent variables was fitted for 30-day mortality and MACEs.

3 950 Circulation September 16, 2014 We generated a third model by including the variables added recently to the EuroSCORE II 26 (ie, angina pectoris Canadian Cardiovascular Society grade 4, New York Heart Association class, urgency of the procedure in 3 categories [urgent, emergent, and salvage], insulindependent diabetes mellitus, creatinine clearance, and the new procedure scores) as independent variables in our model. The focus of this analysis was any on-pump cardiac surgery procedures; however, we also analyzed the distribution of TNT and BNP concentrations following specific procedures. Furthermore, we explored the potential interaction in the association between cardiac biomarkers and outcome by the type of procedure. We opted to test for interaction within the subgroups isolated CABG versus other procedures to maintain power for the interaction analysis rather than to define subgroups by specific types of non-cabg procedures. Statics were calculated by SPSS 20 (IBM SPSS Statistics, Armonk, NY), R , and by Microsoft Excel 2010 (Microsoft, Redmond, WA). We considered the null hypothesis refuted if the 2-sided P value was <0.05. Sample Size For the 36-month enrollment period, we anticipated a sample size of 1500 patients. Assuming a 1-year event rate of 12%, 14 we expected up to 180 events, allowing for a statistically robust multivariate model based on up to 18 variables. 27 Results Descriptive Analysis Of the 1713 consecutive patients who underwent cardiac surgery in our institution between 2007 and 2010, 1569 fulfilled the eligibility criteria. In 10 patients (0.6%), no TNT values were available, of whom 8 (0.5%) had died before the first sampling time (6:00 am on postoperative day 1; Figure 1). Follow-up was completed in 1545 patients (99.1%); the remaining 14 patients were censored at last contact date. A total of 115 patients (7.3%) underwent cardiac surgery within 7 days of an acute coronary syndrome with detectable preoperative TNT concentrations. Their median preoperative TNT was 0.28 μg/l (Q1 Q3, μg/l). BNP concentrations were available in 1364 patients (86.9%). In patients missing postoperative day 1 or day 2 samples, we assumed that the available concentration represented the higher concentration. Within the 12-month postoperative follow-up, 176 patients (11.3%) experienced the composite end point event, including 103 deaths (6.6%). We recorded 86 MACEs (5.5%), including 17 acute coronary syndromes (1.1%), 14 cardiac arrests (0.9%), 20 percutaneous coronary interventions (1.3%), 6 subsequent CABGs (0.4%), and 29 acute congestive heart failures (1.9%) requiring hospitalization. Eighty-three events (5.3%) occurred within 30 days of surgery, of which there were 58 deaths (3.7%). Compared with patients with uneventful follow-up, those who experienced an event had a higher EuroSCORE, higher TNT, higher BNP, and had presented with more postoperative complications (Table 1 and Figure 2). Troponin and BNP were higher after procedures other than isolated CABG (Table 2). Association Between Postoperative TNT and BNP With 1-Year Events and Reclassification Figure 3 plots the likelihood ratio of TNT and BNP for the 1-year composite event by increasing concentrations (groups defined by 10 percentiles). The area under the receiver operating characteristic curve of TNT was (95% confidence interval [CI], ) with an optimal cutoff at 0.8 μg/l. The corresponding values for BNP were (95% CI, ) and 790 ng/l. The median duration from surgery to event was 34 days (Q1 Q3, ) in the whole population. In patients with elevated cardiac biomarkers, the Figure 1. Flow chart of patient inclusion. CABG indicates coronary artery bypass graft; CPB, cardiopulmonary bypass; DHCA, deep hypothermic circulatory arrest; and MACE, major adverse cardiac events.

4 Lurati Buse et al Biomarkers and Outcome After Cardiac Surgery 951 Table 1. Patient Characteristics and Surgical and Postoperative Data All Patients (n=1559) No Events (n=1383) Events (n=176) P Value Preoperative data Age, mean (SD), y 67 (10) 67 (11) 71 (10) 0.64 Men 1151 (73.8) 1031 (74.5) 120 (68.2) 0.07 Recent ACS ( 30 d) 234 (15.0) 199 (14.4) 35 (19.9) Left main stem stenosis >50% 345 (22.2) 291 (21.1) 54 (30.7) 0.04 Acute endocarditis 32 (2.1) 28 (2.0) 4 (2.3) Pulmonary hypertension 60 mm Hg 86 (5.5) 67 (4.9) 19 (10.8) Preoperative EF 60 (49 65) 60 (50 65) 54 (40 60) Arterial hypertension 1092 (70.1) 965 (69.9) 127 (72.2) Diabetes mellitus (insulin or OAD) 359 (23.1) 306 (22.2) 53 (30.1) BMI, mean (SD), kg/m (10.8) 27.6 (9.1) 32.7 (27.1) History of TIA/CVI 137 (8.8) 122 (8.8) 15 (8.5) History of PVD 195 (12.5) 157 (11.4) 38 (21.6) <0.001 Preoperative GRF, ml/min 82 (61 106) 83 (63 107) 71 (58 93) COPD under treatment 87 (5.6) 72 (5.2) 15 (8.6) Logistic EuroSCORE 2.25 ( ) 2.14 ( ) 3.68 ( ) <0.001 Additive EuroSCORE 5 (3 7) 5 (3 7) 7 (4 9) <0.001 Procedural data Isolated CABG 778 (50.0) 695 (50.3) 83 (47.2) Isolated aortic valve replacement 196 (12.6) 187 (13.5) 9 (5.1) CABG and aortic valve surgery 176 (11.3) 153 (11.1) 23 (13.1) CABG and 2-valve procedures 34 (2.2) 31 (2.3) 3 (1.8) Emergent surgery 46 (3.0) 37 (2.7) 9 (5.1) Postoperative data Troponin T POD 1, μg/l 0.45 ( ) 0.42 ( ) 0.87 ( ) <0.001 Troponin T POD 2, μg/l 0.40 ( ) 0.37 ( ) 0.83 ( ) <0.001 BNP POD 1, ng/l 433 ( ) 417 ( ) 870 ( ) <0.001 BNP POD 2, ng/l 498 ( ) 470 ( ) 892 ( ) <0.001 ECMO 6 (0.40) 1 (0.07) 5 (2.80) <0.001 IABP 55 (3.5) 28 (2.0) 27 (15.3) <0.001 Postoperative acute kidney injury/failure 141 (9.0) 89 (6.5) 52 (29.5) <0.001 In-hospital sepsis 33 (2.1) 15 (1.1) 18 (10.2) <0.001 In-hospital respiratory infection* 73 (4.7) 43 (3.1) 30 (17.0) <0.001 Sternal infection 49 (3.1) 34 (2.5) 15 (8.5) <0.001 ICU LOS, d 2 (2 3) 2 (2 2) 4 (2 7) <0.001 Hospital LOS, d 11.0 ( ) 11.0 ( ) 15.5 ( ) <0.001 Data are number (%) unless stated otherwise; continuous variables are reported as median (quartile 1 to quartile 3) unless stated otherwise. ACS indicates acute coronary syndrome; BMI, body mass index; BNP, B-type natriuretic peptide; CABG, coronary artery bypass graft; COPD, chronic obstructive pulmonary disease; CVI, cerebrovascular insult; ECMO, extracorporal membrane oxygenation; EF, ejection fraction; GFR, glomerular filtration rate (Cockcroft-Gault estimate); IABP, intra-aortic balloon pump; ICU, intensive care unit; LOS, length of stay; OAD, oral antidiabetics; POD, postoperative day; PVD, peripheral vascular disease; Q1 Q3, first to third quartile; and TIA, transient ischemic attack. *Respiratory infections included pneumonia, ventilator associated pneumonia, or purulent tracheobronchitis. composite event occurred after a median of 22 days (Q1 Q3, 6 90 days). Patients without elevated cardiac markers experienced their event after a median of 87 days (Q1 Q3, days; P<0.001). Table 3 reports model information. Concentrations of both TNT and BNP above the cutoff within the first 2 postoperative days were independently associated with the composite end point at 1 year after on-pump cardiac surgery after adjustment by the EuroSCORE, postoperative infectious complications, and acute kidney injury (Table 4). Variance inflation factors were not suggestive of collinearity (variance inflation factor, <2 for all variables). Figure 4 shows the calibration results at various risk levels. The overall NRI for 1-year events with the addition of TNT and BNP information to the EuroSCORE versus the EuroSCORE alone was (95% CI, ). The addition of biomarkers performed well for detecting increased risk (NRI, [95% CI, ]) but poorly for

5 952 Circulation September 16, 2014 Figure 2. Kaplan Meier curves and survival table stratified by troponin T (TNT; A), B-type natriuretic peptide (BNP; B), and postoperative complications (C).

6 Lurati Buse et al Biomarkers and Outcome After Cardiac Surgery 953 Table 2. Troponin T and BNP Concentrations by Type of Procedure Isolated CABG Other Procedures Difference of the Median (95%CI) P Value Higher TNT POD 1 and 2 at 6:00 am, μg/l 0.38 ( ) 0.55 ( ) 0.17 ( ) <0.001 TNT POD 1, μg/l 0.37 ( ) 0.54 ( ) 0.17 ( ) <0.001 TNT POD 2, μg/l 0.31 ( ) 0.46 ( ) 0.16 ( ) <0.001 Higher BNP POD 1 and 2 at 6:00 am, ng/l 400 ( ) 532 ( ) 132 (75 189) <0.001 BNP POD 1, ng/l 369 ( ) 493 ( ) 124 (79 169) <0.001 BNP POD 2, ng/l 434 ( ) 541 ( ) 107 (38 176) <0.001 Data are median (quartile 1 to quartile 3). P values were calculated using Mann Whitney U test. BNP indicates B-type natriuretic peptide; CABG, coronary artery bypass grafting; CI, confidence interval; POD, postoperative day; and TNT, troponin T. decreased risk (NRI, [95% CI, to 0.393]). The same pattern was present for the final model. The overall NRI of the final model was (95% CI, ) with an NRI for higher risk amounting to (95% CI, ) and to (95% CI, to 0.283) for decreased risk (Tables I IV in the online-only Data Supplement). The models fitted to predict 30-day events, and, with the addition of EuroSCORE II, the variables showed similar results (Tables V and VI in the online-only Data Supplement). We did not find any evidence of significant interaction in the association between TNT (P=0.153) and BNP (P=0.785) or outcome by the type of procedure (isolated CABG versus other procedures). Discussion Main Results This cohort study showed that the association between an increase in TNT concentration within the first 2 postoperative days and 1-year all-cause mortality and MACEs after cardiac surgery persisted after adjustment considering both preoperative and postoperative confounders. Furthermore, we detected a significant independent association between postoperative BNP concentrations and 1-year mortality and MACEs. Extending the EuroSCORE to include postoperative TNT and BNP information allowed for improved risk stratification in every fourth patient. The time window of 1 month between events and elevated biomarkers suggests that measurement of cardiac biomarkers flags high-risk patients early (ie, at a time point when management optimization holds promise for outcome improvement). Preoperative cardiac surgery risk stratification models, 1 including the EuroSCORE, 2,3 that exclusively consider preoperative and intraoperative risk factors are the most useful for healthcare planning and for preoperative risk/benefit assessments. However, there is a need for reevaluation and risk reassessment after the surgical procedure itself. The data generated in this cohort study suggest that postoperative TNT and BNP concentrations within the first 2 postoperative days are helpful tools for risk reassessment after cardiac surgery. They are noninvasive, easily obtained, and allow for early flagging of patients at risk for death and MACEs within the next (few) month(s). This observational study did not address the question of how to optimize the management of patients Figure 3. Likelihood ratio of the 1-year composite by increasing troponin T (TNT) and B-type natriuretic peptide (BNP) concentrations.

7 954 Circulation September 16, 2014 Table 3. Cox Regression Model Information Model AIC C Index Total Model χ 2 (Wald Test) Adjusted R 2 Univariate models Peak TNT, linear Peak TNT, nonlinear Peak TNT, binary (0.8) BNP, linear BNP, nonlinear BNP, binary (790) Models with BNP and TNT BNP binary, TNT binary BNP nonlinear, TNT nonlinear (original) (corrected, bootstrap samples) Multivariable models EuroSCORE linear, postoperative complications, TNT binary, BNP binary (final model) (original) (corrected, 150 bootstrap samples) EuroSCORE linear, postoperative complications, TNT nonlinear, BNP nonlinear (original) (corrected, 150 bootstrap samples) AIC indicates Akaike information criterion; BNP, B-type natriuretic peptide; EuroSCORE, European System for Cardiac Operative Risk Evaluation; and TNT, troponin T. with increased risk after cardiac surgery. Potential approaches might include more stringent in-hospital monitoring in stepdown units or systematic telemetry monitoring on the regular ward, dedicated postcardiac surgery outpatients clinics, or virtual wards. Comparison With Previous Studies Several studies evaluated the association between troponin elevations after cardiac surgery and short- 9,10 and long-term adverse events Adabag et al 9 reported on a concentrationresponse relationship between troponin I release and shortterm mortality (odds ratio [OR], 1.3 [95% CI, ] per 0.05-μg/L increase in troponin I levels in valvular patients and 1.4 [95% CI, ] in CABG patients). Similarly, Nesher et al 10 described a concentration-dependent association of TNT and short-term MACEs. In addition, their analysis revealed an association between a TNT cutoff concentration of 0.8 μg/l Table 4. Cox Regression Model for Composite End Point Within 1 Year After Surgery Model Derivation and short-term mortality rate. In our data, the optimal cutoff for the prediction of 1-year mortality or MACE was also 0.8 μg/l. Croal et al 12 showed an OR of 1.10 (95% CI, ) per 10-μg/L troponin I increase for 1-year mortality rate. Our group described previously in 750 patients (enrolled in ) an area under the curve of 0.78 (95% CI, ) using TNT for 12-month mortality rate and an independent association between elevated TNT and 12-month mortality rate (OR, 3.83 [95% CI, ]) after adjustment using the EuroSCORE. 14 Previous studies (including our own data), however, disregarded potential confounding by postoperative complications and did not assess BNP concentrations. In addition, they did not quantify the prognostic gain compared with preoperative risk stratification tools in terms of reclassification (eg, the EuroSCORE). Fox et al 19 assessed the association between postoperative BNP and 5-year mortality rate in 1183 patients undergoing Model Validation (1000 Bootstrap Samples) HR 95% CI P Value HR 95% CI P Value Logistic EuroSCORE Peak TNT 0.8 μg/l Peak BNP 790 ng/l In-hospital sepsis In-hospital respiratory infection Sternal infection Acute kidney injury Logistic European System for Cardiac Operative Risk Evaluation (EuroSCORE) was entered continuously per percentage increase. Respiratory infections included pneumonia, ventilator associated pneumonia, or purulent tracheobronchitis. Acute kidney injury included the following categories: (1) risk, serum creatinine (SCr) increased by 1.5 or glomerular filtration rate (GFR) decreased >25%; (2) injury, SCr increased by 2 or GFR decreased >50%; (3) failure, SCr increased by 3 or GFR decreased 75% or SCr 4 mg/dl or acute rise 0.5 mg/dl; and (4) loss, persistent acute renal failure, complete loss of kidney function >4 weeks. 24 BNP indicates B-type natriuretic peptide; CI, confidence interval; HR, hazard ratio; LR, likelihood ratio; and TNT, troponin T.

8 Lurati Buse et al Biomarkers and Outcome After Cardiac Surgery 955 Figure 4. Calibration of the final model by risk groups. elective CABG. In contrast to our data, they did not detect any significant independent association between postoperative BNP and long-term mortality (hazard ratio, 1.62 [95% CI, ]). This fact might be related to their choice of a longer follow-up duration. It is plausible that, over a 5-year period, in an elderly population with a high burden of risk factors, subsequent cardiac and noncardiac events might have diluted the strength of the association between early postoperative BNP concentrations and all-cause death. This hypothesis is supported, first, by the report of the association between congestive heart failure episodes and heart failure mortality in the same population 20 and, second, by the significant independent association between postoperative BNP and poor physical function within a shorter follow-up duration in a subgroup of the same patients. 18 Three smaller cohort studies (<400 patients in total) 17,21,22 described higher postoperative BNP concentrations in patients who died after cardiac surgery. However, they were underpowered to conduct multivariable adjustment. Strength and Limitations The strength of our study stems from the following: (1) inclusion of a broad, unselected population of consecutive patients undergoing on-pump cardiac surgery, (2) the minimal number of missing TNT data (0.6%), (3) achieving >99% follow-up completeness, (4) the blinded event adjudication by 2 independent readers with extensive knowledge and experience of the perioperative cardiac surgery setting (cardiac anesthesiologists and intensive care physicians), (5) the high data quality with extensive data consistency checks and quality controls, and (6) our analysis plan that considers not only preoperative risk factors but also postoperative complications in the prediction model given their role as confounders and with the quantification of reclassification by cardiac biomarkers. We are aware of some limitations of our study. First, we assessed the concentration of biomarkers in all of the patients at 6:00 am on postoperative day 1, independent of the time of surgery. Therefore, the samples of postoperative day 1 were collected between 12 and 20 hours after surgery and the samples of postoperative day 2 between 36 and 44 hours after surgery. Blood sample collection at a fixed point of the day, however, conforms to clinical routine rather than to sample collection at a defined time after surgery. A further disadvantage of our set time for blood sample collection is the fact that cardiac biomarkers were not measured in the few patients who died before 6:00 am on the first postoperative morning. Second, in contrast to the standardized approach to cardiac biomarker sampling, the study protocol did not mandate any screening standard for infectious or renal complications; rather, the detection relied on diagnostic workup ordered by the attending physicians. Therefore, we cannot exclude that some patients might have been misclassified. However, this approach mimicked clinical routine, and as such it better complied with the pragmatic approach of our study question. In addition, the large sample size can be expected to have mitigated this issue. Third, follow-up assessment on the basis of mailed questionnaires may be subjected to nonresponse bias. However, telephone interviews with patients or their general practitioners allowed us to achieve almost complete follow-up (>99%). Our follow-up assessment may also have been biased by misclassification of outcomes, but the probability of forgetting a hospitalization seemed remote. Moreover, misclassification bias could be expected to be nondifferential, because the patients were not aware of their exposure status (ie, TNT and BNP readings). Fourth, the choice of equal weight for sensitivity and specificity to define the optimal cutoff point resulted in the fact that, at the proposed cutoff value, sensitivity was higher than specificity. This in turn led to a good reclassification in patients with increased risk, whereas risk downrating was poor. Fifth, we focused on postoperative TNT concentrations only. Thielmann et al 28 showed a prognostic value of elevated preoperative troponin I concentrations for short-term mortality. However, the high baseline risk of patients presenting a recent acute coronary syndrome or emergent surgery is already considered by preoperative risk stratification (EuroSCORE). 2 In addition, <8% of the included patients demonstrated

9 956 Circulation September 16, 2014 detectable TNT concentrations before surgery. We also did not measure preoperative BNP concentrations. Multiple studies 15,16,19 established the significant independent association between preoperative BNP concentrations and outcome after cardiac surgery. Sixth, for adjustment by preoperative risk estimation, we used the EuroSCORE that was initially generated to predict short-term outcome. However, its prognostic value was validated for long-term events. 23 In addition, the independent prognostic impact of cardiac biomarkers was maintained when restricting the analysis to short-term mortality and MACE. An additional limitation of the EuroSCORE is its overestimation of mortality, particularly with higher scores. 1,29 Seventh, a new version of the EuroSCORE (II) was proposed after completing enrollment. 26 We had information on the variables that were newly included in the model with the exception of pulmonary arterial pressure, which was collected as pulmonary hypertension (>60 mm Hg) instead of the categories (31 55 mm Hg and >55 mm Hg) as per the EuroSCORE II. Therefore, we were not able to calculate the EuroSCORE II in our patients. However, the addition of the new variables included in the EuroSCORE II (eg, angina pectoris Canadian Cardiovascular Society grade 4, New York Heart Association, and creatinine clearance) into our model did not affect the association between TNT and BNP with the 30-day or 1-year end point. In addition, the improvement in predictive performance of the EuroSCORE II is still not fully established. 1,29 31 Eighth, the sample size in the subgroups was not sufficient to test for interaction in the association between cardiac markers and outcome within specific surgical procedures. However, we did not find any evidence for interaction in the subgroup with isolated CABG versus other procedures. Finally, 12% of postoperative BNP concentrations were missing. However, with a total of >1300 patients, this cohort still represents one of the largest to address this topic. Postoperative TNT and BNP elevations are independently and strongly associated with 1-year mortality and MACEs after on-pump cardiac surgery. The extension of the EuroSCORE, a preoperative risk stratification tool, with postoperative TNT and BNP concentrations allowed for a better prediction of 1-year all-cause mortality and the occurrence of MACEs after on-pump cardiac surgery. Acknowledgments We thank Dr Salome Dell-Kuster for statistical advice; Dr Tanja Schmidt, Claudia Werner, and Anne-Michelle Droux for data collection; Domenik Zwahlen for data management; and Allison Dwileski for expert editorial assistance. Sources of Funding This study was supported by a research grant of the Swiss National Science Foundation (SNF No. 3200B0_121943/1), the European Association of Cardiothoracic Anaesthesiologists, and by the Foundation for Research and Education, Department for Anesthesia, Surgical Intensive Care, Prehospital Emergency Medicine and Pain Therapy, University Hospital Basel, Basel, Switzerland. None. Disclosures References 1. Nilsson J, Algotsson L, Höglund P, Lührs C, Brandt J. Comparison of 19 pre-operative risk stratification models in open-heart surgery. Eur Heart J. 2006;27: Nashef SA, Roques F, Michel P, Gauducheau E, Lemeshow S, Salamon R. European System for Cardiac Operative Risk Evaluation (EuroSCORE). Eur J Cardiothorac Surg. 1999;16: Roques F, Michel P, Goldstone AR, Nashef SA. The logistic EuroSCORE. Eur Heart J. 2003;24: Gensini GF, Fusi C, Conti AA, Calamai GC, Montesi GF, Galanti G, Noferi D, Carbonetto F, Palmarini MF, Abbate R, Vaccari M. Cardiac troponin I and Q-wave perioperative myocardial infarction after coronary artery bypass surgery. Crit Care Med. 1998;26: Quaini E, Colombo T, Russo C, Vitali E, Pellegrini A. Hospital morbidity and mortality after myocardial revascularisation surgery: current changes in risk factors. Eur J Cardiothorac Surg. 1995;9: Kollef MH, Sharpless L, Vlasnik J, Pasque C, Murphy D, Fraser VJ. The impact of nosocomial infections on patient outcomes following cardiac surgery. Chest. 1997;112: Kubota H, Miyata H, Motomura N, Ono M, Takamoto S, Harii K, Oura N, Hirabayashi S, Kyo S. Deep sternal wound infection after cardiac surgery. J Cardiothorac Surg. 2013;8: Tolpin DA, Collard CD, Lee VV, Virani SS, Allison PM, Elayda MA, Pan W. Subclinical changes in serum creatinine and mortality after coronary artery bypass grafting. J Thorac Cardiovasc Surg. 2012;143: e1. 9. Adabag AS, Rector T, Mithani S, Harmala J, Ward HB, Kelly RF, Nguyen JT, McFalls EO, Bloomfield HE. Prognostic significance of elevated cardiac troponin I after heart surgery. Ann Thorac Surg. 2007;83: Nesher N, Alghamdi AA, Singh SK, Sever JY, Christakis GT, Goldman BS, Cohen GN, Moussa F, Fremes SE. Troponin after cardiac surgery: a predictor or a phenomenon? Ann Thorac Surg. 2008;85: Sellgren A, Nilsson F, Jeppsson A. The relationship between ASAT, CKMB, troponin-t and mortality after cardiac surgery. Scand Cardiovasc J. 2007: Croal BL, Hillis GS, Gibson PH, Fazal MT, El-Shafei H, Gibson G, Jeffrey RR, Buchan KG, West D, Cuthbertson BH. Relationship between postoperative cardiac troponin I levels and outcome of cardiac surgery. Circulation. 2006;114: Lurati Buse GA, Koller MT, Grapow M, Bolliger D, Seeberger M, Filipovic M. The prognostic value of troponin release after adult cardiac surgery: a meta-analysis. Eur J Cardiothorac Surg. 2010;37: Lurati Buse GA, Koller MT, Grapow M, Brüni CM, Kasper J, Seeberger MD, Filipovic M. 12-Month outcome after cardiac surgery: prediction by troponin T in combination with the European system for cardiac operative risk evaluation. Ann Thorac Surg. 2009;88: Fox AA, Shernan SK, Collard CD, Liu KY, Aranki SF, DeSantis SM, Jarolim P, Body SC. Preoperative B-type natriuretic peptide is as independent predictor of ventricular dysfunction and mortality after primary coronary artery bypass grafting. J Thorac Cardiovasc Surg. 2008;136: Lurati Buse GA, Koller MT, Burkhart C, Seeberger MD, Filipovic M. The predictive value of preoperative natriuretic peptide concentrations in adults undergoing surgery: a systematic review and meta-analysis. Anesth Analg. 2011;112: Crescenzi G, Landoni G, Bignami E, Belloni I, Biselli C, Rosica C, Guarracino F, Marino G, Zangrillo A. N-terminal B-natriuretic peptide after coronary artery bypass graft surgery. J Cardiothorac Vasc Anesth. 2009;23: Fox AA, Marcantonio ER, Collard CD, Thoma M, Perry TE, Shernan SK, Muehlschlegel JD, Body SC. Increased peak postoperative B-type natriuretic peptide predicts decreased longer-term physical function after primary coronary artery bypass graft surgery. Anesthesiology. 2011;114: Fox AA, Muehlschlegel JD, Body SC, Shernan SK, Liu KY, Perry TE, Aranki SF, Cook EF, Marcantonio ER, Collard CD. Comparison of the utility of preoperative versus postoperative B-type natriuretic peptide for predicting hospital length of stay and mortality after primary coronary artery bypass grafting. Anesthesiology. 2010;112: Fox AA, Nascimben L, Body SC, Collard CD, Mitani AA, Liu KY, Muehlschlegel JD, Shernan SK, Marcantonio ER. Increased perioperative b-type natriuretic peptide associates with heart failure hospitalization or heart failure death after coronary artery bypass graft surgery. Anesthesiology. 2013;119: Provenchère S, Berroeta C, Reynaud C, Baron G, Poirier I, Desmonts JM, Iung B, Dehoux M, Philip I, Bénessiano J. Plasma brain natriuretic peptide and cardiac troponin I concentrations after adult cardiac surgery:

10 Lurati Buse et al Biomarkers and Outcome After Cardiac Surgery 957 association with postoperative cardiac dysfunction and 1-year mortality. Crit Care Med. 2006;34: Suttner S, Boldt J, Lang K, Röhm KD, Piper SN, Mayer J. Association of N-terminal pro-brain natriuretic peptide and cardiac troponin T with in-hospital cardiac events in elderly patients undergoing coronary artery surgery. Eur J Anaesthesiol. 2008;25: Toumpoulis IK, Anagnostopoulos CE, DeRose JJ, Swistel DG. European system for cardiac operative risk evaluation predicts long-term survival in patients with coronary artery bypass grafting. Eur J Cardiothorac Surg. 2004;25: Bellomo R, Ronco C, Kellum JA, Mehta RL, Palevsky P; Acute Dialysis Quality Initiative workgroup. Acute renal failure: definition, outcome measures, animal models, fluid therapy and information technology needs: the Second International Consensus Conference of the Acute Dialysis Quality Initiative (ADQI) Group. Crit Care. 2004;8:R204 R Pencina MJ, D Agostino RB Sr, D Agostino RB Jr, Vasan RS. Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med. 2008;27:157 72; discussion Nashef SA, Roques F, Sharples LD, Nilsson J, Smith C, Goldstone AR, Lockowandt U. EuroSCORE II. Eur J Cardiothorac Surg. 2012;41:734 44; discussion Bagley SC, White H, Golomb BA. Logistic regression in the medical literature: standards for use and reporting, with particular attention to one medical domain. J Clin Epidemiol. 2001;54: Thielmann M, Massoudy P, Neuhäuser M, Knipp S, Kamler M, Piotrowski J, Mann K, Jakob H. Prognostic value of preoperative cardiac troponin I in patients with non-st-segment elevation acute coronary syndromes undergoing coronary artery bypass surgery. Chest. 2005;128: Howell NJ, Head SJ, Freemantle N, van der Meulen TA, Senanayake E, Menon A, Kappetein AP, Pagano D. The new EuroSCORE II does not improve prediction of mortality in high-risk patients undergoing cardiac surgery: a collaborative analysis of two European centres. Eur J Cardiothorac Surg. 2013;44: Chalmers J, Pullan M, Fabri B, McShane J, Shaw M, Mediratta N, Poullis M. Validation of EuroSCORE II in a modern cohort of patients undergoing cardiac surgery. Eur J Cardiothorac Surg. 2013;43: Grant SW, Hickey GL, Dimarakis I, Trivedi U, Bryan A, Treasure T, Cooper G, Pagano D, Buchan I, Bridgewater B. How does EuroSCORE II perform in UK cardiac surgery; an analysis of patients from the Society for Cardiothoracic Surgery in Great Britain and Ireland National Database. Heart. 2012;98: Clinical Perspective Surgical risk stratification models focus on preoperative and intraoperative risk factors. This view may prove fragmentary postoperatively, because it disregards perioperative and early postoperative complications. Therefore, there is a need for tools that allow for early postoperative risk reassessment to guide management decisions. This prospective cohort study in 1559 patients undergoing on-pump cardiac surgery showed the independent association between troponin T 0.8 μg/l and B-type natriuretic peptides 790 ng/l within the first 2 postoperative days and 1-year all-cause mortality and major adverse cardiac events. This association persisted after adjustment considering preoperative and postoperative confounders. Extending the European System for Cardiac Operative Risk Evaluation to include postoperative troponin T and B-type natriuretic peptide information allowed for improved risk stratification in every fourth patient. The time window of 1 month between elevated biomarkers and events suggests that measurement of cardiac biomarkers flags high-risk patients early, that is, at a time point when management optimization holds promise for outcome improvement.