Postoperative Cardiac Troponin I Thresholds Associated With 1-Year Cardiac Mortality After Adult Cardiac Surgery: An Attempt to Link Risk Stratification With Management Stratification in an Observational Study

Accepted Manuscript

Postoperative Cardiac Troponin I Thresholds Associated with One-Year Cardiac Mortality after Adult Cardiac Surgery: An Attempt to link Risk Stratification with Management Stratification in an Observational Study Sophie Provenchere , ` , Jean Guglielminotti , Aurelie ´ Gouel-Cheron ´ Edouard Bresson , Laetitia Desplanque , Claire Bouleti , Bernard Iung , Philippe Montravers , Monique Dehoux , Dan Longrois PII: DOI: Reference:

S1053-0770(19)30593-2 https://doi.org/10.1053/j.jvca.2019.06.039 YJCAN 5357

To appear in:

Journal of Cardiothoracic and Vascular Anesthesia

Please cite this article as: Sophie Provenchere , ` , Jean Guglielminotti , Aurelie ´ Gouel-Cheron ´ Edouard Bresson , Laetitia Desplanque , Claire Bouleti , Bernard Iung , Philippe Montravers , Monique Dehoux , Dan Longrois , Postoperative Cardiac Troponin I Thresholds Associated with OneYear Cardiac Mortality after Adult Cardiac Surgery: An Attempt to link Risk Stratification with Management Stratification in an Observational Study, Journal of Cardiothoracic and Vascular Anesthesia (2019), doi: https://doi.org/10.1053/j.jvca.2019.06.039

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Postoperative Cardiac Troponin I Thresholds Associated with One-Year Cardiac Mortality after Adult Cardiac Surgery: An Attempt to link Risk Stratification with Management Stratification in an Observational Study Running head: Provenchere et al.; Troponin I and Mortality after Cardiac Surgery

1

CR IP T

Sophie Provenchère 1,2 ; Jean Guglielminotti 1,3,9 ; Aurélie Gouel-Chéron 1 ; Edouard Bresson ; Laetitia Desplanque 1 ; Claire Bouleti 4 ; Bernard Iung 4,5,7 ; Philippe Montravers 1,5 ;

Monique Dehoux 6,8 ; Dan Longrois 1,5,7.

(1) Département d‟Anesthésie-Réanimation, APHP, Hôpital Bichat-Claude Bernard, 46 rue

AN US

Henri Huchard, 75877, Paris Cedex 18, France.

(2) INSERM, Centre d‟Investigation Clinique 1425, APHP, Hôpital Bichat-Claude Bernard,

M

46 rue Henri Huchard, 75877, Paris Cedex 18, France.

(3) INSERM, UMR 1137, IAME, 16 rue Henri Huchard, 75018 Paris, France.

ED

(4) Département de Cardiologie, APHP, Hôpital Bichat-Claude Bernard, 46 rue Henri

PT

Huchard, 75877 Paris Cedex 18, France.

CE

(5) Université Paris 7-Diderot, Paris, France. (6) Département de Biochimie, APHP, Hôpital Bichat-Claude Bernard, 46 rue Henri Huchard,

AC

75877 Paris Cedex 18, France. (7) INSERM 1148, 16 rue Henri Huchard, 75018 Paris, France (8) INSERM 1152, 16 rue Henri Huchard, 75018 Paris, France (9) Department of Anesthesiology, Columbia University College of Physicians and Surgeons, 622 West 168th Street, PH5-505, New York, NY 10032, USA.

ACCEPTED MANUSCRIPT

Sophie Provenchère (Corresponding author): Département d‟Anesthésie-Réanimation, APHP, Hôpital Bichat-Claude Bernard, 46 rue Henri Huchard, 75877, Paris Cedex 18,

CR IP T

France. [email protected] Telephone number: +33140258355

AN US

Fax number: +33140258351

Jean Guglielminotti: Department of Anesthesiology, Columbia University College of Physicians and Surgeons, 622 West 168th Street, PH5-505, New York, NY 10032, USA.

ED

Fax: 212-305-3296

M

Phone: 212-305-5752

PT

Email: [email protected]

CE

Aurélie Gouel-Chéron: Département d‟Anesthésie-Réanimation, APHP, Hôpital BichatClaude Bernard, 46 rue Henri Huchard, 75877, Paris Cedex 18, France.

AC

[email protected]

Telephone number: +33140258355 Fax number: +33140258351 Edouard Bresson: Département d‟Anesthésie-Réanimation, APHP, Hôpital Bichat-Claude Bernard, 46 rue Henri Huchard, 75877, Paris Cedex 18, France.

ACCEPTED MANUSCRIPT [email protected] Telephone number: +33140258355 Fax number: +33140258351 Laetitia Desplanque: Département d‟Anesthésie-Réanimation, APHP, Hôpital Bichat-

CR IP T

Claude Bernard, 46 rue Henri Huchard, 75877, Paris Cedex 18, France. [email protected]

AN US

Telephone number: +33140258355 Fax number: +33140258351

Claire Bouleti: Département de Cardiologie, APHP, Hôpital Bichat-Claude Bernard, 46 rue

M

Henri Huchard, 75877 Paris Cedex 18, France.

ED

[email protected]

Telephone number: +33 1 40 25 67 60

PT

Fax number: +33 1 40 25 67 32

CE

Bernard Iung: Département de Cardiologie, APHP, Hôpital Bichat-Claude Bernard, 46 rue

AC

Henri Huchard, 75877 Paris Cedex 18, France. [email protected] Telephone number: +33 1 40 25 67 60 Fax number: +33 1 40 25 67 32 Philippe Montravers: Département d‟Anesthésie-Réanimation, APHP, Hôpital BichatClaude Bernard, 46 rue Henri Huchard, 75877, Paris Cedex 18, France.

ACCEPTED MANUSCRIPT [email protected] Telephone number: +33140258355 Fax number: +33140258351

Henri Huchard, 75877 Paris Cedex 18, France. [email protected]

AN US

Telephone number: +331 40 25 85 48

CR IP T

Monique Dehoux: Département de Biochimie, APHP, Hôpital Bichat-Claude Bernard, 46 rue

Fax number: +331 40 25 88 21

Dan Longrois: Département d‟Anesthésie-Réanimation, APHP, Hôpital Bichat-Claude

M

Bernard, 46 rue Henri Huchard, 75877, Paris Cedex 18, France.

ED

[email protected] Telephone number: +33140257427

CE

PT

Fax number: +33140258351

AC

Declarations:

Trial registration: ClinicalTrials.gov ID: NCT03393169 Ethics approval:

ACCEPTED MANUSCRIPT The study was approved by Bichat hospital Institutional Review Board (“Comité d‟Evaluation des Projets de Recherche Biomédicale Paris Nord”; IRB 00006477). It waived need for patient‟s informed consent. Consent to participate:

CR IP T

not applicable Availability of data and materials:

All data generated or analyzed during the study are included in this published article and its

AN US

supplementary information files. Conflict of Interest: None declared

M

Funding Sources:

ED

Support was provided by the Department of Anaesthesiology, Bichat Hospital, Paris, France

PT

Jean Guglielminotti is supported by an R03 from the Agency for Healthcare Research and

CE

Quality (1 R03 HS025787-01).

AC

Authors’ contributions: SP initiated and conducted the study, performed data acquisition, analyzed and interpreted the data and wrote the first manuscript draft. JG performed the statistical analysis, analyzed and interpreted the data and helped with drafting and revising the manuscript. AGC, EB, LD, CB were involved in data acquisition and helped with drafting the manuscript. BI, PM, MD, were involved in revising the manuscript. DL was involved in study design, in statistical analysis,

ACCEPTED MANUSCRIPT analyzed and interpreted the data, and drafting and revising the manuscript. The statistical analysis was performed and interpreted by JG (PhD and postdoc in Clinical Epidemiology), SP and DL. All authors read and approved the final manuscript. Acknowledgements relating to this article:

CR IP T

The authors thank INSERM CIC 1425, for monitoring and data management of the study.

Abstract

AN US

Background: Cardiac troponin (cTn) concentrations are routinely measured in some centers after cardiac surgery as part of risk stratification, but there are no data on how “increased cTn concentrations” could change patients‟ management.

Objectives: To estimate relevant cTnI thresholds and identify potential interventions

M

(additional monitoring/therapeutic interventions) that could be part of management changes of

ED

patients with cTnI > relevant thresholds.

Design: Retrospective single-center observational study.

PT

Setting: Bichat-Claude Bernard Hospital, Paris, France, between January 1, 2009 and December 31, 2012.

CE

Patients: Consecutive adult patients undergoing cardiac surgery.

AC

Main outcome measures: cTnI was measured on the 20th postoperative hour. Causes of death and possible interventions were determined by individual medical records analysis. cTnI thresholds for 1-year cardiac mortality with a specificity > 80% were calculated. Results: We analyzed 3,228 procedures: 129 deaths occurred (4%) out of which 83 (2.6%) were cardiac deaths. Thresholds cTnI values were 4.2 µg/L for coronary artery bypass grafting (CABG) (95% CI, 3.9-4.5) and 10.7 µg/l for non-CABG (95% CI, 10.0-11.3). In multivariable analysis, the EuroSCORE II (OR 1.1; 95% CI, 1.06-1.13; P < 0.001) and cTnI >

ACCEPTED MANUSCRIPT thresholds (OR 5.62; 95 % CI, 3.37-9.37; P < 0.001) were associated with significantly increased risk of death. The additive and absolute Net Reclassification Index were 0.288 and 14.1%, respectively, for a logistic model including cTnI and EuroSCORE II [AUC C-index 0.82 (95% CI, 0.77 to 0.87)] compared with a model including only EuroSCORE II [AUC Cindex 0.80 (95% CI, 0.75 to 0.84)]. Fifty three of the 83 cardiac deaths (64%) had a cTnI >

CR IP T

thresholds and an intervention was deemed possible in 47 out of 53 (89%) of them (mostly patients with mild postoperative cardiac dysfunction). For non-cardiac deaths, 28% had a cTnI > thresholds and no interventions were deemed possible.

Conclusions: In an attempt to evolve from risk- to management stratification, our results

AN US

identified a subgroup of patients with mild cardiac dysfunction and a cTnI > thresholds who could be the target for interventions in future validation studies concerning changes in

M

patients „management.

CE

Introduction

PT

ED

Trial registration: ClinicalTrials.gov ID: NCT03393169

Cardiac troponin I or T (cTnI/T) isoforms are sensitive and specific biomarkers of

AC

myocardial damage, independently of the causative injury [1-4]. Previous research in adult cardiac surgery patients [5-12], including meta-analyses [13, 14] have reported an association between increased postoperative cTnI/T concentrations and short- or intermediate-term allcause mortality. Many, but not all, cardiac surgery centers routinely measure postoperative cTnI/T concentrations as part of postoperative risk stratification [15].

ACCEPTED MANUSCRIPT However, no study has specifically examined whether the information “increased postoperative cTnI/T concentration” may help clinicians to decide interventions (e.g., changes in diagnostic/monitoring or therapeutic strategies) that could result in improved patient outcome. This approach is sometimes referred as management reasoning (i.e., the process of making decisions about patient management) as opposed to diagnosis reasoning and relies on

CR IP T

specific tools [16]. There is a clear lack of management reasoning and stratification as compared to diagnosis reasoning and risk stratification concerning the use of troponin measurements following cardiac surgery. Several explanations may account for this absence of management stratification triggered by an increased postoperative cTnI/T concentration.

AN US

The first one is the absence of a universally accepted cTnI/T threshold associated with an increased risk of adverse postoperative outcome. The definition of such a threshold is complex because it depends on: (i) patients populations [17]; (ii) type of cardiac surgical

M

procedures (coronary artery bypass grafting (CABG) procedures versus non-CABG procedures with myotomy) [18]; (iii) timing of postoperative measurements [19]; (iv)

ED

definitions of the outcomes; (v) and the troponin assays used [20].The second one is that studies that defined cTnI/T thresholds associated with an increased risk of postoperative

PT

mortality were based on a statistical approach that maximized statistical metrics such as the

CE

Youden index (i.e. that maximizes both sensitivity and specificity) [21]. However, this approach attributes the same weight to a false positive and a false negative result. Maximizing

AC

both sensitivity and specificity makes sense only if consequences for the patient of an intervention based on a false positive result or of a non-intervention based on a false negative result are equal, which is a rare situation in medicine [22]. A clinical-based approach for defining a threshold value of “increased postoperative cTnI/T” and the subsequent decisionmaking process is probably more appropriate. When interventions do not exist, or their efficacy is not documented or are associated with potential adverse effects, it is desirable to

ACCEPTED MANUSCRIPT choose a threshold value associated with lower sensitivity and higher specificity. Finally, the third possible explanation for the lack of stratification management based on increased cTnI/T values (whatever the threshold value) after cardiac surgery is that the outcome used in previous studies is all-cause mortality, including cardiac and non-cardiac causes [5, 7, 11, 12]. Because interventions to prevent cardiac and non-cardiac mortality triggered by an increase in

CR IP T

postoperative cTnI/T values may not be the same, preventing all-cause mortality is challenging if not impossible [23-25]. On the contrary, prevention of cardiac mortality is a credible challenge and examples of such interventions have been recently published in non-

AN US

cardiac surgery [23-26].

The goal of this study was to calculate cTnI thresholds for one-year cardiac mortality with specificity greater than 80% in CABG and non-CABG procedures. From these threshold cTnI values we conducted an a posteriori exploratory analysis to identify potential

M

interventions (e.g., changes in monitoring duration/intensity and or changes in

ED

drugs/therapeutic regimens) that could have been undertaken in deceased patients (separated into cardiac and non-cardiac deaths) with cTnI greater than this threshold. To the best of our

PT

knowledge, this study is the first attempt, in cardiac surgery patients, to shift from statistical associations between increased postoperative cTn values and mortality (risk stratification) to

AC

CE

proposals for management stratification.

ACCEPTED MANUSCRIPT Material and Methods TRIPOD and STROBE statements for reporting were followed. The study was approved by Bichat Hospital Institutional Review Board (IRB 00006477- of HUPNVS, Paris 7 University, AP-HP), Pr. Michel LEJOYEUX, Chair of the Institutional Review Board on March 19, 2015 (“Comité d‟Evaluation des Projets de Recherche

CR IP T

Biomédicale Paris Nord”). It waived need for patient‟s informed consent because cTn measurements after cardiac surgery are part of routine care in our Institution. Source of data/Participant

AN US

Bichat-Claude Bernard Hospital is a 900-bed teaching hospital performing 1200 annual cardiac surgical procedures with cardiopulmonary bypass (CPB). Patients and surgery characteristics as well as outcomes are entered into an institutional database by trained research personnel since 2005. It allows calculation of the EuroSCORE II [27]. Death during

M

the hospital stay and up to one year after surgery is also collected in this database that is

ED

interfaced with the hospital information system where deaths are systematically recorded. Data from consecutive patients who underwent cardiac surgery between January 1, 2009 and

PT

December 31, 2012 were first selected. This time period corresponds to the same cTnI measurement assay. Then patients with the following criteria were excluded: 1) emergent

CE

cardiac surgery (within 24 hours after hospital admission); 2) complex or unusual procedures

AC

(e.g., aortic arch surgery with circulatory arrest and deep hypothermia, congenital heart disease such as atrial septal defect); 3) death on the day of surgery or death more than 365 days after the procedure; and 4) missing postoperative cTn values (Fig 1). For the purpose of the study, procedures were defined as isolated CABG (CABG group) or non-isolated CABG. The latter group included isolated valve procedures, valve procedures combined with CABG, and ascending aorta procedures.

ACCEPTED MANUSCRIPT Plasma cTnI concentrations were measured on the 20th postoperative hour with a sensitive assay on the Dimension Vista 1500 analyser (Siemens Healthcare Diagnostics, Saint Denis, France). The limit of detection, the limit of quantification (coefficient of variation <10 %), and the 99th percentile value provided by the manufacturer were 15 ng/L, 40 ng/L and 45 ng/L, respectively. They were confirmed by our laboratory [28].

CR IP T

Aims of the study

The primary aim of the study was to calculate the postoperative cTn I concentration associated with specificity greater than 80% for one-year cardiac mortality. The secondary

AN US

aim was to identify potential interventions (as part of an attempt to evolve from risk

stratification to management stratification) that could be implemented in patients with cTnI values higher than the defined threshold. records

of

non-survivors

were

assessed

by two

senior

cardiac

M

Medical

anesthesiologists (SP, DL) to determine the causes of death and identify the potential

ED

interventions that could have been proposed to prevent death. The reviewers were blinded to

PT

cTnI concentrations. Causes of death were a priori categorized, based on the literature [29] as cardiac and non-cardiac. These definitions are presented in Table 1. Each of the two reviewers

CE

reviewed the medical files independently. Possible discordance was solved by consensus with a third senior anesthesiologist (EB). Potential interventions were categorized as hemodynamic

AC

optimization, enhanced diagnosis/monitoring/therapeutics or none. This approach is similar to that published for non-cardiac surgery patients [23-26]. Predictors A multivariable logistic regression model was developed to identify risk factors for cardiac deaths. The candidate variables were selected a priori based on clinical relevance and included: (i) preoperative factors that are consolidates in the EuroSCORE II values (all

ACCEPTED MANUSCRIPT preoperative data with few modifiable parameters); (ii) the duration of aortic cross-clamping (that reflects the ischemic burden); (iii) type of surgery (CABG or non-CABG),and (iv) cTnI values greater than the threshold with an 80% specificity. A priori and a posteriori power of the study

CR IP T

The required sample size was calculated to observe an odds ratio ≥ 1.5 for the risk of death associated with cTnI values greater than the threshold with an 80% specificity and the following hypotheses: in-hospital mortality 5%, alpha 5%, power 90%, and a two-tailed test. At least 1300 patients had to be included. The power calculated a posteriori was 99%

AN US

Missing data analysis

One hundred and seventy patients with missing cTn value were excluded from the statistical analysis (Fig 1). They were subsequently compared with patients without missing

M

postoperative cardiac troponin I (cTnI) values (Supplemental Table 1).

ED

Statistical analyses

PT

The statistical analyses were performed with R version 3.0.2 (R Foundation for Statistical Computing, Vienna, Austria). Results are expressed as median (interquartile range)

CE

or count (percentages). Comparisons of continuous variables used unpaired Wilcoxon tests and comparisons of discrete variables Chi-square or Fisher exact tests. For comparisons of

AC

continuous variables among three groups, Kruskall-Wallis tests were used followed by unpaired Wilcoxon tests. The P-value for statistical significance was 0.05. The Bonferroni correction was applied when necessary. The ability of postoperative cTnI concentration to discriminate 1-year survivors from non-survivors used the area under the receiver operator characteristic curve (AUCROC). Comparisons of 2 AUCROCs used the DeLong method. Confidence intervals for the

ACCEPTED MANUSCRIPT difference between 2 AUCROCs were calculated with bootstrap resampling (B=2000) with replacement and the percentiles method. Sensitivity and specificity were calculated with standard formulas: true positive / (true positive + false negative) and true negative / (false positive + true negative), respectively. The cTnI threshold value associated with a specificity ≥ 80, 85, and 90% was the lowest value

CR IP T

with a specificity ≥ 80, 85, and 90%, respectively. Confidence intervals for the threshold value were calculated with bootstrap resampling (B=2000) with replacement and the percentiles method.

AN US

A multivariable logistic regression model with backward selection was developed to identify risk factors for cardiac deaths. The candidate variables were selected a priori based on clinical relevance and included: cTnI greater than the threshold with an 80% specificity for cardiac mortality, type of surgery (CABG or non-CABG), EuroSCORE II values, and

M

duration of aortic cross-clamping. The final regression model was used to calculate the

ED

adjusted odds ratio for the risk of cardiac death associated with a cTnI>threshold.

PT

The value of adding the cTnI>threshold to the EuroSCORE II to estimate the risk of cardiac death was estimated by comparing 2 logistic regression models with cardiac death as

CE

the dependent variable. The first model included the EuroSCORE II as the only independent variable and the second model included the EuroSCORE II and the cTnI>threshold as the 2

AC

independent variables. The 2 models were compared using the C-index, the HosmerLemeshow test, and the additive and absolute net reclassification index (NRI) [30, 31]. The additive NRI quantifies the extent to which a biomarker correctly reclassifies individuals to a higher or lower category of risk. The absolute NRI indicates the proportion of patients correctly or incorrectly reclassified. Three groups of risk were considered: low-, intermediate, and high-risk based on a risk of death as calculated by the EuroSCORE II less than 4%,

ACCEPTED MANUSCRIPT between 4 and 8, and greater than 8 [32]. No internal validation was performed (e.g., splitting the dataset into a training and a validation sets) because of the relatively low number of cardiac deaths and the controversy surrounding this approach [33]. Perioperative management

CR IP T

Perioperative management and the anesthesia/intensive care and surgical teams were unchanged during the study period. On the morning of surgery, all the patients received hydroxyzine and those on chronic beta-blocker therapy received their usual dose. Anesthesia was induced and maintained with effect-site target-controlled infusions of propofol [34] and

AN US

sufentanil [35] without use of inhaled anesthetics that could have confounded the interpretation of post-operative troponin values [36]. Muscular relaxation was provided by atracurium. Patients were monitored with invasive radial artery blood pressure, 5-lead electrocardiogram (EKG), pulse oximetry, and bispectral index (BIS Vista®, Medtronic,

M

USA). Trans esophageal echocardiography was used selectively. Pulmonary artery catheter

ED

monitoring was limited to patients with severe left ventricular dysfunction, i.e. left ventricular ejection fraction (LVEF) < 40% or pulmonary arterial hypertension. An autologous blood

PT

recovery system was used for every patient. In the absence of contraindications (e.g.,

CE

idiopathic thromboembolic disease; glomerular filtration rate < 30 ml/min/m2), all patients received a 20 mg/kg tranexamic acid intravenous bolus followed by a continuous infusion of

AC

2 mg/kg/h. In case of arterial hypotension [37] boluses of phenylephrine were given to maintain mean arterial blood pressure between 50 and 80 mm Hg. Cardiopulmonary bypass (CPB) was conducted under normothermia (central

temperature > 35.5°C). An initial dose of 300 UI/kg of unfractionated heparin (UFH) was administrated, before the initiation of CPB. Efficacy of UFH was checked with the activated coagulation time (ACT, Hemochron®, Gamida, Eaubonne, France) with the goal of ACT >

ACCEPTED MANUSCRIPT 400 seconds and controlled every 20 minutes. Myocardial protection (unique solution) was achieved by intermittent anterograde or retrograde warm blood cardioplegia. Patients who underwent CABG had mainly anterograde cardioplegia whereas patients who underwent complex surgery had anterograde and/or retrograde cardioplegia according to surgical constraints. Management of CPB was targeted both saturation of the venous line > 80 % and

CR IP T

mean arterial pressure (MAP) values > 60-80 mmHg according to the preoperative MAP values. After aortic unclamping, if necessary, internal cardioversion (given at 20 J, increased to 30 J after the second shock) was used. Inotropic support following aortic unclamping was mainly based on norepinephrine and/or dobutamine infusion to maintain a stable

AN US

hemodynamic state. As soon as patients were weaned from CPB, protamine was administered (1:1 ratio for the total UFH dose). When required, blood products were administrated according to the international recommendations [38, 39] and adapted to the results of the

M

coagulation tests performed after protamine infusion.

ED

Postoperative care in the ICU was provided by cardiac anesthesiologists who were also intensivists, and in the telemetry unit and on the ward by cardiologists who routinely and

PT

exclusively take care of cardiac surgery patients. EKGs were performed upon arrival in the ICU and daily until discharge. Postoperative echocardiography was performed in the ICU in

CE

case of hemodynamic instability, EKG changes or major bleeding, or later on the ward by the

AC

cardiologists. Routine biologic testing, arterial blood gases and chest radiographs were obtained upon arrival in the ICU, and then daily and when indicated thereafter. Patients were weaned from mechanical ventilation 6-8 hours after admission to the ICU, except in cases of hemodynamic, respiratory or of bleeding complications. Standard postoperative care included blood glucose control, intravenous heparin (100U/Kg) beginning no sooner than 6 h after surgery in the absence of significant bleeding followed by the application of the institutional anticoagulation protocols. In chronically treated patients, beta-blocking agents, statins and

ACCEPTED MANUSCRIPT oral antiplatelet agents were resumed as soon as possible postoperatively (mainly following day 1 after surgery).

Results The selection of the study sample is presented in Figure 1 and characteristics of

CR IP T

patients excluded for missing cTnI values are summarized in Supplemental Table 1. During the study period, 3,228 procedures were analyzed, including 1,247 CABG (39%). Characteristics of the CABG and non-CABG groups are presented in Table 2. Compared with

AN US

patients in the CABG group, patients in the non-CABG group had higher postoperative cTnI concentrations: 4.8 (2.7-9.4) versus 2.2 (1.4-3.6) µg/l, respectively (P <0.0001). In the nonCABG group postoperative cTnI concentrations did not differ across the 3 types of procedures (Supplemental Table 2).

M

During the study period, 129 deaths (4.0%) were recorded in the institutional database

ED

that undergoes regular quality controls during the first year following surgery. There were 83 cardiac deaths (64.4%), including 48 deaths (57.8%) with initially obvious/severe cardiac

PT

dysfunction and 35 deaths (42.2%) with initially absent/mild cardiac dysfunction (Table 3).

CE

Postoperative cTnI concentrations did not differ between cardiac deaths with initially obvious/severe cardiac dysfunction and cardiac deaths with initially absent/mild cardiac

AC

dysfunction (13.7 versus 12.3 µg/l, respectively; P=0.10) but were higher than in non-cardiacdeaths (4.0 µg/l). EuroSCORE II values and type of surgery did not differ across the causes of death (Table 3). Interestingly, the time to deaths was significantly different between the 48 deaths with initially obvious/severe cardiac dysfunction versus the 35 deaths with initially absent/mild cardiac dysfunction (see Table 3). Postoperative cTnI concentrations and one-year all-cause mortality:

ACCEPTED MANUSCRIPT Postoperative cTnI concentrations were higher in non-survivors (all-cause mortality) than in survivors both in the CABG and in the non-CABG groups: 5.2 (3.1-11.5) versus 2.1 (1.4-3.6) µg/l in the CABG group (P<0.0001) and 15.1 (7.1-38.4) versus 4.7 (2.6-8.9) µg/l in the non-CABG group (P<0.0001). The ability of cTnI concentration to discriminate survivors from non-survivors (all-cause mortality) assessed with the AUCROC curve was 0.80 (95%

CR IP T

CI: 0.71 to 0.89) for the CABG group and 0.76 (95% CI: 0.69 to 0.84) for the non-CABG group (P=0.53 for comparison between the two AUCROCs; difference between the 2 AUCROCs: 0.04; 95% CI: -0.08 to 0.14).

AN US

Postoperative cTnI concentrations and one-year cardiac mortality

In the CABG group, cTnI threshold values with an 80% specificity for one-year cardiac mortality was 4.2 (95% CI: 3.9 to 4.5) µg/l (Table 4). In the non-CABG group, the

ED

Risk factors for cardiac mortality

M

threshold value was 10.7 (95% CI: 10.0 to 11.3) µg/l.

In the final multivariable logistic regression model of risk factors for cardiac mortality,

PT

two factors were identified: EuroSCORE II (adjusted OR 1.09 per 1-unit increase; 95% CI,

3).

CE

1.06 to 1.12) and cTnI>threshold (adjusted OR 5.6; (95% CI: 3.4 to 9.4) (Supplemental Table

AC

Added value of the cTnI concentration: Compared with a logistic model including only the EuroSCORE II, adding cTnI

concentration increased the C-index of the logistic model from 0.801 (95% CI, 0.754 to 0.848) to 0.825 (95% CI, 0.778 to 0.872) and the Hosmer-Lemeshow test P-value from 0.001 to 0.006 (Table 5 and Supplemental Figure 1). The additive NRI was 0.288 and the absolute NRI 14.1% (Table 5 and Supplemental Table 4).

ACCEPTED MANUSCRIPT Potential interventions and cTnI thresholds values In the 129 deceased patients, an intervention was estimated possible in 71 (55%) (Table 6). It could have been enhanced postoperative monitoring in 22 (31%) and hemodynamic (drugs) optimization in 49 (69%) patients. When limiting the analysis to the 66 deceased patients with a cTnI value greater than the defined threshold (51%), a potential

CR IP T

intervention was estimated possible in 47 (71%). It could have been enhanced postoperative monitoring in 12 (25%) and hemodynamic (drugs) optimization in 35 (75%). For the 13 noncardiac deaths with a cTnI greater than the threshold, no intervention was deemed possible.

AN US

Of notice, in survivors, 607 out of 3,099 (20%) had a cTnI value greater than this threshold. When examining the cause of death, 53 out of the 83 cardiac deaths (64%) had a cTnI value greater than the cTnI threshold and an intervention was considered possible in 47 out of these

M

53 cases (89%).

ED

Discussion

Main findings and clinical context

PT

The main findings of this manuscript is that by defining cTnI thresholds associated

CE

with a specificity of > 80 % for cardiac mortality we identified a subgroup of patients that could mostly benefit from potential interventions: patients with absent/ mild postoperative

AC

cardiovascular dysfunction. Clinicians that routinely prescribe cTnI/T measurements after cardiac surgery for risk

stratification would like that this be followed by at least a change in medical practice (management stratification) if not improved patient outcome. The transition from risk stratification to management stratification in medical practice faces several challenges: the availability of a validated threshold to define “increased cTnI/T” and the proposal followed

ACCEPTED MANUSCRIPT by validation of potential interventions (management stratification) for patients with cTnI/T values above this threshold [16]. The conceptual difficulties of the transition from risk stratification to management stratification have been described in the Introduction section and are mainly related to the issues of:(i) all-cause mortality versus cardiac mortality; (ii) the clinical “value” or implications of false positive/false negative results.

CR IP T

When comparing our results with previous articles, it is obvious that the strategy to define threshold values for increased postoperative cTnI in this study is different from previous research [18]. First, previous research used a statistical approach for cTnI/T

AN US

thresholds that maximized sensitivity and specificity but attributed the same clinical value to a false positive and a false negative result. When interventions do not exist, or are associated with potential adverse effects or when their efficacy is not documented, it is desirable to choose a threshold value associated with higher specificity even at the price of lower

M

sensitivity. This is the reason why we calculated thresholds with a specificity greater than 80

ED

%. Second, because interventions that could address all-cause mortality especially noncardiac mortality are difficult to imagine, we calculated cTnI threshold values for cardiac

PT

deaths defined from the analysis of individual medical records and the published definitions of cardiac/non cardiac deaths [29]. This analysis was performed without knowledge of the

CE

cTnI value.

AC

Threshold values for cTnI in this study were similar to those reported by Fellahi et al.[18] for the composite outcome of severe cardiac events and/or cardiac deaths based on maximization of the Youden index: 5.2 µg/l for CABG (versus 4.2 µg/l in our study) and 11.0 µg/l for non-CABG (versus 10.7 µg/l in our study). We chose to combine all non-CABG procedures into a single group because we did not observe significant differences in cTnI concentration across valve, combined valve and CABG, or ascending aorta procedures. Higher cTnI values and thresholds in non-CABG procedures are probably related to the

ACCEPTED MANUSCRIPT myotomy performed during these procedures and the longer duration of aortic cross-clamping [18]. If a threshold value for increased cTnI has to be defined and used for clinical decision making, our results clearly highlight that this must take into consideration the type of surgical procedure. In our study, the proportion of deceased patients with a cTnI greater than these thresholds was 51% overall (all-cause mortality) and 64% for cardiac deaths. In survivors,

CR IP T

this proportion was 20%. Similar proportions were reported by Fellahi et al. with 69% of patients with severe cardiac events or deaths having cTnI concentration greater than the threshold and 27% of patients without such events having cTnI concentration greater than the

AN US

threshold.

We also confirmed the independent statistical association between cTnI and one-year cardiac mortality with an adjusted odds ratio of 5.6 for cardiac deaths in patients with cTnI concentration greater than the defined thresholds (Supplemental Table 3). Importantly, in the

M

multivariable model, the duration of aortic cross-clamping was not statistically associated

ED

with an increased risk of death thus rendering the information “postoperative cTnI values > threshold” even more important. The cTnI value > threshold probably reflects individual

PT

susceptibility to ischemia/reperfusion and in this context, and adds independent prognostic value thus justifying its measurement. Indeed, the NRI is 0.288 for a model combining cTnI

CE

and EuroSCORE II compared with EuroSCORE II alone. This additive NRI value is in the

AC

high range of overall published NRI [40]. As a specific example concerning the use of cTn to improve risk prediction, one article reported an NRI was only 0.04 [41]. More importantly, for deceased patients, 42 % of the patients were reclassified to a higher risk by the cTnI values > threshold. This is a strong argument to support the clinical utility of measuring cTnI after cardiac surgery (risk stratification) provided that interventions are implemented (shift to management stratification).

ACCEPTED MANUSCRIPT Strengths and limitations of this study: an attempt to evolve from risk stratification to management stratification There is an obvious contradiction of this study since it makes proposals for a transition from risk stratification to management stratification. The proposals are by definition not validated but they open the way for future validation studies. In this respect, these proposals

CR IP T

targeted on a specific type of patients (see below) can be considered a strength. Forty-seven of the 53 patients (89 %) with cardiac deaths and cTnI values > threshold could have benefited from an intervention. For example, for patients with initially obvious/severe cardiac

AN US

dysfunction (e.g. cardiogenic shock, massive myocardial infarction), an intervention was actually implemented (several catecholamines, extracorporeal circulatory support) in 97% of patients with a cTnI greater than the thresholds (Table 6).

It is obvious that these

interventions would not have required cTnI results to be decided. For non-cardiac deaths, for

M

none of the patients with cTnI was greater than the thresholds an intervention was estimated

ED

as possible because of the diverse clinical contexts (e.g. delayed infection or abdominal complications) (Table 6). Our results suggest that if a decision to implement an intervention

PT

based on cTnI values above the threshold could have been taken, it could have been most relevant for patients with initially absent/mild cardiac dysfunction (Table 6). For those

CE

patients, the immediate postoperative period was characterized by mild cardiac dysfunction

AC

that could have been overlooked if the information “increased cTnI values” had not existed. If the practice of cTnI measurements after cardiac surgery is maintained, we suggest that a cTnI greater than the threshold value without an alarming clinical context (e.g. severe cardiac dysfunction) could be a warning sign to further assess cardiac function (e.g. repeated echocardiography) and decide interventions such as optimized monitoring (e.g. repeated measurements of cTn or natriuretic peptides) or optimized pharmacological support (e.g. vasodilators, angiotensin converting enzyme inhibitors) [42, 43]. Interestingly, in patients

ACCEPTED MANUSCRIPT with initially absent/mild cardiac dysfunction the median time to deaths was 17 days, suggesting that enough time would be available for interventions to be implemented. These numbers of deceased patients that could have benefited from an intervention if their cTnI values had been above the threshold must be compared to the 20% (N=629) of the survivors who had a cTnI above the threshold. Although one can argue that “all efforts are performed

CR IP T

for all patients at any time”, we consider that this may not reflect routine clinical practice. In this context, a patient without an alarming clinical context would benefit from cTnI measurements early (20th) after surgery thus allowing an optimization of care.

AN US

We acknowledge several limitations of our study. Some are technical: first, 5% of cTnI values were missing which may have biased our results. However, only slight differences were observed between patients with and without missing cTnI values. Second, the medical records of the 3,099 survivors were not analyzed and possible interventions in

M

survivors were not explored. This prevents us from conducting a definitive cost-consequence

ED

analysis to estimate the benefit of routine cTnI measurement in all postoperative patients to decide an intervention. Third, we did not analyze cardiac morbidity because most of previous

PT

research has focused on mortality. We cannot exclude that results would have been different for cTnI if morbidity had been taken into consideration. Fourth, we used a single-biomarker

CE

approach and inclusion of multiple markers such as BNP in the decision making process may lead to different results [7, 11, 44, 45]. In addition we used a single time-point measurement

AC

of cTnI (at the 20th postoperative hour) based on the fact that: (i) we wanted an early information thus allowing potential subsequent interventions and (ii) it was demonstrate that a single time-point value had similar prognostic values as multiple time-point [19]. Fifth, our proposals for interventions are speculative but we consider that they are the first step in the transition from risk stratification to risk management [16]. A similar analysis was recently performed for patients with increased cTn values after non cardiac surgery [26]. In that article

ACCEPTED MANUSCRIPT the authors enumerated the implemented interventions targeted at treating the triggering events (tachycardia, anemia, hypertension, hypotension, inflammation and sepsis, hypovolemia, fluid overload). Similar “basic” interventions should be implemented for all patients after cardiac surgery. They could be complemented by more “specific” interventions such as measurements of natriuretic peptides concentrations to titrate therapy or prescription

CR IP T

of addition cardiovascular medication/monitoring. A study published in 2014 used afterload reduction as goal-directed therapy in cardiac failure patients with optimized drug therapy: this

AN US

is an example of intervention that could be targeted on well-defined patients [46].

Conclusion

Our results, based on a novel approach of defining thresholds of postoperative cTnI

M

(prediction of cardiac deaths with a specificity > 80 %) is a first step in the transition from risk stratification to management stratification. To the best of our knowledge this is the first

ED

attempt to explore potential interventions following the availability of the information “cTnI

PT

values above the threshold”. Our exploratory analyses suggest that the sub-population of patients that could benefit from interventions is defined by mild alteration of cardiac function.

CE

This would allow time for both enhanced monitoring and/or improved pharmacological interventions. We hypothesize that any prospective study of potential interventions based on

AC

the information “cTnI values above the threshold” should in priority focus on this type of patients.

ACCEPTED MANUSCRIPT List of abbreviation: cTn: cardiac troponin, CABG: coronary artery bypass grafting; CPB: cardiopulmonary bypass, LVEF: left ventricular ejection fraction; UFH: unfractionated heparin, ACT: activated coagulation time; ICU: intensive care unit; AUCROC: area under the receiver operator

AC

CE

PT

ED

M

AN US

CR IP T

characteristic curve; NRI: net reclassification index

ACCEPTED MANUSCRIPT References: 1.

Writing Committee for the VSI, Devereaux PJ, Biccard BM, Sigamani A, Xavier D, Chan MTV, Srinathan SK, Walsh M, Abraham V, Pearse R et al: Association of Postoperative High-Sensitivity Troponin Levels With Myocardial Injury and 30-Day Mortality Among Patients Undergoing Noncardiac Surgery. JAMA 2017,

2.

CR IP T

317(16):1642-1651.

Haaf P, Drexler B, Reichlin T, Twerenbold R, Reiter M, Meissner J, Schaub N, Stelzig C, Freese M, Heinzelmann A et al: High-sensitivity cardiac troponin in the distinction of

Circulation 2012, 126(1):31-40. 3.

AN US

acute myocardial infarction from acute cardiac noncoronary artery disease.

Vascular Events In Noncardiac Surgery Patients Cohort Evaluation Study I, Devereaux PJ, Chan MT, Alonso-Coello P, Walsh M, Berwanger O, Villar JC, Wang CY, Garutti RI,

M

Jacka MJ et al: Association between postoperative troponin levels and 30-day

ED

mortality among patients undergoing noncardiac surgery. JAMA 2012, 307(21):2295-2304.

McCarthy CP, Vaduganathan M, Januzzi JL, Jr.: Type 2 Myocardial Infarction-

PT

4.

5.

CE

Diagnosis, Prognosis, and Treatment. JAMA 2018. Lasocki S, Provenchere S, Benessiano J, Vicaut E, Lecharny JB, Desmonts JM, Dehoux

AC

M, Philip I: Cardiac troponin I is an independent predictor of in-hospital death after adult cardiac surgery. Anesthesiology 2002, 97(2):405-411.

6.

Fellahi JL, Gue X, Richomme X, Monier E, Guillou L, Riou B: Short- and long-term prognostic value of postoperative cardiac troponin I concentration in patients undergoing coronary artery bypass grafting. Anesthesiology 2003, 99(2):270-274.

ACCEPTED MANUSCRIPT 7.

Provenchere S, Berroeta C, Reynaud C, Baron G, Poirier I, Desmonts JM, Iung B, Dehoux M, Philip I, Benessiano J: Plasma brain natriuretic peptide and cardiac troponin I concentrations after adult cardiac surgery: association with postoperative cardiac dysfunction and 1-year mortality. Crit Care Med 2006, 34(4):995-1000. van Geene Y, van Swieten HA, Noyez L: Cardiac troponin I levels after cardiac

CR IP T

8.

surgery as predictor for in-hospital mortality. Interact Cardiovasc Thorac Surg 2010, 10(3):413-416.

Vikenes K, Andersen KS, Melberg T, Farstad M, Nordrehaug JE: Long-term prognostic

AN US

9.

value of cardiac troponin I and T versus creatine kinase-MB mass after cardiac surgery in low-risk patients with stable symptoms. Am J Cardiol 2010, 106(6):780786.

Lurati Buse GA, Koller MT, Grapow M, Bolliger D, Seeberger M, Filipovic M: The

M

10.

ED

prognostic value of troponin release after adult cardiac surgery - a meta-analysis. Eur J Cardiothorac Surg 2010, 37(2):399-406. Lurati Buse GA, Bolliger D, Seeberger E, Kasper J, Grapow M, Koller MT, Seeberger

PT

11.

CE

MD, Filipovic M: Troponin T and B-type natriuretic peptide after on-pump cardiac surgery: prognostic impact on 12-month mortality and major cardiac events after

AC

adjustment for postoperative complications. Circulation 2014, 130(12):948-957.

12.

Mauermann E, Bolliger D, Fassl J, Grapow M, Seeberger EE, Seeberger MD, Filipovic M, Lurati Buse GAL: Postoperative High-Sensitivity Troponin and Its Association With 30-Day and 12-Month, All-Cause Mortality in Patients Undergoing On-Pump Cardiac Surgery. Anesth Analg 2017, 125(4):1110-1117.

ACCEPTED MANUSCRIPT 13.

Domanski MJ, Mahaffey K, Hasselblad V, Brener SJ, Smith PK, Hillis G, Engoren M, Alexander JH, Levy JH, Chaitman BR et al: Association of myocardial enzyme elevation and survival following coronary artery bypass graft surgery. JAMA 2011, 305(6):585-591.

14.

Lurati Buse GA, Koller MT, Grapow M, Bolliger D, Seeberger M, Filipovic M: The

Eur J Cardiothorac Surg, 37(2):399-406. 15.

Shahian DM, Grover FL: Biomarkers and risk models in cardiac surgery. Circulation

Cook DA, Sherbino J, Durning SJ: Management Reasoning: Beyond the Diagnosis. JAMA 2018, 319(22):2267-2268.

17.

AN US

2014, 130(12):932-935. 16.

CR IP T

prognostic value of troponin release after adult cardiac surgery - a meta-analysis.

Collinson PO, Heung YM, Gaze D, Boa F, Senior R, Christenson R, Apple FS: Influence

M

of population selection on the 99th percentile reference value for cardiac troponin

18.

ED

assays. Clin Chem 2012, 58(1):219-225. Fellahi JL, Hedoire F, Le Manach Y, Monier E, Guillou L, Riou B: Determination of the

PT

threshold of cardiac troponin I associated with an adverse postoperative outcome

CE

after cardiac surgery: a comparative study between coronary artery bypass graft, valve surgery, and combined cardiac surgery. Crit Care 2007, 11(5):R106. Fellahi JL, Hanouz JL, Gue X, Monier E, Guillou L, Riou B: Kinetic analysis of cardiac

AC

19.

troponin I release is no more accurate than a single 24-h measurement in predicting in-hospital outcome after cardiac surgery. Eur J Anaesthesiol 2008, 25(6):490-497.

20.

Apple FS, Collinson PO, Biomarkers ITFoCAoC: Analytical characteristics of highsensitivity cardiac troponin assays. Clin Chem 2012, 58(1):54-61.

ACCEPTED MANUSCRIPT 21.

Ray P, Le Manach Y, Riou B, Houle TT: Statistical evaluation of a biomarker. Anesthesiology 2010, 112(4):1023-1040.

22.

Vickers AJ, Elkin EB: Decision curve analysis: a novel method for evaluating prediction models. Med Decis Making 2006, 26(6):565-574.

23.

Ausset S, Auroy Y, Lambert E, Vest P, Plotton C, Rigal S, Lenoir B, Benhamou D:

outcome. Eur J Anaesthesiol 2008, 25(2):158-164. 24.

CR IP T

Cardiac troponin I release after hip surgery correlates with poor long-term cardiac

Foucrier A, Rodseth R, Aissaoui M, Ibanes C, Goarin JP, Landais P, Coriat P, Le Manach

AN US

Y: The long-term impact of early cardiovascular therapy intensification for

postoperative troponin elevation after major vascular surgery. Anesth Analg 2014, 119(5):1053-1063. 25.

Ausset S, Auroy Y, Verret C, Benhamou D, Vest P, Cirodde A, Lenoir B: Quality of

M

postoperative care after major orthopedic surgery is correlated with both long-

113(3):529-540.

van Waes JA, Grobben RB, Nathoe HM, Kemperman H, de Borst GJ, Peelen LM, van

PT

26.

ED

term cardiovascular outcome and troponin Ic elevation. Anesthesiology 2010,

CE

Klei WA, Cardiac Health After Surgery I: One-Year Mortality, Causes of Death, and Cardiac Interventions in Patients with Postoperative Myocardial Injury. Anesth

AC

Analg 2016, 123(1):29-37.

27.

Nashef SA, Roques F, Sharples LD, Nilsson J, Smith C, Goldstone AR, Lockowandt U: EuroSCORE II. Eur J Cardiothorac Surg 2012, 41(4):734-744; discussion 744-735.

28.

Bruneel A, Dehoux M, Barnier A, Boutten A: External evaluation of the Dimension Vista 1500(R) intelligent lab system. J Clin Lab Anal 2012, 26(5):384-397.

ACCEPTED MANUSCRIPT 29.

Hicks KA, Tcheng JE, Bozkurt B, Chaitman BR, Cutlip DE, Farb A, Fonarow GC, Jacobs JP, Jaff MR, Lichtman JH et al: 2014 ACC/AHA Key Data Elements and Definitions for Cardiovascular Endpoint Events in Clinical Trials: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Data Standards (Writing Committee to Develop Cardiovascular Endpoints Data Standards).

30.

CR IP T

Circulation 2015, 132(4):302-361.

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

31.

AN US

reclassification and beyond. Stat Med 2008, 27(2):157-172; discussion 207-112. Alba AC, Agoritsas T, Walsh M, Hanna S, Iorio A, Devereaux PJ, McGinn T, Guyatt G: Discrimination and Calibration of Clinical Prediction Models: Users' Guides to the Medical Literature. JAMA 2017, 318(14):1377-1384.

Nishimura RA, Otto CM, Bonow RO, Carabello BA, Erwin JP, 3rd, Guyton RA, O'Gara

M

32.

ED

PT, Ruiz CE, Skubas NJ, Sorajja P et al: 2014 AHA/ACC Guideline for the Management of Patients With Valvular Heart Disease: executive summary: a report of the

PT

American College of Cardiology/American Heart Association Task Force on Practice

33.

CE

Guidelines. Circulation 2014, 129(23):2440-2492. Altman DG, Vergouwe Y, Royston P, Moons KG: Prognosis and prognostic research:

AC

validating a prognostic model. BMJ 2009, 338:b605.

34.

Levitt DG, Schnider TW: Human physiologically based pharmacokinetic model for propofol. BMC Anesthesiol 2005, 5(1):4.

35.

Gepts E, Shafer SL, Camu F, Stanski DR, Woestenborghs R, Van Peer A, Heykants JJ: Linearity of pharmacokinetics and model estimation of sufentanil. Anesthesiology 1995, 83(6):1194-1204.

ACCEPTED MANUSCRIPT 36.

Bignami E, Guarnieri M, Pieri M, De Simone F, Rodriguez A, Cassara L, Lembo R, Landoni G, Zangrillo A: Volatile anaesthetics added to cardiopulmonary bypass are associated with reduced cardiac troponin. Perfusion 2017, 32(7):547-553.

37.

Forestier F, Hirschi M, Rouget P, Rigal JC, Videcoq M, Girardet P, Durand M, Maitrasse B, Girard C, Lehot JJ et al: Propofol and sufentanil titration with the bispectral index

CR IP T

to provide anesthesia for coronary artery surgery. Anesthesiology 2003, 99(2):334346. 38.

Menkis AH, Martin J, Cheng DC, Fitzgerald DC, Freedman JJ, Gao C, Koster A,

AN US

Mackenzie GS, Murphy GJ, Spiess B et al: Drug, devices, technologies, and

techniques for blood management in minimally invasive and conventional cardiothoracic surgery: a consensus statement from the International Society for Minimally Invasive Cardiothoracic Surgery (ISMICS) 2011. Innovations (Phila) 2012,

Yaffee DW, Smith DE, 3rd, Ursomanno PA, Hill FT, Galloway AC, DeAnda A, Grossi EA:

ED

39.

M

7(4):229-241.

Management of blood transfusion in aortic valve surgery: impact of a blood

Leening MJ, Vedder MM, Witteman JC, Pencina MJ, Steyerberg EW: Net

CE

40.

PT

conservation strategy. Ann Thorac Surg 2014, 97(1):95-101.

reclassification improvement: computation, interpretation, and controversies: a

AC

literature review and clinician's guide. Ann Intern Med 2014, 160(2):122-131.

41.

deFilippi CR, de Lemos JA, Christenson RH, Gottdiener JS, Kop WJ, Zhan M, Seliger SL: Association of serial measures of cardiac troponin T using a sensitive assay with incident heart failure and cardiovascular mortality in older adults. JAMA 2010, 304(22):2494-2502.

ACCEPTED MANUSCRIPT 42.

Ekeloef S, Alamili M, Devereaux PJ, Gogenur I: Troponin elevations after non-cardiac, non-vascular surgery are predictive of major adverse cardiac events and mortality: a systematic review and meta-analysis. Br J Anaesth 2016, 117(5):559-568.

43.

Gordin JS, Fonarow GC: New medications for heart failure. Trends Cardiovasc Med 2016, 26(6):485-492. Mauermann E, Bolliger D, Fassl J, Grapow M, Seeberger EE, Seeberger MD, Filipovic

CR IP T

44.

M, Lurati Buse GAL: Absolute Postoperative B-Type Natriuretic Peptide

Concentrations, but Not Their General Trend, Are Associated With 12-Month, All-

AN US

Cause Mortality After On-Pump Cardiac Surgery. Anesth Analg 2017, 125(3):753761. 45.

Brynildsen J, Petaja L, Pettila V, Nygard S, Vaara ST, Linko R, Okkonen M, Hagve TA, Soininen L, Suojaranta-Ylinen R et al: The predictive value of NT-proBNP and hs-TnT

Borlaug BA, Olson TP, Abdelmoneim SS, Melenovsky V, Sorrell VL, Noonan K, Lin G,

ED

46.

M

for risk of death in cardiac surgical patients. Clin Biochem 2018, 53:65-71.

Redfield MM: A randomized pilot study of aortic waveform guided therapy in

AC

CE

PT

chronic heart failure. J Am Heart Assoc 2014, 3(2):e000745.

ACCEPTED MANUSCRIPT Figure Legend:

AC

CE

PT

ED

M

AN US

CR IP T

Figure 1: Selection of the study sample.

Abbreviations: cTnI: postoperative cardiac troponin I concentration; CABG: coronary artery bypass grafting

ACCEPTED MANUSCRIPT Table 1: Definitions of the causes of death and potential interventions identified during the assessment of medical records (adapted from van Waes et al.(26)) Causes of death 1- Cardiac death

CR IP T

- Myocardial infarction - Or de novo severe and persistent myocardial dysfunction requiring sustained therapy (e.g., multiple high dose inotropes, mechanical cardio-circulatory support) - Sudden death

AN US

- Less severe or transient cardiac dysfunction that could have worsened the consequences of a postoperative complication (e.g., tamponnade, sepsis or septic shock) or of a preoperative organ dysfunction (e.g., chronically altered renal function). 2- Non-cardiac death

M

No myocardial dysfunction or mild myocardial dysfunction that did not contribute to death or

ED

death that could be unequivocally attributed to another postoperative event (e.g., mediastinitis, severe bleeding with massive transfusion without myocardial dysfunction

PT

documented by echocardiography). Potential interventions

CE

1-Haemodynamic optimization: any enhanced/optimized cardiac support such as inotropes,

AC

vasodilators, or mechanical cardio-circulatory support 2-Optimized basic interventions: correction of perioperative triggers of myocardial oxygen imbalance such as anemia, tachycardia, hypo/hypertension, inflammation/sepsis, fluid overload, hypoxemia (24). 3-Optimization of drugs therapy: β-blockers, antiplatelet agents, statins, angiotensinconverting enzyme inhibitors, diuretics, vasodilators and others new drugs (41)

ACCEPTED MANUSCRIPT 2- Enhanced or prolonged cardiac diagnosis/ monitoring Repeated/ more frequent echocardiographic examination; postoperative natriuretic peptides measurements, prolonged stay in ICU or in telemetry unit, out-of-hospital follow-up for titration of cardiovascular medication on specific hemodynamic goals (42)

CR IP T

3- None

Table 2: Characteristics of the two groups of surgical procedures CABG

Non-CABG

P-value

(N=1,981)

AN US

(N=1,247)

Patients characteristics Age (year)

65 (58-73)

66 (54-76)

0.78

1,047 (84.0%)

1,160 (58.6%)

< 0.0001

27.0 (24.5-30.0)

25.7 (23.1-28.9)

< 0.0001

886 (71.1%)

1,006 (50.8%)

< 0.0001

525 (42.1%)

338 (17.1%)

< 0.0001

116 (9.3%)

182 (9.2%)

0.95

History of coronary artery disease

961 (77.1%)

222 (11.2%)

< 0.0001

History of peripheral vascular disease

266 (21.3%)

148 (7.5%)

< 0.0001

Previous cardiac surgical intervention ≥ 1

15 (1.2%)

253 (12.8%)

< 0.0001

Male gender

ED

Body mass index (kg/m²)

M

(missing=4)

(missing=14)

PT

Medical and surgical history

CE

Hypertension Diabetes

AC

History of chronic obstructive pulmonary disease

Preoperative testing

ACCEPTED MANUSCRIPT Haemoglobin concentration (g/dl) 13.9 (12.6-14.8)

13.3 (12.0-14.4)

< 0.0001

76 (61-91)

74 (58-91)

0.018

55 (45-62)

60 (55-69)

< 0.0001

(missing=13) Creatinine clearance (ml/min) (missing=26) (*)

(missing=133) Current procedure EuroSCORE II

1.5 (0.9-2.8)

Duration of aortic cross-clamping (min)

Postoperative cTnI concentration (µg/l) Deaths

Cause of death

<0.0001

50 (37-69)

<0.0001

2.2 (1.4-3.6)

4.8 (2.7-9.4)

< 0.0001

34 (2.7%)

95 (4.8%)

0.003

M

Any death

3.2 (1.7-6.2)

AN US

37 (30-44) (missing=100)

CR IP T

Left ventricular ejection fraction (%)

0.90

13 (38.2%)

35 (36.8%)

PT

ED

Cardiac deaths with initially

10 (29.4%)

25 (26.3%)

- Non cardiac-deaths

11 (32.4%)

35 (36.8%)

Time to death (days)

23 (8-48)

21 (8-57)

obvious/severe cardiac dysfunction

Cardiac deaths with initially absent/mild

AC

CE

cardiac dysfunction

0.86 0.27

Possible interventions - None

12 (35.3%)

46 (48.4%)

- Enhanced monitoring

5 (14.7%)

17 (17.9%)

- Haemodynamic optimization

17 (50.0%)

32 (48.4%)

ACCEPTED MANUSCRIPT Footnote: Results are expressed as median (interquartile range) or number (%). P-values are from Wilcoxon tests and Chi-square or Fisher exact tests Abbreviations: CABG: coronary artery bypass grafting; cTnI: postoperative cardiac troponin I (*) Creatinine clearance was calculated with the Modification of Diet in Renal Disease

CR IP T

formula Table 3: Postoperative cTnI values, time to death, EuroSCORE II values, and type of surgery according to causes of death Cardiac

Non

deaths with

deaths with

cardiac-

initially

initially

deaths

obvious/severe

absent/mild

(3)

cardiac

cardiac

M

dysfunction (1) dysfunction (2)

cTnI

13.7

concentration

12.3

4.0

versus versus versus 2

3

9

17

45

death (days)

(3-24)

(10-45)

(26-97)

EuroSCORE

8.0

7.1

6.8

II

(4.7-14.2)

(4.7-11.6)

(3.6-10.9)

CE

0.10 0.0004

(2.2-16.6)

AC

2

3

N=35

(3.6-18.9)

Time to

1

N=46

(5.3-65.9)

(µg/l)

value

1

ED

PT

N=48

P-

AN US

Cardiac

Type of

0.053 <0.0001

(ns)

(ns)

<0.0001

0.019

<0.0001

0.0006

0.42

--

--

--

--

--

--

0.88

surgery - CABG

13 (27%)

10 (29%)

11 (24%)

ACCEPTED MANUSCRIPT - Non-CABG

35 (73%)

25 (71%)

35 (76%)

--

--

--

Footnote: Results are expressed as median (interquartile range) or number (%). The P-values are from of a Chi-square test or a Kruskal-Wallis test followed by Wilcoxon tests (ns indicate not statistically significant with Bonferroni corrections for the 3 comparisons)

CR IP T

Abbreviations: CABG: coronary artery bypass grafting; cTnI: cardiac troponin I; N: number. Table 4: Postoperative cardiac troponin I thresholds for 3 selected specificity values for oneyear cardiac mortality according to the type of surgery

(N=1,224 survivors and 23 non-

80%

85%

90%

4.2 (3.9 to 4.5)

5.0 (4.6 to 5.4)

6.8 (5.9 to 7.4)

60.9%

56.6%

39.1%

10.7

13.0

17.1

(10 to 11.3)

(12.0 to 14.1)

(15.6 to 18.4)

65.0%

63.3%

45.0%

ED

CE

PT

Specificity≥

M

survivors)

Sensitivity

Specificity≥

AN US

CABG procedures

cTnI threshold (µg/l) (95% CI)

Specificity≥

Non-CABG procedures

AC

(N=1,921 survivors and 60 nonsurvivors)

cTnI threshold (µg/l) (95% CI)

Sensitivity

ACCEPTED MANUSCRIPT

Footnote: Results are expressed as median (interquartile range) or number Abbreviations: CABG: coronary artery bypass grafting; CI: confidence interval; cTnI: cardiac troponin I; N: number.

Model 1

Model 2 OR (95% CI)

EuroSCORE II (per unit)

1.12 (1.09-1.14)

Model performance

C-index (95% CI)

1.09 (1.06-1.12)

cTnI>threshold (*)

5.46 (3.40-8.77)

ED

(0.75-0.84) 0.001

PT

Hosmer-Lemeshow test P-value

OR (95% CI)

EuroSCORE II

M

0.80

Variables in the model

AN US

Variables in the model

CR IP T

Table 5: Logistic regression models used to compute the net reclassification index

0.82 (0.77-0.87) 0.006 0.288

Absolute net reclassification index

14.1%

AC

CE

Additive net reclassification index

Abbreviations: CI: confidence interval; OR: odds ratio (*) The threshold value is 4.2 µg/l for CABG patients and 10.7 µg/l for non-CABG patients

ACCEPTED MANUSCRIPT Table 6: Potential interventions according to the causes of death and type of surgery and proportion of non-survivors and survivors with a cTnI concentration greater than the threshold with 80% specificity for one-year cardiac mortality Non-survivors

Survivors

Cardiac deaths

Non-

with initially

with initially

cardiac-

obvious/severe

absent/mild

cardiac

cardiac

dysfunction (1)

dysfunction (2)

CABG and non-CABG

N=48

N=35

- Potential interventions

45 (94%)

- cTnI>threshold (*)

31 (65%)

cTnI>threshold

N=13

25 (71%)

1 (2%)

<0.0001

--

22 (63%)

13 (28%)

0.0005

607

17/22 (77%)

0/13 (0%)

(20%) <0.0001

--

13 (100%)

8 (80%)

1 (9%)

<0.0001

--

9 (69%)

5 (50%)

0 (0%)

0.002

237

ED

N=1,213

(20%)

9/9 (100%)

5/5 (100%)

0/0 (0%)

N=35

N=25

N=35

- Potential intervention

32 (91%)

17 (68%)

0 (0%)

<0.0001

--

- cTnI>threshold (*)

22 (63%)

17 (68%)

13 (37%)

0.03

370

CE

- Potential intervention when

N=3,099

N=11

PT

- cTnI>threshold (*)

N=46

N=10

CABG - Potential intervention

deaths (3)

AN US

30/31 (97%)

M

- Potential interventions when

P-value

CR IP T

Cardiac deaths

--

--

cTnI>threshold

AC

Non-CABG

N=1,886

(20%) - Potential intervention when cTnI> threshold

21/22 (95%)

12/17 (71%)

0/13 (0%)

<0.0001

--

ACCEPTED MANUSCRIPT Footnote: Results are expressed as number (%). Interventions include monitoring or optimization. The threshold value is 4.2 µg/l for CABG and 10.7 µg/l for non-CABG. Pvalues are from Chi-square tests.

AC

CE

PT

ED

M

AN US

CR IP T

Abbreviations: cTnI: cardiac troponin I; N: number