Pyogenic liver abscess in patients with inflammatory
bowel disease: a
nationwide cohort study
Jiun-Nong Lin1,2,3, Cheng-Li Lin4,5, Ming-Chia Lin6, Chung-Hsu Lai3, Hsi-Hsun Lin3 and
Chia-Hung Kao7,8
Inflammatory bowel diseases (IBDs), primarily including ulcerative colitis and Crohn’s disease, are characterized by chronic and relapsing immune-mediated
inflammation of the gastrointestinal tract (1–3). Increased incidence and prevalence of IBD have
been observed in many regions and countries (4, 5). In addition to intestinal manifestations of IBD, extraintestinal
complications may affect various organ systems
and body parts, such as joints, skin, eyes, the hepatobiliary system, lungs, the heart and the vascular system (6, 7). These extraintestinal complications hinder IBD treatment and substantially influence the quality of life of affected patients.
Pyogenic liver abscess (PLA) is a severe infection in the hepatobiliary system. Population-based studies in Denmark, Canada and the United States have reported annual incidence rates of PLA ranging from 1.1 to 3.6
per 100 000 population (8–11). However, a populationbased study conducted in Taiwan reported an annual
incidence rate of PLA of 17.6 per 100 000 population (12). The etiology of PLA primarily includes intraabdominal bowel contents that spread to the liver through
portal circulation, ascending from biliary infection, haematogenous seeding or cryptogenic infection (nonspecific
lesion causing PLA) (13–15). The PLA incidence
increases in patients with diabetes mellitus, hepatobiliary diseases, pancreatic diseases, intraabdominal infections, colorectal cancer, and hepatobiliary and
pancreatic system interventions (13–18). Clinical manifestations of PLA are typically nonspecific. Delay in
prognosis (14, 17).
Chronic inflammation of intestinal mucosa with IBD results in friable mucosa and the loss of mucosal architecture (2, 3, 19). In theory, intestinal microorganisms can invade the damaged mucosa, reach portal
circulation and induce PLA. Previous studies have
demonstrated an increased incidence of portal bacteraemia in patients with IBD (20–22). However, PLA
remains a rare extraintestinal manifestation of IBD and only sporadic reports of PLA in IBD cases have been described in literature (23–27). The incidence rate of PLA in patients with IBD has never been assessed. Therefore, we conducted a nationwide retrospective cohort study to explore the association between IBD and PLA among adult patients, analysing data from the National Health Insurance Research Database (NHIRD) of Taiwan.
Materials and methods Data source
The National Health Insurance (NHI) system, initiated in 1995 by the Ministry of Health in Taiwan, covers approximate 99% of the 23 million residents in the country (http://nhird.nhri.org.tw/en/index.htm). The National Health Research Institute (NHRI) is responsible for managing the claims data and established the
NHIRD for public use. The NHIRD includes comprehensive information on inpatient care, ambulatory care,
dental care and prescription drugs, and provides
researchers with scrambled identification numbers associated with relevant claims information, including
patient gender, date of birth, registry of medical services and medication prescriptions. In this study, we used the inpatient dataset and registry of the beneficiaries in the NHIRD. Diagnoses and disease management were defined according to the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM).
We used inpatient claims data to identify patients with newly diagnosed IBD, including ulcerative colitis (ICD-9-CM code 556) and Crohn’s disease (ICD-(ICD-9-CM code 555), during 2000 to 2011 (Figure S1). The date of initial IBD diagnosis was defined as the index date. Patients with missing data for gender or date of birth (N = 1), an age of <20 years (N = 4264) and preexisting PLA (ICD-9-CM code 572.0; N = 25) or amebic liver abscess (ICD-9-CM code 006.3; N = 4) were excluded from our study. For each patient with IBD, four control participants were randomly selected from the pool of participants without IBD and PLA at the baseline, and
matched according to year of hospitalization, age (every 5 years), gender and history of diabetes mellitus (ICD-9-CM code 250) by using the same exclusion criteria.
Outcome and comorbidities
All study participants were observed until they were diagnosed with PLA, which was identified according to hospitalization records, censored for loss to follow-up, withdrew from the NHI system or until December 31, 2011. Regarding the primary outcome, amebic liver abscess was excluded and only pyogenic liver abscess was included.
Risk factors for PLA were identified in all participants. Diabetes mellitus (ICD-9-CM code 250), hypertension (ICD-9-CM codes 401-405), hyperlipidaemia
(ICD-9-CM code 272), cancer (ICD-9-CM codes 140-208), chronic obstructive pulmonary disease (COPD) CM codes 491, 492, 496), heart failure (ICD-9-CM code 428), choledocholithiasis (ICD-9-(ICD-9-CM codes
560.31, 574), alcoholic liver disease (ICD-9-CM codes 571.0, 571.1, 571.3), cirrhosis (ICD-9-CM codes 571.2,
571.5, 571.6), cholangitis/primary sclerosing cholangitis (ICD-9-CM code 576.1), cholecystitis (ICD-9-CM code 575) and pancreatic diseases (ICD-9-CM code 577) were identified according to diagnoses in hospitalization records made before the index date. Interventions
within 1 month before PLA occurrence were
also considered a risk factor for PLA. Such interventions included percutaneous aspiration of the gallbladder and biliary tract (ICD-9-CM procedure codes
510.1, 519.8), endoscopic insertion of biliary drainage tube (ICD-9-CM procedure codes 518.6, 518.7), drainage of pancreatic cyst by catheter (ICD-9-CM procedure code 520.1) and endoscopic insertion of
pancreatic duct drainage tube/stent (ICD-9-CM procedure codes 529.3, 529.7).
To address whether the risk of PLA was associated with the severity of IBD, we used the number of annual IBD-related hospitalizations (28), surgery of the intestine (ICD-9-CM procedure codes 45.0-45.9, 46.0-46.9,
48.0-48.9) and laparotomy (ICD-9-CM procedure codes 54.11, 54.19) as indicators of IBD activity. The IBDrelated hospitalizations were defined as hospital admissions before the endpoint with a main diagnosis of IBD, for which the ICD-9-CM diagnoses codes of IBD were listed in the primary or secondary position. For each IBD patient, the number of annual hospitalizations was calculated using the number of IBD-related hospitalizations during the entire follow-up period divided by
number of follow-up person-years.
Ethics statement
The NHIRD encrypts patient personal information to protect privacy and provides researchers with anonymous identification numbers associated with relevant
claims information, including gender, date of birth,
medical services received and prescriptions. Patient consent is not required to access the NHIRD. This study
was approved by the Institutional Review Board of China Medical University (CMU-REC-101-012). The IRB specifically waived the consent requirement.
Statistical analysis
We used SAS software (Version 9.3 for Windows; SAS Institute Inc., Cary, NC, USA) for all data analyses. A P value <0.05 was considered statistically significant.
A chi-square test and Student’s t test were used to evaluate the allocation of category and continuous variables, respectively, between the IBD and control cohorts. Because the NHI system is compulsory in Taiwan, the censored patients are presumably because of death and cannot possibly be at future risk of PLA. Failure to account for the competing risk of death will overestimate the incidence and relative risk of PLA associated with IBD (29). Therefore, we estimated the 12-year cumulative incidence of PLA by using the cumulative risk method, considering death events as a competing risk. After accounting for the competing risk of death, the Fine and Gray model, which extends the standard Cox proportional hazards regression
model, was used to estimate the PLA risk (30). Univariable and multivariable competing-risks regression
models were performed to measure the overall risk and gender-, age-, comorbidity- and intervention-specific risks of developing PLA. Subhazard ratios (SHRs)
and 95% confidence intervals (CIs) were estimated
using the competing-risks regression models. The multivariable models were simultaneously adjusted for age,
gender and the aforementioned comorbidities and interventions involving the hepatobiliary and pancreatic systems.
Results
This study consisted of 11 504 patients with IBD and 46 016 control participants without IBD. Table 1 shows the demographic characteristics. Both cohorts had a similar age and gender distribution; 54.3% of both
cohorts were men and approximately half of all participants were younger than 55 years of age. The mean follow-up period was significantly shorter in the IBD
cohort (5.69 ± 3.7 years) than in the control cohort (6.25 ± 3.47 years; P < 0.001). Patients in the IBD
cohort were more likely to have hypertension, hyperlipidaemia, cancer, COPD, heart failure, choledocholithiasis,
cholecystitis, pancreatic diseases and percutaneous aspiration of the gallbladder and biliary tract than were
those in the control cohort. Figure 1 shows the cumulative incidence curves of PLA for the IBD and control
cohorts after the competing risk of death was accounted for.
Table 2 shows the incidence and relative risk of PLA
between the IBD and control cohorts. The overall incidence of PLA was higher in the IBD cohort than in the
control cohort (6.72 vs 4.06 per 10 000 person-years), with an adjusted SHR of 1.46 (95% CI, 1.01–2.12). The adjusted SHR of PLA in the IBD cohort was significantly higher in male patients (adjusted SHR, 1.78; 95% CI, 1.08–2.93) than in female patients. Compared with patients in the control cohort, the youngest patients with IBD exhibited the highest relative risk of developing PLA (adjusted SHR, 4.2; 95% CI, 1.21–14.6). In
patients without comorbidities, the risk of PLA was 2.07-fold higher in the IBD cohort than in the control cohort (95% CI, 1.24–3.45).
Table 3 shows the PLA risks related to IBD severity. The incidence of PLA was significantly increased in patients who required hospitalization twice or more per year (290.1 per 10 000 person-years; P < 0.001). Compared with the controls, patients with IBD who were
hospitalized twice or more per year had an adjusted
SHR of 4.92 for PLA (95% CI, 2.72–8.89). The IBD patients who underwent laparotomy exhibited a higher
risk of PLA than did the controls (adjusted SHR, 4.47; 95% CI, 1.55–12.9).
The incidence and relative risks of PLA in patients with ulcerative colitis and Crohn’s disease are shown in Table S1. Patients with ulcerative colitis had a 1.65-fold increased risk of PLA compared with the control participants (95% CI, 1.03–2.65); this was higher than the
increased risk of PLA in patients with Crohn’s disease (adjusted SHR, 1.32; 95% CI, 0.82–2.14). However, no specific association between PLA and the involved intestinal
sites of IBD was observed (Table S1).
We further analysed the risk of PLA in the IBD and control cohorts, which were stratified by age, gender, comorbidity, and hepatobiliary and pancreatic system interventions (Table 4). Among the IBD cohort, age
and gender did not significantly impact PLA risk. However, IBD patients with diabetes mellitus (adjusted SHR,
3.14; 95% CI, 1.38–7.1) or percutaneous aspiration of the gallbladder and biliary tract (adjusted SHR, 39; 95% CI, 7.62–199.9) or who underwent endoscopic insertion of a biliary drainage tube (adjusted SHR, 488.5; 95% CI, 216.5–1102.1) exhibited a significantly increased risk of PLA. Regarding the control cohort, an older age, diabetes mellitus and percutaneous aspiration of the gallbladder and biliary tract were significantly related to
developing PLA. Discussion
A comprehensive literature review suggested that this study is the first cohort investigation to examine the incidence and risk of PLA in patients with IBD. Our study determined that the incidence of PLA in patients with IBD was 6.72 per 10 000 person-years, with a 1.46-fold increased risk for PLA compared with the general population. In addition, we observed that PLA risk increased substantially according to the frequency of IBD-related hospitalizations and laparotomy,
possibly reflecting the flare-up frequency and severity of IBD. Patients with ulcerative colitis were more likely to develop PLA than were those with Crohn’s disease (incidence, 8.56 vs 5.45, per 10 000
person-years). Previous studies have indicated that men and younger patients with IBD have a higher risk of developing
PLA (14, 23, 31, 32). In the present study, PLA was more prevalent in male patients and older patients with IBD. Although male and younger IBD patients had a significantly increased risk of PLA compared with the controls, gender and age did not additionally influence the risk of developing PLA in patients with IBD.
Several factors are associated with developing PLA, including intraabdominal infection, choledocholithiasis, biliary tract disorders, pancreatitis, malignancy, and hepatobiliary and pancreatic system interventions (13– 18). In Asian countries and particularly in Taiwan, cryptogenic PLA caused by Klebsiella pneumoniae has
emerged as a new invasive syndrome in patients with diabetes mellitus (16, 33). K. pneumoniae is the leading pathogen for PLA in Taiwan, and approximately 60% of patients with K. pneumoniae PLA have diabetes mellitus (16, 34). PLA caused by this highly virulent microorganism has also increased in Western countries (33–36).
During recent years, an increasing number of K. pneumoniae PLA has been reported in Europe and the United
States (33–36). To avoid this confounding factor, we matched diabetes mellitus among the IBD and control cohorts in addition to the year of hospitalization, age and gender. We suppose that the results of this study are generalizable to a Western population.
Ulcerative colitis and Crohn’s disease are the two primary disorders of IBD, and both have the same characteristics of chronic and relapsing inflammation of the
gastrointestinal tract (1, 31). However, the epidemiology, genetic predisposition, clinical manifestations,
endoscopic results and histological features are dissimilar (1, 31). Some extraintestinal manifestations are more prevalent in patients with Crohn’s disease, whereas others are more prevalent in patients with ulcerative colitis (37, 38). Among the extraintestinal complications of IBD, PLA is a rare but severe disorder. Previous case reports, mostly from Western countries, described that PLA was observed more frequently in patients with Crohn’s disease (23–27). In a review article in 2004, only
seven (11.7%) cases of PLA occurred in patients with ulcerative colitis among the 60 cases of PLA in patients with IBD (23). However, our study revealed that incidence and risk of developing PLA was higher in patients
The exact cause for this difference remains unclear. Genetic, ethnic, environmental or geographic differences may affect the development of PLA in patients
with IBD.
The main strengths of this study include the large nationwide samples, comprehensive demographic characteristics and complete histories of follow-up. However, this study has several limitations. First, PLA can be mistaken as an acute exacerbation of IBD because the symptoms of PLA are nonspecific. Therefore, the incidence of PLA may have been underestimated. Second, we used inpatient claims data to analyse the risk of PLA in patients with IBD. Although most of these patients required hospitalization, patients with mild disease episodes might have
be excluded in this study. Third, IBD and PLA diagnoses in the NHIRD cannot be validated by performing
chart reviews because the insured medical records are released as deidentified data to the public for research purposes. However, a committee in NHI
reviews the medical records and assessed the accuracy of the claims files. Institutions will be fined and not receive reimbursement if incorrect coding is identified. Thus, we are confident using the coding system, and several studies have proven the accuracy of NHIRD data (39–42). Fourth, the inpatient dataset and registry of the beneficiaries in the NHIRD do not include information on medications, symptoms and laboratory data. Therefore, although we used the number of annual IBD-related hospitalizations, surgeries of intestine and laparotomies to represent disease activity, the actual severity of IBD could not be evaluated accurately. In addition, the relationship between IBD-related therapy and PLA risk could not be assessed. Finally, primary sclerosing cholangitis, an extraintestinal manifestation of IBD, has no unique ICD-9-CM code. Therefore, we could not differentiate primary sclerosing cholangitis from other cholangites
in this ICD-9-CM code-based study.
In conclusion, PLA is a rare complication of IBD. This extraintestinal disorder can increase morbidity and mortality in patients with IBD. Our study determined the incidence and relative risk of PLA in
patients with IBD. Because the symptoms of PLA are
typically nonspecific and may appear to be the exacerbation of IBD, PLA diagnosis requires a high degree
of clinical suspicion. Early diagnosis and adequate treatment can minimize the complications of this severe infection.
