1. Introduction
Pyogenic liver abscess (PLA), the most common type of visceral abscess, is a potentially life-threatening condition, with a mortality rate of up to 70% [
1,
2]. Although the advances in diagnostic techniques and therapeutic modalities have improved patient prognosis, the mortality rate remains within the range of 4–20% [
3,
4,
5]. Previous studies have reported advanced age, the presence of a malignancy, biliary origin and high Acute Physiology and Chronic Health Evaluation II (APACHE II) scores as factors associated with a poor prognosis [
4,
5,
6]. Most of these factors are related to host or clinical conditions rather than the pathogen. Liver abscesses may be caused by bacterial, fungal, or parasitic organisms. Prior to the 1980s, the most frequently isolated organism was
Escherichia coli (E. coli) [
7]. Since the first report of a liver abscess caused by
Klebsiella pneumoniae (K. pneumoniae) leading to metastatic complications, such as endophthalmitis in Taiwan [
8], a shift from
E.coli to
K. pneumoniae has occurred in Asia and some Western countries [
9]. In general, antibiotic therapy composed of either a third-generation cephalosporin plus metronidazole or piperacillin/tazobactam has been empirically recommended for the treatment of PLA [
10]. However, antibiotic resistance has become a major global public health problem, and some cases of PLA caused by extended-spectrum beta-lactamase (ESBL)-producing Enterobacteriaceae have been reported [
11,
12]. Multidrug-resistant organisms (MDROs) are defined as those with acquired nonsusceptibility to at least one agent in three or more antibiotic categories. The main types of MDROs include vancomycin-resistant enterococci (VRE), methicillin-resistant
Staphylococcus aureus (MRSA), ESBL-producing Enterobacteriaceae and carbapenemase-producing Enterobacteriaceae. The emergence and rise of MDROs have become a major global public health concern [
13]. Recently, Lo et al. reported an increasing trend in the prevalence of ESBL-producing
K. pneumoniae in a series of Asian patients with PLA [
14]. However, they did not investigate risk factors for ESBL-producing
K. pneumoniae or the effect of MDROs on clinical outcomes. In Western countries, Mücke et al. focused on Enterococcus species as the main cause of PLA and reported the risk factors associated with VRE in a relatively small study population (86 patients in total) [
15].
In this study, we investigated the clinical features and outcomes of PLA by reviewing clinical data collected over a 10-year period from patients in a large tertiary reference center in Korea. The primary endpoint was the identification of the risk factors for MDRO-induced PLA, and the secondary endpoint was the identification of the factors associated with in-hospital mortality.
2. Materials and Methods
2.1. Ethical Approval
Informed consent was waived because of the retrospective nature of the study and the analysis used anonymous clinical data, and this was approved by the ethics committee of Hallym University Medical Center. The study was conducted with the approval of the Institutional Review Board of Hallym University Medical Center (2019-06-007-001). All methods were carried out in accordance with relevant guidelines and regulations.
2.2. Study Population
This retrospective study was conducted in Korea at Hallym University Medical Center, which consists of six tertiary teaching hospitals, between January 2008 and December 2017. To identify suitable patients, the medical records of Hallym University Medical Center were systematically searched for code K75.0 of the International Classification of Disease, Tenth Revision. Patients with the following characteristics were included: (1) ultrasound (US), computed tomography (CT), or magnetic resonance imaging (MRI) findings compatible with a liver abscess and (2) either positive culture results from material retrieved from the abscess or the resolution of symptoms after antibiotic therapy. Patients with the following characteristics were excluded: (1) age < 18 years old and (2) parasitic/fungal/amoebic abscesses.
2.3. Clinical Data Collection and Definition
The medical records were systemically reviewed. Clinical information including sex, age, underlying medical conditions, initial symptoms, recent exposure to antibiotics and use of health care facilities within the 12 months before admission was gathered. Patients with a clinical picture indicative of cholecystitis, cholangitis, or documented bile duct disease were considered to have PLA secondary to biliary tract disease. Cryptogenic abscesses were diagnosed when there was no obvious source of infection after appropriate investigations. Initial laboratory values were defined as the values obtained on the first day of hospital admission or within 24 h after the clinical diagnosis of PLA, if PLA was not the initial cause of hospitalization. Blood cultures were performed for all patients. If patients were treated with percutaneous drainage or aspiration of the abscess, pus cultures were performed with the collected material. The percutaneous aspiration of the abscess cavity was performed with an 18- to 20-gauge needle under ultrasound guidance. For catheter drainage, catheters ranging in size from 7 to 12 French were introduced into the abscess cavity using the Seldinger technique. The abscess cavity was punctured with a sharp hollow needle under ultrasound guidance and fluoroscopy was used to confirm the position of the catheter. MDROs were considered present when strains showed nonsusceptibility to at least one agent in three or more classes of antimicrobial drugs. In-hospital mortality was defined as death during hospitalization.
2.4. Statistical Analysis
The statistical analysis was performed with SPSS version 24 (IBM Co., Armonk, NY, USA). Quantitative variables are expressed as the means ± standard deviations (SDs). A Student’s t-test was used to analyze continuous variables, and Fisher’s exact or Pearson’s chi-square tests were used to compare categorical variables. Factors that were significant in the univariate analysis were entered into a stepwise multivariate analysis to determine the independent risk factors before propensity score matching (PSM). Given the differences in baseline characteristics between patients with MDRO- and non-MDRO-induced PLA, PSM was used to enable the comparison of the variables of interest by matching participants based on confounding variables, such as age and comorbidities. Before PSM, our data showed a statistically significant difference in ‘age’ and ‘malignancy’ between MDRO and non-MDRO groups. The standardized difference was found to be greater than 0.1 in both variables; ‘age’ and ‘malignancy’. Therefore, we included ‘age’ and ‘malignancy’ in estimating the propensity score. Matching was performed in a 1:3 using the optimal matching method. After PSM, univariate analysis was performed to identify the factors associated with MDROs using Pearson’s chi-square or Fisher’s exact test analysis of variance. Variables significantly associated with MDROs in the univariate analysis were included in the multivariate logistic regression analysis, which was performed to identify the independent risk factors. Odds ratios and their 95% confidence intervals were calculated. All p-values < 0.05 were considered statistically significant and all p-values were two-tailed. Predictive factors for in-hospital mortality were analyzed with uni- and multivariate logistic regression models. Variables with p < 0.05 in the univariate model were entered into the multivariate stepwise logistic regression model (forward selection, likelihood ratio) to assess the independent predictive effect of the variables of interest.
4. Discussion
In this study, we found that 6.6% (55/833) of the patients had PLAs caused by MDROs. ESBL- producing
K. pneumoniae and
E.coli accounted for 76.4% of the MDROs. Previously, Shi et al. found that among 817 patients with PLAs assessed from 2005 to 2015, 8.2% had PLAs caused by ESBL-producing Enterobacteriaceae isolates [
16]. Our study showed that the number of PLA cases increased from 2008 to 2017 but that infection with MDROs has not increased in Korea. The incidence of ESBL-producing bacteria in our data appeared to be lower than that reported in a previous study (5% vs. 8.2%) [
16]. In Western studies, the proportion of PLAs caused by antimicrobial-resistant organisms was reported to be 24.4–34.1% [
15,
17]. A recent study performed in the USA reported that there were no cases of infection with ESBL-producing
K. pneumoniae [
18]. These differences could be because microbes and resistance patterns show both regional and temporal variations.
In the present study, a previous history of hepatobiliary procedures, recent exposure to antibiotics and elevated ALP levels were risk factors for PLAs caused by MDROs. Due to a current lack of research focused on the difference between PLAs caused by MDROs and those not caused by MDROs, the characteristics of PLAs caused by MDROs have not been well described. Shi et al. found that the presence of biliary disorders, a history of treatment for malignancy, pulmonary infection at admission, and the absence of DM were associated with the occurrence of PLAs caused by ESBL-producing Enterobacteriaceae [
16]. They reported that the mortality rates in patients with PLA caused by ESBL-producing Enterobacteriaceae were significantly higher than those in patients with PLA caused by non-ESBL-producing Enterobacteriaceae. However, the ESBL-producing group was older and more likely to have a history of treatment for malignancies than the non-ESBL-producing group. Similarly, our data before PSM showed that history of malignancies and advanced age were more common in the MDRO group. Advanced age and history of malignancy are already known factors predictive of a poor prognosis of PLAs and could influence mortality. Therefore, in our study, to correct this bias, PSM analysis was performed. In the present study, there was no relationship between MDRO-induced PLAs and abscess-related complications or DM. Instead, elevated ALP levels were found to be associated with MDRO-induced PLA. The serum ALP level is frequently elevated in a variety of liver diseases. A great many tissues, including the liver, the bones, the intestinal mucosa, and the placenta are particularly rich sources of the enzyme [
19]. The increase in serum ALP levels is observed during cholestasis with increased bile acid concentration, due either to liver parenchymal damage or to biliary obstruction Abnormally high ALP level is considered the most reliable and consistent biochemical indicator of PLA. Although in our study, the ALP level had no relationship with the abscess size, previous studies reported that the ALP level was correlated with the abscess volume and disease duration [
20,
21]. According to a commonly held view, the reason for the prominent elevation in the level of ALP in the MDRO group was assumed to be the fact that the major cause of MDRO-induced PLAs was biliary disease. However, when we analyzed the clinical characteristics of MDRO-induced PLA only in patients with non-biliary causes, a significant increase in serum ALP levels was observed in the MDRO group (285.7 ± 175.9 vs. 197.2 ± 137.7,
p = 0.005). Therefore, ALP elevation is likely to be related to the unrevealed pathologic condition of MDRO-induced PLA, further research on the mechanism of ALP elevation is required.
In the PSM analysis, the overall proportions of patients with malignancies were not different between the MDRO and non-MDRO groups. Although the presence of hepatopancreatobiliary malignancy was not a risk factor for MDRO-induced PLAs, empirical antibiotic treatment for an infection of unknown etiology, such as neutropenic fever following systemic chemotherapy or malignant biliary obstruction requiring repeated hepatobiliary procedures, might be related to the emergence of MDROs. In the present study, PLAs caused by MDROs had a significant relationship with prior hepatobiliary procedures, such as percutaneous transhepatic gallbladder drainage and/or ERCP. Sand et al. reported that ERCP, especially when performed with endoscopic sphincterotomy, may result in long-term bacterobilia [
22]. Bacterobilia is an important factor in the pathogenesis of recurrent cholangitis and may be associated with gallstone formation, acute pancreatitis and cholangiocarcinogenesis [
23,
24,
25]. Ultimately, the repeated administration of antibiotics to treat recurrent biliary tract infections can precipitate the emergence of MDROs by applying selective pressure.
The overall in-hospital mortality rate was 4.1%, which was slightly lower than previously reported rates (5–13%) [
14,
16,
26,
27], and the presence of MDRO alone was not an independent factor predictive of in-hospital mortality. In the present study, the initial administration of antibiotics that were later determined to be unsuitable based on the results of antibiotic susceptibility testing appeared to be associated with in-hospital mortality. Although the presence of MDROs was not associated with in-hospital mortality, inadequate initial antibiotic treatment was prescribed to a large portion of the patients with MDRO-induced PLAs. All patients except for 42 patients given carbapenem from the beginning received empirical antibiotics generally recommended for PLA treatment. In 42 patients given carbapenem as initial antibiotic treatment, only four patients were confirmed to be infected by ESBL-producing organisms, and twenty-two patients (52.4%, 22/42) were admitted to the ICU. This finding reflects that in clinical practice, admission to the ICU could be the main factor affecting the selection of carbapenem as an empirical antibiotic treatment. However, this is an inappropriate criterion. Therefore, accurate MDRO risk factor assessment is necessary to guide the selection of empirical antibiotics for the treatment of PLA.
This study had some limitations. First, this was a retrospective study based on the review of medical records. Therefore, previous antibiotic exposure and the use of prior healthcare facilities may have been underestimated in our study, and the patients lost to follow-up after hospital discharge may have affected the recurrence rate of PLA. Nevertheless, recent exposure to antibiotics was verified to be a factor predictive of MDRO-induced PLA, and the rate of recurrence of PLA also appeared to be similar to that in a previous study (4.5% vs. 5.3%) [
28]. Second, the information was lacking: the types or duration of antibiotics or healthcare facilities used before admission, so a definite conclusion about these factors cannot be made. The information collected from medical records may inevitably have insufficiency that could impact the validity and attenuate the findings. To minimize bias and overcome the shortcomings of the retrospective study, we used PSM and subgroup analyses and found MDRO-related risk factors and prognostic factors associated with in-hospital mortality.
This was a large study involving more than 800 patients with PLA and one of the few studies to report the incidence and characteristics of MDRO-induced PLA. Although the present study sample is not representative of the distribution of patients with MDRO-induced PLAs worldwide, our data may be helpful for identifying patients in Asia who are at high risk of developing MDRO-induced PLA due to infection with the two leading causative pathogens, namely, K. pneumoniae and E. coli. In the future, a prospective study will be needed to discover more specific risk factors and create a MDRO-predictive model using risk factors in patients with PLA.