Continuous flow left ventricular assist devices (LVADs) are increasingly used in the management of patients with end-stage heart failure, as either a bridge to transplantation (BTT) or as destination therapy (DT). These new generation devices have virtually replaced the previous pulsatile-flow devices, comprising 99% of the LVAD implantations in late 2011.1 They offer the advantages of increased durability, smaller size, and decreased incidence of adverse events.2 There has also been continued improvement in postoperative results with reported 1 and 2 year survival of 79% and 66%, respectively.3 The application of LVADs for DT has doubled since 2006 to 34% of all implants.1 The American College of Cardiology and American Heart Association recommend using LVAD as DT for advanced heart failure patients that do not respond to standard therapy, are not transplantation candidates, and have <50% predicted 1 year survival.4
Despite the remarkable progress over the past several years, adverse events such as infection, bleeding, transient ischemic attacks, and device malfunction still limit the success of LVAD therapy.5 Infection and bleeding still occur in >40% of patients after implantation.3 These adverse events contribute to increased hospital readmissions and costs, decreased patient satisfaction, as well as potential for increased mortality. Although improvements in device durability and reliability have contributed to recent success, it is vital to consider patient selection to further improve outcomes. While interagency registry for mechanically assisted circulatory support (INTERMACS) profiling and New York Heart Association (NYHA) classification are current standards for assessing degree of heart failure, other methods may also find utility to improve patient selection.
Sequential organ failure assessment (SOFA) score is commonly used in critical care settings to quantify the severity of end-organ dysfunction.6,7 This scoring system takes into account cardiovascular, respiratory, hepatic, renal, coagulation, and neurologic function (Table 1). Because it uses simple laboratory values that are commonly collected, it is quick and easy to calculate. In this study, we sought to examine the relationship between preoperative SOFA score and postoperative outcomes, such as mortality, length of stay, and readmission rates. We also sought to determine whether SOFA score is a useful prognostic indicator in patient selection for LVAD implantation.
Systematic chart review was conducted on 97 consecutive patients who received LVADs at our institution from January 2007 to April 2012. Both HeartMate II (HMII) and HeartWare (HW) devices were used in the study. The decision to implant HMII versus HW was based on US Food and Drug Administration approval of devices, as well as surgeon preference. In the DT group, the HMII device was used because HW is not approved for this purpose. In the BTT patients, either the HMII or HW devices were used, and this was based on surgeon preference. Patient selection for BTT versus DT is based on published criteria.8,9 Two patients were excluded from the study because they received a pulsatile LVAD (HeartMate XVE) before receiving a continuous flow LVAD. Preoperative data necessary to calculate SOFA score, as depicted in Table 1, were available for all patients. Outcome data including mortality (1 month, 3 months, 6 months, 9 months, 1 year, 2 years, and 3 years), length of stay, occurrence of adverse events, and hospital readmissions were collected. Respiratory scores were calculated using the point-of-care values obtained from the anesthesia record. Glasgow coma scores were not explicitly reported in patient charts, so a SOFA score of 0 was assigned to patients that were not mechanically ventilated, and a score of 4 was assigned to patients that were mechanically ventilated.
Continuous variables are presented as means with standard deviations with significance level determined using Student’s t-test. Categorical variables are presented as rates and percentages with 95% confidence intervals (CIs) with significance level determined using Fisher’s exact test. Kaplan-Meier analysis was performed to demonstrate survival for assigned SOFA groupings. The log-rank test was used to assess Kaplan-Meier plots. Receiver operating characteristic (ROC) analysis was used to assess sensitivity and specificity of preoperative SOFA score in predicting survival. The Youden index was used to determine the critical cutoff point on the ROC curves.
Baseline characteristics of the cohort are depicted in Table 2. There were 80 patients implanted with HeartMate II (HMII) and 17 patients with HW. Fifty-five patients were BTT, while 42 were receiving DT. The mean SOFA score for the entire group was 4.60 ± 2.82. The average INTERMACS classification was 2.68 ± 1.15. Sixteen patients were INTERMACS classification 1, 32 were class 2, 17 were class 3, 30 were class 4, and 1 was class 6. Of the 16 class 1 patients, the average SOFA score was 6.50 ± 4.18, which was significantly higher than INTERMACS class 2–6 patients average SOFA score of 4.23 ± 2.34 (p = 0.001). Table 2 depicts preoperative laboratory values and subcomponents of the SOFA score. There was no significant difference (p = 0.052) between SOFA scores for patients receiving HMII (3.53 ± 2.27) and patients receiving HW (4.83 ± 2.89). The SOFA score was significantly higher in the DT group compared with the BTT group (5.38 ± 2.29 vs. 4.00 ± 3.06; p = 0.013). Of the 38 patients requiring inotropic support before LVAD implantation, 28 required a low level of support (dopamine < 5 μg/kg/min or dobutamine any dose), nine required a moderate level of support (dopamine > 5 μg/kg/min or epinephrine/norepinephrine < 0.1 μg/kg/min), and one required a high level of support (norepinephrine 0.12 μg/kg/min).
The overall 30 day mortality after LVAD implantation was 10.1% (95% CI, 4.2–17.6]. The 30 day mortality for HMII recipients (10.0% [95% CI, 4.3–19.7]) was not significantly different (p = 0.052) from the 30 day mortality rate for HW recipients (5.8% [95% CI, 0.1–32.7]). The 30 day mortality for BTT patients was (5.5% [95% CI, 0–15.9%]), which was not significantly different (p = 0.157) from the rate for DT patients (14.3% [95% CI, 5.2–31.1]).
The mean SOFA score of survivors at 30 days was 4.50 ± 2.88, which was not significantly different (p = 0.178) from the mean SOFA score of nonsurvivors (5.56 ± 2.01). The 30 day mortality for SOFA scores 0–2 (n = 22) was zero (95% CI, 0–16.8); for scores 3–5 (n = 46), the rate was 13.0% (95% CI, 4.8–28.4); for scores 6–8 (n = 20), it was 10.0% (95% CI, 1.2–36.1); and for SOFA scores ≥ 9 (n = 9), it was 11.1% (95% CI, 0.2–62.0). There was no statistically significant difference in 30 day mortality between the subgroups (Figure 1).
To assess the sensitivity and specificity of SOFA score for predicting 30 day mortality, we performed a ROC analysis. The area under the curve (AUC) was 0.667 (p = 0.024). The criterion with the highest Youden index was 3. This criterion was subsequently used to compare mortality rates between two subgroups. The 30 day mortality rate for SOFA score ≤3 was 2.6% (0–14.7%), whereas the 30 day mortality rate for SOFA score >3 was 13.6% (5.9–26.7%). However, there was no significant difference between these rates using Fisher’s exact test (p = 0.150).
In addition to mortality, correlation between preimplant SOFA score and length of postoperative hospital stay was assessed. Eleven patients did not survive to discharge and were not included. The average postoperative stay for the entire group was 18.6 ± 9.7 days. ROC analysis was used to assess sensitivity and specificity for SOFA score predicting length of stay being greater than the average length of stay of 18 days. The AUC was 0.568 (p = 0.313). The criterion with the highest Youden index was 9. This criterion was then used to compare average length of stay between two subgroups. The average length of stay for SOFA score ≤9 was 18.1 ± 9.7 days, whereas the average length of stay for SOFA score >9 was 26.2 ± 6.1 days. There was a significant difference between these values (p = 0.039).
The overall 1 year survival after LVAD implantation was 73.3% (95% CI, 55.2–95.5). The mean preimplant SOFA score of survivors at 1 year was significantly lower than that of nonsurvivors (3.67 ± 2.19 vs. 6.60 ± 3.65; p = 0.002). For SOFA scores 0–2 (n = 18), the survival rate was 94.4% (95% CI, 55.0–100). For SOFA scores 3–5 (n = 36), the 1 year survival rate was 75.0% (95% CI, 49.4–100). For scores 6–8 (n = 14), the survival rate was 64.3% (95% CI, 29.4–100) and for scores ≥9 (n = 7), the survival rate was 28.6% (95% CI, 3.5–100) (Figure 2).
ROC analysis was used to assess the sensitivity and specificity of SOFA score for predicting 1 year survival (Figure 3). The AUC was 0.715 (p = 0.002). The criterion with the highest Youden index was 4. This criterion was subsequently used to compare 1 year survival rates. The 1 year survival rate for SOFA score ≤4 was 87.0% (95% CI, 62.1–100), whereas the 1 year mortality rate for SOFA score >4 was 51.7% (95% CI, 27.0–85.3). There was significant difference between these rates using Fisher’s exact test (p = 0.001).
In addition to 30 day and 1 year mortality, mortality rates and average SOFA scores for survivors versus nonsurvivors were also calculated for 3 months, 6 months, 9 months, 2 years, and 3 years (Figure 4). In addition to a significant correlation between SOFA score and survival at 1 year, there was significant correlation at 6 months, 9 months, 2 years, and 3 years. The causes of death for the SOFA score subgroups included hemorrhage, stroke, sepsis, electromechanical failure, and other causes. In addition, the cause of death was indeterminate for 13 of the patients. These causes are delineated for the different SOFA score subgroups in Figure 7A and for the different lengths of survival are delineated in Figure 7B.
Kaplan-Meier analysis was performed to assess long-term survival (Figure 5). The median survival for SOFA score 6–8 was 966 days, whereas the median survival for SOFA score 9+ was 339 days. A median survival could not be calculated for SOFA scores 0–2 and 3–5 because >50% of the patients in these groupings were alive at the time of the study. For SOFA score 0–2, survival after mean follow-up of 835 ± 436 days was 81.8% (95% CI, 5.0–46.6); for SOFA score 3–5, survival after mean follow-up of 632 ± 546 days was 79.4% (95% CI, 9.0–37.1).
In addition to considering mortality, adverse outcomes including gastrointestinal bleed, infection, neurologic events, and pump replacement were examined to determine whether SOFA score could predict any of these outcomes (Figure 6). There was no significant difference in SOFA scores for patients who had to be readmitted for any of the aforementioned adverse outcomes.
Bridge to Transplant and Destination Therapy Patients
Data analysis was also performed on the BTT and DT patients separately. The average SOFA score of survivors was not significantly different from nonsurvivors in the BTT group at 30 days (3.98 ± 3.13 vs. 4.33 ± 1.15; p = 0.680), but there was a significant difference at 1 year (2.74 ± 1.71 vs. 6.80 ± 4.51; p = 0.025). The average SOFA score of survivors was not significantly different from nonsurvivors in the DT group at 30 days (5.25 ± 2.32 vs. 6.17 ± 2.14; p = 0.368) or at 1 year (4.88 ± 2.17 vs. 6.40 ± 1.15; p = 0.094).
ROC analysis was also performed for BTT and DT mortality at 30 days and 1 year. For BTT patients, the AUC for 30 days was 0.651 (p = 0.152) and the AUC for 1 year was 0.827 (p < 0.001). The criterion with the highest Youden index for 30 day mortality was 2. For SOFA score ≤2, the mortality rate was 0 (95% CI, 0–18.4), whereas for SOFA score >2, the mortality rate was 8.5% (95% CI, 0–25.0). There was no significant difference between these rates using Fisher’s exact test (p = 0.293). The criterion with the highest Youden index for 1 year survival was 4. For SOFA score ≤4, the 1 year survival was 90.0% (95% CI, 59.3–100], whereas for SOFA score >4, the 1 year survival was 36.4% (95% CI, 10.0–93.1). There was significant difference between these rates using Fisher’s exact test (p = 0.001).
For DT patients, the AUC for 30 days was 0.637 (p = 0.232) and the AUC for 1 year was 0.694 (p = 0.069). The criterion with the highest Youden index for 30 day mortality was 3. For SOFA score ≤3, the mortality rate was zero (95% CI, 0–46.1), whereas for SOFA score >3, the mortality rate was 17.6% (95% CI, 6.5–38.4). There was no significant difference between these rates using Fisher’s exact test (p = 0.577). The criterion with the highest Youden index for 1 year survival was 6. For SOFA score ≤6, the 1 year survival was 83.3% (95% CI, 50.9–100], whereas for SOFA score >6, the 1 year survival was 40.0% (95% CI, 10.9–100). There was significant difference between these rates using Fisher’s exact test (p = 0.034).
In this study, we investigated the correlation between preoperative SOFA score and postoperative outcomes after LVAD implantation. We found that the preimplant SOFA score reliably predicts survival after 6, 9, 12, 24, and 36 months. Interestingly, the preimplant SOFA score did not predict 30 day mortality in our cohort, although there was a correlation with postoperative length of stay. While long-term survival can be predicted, our results showed that there was no correlation between SOFA score and long-term events such as neurologic, bleeding, infection, or pump replacement.
SOFA score is typically high in critically ill patients, and this would be expected to result in higher perioperative mortality. It may therefore seem counterintuitive that SOFA score appears to be a better predictor of long-term rather than short-term outcomes after LVAD implantation. This discrepancy may reflect the fact that a high SOFA score may represent poor functional reserve of liver, kidney, and other organs, which may in turn impact long-term survival even after heart failure has been remedied by LVAD.
Other single-center studies have found utility in using SOFA scores to predict outcomes after LVAD implantation.10,11 A study by Shahzad et al.10 reported similar results to our study, finding significance between SOFA score and survival at 3, 6, 9, and 12 months. However, their study differed in short-term outcomes compared with ours. Although there was no significant increase in operative mortality with increasing SOFA score at our center, their study showed increasing mortality with increasing SOFA score. Specifically, patients with a SOFA score of >9 had an operative mortality of 95% in their study, which was significantly different than patients with lower SOFA scores. Another study also found a correlation between SOFA score and short-term outcomes after LVAD implantation. Qedra et al.11 reported a significant difference in 7 day mortality between groups with SOFA score ≥11 and <11.
SOFA scores may have utility in patient selection for LVAD implantation either as BTT or as DT. Because LVAD therapy is increasingly being used as DT, the long-term survival of these patients is of great importance. Based on our findings, selecting patients for an LVAD before there is significant loss of organ function can improve the long-term survival of patients after implantation. Other studies have suggested that implanting LVAD in earlier stages of heart failure (NYHA class IIIb) may be recommended.5 For example, patients with a SOFA score of <3 had a 1 year survival rate of 95%, whereas those with SOFA scores >9 had a 1 year survival rate of only 38.6%.
SOFA scores should also be considered in patients who are BTT candidates. Based on our data, patients with lower SOFA scores at time of implantation have better long-term outcomes than those with higher SOFA scores. Patients who are bridged with a SOFA score of 4 or below have a 90% survival at 1 year, whereas those with SOFA scores 5 and higher only have a 36% survival. Therefore, bridging candidates to transplantation before development of end-organ damage may result in better long-term outcomes after the patient is transplanted.
Being a retrospective study, one possible limitation of this study was that not all parameters used to calculate SOFA score were taken at the same time preoperatively. While most of the laboratory values used were obtained within a day before LVAD implantation, sometimes a value was used up to a week before operation. Because of this, some patients’ SOFA scores may have underestimated the severity of organ dysfunction before LVAD implantation. Values that were obtained at hospital admission may have not accurately reflected the declining condition of the patient before implantation. In addition to the current data presented, we continue to consider outcomes beyond 3 years. Because LVAD implantation is a relatively new therapeutic option, we do not currently have enough data to report beyond 3 years in our cohort of patients.
1. Kirklin JK, Naftel DC, Kormos RL, et al. The Fourth INTERMACS Annual Report: 4,000 implants and counting. J Heart Lung Transplant. 2012;31:117–126
2. Slaughter MS, Pagani FD, Rogers JG, et al.HeartMate II Clinical Investigators. Clinical management of continuous-flow left ventricular assist devices in advanced heart failure. J Heart Lung Transplant. 2010;29(4 Suppl):S1–39
3. Kilic A, Ailawadi G. Left ventricular assist devices in heart failure. Expert Rev Cardiovasc Ther. 2012;10:649–656
4. Jessup M, Abraham WT, Casey DE, et al. 2009 focused update: ACCF/AHA Guidelines for the Diagnosis and Management of Heart Failure in Adults: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines: Developed in collaboration with the International Society for Heart and Lung Transplantation. Circulation. 2009;119:1977–2016
5. John R, Kamdar F, Eckman P, et al. Lessons learned from experience with over 100 consecutive HeartMate II left ventricular assist devices. Ann Thorac Surg. 2011;92:1593–1599; discussion 1599
6. Ferreira FL, Bota DP, Bross A, Mélot C, Vincent JL. Serial evaluation of the SOFA score to predict outcome in critically ill patients. JAMA. 2001;286:1754–1758
7. Vincent JL, Moreno R, Takala J, et al. The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine. Intensive Care Med. 1996;22:707–710
8. Mehra MR, Kobashigawa J, Starling R, et al. Listing criteria for heart transplantation: International Society for Heart and Lung Transplantation guidelines for the care of cardiac transplant candidates–2006. J Heart Lung Transplant. 2006;25:1024–1042
9. Feldman D, Pamboukian SV, Teuteberg JJ, et al.International Society for Heart and Lung Transplantation. The 2013 International Society for Heart and Lung Transplantation Guidelines for mechanical circulatory support
: Executive summary. J Heart Lung Transplant. 2013;32:157–187
10. Shahzad K, Li Q, Akashi H, et al. Degree of end organ damage as a predictor of outcomes after mechanical circulatory support
device implantation—A more quantitative assessment. J Heart Lung Transplant. 2011;30(4S):S32
11. Qedra N, Drews T, Yilmaz K, et al. Preoperative application of modified sequential organ failure assessment score (SOFA) in prediction of early postoperative mortality after implantation of ventricular assist devices (VAD). J Heart Lung Transplant. 2008;27:S61