Risk Factors of Septic Metastatic Infection among Patients with Klebsiella pneumoniae Liver Abscess in Singapore: A Case–Control Study

Klebsiella pneumoniae is the most common etiology of liver abscess in Singapore and most of Asia.^1–4 Klebsiella pneumoniae liver abscess (KPLA) complications include bacteremia, sepsis, and metastatic infection in 10% to 45% of cases.^5–7 Extremely poor clinical outcome and a high mortality rate are common among severe metastatic complications in patients with KPLA.^5,8,9

Risk factors for metastatic infection include diabetes,^10–12 bacteremia,^10,13 and thrombophlebitis.¹⁴ Other risk factors for septic metastasis include the severity of the disease, such as septic shock, and a high Acute Physiology And Chronic Health Evaluation (APACHE) score.⁵ Patients with metastatic infection often present with a low platelet count and albumin level, and high level of C-reactive protein.⁵ Recent studies indicate that the KPLA patient mortality rate with metastatic infection is greater compared with non-metastatic infection.^12,15 The size of an abscess (5–6 cm) was shown to be a significant predictor of extrapulmonary metastatic infection among KPLA patients.¹²

Studies exploring metastatic infection associated with K. pneumoniae are primarily case reviews. There were no major local studies focusing on the risk factors of septic metastatic infection among KPLA patients. This study aimed to identify risk factors of septic metastatic infection and the site of occurrence among KPLA patients in Singapore.

A retrospective case–control study was conducted among adult patients admitted to National University Hospital with KPLA between November 1, 2013 and November 31, 2017. Domain-specific review board approval with a waiver of consent (2018/01265) was obtained for the use of de-identified clinical data. The data were collected using standardized electronic data collection forms. Inclusion criteria were adult patients (age, > 21 years) with K. pneumoniae bacteria isolated from either blood or liver abscess fluid culture, and radiological evidence of liver abscess, as defined by the presence of one or more focal areas of hypo- or hyper-attenuation in the liver.² Patients were further classified into two groups: 1) patients who satisfied the criteria of KPLA and showed evidence of metastatic infection to one or more sites during the entire duration of the hospital stay (case group) and 2) patients with liver abscess only and no evidence of metastatic spread (control group). Metastatic infection was defined as extrahepatic complications such as central nervous system (CNS) infection, endophthalmitis, lung metastasis, skin and soft tissue infections, and other organs as defined by the clinical definition described below. Exclusion criteria were patients with liver abscess who had polymicrobial infection or abscess from other sources.

Data on demographic characteristics and comorbidities, and clinical data on admission were collected. The Charlson comorbidity index was calculated. Systolic blood pressure was further converted into a categorical variable of sepsis-induced hypotension, defined as a systolic blood pressure of < 90 mmHg in the absence of other causes of hypotension.¹⁰ Hospital admission data on length of stay (LOS) in the hospital and the intensive care unit (ICU), if applicable, were collected. Laboratory data on hematology, biochemistry, liver function tests, and microbiology were also collected. Radiological data such as size, multiple abscesses, abscess drainage, and presence of thrombophlebitis, if any, were collected retrospectively. All these data were collected from hospital medical records objectively. Susceptibility data for antibiotics (ceftriaxone, ciprofloxacin, ertapenem, and cotrimoxazole) were collected; however, none of the isolates from the cohort of 116 patients were resistant to these antimicrobials.

The clinical definition of CNS metastatic infections included 1) presence of CNS symptoms and signs such as headache, dizziness, fever (> 38°C), localizing neurological signs, changing level of consciousness, confusion, meningeal signs or 2) imaging evidence of meningeal enhancement, meningitis, or abscess on computed tomography or magnetic resonance imaging.¹⁶ The clinical definition of ophthalmological metastasis included an eye examination and diagnosis of endophthalmitis by ophthalmologists, or eye signs and symptoms such as blurring of vision and pain.¹⁶ The clinical definition of pulmonary metastasis included the presence of septic pulmonary embolism and/or focal or multifocal lung infiltrates, liver abscess as an embolic source,¹⁷ and resolution of lung infiltrates with appropriate antimicrobial therapy.^18,19 The clinical definition of other organs (namely, the spleen) included focal hypo- or hyper-attenuating lesion on ultrasound, computed tomography, or abdominal magnetic resonance imaging, or bacterium identified from the abscess by culture or by a non-culture-based microbiological testing method.

Data analysis was performed using statistical software (STATA version 15; StataCorp LLC, College Station, TX). Depending on data distribution, categorical data were represented as a frequency whereas continuous data were represented by mean (sd) and median (interquartile range). Wilcoxon’s rank sum test and the Student t-test were used to analyze continuous variables. Categorical data were compared using Pearson’s χ² or Fisher’s exact tests.⁵ Statistical significance was defined by a P value of < 0.05. Odds ratio (OR) and CI were used to measure the strength of the association between exposure and outcome. Risk factors for septic metastatic infection were analyzed using logistic regression. Variables in the multivariable regression model were selected using a forward stepwise selection method, with the likelihood ratio test at each step. The model was further compared by the nested model using the likelihood ratio test to achieve the most parsimonious model. The final model was post-estimated using the goodness-of-fit test, and was adjusted for age and diabetes as confounding factors.

The metastatic group included 45 patients; the non-metastatic group included 71 patients. Of the 45 patients in the metastatic group, 100% presented lung metastasis, of which two patients also had splenic metastasis. The population consisted primarily of elderly individuals (median age, 65 years) with a background history of diabetes (60% in the metastatic group versus 56.3% in the non-metastatic group) and hypertension (46.7% versus 50.7%; Table 1). In the metastatic group, 40% of the patients were admitted to the ICU compared with 22.5% in the non-metastatic group (P = 0.04; Table 1). The median ICU LOS was 4.5 days in the metastatic group compared with 4 days in the non-metastatic group (P = 0.049). The median total LOS in the hospital was 10 days in the metastatic group versus 8 days in the non-metastatic group (P = 0.001). Sepsis-induced hypotension was also significantly greater in the metastatic group—15.6% compared with 4.2% in the non-metastatic group. The majority of the patients in the metastatic group underwent abscess drainage (75.6%), compared with half the patients in the non-metastatic group (54.9%; Table 1).

Fever, chills, and rigor and malaise were the three most common symptoms with which patients in both groups presented, with fever being the most common (91.1% in metastatic group versus 98.6% in non-metastatic group; Table 2). Breathlessness was significantly greater in the metastatic group at 24.4%, compared with 5.6% in the non-metastatic group (P = 0.003). Cough and chest pain were common in the metastatic group (22.2% and 11.1% of patients, respectively). Systolic and diastolic blood pressures were significantly less in the metastatic group, with mean systolic blood pressure of 112.89 mmHg (P = 0.003) and a diastolic blood pressure of 64.24 mmHg (P = 0.016), compared with 127.59 mmHg and 70.87 mmHg in the non-metastatic group, respectively. The mean C-reactive protein level in the metastatic group was 228.32 mg/L compared with 194.76 mg/L in the non-metastatic group (P = 0.049). The mean abscess size was significantly larger in the metastatic group, measuring 65.38 mm, compared with 52.70 mm in the non-metastatic group (P = 0.032).

On univariate analysis, statistically significant risk factors of metastatic infection were sepsis-induced hypotension (OR, 4.18; 95% CI, 1.02–17.09), breathlessness (OR, 5.42; 95% CI, 1.61–18.29), albumin (OR, 0.91; 95% CI, 0.86–0.97), total bilirubin (OR, 1.03; 95% CI, 1.01–1.04), and abscess size (OR, 1.02; 95% CI, 1.01–1.03; Table 3). Multivariate analysis showed sepsis-induced hypotension (adjusted OR [AOR], 4.62; 95% CI, 1.04–20.5), breathlessness (AOR, 5.05; 95% CI, 1.41–18), and abscess size (AOR, 1.01; 95% CI, 1.001–1.03) were independent risk factors of metastatic infection among KPLA patients. Sepsis-induced hypotension was reported as a key risk factor in another study.⁵ Breathlessness was one of the most prevalent symptoms in patients with septic metastatic infection.¹⁹ However, these risk factors were not statistically significant in other study.¹⁰ Abscess size is associated with a greater chance of rupture, leading to spread.²⁰ An abscess size of 50 mm or more has been associated with poor prognosis.²¹ The more virulent K1 capsule type is associated with metastatic spread and abscess rupture.^20,22 Singapore has a predominantly Chinese population in which more virulent K1 encapsulation is found.²³ The Chinese population has greater intestinal carriage rate of the K1 capsule type, possibly resulting from different environmental characteristics (such as food type), and sources may result in a greater risk of a larger abscess.²³ Thrombophlebitis, the presence of gas in the liver, and having diabetes mellitus have been reported as important signs of metastatic infection.^10,14 However, these factors were not statistically significant in our study.

Our existing data involves only one hospital, which potentially limits generalizability. However, there were no data on the demographic characteristics of KPLA patients with metastatic infection in Singapore for comparison. There is the possibility of misclassification; however, both the outcome and exposure variables were collected from hospital records objectively. Hence, the chance is low. The Pitt bacteremia and APACHE scores are better representations of the severity of the disease. However, these data and the virulence factors, including the serotype of pneumonia, were not available. The date of drainage was not collected as part of the study; hence, the timeline and interval between abscess drainage and metastatic infection could not be established and assessed for a potential confounding effect. Our existing sample size was small, which resulted in limited statistical power to present some additional significant variables. Among the cases, 77.8% were bacteremic on admission. Although there is the possibility the liver was seeded from the systemic circulation, the possibility is low because the pathogenesis of Klebsiella liver abscess syndrome seen locally and in Southeast Asia is primarily through gut translocation and the splanchnic circulation. None of the patients developed symptoms consistent with CNS metastatic infection; hence, none underwent brain imaging to rule this out. Although it is possible an early or mild CNS manifestation may have been missed, this is consistent with clinical practice and clinical indications for brain imaging.

The risk factors highlighted may guide clinicians to triage patients with a greater risk of metastatic infection at admission. Early detection of increasing abscess size may reduce morbidity.