Alimentary Pharmacology and Therapeutics
Appropriate and timely antimicrobial therapy in cirrhotic patients with spontaneous bacterial peritonitis-associated septic shock: a retrospective cohort study C. J. Karvellas*,†, J. G. Abraldes†, Y. M. Arabi‡ & A. Kumar§ For the Cooperative Antimicrobial Therapy of Septic Shock (CATSS) Database Research Group
*Division of Critical Care Medicine and Gastroenterology/Hepatology, University of Alberta, Edmonton, AB, Canada. † Division of Gastroenterology and Hepatology, University of Alberta, Edmonton, AB, Canada. ‡ Intensive Care Department, College of Medicine, King Saud Bin Abdulaziz University for Health Sciences and King Abdullah International Medical Research Center, Riyadh, Kingdom of Saudi Arabia. § Section of Critical Care Medicine and Section of Infectious Diseases, Health Sciences Center and St. Boniface Hospital, University of Manitoba, Winnipeg, MB, Canada.
SUMMARY
Correspondence to: Dr C. J. Karvellas, Division of Gastroenterology (Liver Unit), Division of Critical Care Medicine, University of Alberta, 1-40 Zeidler Ledcor Building, Edmonton, AB T6G-2X8, Canada. E-mail: [email protected]
Results
Publication data Submitted 27 November 2014 First decision 19 December 2014 Resubmitted 7 January 2015 Resubmitted 28 January 2015 Accepted 2 February 2015 EV Pub Online 20 February 2015 This paper followed the STROBE guideline for reporting retrospective studies (BMJ 2007), File S1.
Background Spontaneous bacterial peritonitis (SBP)-associated septic shock carries significant mortality in cirrhosis.
Aim To determine whether practice-related aspects of antimicrobial therapy contribute to high mortality.
Methods Retrospective cohort study of all (n = 126) cirrhotics with spontaneous bacterial peritonitis (neutrophil count >250 or positive ascitic culture)-associated septic shock (1996–2011) from an international, multicenter database. Appropriate antimicrobial therapy implied either in vitro activity against a subsequently isolated pathogen (culture positive) or empiric management consistent with broadly accepted norms (culture negative). Overall hospital mortality was 81.8%. Comparing survivors (n = 23) with non-survivors (n = 103), survivors had lower Acute Physiology and Chronic Health Evaluation (APACHEII) (mean s.d.; 22 7 vs. 32 8) and model for end-stage liver disease (MELD) (24 9 vs. 34 11) scores and serum lactate on admission (4.9 3.1 vs. 8.9 5.3), P < 0.001 for all. Survivors were less likely to receive inappropriate initial antimicrobial therapy (0% vs. 25%, P = 0.013) and received appropriate antimicrobial therapy earlier [median 1.8 (1.1–5.2) vs. 9.5 (3.9–14.3) h, P < 0.001]. After adjusting for covariates, APACHEII [OR, odds ratio 1.45 (1.04– 2.02) per 1 unit increment, P = 0.03], lactate [OR 2.34 (1.04–5.29) per unit increment, P = 0.04] and time delay to appropriate antimicrobials [OR 1.86 (1.10–3.14) per hour increment, P = 0.02] were significantly associated with increased mortality.
Conclusions Cirrhotic patients with septic shock secondary to spontaneous bacterial peritonitis have high mortality (>80%). Each hour of delay in appropriate antimicrobial therapy was associated with a 1.86 times increased hospital mortality. Admission APACHEII and serum lactate also significantly impacted hospital mortality. Earlier initiation of appropriate antimicrobial therapy could substantially improve outcome.
Aliment Pharmacol Ther 2015; 41: 747–757
ª 2015 John Wiley & Sons Ltd doi:10.1111/apt.13135
747
C. J. Karvellas et al. INTRODUCTION Patients with cirrhosis and acute on chronic liver failure (ACLF) are abnormally susceptible to infection as a result of persistent defects in both humoral and cell mediated immunity.1 Contributory mechanisms include decreased hepatic production of complement (reduced C3 and C5 levels), impaired Kupffer cell function (phagocytosis), altered neutrophil chemotaxis (presence of chemotactic inhibitors) and impaired clearance of inflammatory cytokines.2–5 These patients may also demonstrate decreased bactericidal activities of immunoglobulins against Gram-negative pathogens.6 In the presence of cirrhosis, filtering of portal blood by hepatic reticuloendothelial cells is impaired with diminished clearance of microbial pathogens.7 Recent studies have shown that patients with decompensated cirrhosis (Child Pugh C) also suffer from down regulated HLA-DR expression on monocytes leading to further immune paralysis.8, 9 Spontaneous bacterial peritonitis (SBP), a complication of cirrhosis, can have devastating consequences. Cirrhotics have a specific predisposition to SBP due to disease-associated intestinal bacterial overgrowth, intestinal dysmotility and increased intestinal permeability leading to an increase in bacterial translocation.10 Together with portosystemic shunting and reduced hepatic clearance, this leads to SBP and bloodstream infection (BSI) that can culminate in septic shock.11–13 As infection in cirrhosis is often initially asymptomatic but potentially life threatening, withholding treatment until a causative organism is identified and antimicrobial susceptibility data made available or until the patient develops hypotension is not an option.14, 15 Previous studies have revealed that the timing and appropriateness of empiric antimicrobial therapy has a significant impact on patients with septic shock with a broad range of infections including those with cirrhosis.16, 17 However little has been published directly related to SBP. Given its insidious onset and significant clinical consequences, early recognition and appropriate management is vital. In this study, we attempted to determine correlates of outcome in SBP-associated septic shock among cirrhotic patients. Specific covariates that were examined included time to appropriate antimicrobial therapy, co-existing BSI and indicators/scores of severity of illness including the modified end-stage liver disease (MELD) and acute physiology and chronic health evaluation (APACHEII) scores. MATERIALS AND METHODS This was a nested cohort study within a retrospective database [the Cooperative Antimicrobial Therapy of 748
Septic Shock (CATSS) Database] of cases of septic shock collected from 28 medical centres in Canada, the United States, and Saudi Arabia for discrete, continuous periods between 1996 and 2011. The details of the setting have been described elsewhere.18 Approval was obtained from the Institutional Review Boards of all participating institutions. This study was written according to the STROBE guideline for reporting retrospective studies.19
Study design: patients and setting Data were extracted for all (n = 126) adult cirrhotic (biopsy-proven cirrhosis, documented variceal haemorrhage or portal hypertension, hepatic ascites, or encephalopathy) patients with SBP (neutrophil count >250 or positive ascitic culture)-associated septic shock in the CATSS database. Exposures and outcomes Baseline patient characteristics including demographics and comorbid conditions were obtained from all participants at enrolment into the registry. The following data were obtained on day 1 of septic shock (most aberrant values in the initial 24 h following presentation): serum lactate, bilirubin, creatinine, and bicarbonate levels, platelet count, international normalised ratio and white blood cell (WBC) count, and acute physiology and chronic health evaluation (APACHE) II score.20 The model for end-stage liver disease (MELD) score was calculated on day 1 as described previously.21 The primary outcome of this study was in-hospital mortality Secondary outcomes were hospital and ICU lengths of stay. Operational definitions Septic shock was defined using the 1992 ACCP/SCCM guidelines.22 As per that definition, case patients were required to have documented or suspected infection, persistent hypotension requiring vasopressors, and two of the following four elements: (i) a heart rate of >90 beats/min; (ii) a respiratory rate of >20 breaths/min or arterial partial pressure of carbon dioxide (PaCO2) of 10%. An episode of hypotension was considered to represent the initial onset of septic shock when hypotension persisted from onset despite fluid (2 L of saline or equivalent) administration (persistent hypotension) or hypotension was only transiently improved (hypotension resolution for 48 h after hospital admission. Cases not meeting this definition were considered to be septic shock associated with community-acquired infections. We used the term ‘immunocompromised patients’ for a subgroup of patients who had one of the following comorbidities: acquired immunodeficiency syndrome, acute or chronic lymphoma, acute or chronic leukaemia/multiple myeloma, metastatic solid cancer, immunosuppressive chemotherapy, or long-term steroid therapy (>10 mg prednisone equivalent daily). Other patients were labelled as ‘non-immunocompromised’. Predetermined rules were used to assess the appropriateness and delays of initial antimicrobial therapy.16, 18, 23 For culture-positive septic shock, initial antimicrobial therapy was considered appropriate if an antimicrobial with in vitro activity appropriate for the isolated pathogen or pathogens was the first new antimicrobial agent given after the onset of recurrent or persistent hypotension or was initiated within 6 h of the administration of the first new antimicrobial agent. Otherwise, the initial therapy was considered inappropriate.18 For culture-negative septic shock, initial therapy was considered appropriate when an antimicrobial agent consistent with broadly accepted norms for empiric management of the typical pathogens for the clinical syndrome was the new antimicrobial agent given after the onset of recurrent or persistent hypotension or was initiated within 6 h of administration of the first new antimicrobial agent.18 At each participating institution, infectious disease physicians/microbiologists were consulted to account for the local community and nosocomial flora when considering appropriateness of empiric therapy during the period covered by data collection. Otherwise, appropriate empiric therapy of culture-negative infections leading to septic shock was based on the recommendations listed in the ‘Clinical Approach to Initial Choice of Antimicrobial Therapy’ in the Sanford Guide to Antimicrobial Therapy 2004 (34th edition).24 For the purposes of this study, antibiotic monotherapy was defined as the administration of any single, appropriate, intravenous, preferably bactericidal antibiotic at Aliment Pharmacol Ther 2015; 41: 747–757 ª 2015 John Wiley & Sons Ltd
any point after the onset of recurrent or persistent hypotension. Combination therapy was defined as the concomitant use of two or more antibiotics of different mechanistic classes with activity for the isolated or suspected pathogens. The second agent had to be started within 24 h of the first antibiotic or within 24 h of the onset of hypotension (if the first agent was initiated before hypotension was documented).23 Patients with septic shock due to yeast, anaerobic or mycobacterial infection were excluded from this analysis of single vs. combined antibiotics.
Statistical analysis Statistical analysis was performed using SPSS version 20 (Chicago, IL, USA) and R (www.r-project.org). Continuous variables are reported as the mean (s.d.) and median (interquartile, IQR range) following testing for normality. Measure of central tendency for continuous variables was compared using the Student’s t-test (normal distribution) and Wilcoxon Rank sum test (nonparametric variables) following normality testing. Categorical variables are reported as numbers and percentage and were analysed using the v2 test and Fisher’s exact test for number of cases less than 5. Statistical significance (95% confidence interval, CI) was reported for a P < 0.05 with a trend towards significance for a P < 0.15. To study the association between timing and appropriateness of antimicrobials and in-hospital mortality (dependent variable), multivariable logistic regression analyses were performed. Covariates were included based on their significance in the univariable regression analysis (P < 0.05), demographic/physiological plausibility and colinearity: age, sex, APACHE II, lactate (admission), number of organ failures (day 1) and presence of concomitant BSI. Collinear variables were excluded (MELD). Missing data were neither imputed nor replaced. Results were reported as adjusted odds ratios (OR) and 95% CI. P < 0.05 was considered significant. Model performance was assessed using the c-statistic (area under the receiver operator curve ~AUROC) and Hosmer–Lemeshow goodness of fit test. Unadjusted and adjusted risk plots were constructed with the visreg and rms R packages. RESULTS Participants and descriptive data In total, 126 cirrhotic patients with SBP [mean age (s.d.) 55 (13) years, 60% male] were included in this analysis (see Table 1). On presentation with septic shock, mean APACHEII score was 30 (9) and MELD 32 (11). Patients 749
C. J. Karvellas et al. Table 1 | Baseline characteristics of 126 patients with cirrhosis, spontaneous bacterial peritonitis and septic shock
Demographics Age (years) Sex (Male) Body mass index (kg/m2) Severity scores APACHE II MELD Organ failure Number of organ failures on day 1 Day 3 Day 5 RRT (any day) Vasopressor use (any day) Mechanical ventilation (any day) Organ failures (day 1) Cardiovascular Renal Respiratory Haematological Metabolic CNS Hepatic Coagulation Most abnormal lab findings within first 24 h of shock Sodium (mmol/L) Bicarbonate (mmol/L) Lactate (mmol/L) Bilirubin (lmol/L) Creatinine (lmol/L) Albumin (lmol/L) Platelet count (9109/L) INR White blood count (9109/L) Therapies APC Corticosteroids (hydrocortisone) Lengths of Stay: ICU Lengths of stay: in-hospital Overall Survivors Non-survivors Survival ICU Hospital 90-day (overall)* Appropriate antimicrobials administered Pre-shock Appropriate antimicrobials administered Post-shock
N (total)
Mean s.d. or Median (IQR)
126 126 50
54.5 12.6
118 97
29.8 8.8 32 (11)
126 126 126 126 126 122
5.7 1.6 6.2 1.5 6.3 1.5
76 (60)
26.5 6.0
15 (11) 126 (100) 106 (87)
126 126 126 126 126 126 126 126 46 81 95 120 124 25 113 110 121
N (%)
126 97 83 68 85 62 94 100
(100)/126 (100) (77.0)/103 (81.7) (65.9)/100 (79.4) (54.0)/103 (81.7) (67.5)/97 (77.0) (49.2)/67 (53.2) (74.6)/97 (77.0) (79.4)/105 (83)
137 8 13.7 5.4 8.2 5.2 99 (48–289) 239 (156–366) 21 5 86 (53–136) 2.3 (1.8–3.2) 13.9 (14.8–19)
126 126
1 28 5 (3–10)
126 23 103 126 126 126 25 97
8 (3–14) days 17 days 6 days 35 23 23/126 3/25 20/97
(27.8) (18.2) (18.3) (12) (21)
* Assuming patients discharged home lived to 90 days. Organ failure definitions: Cardiovascular: vasopressors required to support target MAP; Renal: elevation of normal baseline serum creatinine to >1.5 9 normal value; Respiratory: mechanical ventilation required (more than immediate for surgery); Haematologic: platelet count 3 mmol/L (1.5 times the upper limit of the normal); Central nervous system: an acute alteration in mental status not attributable to sedation; Hepatic: elevation of serum total bilirubin to >35 lmol/L.
750
Aliment Pharmacol Ther 2015; 41: 747–757 ª 2015 John Wiley & Sons Ltd
Antimicrobial therapy in spontaneous bacterial peritonitis-septic shock had a mean of 5.7 (1.6), 6.2 (1.5) and 6.3 (1.5) organ failures on days 1, 3 and 5 of study respectively. During hospital stay, 15 (11%) were placed on renal replacement therapy (RRT), 106 (87%) were mechanically ventilated and 100% of patients required vasopressor support (as per definition). The peak mean serum lactate level (within 24 h) was 8.2 (5.2) mmol/L. For the overall cohort, the median (IQR) bilirubin was 120 (48– 289) lmol/L, creatinine 239 (156–366) lmol/L and INR 2.3 (1.8–3.2) (first 24 h maximum). The median overall length of stay in ICU was 5 (3–10) days and 8 (3–14) days in hospital; 6 (3–13) and 17 (9–33) days respectively for survivors and 5 (2–9) and 6 (3–12) days respectively for non-survivors. Thirty-five patients (27.8%) survived to ICU discharge while only 23 (18.2%) survived to hospital discharge (primary endpoint).
Microbiology descriptive data Pathogen and diagnostic information is shown in Table 2. Of 126 patients, 86 (68%) were culture positive (either ascitic or bloodstream) while 40 (32%) were culture-negative (i.e. >250 neutrophils on ascitic tap). Fifty-six (44%) had a positive bloodstream culture. Seventy-three (58%) patients had community-acquired SBP (250 PMNs 40/126 in ascitic fluid) Positive (+Ascitic tap or 86/126 bacteremia + >250 PMNs) Bacteremia positive 56/126 Bacteremia negative 70/126 Infection type Community acquired 73 (58%) (SBP 48 h of admission) Positive cultures N = 86 Ascites only 31 Blood only (high WBC 17 in ascites) Ascites and blood 38 Gram-negative (56%) 48 E. coli 27 Klebsiella sp. 9 Pseudomonas aeruginosa 3 Enterobacter sp. 3 Citrobacter sp. 2 Acinetobacter sp. 2 Stenotrophomonas sp. 1 Other Gram-negative 1 Gram positive (29%) 26 Staphylococcus aureus 4 Staphylococcus pneumoniae 9 Staphylococcus faecalis 3 Group A Streptococci 0 Other b-haemolytic strep 3 Strep viridans 3 Strep faecium 2 Other Gram-positive 2 Anaerobes (2%) 1 Fungal/Candida (13%) 11 Candida albicans 9 Candida glabrata/tropicalis 2 Timing of appropriate antimicrobials Administered pre-shock 25 Administered post-shock 0–3 h 27 3–6 h 16 6–12 h 28 >12 h or never* 30 Pathogen identified 86 Appropriate antimicrobials 67 (78%) Inappropriate antimicrobials 19 (22%)
Survivors (%) 11/40 (28) 12/86 (14)
5/56 (9) 18/70 (26) 15/73 (21)
8/53 (15)
7/48 (15)
4/26 (15)
1/1 (100) 0/11 (0)
3/25 (12) 14/27 3/16 3/28 0/30
(52) (19) (11) (0)
12/67 (18) 0/19 (0)
* Four patients never received appropriate antimicrobial therapy.
751
C. J. Karvellas et al. who received appropriate antimicrobials only after diagnosis of shock, 27 received them within 3 h with 14 (52%) surviving until hospital discharge. Sixteen patients received antimicrobials from 3 to 6 h (3, 19% survived) and 28 received antimicrobials between 6 and 12 h (3, 11% survived). Twenty-six patients received appropriate antimicrobial agents after 12 h with none surviving until hospital discharge. Overall delays of >6 h was associated with 5.5% survival to hospital discharge.
Unadjusted comparisons: hospital survivors vs. non-survivors. Unadjusted comparisons of hospital survivors (n = 23) with non-survivors (n = 103) are shown in Table 3. Survivors had lower mean APACHE II (22 vs. 32) and MELD (24 vs. 34, P < 0.001 for both) on admission. Survivors also had lower number of organ failures on day 1 (mean 4.6 vs. 5.9), day 3 (5.5 vs.
6.4) and day 5 (5.5. vs. 6.5, P < 0.02 for all) respectively. Within initial 24 h of septic shock, cirrhotic survivors had lower serum median bilirubin (50 vs. 127 lmol/L), creatinine (159 vs. 246 lmol/L) and INR values (1.9 vs. 2.4, P < 0.01 for all). Survivors also had lower mean serum lactate (4.9 vs. 8.9 mmol/L, P < 0.001). Survivors less frequently had concomitant BSI (22% vs. 50%, P = 0.015). Survivors had a significantly shorter median delay to appropriate antimicrobial administration following the initial documentation of hypotension (1.8 vs. 9.5 h, P < 0.001). Finally, survivors had significantly longer median lengths of stay in hospital (17 vs. 6 days, P < 0.001). Figure 1 shows the relation between APACHEII, serum lactate and time delay to antimicrobial therapy and in-hospital mortality. Figure 2 shows the relationship between predicted death rates according to increase time to antimicrobials is synergistic (worse outcomes) with
Table 3 | Univariate (crude) comparisons of 23 cirrhotics presenting with spontaneous bacterial peritonitis and septic shock that survived to hospital discharge with 103 non-survivors
Male Age BMI (m/kg2) APACHEII Number of organ failures Day 1 Day 3 Day 5 MELD Vitals Temperature (°C) Heart rate Respiratory rate Biochemistry White blood count (910/L) Platelet count (910/L) Bilirubin (lmol/L) Creatinine (lmol/L) Bicarbonate (mmol/L) Lactate (mmol/L) INR Infection Nosocomial Positive blood culture Positive site or blood culture Appropriate empiric antimicrobials administered Time delay to appropriate antimicrobials (h) Drotrecogin-alfa (activated) Corticosteroids ICU LOS Hospital LOS 752
N
Survivors (n = 23)
23 23 10 22
15 53.3 (14.8) 26.0 (6.2) 22.2 (7.4)
23 23 23 21
Non-survivors (n = 103)
P
103 103 40 96
61 54.8 (12.1) 26.6 (6.0) 31.6 (8.1)
0.595 0.623 0.318
Appropriate and timely antimicrobial therapy in cirrhotic patients with spontaneous bacterial peritonitis-associated septic shock: a retrospective cohort study C. J. Karvellas*,†, J. G. Abraldes†, Y. M. Arabi‡ & A. Kumar§ For the Cooperative Antimicrobial Therapy of Septic Shock (CATSS) Database Research Group
*Division of Critical Care Medicine and Gastroenterology/Hepatology, University of Alberta, Edmonton, AB, Canada. † Division of Gastroenterology and Hepatology, University of Alberta, Edmonton, AB, Canada. ‡ Intensive Care Department, College of Medicine, King Saud Bin Abdulaziz University for Health Sciences and King Abdullah International Medical Research Center, Riyadh, Kingdom of Saudi Arabia. § Section of Critical Care Medicine and Section of Infectious Diseases, Health Sciences Center and St. Boniface Hospital, University of Manitoba, Winnipeg, MB, Canada.
SUMMARY
Correspondence to: Dr C. J. Karvellas, Division of Gastroenterology (Liver Unit), Division of Critical Care Medicine, University of Alberta, 1-40 Zeidler Ledcor Building, Edmonton, AB T6G-2X8, Canada. E-mail: [email protected]
Results
Publication data Submitted 27 November 2014 First decision 19 December 2014 Resubmitted 7 January 2015 Resubmitted 28 January 2015 Accepted 2 February 2015 EV Pub Online 20 February 2015 This paper followed the STROBE guideline for reporting retrospective studies (BMJ 2007), File S1.
Background Spontaneous bacterial peritonitis (SBP)-associated septic shock carries significant mortality in cirrhosis.
Aim To determine whether practice-related aspects of antimicrobial therapy contribute to high mortality.
Methods Retrospective cohort study of all (n = 126) cirrhotics with spontaneous bacterial peritonitis (neutrophil count >250 or positive ascitic culture)-associated septic shock (1996–2011) from an international, multicenter database. Appropriate antimicrobial therapy implied either in vitro activity against a subsequently isolated pathogen (culture positive) or empiric management consistent with broadly accepted norms (culture negative). Overall hospital mortality was 81.8%. Comparing survivors (n = 23) with non-survivors (n = 103), survivors had lower Acute Physiology and Chronic Health Evaluation (APACHEII) (mean s.d.; 22 7 vs. 32 8) and model for end-stage liver disease (MELD) (24 9 vs. 34 11) scores and serum lactate on admission (4.9 3.1 vs. 8.9 5.3), P < 0.001 for all. Survivors were less likely to receive inappropriate initial antimicrobial therapy (0% vs. 25%, P = 0.013) and received appropriate antimicrobial therapy earlier [median 1.8 (1.1–5.2) vs. 9.5 (3.9–14.3) h, P < 0.001]. After adjusting for covariates, APACHEII [OR, odds ratio 1.45 (1.04– 2.02) per 1 unit increment, P = 0.03], lactate [OR 2.34 (1.04–5.29) per unit increment, P = 0.04] and time delay to appropriate antimicrobials [OR 1.86 (1.10–3.14) per hour increment, P = 0.02] were significantly associated with increased mortality.
Conclusions Cirrhotic patients with septic shock secondary to spontaneous bacterial peritonitis have high mortality (>80%). Each hour of delay in appropriate antimicrobial therapy was associated with a 1.86 times increased hospital mortality. Admission APACHEII and serum lactate also significantly impacted hospital mortality. Earlier initiation of appropriate antimicrobial therapy could substantially improve outcome.
Aliment Pharmacol Ther 2015; 41: 747–757
ª 2015 John Wiley & Sons Ltd doi:10.1111/apt.13135
747
C. J. Karvellas et al. INTRODUCTION Patients with cirrhosis and acute on chronic liver failure (ACLF) are abnormally susceptible to infection as a result of persistent defects in both humoral and cell mediated immunity.1 Contributory mechanisms include decreased hepatic production of complement (reduced C3 and C5 levels), impaired Kupffer cell function (phagocytosis), altered neutrophil chemotaxis (presence of chemotactic inhibitors) and impaired clearance of inflammatory cytokines.2–5 These patients may also demonstrate decreased bactericidal activities of immunoglobulins against Gram-negative pathogens.6 In the presence of cirrhosis, filtering of portal blood by hepatic reticuloendothelial cells is impaired with diminished clearance of microbial pathogens.7 Recent studies have shown that patients with decompensated cirrhosis (Child Pugh C) also suffer from down regulated HLA-DR expression on monocytes leading to further immune paralysis.8, 9 Spontaneous bacterial peritonitis (SBP), a complication of cirrhosis, can have devastating consequences. Cirrhotics have a specific predisposition to SBP due to disease-associated intestinal bacterial overgrowth, intestinal dysmotility and increased intestinal permeability leading to an increase in bacterial translocation.10 Together with portosystemic shunting and reduced hepatic clearance, this leads to SBP and bloodstream infection (BSI) that can culminate in septic shock.11–13 As infection in cirrhosis is often initially asymptomatic but potentially life threatening, withholding treatment until a causative organism is identified and antimicrobial susceptibility data made available or until the patient develops hypotension is not an option.14, 15 Previous studies have revealed that the timing and appropriateness of empiric antimicrobial therapy has a significant impact on patients with septic shock with a broad range of infections including those with cirrhosis.16, 17 However little has been published directly related to SBP. Given its insidious onset and significant clinical consequences, early recognition and appropriate management is vital. In this study, we attempted to determine correlates of outcome in SBP-associated septic shock among cirrhotic patients. Specific covariates that were examined included time to appropriate antimicrobial therapy, co-existing BSI and indicators/scores of severity of illness including the modified end-stage liver disease (MELD) and acute physiology and chronic health evaluation (APACHEII) scores. MATERIALS AND METHODS This was a nested cohort study within a retrospective database [the Cooperative Antimicrobial Therapy of 748
Septic Shock (CATSS) Database] of cases of septic shock collected from 28 medical centres in Canada, the United States, and Saudi Arabia for discrete, continuous periods between 1996 and 2011. The details of the setting have been described elsewhere.18 Approval was obtained from the Institutional Review Boards of all participating institutions. This study was written according to the STROBE guideline for reporting retrospective studies.19
Study design: patients and setting Data were extracted for all (n = 126) adult cirrhotic (biopsy-proven cirrhosis, documented variceal haemorrhage or portal hypertension, hepatic ascites, or encephalopathy) patients with SBP (neutrophil count >250 or positive ascitic culture)-associated septic shock in the CATSS database. Exposures and outcomes Baseline patient characteristics including demographics and comorbid conditions were obtained from all participants at enrolment into the registry. The following data were obtained on day 1 of septic shock (most aberrant values in the initial 24 h following presentation): serum lactate, bilirubin, creatinine, and bicarbonate levels, platelet count, international normalised ratio and white blood cell (WBC) count, and acute physiology and chronic health evaluation (APACHE) II score.20 The model for end-stage liver disease (MELD) score was calculated on day 1 as described previously.21 The primary outcome of this study was in-hospital mortality Secondary outcomes were hospital and ICU lengths of stay. Operational definitions Septic shock was defined using the 1992 ACCP/SCCM guidelines.22 As per that definition, case patients were required to have documented or suspected infection, persistent hypotension requiring vasopressors, and two of the following four elements: (i) a heart rate of >90 beats/min; (ii) a respiratory rate of >20 breaths/min or arterial partial pressure of carbon dioxide (PaCO2) of 10%. An episode of hypotension was considered to represent the initial onset of septic shock when hypotension persisted from onset despite fluid (2 L of saline or equivalent) administration (persistent hypotension) or hypotension was only transiently improved (hypotension resolution for 48 h after hospital admission. Cases not meeting this definition were considered to be septic shock associated with community-acquired infections. We used the term ‘immunocompromised patients’ for a subgroup of patients who had one of the following comorbidities: acquired immunodeficiency syndrome, acute or chronic lymphoma, acute or chronic leukaemia/multiple myeloma, metastatic solid cancer, immunosuppressive chemotherapy, or long-term steroid therapy (>10 mg prednisone equivalent daily). Other patients were labelled as ‘non-immunocompromised’. Predetermined rules were used to assess the appropriateness and delays of initial antimicrobial therapy.16, 18, 23 For culture-positive septic shock, initial antimicrobial therapy was considered appropriate if an antimicrobial with in vitro activity appropriate for the isolated pathogen or pathogens was the first new antimicrobial agent given after the onset of recurrent or persistent hypotension or was initiated within 6 h of the administration of the first new antimicrobial agent. Otherwise, the initial therapy was considered inappropriate.18 For culture-negative septic shock, initial therapy was considered appropriate when an antimicrobial agent consistent with broadly accepted norms for empiric management of the typical pathogens for the clinical syndrome was the new antimicrobial agent given after the onset of recurrent or persistent hypotension or was initiated within 6 h of administration of the first new antimicrobial agent.18 At each participating institution, infectious disease physicians/microbiologists were consulted to account for the local community and nosocomial flora when considering appropriateness of empiric therapy during the period covered by data collection. Otherwise, appropriate empiric therapy of culture-negative infections leading to septic shock was based on the recommendations listed in the ‘Clinical Approach to Initial Choice of Antimicrobial Therapy’ in the Sanford Guide to Antimicrobial Therapy 2004 (34th edition).24 For the purposes of this study, antibiotic monotherapy was defined as the administration of any single, appropriate, intravenous, preferably bactericidal antibiotic at Aliment Pharmacol Ther 2015; 41: 747–757 ª 2015 John Wiley & Sons Ltd
any point after the onset of recurrent or persistent hypotension. Combination therapy was defined as the concomitant use of two or more antibiotics of different mechanistic classes with activity for the isolated or suspected pathogens. The second agent had to be started within 24 h of the first antibiotic or within 24 h of the onset of hypotension (if the first agent was initiated before hypotension was documented).23 Patients with septic shock due to yeast, anaerobic or mycobacterial infection were excluded from this analysis of single vs. combined antibiotics.
Statistical analysis Statistical analysis was performed using SPSS version 20 (Chicago, IL, USA) and R (www.r-project.org). Continuous variables are reported as the mean (s.d.) and median (interquartile, IQR range) following testing for normality. Measure of central tendency for continuous variables was compared using the Student’s t-test (normal distribution) and Wilcoxon Rank sum test (nonparametric variables) following normality testing. Categorical variables are reported as numbers and percentage and were analysed using the v2 test and Fisher’s exact test for number of cases less than 5. Statistical significance (95% confidence interval, CI) was reported for a P < 0.05 with a trend towards significance for a P < 0.15. To study the association between timing and appropriateness of antimicrobials and in-hospital mortality (dependent variable), multivariable logistic regression analyses were performed. Covariates were included based on their significance in the univariable regression analysis (P < 0.05), demographic/physiological plausibility and colinearity: age, sex, APACHE II, lactate (admission), number of organ failures (day 1) and presence of concomitant BSI. Collinear variables were excluded (MELD). Missing data were neither imputed nor replaced. Results were reported as adjusted odds ratios (OR) and 95% CI. P < 0.05 was considered significant. Model performance was assessed using the c-statistic (area under the receiver operator curve ~AUROC) and Hosmer–Lemeshow goodness of fit test. Unadjusted and adjusted risk plots were constructed with the visreg and rms R packages. RESULTS Participants and descriptive data In total, 126 cirrhotic patients with SBP [mean age (s.d.) 55 (13) years, 60% male] were included in this analysis (see Table 1). On presentation with septic shock, mean APACHEII score was 30 (9) and MELD 32 (11). Patients 749
C. J. Karvellas et al. Table 1 | Baseline characteristics of 126 patients with cirrhosis, spontaneous bacterial peritonitis and septic shock
Demographics Age (years) Sex (Male) Body mass index (kg/m2) Severity scores APACHE II MELD Organ failure Number of organ failures on day 1 Day 3 Day 5 RRT (any day) Vasopressor use (any day) Mechanical ventilation (any day) Organ failures (day 1) Cardiovascular Renal Respiratory Haematological Metabolic CNS Hepatic Coagulation Most abnormal lab findings within first 24 h of shock Sodium (mmol/L) Bicarbonate (mmol/L) Lactate (mmol/L) Bilirubin (lmol/L) Creatinine (lmol/L) Albumin (lmol/L) Platelet count (9109/L) INR White blood count (9109/L) Therapies APC Corticosteroids (hydrocortisone) Lengths of Stay: ICU Lengths of stay: in-hospital Overall Survivors Non-survivors Survival ICU Hospital 90-day (overall)* Appropriate antimicrobials administered Pre-shock Appropriate antimicrobials administered Post-shock
N (total)
Mean s.d. or Median (IQR)
126 126 50
54.5 12.6
118 97
29.8 8.8 32 (11)
126 126 126 126 126 122
5.7 1.6 6.2 1.5 6.3 1.5
76 (60)
26.5 6.0
15 (11) 126 (100) 106 (87)
126 126 126 126 126 126 126 126 46 81 95 120 124 25 113 110 121
N (%)
126 97 83 68 85 62 94 100
(100)/126 (100) (77.0)/103 (81.7) (65.9)/100 (79.4) (54.0)/103 (81.7) (67.5)/97 (77.0) (49.2)/67 (53.2) (74.6)/97 (77.0) (79.4)/105 (83)
137 8 13.7 5.4 8.2 5.2 99 (48–289) 239 (156–366) 21 5 86 (53–136) 2.3 (1.8–3.2) 13.9 (14.8–19)
126 126
1 28 5 (3–10)
126 23 103 126 126 126 25 97
8 (3–14) days 17 days 6 days 35 23 23/126 3/25 20/97
(27.8) (18.2) (18.3) (12) (21)
* Assuming patients discharged home lived to 90 days. Organ failure definitions: Cardiovascular: vasopressors required to support target MAP; Renal: elevation of normal baseline serum creatinine to >1.5 9 normal value; Respiratory: mechanical ventilation required (more than immediate for surgery); Haematologic: platelet count 3 mmol/L (1.5 times the upper limit of the normal); Central nervous system: an acute alteration in mental status not attributable to sedation; Hepatic: elevation of serum total bilirubin to >35 lmol/L.
750
Aliment Pharmacol Ther 2015; 41: 747–757 ª 2015 John Wiley & Sons Ltd
Antimicrobial therapy in spontaneous bacterial peritonitis-septic shock had a mean of 5.7 (1.6), 6.2 (1.5) and 6.3 (1.5) organ failures on days 1, 3 and 5 of study respectively. During hospital stay, 15 (11%) were placed on renal replacement therapy (RRT), 106 (87%) were mechanically ventilated and 100% of patients required vasopressor support (as per definition). The peak mean serum lactate level (within 24 h) was 8.2 (5.2) mmol/L. For the overall cohort, the median (IQR) bilirubin was 120 (48– 289) lmol/L, creatinine 239 (156–366) lmol/L and INR 2.3 (1.8–3.2) (first 24 h maximum). The median overall length of stay in ICU was 5 (3–10) days and 8 (3–14) days in hospital; 6 (3–13) and 17 (9–33) days respectively for survivors and 5 (2–9) and 6 (3–12) days respectively for non-survivors. Thirty-five patients (27.8%) survived to ICU discharge while only 23 (18.2%) survived to hospital discharge (primary endpoint).
Microbiology descriptive data Pathogen and diagnostic information is shown in Table 2. Of 126 patients, 86 (68%) were culture positive (either ascitic or bloodstream) while 40 (32%) were culture-negative (i.e. >250 neutrophils on ascitic tap). Fifty-six (44%) had a positive bloodstream culture. Seventy-three (58%) patients had community-acquired SBP (250 PMNs 40/126 in ascitic fluid) Positive (+Ascitic tap or 86/126 bacteremia + >250 PMNs) Bacteremia positive 56/126 Bacteremia negative 70/126 Infection type Community acquired 73 (58%) (SBP 48 h of admission) Positive cultures N = 86 Ascites only 31 Blood only (high WBC 17 in ascites) Ascites and blood 38 Gram-negative (56%) 48 E. coli 27 Klebsiella sp. 9 Pseudomonas aeruginosa 3 Enterobacter sp. 3 Citrobacter sp. 2 Acinetobacter sp. 2 Stenotrophomonas sp. 1 Other Gram-negative 1 Gram positive (29%) 26 Staphylococcus aureus 4 Staphylococcus pneumoniae 9 Staphylococcus faecalis 3 Group A Streptococci 0 Other b-haemolytic strep 3 Strep viridans 3 Strep faecium 2 Other Gram-positive 2 Anaerobes (2%) 1 Fungal/Candida (13%) 11 Candida albicans 9 Candida glabrata/tropicalis 2 Timing of appropriate antimicrobials Administered pre-shock 25 Administered post-shock 0–3 h 27 3–6 h 16 6–12 h 28 >12 h or never* 30 Pathogen identified 86 Appropriate antimicrobials 67 (78%) Inappropriate antimicrobials 19 (22%)
Survivors (%) 11/40 (28) 12/86 (14)
5/56 (9) 18/70 (26) 15/73 (21)
8/53 (15)
7/48 (15)
4/26 (15)
1/1 (100) 0/11 (0)
3/25 (12) 14/27 3/16 3/28 0/30
(52) (19) (11) (0)
12/67 (18) 0/19 (0)
* Four patients never received appropriate antimicrobial therapy.
751
C. J. Karvellas et al. who received appropriate antimicrobials only after diagnosis of shock, 27 received them within 3 h with 14 (52%) surviving until hospital discharge. Sixteen patients received antimicrobials from 3 to 6 h (3, 19% survived) and 28 received antimicrobials between 6 and 12 h (3, 11% survived). Twenty-six patients received appropriate antimicrobial agents after 12 h with none surviving until hospital discharge. Overall delays of >6 h was associated with 5.5% survival to hospital discharge.
Unadjusted comparisons: hospital survivors vs. non-survivors. Unadjusted comparisons of hospital survivors (n = 23) with non-survivors (n = 103) are shown in Table 3. Survivors had lower mean APACHE II (22 vs. 32) and MELD (24 vs. 34, P < 0.001 for both) on admission. Survivors also had lower number of organ failures on day 1 (mean 4.6 vs. 5.9), day 3 (5.5 vs.
6.4) and day 5 (5.5. vs. 6.5, P < 0.02 for all) respectively. Within initial 24 h of septic shock, cirrhotic survivors had lower serum median bilirubin (50 vs. 127 lmol/L), creatinine (159 vs. 246 lmol/L) and INR values (1.9 vs. 2.4, P < 0.01 for all). Survivors also had lower mean serum lactate (4.9 vs. 8.9 mmol/L, P < 0.001). Survivors less frequently had concomitant BSI (22% vs. 50%, P = 0.015). Survivors had a significantly shorter median delay to appropriate antimicrobial administration following the initial documentation of hypotension (1.8 vs. 9.5 h, P < 0.001). Finally, survivors had significantly longer median lengths of stay in hospital (17 vs. 6 days, P < 0.001). Figure 1 shows the relation between APACHEII, serum lactate and time delay to antimicrobial therapy and in-hospital mortality. Figure 2 shows the relationship between predicted death rates according to increase time to antimicrobials is synergistic (worse outcomes) with
Table 3 | Univariate (crude) comparisons of 23 cirrhotics presenting with spontaneous bacterial peritonitis and septic shock that survived to hospital discharge with 103 non-survivors
Male Age BMI (m/kg2) APACHEII Number of organ failures Day 1 Day 3 Day 5 MELD Vitals Temperature (°C) Heart rate Respiratory rate Biochemistry White blood count (910/L) Platelet count (910/L) Bilirubin (lmol/L) Creatinine (lmol/L) Bicarbonate (mmol/L) Lactate (mmol/L) INR Infection Nosocomial Positive blood culture Positive site or blood culture Appropriate empiric antimicrobials administered Time delay to appropriate antimicrobials (h) Drotrecogin-alfa (activated) Corticosteroids ICU LOS Hospital LOS 752
N
Survivors (n = 23)
23 23 10 22
15 53.3 (14.8) 26.0 (6.2) 22.2 (7.4)
23 23 23 21
Non-survivors (n = 103)
P
103 103 40 96
61 54.8 (12.1) 26.6 (6.0) 31.6 (8.1)
0.595 0.623 0.318
