J Antimicrob Chemother 2012; 67: 1508 – 1513 doi:10.1093/jac/dks062 Advance Access publication 9 March 2012
Epidemiology and prognostic determinants of bacteraemic biliary tract infection M. Ortega1*, F. Marco2, A. Soriano3, M. Almela2, J. A. Martı´nez3, J. Lo´pez3, C. Pitart2 and J. Mensa3 1
Emergency Department and Infectious Diseases Unit, Hospital Clı´nic, University of Barcelona, Barcelona, Spain; 2Laboratory of Clinical Microbiology, Hospital Clı´nic, University of Barcelona, Barcelona, Spain; 3Infectious Diseases Unit, Hospital Clı´nic, University of Barcelona, Barcelona, Spain *Corresponding author. Tel/Fax: +34-93-2279830; E-mail: [email protected]
Received 17 November 2011; returned 7 January 2012; revised 30 January 2012; accepted 4 February 2012 Objectives: To determine the epidemiology of bacteraemia due to biliary tract infection (BTI) and to identify independent predictors of mortality. Methods: This study was part of a bloodstream infection surveillance study that prospectively collected data on consecutive patients with bacteraemia in our institution from 1991 to 2010. BTI was the confirmed source of 1373 patients with bacteraemia, and the independent prognostic factors of 30 day mortality were determined. Results: The mean age of patients with biliary sepsis was 71 years (+14 years). The most frequent comorbidities were biliary lithiasis and solid-organ cancer [484 cases (35%) and 362 cases (26%), respectively]. The BTI was healthcare-associated in 33% of patients. Shock and mortality accounted for 209 and 126 cases, respectively (15% and 9%). The most frequent microorganisms isolated were Escherichia coli (749, 55%), Klebsiella spp. (240, 17%), Enterococcus spp. (171, 12%), Pseudomonas aeruginosa (86, 6%) and Enterobacter spp. (63, 5%). There were 47 (3%) cefotaxime-resistant (CTX-R) E. coli or Klebsiella spp. Inappropriate empirical antibiotic treatment was an independent factor associated with mortality (OR 1.4, 95% CI 1.1 –1.7). Inappropriate empirical treatment was more frequent in P. aeruginosa and CTX-R Enterobacteriaceae bacteraemia. These microorganisms were significantly more common in patients with previous antibiotic therapy, solid-organ cancer or transplantation and in healthcare-associated bacteraemia. Conclusions: In patients with bacteraemic BTI, inappropriate empirical therapy was more frequent in P. aeruginosa and CTX-R Enterobacteriaceae infection and was associated with a higher mortality rate. In patients with bacteraemia due to BTI and solid-organ cancer or transplantation, healthcare-associated infection or previous antibiotic treatment, initial therapy with piperacillin/tazobactam or a carbapenem would be advisable. Keywords: cholangitis, cholecystitis, bloodstream infections, mortality, inappropriate empirical treatment
Introduction Biliary tract infection (BTI), including cholecystitis and cholangitis, remains a significant cause of morbidity, especially in elderly patients.1 – 5 More than half of all people older than 70 years have gallstones. Although most cases are asymptomatic, infectious complications may develop in 20% of cases.6 Severe BTI complicated with bloodstream infection (BSI) represents 8%–20% of the community-acquired bacteraemias in older patients2,3 and it is the second most common cause of sepsis in elderly patients after urinary tract infection.7 Initial treatment of BTI includes antibiotics and hydration. Early biliary decompression is suggested when the patient has not improved or if they worsen despite medical treatment. Three decompressive procedures (endoscopic, percutaneous and surgical) are widely available in most institutions. The preference for
percutaneous or endoscopic stent is due to the minimal invasiveness of this procedure.4 Despite the advancement of treatment, bacteraemic BTI still carries a mortality of 10% – 20%.8 – 11 Most epidemiological and outcome analysis on bacteraemic BTI has been described in earlier reports, but recent data on bacteraemic BTI are scarce. The aim of this study was to determine the epidemiology and susceptibility pattern of microorganisms isolated in bacteraemic BTI during 1991–2010 in our setting and to identify the independent prognostic factors of mortality in these cases.
Patients and methods Setting and data collection The setting was the Hospital Clı´nic in Barcelona, Spain, a 700 bed university tertiary centre that provides specialized and broad medical, surgical
# The Author 2012. Published by Oxford University Press on behalf of the British Society for Antimicrobial Chemotherapy. All rights reserved. For Permissions, please e-mail: [email protected]
Bacteraemia due to biliary tract infection
and intensive care for an urban population of 500000 people. This hospital has followed a blood culture surveillance programme since 1991. Briefly, an infectious disease specialist and one microbiologist review the charts of patients with positive blood cultures and recommend antibiotic therapy according to the clinical context and the results of the Gram’s stain, organism identification and antimicrobial susceptibility test. Patients were observed from the diagnosis of bacteraemia until 30 days afterwards, until death in hospital or until discharge. Data regarding the episode of bacteraemia are thus collected and entered in a database designed specifically for the blood culture surveillance programme.
Study design and inclusion criteria The type of study was based on an analysis of cases of bacteraemia due to BTI prospectively collected through the previously described blood culture surveillance programme from January 1991 to December 2010. We included the first episode of patients with bacteraemia due to BTI. The Ethics Committee of the hospital approved the study.
Diagnosis of cholecystitis or cholangitis The diagnostic criteria for cholangitis were: (i) fever or abdominal pain in the right upper quadrant; (ii) endoscopic or radiological (sonography or CT) evidence of biliary tract obstruction owing to stones, stricture or tumour; and (iii) laboratory evidence of hyperbilirubinaemia and elevated alkaline phosphatase. When cholangitis occurred after endoscopic retrograde cholangiopancreatography (ERCP) was classified separately. Cholecystitis was suggested clinically by fever or leucocytosis and right upper quadrant tenderness. The diagnosis was confirmed by ultrasonography or histological confirmation of acute inflammation of the gallbladder.
Microbiological methods Between 1991 and 1997, blood samples were processed by the BACTEC NR-730 system (Becton-Dickinson Microbiology Systems) and maintained routinely for 7 days. Since 1998 we have used the BACTEC 9240 system (Becton-Dickinson Microbiology Systems), with an incubation period of 5 days. Isolates were identified by standard techniques.12 Antimicrobial susceptibility testing was performed by a microdilution system (Microscan, Dade Behring, West Sacramento, CA, USA, or Sensititre, Trek Diagnostic Systems, East Grinstead, West Sussex, UK, or Phoenix System, Becton Dickinson, Franklin Lakes, NJ, USA). MICs observed with the three systems were interpreted using CLSI guidelines.13 Microorganisms reported as intermediate were considered resistant.
Patient characteristics The following data were obtained for all patients: age, sex, pre-existing comorbidities, prognosis of the underlying disease, prior antibiotic therapy, prior surgery (within the last month), current administration of ≥20 mg of corticosteroids every day, current administration of antineoplastic chemotherapy, source of bacteraemia, leucocyte count, origin of the infection (community-acquired or healthcare-associated, including last conventional hospitalization or outpatient visit), length of hospitalization before diagnosis of bacteraemia, intensive care unit (ICU) admission, need for mechanical ventilation, empirical and definitive antibiotic treatment, susceptibility to antibiotics of the isolates, presence of shock, and mortality.
symptoms of sepsis (as described previously).14 The origins of bloodstream infections were considered as: (i) nosocomial when cultures of blood specimens obtained .48 h after admission were positive; (ii) community-acquired when cultures were obtained prior to admission or during the first 48 h without hospitalization or healthcare contact during the month before bacteraemia; and (iii) healthcare-related when there was hospitalization or healthcare contact during the month before bacteraemia (i.e. dialysis patients). For analysis purposes we divided the origin of bacteraemia into two categories: communityacquired and healthcare-associated (this included nosocomial and healthcare-related).15 The source of infection was determined by a senior infectious disease specialist who considered the patient’s medical history, physical examination and the results of other microbiological tests and complementary imaging exploration. Comorbidity was defined as a disease or therapy that could predispose patients to infection, alter defence mechanisms or cause functional impairment, such as the following: diabetes, liver cirrhosis, renal failure, alcoholism (.100 g of alcohol every day), active neoplastic disease, severe chronic obstructive pulmonary disease, severe cardiac disease with symptomatic heart failure, severe dementia, and administration of immunosuppressive drugs (≥20 mg of corticosteroids every day on a regular basis or anti-neoplastic chemotherapy). The prognosis of the underlying disease was classified, according to the
Table 1. Demographic and clinical data of patients with bacteraemia due to BTI included in the study; N¼1373 Age (years), mean+SD
Male gender, n (%)
Comorbidity, n (%) biliary lithiasis solid-organ cancer diabetes mellitus chronic heart disease solid organ transplantation liver cirrhosis chronic lung disease chronic renal insufficiency haematological cancer HIV infection
484 (35) 362 (26) 239 (17) 161 (12) 131 (10) 98 (7) 66 (5) 49 (4) 25 (2) 11 (1)
Prognosis of underlying disease, n (%) non-fatal ultimately fatal rapidly fatal
917 (67) 381 (28) 75 (5)
Origin of bacteraemia, n (%) community-acquired healthcare-associated Corticosteroid use, n (%)
920 (67) 453 (33) 149 (11)
Bacteraemia source, n (%) cholangitis cholecystitis cholangitis post-ERCP
1124 (82) 207 (15) 42 (3)
Definition of terms
Significant bacteraemia was defined as one or more blood cultures positive for one or more primary pathogens and clinically apparent signs and
Ortega et al.
criteria of McCabe and Jackson, as rapidly fatal (when death was expected within ≤3 months), ultimately fatal (when death was expected within a period of .3 months but ,5 years) and non-fatal (when life expectancy was ≥5 years).16 Prior antibiotic therapy was defined as the use of any antimicrobial agent for ≥3 days during the month prior to the occurrence of the bacteraemic episode. Antibiotic treatment, either empirical or definitive (before or after the microbiological results and susceptibilities were known, respectively), was considered appropriate if at least one of the antibiotics involved had in vitro activity against the bacteria and the dose and route of administration were adequate. Shock was defined as a systolic pressure of ,90 mmHg that was unresponsive to fluid treatment or required vasoactive drug therapy.14 Death was considered related to the bloodstream infection if it occurred before the resolution of symptoms or signs or within 7 days of the onset of bacteraemia, and if there was no other explanation; otherwise, death within 30 days of the beginning of bacteraemia was considered unrelated to the episode.
Follow-up Patients were observed from the diagnosis of bacteraemia until 30 days afterwards.
Statistical analysis Statistical analyses were carried out using the SPSS software (version 14.0; SPSS, Chicago, IL, USA). Continuous variables are expressed as mean+SD or median (range) according to their homogeneity. Categorical variables were compared using the x 2 test or Fisher’s exact test (when necessary). The quantitative variables were compared using the Student’s– Fisher t-test or analysis of variance (ANOVA). Non-parametric tests were used when the application conditions were not applicable. Statistical significance was defined as a two-tailed P value ,0.05. Variables with P≤0.2 in the univariate analysis were further analysed by using a stepwise non-conditional (logistic
regression) multivariate analysis to find the independent factors associated with mortality. For analysis purposes, we considered related and unrelated mortality (within 30 days of bloodstream infection) together.
Results Demographic and clinical data BTI was the confirmed source of 1373 patients with bacteraemia identified during the period of study, 1991 –2010. Demographic and clinical data are presented in Table 1. The average age was 71 years (SD 14), with 60% .65 years old and 18% .80 years old. Biliary lithiasis and solid-organ cancer were the most frequent underlying comorbidities. There were 456 patients (33%) with ultimately or rapidly fatal prognosis of underlying disease. Four-hundred-and-fifty-three (33%) cases of bacteraemia due to BTI were healthcare-associated. There were 209 (15%) cases with shock on presentation, and mortality accounted for 126 cases (9%).
Microbiological data Analysis of positive blood cultures showed a predominance of Gram-negative infection. Bacteraemia episodes were polymicrobial in 207 cases (15%). Microbiological results are summarized in Table 2. According to our data, the most common microorganisms isolated were Escherichia coli and Klebsiella spp., and their frequency remained stable throughout the years of study. Enterococcus spp. and Pseudomonas aeruginosa have been isolated more frequently in cases of healthcare-associated bacteraemia. There were 218 (22%) fluoroquinolone-resistant E. coli and Klebsiella spp. strains and 47 (3%) cefotaxime-resistant (CTX-R) strains. Of these, 40 cases were isolated since 2001.
Table 2. Comparison of community-acquired or healthcare-associated microorganisms most frequently isolated in bacteraemia due to BTI
Total bacteraemia episodes (N¼1373), n (%)
Community-acquired (N¼920), n (%)
Healthcare-associated (N ¼453), n (%)
E. coli CTX-R E. coli
749 (55) 38 (5)a
568 (62) 8 (1)a
181 (40) 30 (17)a
Klebsiella spp. K. pneumoniae K. oxytoca CTX-R Klebsiella spp.
240 (17) 180 (13) 60 (4) 9 (4)a
172 (19) 119 (13) 53 (6) 0
68 (15) 61 (13) 7 (2) 9 (13)a
0.06 NS NS 0.001
Enterococcus spp. E. faecalis E. faecium other Enterococcus spp.
163 (12) 78 (6) 54 (4) 31 (2)
91 41 36 14
(20) (9) (8) (3)
0.001 0.08 0.06 NS
72 37 18 17
(8) (4) (2) (2)
NS, not significant. Other microorganisms that represent ,5% of isolates are Citrobacter spp., Bacteroides fragilis, Proteus mirabilis, Staphylococcus aureus, Serratia spp. and other Pseudomonas spp. a Percentage of specific microorganism.
Bacteraemia due to biliary tract infection
Table 3. Microorganisms most frequently isolated according to BTI type
Microorganisms E. coli Klebsiella spp. Enterococcus spp. P. aeruginosa Enterobacter spp.
Total (N¼1373), n (%)
Cholangitis (N ¼1124), n (%)
Cholecystitis (N¼207), n (%)
Cholangitis post-ERCP (N¼42), n (%)
749 (55) 240 (17) 163 (12) 86 (6) 63 (5)
618 (55) 204 (18) 141 (13) 78 (7) 54 (5)
114 (55) 29 (14) 15 (7) 4 (2) 8 (4)
17 (40) 7 (17) 7 (17) 4 (10) 1 (2)
Table 4. Mortality analysis in bacteraemia due to BTI (N¼1360) Alive (N¼1234; 91%)
Dead (N¼126; 9%)
Multivariate OR (95% CI)
1.2 (1.1– 1.4)
Comorbidity, n (%) solid-organ cancer solid organ transplantation liver cirrhosis chronic renal insufficiency haematological cancer
291 (82) 126 (96) 90 (93) 42 (86) 22 (88)
63 (18) 5 (4) 7 (7) 7 (14) 3 (12)
0.001 0.01 0.2 0.1 0.2
2.4 (0.9– 4.2) 0.8 (0.6– 1.2) — 1.9 (0.7– 2.5) —
Prognosis of underlying disease, n (%) non-fatal ultimately fatal rapidly fatal
857 (94) 337 (89) 40 (54)
50 (6) 42 (11) 34 (46)
0.2 0.1 ,0.001
1 2.5 (1.5– 4.2) 28.9 (14.8–56.7)
Origin of bacteraemia, n (%) community-acquired healthcare-associated
849 (93) 385 (86)
61 (7) 65 (14)
1 2.3 (0.9– 3.4)
Polymicrobial bacteraemia, n (%)
2.1 (0.8– 3.1)
Microorganism, n (%) E. coli Klebsiella spp. CTX-R Enterobacteriaceae Enterococcus spp. P. aeruginosa Enterobacter spp.
670 (92) 215 (92) 40 (89) 141 (87) 68 (79) 54 (86)
60 (8) 17 (8) 5 (11) 22 (13) 18 (21) 9 (14)
0.2 0.2 0.1 0.07 0.001 0.1
— — 2.3 (0.8– 3.5) 2.7 (0.9– 5.8) 2.3 (1.2– 4.6) 1.7 (0.6– 3.0)
NS NS 0.02 NS
Shock, n (%)
11.5 (7.1– 18.5)
Empirical therapy, n (%) appropriate inappropriate not available
842 (92) 111 (85) 281 (89)
70 (8) 20 (15) 36 (11)
0.2 0.01 0.3
— 1.4 (1.1– 1.7) —
Age (years), mean+SD
P 0.001 NS NS NS
NS, not significant.
During 2006–10, CTX-R Enterobacteriaceae represented 9% of the bacteraemia episodes.
differences between isolates in cholangitis or cholecystitis cases; however, Enterococcus spp. and P. aeruginosa were more common in cases of cholangitis post-ERCP.
Characteristics related to BTI Of 1373 bacteraemic BTI, a total of 1124 (82%) were cholangitis, 207 (15%) cholecystitis and 42 (3%) cholangitis post-ERCP. The different microorganisms isolated in each type of BTI are shown in Table 3. There were no statistically significant
Mortality Univariate and multivariate analysis of mortality in bacteraemia due to BTI are shown in Table 4. Independent risk factors associated with mortality were age, ultimately or rapidly fatal
Ortega et al.
Table 5. Factors associated with P. aeruginosa or CTX-R Enterobacteriaceae in bacteraemia due to BTI Characteristic Solid-organ transplantation Solid-organ cancer Healthcare-associated bacteraemia Previous antibiotic therapy
1.58 1.60 1.66 1.92
1.03 –2.81 1.11 –2.29 1.28 –2.15 1.34 –2.75
0.04 0.01 0.001 0.001
prognosis of underlying disease, shock on presentation, bacteraemia due to P. aeruginosa and an inappropriate empirical antibiotic treatment. The rate of inappropriate empirical treatment was 10%. However, this percentage was higher in the case of bacteraemia produced by Enterococcus spp. (22%), P. aeruginosa (23%) and CTX-R E. coli or Klebsiella spp. (33%). Factors associated with P. aeruginosa or CTX-R Enterobacteriaceae in bacteraemia due to BTI were previous antibiotic therapy, healthcare-associated bacteraemia and solid-organ cancer or transplantation (Table 5).
Discussion BTI remains a significant cause of morbidity in all age groups. However, few studies have investigated bacteraemic BTI. Consistent with previous observations, the predominance of Gramnegative pathogens and high percentage of polymicrobial infections are typical blood culture findings.10,11,17 The most common microorganisms isolated were E. coli and Klebsiella spp., and their frequency remained stable throughout the years of study. However, Enterococcus spp. and P. aeruginosa have been isolated more frequently in cases of healthcare-associated or post-ERCP cholangitis. Similar to previously reported data, we observed a higher proportion of CTX-R isolates in the last years of the study.18,19 During the 2006–10 period, CTX-R E. coli and Klebsiella spp. strains represented 9% of bacteraemia due to BTI. The 30 day crude mortality rate in our series was 9%, roughly comparable to that in previous series.17 According to our results, age, ultimately or rapidly fatal prognosis of underlying disease, shock on presentation, an isolate of P. aeruginosa and an inappropriate empirical antibiotic treatment were independently associated with mortality. Septic shock was a strong welldocumented predictor of mortality among bacteraemic patients, reflecting haemodynamic compromise.20,21 Multiple comorbidity may be a marker of poor physical health rather than a risk factor for mortality per se. The rate of incorrect empirical treatment could be improved by knowing the cases where empirical therapy is inappropriate. The rate of inappropriate empirical treatment was higher in cases of infection caused by P. aeruginosa, Enterococcus spp. or CTX-R Enterobacteriaceae. The factors that identified patients at increased risk of infection with these microorganisms were previous antibiotic therapy, healthcareassociated bacteraemia or solid-organ cancer or transplantation. The results of this study should be interpreted in the light of some limitations. First, this is a single-centre study with inherent selection bias. Second, we have reported the results of a retrospective analysis of a database prospectively collected for .18 years. Third, although ERCP was the choice in most cases, information on the drainage procedure done in every patient
was not available. Finally, we considered related and unrelated mortality together. This is an analysis option used almost always in bloodstream infection outcome studies, but the factors associated with mortality in the first week after the bacteraemia diagnosis could be different from those associated with a fatal outcome during the following 30 days. In conclusion, biliary sepsis continues to be a life-threatening condition with a substantial mortality rate. E. coli and Klebsiella spp. are the most frequently isolated microorganisms. However, in cases of previous antibiotic therapy, healthcareassociated bacteraemia or solid-organ cancer or transplantation may indicate the need to initiate empirical therapy activity against P. aeruginosa or CTX-R Enterobacteriaceae. In patients with such risk factors, initial therapy with piperacillin/tazobactam or a carbapenem would be advisable. In geographical areas with a prevalence of CTX-R Enterobacteriaceae similar to ours and in cases of BTI with severe sepsis criteria it would be advisable to start empirical treatment with a carbapenem.
Acknowledgements This study was partially presented at the Twenty-second European Congress of Clinical Microbiology and Infectious Diseases, London, 2012. (Abstract no. P-1531).
Funding This study was carried out as part of our routine work.
Transparency declarations None to declare.
Author contributions M. O.: study concept and design, analysis and interpretation of data, statistical analysis, and drafting of the manuscript. F. M. and M. A.: critical revision of the manuscript for important intellectual content. J. A. M., A. S. and J. M.: acquisition of data and critical revision of the manuscript for important intellectual content. J. L. and C. P.: acquisition of data.
References 1 Hanau LH, Steigbigel NH. Acute (ascending) cholangitis. Infect Dis Clin North Am 2000; 14: 521–46. 2 Stewart L, Griffiss JM, Way LW. Spectrum of gallstone disease in the veterans population. Am J Surg 2005; 190: 746–51. 3 Siegman-Igra Y, Schwartz D, Konforti N et al. Septicemia from biliary tract infection. Arch Surg 1988; 123: 366– 8. 4 Sugiyama M, Atomi Y. Treatment of acute cholangitis due to choledocholithiasis in elderly and younger patients. Arch Surg 1997; 132: 1129– 33. 5 Podnos YD, Jimenez JC, Wilson SE. Intra-abdominal sepsis in elderly persons. Clin Infect Dis 2002; 35: 62– 8. 6 Margiotta SJ, Willis IH, Wallack MK. Cholecystectomy in the elderly. Am Surg 1988; 54: 34– 9. 7 Esposito AL, Gleckman RA, Cram S et al. Community-acquired bacteraemia in the elderly: analysis of one hundred consecutive episodes. J Am Geriatr Soc 1980; 28: 315– 9.
Bacteraemia due to biliary tract infection
8 Thompson JE, Tompkins RK, Longmire WP. Factors in management of acute cholangitis. Ann Surg 1982; 195: 137–45.
16 McCabe WR, Jackson GG. Gram-negative bacteraemia. II. Clinical, laboratory and therapeutic observations. Arch Intern Med 1962; 110: 847–55.
9 McSherry CK, Glenn F. The incidence and causes of death following surgery for non-malignant biliary tract disease. Ann Surg 1980; 191: 271–5.
17 Lee CC, Chang IJ, Lai YC et al. Epidemiology and prognostic determinants of patients with bacteremic cholecystitis or cholangitis. Am J Gastroenterol 2007; 102: 563–9.
10 Kuo CH, Changchien CS, Chen JJ et al. Septic acute cholecystitis. Scand J Gastroenterol 1995; 30: 272–5.
18 Ortega M, Marco F, Soriano A et al. Analysis of 4758 Escherichia coli bacteraemia episodes: predictive factors for isolation of an antibiotic-resistant strain and their impact on the outcome. J Antimicrob Chemother 2009; 63: 568– 74.
11 Shimada K, Noro T, Inamatsu T et al. Bacteriology of acute obstructive suppurative cholangitis of the aged. J Clin Microbiol 1981; 14: 522–6. 12 Murray PR, Baron EJ, Jorgensen JH (eds). Manual of Clinical Microbiology, 9th edn. Washington, DC: ASM Press, 2007. 13 Clinical and Laboratory Standards Institute. Performance Standards for Antimicrobial Susceptibility Testing: Sixteenth Informational Supplement M100-S16. CLSI, Wayne, PA, USA, 2006. 14 Annane D, Bellisant E, Cavaillon JM. Septic shock. Lancet 2005; 365: 63– 78. 15 Garner JS, Jarvis WR, Emori TG et al. CDC definitions for nosocomial infections, 1988. Am J Infect Control 1988; 16: 128–40.
19 Ortega M, Marco F, Soriano A et al. Cefotaxime resistance and outcome of Klebsiella spp bloodstream infection. Eur J Clin Microbiol Infect Dis 2011; 30: 1599 –605. 20 Shmuely H, Pitlik S, Drucker M et al. Prediction of mortality in patients with bacteraemia: the importance of preexisting renal insufficiency. Ren Fail 2000; 22: 99–108. 21 Hugonnet S, Harbarth S, Ferriere K et al. Bacteremic sepsis in intensive care: temporal trends in incidence, organ dysfunction, and prognosis. Crit Care Med 2003; 31: 390–4.