- Email: [email protected]
Comparison of Escherichia coli and Klebsiella pneumoniae Liver Abscesses
Comparison of Escherichia coli and Klebsiella pneumoniae Liver Abscesses SHIUAN-CHIH CHEN, MD, PHD; WEI-YA WU, MD, MSC; CHUNG-HSIN YEH, MD, MSC; KUANG-CHI LAI, MD, PHD; KEN-SHENG CHENG, MD; LONG-BIN JENG, MD; PO-HUI WANG, MD, PHD; DING-BANG LIN, PHD; CHUN-CHIEH CHEN, MD, MSC; MENG-CHIH LEE, MD, PHD, MPH; WILLIAM R. BELL, PHD, MD
ABSTRACT: Background: Escherichia coli and Klebsiella pneumoniae are the most common causative pathogens of pyogenic liver abscesses. The objective of this study was to compare outcome between patients with liver abscesses due to E coli and those with liver abscesses caused by K pneumoniae; we also aimed to identify separately the predictors of mortality in the 2 groups. Methods: We conducted a retrospective study of 202 patients who presented with pyogenic liver abscesses caused by either E coli or K pneumoniae from July 2000 to June 2005. Outcome of the patients was analyzed by exact logistic regression with adjustment for baseline and clinical covariates. Significant predictors of mortality in the E coli and the K pneumoniae groups were investigated by multivariate analysis of demographic and clinical variables in each group. Results: Of the 202 patients (128 men and 74 women; age range, 19 to 89 years), pyogenic liver abscess was due to E coli infection in 55 patients and K pneumoniae in 147 patients. In contrast to patients with K pneumoniae, patients with E coli liver abscess were more likely to be older and female, have a biliary abnormality or malignancy, pleural effusion, polymicrobial infection with anaerobic or multi– drug-resistant organisms, a higher APACHE II score, and to have been treated initially with ineffective antibiotics; they were also less likely to have diabetes mellitus. The cause of K pneumoniae liver abscess was often cryptogenic. The sensitivity, specificity, positive predictive value, and likelihood ratio of the presence of biliary disorders and coexisting malignancy as a predictive
parameter of E coli liver abscess were 25%, 96%, 67%, and 5.45/1, respectively. The sensitivity, specificity, positive predictive value, and likelihood ratio of the presence of diabetes mellitus with an abscess of cryptogenic origin as a predictive parameter of K pneumoniae liver abscess were 39%, 84%, 81%, and 2.36/1, respectively. There was no significant difference in mortality between patients with E coli and those with K pneumoniae infections (26% vs 4%; adjusted OR, 4.2; 95% CI, 0.63 to 27; P ⫽ 0.105). However, for patients with liver abscess caused by E coli, the APACHE II score at admission (OR, 1.7; 95% CI, 1.1 to 2.6; P ⫽ 0.021), malignancy (OR, 26; 95% CI, 1.8 to 370; P ⫽ 0.016), and right-lobe abscess (OR, 0.0029; 95% CI, 0.00010 to 0.15; P ⫽ 0.004) were significant predictors of death, whereas uremia (OR, 52; 95% CI, 3.5 to 750; P ⫽ 0.004) and multi– drug-resistant isolates (OR, 26; 95% CI, 2.3 to 290; P ⫽ 0.009) were significant predictors of death in the K pneumoniae group. Conclusions: A higher APACHE II score at admission and a higher frequency of coexisting malignancy may have contributed to the higher, although not significant, mortality rate in patients with liver abscess caused by E coli infection. Clinicians should begin with broad antibiotic coverage such as a second-generation cephalosporin and an aminoglycoside with metronidazole when treating liver abscesses with E coli as the likely pathogen due to the high frequency of multi–drug-resistant isolates among E coli isolates. KEY INDEXING TERMS: Escherichia coli; Klebsiella pneumoniae; Pyogenic liver abscess; Outcome. [Am J Med Sci 2007;334(2):97–105.]
P
Escherichia coli was the predominant organism isolated from pyogenic liver abscess worldwide until the 1980s,1,3,5 when Klebsiella pneumoniae began to be
yogenic liver abscess continues to be a potentially life-threatening disease despite the advances in microbiology methods, imaging techniques, and therapeutic modalities.1– 8 In terms of causative pathogens,
Institute of Medicine (SCC, WYW, KCL, MCL) and Schools of Medicine (SCC, PHW, CCC) and Medical Laboratory and Biotechnology (DBL), Chung Shan Medical University; Department of Family Medicine (SCC, WYM, CCC, MCL), Chung Shan Medical University Hospital, Taichung, Taiwan; Department of Neurology (CHY), Show Chuan Memorial Hospital, Changhua, Taiwan; Department of Life Sciences (CHY), National Chung Hsing University, Taichung, Taiwan; Departments of Surgery (KCL, LBJ) and Internal Medicine (KSC), China Medical University Hospital, Taichung, Taiwan; and Department of Medicine, Division of Hematology (WRB), THE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
The Johns Hopkins University School of Medicine, Baltimore, Maryland. Submitted July 27, 2006; accepted in revised form January 17, 2007. None of the authors have any financial involvement with any organization. Correspondence: Dr. Chun-Chieh Chen, Faculty of Medicine, Chung Shan Medical University School of Medicine, Department of Family Medicine, Chung Shan Medical University Hospital, No. 110, Section 1, Chien Kuo N. Road, Taichung 402, Taiwan (E-mail: [email protected]).
97
E coli and K pneumoniae Liver Abscesses
reported as the leading causative pathogen of pyogenic liver abscess in some Asian countries.7,9 –13 There is evidence that a shift from E coli to K pneumoniae liver abscesses is occurring gradually in some Western countries.4,14 –18 E coli and K pneumoniae liver abscesses have been described individually in several studies1,7,9 –13; however, there is still a lack of a comprehensive comparison between the 2 leading causative pathogens involved in the development of pyogenic liver abscesses. Hence, we conducted a retrospective study to delineate the clinical features of patients with E coli liver abscesses and those with K pneumoniae liver abscesses to compare the outcomes between these 2 types of patients and to identify the predictors of mortality in the 2 groups. Methods Study Patients The medical records of consecutive patients aged 18 years and older who had been admitted to the China Medical University Hospital, a 1600-bed primary and tertiary care medical center in Taichung, Taiwan, from July 2000 to June 2005, and discharged with a diagnosis of pyogenic liver abscess were reviewed by a systematic search of the patients’ records for diagnostic code 572.0 (International Classification of Diseases, 9th Revision). Patients were included in the study if they met both of the following criteria: 1) one or more discrete abscess cavities of the liver had been identified by imaging studies [endoscopic retrograde cholangiopancreatography (ERCP), abdominal ultrasonography (US) and/or computerized axial tomography (CT) scans with contrast enhancement], and 2) either K pneumoniae or E coli had been isolated from abscess culture. Of the 252 patients with pyogenic liver abscess that were screened, 202 patients fulfilled the inclusion criteria; they were divided into 2 groups depending on whether they had E coli liver abscess (E coli group) or K pneumoniae liver abscess (K pneumoniae group). A total of 50 patients, of the 252 patients with pyogenic liver abscess that were screened, were excluded for the following reasons: amebic liver abscess (n ⫽ 2), liver abscess due to ascarid infection (n ⫽ 1), fungal liver abscess (n ⫽ 2), tuberculous liver abscess (n ⫽ 1), incomplete medical records (n ⫽ 10), necrotic tumors (n ⫽ 1), biloma (n ⫽ 1), liver abscess due to concomitant E coli and K pneumoniae infections (n ⫽ 11), and liver abscess due to the causative organisms isolated only from blood culture, but not the liver abscess (n ⫽ 21). This study was approved by the Institutional Review Board of the China Medical University Hospital. Data Collection Demographic and medical information were collected. The medical information included clinical, imaging, microbiological, and laboratory parameters. The clinical parameters comprised underlying medical conditions, presenting symptoms/signs, type of treatment, and outcome. Abscess material had been obtained by image-guided (CT or US) percutaneous needle aspiration (PNA), image-guided percutaneous catheter drainage (PCD), or direct surgical intervention. The specimens obtained had been processed for Gram stain, bacterial cultures (standard aerobic and anaerobic diagnostic methods), and tests for antimicrobial susceptibility. Antimicrobial susceptibility had been determined by the disk diffusion method (BD BBL, Sensi-Disc Antimicrobial Susceptibility Test Discs, Sparks, MD), based on the microorganism isolated. The results were evaluated according to the recommendations set forth by the Clinical and Laboratory Standards Institute (formerly known as the National Committee for Clinical Laboratory Standards).19,20 Multi– drug resistance was defined as resistance to 3 or more of the antimicrobial classes. After the cultures of blood and/or liver
98
abscess aspirate had been obtained, initial empirical broad-spectrum antibiotic treatment was administered intravenously, and adjusted as necessary when the results of the culture and susceptibility test were known. When isolates from patients were resistant to all of the initial antibiotics that had been given, the initial treatment was regarded to have been ineffective. Response to treatment was evaluated in each patient by intensive follow-up US or CT scan of the liver during hospitalization and/or subsequent out-patient visits after discharge. The origin of the abscess was based on the available imaging, clinical, pathological and/or surgical findings. If no obvious extra-hepatic source of infection could be identified after appropriate investigation, the abscess was considered cryptogenic in origin. The severity of illness at admission was evaluated by the first-day Acute Physiology and Chronic Health Evaluation II (APACHE II) score.21 Patient outcome assessments included the recurrence of liver abscess, hospital stay, and death. Recurrence of liver abscess was defined as the development of new clinical and radiological changes after discharge. The median follow-up of patients after discharge was 3 months (range 1 to 9 months). Death was defined as death within 30 days of treatment or before discharge from the hospital. The cause of death was ascertained from the death certificate. Statistical Analysis All statistical analyses were performed using a statistical software package (SAS version 8.2, SAS Institute Inc, Cary, NC). For analysis, patients with pyogenic liver abscess were divided into two groups according to bacterial etiology: the E coli group and the K pneumoniae group. Descriptive data are presented as medians with interquartile ranges (IQRs) for continuous data and percentages for categorical data. Continuous variables were compared by the Mann-Whitney U test. Categorical variables were compared by either the 2 test or Fisher exact test (if more than 20% of the expected frequencies were ⬍5), as appropriate. Because of the imbalanced data set in our study, data were analyzed by exact logistic regression rather than by conventional logistic regression.22 We constructed an exact logistic regression model with covariate adjustment to estimate odds ratios (ORs) and 95% confidence intervals (CIs) to analyze associations between dependent and independent variables. The baseline variables, including sex, age, duration from the onset of symptoms to the time of diagnosis, and severity of illness at admission, were tested between the 2 bacterial groups. The significant variables above were regarded as baseline covariates. Associations between the bacterial group (E coli vs K pneumoniae) and clinical variables at admission were analyzed after adjusting the same set of baseline covariates. Outcome was analyzed between the 2 groups of patients by exact logistic regression with adjustment for the baseline covariates and the significant clinical variables. Data were then analyzed by bivariate and multivariate analyses to identify significant independent predictive factors of death associated with E coli liver abscess and those associated with K pneumoniae abscess. The significant independent variables obtained by bivariate analyses were entered into the multiple stepwise logistic regression model, which determined the prognostic factors associated with death. All P values were two-tailed, and P ⬍ 0.05 was considered statistically significant in all analyses. In addition, P ⬍ 0.001 was particularly shown among multiple tests in tables.
Results Demographic Data, Underlying Diseases, and Symptoms/Signs Of the 202 patients with pyogenic liver abscess, E coli had been diagnosed in 55 patients (E coli group) and K pneumoniae in 147 patients (K pneumoniae group). The median age of patients with E coli liver August 2007 Volume 334 Number 2
Chen et al
Table 1. Underlying Diseases and Presenting Symptoms/Signs at Admission in the Escherichia coliand K lebsiella pneumoniae Groups Variable
E coli Liver Abscess (n ⫽ 55)
K pneumoniae Liver Abscess (n ⫽ 147)
E coli vs K pneumoniae Adjusted ORf (95% CI)
19 (35) 42 (76) 18 (33) 4 (7) 8 (15) 3 (6) 17 (31) 4 (7)
82 (56) 47 (32) 15 (10) 22 (15) 14 (10) 8 (5) 112 (76) 4 (3)
0.31 (0.15–0.64)* 6.4 (3.0–14)† 2.9 (1.2–7.1)* 0.97 (0.28–3.3) 1.7 (0.63–4.7) 0.32 (0.061–1.7) 0.14 (0.068–0.29)† 2.5 (0.41–16)
49 (91) 38 (69) 26 (47) 23 (42) 20 (36) 20 (36) 19 (35) 11 (20) 8 (15) 5 (9) 2 (4)
140 (95) 77 (52) 61 (42) 58 (40) 50 (34) 42 (29) 45 (31) 17 (12) 17 (12) 8 (5) 11 (8)
0.44 (0.12–1.6) 2.9 (1.4–6.1)* 1.6 (0.83–3.3) 1.3 (0.65–2.6) 1.0 (0.52–2.1) 2.2 (1.1–4.6)* 0.96 (0.47–2.0) 2.5 (1.1–6.5)* 1.1 (0.36–3.2) 2.6 (0.74–8.9) 0.18 (0.031–1.1)
Underlying condition,a No. (%) Diabetes mellitus Biliary disorderb Malignancy Alcoholism Liver cirrhosis Uremia Cryptogenic origin of abscess Other origin of abscessc Symptoms/signs,a No. (%) Fever/chills RUQ pain/tenderness Malaise Anorexia Nausea/emesis Jaundice Chest symptomsd Murphy signe Ascites Weight loss Mental confusion
CI indicates confidence interval; OR, odds ratio; RUQ, right upper quadrant. When patients fit into more than one category, they were counted in each category. b Biliary disorder: Cholelithiasis, choledocholithiasis, hepatolithiasis, cholecystitis, or purulent cholangitis with/without malignancy. c Others: Pancreatitis, empyema of the gall bladder, subphrenic abscess ulcerative colitis, colonic diverticulitis, liver injury, prior transcatheter arterial embolization for hepatocellular carcinoma, or recent biliary surgery. d Chest symptoms: Cough, chest pain, and/or short of breath. e Murphy sign: Deep inspiration or cough during subcostal palpation of the RUQ producing increased pain and inspiratory arrest. f The exact logistic regression model included adjustment for age, sex, and severity of illness at admission. * P ⬍ 0.05. † P ⬍ 0.001. a
abscess was significantly greater than that of patients with K pneumoniae liver abscess (67 years [IQR, 61 to 74 years] vs 54 years [IQR, 45 to 68 years]; P ⬍ 0.001; using the Mann-Whitney U test). There was a significant difference in sex between the E coli and K pneumoniae groups (male/female ⫽ 27/28 vs 101/36; P ⫽ 0.01; using the 2 test). The median APACHE II score at admission among patients in the E coli group was significantly higher than that among patients in the K pneumoniae group (12 points [IQR, 9 to 15 points] vs 8 points [IQR, 5 to 12 points]; P ⬍ 0.001; using the MannWhitney U test). The median duration from the onset of symptoms to the time of diagnosis was 5 days (IQR, 3 to 10 days) in the E coli group and 6 days (IQR, 3–9 days) in the K pneumoniae group; however, the difference between groups was not significant (P ⫽ 0.468; using the Mann-Whitney U test). Age, sex, and APACHE II score at admission were considered the baseline covariates. After adjusting for the baseline covariates, patients in the E coli group were more likely to have biliary disorders (cholelithiasis, choledocholithiasis, hepatolithiasis, cholecystitis, and/or purulent cholangitis) and underlying malignancy than those in the K pneumoniae group. Patients in the K pneumoniae group THE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
were more likely to have diabetes mellitus and abscesses that were cryptogenic in origin than patients in the E coli group. The underlying diseases and presenting symptoms/signs at admission of these patients are summarized in Table 1. Among the 242 patients with complete medical records, an initial pending diagnosis of liver abscess and available diagnostic radiological studies, the significant variables regarding underlying diseases and origin of abscess were analyzed to see whether they could be used to predict the etiology of the abscess. The sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of the presence of biliary disorder as a predictive factor of E coli liver abscess were 80%, 68%, 48%, and 90%, respectively. The sensitivity, specificity, PPV and NPV of the presence of biliary disorder and coexisting malignancy as a predictive parameter of E coli liver abscess were 25%, 96%, 67%, and 76%, respectively. The likelihood ratio for the prediction of E coli as the etiology of liver abscesses in patients with biliary disorder and concomitant malignancy was 5.45/1. The sensitivity, specificity, PPV and NPV of the presence of diabetes mellitus as a predictive factor of K pneumoniae liver abscess were 57%, 61%, 73%, and 43%, respectively. The sensitivity, specificity, PPV and 99
E coli and K pneumoniae Liver Abscesses
NPV of abscesses of unknown origin as a predictive factor of K pneumoniae liver abscess were 75%, 65%, 80%, and 59%, respectively. The sensitivity, specificity, PPV and NPV of the presence of diabetes mellitus with an abscess of cryptogenic origin as a predictive parameter of K pneumoniae liver abscess were 39%, 84%, 81%, and 43%, respectively. The likelihood ratio for the prediction of K pneumoniae as the etiology of liver abscesses in patients with diabetes mellitus and liver abscesses of cryptogenic origin was 2.36/1. Microbiological, Imaging, and Laboratory Findings and Initial Therapeutic Modalities The species of organisms that had been isolated from patients in the E coli and K pneumoniae groups are summarized in Table 2. Polymicrobial infection, anaerobes, and multi– drug-resistant (MDR) isolates (from abscess or blood cultures) were found more often in the E coli group than in the K pneumoniae group (Table 3). Of the 74 E coli isolates from blood and abscess cultures, 19 (26%) were susceptible to all antimicrobials tested, 31 (42%) were resistant to 1 or 2 antimicrobial classes, and 24 (32%) were MDR strains. Among the E coli isolates, the majority were resistant to ampicillin (73%), followed by trimethoprim-sulfamethoxazole (45%), cephalothin (36%), piperacillin (26%), amoxicillin/
clavulanic acid (22%), gentamicin (18%), ciprofloxacin (14%), and cefuroxime (8%). Only one E coli isolate was resistant to ceftazidime. Most (218/228, 96%) of the K pneumoniae isolates from blood and abscess cultures were susceptible to all antimicrobial agents tested, except ampicillin. Of the remaining 10 K pneumoniae isolates, 1 was susceptible to all antimicrobials tested, 5 were resistant to 1 or 2 antimicrobial classes, and 4 were MDR isolates. Image findings revealed that patients in the E coli group had a higher frequency of pleural effusion than patients in the K pneumoniae group; however, no other differences in image findings were seen between the 2 groups (Table 3). Patients in the E coli group had a higher frequency of anemia compared with those in the K pneumoniae group; however, no other differences in frequency of abnormal laboratory values were noted between the groups (Table 3). Patients were initially managed with 1 of 4 therapeutic approaches: antibiotics alone, antibiotics plus PNA, antibiotics plus PCD or antibiotics plus surgical intervention. The most common primary treatment modality was percutaneous drainage (PNA or PCD) (E coli group, 89%; K pneumoniae group, 94%) All patients initially received intravenous empirical antibiotics (cephalosporins, penicillin group, aminoglycosides, and metronidazole).
Table 2. Microbiological Spectrum in E coli andK pneumonia Liver Abscesses E coli Liver Abscess (n ⫽ 55) Organisms Gram-negative aerobes Klebsiella pneumoniae Escherichia coli Klebsiella oxytoca Pseudomonas aeruginosa Morganella morganii Proteus spp Enterobacter spp Aeromonas spp Acinetobacter spp Citrobacter spp Burkholderia cepacia Edwardsiella tarda Gram-positive aerobes Enterococcus spp Streptococcus spp ORSA CoNS Gemella morbillorum Micrococcus spp Anaerobes Bacteroides spp Prevotella spp Fusobacterium spp Peptostreptococcus spp Bifidobacterium spp Total isolates
K pneumoniae Liver Abscess (n ⫽ 147)
Blood (n ⫽ 55)
Abscess (n ⫽ 55)
19 1 1
55 2 5
2
8 1 2 2 1
Blood (n ⫽ 143)
Abscess (n ⫽ 147)
81
147
1 1 1 1 3 1
1 3 1
6 10 1
4
1 1
3 2 1
1 1 1
6 2
39
14 2 3 5 1 120
2
2
89
1 1 163
CoNS indicates coagulase negative Staphylococci; ORSA, oxacillin-resistant Staphylococcus aureus.
100
August 2007 Volume 334 Number 2
Chen et al
Table 3. Microbiological, Imaging Findings, Laboratory Findings at Admission, and Initial Therapeutic Modalities in the E Coliand the K Pneumoniae Groups Variable Microbiologic characteristics, No. (%) Bacteremia Polymicrobialinfectiona Anaerobicinfectionb MDR isolates Imaging finding, No. (%) Pleural effusion Unilobar abscess Right-lobe abscess Solitary abscess Abscess ⬎5 cm in diameter Gas-forming abscess Rupture of abscess Laboratory tests, No. (%) WBC count (⬎104 cells/mm3) Serum albumin (⬍3.5 g/dL) Hb (⬍14 g/dL in male,⬍12 g/dL in female) Serum AST (⬎40 U/L) Serum ALP (⬎126 U/L) Total bilirubin (⬎1.3 mg/dL) PT (⬎13.1 seconds) Initial treatment method, No. (%) Antibiotics alone Invasive procedurec plus antibiotics Ineffective initial antibiotics
E coli Liver Abscess
K pneumoniae Liver Abscess
E coli vs K pneumoniae Adjusted ORf (95% CI)
35/55 (64) 38/55 (69) 21/55 (38) 23/55 (42)
87/143 (61) 11/147 (8) 7/147 (5) 6/147 (4)
0.92 (0.44–1.9) 22 (9.4–53)† 10 (3.8–26)† 15 (5.2–43)†
38 (69) 45 (82) 29 (53) 37 (67) 32 (58) 9 (16) 2 (4)
60 (41) 124 (84) 98 (67) 115 (78) 104 (71) 36 (25) 9 (6)
2.7 (1.3–5.4)* 0.93 (0.39–2.2) 0.55 (0.28–1.1) 0.60 (0.29–1.3) 0.50 (0.25–1.01) 0.43 (0.18–1.1) 0.71 (0.13–3.8)
44/55 (82) 40/42 (95) 47/55 (86) 33/52 (64) 31/46 (67) 32/47 (68) 29/46 (63)
118/147 (80) 76/77 (99) 88/147 (60) 90/131 (69) 64/112 (57) 67/107 (63) 96/142 (68)
0.78 (0.33–1.8) 0.16 (0.012–2.1) 5.2 (2.0–13)† 0.88 (0.42–1.9) 1.3 (0.61–2.9) 1.5 (0.65–3.4) 0.61 (0.28–1.3)
6/55 (11) 49d/55 (89) 12/55 (22)
6/147 (4) 141e/147 (96) 6/147 (4)
3.0 (0.82–11) 0.33 (0.088–1.2) 4.9 (1.6–14)*
ALP indicates alkaline phosphatase; AST, aspartate aminotransferase; CI, confidence interval; Hb, hemoglobin; MDR, multi–drugresistant; OR, odds ratio; PT, prothrombin time; WBC, white blood cell. a Polymicrobial infection: Mixed bacterial flora were cultured in blood and/or abscess cultures. b Anaerobic infection: Anaerobic isolateswere cultured in blood and/or abscess cultures. c Invasive procedure: percutaneous needle aspiration (PNA), percutaneous catheter drainage (PCD), or surgical intervention. d Of the 49 patients, 4 underwent PNA and 45 underwent PCD. e Of the 141 patients, 12 underwent PNA, 126 underwent PCD, and 3 were managed surgically. f The exact logistic regression model included adjustment for age, sex, and severity of illness at admission. * P ⬍ 0.05. † P ⬍ 0.001.
A total of 34 patients (62%) in the E coli group and 102 patients (69%) in the K pneumoniae group initially received either first or second-generation cephalosporins with or without gentamicin. The frequency of poor initial antibiotic response was higher in the E coli group than in the K pneumoniae group (22% vs 4%; adjusted OR, 4.9; 95% CI, 1.6 –14) (Table 3). Morbidity and Outcome In the E coli group, spread of infection from liver abscess occurred in 5 patients; all were to contiguous regions and led to subcutaneous abscess of the abdomen in 1 patient and splenic abscess in 4 patients. In the K pneumoniae group, spread of infection from the liver abscess resulted in subsequent subcutaneous/ muscular abscesses, intra-abdominal abscesses, septic endophthalmitis, meningitis, pulmonary abscess, septic pulmonary embolism, and osteomyelitis (Table 4). No patients died of contiguous infections in the E coli group; one patient with septic pulmonary embolism in the K pneumoniae group died of uncontrolled sepsis. Subsequent invasive intervention was required for 9 patients with E coli liver abscess (5 with initial antiTHE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
biotics alone and 4 with initial PCD) and 13 patients with K pneumoniae liver abscess (5 with initial antibiotics alone, 5 with initial PCD, and 3 with initial PNA). Causes of death included uncontrolled sepsis (12 in the E coli group and 3 in the K pneumoniae group), uncontrolled sepsis with concomitant acute respiratory distress syndrome (1 in the E coli group and 2 in the K pneumoniae group), and uncontrolled sepsis with concomitant nosocomial pneumonia (1 in the E coli group and 1 in the K pneumoniae group). The potential confounders, which included baseline covariates and significant clinical variables, were adjusted for final outcome analysis. The significant clinical variables included diabetes mellitus, biliary disorder, malignancy, cryptogenic origin of abscess, polymicrobial infection, anaerobic infection, MDR isolates, pleural effusion, anemia, and initial ineffective antibiotic treatment. After adjusting for baseline covariates and the significant clinical variables, we found that patients in the E coli group had a higher frequency of time to defervesce ⬎ 1 week after admission (advanceadjusted OR, 3.8; 95% CI, 1.3 to 11) than those in the 101
E coli and K pneumoniae Liver Abscesses
Table 4. Morbidity and Clinical Outcome in the E Coliand K Pneumoniae Groups E coli vs K pneumoniae
Variable Metastatic/contiguous infection,a No. (%) Septic endophthalmitis Meningitis Septic pulmonary embolism Pulmonary abscess Intra-abdominal abscessb Subcutaneous/muscular abscess Osteomyelitis Change therapeutic modality, No. (%) Time to defervesce ⬎1 week after admission, No. (%) (median; IQR; days) Hospital stay ⬎3 weeks, No. (%) (median; IQR; days) Length of antibiotic treatment ⬎6 weeks, No. (%) (median; IQR; days) Recurrence, No. (%) Death, No. (%)
E coli Liver Abscess (n ⫽ 55)
K pneumoniae Liver Abscess (n ⫽ 147)
Baseline-Adjusted ORc (95% CI)
Advance-Adjusted ORd (95% CI)
5 (9)
20 (14)
0.41 (0.11–1.5)
0.60 (0.12–3.0)
0 0 0 0 4 1 0 9 (16) 25 (46)
5 3 4 1 4 5 1 13 (9) 39 (27)
2.0 (0.76–5.5) 2.4 (1.2–4.9)*
1.6 (0.37–6.7) 3.8 (1.3–11)*
(7; 2–13) 31 (56) (22; 15–36) 18 (33)
(4; 1–8) 61 (42) (19; 14–25) 70 (48)
1.4 (0.73–2.8)
0.97 (0.35–2.7)
0.54 (0.27–1.1)
0.65 (0.25–1.7)
(31; 22–45) 3 (6) 14 (26)
(42; 30–53) 2 (1) 6 (4)
5.4 (0.69–42) 9.3 (2.7–32)†
7.0 (0.47–97) 4.2 (0.63–27)
CI indicates confidence interval; IQR, interquartile range; OR, odds ratio. When patients fit into more than one category, they were counted in each category. Intra-abdominal abscess: Splenic abscess or renal abscess. c The exact logistic regression model included adjustment for age, sex, and severity of illness at admission. c The exact logistic regression model included adjustment for baseline covariates, diabetes mellitus, biliary disorder, malignancy, cryogenic origin of abscess, polymicrobial infection, anaerobic infection, MDR isolate, pleural effusion, anemia, and initial ineffective antibiotic treatment. * P ⬍ 0.05. † P ⬍ 0.001. a b
K pneumoniae group (Table 4). The incidence of metastatic/contiguous infection, the frequency of requirement to change treatment method, the recurrence rate, the length of hospital stay, and the duration of treatment with antibiotics did not differ between the 2 groups. After adjusting for the baseline covariates, we found that patients in the E coli group (26%) had a higher mortality rate than those in the K pneumoniae group (4%); however, there was no significant difference in mortality rate between the 2 groups after adjusting for both the baseline covariates and significant clinical variables (Table 4). Analyses of Prognostic Factors Related to Death in the E Coli and K Pneumoniae Groups Separately In the E coli group, bivariate analysis revealed that the APACHE II score, the presence of right-lobar abscess involvement, multiple abscesses, MDR isolates, and underlying malignancy were significant prognostic variables; however, only the APACHE II score, right-lobar abscess involvement, and underlying malignancy fit the stepwise, logistic regression model and reached statistical significance (Table 5). In the K pneumoniae group, bivariate analysis revealed that uremia, gas-forming abscess, MDR isolates, and ineffective initial antibiotic treatment were significant prognostic variables. However, only underlying ure102
mia and MDR isolates were statistically independent prognostic factors for death in the multivariate analysis (Table 5). Discussion A number of features differed between patients with E coli liver abscess and those with K pneu-
Table 5. Multivariate Analysis of Prognostic Factors in Relation to Death in the E Coliand the K Pneumoniae Groups Separately Variable E coli group (n ⫽ 55) APACHE II score Malignancy Right-lobar abscess involvement K Pneumoniae Group (n ⫽ 147) Uremia MDR isolates
ORa
(95% CI)
P Value
1.7 26 0.0029
(1.1–2.6) (1.8–370) (0.00010–0.15)
0.021 0.016 0.004
(3.5–750) (2.3–290)
0.004 0.009
52 26
APACHE indicates Acute Physiology and Chronic Health Evaluation; OR, odds ratio; CI, confidence interval. a Using stepwise logistic regression model.
August 2007 Volume 334 Number 2
Chen et al
moniae liver abscesses. We found that patients in the K pneumoniae group were more likely to be men and younger than patients in the E coli group. The difference in sex ratio can be attributed to the fact that patients in the E coli group had a higher proportion of underlying biliary disease, which is more prevalent in women. Current techniques and procedures used for culturing pathogenic organisms, especially anaerobes, do not always lead to perfect cultures. Anaerobic organisms were recovered from approximately twofifths of the patients with E coli liver abscess in the present study, which is within the range (9 to 46%) of the anaerobic recovery rates among pyogenic liver abscess in recent reports.3,5,6,8,13,15,23,24 MDR isolates from K pneumoniae liver abscesses in our study were uncommon, which is consistent with previous reports on K pneumoniae liver abscess.9,10,16,25 On the other hand, the frequency of MDR isolates (E coli isolates and concomitant microbial isolates) and the ineffective initial antibiotic treatment was higher in patients with E coli liver abscess. Old age and underlying malignancy have been shown to be risk factors for acquisition of antibiotic resistant pathogens,26 –28 which may explain the high incidence of MDR isolates in the E coli group. Furthermore, many of the patients in the E coli group had coexisting biliary tract infection, and may have been taking antibiotics to treat their conditions before admission. However, information regarding antibiotic use among patients prior to admission had not been noted in the medical records, and therefore, a definite conclusion cannot be drawn. Further prospective studies should be conducted to clarify it. The optimal choice of initial antibiotics for patients with pyogenic liver abscess is still unknown. Johannsen et al8 recommended that the empirical initial antibiotics should be composed of ampicillin and gentamicin with/without metronidazole, especially for patients with liver abscesses of biliary origin. Chu et al6 reported that the combination of a second-generation cephalosporin and metronidazole would cover most pathogens. Our data indicate that initial antibiotic regimens should comprise a combination of 2 or 3 antimicrobial classes (eg, a secondgeneration cephalosporin and an aminoglycoside with/without metronidazole) for patients with E coli liver abscesses, given the high frequency of MDR and anaerobic bacteria isolated from those patients. Our data also show that an initial antimicrobial treatment comprising first-generation cephalosporin should cover most of the pathogens found in K pneumoniae liver abscesses. The suspected origin of abscess and underlying diseases should be considered when choosing initial antimicrobial agents before the culture results are known because it will help in predicting the most likely pathogens. For instance, in our study, E coli was the most common pathogen isolated from patients with liver abscesses of biliary THE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
origin, and from those with coexisting malignancy. The presence of biliary disorders and concomitant malignancy had a high specificity and likelihood ratio, and a modest positive predictive value for the prediction of E coli as the etiology of liver abscess. In contrast, K pneumoniae was the most prevalent pathogen isolated from patients with liver abscesses of cryptogenic origin, and from those with concomitant diabetes mellitus. The presence of diabetes mellitus and liver abscesses of cryptogenic origin had a good specificity and positive predictive value, and a modest likelihood ratio for the prediction of K pneumoniae as the etiology of liver abscess. The effectiveness of such initial antibiotics strategies needs to be evaluated further. The incidence of metastatic infection from K pneumoniae liver abscess was 14% in our study, which is consistent with the incidence rate of metastatic infection from K pneumoniae liver abscess reported previously.9,10,13,29 Although very few reports of the development of outspreading infection in patients with E coli liver abscess have been published, the incidence in our study was 9%. The most common outspreading infection from liver abscess in the E coli group occurred in the abdominal region. In the K pneumoniae group, outspreading infection from liver abscess tended to occur at more distant locations, such as eyes, meninges, lungs, and bones. Hematogenous seeding has been reported to be a possible route of outspreading infection from pyogenic liver abscesses, especially those caused by K pneumoniae.9,10,29 We suggest that contiguous spreading might be an alternative route of infectious spread from pyogenic liver abscess, especially those caused by E coli. This may explain why the spread of infections from E coli liver abscesses were more likely to the abdominal region than to distant sites. In addition, contiguous infections in patients with E coli liver abscesses are often in the abdominal region; it is feasible that these foci can be detected early using abdominal US/CT scan during hospitalization. This may explain why there were no deaths among patients with contiguous infection in the E coli group. Some investigators17,18 have suggested that the decreased mortality rates for pyogenic liver abscess in recent years may be attributed, in part, to a change in the predominant causative pathogens, especially the emergence of K pneumoniae. After a head-to-head comparison of the 2 groups, we found no statistical difference in mortality rate between patients in the E coli and those in the K pneumoniae groups after adjusting for potential confounders. Many studies have attempted to define the risk factors for death among patients with pyogenic live abscesses.3,6,8,10,12,24 However, no general consensus has been reached regarding risk factors. In the present study, the differences in mortality predictors between the 2 groups were not organism-specific; instead, the differences appeared to be due to the heterogeneity of the 2 groups. Variables 103
E coli and K pneumoniae Liver Abscesses
such as high APACHE scores, uremia, MDR isolates, and malignancy are seen in a variety of infectious conditions. Interestingly, right-lobar abscess involvement was found herein to be a protective factor in the E coli group. Right-sided liver abscess is relatively easy to locate and drain. This is because the drainage route can be easily planned to avoid major vascular and biliary structures. In the present study, patients in the E coli group had a longer time to defervesce after admission than those in the K pneumoniae group despite the fact that no difference in mortality rate was noted between the both groups. This finding was similar to that reported in our prior study which compared pyogenic liver abscesses of biliary origin with those of cryptogenic origin.13 This finding implies that the long period for defervesce may be related to coexisting biliary tract infection rather than to liver abscess per se. One of the limitations of our study was its retrospective design. Although bivariate analyses did not detect differences among some of the covariates, that statistical method may have been insufficient at detecting differences in covariate values between the 2 groups of patients. Another limitation of our study was the sample size of patients; however, the fact that pyogenic liver abscess is relatively rare makes it difficult to collect a large sample during a finite study period. Hence, we used the exact method instead of the asymptotic method for logistic regression, because inference from the exact method is more reliable and valid in such a situation.22 Additionally, to compare outcome between patients with E coli and those with K pneumoniae, we adjusted for baseline covariates and the significant clinical variables. Finally, we must draw the readers’ attention to the wide confidence intervals for many of the variables, and point out that such intervals do not rule out important differences between groups. In conclusion, our data suggest that the 2 leading causative pathogens of pyogenic liver abscess, E coli and K pneumoniae, are dissimilar in many ways, including demography, underlying diseases, origin of abscess, microbiological characteristics, and clinical response. Moreover, the higher mortality rate associated with E coli liver abscess may be attributed to the severity of illness at admission and high frequency of underlying malignancy rather than to the pathogen itself. Knowledge of the underlying diseases and the origin of liver abscess is useful for predicting the microbial etiology of the liver abscess before the culture results are known. Clinicians should begin with broad antibiotic coverage such as a second-generation cephalosporin and an aminoglycoside with metronidazole when treating liver abscesses given the high frequency of MDR isolates seen among E coli isolates. 104
Acknowledgments We would like to thank Professors H.S. Lee, PhD, and R.H. Wong, PhD, School of Public Health, Chung Shan Medical University, for their expert assistance in statistical analysis. References 1. Gyorffy EJ, Frey CF, Silva J Jr. Pyogenic liver abscess: diagnostic and therapeutic strategies. Ann Surg 1987;206: 699–705. 2. Cohen JL, Martin FM, Rossi RL, et al. Liver abscess: the need for complete gastrointestinal evaluation. Arch Surg 1989;124:561–4. 3. Frey CF, Zhu Y, Suzuki M, et al. Liver abscesses. Surg Clin N Am 1989;69:259–71. 4. Branum GD, Tyson GS, Branun MA, et al. Hepatic abscess: changes in etiology, diagnosis, and management. Ann Surg 1990;212:655–62. 5. Seeto RK, Rockey DC. Pyogenic liver abscess: changes in etiology, management, and outcome. Medicine 1996;75:99–113. 6. Chu KM, Fan ST, Lai ECS, et al. Pyogenic liver abscess: an audit of experience over the past decade. Arch Surg 1996;131: 148–52. 7. Yu SC, Lo RH, Kan PS, et al. Pyogenic liver abscess: treatment with needle aspiration. Clin Radiol 1997;52:912–6. 8. Johannsen EC, Sifri CD, Madoff LC. Pyogenic liver abscesses. Infect Dis Clin N Am 2000;14:547–63. 9. Chang FY, Chou MY. Comparison of pyogenic liver abscesses caused by Klebsiella pneumoniae and non-K pneumoniae pathogens. J Formos Med Assoc 1995;94:232–7. 10. Wang JH, Liu YC, Lee SSJ, et al. Primary liver abscess due to Klebsiella pneumoniae in Taiwan. Clin Infect Dis 1998;26: 1434–8. 11. Wiwanitkit V, Suwansaksri N, Suwansaksri J. Causative agents of liver abscess in those with liver cirrhosis: a 10-year case review of hospitalized patients in Thailand. Ann Trop Med Parasitol 2002;96:513–6. 12. Lee KT, Wong SR, Sheen PC. Pyogenic liver abscess: an audit of 10 years’ experience and analysis of risk factors. Digest Surg 2001;18:459–66. 13. Chen SC, Yen CH, Tsao SM, et al. Comparison of pyogenic liver abscesses of biliary and cryptogenic origin: an eight-year analysis in a University Hospital. Swiss Med Wkly 2005;135: 344–51. 14. Moulds-Merritt C, Frazee RC. Therapeutic approach to hepatic abscesses. South Med J 1994;87:884–8. 15. Huang CJ, Pitt HA, Lipsett PA, et al. Pyogenic hepatic abscess: changing trends over 42 years. Ann Surg 1996;223: 600–9. 16. Rahimian J, Wilson T, Oram V, et al. Pyogenic liver abscess: recent trends in etiology and mortality. Clin Infect Dis 2004;39:1654–9. 17. Lederman ER, Crum NF. Pyogenic liver abscess with a focus on Klebsiella pneumoniae as a primary pathogen: an emerging disease with unique clinical characteristics. Am J Gastroenterol 2005;100:322–31. 18. Lederman ER, Crum NF. Klebsiella pneumoniae liver abscess: a coast-to-coast phenomenon. Clin Infect Dis 2005;41:273. 19. National Committee for Clinical Laboratory Standards. 2000. Performance standards for antimicrobial susceptibility testing. Approved standard M100-S10. National Committee for Clinical Laboratory Standards, Wayne, Pa. 20. National Committee for Clinical Laboratory Standards. 2004. Performance standards for antimicrobial susceptibility test-
August 2007 Volume 334 Number 2
Chen et al
21.
22. 23.
24.
25.
ing. Approved standard M100-S14. National Committee for Clinical Laboratory Standards, Wayne, PA. Knaus WA, Draper EA, Wagner DP, et al. APACHE II: a severity of disease classification system. Crit Care Med 1985; 13:818–29. Mehta CR, Patel NR. Exact logistic regression: theory and examples. Stat Med 1995;14:2143–60. Monsen AH, Green ST, Read RC, et al. Liver abscess in adults: ten years experience in an UK centre. QJM 2002;95: 797–802. Alvarez Perez JA, Gonzalez JJ, Baldonedo RF, et al. Clinical course, treatment, and multivariate analysis of risk factors for pyogenic liver abscess. Am J Surg 2001;181:177–86. Cheng HP, Siu LK, Chang FY. Extended-spectrum cephalosporin compared to cefazolin for treatment of Klebsiella pneumoniae-caused liver abscess. Antimicrob Agents Chemother 2003;47:2088–92.
THE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
26. Carratala J, Fernandez-Sevilla A, Tubau F, et al. Emergence of quinolone-resistant Escherichia coli bacteremia in neutropenic patients with cancer who have received prophylactic norfloxacin. Clin Infect Dis 1995;20:557–60. 27. Perea S, Hidalgo M, Arcediano A, et al. Incidence and clinical impact of fluoroquinolone-resistant Escherichia coli in the fecal flora of cancer patients treated with high dose chemotherapy and ciprofloxacin prophylaxis. J Antimicrob Chemother 1999;44:117–20. 28. Lautenbach E, Fishman NO, Bilker WB, et al. Risk factors for fluoroquinolone resistance in nosocomial Escherichia coli and Klebsiella pneumoniae infections. Arch Intern Med 2002;162:2469–77. 29. Cheng DL, Liu YC, Yen MY, et al. Septic metastatic lesions of pyogenic liver abscess: their association with Klebsiella pneumoniae bacteremia in diabetic patients. Arch Intern Med 1991;151:1557–9.
105
ABSTRACT: Background: Escherichia coli and Klebsiella pneumoniae are the most common causative pathogens of pyogenic liver abscesses. The objective of this study was to compare outcome between patients with liver abscesses due to E coli and those with liver abscesses caused by K pneumoniae; we also aimed to identify separately the predictors of mortality in the 2 groups. Methods: We conducted a retrospective study of 202 patients who presented with pyogenic liver abscesses caused by either E coli or K pneumoniae from July 2000 to June 2005. Outcome of the patients was analyzed by exact logistic regression with adjustment for baseline and clinical covariates. Significant predictors of mortality in the E coli and the K pneumoniae groups were investigated by multivariate analysis of demographic and clinical variables in each group. Results: Of the 202 patients (128 men and 74 women; age range, 19 to 89 years), pyogenic liver abscess was due to E coli infection in 55 patients and K pneumoniae in 147 patients. In contrast to patients with K pneumoniae, patients with E coli liver abscess were more likely to be older and female, have a biliary abnormality or malignancy, pleural effusion, polymicrobial infection with anaerobic or multi– drug-resistant organisms, a higher APACHE II score, and to have been treated initially with ineffective antibiotics; they were also less likely to have diabetes mellitus. The cause of K pneumoniae liver abscess was often cryptogenic. The sensitivity, specificity, positive predictive value, and likelihood ratio of the presence of biliary disorders and coexisting malignancy as a predictive
parameter of E coli liver abscess were 25%, 96%, 67%, and 5.45/1, respectively. The sensitivity, specificity, positive predictive value, and likelihood ratio of the presence of diabetes mellitus with an abscess of cryptogenic origin as a predictive parameter of K pneumoniae liver abscess were 39%, 84%, 81%, and 2.36/1, respectively. There was no significant difference in mortality between patients with E coli and those with K pneumoniae infections (26% vs 4%; adjusted OR, 4.2; 95% CI, 0.63 to 27; P ⫽ 0.105). However, for patients with liver abscess caused by E coli, the APACHE II score at admission (OR, 1.7; 95% CI, 1.1 to 2.6; P ⫽ 0.021), malignancy (OR, 26; 95% CI, 1.8 to 370; P ⫽ 0.016), and right-lobe abscess (OR, 0.0029; 95% CI, 0.00010 to 0.15; P ⫽ 0.004) were significant predictors of death, whereas uremia (OR, 52; 95% CI, 3.5 to 750; P ⫽ 0.004) and multi– drug-resistant isolates (OR, 26; 95% CI, 2.3 to 290; P ⫽ 0.009) were significant predictors of death in the K pneumoniae group. Conclusions: A higher APACHE II score at admission and a higher frequency of coexisting malignancy may have contributed to the higher, although not significant, mortality rate in patients with liver abscess caused by E coli infection. Clinicians should begin with broad antibiotic coverage such as a second-generation cephalosporin and an aminoglycoside with metronidazole when treating liver abscesses with E coli as the likely pathogen due to the high frequency of multi–drug-resistant isolates among E coli isolates. KEY INDEXING TERMS: Escherichia coli; Klebsiella pneumoniae; Pyogenic liver abscess; Outcome. [Am J Med Sci 2007;334(2):97–105.]
P
Escherichia coli was the predominant organism isolated from pyogenic liver abscess worldwide until the 1980s,1,3,5 when Klebsiella pneumoniae began to be
yogenic liver abscess continues to be a potentially life-threatening disease despite the advances in microbiology methods, imaging techniques, and therapeutic modalities.1– 8 In terms of causative pathogens,
Institute of Medicine (SCC, WYW, KCL, MCL) and Schools of Medicine (SCC, PHW, CCC) and Medical Laboratory and Biotechnology (DBL), Chung Shan Medical University; Department of Family Medicine (SCC, WYM, CCC, MCL), Chung Shan Medical University Hospital, Taichung, Taiwan; Department of Neurology (CHY), Show Chuan Memorial Hospital, Changhua, Taiwan; Department of Life Sciences (CHY), National Chung Hsing University, Taichung, Taiwan; Departments of Surgery (KCL, LBJ) and Internal Medicine (KSC), China Medical University Hospital, Taichung, Taiwan; and Department of Medicine, Division of Hematology (WRB), THE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
The Johns Hopkins University School of Medicine, Baltimore, Maryland. Submitted July 27, 2006; accepted in revised form January 17, 2007. None of the authors have any financial involvement with any organization. Correspondence: Dr. Chun-Chieh Chen, Faculty of Medicine, Chung Shan Medical University School of Medicine, Department of Family Medicine, Chung Shan Medical University Hospital, No. 110, Section 1, Chien Kuo N. Road, Taichung 402, Taiwan (E-mail: [email protected]).
97
E coli and K pneumoniae Liver Abscesses
reported as the leading causative pathogen of pyogenic liver abscess in some Asian countries.7,9 –13 There is evidence that a shift from E coli to K pneumoniae liver abscesses is occurring gradually in some Western countries.4,14 –18 E coli and K pneumoniae liver abscesses have been described individually in several studies1,7,9 –13; however, there is still a lack of a comprehensive comparison between the 2 leading causative pathogens involved in the development of pyogenic liver abscesses. Hence, we conducted a retrospective study to delineate the clinical features of patients with E coli liver abscesses and those with K pneumoniae liver abscesses to compare the outcomes between these 2 types of patients and to identify the predictors of mortality in the 2 groups. Methods Study Patients The medical records of consecutive patients aged 18 years and older who had been admitted to the China Medical University Hospital, a 1600-bed primary and tertiary care medical center in Taichung, Taiwan, from July 2000 to June 2005, and discharged with a diagnosis of pyogenic liver abscess were reviewed by a systematic search of the patients’ records for diagnostic code 572.0 (International Classification of Diseases, 9th Revision). Patients were included in the study if they met both of the following criteria: 1) one or more discrete abscess cavities of the liver had been identified by imaging studies [endoscopic retrograde cholangiopancreatography (ERCP), abdominal ultrasonography (US) and/or computerized axial tomography (CT) scans with contrast enhancement], and 2) either K pneumoniae or E coli had been isolated from abscess culture. Of the 252 patients with pyogenic liver abscess that were screened, 202 patients fulfilled the inclusion criteria; they were divided into 2 groups depending on whether they had E coli liver abscess (E coli group) or K pneumoniae liver abscess (K pneumoniae group). A total of 50 patients, of the 252 patients with pyogenic liver abscess that were screened, were excluded for the following reasons: amebic liver abscess (n ⫽ 2), liver abscess due to ascarid infection (n ⫽ 1), fungal liver abscess (n ⫽ 2), tuberculous liver abscess (n ⫽ 1), incomplete medical records (n ⫽ 10), necrotic tumors (n ⫽ 1), biloma (n ⫽ 1), liver abscess due to concomitant E coli and K pneumoniae infections (n ⫽ 11), and liver abscess due to the causative organisms isolated only from blood culture, but not the liver abscess (n ⫽ 21). This study was approved by the Institutional Review Board of the China Medical University Hospital. Data Collection Demographic and medical information were collected. The medical information included clinical, imaging, microbiological, and laboratory parameters. The clinical parameters comprised underlying medical conditions, presenting symptoms/signs, type of treatment, and outcome. Abscess material had been obtained by image-guided (CT or US) percutaneous needle aspiration (PNA), image-guided percutaneous catheter drainage (PCD), or direct surgical intervention. The specimens obtained had been processed for Gram stain, bacterial cultures (standard aerobic and anaerobic diagnostic methods), and tests for antimicrobial susceptibility. Antimicrobial susceptibility had been determined by the disk diffusion method (BD BBL, Sensi-Disc Antimicrobial Susceptibility Test Discs, Sparks, MD), based on the microorganism isolated. The results were evaluated according to the recommendations set forth by the Clinical and Laboratory Standards Institute (formerly known as the National Committee for Clinical Laboratory Standards).19,20 Multi– drug resistance was defined as resistance to 3 or more of the antimicrobial classes. After the cultures of blood and/or liver
98
abscess aspirate had been obtained, initial empirical broad-spectrum antibiotic treatment was administered intravenously, and adjusted as necessary when the results of the culture and susceptibility test were known. When isolates from patients were resistant to all of the initial antibiotics that had been given, the initial treatment was regarded to have been ineffective. Response to treatment was evaluated in each patient by intensive follow-up US or CT scan of the liver during hospitalization and/or subsequent out-patient visits after discharge. The origin of the abscess was based on the available imaging, clinical, pathological and/or surgical findings. If no obvious extra-hepatic source of infection could be identified after appropriate investigation, the abscess was considered cryptogenic in origin. The severity of illness at admission was evaluated by the first-day Acute Physiology and Chronic Health Evaluation II (APACHE II) score.21 Patient outcome assessments included the recurrence of liver abscess, hospital stay, and death. Recurrence of liver abscess was defined as the development of new clinical and radiological changes after discharge. The median follow-up of patients after discharge was 3 months (range 1 to 9 months). Death was defined as death within 30 days of treatment or before discharge from the hospital. The cause of death was ascertained from the death certificate. Statistical Analysis All statistical analyses were performed using a statistical software package (SAS version 8.2, SAS Institute Inc, Cary, NC). For analysis, patients with pyogenic liver abscess were divided into two groups according to bacterial etiology: the E coli group and the K pneumoniae group. Descriptive data are presented as medians with interquartile ranges (IQRs) for continuous data and percentages for categorical data. Continuous variables were compared by the Mann-Whitney U test. Categorical variables were compared by either the 2 test or Fisher exact test (if more than 20% of the expected frequencies were ⬍5), as appropriate. Because of the imbalanced data set in our study, data were analyzed by exact logistic regression rather than by conventional logistic regression.22 We constructed an exact logistic regression model with covariate adjustment to estimate odds ratios (ORs) and 95% confidence intervals (CIs) to analyze associations between dependent and independent variables. The baseline variables, including sex, age, duration from the onset of symptoms to the time of diagnosis, and severity of illness at admission, were tested between the 2 bacterial groups. The significant variables above were regarded as baseline covariates. Associations between the bacterial group (E coli vs K pneumoniae) and clinical variables at admission were analyzed after adjusting the same set of baseline covariates. Outcome was analyzed between the 2 groups of patients by exact logistic regression with adjustment for the baseline covariates and the significant clinical variables. Data were then analyzed by bivariate and multivariate analyses to identify significant independent predictive factors of death associated with E coli liver abscess and those associated with K pneumoniae abscess. The significant independent variables obtained by bivariate analyses were entered into the multiple stepwise logistic regression model, which determined the prognostic factors associated with death. All P values were two-tailed, and P ⬍ 0.05 was considered statistically significant in all analyses. In addition, P ⬍ 0.001 was particularly shown among multiple tests in tables.
Results Demographic Data, Underlying Diseases, and Symptoms/Signs Of the 202 patients with pyogenic liver abscess, E coli had been diagnosed in 55 patients (E coli group) and K pneumoniae in 147 patients (K pneumoniae group). The median age of patients with E coli liver August 2007 Volume 334 Number 2
Chen et al
Table 1. Underlying Diseases and Presenting Symptoms/Signs at Admission in the Escherichia coliand K lebsiella pneumoniae Groups Variable
E coli Liver Abscess (n ⫽ 55)
K pneumoniae Liver Abscess (n ⫽ 147)
E coli vs K pneumoniae Adjusted ORf (95% CI)
19 (35) 42 (76) 18 (33) 4 (7) 8 (15) 3 (6) 17 (31) 4 (7)
82 (56) 47 (32) 15 (10) 22 (15) 14 (10) 8 (5) 112 (76) 4 (3)
0.31 (0.15–0.64)* 6.4 (3.0–14)† 2.9 (1.2–7.1)* 0.97 (0.28–3.3) 1.7 (0.63–4.7) 0.32 (0.061–1.7) 0.14 (0.068–0.29)† 2.5 (0.41–16)
49 (91) 38 (69) 26 (47) 23 (42) 20 (36) 20 (36) 19 (35) 11 (20) 8 (15) 5 (9) 2 (4)
140 (95) 77 (52) 61 (42) 58 (40) 50 (34) 42 (29) 45 (31) 17 (12) 17 (12) 8 (5) 11 (8)
0.44 (0.12–1.6) 2.9 (1.4–6.1)* 1.6 (0.83–3.3) 1.3 (0.65–2.6) 1.0 (0.52–2.1) 2.2 (1.1–4.6)* 0.96 (0.47–2.0) 2.5 (1.1–6.5)* 1.1 (0.36–3.2) 2.6 (0.74–8.9) 0.18 (0.031–1.1)
Underlying condition,a No. (%) Diabetes mellitus Biliary disorderb Malignancy Alcoholism Liver cirrhosis Uremia Cryptogenic origin of abscess Other origin of abscessc Symptoms/signs,a No. (%) Fever/chills RUQ pain/tenderness Malaise Anorexia Nausea/emesis Jaundice Chest symptomsd Murphy signe Ascites Weight loss Mental confusion
CI indicates confidence interval; OR, odds ratio; RUQ, right upper quadrant. When patients fit into more than one category, they were counted in each category. b Biliary disorder: Cholelithiasis, choledocholithiasis, hepatolithiasis, cholecystitis, or purulent cholangitis with/without malignancy. c Others: Pancreatitis, empyema of the gall bladder, subphrenic abscess ulcerative colitis, colonic diverticulitis, liver injury, prior transcatheter arterial embolization for hepatocellular carcinoma, or recent biliary surgery. d Chest symptoms: Cough, chest pain, and/or short of breath. e Murphy sign: Deep inspiration or cough during subcostal palpation of the RUQ producing increased pain and inspiratory arrest. f The exact logistic regression model included adjustment for age, sex, and severity of illness at admission. * P ⬍ 0.05. † P ⬍ 0.001. a
abscess was significantly greater than that of patients with K pneumoniae liver abscess (67 years [IQR, 61 to 74 years] vs 54 years [IQR, 45 to 68 years]; P ⬍ 0.001; using the Mann-Whitney U test). There was a significant difference in sex between the E coli and K pneumoniae groups (male/female ⫽ 27/28 vs 101/36; P ⫽ 0.01; using the 2 test). The median APACHE II score at admission among patients in the E coli group was significantly higher than that among patients in the K pneumoniae group (12 points [IQR, 9 to 15 points] vs 8 points [IQR, 5 to 12 points]; P ⬍ 0.001; using the MannWhitney U test). The median duration from the onset of symptoms to the time of diagnosis was 5 days (IQR, 3 to 10 days) in the E coli group and 6 days (IQR, 3–9 days) in the K pneumoniae group; however, the difference between groups was not significant (P ⫽ 0.468; using the Mann-Whitney U test). Age, sex, and APACHE II score at admission were considered the baseline covariates. After adjusting for the baseline covariates, patients in the E coli group were more likely to have biliary disorders (cholelithiasis, choledocholithiasis, hepatolithiasis, cholecystitis, and/or purulent cholangitis) and underlying malignancy than those in the K pneumoniae group. Patients in the K pneumoniae group THE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
were more likely to have diabetes mellitus and abscesses that were cryptogenic in origin than patients in the E coli group. The underlying diseases and presenting symptoms/signs at admission of these patients are summarized in Table 1. Among the 242 patients with complete medical records, an initial pending diagnosis of liver abscess and available diagnostic radiological studies, the significant variables regarding underlying diseases and origin of abscess were analyzed to see whether they could be used to predict the etiology of the abscess. The sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of the presence of biliary disorder as a predictive factor of E coli liver abscess were 80%, 68%, 48%, and 90%, respectively. The sensitivity, specificity, PPV and NPV of the presence of biliary disorder and coexisting malignancy as a predictive parameter of E coli liver abscess were 25%, 96%, 67%, and 76%, respectively. The likelihood ratio for the prediction of E coli as the etiology of liver abscesses in patients with biliary disorder and concomitant malignancy was 5.45/1. The sensitivity, specificity, PPV and NPV of the presence of diabetes mellitus as a predictive factor of K pneumoniae liver abscess were 57%, 61%, 73%, and 43%, respectively. The sensitivity, specificity, PPV and 99
E coli and K pneumoniae Liver Abscesses
NPV of abscesses of unknown origin as a predictive factor of K pneumoniae liver abscess were 75%, 65%, 80%, and 59%, respectively. The sensitivity, specificity, PPV and NPV of the presence of diabetes mellitus with an abscess of cryptogenic origin as a predictive parameter of K pneumoniae liver abscess were 39%, 84%, 81%, and 43%, respectively. The likelihood ratio for the prediction of K pneumoniae as the etiology of liver abscesses in patients with diabetes mellitus and liver abscesses of cryptogenic origin was 2.36/1. Microbiological, Imaging, and Laboratory Findings and Initial Therapeutic Modalities The species of organisms that had been isolated from patients in the E coli and K pneumoniae groups are summarized in Table 2. Polymicrobial infection, anaerobes, and multi– drug-resistant (MDR) isolates (from abscess or blood cultures) were found more often in the E coli group than in the K pneumoniae group (Table 3). Of the 74 E coli isolates from blood and abscess cultures, 19 (26%) were susceptible to all antimicrobials tested, 31 (42%) were resistant to 1 or 2 antimicrobial classes, and 24 (32%) were MDR strains. Among the E coli isolates, the majority were resistant to ampicillin (73%), followed by trimethoprim-sulfamethoxazole (45%), cephalothin (36%), piperacillin (26%), amoxicillin/
clavulanic acid (22%), gentamicin (18%), ciprofloxacin (14%), and cefuroxime (8%). Only one E coli isolate was resistant to ceftazidime. Most (218/228, 96%) of the K pneumoniae isolates from blood and abscess cultures were susceptible to all antimicrobial agents tested, except ampicillin. Of the remaining 10 K pneumoniae isolates, 1 was susceptible to all antimicrobials tested, 5 were resistant to 1 or 2 antimicrobial classes, and 4 were MDR isolates. Image findings revealed that patients in the E coli group had a higher frequency of pleural effusion than patients in the K pneumoniae group; however, no other differences in image findings were seen between the 2 groups (Table 3). Patients in the E coli group had a higher frequency of anemia compared with those in the K pneumoniae group; however, no other differences in frequency of abnormal laboratory values were noted between the groups (Table 3). Patients were initially managed with 1 of 4 therapeutic approaches: antibiotics alone, antibiotics plus PNA, antibiotics plus PCD or antibiotics plus surgical intervention. The most common primary treatment modality was percutaneous drainage (PNA or PCD) (E coli group, 89%; K pneumoniae group, 94%) All patients initially received intravenous empirical antibiotics (cephalosporins, penicillin group, aminoglycosides, and metronidazole).
Table 2. Microbiological Spectrum in E coli andK pneumonia Liver Abscesses E coli Liver Abscess (n ⫽ 55) Organisms Gram-negative aerobes Klebsiella pneumoniae Escherichia coli Klebsiella oxytoca Pseudomonas aeruginosa Morganella morganii Proteus spp Enterobacter spp Aeromonas spp Acinetobacter spp Citrobacter spp Burkholderia cepacia Edwardsiella tarda Gram-positive aerobes Enterococcus spp Streptococcus spp ORSA CoNS Gemella morbillorum Micrococcus spp Anaerobes Bacteroides spp Prevotella spp Fusobacterium spp Peptostreptococcus spp Bifidobacterium spp Total isolates
K pneumoniae Liver Abscess (n ⫽ 147)
Blood (n ⫽ 55)
Abscess (n ⫽ 55)
19 1 1
55 2 5
2
8 1 2 2 1
Blood (n ⫽ 143)
Abscess (n ⫽ 147)
81
147
1 1 1 1 3 1
1 3 1
6 10 1
4
1 1
3 2 1
1 1 1
6 2
39
14 2 3 5 1 120
2
2
89
1 1 163
CoNS indicates coagulase negative Staphylococci; ORSA, oxacillin-resistant Staphylococcus aureus.
100
August 2007 Volume 334 Number 2
Chen et al
Table 3. Microbiological, Imaging Findings, Laboratory Findings at Admission, and Initial Therapeutic Modalities in the E Coliand the K Pneumoniae Groups Variable Microbiologic characteristics, No. (%) Bacteremia Polymicrobialinfectiona Anaerobicinfectionb MDR isolates Imaging finding, No. (%) Pleural effusion Unilobar abscess Right-lobe abscess Solitary abscess Abscess ⬎5 cm in diameter Gas-forming abscess Rupture of abscess Laboratory tests, No. (%) WBC count (⬎104 cells/mm3) Serum albumin (⬍3.5 g/dL) Hb (⬍14 g/dL in male,⬍12 g/dL in female) Serum AST (⬎40 U/L) Serum ALP (⬎126 U/L) Total bilirubin (⬎1.3 mg/dL) PT (⬎13.1 seconds) Initial treatment method, No. (%) Antibiotics alone Invasive procedurec plus antibiotics Ineffective initial antibiotics
E coli Liver Abscess
K pneumoniae Liver Abscess
E coli vs K pneumoniae Adjusted ORf (95% CI)
35/55 (64) 38/55 (69) 21/55 (38) 23/55 (42)
87/143 (61) 11/147 (8) 7/147 (5) 6/147 (4)
0.92 (0.44–1.9) 22 (9.4–53)† 10 (3.8–26)† 15 (5.2–43)†
38 (69) 45 (82) 29 (53) 37 (67) 32 (58) 9 (16) 2 (4)
60 (41) 124 (84) 98 (67) 115 (78) 104 (71) 36 (25) 9 (6)
2.7 (1.3–5.4)* 0.93 (0.39–2.2) 0.55 (0.28–1.1) 0.60 (0.29–1.3) 0.50 (0.25–1.01) 0.43 (0.18–1.1) 0.71 (0.13–3.8)
44/55 (82) 40/42 (95) 47/55 (86) 33/52 (64) 31/46 (67) 32/47 (68) 29/46 (63)
118/147 (80) 76/77 (99) 88/147 (60) 90/131 (69) 64/112 (57) 67/107 (63) 96/142 (68)
0.78 (0.33–1.8) 0.16 (0.012–2.1) 5.2 (2.0–13)† 0.88 (0.42–1.9) 1.3 (0.61–2.9) 1.5 (0.65–3.4) 0.61 (0.28–1.3)
6/55 (11) 49d/55 (89) 12/55 (22)
6/147 (4) 141e/147 (96) 6/147 (4)
3.0 (0.82–11) 0.33 (0.088–1.2) 4.9 (1.6–14)*
ALP indicates alkaline phosphatase; AST, aspartate aminotransferase; CI, confidence interval; Hb, hemoglobin; MDR, multi–drugresistant; OR, odds ratio; PT, prothrombin time; WBC, white blood cell. a Polymicrobial infection: Mixed bacterial flora were cultured in blood and/or abscess cultures. b Anaerobic infection: Anaerobic isolateswere cultured in blood and/or abscess cultures. c Invasive procedure: percutaneous needle aspiration (PNA), percutaneous catheter drainage (PCD), or surgical intervention. d Of the 49 patients, 4 underwent PNA and 45 underwent PCD. e Of the 141 patients, 12 underwent PNA, 126 underwent PCD, and 3 were managed surgically. f The exact logistic regression model included adjustment for age, sex, and severity of illness at admission. * P ⬍ 0.05. † P ⬍ 0.001.
A total of 34 patients (62%) in the E coli group and 102 patients (69%) in the K pneumoniae group initially received either first or second-generation cephalosporins with or without gentamicin. The frequency of poor initial antibiotic response was higher in the E coli group than in the K pneumoniae group (22% vs 4%; adjusted OR, 4.9; 95% CI, 1.6 –14) (Table 3). Morbidity and Outcome In the E coli group, spread of infection from liver abscess occurred in 5 patients; all were to contiguous regions and led to subcutaneous abscess of the abdomen in 1 patient and splenic abscess in 4 patients. In the K pneumoniae group, spread of infection from the liver abscess resulted in subsequent subcutaneous/ muscular abscesses, intra-abdominal abscesses, septic endophthalmitis, meningitis, pulmonary abscess, septic pulmonary embolism, and osteomyelitis (Table 4). No patients died of contiguous infections in the E coli group; one patient with septic pulmonary embolism in the K pneumoniae group died of uncontrolled sepsis. Subsequent invasive intervention was required for 9 patients with E coli liver abscess (5 with initial antiTHE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
biotics alone and 4 with initial PCD) and 13 patients with K pneumoniae liver abscess (5 with initial antibiotics alone, 5 with initial PCD, and 3 with initial PNA). Causes of death included uncontrolled sepsis (12 in the E coli group and 3 in the K pneumoniae group), uncontrolled sepsis with concomitant acute respiratory distress syndrome (1 in the E coli group and 2 in the K pneumoniae group), and uncontrolled sepsis with concomitant nosocomial pneumonia (1 in the E coli group and 1 in the K pneumoniae group). The potential confounders, which included baseline covariates and significant clinical variables, were adjusted for final outcome analysis. The significant clinical variables included diabetes mellitus, biliary disorder, malignancy, cryptogenic origin of abscess, polymicrobial infection, anaerobic infection, MDR isolates, pleural effusion, anemia, and initial ineffective antibiotic treatment. After adjusting for baseline covariates and the significant clinical variables, we found that patients in the E coli group had a higher frequency of time to defervesce ⬎ 1 week after admission (advanceadjusted OR, 3.8; 95% CI, 1.3 to 11) than those in the 101
E coli and K pneumoniae Liver Abscesses
Table 4. Morbidity and Clinical Outcome in the E Coliand K Pneumoniae Groups E coli vs K pneumoniae
Variable Metastatic/contiguous infection,a No. (%) Septic endophthalmitis Meningitis Septic pulmonary embolism Pulmonary abscess Intra-abdominal abscessb Subcutaneous/muscular abscess Osteomyelitis Change therapeutic modality, No. (%) Time to defervesce ⬎1 week after admission, No. (%) (median; IQR; days) Hospital stay ⬎3 weeks, No. (%) (median; IQR; days) Length of antibiotic treatment ⬎6 weeks, No. (%) (median; IQR; days) Recurrence, No. (%) Death, No. (%)
E coli Liver Abscess (n ⫽ 55)
K pneumoniae Liver Abscess (n ⫽ 147)
Baseline-Adjusted ORc (95% CI)
Advance-Adjusted ORd (95% CI)
5 (9)
20 (14)
0.41 (0.11–1.5)
0.60 (0.12–3.0)
0 0 0 0 4 1 0 9 (16) 25 (46)
5 3 4 1 4 5 1 13 (9) 39 (27)
2.0 (0.76–5.5) 2.4 (1.2–4.9)*
1.6 (0.37–6.7) 3.8 (1.3–11)*
(7; 2–13) 31 (56) (22; 15–36) 18 (33)
(4; 1–8) 61 (42) (19; 14–25) 70 (48)
1.4 (0.73–2.8)
0.97 (0.35–2.7)
0.54 (0.27–1.1)
0.65 (0.25–1.7)
(31; 22–45) 3 (6) 14 (26)
(42; 30–53) 2 (1) 6 (4)
5.4 (0.69–42) 9.3 (2.7–32)†
7.0 (0.47–97) 4.2 (0.63–27)
CI indicates confidence interval; IQR, interquartile range; OR, odds ratio. When patients fit into more than one category, they were counted in each category. Intra-abdominal abscess: Splenic abscess or renal abscess. c The exact logistic regression model included adjustment for age, sex, and severity of illness at admission. c The exact logistic regression model included adjustment for baseline covariates, diabetes mellitus, biliary disorder, malignancy, cryogenic origin of abscess, polymicrobial infection, anaerobic infection, MDR isolate, pleural effusion, anemia, and initial ineffective antibiotic treatment. * P ⬍ 0.05. † P ⬍ 0.001. a b
K pneumoniae group (Table 4). The incidence of metastatic/contiguous infection, the frequency of requirement to change treatment method, the recurrence rate, the length of hospital stay, and the duration of treatment with antibiotics did not differ between the 2 groups. After adjusting for the baseline covariates, we found that patients in the E coli group (26%) had a higher mortality rate than those in the K pneumoniae group (4%); however, there was no significant difference in mortality rate between the 2 groups after adjusting for both the baseline covariates and significant clinical variables (Table 4). Analyses of Prognostic Factors Related to Death in the E Coli and K Pneumoniae Groups Separately In the E coli group, bivariate analysis revealed that the APACHE II score, the presence of right-lobar abscess involvement, multiple abscesses, MDR isolates, and underlying malignancy were significant prognostic variables; however, only the APACHE II score, right-lobar abscess involvement, and underlying malignancy fit the stepwise, logistic regression model and reached statistical significance (Table 5). In the K pneumoniae group, bivariate analysis revealed that uremia, gas-forming abscess, MDR isolates, and ineffective initial antibiotic treatment were significant prognostic variables. However, only underlying ure102
mia and MDR isolates were statistically independent prognostic factors for death in the multivariate analysis (Table 5). Discussion A number of features differed between patients with E coli liver abscess and those with K pneu-
Table 5. Multivariate Analysis of Prognostic Factors in Relation to Death in the E Coliand the K Pneumoniae Groups Separately Variable E coli group (n ⫽ 55) APACHE II score Malignancy Right-lobar abscess involvement K Pneumoniae Group (n ⫽ 147) Uremia MDR isolates
ORa
(95% CI)
P Value
1.7 26 0.0029
(1.1–2.6) (1.8–370) (0.00010–0.15)
0.021 0.016 0.004
(3.5–750) (2.3–290)
0.004 0.009
52 26
APACHE indicates Acute Physiology and Chronic Health Evaluation; OR, odds ratio; CI, confidence interval. a Using stepwise logistic regression model.
August 2007 Volume 334 Number 2
Chen et al
moniae liver abscesses. We found that patients in the K pneumoniae group were more likely to be men and younger than patients in the E coli group. The difference in sex ratio can be attributed to the fact that patients in the E coli group had a higher proportion of underlying biliary disease, which is more prevalent in women. Current techniques and procedures used for culturing pathogenic organisms, especially anaerobes, do not always lead to perfect cultures. Anaerobic organisms were recovered from approximately twofifths of the patients with E coli liver abscess in the present study, which is within the range (9 to 46%) of the anaerobic recovery rates among pyogenic liver abscess in recent reports.3,5,6,8,13,15,23,24 MDR isolates from K pneumoniae liver abscesses in our study were uncommon, which is consistent with previous reports on K pneumoniae liver abscess.9,10,16,25 On the other hand, the frequency of MDR isolates (E coli isolates and concomitant microbial isolates) and the ineffective initial antibiotic treatment was higher in patients with E coli liver abscess. Old age and underlying malignancy have been shown to be risk factors for acquisition of antibiotic resistant pathogens,26 –28 which may explain the high incidence of MDR isolates in the E coli group. Furthermore, many of the patients in the E coli group had coexisting biliary tract infection, and may have been taking antibiotics to treat their conditions before admission. However, information regarding antibiotic use among patients prior to admission had not been noted in the medical records, and therefore, a definite conclusion cannot be drawn. Further prospective studies should be conducted to clarify it. The optimal choice of initial antibiotics for patients with pyogenic liver abscess is still unknown. Johannsen et al8 recommended that the empirical initial antibiotics should be composed of ampicillin and gentamicin with/without metronidazole, especially for patients with liver abscesses of biliary origin. Chu et al6 reported that the combination of a second-generation cephalosporin and metronidazole would cover most pathogens. Our data indicate that initial antibiotic regimens should comprise a combination of 2 or 3 antimicrobial classes (eg, a secondgeneration cephalosporin and an aminoglycoside with/without metronidazole) for patients with E coli liver abscesses, given the high frequency of MDR and anaerobic bacteria isolated from those patients. Our data also show that an initial antimicrobial treatment comprising first-generation cephalosporin should cover most of the pathogens found in K pneumoniae liver abscesses. The suspected origin of abscess and underlying diseases should be considered when choosing initial antimicrobial agents before the culture results are known because it will help in predicting the most likely pathogens. For instance, in our study, E coli was the most common pathogen isolated from patients with liver abscesses of biliary THE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
origin, and from those with coexisting malignancy. The presence of biliary disorders and concomitant malignancy had a high specificity and likelihood ratio, and a modest positive predictive value for the prediction of E coli as the etiology of liver abscess. In contrast, K pneumoniae was the most prevalent pathogen isolated from patients with liver abscesses of cryptogenic origin, and from those with concomitant diabetes mellitus. The presence of diabetes mellitus and liver abscesses of cryptogenic origin had a good specificity and positive predictive value, and a modest likelihood ratio for the prediction of K pneumoniae as the etiology of liver abscess. The effectiveness of such initial antibiotics strategies needs to be evaluated further. The incidence of metastatic infection from K pneumoniae liver abscess was 14% in our study, which is consistent with the incidence rate of metastatic infection from K pneumoniae liver abscess reported previously.9,10,13,29 Although very few reports of the development of outspreading infection in patients with E coli liver abscess have been published, the incidence in our study was 9%. The most common outspreading infection from liver abscess in the E coli group occurred in the abdominal region. In the K pneumoniae group, outspreading infection from liver abscess tended to occur at more distant locations, such as eyes, meninges, lungs, and bones. Hematogenous seeding has been reported to be a possible route of outspreading infection from pyogenic liver abscesses, especially those caused by K pneumoniae.9,10,29 We suggest that contiguous spreading might be an alternative route of infectious spread from pyogenic liver abscess, especially those caused by E coli. This may explain why the spread of infections from E coli liver abscesses were more likely to the abdominal region than to distant sites. In addition, contiguous infections in patients with E coli liver abscesses are often in the abdominal region; it is feasible that these foci can be detected early using abdominal US/CT scan during hospitalization. This may explain why there were no deaths among patients with contiguous infection in the E coli group. Some investigators17,18 have suggested that the decreased mortality rates for pyogenic liver abscess in recent years may be attributed, in part, to a change in the predominant causative pathogens, especially the emergence of K pneumoniae. After a head-to-head comparison of the 2 groups, we found no statistical difference in mortality rate between patients in the E coli and those in the K pneumoniae groups after adjusting for potential confounders. Many studies have attempted to define the risk factors for death among patients with pyogenic live abscesses.3,6,8,10,12,24 However, no general consensus has been reached regarding risk factors. In the present study, the differences in mortality predictors between the 2 groups were not organism-specific; instead, the differences appeared to be due to the heterogeneity of the 2 groups. Variables 103
E coli and K pneumoniae Liver Abscesses
such as high APACHE scores, uremia, MDR isolates, and malignancy are seen in a variety of infectious conditions. Interestingly, right-lobar abscess involvement was found herein to be a protective factor in the E coli group. Right-sided liver abscess is relatively easy to locate and drain. This is because the drainage route can be easily planned to avoid major vascular and biliary structures. In the present study, patients in the E coli group had a longer time to defervesce after admission than those in the K pneumoniae group despite the fact that no difference in mortality rate was noted between the both groups. This finding was similar to that reported in our prior study which compared pyogenic liver abscesses of biliary origin with those of cryptogenic origin.13 This finding implies that the long period for defervesce may be related to coexisting biliary tract infection rather than to liver abscess per se. One of the limitations of our study was its retrospective design. Although bivariate analyses did not detect differences among some of the covariates, that statistical method may have been insufficient at detecting differences in covariate values between the 2 groups of patients. Another limitation of our study was the sample size of patients; however, the fact that pyogenic liver abscess is relatively rare makes it difficult to collect a large sample during a finite study period. Hence, we used the exact method instead of the asymptotic method for logistic regression, because inference from the exact method is more reliable and valid in such a situation.22 Additionally, to compare outcome between patients with E coli and those with K pneumoniae, we adjusted for baseline covariates and the significant clinical variables. Finally, we must draw the readers’ attention to the wide confidence intervals for many of the variables, and point out that such intervals do not rule out important differences between groups. In conclusion, our data suggest that the 2 leading causative pathogens of pyogenic liver abscess, E coli and K pneumoniae, are dissimilar in many ways, including demography, underlying diseases, origin of abscess, microbiological characteristics, and clinical response. Moreover, the higher mortality rate associated with E coli liver abscess may be attributed to the severity of illness at admission and high frequency of underlying malignancy rather than to the pathogen itself. Knowledge of the underlying diseases and the origin of liver abscess is useful for predicting the microbial etiology of the liver abscess before the culture results are known. Clinicians should begin with broad antibiotic coverage such as a second-generation cephalosporin and an aminoglycoside with metronidazole when treating liver abscesses given the high frequency of MDR isolates seen among E coli isolates. 104
Acknowledgments We would like to thank Professors H.S. Lee, PhD, and R.H. Wong, PhD, School of Public Health, Chung Shan Medical University, for their expert assistance in statistical analysis. References 1. Gyorffy EJ, Frey CF, Silva J Jr. Pyogenic liver abscess: diagnostic and therapeutic strategies. Ann Surg 1987;206: 699–705. 2. Cohen JL, Martin FM, Rossi RL, et al. Liver abscess: the need for complete gastrointestinal evaluation. Arch Surg 1989;124:561–4. 3. Frey CF, Zhu Y, Suzuki M, et al. Liver abscesses. Surg Clin N Am 1989;69:259–71. 4. Branum GD, Tyson GS, Branun MA, et al. Hepatic abscess: changes in etiology, diagnosis, and management. Ann Surg 1990;212:655–62. 5. Seeto RK, Rockey DC. Pyogenic liver abscess: changes in etiology, management, and outcome. Medicine 1996;75:99–113. 6. Chu KM, Fan ST, Lai ECS, et al. Pyogenic liver abscess: an audit of experience over the past decade. Arch Surg 1996;131: 148–52. 7. Yu SC, Lo RH, Kan PS, et al. Pyogenic liver abscess: treatment with needle aspiration. Clin Radiol 1997;52:912–6. 8. Johannsen EC, Sifri CD, Madoff LC. Pyogenic liver abscesses. Infect Dis Clin N Am 2000;14:547–63. 9. Chang FY, Chou MY. Comparison of pyogenic liver abscesses caused by Klebsiella pneumoniae and non-K pneumoniae pathogens. J Formos Med Assoc 1995;94:232–7. 10. Wang JH, Liu YC, Lee SSJ, et al. Primary liver abscess due to Klebsiella pneumoniae in Taiwan. Clin Infect Dis 1998;26: 1434–8. 11. Wiwanitkit V, Suwansaksri N, Suwansaksri J. Causative agents of liver abscess in those with liver cirrhosis: a 10-year case review of hospitalized patients in Thailand. Ann Trop Med Parasitol 2002;96:513–6. 12. Lee KT, Wong SR, Sheen PC. Pyogenic liver abscess: an audit of 10 years’ experience and analysis of risk factors. Digest Surg 2001;18:459–66. 13. Chen SC, Yen CH, Tsao SM, et al. Comparison of pyogenic liver abscesses of biliary and cryptogenic origin: an eight-year analysis in a University Hospital. Swiss Med Wkly 2005;135: 344–51. 14. Moulds-Merritt C, Frazee RC. Therapeutic approach to hepatic abscesses. South Med J 1994;87:884–8. 15. Huang CJ, Pitt HA, Lipsett PA, et al. Pyogenic hepatic abscess: changing trends over 42 years. Ann Surg 1996;223: 600–9. 16. Rahimian J, Wilson T, Oram V, et al. Pyogenic liver abscess: recent trends in etiology and mortality. Clin Infect Dis 2004;39:1654–9. 17. Lederman ER, Crum NF. Pyogenic liver abscess with a focus on Klebsiella pneumoniae as a primary pathogen: an emerging disease with unique clinical characteristics. Am J Gastroenterol 2005;100:322–31. 18. Lederman ER, Crum NF. Klebsiella pneumoniae liver abscess: a coast-to-coast phenomenon. Clin Infect Dis 2005;41:273. 19. National Committee for Clinical Laboratory Standards. 2000. Performance standards for antimicrobial susceptibility testing. Approved standard M100-S10. National Committee for Clinical Laboratory Standards, Wayne, Pa. 20. National Committee for Clinical Laboratory Standards. 2004. Performance standards for antimicrobial susceptibility test-
August 2007 Volume 334 Number 2
Chen et al
21.
22. 23.
24.
25.
ing. Approved standard M100-S14. National Committee for Clinical Laboratory Standards, Wayne, PA. Knaus WA, Draper EA, Wagner DP, et al. APACHE II: a severity of disease classification system. Crit Care Med 1985; 13:818–29. Mehta CR, Patel NR. Exact logistic regression: theory and examples. Stat Med 1995;14:2143–60. Monsen AH, Green ST, Read RC, et al. Liver abscess in adults: ten years experience in an UK centre. QJM 2002;95: 797–802. Alvarez Perez JA, Gonzalez JJ, Baldonedo RF, et al. Clinical course, treatment, and multivariate analysis of risk factors for pyogenic liver abscess. Am J Surg 2001;181:177–86. Cheng HP, Siu LK, Chang FY. Extended-spectrum cephalosporin compared to cefazolin for treatment of Klebsiella pneumoniae-caused liver abscess. Antimicrob Agents Chemother 2003;47:2088–92.
THE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
26. Carratala J, Fernandez-Sevilla A, Tubau F, et al. Emergence of quinolone-resistant Escherichia coli bacteremia in neutropenic patients with cancer who have received prophylactic norfloxacin. Clin Infect Dis 1995;20:557–60. 27. Perea S, Hidalgo M, Arcediano A, et al. Incidence and clinical impact of fluoroquinolone-resistant Escherichia coli in the fecal flora of cancer patients treated with high dose chemotherapy and ciprofloxacin prophylaxis. J Antimicrob Chemother 1999;44:117–20. 28. Lautenbach E, Fishman NO, Bilker WB, et al. Risk factors for fluoroquinolone resistance in nosocomial Escherichia coli and Klebsiella pneumoniae infections. Arch Intern Med 2002;162:2469–77. 29. Cheng DL, Liu YC, Yen MY, et al. Septic metastatic lesions of pyogenic liver abscess: their association with Klebsiella pneumoniae bacteremia in diabetic patients. Arch Intern Med 1991;151:1557–9.
105