Original Report: Patient-Oriented, Translational Research American
Journal of
Nephrology
Am J Nephrol 2014;39:195–203 DOI: 10.1159/000358495
Received: August 13, 2013 Accepted: January 3, 2014 Published online: February 15, 2014
Acute Kidney Injury and Mortality following Ventricular Assist Device Implantation Abhijit Naik a Shahab A. Akhter e Savitri Fedson b Valluvan Jeevanandam c Jonathan D. Rich d Jay L. Koyner a Sections of a Nephrology and b Cardiology, Department of Medicine, and c Section of Cardiac and Thoracic Surgery, Department of Surgery, University of Chicago, and d Division of Cardiology, Department of Medicine, Northwestern University, Chicago, Ill., and e Division of Cardiothoracic Surgery, Department of Surgery, University of Wisconsin, Madison, Wisc., USA
Abstract Background: Ventricular assist devices (VADs) are increasingly common, and their surgical implantation predisposes patients to an increased risk of acute kidney injury (AKI). We sought to evaluate the incidence, risk factors and short- and long-term all-cause mortality of patients with AKI following VAD implantation. Methods: We identified all patients who underwent VAD implantation at the University of Chicago between January 1, 2008, and January 31, 2012. We evaluated the incidence of AKI, defined as a ≥50% increase in serum creatinine over the first 7 postoperative days (RIFLE RiskCreatinine). A logistic regression model was used to identify risk factors for the development of AKI, and a Cox proportional hazards model was used to examine factors associated with 30-day and 365-day all-cause mortality. Results: A total of 157 eligible patients had VAD implantations with 44 (28%) developing postimplantation AKI. In a multivariate analysis, only diabetes mellitus [odds ratio = 2.25 (1.03–4.94), p = 0.04] was identified as a significant predictor of postopera-
© 2014 S. Karger AG, Basel 0250–8095/14/0393–0195$39.50/0 E-Mail [email protected] www.karger.com/ajn
tive AKI. Using a multivariable model censored for heart transplantation, only AKI [hazard ratio, HR = 3.01 (1.15–7.92), p = 0.03] and cardiopulmonary bypass time [HR = 1.01 (1.001–1.02), p = 0.02] were independent predictors of 30day mortality. Preoperative body mass index [HR = 0.95 (0.90–0.99), p = 0.03], preoperative diabetes mellitus [HR = 1.89 (1.07–3.35), p = 0.03] and postimplantation AKI [HR = 1.85 (1.06–3.21), p = 0.03] independently predicted 365-day mortality. Conclusion: AKI is common following VAD implantation and is an independent predictor of 30-day and 1-year all-cause mortality. © 2014 S. Karger AG, Basel
Introduction
Acute kidney injury (AKI) is common following traditional cardiac surgery (coronary artery bypass grafting and/or valve replacement) and is an independent risk factor for postoperative mortality [1, 2]. Additionally, AKI has been associated with higher patient morbidity, longer length of stay and higher costs [1, 3–5]. Multiple preoperative risk stratification systems have been developed to predict the development of AKI following traditional carJay L. Koyner, MD Section of Nephrology, University of Chicago 5841 South Maryland Avenue, Suite S-504, MC 5100 Chicago, IL 60637 (USA) E-Mail jkoyner @ uchicago.edu
Downloaded by: University of Pittsburgh 198.143.32.65 - 8/10/2015 6:13:05 PM
Key Words Acute kidney injury · Cardiac surgery · Mortality · Ventricular assist device · RIFLE criteria
Methods The Institutional Review Board of The University of Chicago approved this study. We performed a single-center retrospective chart review of all patients who underwent a VAD placement at our institution from January 1, 2008, through January 31, 2012. Patients with end-stage renal disease, those who were already on renal replacement therapy (RRT) for AKI prior to VAD implantation and those who died intraoperatively were excluded. Observations and Measurements Demographic, biochemical and clinical profiles were obtained from the INTERMACS national registry, Society of Thoracic Surgeons (STS) database and the University of Chicago electronic medical record (EMR) [20]. We collected preoperative data on age, race, sex, body mass index (BMI), baseline renal function (estimated glomerular filtration rate, eGFR, using MDRD) [21], past medical history, hemodynamic numbers, intraoperative data including cardiopulmonary bypass time and receipt of blood products, postoperative data including length of stay, postoperative complications (e.g. sepsis, thromboembolic events), need for RRT and in-hospital mortality. Baseline creatinine was defined as the listing of creatinine data obtained from the INTERMACS registry; if INTERMACS did not contain a preoperative serum creatinine, then the last serum creatinine in the EMR prior to device implantation was used. We collected infor-
196
Am J Nephrol 2014;39:195–203 DOI: 10.1159/000358495
mation about the indication for VAD implantation (bridge to transplant or destination therapy) as well as the type of device implanted. Outcomes The primary outcome of AKI was defined as a 50% rise in serum creatinine over the preoperative baseline during the first 7 postoperative days as per the creatinine-based RIFLE (Risk of renal dysfunction, Injury to the kidney, Failure or Loss of kidney function, and End-stage renal disease; AKI assessed over 7 days) Risk criteria [22]. The Acute Kidney Injury Network (AKIN; AKI assessed over 48 h) creatinine criteria [23, 24] and urine output criteria were utilized for secondary analyses. All-cause mortality was monitored over 1 year after implantation, with data reported at 30 and 365 days. Statistical Methods Qualitative data were recorded in a categorical fashion, and quantitative covariates were measured as continuous variables. Differences between characteristics were analyzed either by an unpaired t test or the Wilcoxon signed-rank test as appropriate. Categorical variables were assessed by χ2 analysis or Fisher’s exact test and were reported with their corresponding odds ratio (OR) and 95% confidence intervals (CI). Continuous variables are represented with mean and their 95% CI values and hazard ratio (HR) when pertaining to survival data. To assess determinants of AKI, we first performed a univariate logistic regression analysis. Significant predictors (p < 0.10) were then included in a multivariate model. Survival data were analyzed using Cox proportional hazards modeling. All survival data were censored for cardiac transplantation. We analyzed survival at 30 and 365 days. Significant variables (p < 0.10) were then included in a multivariate model. All statistical tests were two-sided with α set at 0.05 for statistical significance. STATA 11.2 (StataCorp LP, College Station, Tex., USA) was used for all data analysis.
Results
Baseline Characteristics of Patients with AKI We identified 168 patients who underwent VAD implantation at our institution during the study period. Eleven subjects were excluded from the analysis, with 8 patients having previous end-stage renal disease, 2 requiring preimplant RRT for AKI and 1 subject with intraoperative mortality. In the final cohort of 157, a total of 44 (28%) patients developed AKI. Table 1 demonstrates the pre-, intra- and postoperative characteristics of those with and without AKI. Prior to implantation, there was no difference in age, baseline renal function (serum creatinine and eGFR), New York Heart Association (NYHA) CHF class, preoperative hemodynamics, INTERMACS score, the presence of cardiogenic shock or inotrope use in those with and without AKI. However, the patients with AKI tended to have more diabetes and cerebrovasNaik /Akhter /Fedson /Jeevanandam /Rich / Koyner
Downloaded by: University of Pittsburgh 198.143.32.65 - 8/10/2015 6:13:05 PM
diac surgery; however, their role and application in the setting of ventricular assist device (VAD) implantation is unclear [6–9]. Since the Randomized Evaluation of Mechanical Assistance for the Treatment of Congestive Heart Failure (REMATCH) trial [10], there has been an increased number of VADs being implanted, with nearly 6,000 VADs being implanted since June 2006 [11]. The first-generation VADs were pulsatile and presumably more ‘physiologic’; however, they were associated with higher infection and thrombosis rates, lower durability and even lower cardiac transplant rates in comparison to continuous flow devices [12]. This has led to a transition away from pulsatile flow devices to continuous flow devices which now make up for >90% implants in the United States [13]. While the increasing numbers of implants have improved our understanding of the impact of AKI during the perioperative period, its etiology, predictors and short- and long-term consequences require further investigation. Previous studies have been difficult to interpret, as many have not used consensus definitions of AKI [14–19]. Utilizing consensus definitions of AKI, we performed a single-center, retrospective, longitudinal cohort study to evaluate the incidence of AKI following VAD implantation, identify risk factors as well as assess its impact on 30- and 365-day survival.
Table 1. Pre-, intra- and postoperative characteristics of patients with and without AKI
a Baseline clinical characteristics of VAD patients by AKI status Operative age, years 58.3 ± 12.7 Weight, kg 89.2 ± 23.3 Race 68 (60.2) Caucasian 36 (31.9) African-American Asian 4 (3.5) Unknown 5 (4.4) Males 89 (78.8) BMI 28.9 ± 6.3 Baseline serum creatinine, mg/dl 1.67 ± 1.01 Baseline eGFR (MDRD), ml/min 49.3 ± 26.7 Preoperative hematocrit 34.2 ± 6.2 HbA1c 6.9 ± 1.2 Diabetes 35 (31.0) PVD 7 (6.2) Hypertension 48 (42.5) CVD 9 (8.0) Hyperlipidemia 61 (54) CHF 101 (89.4) Ejection fraction 17.7 ± 7.6 Cardiogenic shock 34 (30.1) VAD device HeartMate II 88 (77.9) HeartWare 19 (16.8) BiVAD 3 (2.7) PVAD/RVAD 0 (0) 3 (2.6) HeartMate XVE Goal of implant Bridge to transplant 54 (47.8) Destination therapy 59 (52.2) NYHA classification (n =133) 1 (1.1) Class I 1 (1.1) Class II 9 (9.4) Class III 84 (88.4) Class IV INTERMACS score (n = 147) 18 (17) 1 46 (43.39) 2 25 (23.58) 3 15 (14.15) 4 2 (1.88) ≥5 Preoperative hemodynamics (n = 111) RAP, mm Hg 13.7 ± 6.6 RVSP, mm Hg 53.3 ± 14.8 PASP, mm Hg 54.3 ± 15.1 PADP, mm Hg 26.9 ± 8.5 PCWP, mm Hg 25.9 ± 8.7 Cardiac output, l/min 3.78 ± 1.7 Cardiac index, l/min/m2 1.84 ± 0.6
AKI and Mortality following VAD Implantation
Am J Nephrol 2014;39:195–203 DOI: 10.1159/000358495
AKI (n = 44)
p value
56.4 ± 12.7 84.3 ± 25.3
0.4 0.25
23 (52.3) 18 (40.9) 0 (0) 3 (6.8) 34 (77.3) 28.4 ± 7.3 1.58 ± 0.71 50.4 ± 22.1 36.5 ± 5.3 6.9 ± 1.2 23 (52.3) 3 (6.8) 23 (52.2) 9 (20.5) 21 (47.7) 43 (97.7) 16.5 ± 7.9 16 (36.36)
0.39
35 (79.6) 5 (11.4) 1 (2.2) 1 (2.2) 2 (4.6)
0.84 0.62 0.57 0.82 0.05 0.84 0.01 0.85 0.27 0.03 0.48 0.09 0.57 0.68
0.48
21 (48.8) 22 (51.2)
0.90
1 (2.6) 1 (2.6) 4 (10.5) 32 (84.2)
0.58
11 (26.82) 14 (34.14) 10 (24.39) 6 (14.63) 0 14.5 ± 7.5 51 ± 12.7 51.6 ± 13.9 26.5 ± 7.2 23.6 ± 10.2 3.60 ± 1.4 1.82 ± 0.7
0.63
0.60 0.45 0.40 0.84 0.25 0.62 0.87
197
Downloaded by: University of Pittsburgh 198.143.32.65 - 8/10/2015 6:13:05 PM
No AKI (n = 113)
Table 1 (continued)
No AKI (n = 113)
AKI (n = 44)
p value
b Intra- and postoperative outcomes in patients with and without AKI following VAD implantation Intraoperative outcomes CPB time, min 131.0 ± 36.5 133.9 ± 49.4 0.69 Patients who received intraoperative blood products 88 (77.9) 26 (59.1) 0.02 Patients who received intraoperative PRBC 61 (54.0) 24 (54.6) 0.94 Units of intraoperative PRBC used 1 (0 – 3) 1 (0 – 6) 0.45 Postoperative outcomes Net fluid balance at 7 days, liters –6.65 ± 5.62 – 4.06 ± 5.78 0.02 ICU length of stay, days 6.4 ± 7.7 11.2 ± 11.0 0.006 Total length of stay, days 29.2 ± 59.4 34.5 ± 46.1 0.60 Pneumonia 4 (3.5) 4 (9.1) 0.16 Reoperation for bleeding 11 (9.7) 3 (6.8) 0.57 Gastrointestinal bleeding 6 (5.3) 3 (6.8) 0.71 Blood product transfusion 70 (68.6) 23 (57.5) 0.21 Units of PRBC transfused 3 (2 – 6) 7 (4–16) 0.002 In-hospital mortality 9 (8.0) 14 (31.8)
Journal of
Nephrology
Am J Nephrol 2014;39:195–203 DOI: 10.1159/000358495
Received: August 13, 2013 Accepted: January 3, 2014 Published online: February 15, 2014
Acute Kidney Injury and Mortality following Ventricular Assist Device Implantation Abhijit Naik a Shahab A. Akhter e Savitri Fedson b Valluvan Jeevanandam c Jonathan D. Rich d Jay L. Koyner a Sections of a Nephrology and b Cardiology, Department of Medicine, and c Section of Cardiac and Thoracic Surgery, Department of Surgery, University of Chicago, and d Division of Cardiology, Department of Medicine, Northwestern University, Chicago, Ill., and e Division of Cardiothoracic Surgery, Department of Surgery, University of Wisconsin, Madison, Wisc., USA
Abstract Background: Ventricular assist devices (VADs) are increasingly common, and their surgical implantation predisposes patients to an increased risk of acute kidney injury (AKI). We sought to evaluate the incidence, risk factors and short- and long-term all-cause mortality of patients with AKI following VAD implantation. Methods: We identified all patients who underwent VAD implantation at the University of Chicago between January 1, 2008, and January 31, 2012. We evaluated the incidence of AKI, defined as a ≥50% increase in serum creatinine over the first 7 postoperative days (RIFLE RiskCreatinine). A logistic regression model was used to identify risk factors for the development of AKI, and a Cox proportional hazards model was used to examine factors associated with 30-day and 365-day all-cause mortality. Results: A total of 157 eligible patients had VAD implantations with 44 (28%) developing postimplantation AKI. In a multivariate analysis, only diabetes mellitus [odds ratio = 2.25 (1.03–4.94), p = 0.04] was identified as a significant predictor of postopera-
© 2014 S. Karger AG, Basel 0250–8095/14/0393–0195$39.50/0 E-Mail [email protected] www.karger.com/ajn
tive AKI. Using a multivariable model censored for heart transplantation, only AKI [hazard ratio, HR = 3.01 (1.15–7.92), p = 0.03] and cardiopulmonary bypass time [HR = 1.01 (1.001–1.02), p = 0.02] were independent predictors of 30day mortality. Preoperative body mass index [HR = 0.95 (0.90–0.99), p = 0.03], preoperative diabetes mellitus [HR = 1.89 (1.07–3.35), p = 0.03] and postimplantation AKI [HR = 1.85 (1.06–3.21), p = 0.03] independently predicted 365-day mortality. Conclusion: AKI is common following VAD implantation and is an independent predictor of 30-day and 1-year all-cause mortality. © 2014 S. Karger AG, Basel
Introduction
Acute kidney injury (AKI) is common following traditional cardiac surgery (coronary artery bypass grafting and/or valve replacement) and is an independent risk factor for postoperative mortality [1, 2]. Additionally, AKI has been associated with higher patient morbidity, longer length of stay and higher costs [1, 3–5]. Multiple preoperative risk stratification systems have been developed to predict the development of AKI following traditional carJay L. Koyner, MD Section of Nephrology, University of Chicago 5841 South Maryland Avenue, Suite S-504, MC 5100 Chicago, IL 60637 (USA) E-Mail jkoyner @ uchicago.edu
Downloaded by: University of Pittsburgh 198.143.32.65 - 8/10/2015 6:13:05 PM
Key Words Acute kidney injury · Cardiac surgery · Mortality · Ventricular assist device · RIFLE criteria
Methods The Institutional Review Board of The University of Chicago approved this study. We performed a single-center retrospective chart review of all patients who underwent a VAD placement at our institution from January 1, 2008, through January 31, 2012. Patients with end-stage renal disease, those who were already on renal replacement therapy (RRT) for AKI prior to VAD implantation and those who died intraoperatively were excluded. Observations and Measurements Demographic, biochemical and clinical profiles were obtained from the INTERMACS national registry, Society of Thoracic Surgeons (STS) database and the University of Chicago electronic medical record (EMR) [20]. We collected preoperative data on age, race, sex, body mass index (BMI), baseline renal function (estimated glomerular filtration rate, eGFR, using MDRD) [21], past medical history, hemodynamic numbers, intraoperative data including cardiopulmonary bypass time and receipt of blood products, postoperative data including length of stay, postoperative complications (e.g. sepsis, thromboembolic events), need for RRT and in-hospital mortality. Baseline creatinine was defined as the listing of creatinine data obtained from the INTERMACS registry; if INTERMACS did not contain a preoperative serum creatinine, then the last serum creatinine in the EMR prior to device implantation was used. We collected infor-
196
Am J Nephrol 2014;39:195–203 DOI: 10.1159/000358495
mation about the indication for VAD implantation (bridge to transplant or destination therapy) as well as the type of device implanted. Outcomes The primary outcome of AKI was defined as a 50% rise in serum creatinine over the preoperative baseline during the first 7 postoperative days as per the creatinine-based RIFLE (Risk of renal dysfunction, Injury to the kidney, Failure or Loss of kidney function, and End-stage renal disease; AKI assessed over 7 days) Risk criteria [22]. The Acute Kidney Injury Network (AKIN; AKI assessed over 48 h) creatinine criteria [23, 24] and urine output criteria were utilized for secondary analyses. All-cause mortality was monitored over 1 year after implantation, with data reported at 30 and 365 days. Statistical Methods Qualitative data were recorded in a categorical fashion, and quantitative covariates were measured as continuous variables. Differences between characteristics were analyzed either by an unpaired t test or the Wilcoxon signed-rank test as appropriate. Categorical variables were assessed by χ2 analysis or Fisher’s exact test and were reported with their corresponding odds ratio (OR) and 95% confidence intervals (CI). Continuous variables are represented with mean and their 95% CI values and hazard ratio (HR) when pertaining to survival data. To assess determinants of AKI, we first performed a univariate logistic regression analysis. Significant predictors (p < 0.10) were then included in a multivariate model. Survival data were analyzed using Cox proportional hazards modeling. All survival data were censored for cardiac transplantation. We analyzed survival at 30 and 365 days. Significant variables (p < 0.10) were then included in a multivariate model. All statistical tests were two-sided with α set at 0.05 for statistical significance. STATA 11.2 (StataCorp LP, College Station, Tex., USA) was used for all data analysis.
Results
Baseline Characteristics of Patients with AKI We identified 168 patients who underwent VAD implantation at our institution during the study period. Eleven subjects were excluded from the analysis, with 8 patients having previous end-stage renal disease, 2 requiring preimplant RRT for AKI and 1 subject with intraoperative mortality. In the final cohort of 157, a total of 44 (28%) patients developed AKI. Table 1 demonstrates the pre-, intra- and postoperative characteristics of those with and without AKI. Prior to implantation, there was no difference in age, baseline renal function (serum creatinine and eGFR), New York Heart Association (NYHA) CHF class, preoperative hemodynamics, INTERMACS score, the presence of cardiogenic shock or inotrope use in those with and without AKI. However, the patients with AKI tended to have more diabetes and cerebrovasNaik /Akhter /Fedson /Jeevanandam /Rich / Koyner
Downloaded by: University of Pittsburgh 198.143.32.65 - 8/10/2015 6:13:05 PM
diac surgery; however, their role and application in the setting of ventricular assist device (VAD) implantation is unclear [6–9]. Since the Randomized Evaluation of Mechanical Assistance for the Treatment of Congestive Heart Failure (REMATCH) trial [10], there has been an increased number of VADs being implanted, with nearly 6,000 VADs being implanted since June 2006 [11]. The first-generation VADs were pulsatile and presumably more ‘physiologic’; however, they were associated with higher infection and thrombosis rates, lower durability and even lower cardiac transplant rates in comparison to continuous flow devices [12]. This has led to a transition away from pulsatile flow devices to continuous flow devices which now make up for >90% implants in the United States [13]. While the increasing numbers of implants have improved our understanding of the impact of AKI during the perioperative period, its etiology, predictors and short- and long-term consequences require further investigation. Previous studies have been difficult to interpret, as many have not used consensus definitions of AKI [14–19]. Utilizing consensus definitions of AKI, we performed a single-center, retrospective, longitudinal cohort study to evaluate the incidence of AKI following VAD implantation, identify risk factors as well as assess its impact on 30- and 365-day survival.
Table 1. Pre-, intra- and postoperative characteristics of patients with and without AKI
a Baseline clinical characteristics of VAD patients by AKI status Operative age, years 58.3 ± 12.7 Weight, kg 89.2 ± 23.3 Race 68 (60.2) Caucasian 36 (31.9) African-American Asian 4 (3.5) Unknown 5 (4.4) Males 89 (78.8) BMI 28.9 ± 6.3 Baseline serum creatinine, mg/dl 1.67 ± 1.01 Baseline eGFR (MDRD), ml/min 49.3 ± 26.7 Preoperative hematocrit 34.2 ± 6.2 HbA1c 6.9 ± 1.2 Diabetes 35 (31.0) PVD 7 (6.2) Hypertension 48 (42.5) CVD 9 (8.0) Hyperlipidemia 61 (54) CHF 101 (89.4) Ejection fraction 17.7 ± 7.6 Cardiogenic shock 34 (30.1) VAD device HeartMate II 88 (77.9) HeartWare 19 (16.8) BiVAD 3 (2.7) PVAD/RVAD 0 (0) 3 (2.6) HeartMate XVE Goal of implant Bridge to transplant 54 (47.8) Destination therapy 59 (52.2) NYHA classification (n =133) 1 (1.1) Class I 1 (1.1) Class II 9 (9.4) Class III 84 (88.4) Class IV INTERMACS score (n = 147) 18 (17) 1 46 (43.39) 2 25 (23.58) 3 15 (14.15) 4 2 (1.88) ≥5 Preoperative hemodynamics (n = 111) RAP, mm Hg 13.7 ± 6.6 RVSP, mm Hg 53.3 ± 14.8 PASP, mm Hg 54.3 ± 15.1 PADP, mm Hg 26.9 ± 8.5 PCWP, mm Hg 25.9 ± 8.7 Cardiac output, l/min 3.78 ± 1.7 Cardiac index, l/min/m2 1.84 ± 0.6
AKI and Mortality following VAD Implantation
Am J Nephrol 2014;39:195–203 DOI: 10.1159/000358495
AKI (n = 44)
p value
56.4 ± 12.7 84.3 ± 25.3
0.4 0.25
23 (52.3) 18 (40.9) 0 (0) 3 (6.8) 34 (77.3) 28.4 ± 7.3 1.58 ± 0.71 50.4 ± 22.1 36.5 ± 5.3 6.9 ± 1.2 23 (52.3) 3 (6.8) 23 (52.2) 9 (20.5) 21 (47.7) 43 (97.7) 16.5 ± 7.9 16 (36.36)
0.39
35 (79.6) 5 (11.4) 1 (2.2) 1 (2.2) 2 (4.6)
0.84 0.62 0.57 0.82 0.05 0.84 0.01 0.85 0.27 0.03 0.48 0.09 0.57 0.68
0.48
21 (48.8) 22 (51.2)
0.90
1 (2.6) 1 (2.6) 4 (10.5) 32 (84.2)
0.58
11 (26.82) 14 (34.14) 10 (24.39) 6 (14.63) 0 14.5 ± 7.5 51 ± 12.7 51.6 ± 13.9 26.5 ± 7.2 23.6 ± 10.2 3.60 ± 1.4 1.82 ± 0.7
0.63
0.60 0.45 0.40 0.84 0.25 0.62 0.87
197
Downloaded by: University of Pittsburgh 198.143.32.65 - 8/10/2015 6:13:05 PM
No AKI (n = 113)
Table 1 (continued)
No AKI (n = 113)
AKI (n = 44)
p value
b Intra- and postoperative outcomes in patients with and without AKI following VAD implantation Intraoperative outcomes CPB time, min 131.0 ± 36.5 133.9 ± 49.4 0.69 Patients who received intraoperative blood products 88 (77.9) 26 (59.1) 0.02 Patients who received intraoperative PRBC 61 (54.0) 24 (54.6) 0.94 Units of intraoperative PRBC used 1 (0 – 3) 1 (0 – 6) 0.45 Postoperative outcomes Net fluid balance at 7 days, liters –6.65 ± 5.62 – 4.06 ± 5.78 0.02 ICU length of stay, days 6.4 ± 7.7 11.2 ± 11.0 0.006 Total length of stay, days 29.2 ± 59.4 34.5 ± 46.1 0.60 Pneumonia 4 (3.5) 4 (9.1) 0.16 Reoperation for bleeding 11 (9.7) 3 (6.8) 0.57 Gastrointestinal bleeding 6 (5.3) 3 (6.8) 0.71 Blood product transfusion 70 (68.6) 23 (57.5) 0.21 Units of PRBC transfused 3 (2 – 6) 7 (4–16) 0.002 In-hospital mortality 9 (8.0) 14 (31.8)
