The association between bowel resection and the risk of nontyphoidal salmonella infection: a nationwide propensity score-matched cohort study

Nontyphoidal salmonella (NTS) infection has a high mortality rate. Bowel resections affect gut microbiota and immune function, and the association between bowel resection and NTS infection in human beings has not been addressed. We conducted a nationwide propensity score (PS)-matched cohort study to clarify this association. Data from the Longitudinal Health Insurance Database of Taiwan were used to establish a case-cohort with bowel resections from 2000 to 2013. Informed consent was waived by the Institutional Review Board of China Medical University Hospital (CMUH104-REC2-115) because all personal identifying information used had been de-identified. Each case was matched with one control without any bowel resection according to age, gender, index date, and propensity score (PS). Cumulative incidences of and hazard ratios (HRs) for NTS infection development were analyzed. The incidence of NTS infection was greater in patients with a bowel resection than in the control group (2.97 vs. 1.92 per 10,000 person-years), with an adjusted hazard ratio (aHR) of 1.64 (95% CI = 1.08–2.48). The incidence of NTS infection increased significantly for cases with small bowel resections and right hemicolectomies. Age (31–40 and > 50 years), hypertension, chronic kidney disease, chronic obstructive pulmonary disease, and autoimmune diseases were significant risk factors of NTS infection. Stratification analysis revealed that patients without comorbidities were prone to NTS infection after bowel resections. The increased risk of developing NTS infection could be related to the bowel resection. Specific age groups and comorbidities also contribute to increased risk of NTS infection.

www.nature.com/scientificreports/ controlling effects on S. enterica, such as Bifidobacterium thermophilum and Lactobacillus casei 6,7 . Argüello et al. discovered that manipulating certain taxa of microbiota could decrease salmonella infection for pigs 8 .
In human beings, bowel resection strongly influences gut microbiota and intestinal healing 9 . In the small intestine, the main reported bacterial species included Bacilli (Firmicutes), Streptococcaceae (Firmicutes), Actinobacteria, Clostridium and Bacteroides 5 . Phyla Bacteroidetes and Firmicutes can dominate the mucosaassociated bacteria from the distal small intestine and the colon 5 . Specialized bacteria (e.g., Clostridium, Lactobacillus, or Enterococcus) able to adhere to mucus (as a nutrient source) and different species in feces (belonging to Bacteroides, Bifidobacterium, Streptococcus, Enterobacteriaceae, Enterococcus, Clostridium, Lactobacillus, and Ruminococcus) were identified 5 . Kunz et al. indicated that gastrectomy could lead to S. enteritis. This might result from relative or absolute achlorhydria, decreased hydrogen ion concentration, rapid emptying of food into the small intestine and the colon and altered bacterial flora 10 .
However, currently the association between bowel resection (except the stomach) and NTS infection was still not clarified. The present nationwide propensity score (PS)-matched cohort study was carried out in Taiwan to evaluate the risk of NTS infection after bowel resection.

Materials and methods
Data source and ethics statement. We extracted the claims-based data from the Longitudinal Health Insurance Database 2000 (LHID), which is a dataset included in Taiwan's National Health Insurance Research Database (NHIRD). The NHIRD contains patients' characteristics, medical information, total expenditure, and diagnoses coded in the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) format. Taiwan's National Health Insurance (NHI) programme is administered by the government and covers more than 99% of the 23.4 million population of Taiwan 11 . Data for the LHID were collected by systematically and randomly sampling from the NHIRD; this database includes the data of one million individuals. Information on demographic data, inpatient and outpatient cares, date of clinic visit or hospitalization, and prescriptions were available in the database for the period from 1997 to 2013. The NHRI reported no significant differences between the patients in the LHID and those in the original NHIRD 12 . The studies involving human participants were reviewed and approved by the Institutional Review Board of China Medical University Hospital (approval number CMUH104-REC2-115). Written informed consent for participation was not required for this study in accordance with national legislation and institutional requirements. Our research adhered strictly in accordance with relevant guidelines/regulations.

Study subjects and design.
A bowel resection-related case group was established, including patients aged 18 years and over who had received a bowel resection (small bowel, large bowel, or both) from 2000 to 2013. We included the cases identified by the procedure codes of 45.6 (small bowel resection), 45.73 (right hemicolectomy), 45.74 (transverse colon resection), 45.75 (left hemicolectomy), 45.76 (sigmoidectomy), 45.8 (total colectomy), or 45.9 (bowel anastomosis). We defined the index date as the operation date. We excluded patients with appendectomy/gastrectomy/cholecystectomy, any kind of cancer, human immunodeficiency virus infection, index dates out of the study period, diagnosis of NTS, or missing gender or aged under 18 before index dates (Fig. 1). The outcome was the development of NTS (ICD-9-CM codes: 003.xx) after the index date.
Propensity score matching and covariates. A matched control was assigned to each case ( Fig. 1).
We used the PS-matching method to account for baseline differences 13

Statistical analysis.
Chi-square test was used to determine the differences of baseline characteristics for categorical variables and Student's t-test was applied to examine continuous variables between case group (bowel resection) and the control group. Standardized mean difference (SMD) was used to assess the difference between the case group and the control group after PS matching; a value of < 0.1 was considered negligible 13 . Incidence of NTS infection was estimated for both groups by the end of 2013. We used the Kaplan-Meier method to measure fractions free of NTS infection during the follow-up period in the two groups and used the log-rank test to examine the difference. Incidence rates (IRs) of NTS infection, per 10,000 person-years, were calculated for patients with or without bowel resections, gender, age group, occupation, and comorbidities. Cox proportionalhazards regression analysis was used to calculate the case group to the control group hazard ratio (HR) of NTS infection and 95% confidence interval (CI). Adjusted hazard ratio (aHR) was estimated after controlling for covariates. Gender-, age-, occupation-, and comorbidities-stratified analyses of NTS infection with or without bowel resection were conducted. All statistical analyses were performed using SAS software version 9.4 (SAS Institute INC., Carey, NC, USA). A two-tailed p-value below 0.05 was considered as significant.
Sensitivity analysis. According to Wu et al. 14 , a temporal association between the use of proton pump inhibitors and increased susceptibility to NTS was demonstrated, which would bias our results. Therefore, we www.nature.com/scientificreports/ performed the sensitivity analysis by excluding peptic ulcer disease patients (ICD9 codes: 531, 532, 533, and 534) who had a high probability of taking proton pump inhibitors and PS matching.

Results
Demographic characteristics of patients. Our study population includes 35,851 cases and 35,851 controls. Table 1 shows that the male ratio and the mean age were 52% and 56.1 years in the case group, and 51% and 55.6 years in the control group, respectively. There was no significant difference in the baseline occupation and comorbidities. The overall median follow-up time was 4.53 years.

Cumulative incidence and risk of NTS infection.
By the end of follow-up, the Kaplan-Meier analysis showed that the NTS infection was significantly more in the bowel resection group than in the control group (p = 0.03) (Fig. 2). The IR of NTS infection was higher in the bowel resection group than in the control group (2.97 vs. 1.92 per 10,000 person-years), with an aHR of 1.64 (95% CI = 1.08-2.48) for the bowel resection group ( www.nature.com/scientificreports/ Results of sensitivity analysis. Results of our sensitivity analysis were almost consistent with those of our primary analyses ( Table 4). The differences are that our sensitivity analysis even indicated that patients would have an increased risk of NTS infection receiving large bowel resection (aHR of 1.69, 95% CI = 1.02-2.80) or left hemicolectomy (aHR of 2.41, 95% CI = 1.01-5.77). The patient baseline characteristics in our sensitivity analysis are shown in Table 5.

Discussion
Our study first demonstrated that bowel resection was associated with a significantly increased risk of NTS infection. Of note, the HR of NTS infection for patients with small bowel resection or right hemicolectomy was nearly two times greater than for those without small bowel resection or right hemicolectomy. We hypothesized that based on the bowel resection site, the microbiota composition change would result in NTS infection. As mentioned above, the reported bacterial species in the intestine included Bacilli (Firmicutes), Streptococcaceae (Firmicutes), Actinobacteria, Clostridium, Staphylococcus, Bacteroides, Lactobacillus, Enterococcus, Bifidobacterium, Enterobacteriaceae, and Ruminococcus. Oh et al. demonstrated the potential protective effects of using Bacillus subtilis CSL2, against Salmonella gallinarum infection on laying hens 15 . Actinomyces might have antimicrobial potency toward NTS 16 . Jacobson et al. discovered that in mice, Bacteroides spp. limit intestinal Salmonella typhimurium expansion and fecal shedding 17 . Lactobacillus has remarkable anti-salmonella activities in vivo and in vitro 18,19 . Strains of Bifidobacterium were antagonistic to Salmonella 20 . Bifidobacterium thermophilum RBL67 influences the transcriptome of Salmonella and causes an imbalanced virulence gene expression, but can protect hosts from infection 21 . Clostridia, including the genera Roseburia and Blautia from the family Lachnospiraceae and the genera Ruminococcus and Anaerovibrio, were more abundant in Salmonellanegative pigs 8 . A metabolite from bacteria-indole, is probably associated with NTS infection in our gut. Indole can prevent NTS infection because of its effects on decreasing Salmonella invasion in vivo, Salmonella motility, virulence gene expression, and increasing epithelial cells' resistance to Salmonella invasion 22 .
Gordon et al. indicated that host risk factors included extremes of age and immunocompromised status such as rheumatologic disease, which corresponded with our results 3 . We discovered that patients aged 31-40 years had an increased risk of NTS infection, which was first reported by us. Our study demonstrated other risk factors for NTS infection like hypertension, chronic kidney disease, and COPD. Atherosclerosis was demonstrated a risk factor for endovascular Salmonella bacteremia 23  www.nature.com/scientificreports/  www.nature.com/scientificreports/ peripheral CD4 + and CD8 + T cells and B cells in the blood decrease 25 , and Fernández-Fresnedo et al. indicated an increased incidence of apoptosis in B cells in chronic kidney disease patients 26 . In end-stage renal disease patients, increased apoptosis and diminished populations of naïve and central memory T cells were found, and their antigen-specific memory CD4 + T cells were impaired 27 . In smokers with COPD, Knobloch's research revealed that interferon-γ release from ex vivo generated CD4 + effector cells of the Th1 subtype upon challenge with lipopolysaccharide (LPS) purified from Salmonella minnesota would decrease 28 .
Our stratification analysis showed patients without hypertension, diabetes, hyperlipidemia, coronary artery disease, cerebrovascular accident, chronic kidney disease, COPD, chronic liver diseases, vasculitis, or IBD would be at higher risk of NTS infection after bowel resection. We might explain this phenomenon by the fact that affected patients usually have underlying diseases, so that NTS poses significant health threats 4 . If underlying covariates did not exist, bowel resection would play a significant predisposing factor of NTS infection. Our results reaffirm that in patients without underlying diseases, bowel resection should be considered carefully to avoid possible NTS infection in the future.
Our study has the advantage that it is a large-scale nationwide analysis with longitudinal follow-up. However, we admit several limitations to our study. First, we lacked information regarding body mass index, alcohol consumption, betel nut use, smoking, and other immunomodulator prescription. However, we used adjusted covariates, such as autoimmune diseases, psoriasis, and IBD, which could represent immune-related medication use. Second, the potential misclassification in establishing study cohorts might exist. We used bowel resection procedure codes to identify the case group. The policy of patient anonymity within the NHIRD prevented us from confirming patients' diagnoses. Third, we demonstrated that right hemicolectomy in our primary analyses and left hemicolectomy in our sensitivity analysis increased the risk of NTS infection, but we did not discover any reference which could explain the mechanism.

Conclusions
Bowel resection causes an increased risk of NTS infection. Patients aged 31-40 and > 50 years or those with comorbidities such as hypertension, chronic kidney disease, COPD, and autoimmune disease have an elevated risk of NTS infection. Furthermore, in patients without comorbidities, bowel resection should be considered carefully.