We retrospectively reviewed electronic medical records and clinical microbiology records in Kurashiki Central Hospital (Okayama, Japan), a 1,166-bed, tertiary-care hospital that provides care to ≈300,000 persons annually. Clinical specimens submitted to the microbiology laboratory included blood, sputum, urine, bile, ascites, feces, placenta, tissue, and pus. Information about identified bacteria and antimicrobial susceptibility were kept as microbiology laboratory records for each specimen. We considered bacteremia to exist when >1 set of blood cultures was positive. We identified all cultures growing E. tarda from clinical specimens submitted during January 2005–December 2016.

We processed blood culture samples using the BacT/Alert system (Sysmex bioMérieux Co. Ltd., https://www.biomerieux.com) and conducted microbial culture using KBM Chocolate HB Agar (Kohjin Bio Co. Ltd., http://www.kohjin-bio.jp/english), KBM Sheep Blood Agar (Kohjin Bio Co. Ltd.), and BTB agar (Kyokuto Pharmaceutical Co. Ltd., https://ssl.kyokutoseiyaku.co.jp/english/index.html). We used different bacterial identification and antimicrobial susceptibility testing methods in our hospital throughout the study period. We used ID test EB-20 Nissui (Nissui Pharmaceutical Co. Ltd., https://www.nissui-pharm.co.jp/english) for bacterial identification and Kirby–Bauer disk (Eiken Chemical Co. Ltd., http://www.eiken.co.jp) for antimicrobial susceptibility testing from January 2005 through June 2007. EB-20 is a system to identify glucose-fermenting gram-negative rods by 20 patterns of biochemical properties, using hydrogen sulfide, indole, lysine, ONPG (2-nitrophenyl-β-D-galactopyranoside), adunit, inositol, rhamnose, mannit, esculin, Voges-Proskauer, arginine, urea, inositol, sorbitol, arabinose, phenylpyruvic acid, citric acid, ornithine, malonic acid, raffinose, and sugar. Thereafter, automatic systems were introduced at our hospital: DPS192 (Eiken Chemical Co. Ltd, http://www.eiken.co.jp) during July 2007–February 2013 and MicroScan WalkAway (Beckman Coulter, Inc, https://www.beckmancoulter.com/en) during March 2013–March 2014. Since April 2014, we have used MALDI Biotyper (Bruker Daltonics GmbH, https://www.bruker.com), using the manufacturer-provided database, for bacterial identification. We judged the drug susceptibility of a microorganism based on clinical breakpoints set by the Clinical and Laboratory Standards Institute; in particular, we used the document M100-S22 (2) during June 1, 2013–December 31, 2016.

We collected all clinical information of patients with positive E. tarda bacteremia results from electronic medical records, including age, sex, underlying diseases, source of infection, antimicrobial drug administered, treatment period, and outcome. We defined chronic kidney disease as a serum creatinine level of >2.0 mg/dL (reference range 0.65–1.07 mg/dL) and chronic liver disease as liver cirrhosis or chronic hepatitis B or C infection. We defined nosocomial bloodstream infection, healthcare-associated bloodstream infection, community-acquired bloodstream infection, and febrile neutropenia according to the previous study and guideline (3,4). We defined 30-day mortality as patient death within 30 days after the onset of E. tarda bacteremia and 90-day mortality as patient death within 90 days after onset. We also collected information of patients with E. tarda nonbacteremic infections.

We described the clinical characteristics and 30-day mortality of patients with E. tarda bacteremia, along with the source of infection and antimicrobial susceptibility. We then compared the characteristics of patients with E. tarda bacteremia by 30-day mortality. We also compared the characteristics of patients with bacteremic and nonbacteremic E. tarda infections. We also conducted an exploratory multivariable logistic regression analysis to investigate the risk for E. tarda bacteremia incidence among all E. tarda infections.

Because a previous literature review suggested seasonal variation in the occurrence of E. tarda bacteremia (5), we thus examined whether such variation or trend existed in the cases in our study by using Cochran-Armitage test. We tested dichotomous variables with Fisher exact test and and continuous variables by Wilcoxon signed-rank test. Statistical analysis was performed using Stata version 15.1 (StataCorp, http://www.stata.com). We considered p<0.05 to be statistically significant.

The Ethics Committee of Kurashiki Central Hospital approved this study (no. 2,527). Only persons with appropriate authorization had access to participants’ records, and patient confidentiality was maintained. Given the nature of a retrospective chart review, written consent from the patients was waived.

We obtained 182,668 sets of blood cultures during the study period, of which 19,234 sets were positive for some organisms and 40 sets from 26 patients were E. tarda–positive. E. tarda bacteremia was diagnosed in 26 patients (13 men and 13 women); their median age was 75 years (range 45–101 years) (Table 1).

Seasonal Variation in E. tarda Bacteremia

Figure

Figure. Seasonal variation in the incidence of Edwardsiella tarda infection, Kurashiki Central Hospital, Okayama, Japan, 2005–2016. Black bars, blood culture; gray bars, all specimens (including blood cultures).

The incidence of E. tarda infection did not vary by season (Figure). We found no trend of E. tarda bacteremia incidence among all E. tarda infections when we examined them by month (p = 0.46) or by season, defined as a set of 3 months (p = 0.53).

E. tarda is associated with freshwater and marine life, including fish, reptiles, and amphibians (1). The organism resembles Salmonella biochemically and clinically (1). Salmonella usually ferments D-mannitol, urease, oxidase, and D-sorbitol, whereas E. tarda produces hydrogen sulfide and indole (6).

E. tarda is a rare human pathogen and is primarily associated with gastrointestinal diseases, including the asymptomatic carrier state (1). Approximately 80% of infections are intestinal. E. tarda causes a Salmonella-like gastrointestinal infection, usually self-limited enteritis, with intermittent watery diarrhea and low-grade fever (1,7).

The pathogenesis of E. tarda and its disease-causing mechanism remain unclear. Twelve classes of bacterial protein secretion systems are known; these systems transport virulence proteins into the cell and, in some cases, directly into the cytoplasm of a target cell (8). The bacterial type III and type VI secretion systems (T3SS and T6SS) are believed to play an essential role in E. tarda survival, replication, and virulence inside the host. In particular, T6SS is proposed to enable E. tarda to establish inside the host, cause severe systemic infection, and eventually kill the host.

We reviewed 26 cases of E. tarda bacteremia. Clinical diagnoses included 15 (58%) biliary tract infections (cholangitis, cholecystitis, and liver abscess). Eight of these patients had hepatobiliary diseases including cholangiocarcinoma, gallbladder cancer, pancreatic cancer, gallstone disease. Therefore, hepatobiliary diseases may be a predisposing factor of E. tarda biliary tract infections. However, our multivariable logistic regression found that only age >65 years was associated with the incidence of E. tarda bacteremia. We acknowledge that the sample size of our study and the number of E. tarda bacteremia incidence were still small, and thus the finding from our multivariable analysis might be only exploratory.

Previous studies reported high rates of death for E. tarda bacteremia, ranging from 22.7% to 44.6% (1,5,9). In contrast, the death rate for patients with E. tarda bacteremia in the cohort reported here was low at 12%. However, 2 of these 3 patients had end-stage liver disease; only 1 death among these patients was attributed to E. tarda bacteremia.

E. tarda is susceptible to most antimicrobial drugs, including tetracyclines, aminoglycosides, quinolones, antifolates, chloramphenicol, nitrofurantoin, fosfomycin, and most β-lactams (10), and is naturally resistant to benzylpenicillin, colistin, and polymyxin B (1,11). In our study, E. tarda was susceptible to most commonly used antimicrobial drugs. E. tarda susceptibilities to colistin and polymyxin B are unknown because susceptibility testing is not routinely performed for these drugs in our institution. Previous studies have shown that all strains of E. tarda were positive for β-lactamase production examined with nitrocefin β-lactamase disks, but an ampicillin-resistant E. tarda strain has not been reported (10,11). Whether E. tarda isolates detected in our institution produced β-lactamase is not clear because we did not perform the β-lactamase test, but 5 cases were successfully treated with ampicillin.

Hirai et al. suggested that E. tarda bacteremia is likely to develop during summer and autumn months in the Northern Hemisphere (8). The authors conducted a literature review of 77 E. tarda bacteremia cases reported from diverse areas and suggested seasonal variation in incidence for 22 cases. Our study of 26 E. tarda bacteremia cases suggests no such seasonal distribution. Several possible reasons might account for this discrepancy. First, E. tarda can colonize. In our study, hepatobiliary infection (such as cholangitis, cholecystitis, and liver abscess) was diagnosed in 58% (15/26) patients, and patients colonizing E. tarda developed E. tarda bacteremia. Second, diversity might exist in the patients’ dietary patterns. E. tarda frequently infects fish. Hirai et al. included patients from many parts of the world, so the intake of fish might have differed according to the season or geographic area across reports. In contrast, our study included only people in a single area of Japan who habitually ate raw seafood, such as sashimi, throughout the year; this tendency might have led to no seasonal variation of E. tarda bacteremia incidence. Third, our study had no missing clinical data for any patients, whereas Hirai et al. examined 22 of all 77 eligible patients, which might have rendered their analysis vulnerable to information bias.

Our study had some strengths. First, we elucidated that no seasonal variation existed in E. tarda bacteremia in this population. Second, we described the characteristics of each patient with E. tarda bacteremia and provided risk factors for E. tarda bacteremia incidence among all E. tarda infections.

Our study also had some limitations. First, the number of blood cultures submitted increased in recent years in our hospital. The number of blood cultures submitted in 2016 nearly doubled that for 2005. This increase might have resulted in the underestimation of E. tarda bacteremia in the earlier years of our study period. Second, ours was a retrospective and single-center study. However, our study had no missing data regarding clinical information. Furthermore, we successfully presented a particularly large case series of E. tarda bacteremia.

In conclusion, E. tarda bacteremia is a rare disease that is not associated with high rates of death. E. tarda bacteremia patients in our cohort in Japan had more severe underlying diseases, such as hepatobiliary disease and solid tumors, than did patients in previous studies. Hepatobiliary infections, such as cholangitis, cholecystitis, and liver abscess, are the most common clinical manifestations in patients with E. tarda bacteremia. The major underlying diseases in this study were hepatobiliary diseases and malignancy. Furthermore, E. tarda strains we isolated were susceptible to most antimicrobial drugs, including β-lactams, aminoglycoside, tetracycline, fosfomycin, fluoroquinolone, and trimethoprim/sulfamethoxazole, and E. tarda bacteremia was successfully treated with ampicillin. Finally, we observed no seasonal distribution of E. tarda bacteremia. Risk factors for E. tarda bacteremia–related death remain to be investigated.