OUP user menu

★ Editor's choice ★

Long-term outcomes and cardiac surgery in critically ill patients with infective endocarditis

Mariana Mirabel, Romain Sonneville, David Hajage, Emmanuel Novy, Florence Tubach, Philippe Vignon, Pierre Perez, Sylvain Lavoué, Achille Kouatchet, Olivier Pajot, Armand Mekontso-Dessap, Jean-Marie Tonnelier, Pierre-Edouard Bollaert, Jean-Pierre Frat, Jean-Christophe Navellou, Herve Hyvernat, Ali Ait Hssain, Jean-François Timsit, Bruno Megarbane, Michel Wolff, Jean-Louis Trouillet,
DOI: http://dx.doi.org/10.1093/eurheartj/eht303 1195-1204 First published online: 20 August 2013

Abstract

Aims To assess long-term outcomes and the management of critical left-sided infective endocarditis (IE) and evaluate the impact of surgery.

Methods and results Among the 198 patients included prospectively for IE across 33 adult intensive care units (ICU) in France from 1 April 2007 to 1 October 2008, 137 (69%) were dead at a median follow-up time of 59.5 months. Characteristics significantly associated with mortality were: Sepsis-related Organ-Failure Assessment (SOFA) score at ICU admission [Hazard ratio (HR), 95% Confidence Interval (CI) of 1.43 (0.79–2.59) for SOFA 5–9; 2.01 (1.05–3.85) for SOFA 10–14; 3.53 (1.75–7.11) for SOFA 15–20; reference category SOFA 0–4; P = 0.003]; prosthetic mechanical valve IE [HR 2.01; 95% CI 1.09–3.69, P = 0.025]; vegetation size ≥15 mm [HR 1.64; 95% CI 1.03–2.63, P = 0.038]; and cardiac surgery [HR (95%CI), 0.33 (0.16–0.67) for surgery ≤1 day after IE diagnosis; 0.61 (0.29–1.26) for surgery 2–7 days after IE diagnosis; 0.42 (0.21–0.83) for surgery >7 days after IE diagnosis; reference category no surgery; P = 0.005]. One hundred and three (52%) patients underwent cardiac surgery after a median time of 6 (16) days. Independent predictors of surgical intervention on multivariate analysis were: age ≤60 years [Odds ratio (OR) 5.30; 95% CI (2.46–11.41), P < 0.01], heart failure [OR 3.27; 95% CI (1.03–10.35), P = 0.04], cardiogenic shock [OR 3.31; 95% CI (1.47–7.46), P = 0.004], septic shock [OR 0.25; 95% CI (0.11–0.59), P = 0.002], immunosuppression [OR 0.15; 95% CI (0.04–0.55), P = 0.004], and diagnosis before or within 24 h of ICU admission [OR 2.81; 95% CI (1.14–6.95), P = 0.025]. SOFA score calculated the day of surgery was the only independently associated factor with long-term mortality [HR (95% CI) 1.59 (0.77–3.28) for SOFA 5–9; 3.56 (1.71–7.38) for SOFA 10–14; 11.58 (4.02–33.35) for SOFA 15–20; reference category SOFA 0–4; P < 0.0001]. Surgical timing was not associated with post-operative outcomes. Of the 158 patients with a theoretical indication for surgery, the 58 deemed not fit had a 95% mortality rate.

Conclusion Mortality in patients with critical IE remains unacceptably high. Factors associated with long-term outcomes are the severity of multiorgan failure, prosthetic mechanical valve IE, vegetation size ≥15 mm, and surgical treatment. Up to one-third of potential candidates do not undergo surgery and these patients experience extremely high mortality rates. The strongest independent predictor of post-operative mortality is the pre-operative multiorgan failure score while surgical timing does not seem to impact on outcomes.

  • Valve surgery
  • Heart valve disease
  • Intensive care

Introduction

Infective endocarditis (IE) is a rare but severe disease that still carries high mortality rates approaching 30% at 1 year.1,2 In the recent years, the epidemiology of IE has significantly changed in western countries with the typical patient increasingly elderly and comorbid. Infective endocarditis is now most often acute, associated with Staphylococcus aureus infection, and characterized by cardiac complications and embolic events.3,4 A growing number of patients require intensive care unit (ICU) admission (referred as critical IE).5,6 Outcomes therefore depend not only on the severity of the IE but also the underlying health status of the patients. Nonetheless very few studies focus on patients with critical IE and multiple complications.58

There have been an increasing number of reports advocating early surgery in complicated IE over the past decade.913 Early surgery has become a mainstay in the therapy of complicated IE.14,15 However, studies reporting the impact of surgery on outcomes in patients who are critically ill are scarce.5,6,8 Moreover, in such patients, there is often a need to weigh up the coexistent risks and benefits.

We describe here the results of a multicentre prospective observational study of critical left-sided IE patients with a special focus on the impact of cardiac surgery. We assess factors associated with long-term outcomes, with surgical vs. medical treatment alone, and with post-operative mortality.

Methods

Patients

The methods of the study has been described elsewhere.6 Briefly, ENDOREA was a multicentre prospective observational study conducted in 33 adult ICUs in 23 university-affiliated, and in 10 general French hospitals. Patients were included when they satisfied the modified Duke criteria for definite IE between 1 April 2007 and 1 October 2008.16 Infective endocarditis was considered to be active if the patient was admitted to the ICU before or within 30 days of starting antibiotics. Patients transferred to the ICU after cardiac surgery for IE and patients who developed IE in the ICU were excluded. The Ethics Committee of the Société de Réanimation de Langue Française approved the study.6

Data collection

For each patient, the following data were collected: demographics, pre-existing cardiac comorbidities, the Charlson score,17 a history of cardiac surgery, chronic haemodialysis, history of intravenous drug use, and immunosuppression (defined as one of the following: human immunodeficiency virus infection, malignancy, and/or long-term use of corticosteroids and/or other immunosuppressant). Each patient's condition at ICU admission was assessed using the Simplified Acute Physiology Score (SAPS) II, the Sepsis-related Organ-Failure Assessment (SOFA), and the Glasgow coma scale (GCS) scores.1820 Neurological failure was defined by a GCS < 10 within the first 24 h of ICU admission before sedative drugs were given. The following characteristics of IE were prospectively collected: microorganism isolated from blood and/or heart valve cultures; echocardiographic findings [left ventricular ejection fraction (LVEF), valve dysfunction, vegetation length, mobility, and location)]. The vegetation length was measured in various planes during the first echocardiogram and on follow-up scans when available, and the maximal length was used for analysis. Patients with both left- and right-sided IE were assigned to the left-sided group. Those with both native valve IE and prosthetic valve IE were assigned to the latter. Infective endocarditis complications included cardiogenic shock, septic shock, heart failure without shock, severe valve dysfunction without heart failure, and systemic embolic event (e.g. stroke).14

In patients who underwent surgery, the following data were collected: indication for surgery; surgical timing; the SOFA score on the day of surgery; type of procedure performed; and duration of extra-corporeal circulation and aortic clamping. Cardiac surgery was considered as: (i) emergency if performed within the first 24 h of IE diagnosis; (ii) urgent if performed between 2 and 7 days following IE diagnosis; (iii) elective if performed after 7 days of IE diagnosis.14

In addition, the adjudication committee (M.M., R.S., J.-L.T., M.W.) reviewed the hospital discharge summaries, and decided for each patient whether surgery was indicated according to contemporary guidelines.15 The committee was not blinded to patients' outcomes.

Outcomes

The main outcome measure was long-term mortality from ICU admission. Cross-sectional follow-up was undertaken between 15 January and 30 March 2013 by contacting the referring physician. We analysed factors associated with long-term mortality, factors associated with surgical treatment, and we determined the reasons for denying surgery. We assessed pre-operative risk factors associated with mortality at long-term follow-up in patients who had been operated on.

Statistical analysis

The results are reported as median and interquartile range (IQR) or as numbers and percentages. Categorical variables were compared using χ2 test or Fisher's exact test, and continuous variables using Student's t-test or Wilcoxon rank sum test, as appropriate. Factors associated with surgical treatment were analysed using a logistic regression model. Factors associated with long-term mortality were analysed using a Cox proportional hazard model. For the analysis of the overall population, time to death was calculated as time from ICU admission to death or last follow-up. Surgery was categorized according to time from IE diagnosis, and entered in the model as a time-dependent variable. For the analysis of surgical treatment patients, time to death was calculated from the date of surgery.

All models were constructed using the same methodology: variables associated (P < 0.10) with outcome in the univariate analysis were considered for the multivariate model, and final model was selected using backward stepwise (P < 0.10). Odds ratios (OR) for logistic regression model, and hazard ratios (HR) for Cox model were calculated accordingly with their 95% confidence intervals (CI). Significance was defined as P-values <0.05. Statistical tests were two-sided. Statistical analyses were performed using R 2.15.2 (http://www.R-project.org).

Results

Patients' characteristics

A total of 198 patients fulfilled the Duke modified criteria for definite left-sided IE and constitute the study cohort. Baseline characteristics have been described elsewhere.6 Briefly, mean age was 61.1 (15.5–71.0) years and 138 (70%) were male. Median (IQR) SOFA and SPA II scores at admission were of 8 (5.0–11.0) and 43 (33–62), respectively. Treatment included mechanical ventilation, vasoactive drugs, and/or renal replacement therapy in 156 (79%), 141 (71%), and 68 (34%) cases, respectively. Median (IQR) ICU stay was 13 (6.0–29.5) days. Eighty-two patients (41%) died during ICU stay, and 80 (40%) were discharged alive from hospital.

Long-term outcomes

Among the 198 patients admitted for critical IE, 137 (69%) were dead at a median follow-up time of 59.5 months. Three patients (1.5%) were lost to follow-up. Characteristics significantly associated with mortality were: SOFA score at ICU admission [HR (95% CI) of 1.43 (0.79–2.59) for SOFA 5–9; 2.01 (1.05–3.85) for SOFA 10–14; 3.53 (1.75–7.11) for SOFA 15–20; reference category SOFA 0–4; P = 0.003]; prosthesis mechanical valve IE [HR 2.01; 95% CI 1.09–3.69, P = 0.025]; vegetation size ≥15 mm [HR 1.64; 95% CI 1.02–2.63, P = 0.038]; and cardiac surgery [HR (95%CI), 0.33 (0.16–0.67) for surgery ≤1 day after IE diagnosis; 0.61 (0.29–1.26) for surgery 2–7 days after IE diagnosis; 0.42 (0.21–0.83) for surgery >7 days after IE diagnosis; reference category no surgery; P = 0.005] (Table 1). Long-term outcomes in patients according to treatment are depicted in Figures 1 and 2.

View this table:
Table 1

Univariate and multivariate analyses of factors associated with long-term mortality in 198 patients admitted to the intensive care unit with left-sided infective endocarditis

CharacteristicSurvivors n = 61Non-survivors n = 137Univariate HR (95% CI)P-valueMultivariate HR (95% CI)P-value
Age >60 years29 (48)85 (62)1.41 (1.00–2.00)0.049
Sex, male n (%)45 (74)93 (68)0.78 (0.55–1.21)0.181
Diagnosis made before or within the first 24 h after ICU admission15 (25)29 (21)0.78 (0.51–1.17)0.224
Charlson Score ≥3, n (%)10 (16)50 (36)1.51 (1.06–2.14)0.022
Immunosuppressiona, n (%)2 (3)9 (7)1.63 (1.00–2.66)0.049
Intravenous drug user, n (%)6 (10)11 (8)0.87 (0.47–1.61)0.651
Chronic renal replacement therapy, n (%)010 (7)1.86 (0.98–3.56)0.059
Prosthetic valve endocarditisb, n (%)7 (11)36 (26)1.50 (1.03–2.20)0.036
Prosthetic mechanical valve endocarditis, n (%)2 (3)18 (13)1.86 (1.13–3.06)0.0152.01 (1.09–3.69)0.025
Site involved, n (%)
 Mitral valve30 (49)67 (49)1.11 (0.76–1.60)0.601
 Aortic valve25 (41)47 (34)
 Aortic and mitral valves6 (10)22 (16)1.49 (0.89–0.25)0.123
Pathogen, n (%)
Staphylococcus aureus22 (36)69 (50)1.55 (1.11–2.18)0.010
Streptococci24 (39)30 (22)0.59 (0.39–0.88)0.010
 Other15 (25)38 (28)0.98 (0.67–1.43)0.916
Acquisition
 Community, n (%)51 (84)102 (74)1.27 (0.86–1.88)0.222
 Health care-related infection, n (%)10 (16)34 (25)
LVEF < 50%, n (%)18 (30)50 (36)1.39 (0.96–2.01)0.083
Heart failure without shock, n (%)16 (26)20 (16)0.59 (0.37–0.95)0.029
Severe valvular regurgitation without heart failure, n (%)3 (5)7 (5)0.87 (0.40–1.85)0.711
Cardiogenic shock, n (%)43 (70)90 (66)1.22 (0.85–1.74)0.276
Septic shock, n (%)21 (34)80 (58)1.91 (1.36–2.69)0.0002
Infection caused by fungi or multiresistant organismsc, n (%)10 (16)22 (16)0.83 (0.53–1.32)0.438
Vegetation size >10 mm and other predictors of bad outcomed, n (%)10 (16)22 (16)0.90 (0.56–1.46)0.679
Vegetation >15 mm, n (%)13 (21)39 (28)1.437 (0.98–2.1)0.0611.64 (1.03–2.63)0.038
Abscess, false aneurysm or fistula on echocardiogram, n (%)9 (15)22 (16)0.94 (0.59–1.49)0.799
Stroke, n (%)24 (39)55 (40)1.06 (0.75–1.49)0.751
Neurological failure (GCS < 10), n (%)12 (20)43 (31)1.70 (1.18–2.45)0.004
Renal failure, n (%)e9 (15)41 (30)1.64 (1.14–2.37)0.008
SOFA at admission, n (%)<0.00010.003
SOFA 0–4 points20 (33)21 (15)
SOFA 5–9 points22 (36)48 (35)1.74 (1.04–2.90)1.43 (0.79–2.59)
SOFA 10–14 points9 (15)38 (28)2.76 (1.61–4.72)2.01 (1.05–3.85)
SOFA 15–20 points3 (5)22 (16)4.21 (2.30–7.71)3.53 (1.75–7.11)
Surgery, n (%)0.0230.005
≤1 day15 (25)19 (14)0.51 (0.29–0.87)0.33 (0.16–0.67)
2–7 days7 (11)14 (10)0.61 (0.36–1.11)0.61 (0.29–1.26)
>7 days22 (36)24 (18)0.57 (0.35–0.95)0.42 (0.21–0.83)
  • HR, hazard ratio; IQR, interquartile range; LVEF, left-ventricular ejection fraction; GCS, Glasgow Coma Scale; SOFA, Sepsis-Related Organ Failure Assessment.

  • aDefined as one of the following: human immunodeficiency virus infection, malignancy, and/or long-term use of corticosteroids and/or other immunosuppressants.

  • bIncluding bioprosthesis, mechanical valves, and mitral valve repair with annular ring.

  • cDefined by Bacilli negative bacteria, Methicillin-resistant Staphylococcus aureus, fungus as Candida and Aspergillus sp.

  • dDefined by heart failure, persistent infection, and/or peri-valvular abscess on echocardiography.

  • eDefined as Serum creatinine >299 µmol/L (2 mg/dL) within the first 24 h of ICU admission and/or urine output <500 mL/24 h, excluding the eight patients on chronic renal replacement therapy.

Figure 1

Flow chart. Emergency surgery: within the first 24 h of infective endocarditis (IE) diagnosis. Urgent surgery: between Day 2 and Day 7 of IE diagnosis. Elective surgery: after 7 days of IE diagnosis. Missing data on surgical timing in two out of 103 patients.

Figure 2

Kaplan–Meier estimates of long-term survival in the overall population and in those treated medically only.

Surgical procedures and factors associated with surgery

Cardiac surgery was performed in 103 (52%) cases. Valve replacement was performed in 89 of 103 patients (92%); a mechanical valve was used in 29 cases. Thirty-four patients underwent mitral valve replacement (MVR) alone, 31 had aortic valve replacement (AVR) alone, 11 had mitral valve repair alone, 18 had both AVR and MVR, and nine had no replacement or repair. Five patients underwent concomitant coronary artery bypass graft surgery. Median time (IQR) to surgery from IE diagnosis was 6 (16) days. Surgery was delayed in 15 (15%) patients due to neurological complications.

Independent factors associated with surgical treatment on multivariate analysis were: age ≤60 years old [OR 5.30; 95% CI (2.46–11.41), P < 0.01], heart failure [OR 3.27; 95% CI (1.03–10.35), P = 0.04], cardiogenic shock [OR 3.31; 95% CI (1.47–7.46), P = 0.004], septic shock [OR 0.25; 95% CI (0.11–0.59), P = 0.002], immunosuppression [OR 0.15; 95% CI (0.04–0.55), P = 0.004], and diagnosis before or within 24 h of ICU admission [OR 2.81; 95% CI (1.14–6.95), P = 0.025] (Table 2).

View this table:
Table 2

Characteristics according to treatment in 198 patients with left-sided infective endocarditis

CharacteristicAll n = 198Surgery n = 103Medical treatment n = 95Univariate OR (95% CI)-valueMultivariate OR (95% CI)P-value
Age ≤60 years59 (30)25 (24)84 (88)3.75 (2.06–6.85)<0.0015.3 (2.46–11.41)<0.001
Immunosuppressiona, n (%)21 (11)6 (6)15 (16)0.33 (0.12–0.89)0.020.15 (0.04–0.55)0.004
Cardiogenic shock, n (%)64 (67)40 (20)24 (23)1.91 (1.04–3.51)0.043.31 (1.47–7.36)0.004
Heart failure without shock, n (%)36 (18)28 (27)8 (8)0.25 (0.11–0.57)0.0063.27 (1.03–10.35)0.044
Septic shock, n (%)101 (51)42 (21)59 (57)2.34 (1.32–4.15)0.030.25 (0.11–0.59)0.002
Diagnosis made before or within the first 24 h after ICU admission3.12 (1.49–6.5)0.0022.81 (1.14–6.95)0.025
Sex, male n (%)138 (70)77 (75)61 (64)1.65 (.09–3.04)0.11
Charlson Score ≥3, n (%)91 (46)36 (35)55 (58)0.64 (0.21–0.68)0.048
Intravenous drug user, n (%)17 (9)16 (16)1 (1)17.29 (2.25–133.11)<0.001
Chronic renal replacement therapy, n (%)10 (5)2 (2)8 (8)0.22 (0.04–1.04)0.08
Native valve endocarditis, n (%)155 (78)78 (76)77 (81)0.73 (0.37–1.44)0.36
Prosthetic valve endocarditis, n (%)0.51
 Mechanical valve20 (10)9 (9)11 (12)1.37 (0.54–3.46)
 Bioprosthesis15 (8)9 (9)6 (6)
 Mitral valve repair8 (4)7 (7)1 (1)
Site involved, n (%)0.97
 Mitral valve97 (49)50 (49)47 (49)0.95 (0.52–1.75)
 Aortic valve72 (36)38 (37)34 (36)
 Aortic and mitral valves28 (14)15 (15)13 (14)1.03 (0.43–2.48)
Pathogen, n (%)
Staphylococcus aureus91 (46)41 (40)50 (53)0.6 (0.34–1.05)0.07
Streptococci54 (27)31 (30)23 (24)1.35 (0.72–2.53)0.35
 Other53 (27)31 (30)22 (23)
Acquisition
 Community, n (%)153 (77)84 (82)69 (73)0.10
 Health care-related infection, n (%)44 (22)18 (17)26 (27)0.57 (0.29–1.12)
 LVEF <50%, n (%)68 (34)33 (32)35 (37)0.79 (0.43–1.46)0.45
Severe valvular regurgitation without heart failure, n (%)10 (5)6 (6)4 (4)0.71 (0.19; 2.6)0.85
Infection caused by fungi or multiresistant organismsb, n (%)32 (16)16 (16)16 (17)0.91 (0.43–1.94)0.8
Vegetation size>10 mm and other predictors of bad outcomec, n (%)32 (16)24 (23)8 (8)3.52 (0.12–0.69)0.004
Vegetation >15 mm, n (%)52 (26)31 (30)21 (22)1.5 (0.78–2.87)0.22
Abscess, false aneurysm, or fistula on echocardiogram31 (16)22 (21)9 (9)2.6 (1.13–5.97)0.02
Stroke, n (%)79 (40)43 (42)36 (38)1.17 (0.66–2.08)0.58
Neurological failure (i.e. GCS < 10), n (%)55 (26)16 (16)39 (41)0.26 (0.13–0.52)<0.001
Renal failure, n (%)d50 (25)20 (19)30 (32)0.52 (0.27–1.00)0.048
SOFA at admission, n (%)0.014
 SOFA 0–4 points41 (21)28 (27)13 (14)
 SOFA 5–9 points70 (35)38 (37)32 (35)0.55 (0.25–1.24)
 SOFA 10–14 points47 (24)17 (17)30 (32)0.26 (0.11–0.64)
 SOFA 15–20 points25 (13)10 (10)15 (16)0.31 (0.11–0.87)
  • OR, odds ratio; IQR, interquartile range; LVEF, left-ventricular ejection fraction; GCS, Glasgow Coma Scale; SOFA, Sepsis-related Organ Failure Assessment.

  • aDefined as one of the following: human immunodeficiency virus infection, malignancy, and/or long-term use of corticosteroids and/or other immunosuppressants.

  • bDefined by Bacilli negative bacteria, Methicillin-resistant Staphylococcus aureus, fungus as Candida and Aspergillus sp.

  • cDefined by heart failure, persistent infection, and/or peri-valvular abscess on echocardiography, missing data in 53 cases.

  • dDefined as serum creatinine >299 µmol/L (2 mg/dL) within the first 24 h of ICU admission and/or urine output <500 mL/24 h, excluding the eight patients on chronic renal replacement therapy.

Post-operative outcomes

Fifty-seven out of 103 patients (55%) were dead at follow-up. Pre-operative SOFA score remained the single factor associated with post-operative mortality on multivariate analysis [HR (95% CI) 1.59 (0.77–3.28) for SOFA 5–9; 3.56 (1.71–7.38) for SOFA 10–14; 11.58 (4.02–33.35) for SOFA 15–20; reference category SOFA 0–4; P < 0.0001] (Table 3, Figure 3A). Post-operative survival according to surgical timing after IE diagnosis (≤1 day, 2–7 days, and >7 days) are illustrated in Figure 3B.

View this table:
Table 3

Factors associated with post-operative outcomes

CharacteristicSurvivors at FU n = 46Non-survivors at FU n = 57Univariate HR (95% CI)P-valueMultivariate HR (95% CI)P-value
Age >60 years, n (%)28 (61)31 (54)1.22 (0.73–2.06)0.450
Male, n (%)35 (76)42 (74)0.82 (0.46–1.48)0.515
Cardiogenic shock, n (%)18 (39)22 (39)1.07 (0.63–1.84)0.80
Septic shock, n (%)16 (35)26 (46)1.43 (0.85–2.42)0.18
Charlson score ≥3, n (%)7 (15)18 (32)1.79 (1.02–3.14)0.044
Immunosuppressiona, n (%)1 (2)5 (9)1.99 (0.79–5.02)0.145
Prosthetic valveª IE, n (%)6 (13)19 (33)2.02 (1.16–3.52)0.013
Site involved0.999
 Mitral valve22 (48)28 (49)1.03 (0.58–1.84)
 Aortic valve18 (39)20 (35)
 Aortic and mitral valves6 (13)9 (16)1.06 (0.48–2.33)
Health care acquired, n (%)5 (11)13 (23)1.62 (0.87–3.01)0.131
Staphylococcus aureus IE, n (%)17 (37)24 (42)1.27 (0.75–2.15)0.388
Streptococcal IE, n (%)17 (37)14 (35)0.63 (0.35–1.16)0.139
LVEF < 50%, n (%)13 (28)20 (35)1.57 (0.87–2.83)0.134
SOFA score at admission0.045
SOFA 0–4 points27 (59)19 (33)
SOFA 5–9 points11 (24)15 (26)1.25 (0.62–2.51)
SOFA 10–14 points3 (7)14 (25)1.65 (0.72–3.77)
SOFA 15–20 points015 (26)3.34 (1.40–8.24)
Heart failure without shock, n (%)15 (32)13 (23)0.70 (0.37–1.29)0.249
Valve dysfunction without heart failure, n (%)2 (4)4 (7)1.21 (0.44–3.35)0.713
Resistant bacteria or fungib, n (%)7 (15)9 (16)0.91 (0.45–1.86)0.799
Vegetation size>15 mm, n (%)13 (28)18 (32)1.34 (0.76–2.37)0.319
Abscess, false aneurysm, or fistula on echocardiogram7 (15)15 (26)1.53 (0.85–2.75)0.161
Diagnosis made before or within the first 24 h of ICU admission, n (%)13 (28)19 (33)1.08 (0.62–1.88)0.783
Stroke, n (%)18 (39)25 (44)1.11 (0.66–1.87)0.703
Neurological failure (i.e. GCS < 10), n (%)7 (15)9 (16)1.14 (0.56–2.33)0.720
Time to surgery from IE diagnosis0.537
 ≤1 day, n (%)15 (32)19 (33)
 1–7 days, n (%)9 (20)14 (25)1.00 (0.50–2.00)
 >7 days, n (%)22 (48)24 (42)0.74 (0.41–1.36)
Surgical procedure, n (%):
Valve replacement41 (89)53 (93)0.91 (0.44–1.85)0.938
 Mechanical16 (35)16 (28)1.24 (0.69–2.21)0.470
 Aortic valve replacement14 (30)17 (30)1.21 (0.41–3.59)
 Mitral valve replacement14 (30)20 (35)1.24 (0.43–3.57)
 Multiple valve replacement7 (15)11 (19)1.43 (0.45–4.48)
Mitral valve repair7 (15)9 (16)
Extra-corporeal circulation time >100 min, n (%)18 (39)31 (54)1.52 (0.86–2.70)0.150
Aortic clamping time >80 min, n (%)16 (35)26 (46)1.36 (0.76–2.41)0.290
Surgery delayed due to a neurological complication, n (%)9 (20)6 (11)1.92 (0.82–4.48)0.132
Pre-operative SOFA scorec,<0.001<0.001
SOFA 0–4 points27 (59)19 (33)
SOFA 5–9 points11 (24)15 (26)1.69 (0.86–3.32)1.59 (0.77–3.28)
SOFA 10–14 points3 (7)14 (25)3.86 (1.92–7.76)3.56 (1.71–7.38)
SOFA 15–20 points05 (9)11.49 (4.03–32.73)11.58 (4.02–33.35)
  • HR, hazard ratio; IQR, interquartile range; LVEF, left-ventricular ejection fraction.

  • aDefined as one of the following: human immunodeficiency virus infection, malignancy, and/or long-term use of corticosteroids and/or other immunosuppressants.

  • bGram negative bacilii, Methicillin-resistant Staphylococcus aureus, fungus as Candida and Aspergillus sp.

  • cCalculated the day surgery was performed.

Figure 3

(A) Kaplan–Meier estimates of post-operative survival according to pre-operative SOFA scores presented in quartiles: SOFA score 0–4; SOFA score 5–9; SOFA score 10–14; SOFA score 15–20. (B) Kaplan–Meier estimates of post-operative survival according to surgical timing after IE diagnosis (emergency surgery, i.e. ≤1 day; urgent surgery, i.e. 2–7 days; elective surgery, i.e. >7 days).

Patients denied surgery: characteristics and outcomes

The adjudication committee found that 158 out of 198 patients fulfilled at least one criterion that prompted cardiac surgery, according to guidelines in use at the time of recruitment (Figure 1).15 Of these, 58 (36.7%) did not undergo surgery, of whom 55 were dead at follow-up (95%). The main reasons for declining surgery were: neurological complication (19 patients, 33%), multiorgan failure (17 patients, 29%), comorbidities (15 patients, 26%), and miscellaneous (18 patients, 31%).

Discussion

We present here the largest contemporary series of patients with critical IE with the longest follow-up to date in this population. We summarize factors associated with long-term mortality in a rare condition and assess prognostic factors for surgery. Across the multicentre cohort, we not only identify independent predictors of surgical intervention but also the leading reasons for denial of surgery in critically ill patients, and the outcome of those patients.

Major findings: long-term outcomes in critical infective endocarditis

In our cohort, overall mortality reached 69% at 5 years' follow-up. Our results differ greatly from large recent surgical series, where unadjusted early mortality rates vary between 8 and 12%.10,21,22 Differences in the severity of IE presentation may explain these apparent discrepancies. Indeed, the high mortality in our cohort is consistent with the prediction of the SOFA and SAPS II scores at ICU admission and with previous reports of critical IE.5,7,8

A major prognostic factor in our study was the severity of multiorgan failure as illustrated by the SOFA score. The SOFA score includes haemodynamic, pulmonary, renal, liver, and neurological failure with a 0 to 4 point scale for each organ. SOFA score ranges between 0 (no failure) up to 20 (total multiorgan failure).19 Our results are in line with a previously reported series of IE in ICU patients.5

Infective endocarditis of mechanical prosthetic valve was another factor associated with poor outcomes. Previous studies have reported higher mortality in patients with prosthetic valve IE when compared with native valve IE,5,23 and in prosthetic mechanical valves in particular.24

The size of the vegetation measured on echocardiogram appeared associated with all-cause mortality. The vegetation's maximum length remained a prognostic factor on multivariate analysis. Our finding underscores how meaningful the extent of locally infected tissue is, even after adjustment for strong factors such as the degree of multiorgan failure or the presence of septic shock. There has been extensive evidence of vegetation size's impact on outcome since the late 1980s,2,2527 subsequently leading to changes in the guidelines of IE management.14

We assessed the relationship between surgery and outcomes in patients ‘conspicuous by their absence’ in most studies assessing cardiac surgery for IE. Consistent with many contemporary reports, mortality was significantly lower in patients who underwent surgery, surgery being analysed as a time-dependent variable.912,28

Unlike other series, we did not identify Staphylococcus aureus as a factor associated with long-term morality.4,29,30 Our results may be due to confounding factors as the severity of sepsis-related multiorgan failure that may have outweighed microbiological findings.

In the present study, factors associated with long-term outcome differ from those previously reported by our group at 3 months [i.e. immunosuppression, neurological failure [GCS < 10], and Staphylococcus aureus infection].6 The long follow-up period and different statistical methods (use of a Cox model instead of a logistic regression model) may account for these discrepancies. Time to follow-up strongly alters the nature of the results, and a minimum of 6 months to 1 year seems necessary in order to accurately assess outcomes in IE.12,31

Factors associated with surgery

Treatment was at the discretion of the physician in charge of the patient in this observational study. As in other uncontrolled studies, older age (over 60 years) and underlying comorbidities as immunosuppression were associated with medical treatment alone.911,32,33 Patients with septic shock were mostly treated medically.

In our series, heart failure and cardiogenic shock were significantly associated with valve surgery. Within the wide spectrum of complicated IE, patients most likely to undergo surgery are also those who benefit the more from intervention,10,34 especially in the presence of heart failure and/or cardiogenic shock.9,32,34 In the largest series of patients with IE complicated by heart failure, valve surgery was associated with lower 1-year mortality including patients in NHYA III and IV class.32 A recent study has shown that emergency surgery for native mitral valve endocarditis in patients with cardiogenic shock achieved satisfactory results. In the same study, the presence of septic shock was linked to dismal prognosis.35 Our study shows that contemporary medical practice across 33 centres in France is in agreement with current guidelines, offering valve surgery to patients with heart failure or cardiogenic shock.14

Finally, early diagnosis (i.e. before or within the first 24 h of ICU admission) was associated with surgical treatment. Our results suggest that early diagnosis strongly impacts on patient's management, patients diagnosed later being less likely to undergo surgery. Our results advocate the use of early diagnostic strategies in order to initiate antibiotics and thereby limit the advent of complications.36

Post-operative outcomes

Patients with higher SOFA scores on the day of surgery were at significant risk of dying post-operatively. By analysing several components of a complex disease, we underscore the impact of multiorgan failure on post-operative mortality. The use of the SOFA score may be of particular interest in the decision-making process, especially to assess patients with critical IE in whom indications and contraindications to heart surgery coexist. Indeed, patients with SOFA scores over 15 seem to bear extremely poor outcomes. Our results do not support surgery in the subset of patients with severe multiorgan failure. While the vast majority of studies are in favour of early surgery, our findings suggest that recently published guidelines may not apply to the subgroup of most severe patients with critical IE.

Most patients underwent urgent surgery within the first week after IE diagnosis, in keeping with the severity of the patients' condition. Mortality was not reduced by urgent or emergency operations (i.e. within the first 24 h or the first week, respectively). Our findings should be interpreted cautiously due to the observational nature of our study. The issue of surgical timing has been addressed in recent years with a significant number of limitations, differences in patients' characteristics and outcomes measures, leading to apparently conflicting results.13,28,37,38 By means of a randomized controlled study, Kang et al. 13 recently showed that surgery within 48 h reduced the risk of embolic events in uncomplicated IE, with no impact on mortality. The results of another study were in favour of surgery within the first 7 days of antimicrobial treatment, especially among young patients with Staphylococcus aureus IE and large vegetation on echocardiography.38 However, emergency or salvage status account for the highest mortality rates in a recent register including over 19 000 patients operated on for IE.21 Unlike patients who are relatively stable pre-operatively, our results suggest that correction of valvular dysfunction and debridement of infected tissue may be insufficient in critical IE receiving salvage surgical therapy to significantly improve outcomes.

Of note, we did not identify factors previously associated with post-operative outcomes such as heart failure or shock,21 likely due to the weight of the SOFA score.

Patients denied surgery

The severity of the population studied here may explain why up to one-third of patients were denied surgery, with subsequent extremely poor outcomes. Decision-making is most difficult in patients in whom indications and contraindications for cardiac surgery coexist. Multiorgan failure was the second most common reason for denying surgery. The practitioner's decisions seem somehow validated by the extremely poor outcomes in those with high SOFA scores, regardless the treatment.

Strength and limitations

This is the first large multicentre prospective study that assesses long-term outcomes in critical IE. While our initial report provided 3-month follow-up data, we report here long-term outcomes.6 The study also addresses surgery in critical IE carried out during a relatively short period, thereby avoiding bias due to changes and advances in the field.

This study underlines the importance of scoring organ failure when considering surgery. However, our study carries a number of limitations. This observational study included primary and tertiary centres possibly leading to different strategies regarding surgery. Therefore, we cannot exclude the possibility that some patients developed complications that could have been prevented by a more timely intervention. We decided not to resort to propensity score analysis because of the limited number of patients included in our study. However, a survivor treatment selection bias was limited by the use of a Cox model in the analysis.

Management should be improved to prevent complications leading to admission to the ICU. Key factors may include early optimal antibiotic regimen and early referral to tertiary medico-surgical centres.24 A multidisciplinary approach should be recommended to assess the likely benefit of early surgery in the group of extremely ill patients.24,39 However, ways to reduce IE mortality in the critically ill still remain to be better defined.

Conclusion

Mortality in patients with critical IE remains unacceptably high. Age, multiorgan failure, prosthetic mechanical valve IE, vegetation size, and surgical treatment were associated with long-term mortality. Immediate pre-operative severe multiorgan failure carries poor post-operative prognosis. The use of the SOFA score may be of interest when assessing patients with critical IE for surgery. Surgical timing was not associated with post-operative outcomes. Our results should foster investigators to include patients admitted to the ICU with IE in future international trials assessing surgery.

Funding

This work was supported by a grant from the Société de Réanimation de Langue Française (SRLF).

Conflict of interest: none declared.

Acknowledgements

We thank Dr Tom Parks for his editorial assistance. We are grateful to all Members of the ENDOcardite en REAnimation Study Group: R. Sonneville and M. Wolff, Service de Réanimation Médicale et Infectieuse, CHU Bichat–Claude-Bernard, Paris, France; M. Mirabel and J.–L. Trouillet, Service de Réanimation Médicale, CHU La Pitié–Salpétrière, Paris, France; A. Mekontso-Dessap, Service de Réanimation Médicale, CHU Henri Mondor, Créteil, France; A. Tabah and J.-F. Timsit, Service de Réanimation Médicale, CHU de Grenoble, Grenoble, France; K. Chergui, Service de Réanimation Médicale, CH Sud Francilien, Corbeil-Essonnes, France; A. Alvarez, Service de Réanimation Polyvalente, CH Delafontaine, Saint-Denis, France; O. Pajot, Service de Réanimation Polyvalente, CH Victor Dupouy, Argenteuil, France; F. Bruneel, Service de Réanimation Médico-Chirurgicale, CH AndreMignot, Le Chesnay, France; J.P. Frat, Service de Réanimation Médicale, CHU de Poitiers, Poitiers, France; J. C. Navellou, Service de Réanimation, CHU, Hôpital Jean-Minjoz, Besançon, France; P.-E. Bollaert, Service de Réanimation Médicale, CHU de Nancy Hôpital Central, Nancy, France; J. Allardet-Servent, Service de Réanimation Médicale, CHU de Marseille–Hôpital Sainte Marguerite, Marseille, France; G. Prat and J.M. Tonnelier, Service de Réanimation Médicale, CHU de la Cavale Blanche, Brest, France; A. Kouatchet, Service de Réanimation Médicale et de Médecine Hyperbare, CHU d'Angers, Angers, France; J. Baudot, Service de Réanimation Polyvalente, CH de Nevers, Nevers, France; A. Rabbat, Service de Réanimation Médicale et Respiratoire, CHU Hôtel-Dieu, Paris, France; I. Coquet, Service de Réanimation Médicale, CHU Saint-Louis, Paris, France; E. Maury, Service de Réanimation Médicale, CHU Saint-Antoine, Paris, France; C. Lamer, Service de Réanimation Polyvalente, Institut Montsouris, Paris, France; A. Novara, Service de Réanimation Médicale, CHU HEGP, Paris, France; P. Fangio, Service de Réanimation Médico-Chirurgicale, CH de Poissy, Poissy, France; B. Megarbane, Service de Réanimation Médicale et Toxicologique, CHU Lariboisière, Paris, France; D. Prat, Service de Réanimation Médicale Polyvalente, CHU Antoine- Béclère, Clamart, France; A. Ouchenir, Service de Réanimation Polyvalente, CH Louis Pasteur, Chartres, France; S. Lavoué, Service des Maladies Infectieuses et Réanimation Médicale, Hopital Pontchaillou, CHU de Rennes, Rennes, France; A. Ait Hssain, Service de Réanimation Médico-Chirurgicale, CHU Gabriel Montpied, Clermont-Ferrand, France; P. Vignon, Service de Réanimation Polyvalente, CIC-P0801 et Université de Limoges, Limoges, France; H. Hyvernat, Réanimation Médicale, CHU de Nice Sophia Antipolis, Nice, France; C. Richard, Service de Réanimation Médicale, CHU de Bicêtre, Le Kremlin-Bicêtre, France; A. Perez and B. Levy, Service de Réanimation Médicale, CHU Nancy–Hôpital Brabois adultes, Vandoeuvre-les-Nancy, France; A. Mahrez, Service de Réanimation Médico-Chirurgicale, CH Simone Veil, Eaubonne, France; J. Charpentier, Service de Réanimation Médicale, CHU Cochin–Saint-Vincent de Paul, Paris, France; D. Combaux, Service de Réanimation Médico-Chirurgicale, CH de Compiègne, Compiègne, France; and M. Misset, Service de Réanimation Médicale Saint-Joseph, Paris.

Footnotes

  • M.W. and J.-L.T. contributed equally.

  • Members listed in the Acknowledgements section.

References

View Abstract