Current Management of Pyogenic Liver Abscess: Surgery is Now Second-Line Treatment James J Mezhir, MD, Yuman Fong, MD, FACS, Lindsay M Jacks, MS, George I Getrajdman, MD, Lynn A Brody, MD, Ann M Covey, MD, Raymond H Thornton, MD, William R Jarnagin, MD, FACS, Stephen B Solomon, MD, Karen T Brown, MD The objective of this study was to examine the current treatment for liver abscess and to assess the factors associated with failure of percutaneous drainage. STUDY DESIGN: Records of 58 patients with pyogenic hepatic abscess, from 1998 to 2009, were examined. Clinicopathologic variables were analyzed as predictors of failure of percutaneous drainage using multivariable logistic regression. The results of surgical intervention after failure of percutaneous treatment were also examined. RESULTS: Fifty-one patients (88%) had a history of malignancy including pancreas (36%), cholangiocarcinoma (17%), colon (12%), and gallbladder (10%). Recent hepatic artery embolization or radiofrequency ablation preceded development of abscess in 13 patients (22%). Fifteen patients (26%) had evidence of biliary tract communication, and 14 of 15 (93%) of these patients had concomitant biliary tract obstruction. Percutaneous drainage was successful in 38 patients (66%) with a median drain dwell time of 26 days (range 3 to 319 days). Five patients (9%) required operative intervention and 2 of these patients (3% overall) died postoperatively from septic complications. Fifteen patients (26%) died with percutaneous drains in place; 9 (60%) of these patients died of cancer progression without evidence of sepsis. Independent predictors of failure of percutaneous drainage included abscesses containing yeast (p ⫽ 0.003) and communication of the abscess cavity with the biliary tree (p ⫽ 0.02). CONCLUSIONS: Pyogenic hepatic abscess was treated successfully in the majority of patients with advanced malignancy, although mortality remained high. The presence of yeast and communication with an untreated obstructed biliary tree were associated with failure of percutaneous drainage. The need for surgical salvage was associated with a high mortality. (J Am Coll Surg 2010;210: 975–983. © 2010 by the American College of Surgeons) BACKGROUND: tients with larger and multiloculated abscesses had better clinical outcomes with surgery compared with PD.6 This study analyzes the results of treatment of liver abscess both in the context of cryptogenic disease and in patients with actively treated advanced malignancy. Patients with malignancy are often malnourished, are being treated with chemotherapy, or have suboptimal performance status precluding major surgery. Furthermore, these patients may be undergoing liver-directed therapies such as hepatic artery embolization, radiofrequency ablation, and resection. The percutaneous approach in these patients, if effective, would be the preferred method of treatment.11,12 A better understanding of which clinicopathologic factors predict failure in this population could be beneficial in guiding therapy. Treatment of pyogenic liver abscess (PLA) has evolved over the last 10 years such that percutaneous drainage (PD) is now the most common approach,1-5 although some still advocate surgical drainage as more effective.6-10 Proponents of early operative intervention generally reported on abscesses related to benign disease, and contended that pa- Disclosure Information: Nothing to disclose. Presented at the American College of Surgeons 95th Annual Clinical Congress, Chicago, IL, October 2009. Received November 13, 2009; Revised February 26, 2010; Accepted March 3, 2010. From the Departments of Surgery (Mezhir, Fong, Jarnagin), Interventional Radiology (Getrajdman, Brody, Covey, Thornton, Solomon, Brown), and Epidemiology and Biostatistics (Jacks), Memorial Sloan-Kettering Cancer Center, New York, NY. Correspondence address: Yuman Fong, MD, FACS, Department of Surgery, Memorial Sloan-Kettering Cancer Center, 1275 York Ave, New York, New York 10065. email: [email protected] © 2010 by the American College of Surgeons Published by Elsevier Inc. METHODS The Memorial Sloan-Kettering Cancer Center interventional radiology and hepatopancreaticobiliary databases 975 ISSN 1072-7515/10/$36.00 doi:10.1016/j.jamcollsurg.2010.03.004 976 Mezhir et al Management of Pyogenic Liver Abscess Abbreviations and Acronyms OR ⫽ odds ratio PD ⫽ percutaneous drainage PLA ⫽ pyogenic liver abscess were queried for patients diagnosed with PLA. Patients met criteria for inclusion based on 2 specific factors: clinical signs of infection such as leukocytosis, fever, bacteremia, or sepsis; and imaging consistent with PLA. All imaging was retrospectively reviewed to confirm PLA; any patient with an extrahepatic abscess or perihepatic fluid collection was excluded. There were 58 patients who met criteria for analysis, and complete data were collected retrospectively from the medical record. All patients were treated with broad-spectrum antibiotics during the initial drainage procedure. The most common antibiotics administered during this interval were ticarcillin/clavulanate (36%), piperacillin/tazobactam (31%), or a fluoroquinolone (21%) with or without metronidazole (22%). Vancomycin was given alone or in combination with other gram-negative and anaerobic coverage in 9 patients (16%). Antibiotic therapy was subsequently tailored based on culture and sensitivity results. For patients with yeast in their culture, treatment with intravenous fluconazole was initiated. If a resistant organism was found, treatment with a ␤-glucan synthase inhibitor such as caspofungin or micafungin was initiated. Patients routinely remained on antibiotic therapy until their sepsis was resolved and there was no clinical evidence of infection (afebrile, no leukocytosis, negative blood cultures). The initial treatment in all patients was with PD and was performed using standard Seldinger technique; 6.5- to 12F locking, pigtail-type drains were placed under CT, ultrasound, and/or fluoroscopic guidance at the discretion of the attending interventional radiologist performing the procedure. All patients received preprocedure antibiotics. Patients with platelet count ⬍50,000/␮L and/or international normalized ratio (INR) ⬎1.9 were typically transfused with appropriate products immediately before drainage. All catheters were connected to gravity drainage and flushed twice daily with 3 to 10 mL of saline. Success of PD was defined as having clinical or radiologic evidence of abscess resolution followed by drain removal. Drain removal was typically performed when the patient demonstrated no signs of infection (fever, leukocytosis, bacteremia) and drainage was nonbilious and less than 30 mL over a 24- to 48-hour period. Drain exchange was reserved for patients who had a decrease in drainage with clinical signs of infection or persistent abscess on imaging. Failure of PD was defined as requiring operative J Am Coll Surg intervention, or death with abscess drains in place (including patients who died of disease unrelated to septic complications). Patients requiring operative intervention either demonstrated no improvement in clinical symptoms with initial drainage or had persistent abscesses despite multiple drainage attempts. Radiologic characteristics Examples of liver abscesses in this study are shown in Figure 1. Abscesses are solitary or multiple well-defined lesions, and may be round, oval, or lobulated. On CT, liver abscesses appear isodense with normal liver parenchyma on unenhanced scans, and with low attenuation on contrastenhanced equilibrium phase images. They may be complex and multiloculated (Fig. 1A) or unilocular (Fig. 1B) with or without an air-fluid level. Abscesses typically have an enhancing rim or peripheral capsule (Fig. 1B). On MRI, they exhibit a low signal on T1-weighted images and appear hyperintense on T2 (Fig. 1C). They are often surrounded by a local increase in signal on T2-weighted images; this “halo” is secondary to edema. Most have peripheral enhancement after the intravenous administration of gadolinium. Biliary fistula, when present, can be demonstrated by fluoroscopic contrast injection study (Fig. 1D). Statistical analysis Characteristics of the cohort were summarized. Linear regression was used to determine predictors of prolonged dwell time for patients treated successfully with PD alone. Dwell time was modeled using a log transformation due to the skewed nature of the distribution. Multivariable logistic regression was used to determine predictors of failure of PD defined as either death or requiring operative debridement. Variables statistically significant (p ⬍ 0.05) on univariate analysis were considered in a multivariable model. All statistical analyses were performed using SAS 9.2 (SAS Institute). RESULTS From January 1998 until May 2009, there were 58 patients with PLA treated with PD. During this time period, 2,738 hepatic resections were performed, and 5 (0.2%) were for treatment of PLA. Clinicopathologic variables are listed in Table 1. The vast majority (88%) of patients in the study had a history of metastatic or primary adenocarcinoma or neuroendocrine neoplasm of the pancreas, bile ducts, duodenum, or liver. There were 17 patients (29%) receiving chemotherapy at the time of their abscess diagnosis; 8 patients (14%) had diabetes mellitus. Hepatic metastases were present in 45% of patients. Nine patients (16%) had a previous liver resection. Among the 7 patients without a Vol. 210, No. 6, June 2010 Mezhir et al Management of Pyogenic Liver Abscess 977 Figure 1. Radiographic presentation of abscesses. (A) Multiloculated abscess seen on contrast-enhanced CT scan. Note extensive loculation; a classic appearance for a liver abscess. (B) Patient with indwelling wall stent (short arrow) for cholangiocarcinoma presented with fever, cholangitis; unilocular liver abscess with air fluid level (long arrow). (C) MRI in patient with acinar cell pancreatic cancer with liver metastases who developed fever and bacteremia. Note 2 hyperintense liver abscesses in the left lateral segment and segment IV (short arrows) on T2-weighted image adjacent to known liver metastasis (long arrow) in right liver. The abscesses were new findings since the previous scan 2 months earlier. (D) Contrast study revealing communication of abscess cavity (short arrow) with biliary tree. diagnosis of cancer, 3 had a previous hepaticojejunostomy for bile duct injury, and 1 had resection of a giant cavernous hemangioma. Three of these 7 patients had biliary obstruction that preceded the development of the hepatic abscess, and 4 abscesses were cryptogenic. Presenting characteristics and diagnostic and therapeutic measures are shown in Table 2. For the entire cohort, median white blood cell count (14,200/␮L, normal range 4,000 to 11,000/␮L) and total bilirubin levels (1.3 mg/dL, normal ⬍1.0 mg/dL) were elevated. Median creatinine and international normalized ratio (INR) were within normal limits. The median albumin level was low (2.6 mg/dL, normal ⱖ 4.0 mg/dL). Most patients (72%) had only 1 liver abscess and the most common method for abscess diagnosis was CT scan (97%). Consistent with earlier reports, the right hepatic lobe was the most common location for abscess (91%).3,6 Communication with the biliary tract was determined by contrast study and was noted in 15 patients (26%). Among the patients with biliary tract communication, 14 of 15 (93%) had concomitant biliary tract obstruction. Seven of these patients underwent treatment for biliary obstruction with either percutaneous or endoscopic drainage. A secondary procedure was defined as requiring an additional interventional procedure to reposition or replace the initial drain or to study the abscess cavity; more than 1 procedure was required in 38 patients (66%). 978 Mezhir et al Management of Pyogenic Liver Abscess Table 1. Clinicopathologic Variables of 58 Patients with Pyogenic Hepatic Abscess Variable Age ⱖ 60 y Gender (male) History of malignancy Pancreatic/duodenal adenocarcinoma Cholangiocarcinoma Colon adenocarcinoma Pancreatic/duodenal neuroendocrine tumor Gallbladder adenocarcinoma Other Current cancer therapy Chemotherapy Octreotide Radiation Diabetes mellitus Liver disease Hepatic metastases Previous liver resection Portal vein thrombosis Bilioenteric anastomosis Pancreaticoduodenectomy Hepaticojejunostomy Choledochojejunostomy or duodenostomy Cholecystojejunostomy Previous biliary intervention ERCP with plastic endobiliary stent PTC with internal/external drain ERCP or PTC with metal stent Previous hepatic artery embolization Previous radiofrequency ablation n % 30 33 51 15 10 7 7 6 6 23* 17 5 2 8 33 26 9 7 30 14 11 4 1 17 7 6 4 12 1 52 57 88 30 20 14 14 12 20 40 74 22 9 14 57 45 16 12 52 47 37 13 3 29 41 35 24 21 2 *One patient was being treated with chemotherapy and radiation. ERCP, endoscopic retrograde cholangiopancreatography; PTC, percutaneous transhepatic cholangiogram. Microbiology Initial drainage specimens for all patients were sent for Gram stain and culture; 3 abscesses (5%), despite aspiration of purulent material with positive Gram stains, did not grow bacteria on final culture (Table 3). Eight patients (14%) had negative Gram stains, but subsequent positive cultures. Culture growth was polymicrobial in 32 patients (55%) and the most common bacterial species cultured was Streptococcus (34%) followed by Enterococcus (22%) and Escherichia coli (22%). Candida was the only yeast that grew from cultures, and included albicans (3 patients), glabrata (5 patients), albicans and glabrata (1 patient), and krusei (1 patient). Blood cultures were positive in 27 patients (47%) and in each instance had bacterial growth of at least 1 of the species from the corresponding abscess culture. J Am Coll Surg Table 2. Presenting Characteristics, Diagnostic and Therapeutic Measures Variable n (%) or median (range) Symptoms at presentation Fever ⫾ chills 57 (98) Leukocytosis 36 (62) Bacteremia 9 (16) Sepsis/multisystem organ failure 5 (9) White blood cell count, K/␮L 14.2 (3.9–42.8) Creatinine, mg/dL 0.9 (0.3–2.0) Albumin, mg/dL 2.6 (1.5–3.8) International normalized ratio 1.3 (1.0–2.9) Total bilirubin, mg/dL 1.3 (0.2–19.1) Method of diagnosis CT scan 56 (97) Other (ultrasound, MRI) 2 (3) Abscesses imaged per patient, n 1 41 (71) ⱖ2 17 (29) Abscesses drained per patient, n 1 55 (95) 2 3 (5) Primary abscess location Right lobe 53 (91) Left lobe 3 (5) Central 2 (3) Loculations present 24 (41) Air in abscess 27 (47) Abscess size, maximum diameter, cm 5.7 (2.3–1.6) ⬍5 23 (40) 5–10 28 (48) ⬎10 7 (12) Initial amount of fluid drained from abscess, mL 35 (3.0–1,300) Communication of abscess with biliary tract 15 (26) Method of drainage CT-guided drain placement 49 (84) Fluoroscopy or ultrasound 9 (16) All purpose drainage catheter used 12F 4 (7) 10F 5 (9) ⱕ 8.5F 49 (84) Procedures required per patient, n 1 20 (34) 2 13 (22) 3 7 (12) 4 12 (21) ⱖ5 6 (10) Drain dwell time, d 32 (2–319) Initial hospital stay, d 13 (2–115) ICU stay during initial drainage, d 6 (10) Vol. 210, No. 6, June 2010 Mezhir et al Management of Pyogenic Liver Abscess 979 Outcomes ther therapy, either percutaneous or operative. In these 6 patients, care was withdrawn and they subsequently died of sepsis. One patient presented with liver failure and sepsis and later died from complications of a perforated duodenal ulcer. As shown in Table 5, the 2 factors that predicted failure of PD were growth of yeast in abscess culture (odds ratio [OR] 14.76; CI 2.49 to 87.30; p ⫽ 0.003) and communication of the abscess cavity with the biliary tree (OR 5.51; CI 1.38 to 21.96; p ⫽ 0.02). On univariate analysis, there was a trend toward significance for current cancer therapy (p ⫽ 0.09) and higher serum bilirubin levels (p ⫽ 0.07). To determine factors associated with failure of PD resulting in operation (5 events) or death (15 events), a separate multivariable analysis was performed for each of these endpoints. The results were similar in that yeast in culture and communication with the biliary tree were the only independent predictors of failure (data not shown). Thirty-eight patients (66%) were treated successfully with PD alone (defined as abscess resolution with drain removal). Median drain dwell time for successfully treated patients was 26 days (range 3 to 319 days). On univariate analysis, younger age was the only factor associated with a prolonged dwell time (p ⫽ 0.02); there was a trend toward significance for previous embolization or radiofrequency ablation (p ⫽ 0.07) (data not shown). Five patients (9%) required operative intervention due to either clinical deterioration or imaging that demonstrated failure of abscess resolution with PD. The median drain dwell time among these patients was 21 days (range 4 to 215 days). Initial operative intervention in all cases involved exploratory laparotomy, debridement, and placement of drains. Three patients subsequently required repeat operations; 2 required right hepatic lobectomy and 1 patient had a repeat drainage procedure. Mortality in patients requiring surgery was 40% (2 of 5 patients). The details of patients requiring operative intervention are listed in Table 4. Fifteen patients (26%) died before resolution of abscess at a median time of 51 days (range 4 to 106 days) from the time of initial drainage. The most common cause of death was progression of disease in patients with advanced malignancy (9 of 15, 60%) with controlled but unresolved abscess. Among these 9 patients, all had advanced disease, including 6 patients with multiple hepatic metastases. There were 6 patients with pancreatic cancer, 2 with gallbladder cancer, and 1 with anal squamous cell carcinoma. None of the patients had evidence of complete abscess resolution before death; however, their sepsis was controlled with PD. Of the remaining 6 patients, 5 (33%) developed signs of recurrent infection (leukocytosis, fever), but due to advanced cancer, were extremely poor candidates for fur- DISCUSSION Therapy for PLA continues to evolve. Early reports demonstrated high morbidity and mortality in patients managed primarily with operative intervention (drainage with or without liver resection).3,13,14 With improvements in antibiotic therapy, the incidence of PLA has decreased and the most common cause has shifted from portal seeding from appendicitis and diverticulitis to malignant biliary tract disease. The treatment paradigm has also shifted, with increased use of PD as the preferred first line management over surgical drainage.1,15-17 Some authors continue to favor open surgical drainage rather than PD. Tan and colleagues6 reported on 80 patients with PLA ⬎ 5 cm in diameter: 36 patients (45%) were managed with PD and 44 (55%) by surgical drainage. The majority (66%) of the abscesses were cryptogenic. Failure of treatment was defined as death or deterioration of clinical status requiring additional procedures. Patients who underwent surgical drainage had a shorter hospital stay and required a secondary procedure less often compared with the percutaneous group. Ten patients in the percutaneous group (28%) versus 3 in the surgical group (7%) met the criteria for failure of treatment (p ⫽ 0.01); mortality was equivalent. Hope and associates7 retrospectively stratified 107 patients with PLA into 3 types: I (small, ⬍3 cm); II (large, ⬎3 cm, unilocular), and III (large, ⬎3 cm, complex multilocular). All type I abscesses were treated successfully with antibiotics alone; type II abscesses were treated successfully with PD in 40 of 48 patients (83%). Type III abscesses were successfully treated with operative drainage in 27 of 27 patients; PD only succeeded in 8 of 24 patients (33%). Table 3. Microbiology from Final Cultures of Initial Abscess Cavity Drainage Growth from culture n % Polymicrobial Streptococcus Enterococcus Escherichia coli Klebsiella Candida Vancomycin-resistant Enterococcus Enterobacter Citrobacter Lactobacillus Bacteroides Other bacterial species Negative 32 20 13 13 12 10 9 8 6 4 3 17 3 55 34 22 22 21 17 16 14 10 7 5 30 5 980 Mezhir et al Management of Pyogenic Liver Abscess J Am Coll Surg Table 4. Details of Patients Who Required Operative Intervention Patient Age (y)/ gender 39F 54M 46M 63F 69F Primary diagnosis Pancreatic NET after Whipple with liver metastases. Abscess developed after embolization. Pancreatic NET with liver metastases. Abscess developed after embolization. Pancreatic adenocarcinoma after biliary bypass. Biliary obstruction. Liver resection for giant cavernous hemangioma. Postoperative abscess. Whipple for duodenal adenocarcinoma. Postoperative liver necrosis. Operative intervention Abscess size, cm Procedures before surgery, n Days after drainage, n Operation Outcomes 11 1 4 Persistent sepsis Incision and drainage; second operation for right hepatectomy. Died of septic shock 9 4 21 Persistent sepsis Discharged home 11 1 21 Persistent sepsis Incision and drainage; second operation for right hepatectomy. Incision and drainage 8 1 22 Persistent sepsis Died of septic shock 6 3 81 Persistent abscess Incision and drainage; second operation for repeat drainage Incision and drainage Indication Discharged home Discharged home NET, neuroendocrine tumor. Eight of the 16 failures (50%) were managed with repeat PD, 5 (31%) were managed with resection, and 3 patients (19%) died. These studies concluded that larger, loculated abscesses should be treated with operative drainage as primary treatment. Systemic antibiotic therapy alone has also been shown to be effective as primary treatment of PLA in well-selected patients with small abscesses (⬍ 3cm).7 Percutaneous aspiration is another technique that has the potential to enhance systemic antibiotic therapy by providing bacterial speciation and antimicrobial sensitivities. Giorgio and coworkers18 treated 39 patients with multiple PLA, including patients with abscesses larger than 4 cm, with percutaneous aspiration. All patients were successfully treated with aspiration alone. Yu and coauthors19 randomized 64 patients with PLA of at least 3 cm to percutaneous needle aspiration or catheter drainage. Abscess resolution, hospital stay, and mortality were not significantly different between groups. Several investigators have reported their experience with PD of pyogenic hepatic abscess.15,16 Ferraioli and colleagues15 treated 148 patients with PLA with either percutaneous (n ⫽ 104) or surgical (n ⫽ 44) drainage. The authors compared the 2 treatment approaches and found that patients treated with surgical drainage had a longer hospital stay and higher cost of care when compared with patients treated with PD; surgical drainage was also associated with higher morbidity. Notably, there were no failures among the 104 patients treated with either percutaneous needle aspiration or drainage. To determine the effectiveness of PD for the management of multiple and/or loculated abscesses, Liu and associates16 analyzed a series of 109 patients treated with PD alone. Primary PD was successful regardless of the presence of loculations or the number of abscesses treated. Many of the previously listed studies do not include patients with advanced malignancy. As demonstrated in earlier reports, patients with pyogenic hepatic abscess in the setting of malignant disease often had poor outcomes.1,11 Yeh and coworkers11 reviewed 52 patients with pyogenic hepatic abscess and underlying malignant disease. Patients with malignancy of the liver, bile duct, or pancreas were compared with patients who had cancer in other sites. Patients with abscesses in the setting of biliary tract malignancy were more likely to have a fistulous connection with the biliary tree. Furthermore, compared with patients with cancer of other sites, those with biliary tract cancer had a higher in-hospital mortality (28% vs 10%; p ⫽ 0.04) and 72% of these patients died of sepsis, disease progression, or both within 6 months of abscess diagnosis. When compared with a cohort of patients without cancer, in-hospital mortality was 21% versus 11%. The majority (88%) of the patients in our study had a history of malignancy and represent a special cohort of patients with high risk for death after diagnosis. Biliary tract communication is a known factor associated with failure of PD.20 Sugiyama and Atomi20 reported a Vol. 210, No. 6, June 2010 Mezhir et al Management of Pyogenic Liver Abscess 981 Table 5. Multivariate Analysis for Failure of Percutaneous Drainage Defined as Either Death or Requiring Operative Debridement Variable Age Gender Female Male Current cancer therapy Yes No Bilioenteric anastomosis Yes No Previous biliary instrumentation Yes No Previous embolization or RFA Yes No Abscesses, n 1 ⬎1 Abscess size White blood cells Total bilirubin Albumin Gram-negative in culture Yes No Yeast in culture No Yes Anaerobic in culture Yes No VRE in culture No Yes Loculations present Yes No Air in abscess cavity Yes No Biliary tract communication No Yes Univariate unadjusted OR (95% CI) p Value 1.01 (0.97–1.05) 0.79 1.00 0.65 (0.22–1.94) 0.44 1.00 0.38 (0.12–1.15) 0.09 1.00 1.11 (0.38–3.28) 0.85 1.00 0.66 (0.21–2.13) 0.49 1.00 0.93 (0.27–3.28) 0.91 1.00 2.15 (0.67–6.89) 0.99 (0.84–1.17) 1.04 (0.97–1.11) 1.25 (0.99–1.59) 0.36 (0.10–1.31) 0.20 0.91 0.31 0.07 0.12 1.00 0.72 (0.21–2.45) 0.60 1.00 12.00 (2.23–64.49) 0.004 1.00 0.74 (0.23–2.35) 0.60 1.00 1.65 (0.39–6.99) 0.50 1.00 3.00 (0.91–9.91) 0.07 1.00 0.60 (0.20–1.77) 0.35 1.00 4.36 (1.26–15.07) 0.02 OR, odds ratio; RFA, radiofrequency ablation; VRE, vancomycin-resistant Enterococcus. Multivariate adjusted OR (95% CI) p Value 1.00 14.76 (2.49–87.30) 0.003 1.00 5.51 (1.38–21.96) 0.02 982 Mezhir et al Management of Pyogenic Liver Abscess series of 61 patients with PLA; 20 had communication with the biliary tree. Among these 20 patients, only 3 (15%) were treated successfully with PD alone compared with 37 of 41 (90%) of patients treated without biliary tract communication. When patients with biliary tract communication had persistent discharge from their abscess drain in the presence of biliary obstruction, endoscopic stenting resulted in resolution. In the Liu and colleagues16 study previously mentioned, 19 of 104 (18%) patients had radiologic evidence of communication with the biliary tree. In patients without concomitant biliary obstruction, PD was successful in 74% of patients. In 4 patients with biliary tract communication with concomitant obstruction, PD failed in all. Two important factors relate to the finding that failure of PD is associated with communication with the biliary tree. First, the majority of patients (14 of 15, 93%) with biliary tract communication in this study also had evidence of biliary obstruction. It is possible that the presence of biliary obstruction rather than communication with the biliary tree alone led to failure of PD. This leads to the conclusion that treatment of biliary tract obstruction, whether percutaneous or endoscopic, is imperative for successful treatment of abscesses that communicate with the biliary tree. Second, an additional 15 patients (26%) in this study did not undergo a subsequent contrast study due to abscess resolution. It is possible that communication with the biliary tree was present in these patients, but not demonstrated. There are several limitations to this study. The study is retrospective in nature and the patient cohort was selected from the interventional radiology and hepatopancreaticobiliary databases, imposing selection bias for the analysis. It is possible that there were patients with small abscesses primarily treated with antibiotics alone or by percutaneous aspiration that were not included in this study. Despite having criteria for antibiotic therapy and abscess catheter removal, the heterogeneous nature of this patient cohort limits our ability to promote a specific antibiotic and drainage duration protocol. Finally, the majority of patients in this study had advanced malignancy, which limits the generalizability of our findings to patients with cryptogenic liver abscesses. We conclude that PD is safe and effective as first-line treatment for the management of PLA. Operative intervention remains second-line therapy, although surgical treatment should be considered when clinically indicated and in those who grow yeast. In addition to percutaneous abscess drainage, initial management of hepatic abscesses in patients who have biliary tract communication and concomitant biliary obstruction should include biliary drainage. J Am Coll Surg Author Contributions Study conception and design: Mezhir, Fong, Brown Acquisition of data: Mezhir, Brown Analysis and interpretation of data: Mezhir, Fong, Jacks, Getrajdman, Brody, Covey, Thornton, Jarnagin, Solomon, Brown Drafting of manuscript: Mezhir, Fong, Brown Critical revision: Mezhir, Fong, Jacks, Getrajdman, Brody, Covey, Thornton, Jarnagin, Solomon, Brown Acknowledgment: We thank Dr Ronald P DeMatteo and Dr Michael I D’Angelica for their critical review of the manuscript. REFERENCES 1. Rintoul R, O’Riordain MG, Laurenson IF, et al. Changing management of pyogenic liver abscess. Br J Surg 1996;83:1215– 1218. 2. Herbert DA, Fogel DA, Rothman J, et al. Pyogenic liver abscesses: successful non-surgical therapy. Lancet 1982;1:134– 136. 3. Huang CJ, Pitt HA, Lipsett PA, et al. Pyogenic hepatic abscess. Changing trends over 42 years. Ann Surg 1996;223:600–607; discussion 607–609. 4. Strong R, Fawcett J, Lynch S, et al. Hepatectomy for pyogenic liver abscess. HPB (Oxford) 2003;5:86–90. 5. Berger LA, Osborne DR. Treatment of pyogenic liver abscesses by percutaneous needle aspiration. Lancet 1982;1: 132–134. 6. Tan YM, Chung AY, Chow PK, et al. An appraisal of surgical and percutaneous drainage for pyogenic liver abscesses larger than 5 cm. Ann Surg 2005;241:485–490. 7. Hope WW, Vrochides DV, Newcomb WL, et al. Optimal treatment of hepatic abscess. Am Surg 2008;74:178–182. 8. Bertel CK, van Heerden JA, Sheedy PF 2nd. Treatment of pyogenic hepatic abscesses. Surgical vs percutaneous drainage. Arch Surg 1986;121:554–558. 9. Chu KM, Fan ST, Lai EC, et al. Pyogenic liver abscess. An audit of experience over the past decade. Arch Surg 1996;131:148– 152. 10. Herman P, Pugliese V, Montagnini AL, et al. Pyogenic liver abscess: the role of surgical treatment. Int Surg 1997;82:98– 101. 11. Yeh TS, Jan YY, Jeng LB, et al. Pyogenic liver abscesses in patients with malignant disease: a report of 52 cases treated at a single institution. Arch Surg 1998;133:242–245. 12. Branum GD, Tyson GS, Branum MA, et al. Hepatic abscess. Changes in etiology, diagnosis, and management. Ann Surg 1990;212:655–662. 13. Northover JM, Jones BJ, Dawson JL, et al. Difficulties in the diagnosis and management of pyogenic liver abscess. Br J Surg 1982;69:48–51. 14. Neoptolemos JP, Macpherson DS, Holm J, et al. Pyogenic liver abscess: a study of forty-four cases in two centres. Acta Chir Scand 1982;148:415–421. Vol. 210, No. 6, June 2010 15. Ferraioli G, Garlaschelli A, Zanaboni D, et al. Percutaneous and surgical treatment of pyogenic liver abscesses: observation over a 21-year period in 148 patients. Dig Liver Dis 2008;40:690– 696. 16. Liu CH, Gervais DA, Hahn PF, et al. Percutaneous hepatic abscess drainage: do multiple abscesses or multiloculated abscesses preclude drainage or affect outcome? J Vasc Interv Radiol 2009;20:1059–1065. 17. Pearce N, Knight R, Irving H, et al. Non-operative management of pyogenic liver abscess. HPB (Oxford) 2003;5:91–95. Mezhir et al Management of Pyogenic Liver Abscess 983 18. Giorgio A, de Stefano G, Di Sarno A, et al. Percutaneous needle aspiration of multiple pyogenic abscesses of the liver: 13-year single-center experience. AJR Am J Roentgenol 2006;187: 1585–1590. 19. Yu SC, Ho SS, Lau WY, et al. Treatment of pyogenic liver abscess: prospective randomized comparison of catheter drainage and needle aspiration. Hepatology 2004;39:932– 938. 20. Sugiyama M, Atomi Y. Pyogenic hepatic abscess with biliary communication. Am J Surg 2002;183:205–208.