Plain Language Summary

Clostridium difficile (C. difficile) is a bacterium that often lives without causing harm in people’s gut. However, when a person has antibiotic treatment for another infection, this can cause an imbalance in the normal levels of bacteria in the gut, and the C. difficile can grow and replace many of the normal bacteria, causing C. difficile infection (CDI). Symptoms include diarrhea, fever and pain. Although CDI is often mild, it can be very serious, particularly in older people, and, if untreated, can be fatal. This review looked at studies published from 2006 to 2017 to investigate patterns of CDI sickness (epidemiology) in Japan. A total of 55 studies were useful for our review and showed that, in general, CDI occurred less commonly in Japan than in Western countries. However, there was wide variation in the tests used to detect infection and the methods used to identify specific types of C. difficile bacteria responsible for the infections. Because of this variety, there was a difference in the reliability of the results from the different studies, which made it difficult to make comparisons between studies. However, there seemed to be consistent results showing that certain types of C. difficile were common in Japan. The studies were not able to tell us whether the types of C. difficile varied over time. More studies that use reliable high-quality tests, and greater detailed analysis in Japan to map patterns of CDI over time are needed. This would help us to understand the importance of CDI in Japan.

Introduction

Clostridium difficile is the most common infective cause of nosocomial diarrhea, implicated in 20–30% of cases of antibiotic-associated diarrhea [1, 2]. Appropriate patient care requires rapid and accurate diagnosis to support optimal management and prevent the spread of infection. Furthermore, knowledge of specific risk factors for C. difficile infection (CDI) in different clinical settings is essential.

No national CDI surveillance system has been implemented in Japan, and therefore it is challenging to grasp the trend in epidemiology over time using a standardized method. A review of CDI in Asia published in 2013 found only a few molecular-typing studies providing contemporary epidemiological information [3]. According to a questionnaire-based survey of 2537 hospitals in Japan in 2013, which had valid responses from 321 hospitals, CDI incidence varied between centers [4], and there was little information on the specific strains causing infection.

There have been several important changes in CDI diagnosis and treatment in Japan. First, a new diagnostic kit detecting toxin A and B plus “common” antigen (glutamate dehydrogenase; GDH) became available in April 2011. Second, oral and injectable metronidazole were indicated for CDI in August 2012 [5] and September 2014 [6], respectively, although unlicensed use of oral metronidazole for CDI had occurred in Japan prior to 2012. Third, in 2015, the Japanese Association for Infectious Diseases and Japanese Society of Chemotherapy released guidelines for the treatment of enteric infection, in which oral metronidazole was designated as the first-line treatment for CDI [7]. Vancomycin was recommended for severe cases and/or second and subsequent recurrences [7].

Considering these recent changes in the diagnosis and treatment of CDI, there is a greater need to understand and update the epidemiology of CDI, the predominant strains causing the infection, and the consequences, risks and resource utilization associated with CDI in clinical settings in Japan. This literature review was undertaken to summarize published epidemiological data on CDI in Japan from January 2006 to November 2017, to describe definitions of CDI applied, molecular typing and diagnostic methods used, and key risk factors and expected outcomes.

Methods

The recent literature was reviewed in a systematic fashion to identify studies and reports relating to the epidemiology of CDI in Japan. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were used to inform search terms, and the literature review process was conducted using the PRISMA Checklist and PRISMA Flow diagram.

Identification

Searches of MEDLINE-PubMed and EMBASE® were made using the following primary search terms: C. difficile infection; pseudomembranous colitis; epidemiology; Japan. Secondary search terms were as follows: C. difficile diarrhea; C. difficile colitis; enterocolitis; toxic megacolon; hospital-acquired diarrhea; nosocomial diarrhea; antibiotic-associated diarrhea; incidence; Japan. The publications were limited to the English language from 1 January 2006 to 27 November 2017.

Selection

Identified abstracts were reviewed by a single reviewer to remove duplicates and to identify publications meeting the pre-defined inclusion and exclusion criteria. Inclusion criteria were: Japanese patients or human samples with CDI; observational or non-randomized interventional studies; cross-sectional surveys; cohort studies; case–control studies; pharmacy records or claims databases; electronic registers or electronic medical/health records; insurance or administrative claims databases studies; registry studies; prospective or retrospective studies; longitudinal or follow-up studies; and reviews. Publications were included if they reported on: CDI epidemiology (incidence/prevalence); CDI risk factors; CDI definitions; diagnostic and laboratory test methods; CDI strains; length of hospital stay (LOS); intensive care unit admission; CDI recurrence; and mortality. There was reliance on the individual publications to define CDI and no minimum (discriminatory) definition was used during the selection process. Exclusion criteria were: animal studies; in vitro studies; case reports; editorials, commentaries and letters; congress abstracts; and non-English language publications. All publications that met criteria for the review were obtained as full articles, reassessed and reviewed.

Quality Determination and Data Extraction

A single reviewer assessed the quality of each paper/study according to Oxford Centre for Evidence-Based Medicine – Levels of Evidence [8] (Enhanced Supplementary Material). As most of the captured studies did not fall strictly within a given category, references were also assessed by the same individual to ensure consistent application of the criteria across all publications.

Data from the selected studies were extracted by a single reviewer and used to populate summary tables.

Compliance with Ethics Guidelines

The analysis in this article is based on previously conducted studies and does not involve any new studies of human or animal subjects performed by either of the authors.

Results

Identification of Relevant Publications

A total of 385 potential articles were identified, of which 55 were defined as relevant, after applying the pre-defined inclusion and exclusion criteria (Fig. 1). One article was an erratum [9] of a previously identified study [10], therefore the study was counted only once. The assigned grades of literature per relevant article are summarized in the Enhanced Supplementary Material.

Fig. 1
figure 1

Assessment of search results to identify key papers for review and data extraction. Asterisk did not meet inclusion criteria in relation to study population or design (see “Selection”). Hash did not include reports of: Clostridium difficile infection (CDI) epidemiology (incidence/prevalence); CDI risk factors; CDI definitions; diagnostic and laboratory test methods; CDI strains; length of hospital stay; intensive care unit admission; CDI recurrence; or mortality. Dagger one identified article was an erratum of a previously identified study, therefore the study was counted only once

Full size image

Many papers were insufficiently specific: for example, several papers reported on the validation of novel C. difficile diagnostic assays or laboratory testing methods in Japan, but did not include clinical data. Others were excluded owing to a small sample size (n ≤ 8), a focus on pre-clinical evaluation of CDI testing methods, or for reporting CDI contamination in non-patient groups [11,12,13,14,15]. Several papers were reviews or editorials with limited relevance to Japanese CDI epidemiology [3, 16, 17].

Incidence and Prevalence

Twenty-four papers reported data relating to the incidence and prevalence of CDI, or to C. difficile-related disease or diarrhea in Japanese cohorts (Table 1). Most reports were based on retrospective chart reviews (n = 16), eight were prospective studies; either observational studies or randomized controlled trials (RCTs), and one was a systematic review and meta-analysis. Most of the papers described and defined CDI in terms of clinical diagnosis (‘diarrhea’) and laboratory findings (‘toxin positivity’). The papers differed in their approach to testing for CDI: some actively investigated C. difficile colonization across cohorts whilst others tested for C. difficile only in patients with clinical symptoms suggestive of CDI. Testing methods also varied. Because of these differences and the heterogeneity in the patient populations examined in the publications, including hematopoietic stem cell transplant (HSCT) patients, rheumatology patients and those with Helicobacter pylori-positive peptic ulcer, it was not possible to examine trends over time in the incidence of CDI.

Table 1 Incidence and prevalence of C. difficile, and risk factors associated with C. difficile infection reported in Japanese patient populations
Full size table

Very few papers reported incidence in terms of cases per 10,000 patient-days; most reported observed CDI ratios (as percentages), with CDI variously defined, or prevalence within specific patient subgroups or populations (Table 1). Figure 2 depicts reported CDI incidence from retrospective chart reviews in different patient cohorts, where CDI was defined based on diarrhea and laboratory detection of fecal C. difficile toxin.

Fig. 2
figure 2

Clostridium difficile infection (defined as diarrhea/CD toxin) reported in retrospective cohorts of Japanese patients. CRC colorectal cancer, GI gastrointestinal, HSCT hematopoietic stem cell transplantation, IBD inflammatory bowel disease, pts patients, RA rheumatoid arthritis. Patient numbers represent those diagnosed with Clostridium difficile infection

Full size image

A chart review of over 22,800 inpatients at a single tertiary care center (of whom 851 were tested for C. difficile) reported on healthcare-facility onset CDI, defined as diarrhea and a positive toxin test (using C DIFF QUIK CHEK COMPLETE®) [18]. The CDI prevalence was 5.5 cases/1000 patients. The incidence of healthcare-facility onset CDI was 3.11 cases/10,000 patient-days, compared with 0.2 cases/10,000 patient-days for community-onset CDI [18]. The authors considered the CDI incidence (hospital- and community-onset) to be rather low, which they suggested may have been attributable to the relatively low frequency of testing for C. difficile and the relatively low sensitivity of the EIA toxin detection method. Another large-scale outpatient study (n = 2193) reported the incidence of community-acquired CDI as 1.4/100,000 patient-years (or 0.14/10,000 patient-years) [19], and a study of both actively tested in- and outpatients provided an incidence of 0.8 cases/10,000 patient-days [20]. A retrospective cohort study based on chart reviews at four tertiary care hospitals reported 160 patients aged at least 14 years with hospital-onset CDI, as defined according to clinical practice guidelines, giving an incidence of 1.04/10,000 patient-days (or 1.61/1000 admissions) [21]. The low incidence of hospital-onset CDI compared with that reported in studies from Western countries was suggested by the authors to be a consequence of the different strains prevalent in different regions, with outbreaks of hospital-acquired CDI in Western countries being attributed to highly virulent strains that are not prevalent in Japan. The study did not explore the strains responsible for the CDI episodes at the four hospitals. The authors also suggested that the low frequency of CDI may be a consequence of the low frequency of testing for C. difficile [21].

Few studies identified in our search described measures to control the incidence of C. difficile in hospitals. Although not strictly an infection control measure, the prophylactic use of pre- and probiotics has been suggested to maintain the colonic microbiota and potentially reduce the development of CDI [22]. A single-center RCT of 379 patients undergoing colorectal surgery evaluated the impact of perioperative synbiotics (combination of pro- and prebiotics) on post-surgical outcomes and fecal microbiota composition, finding that patients administered synbiotics before surgery had a lower incidence of C. difficile in their fecal microbiota compared with control patients. The incidence of CDI was low, with only one patient in the control group developing CDI, while none in the treated group did so. The authors suggested a potential role for synbiotics in suppressing overgrowth of C. difficile after surgery [22]. The literature search also identified a study reporting the impact of infection control interventions on CDI occurrence. A medical record-based C. difficile-associated disease at a single center was reported to have an incidence of 0.47 cases/1000 inpatient days, which fell to 0.11 cases/1000 patient-days after intensive infection control intervention [23]. Infection control measures included carbapenem restriction and continuous instruction to the ward staff on infection control measures [23].

Some of the epidemiological reports highlighted the incidence and prevalence of CDI in particular patient groups (Table 1; Fig. 2). For example, in patients undergoing HSCT, the cumulative incidence of CDI was 6.2% for all patients, compared with 9.2% in the allogeneic HSCT subpopulation, 1.0% in the autologous HSCT subpopulation and 9% in a cohort of HSCT recipients who received unrelated cord blood [24, 25]. In a cohort of liver transplant patients, CDI-associated diarrhea occurred at a rate of 4.5% [26]. Among rheumatology inpatients, CDI was observed in 0.16% [27]; in a large cohort of patients with sepsis, hospital-acquired CDI was observed in 1.3% of patients without and 1.4% with ulcer prophylaxis [28]. A retrospective database review of over 140,000 gastrointestinal (GI) surgery patients reported CDI (ICD-10 definition) in 0.28% of the study population, or a prevalence of 2.8 cases/1000 patients [29]. In a smaller cohort study of patients with active ulcerative colitis, 40.1% tested positive for possible CDI [30]. A study of pediatric patients with cancer who were hospitalized at a single center reported CDI (clinical symptoms and positivity for toxin EIA using TOXA/B QUIK CHEK) in 27% of the study population [31]. The authors suggested that the high incidence of CDI compared with other studies of similar patient populations in non-Japanese settings may be because of the much longer length of stay for patients with cancer in Japan compared with other countries [31].

Among the studies that reported on possible C. difficile colonization of patient groups was an epidemiological study of hospitalized pediatric patients. At least 1 in 10 pediatric patients harbored C. difficile asymptomatically, while fecal cytotoxin was found in 9% of otherwise healthy children and 23.1% of children with underlying disease [32]. However, these findings should be interpreted with caution as the inclusion of patients younger than 3 years old may increase the likelihood that the diarrhea had a cause other than C. difficile [33].

The reviewed studies included reports of both hospital and community patients. No papers were identified that specifically related to patients in long-term healthcare facilities and only one paper reported on changes in infection over time [23]. Therefore, depending on the patient group and methods used to define and assess CDI, the incidence varied between 0.8 and 4.71/10,000 patient-days, while the prevalence was between 0.3 and 5.5 cases/1000 patients.

Risk Factors

Known risk factors for CDI include: co- and previously administered broad-spectrum antibiotics; age and comorbidities; poor infection-control practices; GI tract surgery; and gastric-acid suppressing agents [13, 34,35,36]. Thirteen studies in our search, including a large database study representing 40% of all adult-care hospitalizations, identified risk factors for CDI in Japanese patients, which included older age, higher comorbidity index; gastric acid-suppressing proton pump inhibitors (PPIs); and a longer pre-operative LOS before GI surgery [19, 20, 24,25,26, 29, 31, 32, 36,37,38,39,40] (Table 1). Malignant disease and intensive care unit (ICU) stay were linked with increased risk for CDI recurrence in in- and outpatients [20]. Six articles reported on the number of days spent in hospital prior to surgery or CDI diagnosis, indicating a wide variation in inpatient stay before diagnosis [18, 20, 29, 38, 41, 42]. The only study to formally compare pre-operative number of days in hospital for patients who did and did not develop CDI found no difference between the two patient groups, with a median (interquartile range) of 6 (3–14) days and 5 (3–8) days, respectively (p < 0.001) [29].

Patients undergoing HSCT are recognized to be at particular risk of infection. Allogeneic HSCT, conditioning for HSCT, acute leukemia and prolonged neutropenia in the first 30 days after HSCT may all confer an increased risk for CDI as reported in a single-center study [24]. In contrast, it was noted that among allogeneic HSCT patients, treatment with total body irradiation may reduce post-transplant risk of CDI [25].

A study of pediatric patients reported that tube feeding was significantly associated with higher colonization rates by toxin-positive C. difficile [32].

Antibiotic use was a risk factor for C. difficile diarrhea in a number of studies [19, 31, 37, 38]; however, in a study of liver transplant patients, the intensity of antibiotic use (measured as use of preoperative antibiotics or the number of antibiotics used postoperatively) was not a predictor for C. difficile diarrhea [26]. Among hospitalized pediatric patients with cancer, use of a wide variety of antibiotics (the study specified at least four different types) in the 60 days prior to CDI diagnosis was a significant risk factor for the development of CDI [31].

Specific Strains Responsible for CDI

A review of CDI in Asia in 2013 reported that PCR ribotypes 027 and 078 were rare, while variant toxin A−/toxin B+ strains of ribotype 017 were common. Furthermore, in Japan, common ribotypes include 014, 002 and 001, and ribotype smz/018 has been implicated in widespread disease [3]. The review noted that a variety of typing techniques has been used in Japan, including tcdA and tcdB detection, pulsed field gel electrophoresis (PFGE), PCR ribotyping and slpA typing. Although molecular typing had identified toxigenic A−B+ strains, the authors did not comment on binary (CDT) toxin assessment or C. difficile surveillance in Japan [3].

In our literature search, 16 papers provided further details of testing methods used in Japan and described reports on the isolates and strains associated with CDI in Japanese cohorts (Table 2). The methods used to detect CDI and/or detect, isolate and type C. difficile included stool culture and C DIFF QUIK CHEK COMPLETE®; stool culture, PCR ribotyping and slpA sequence typing; stool culture, PCR and PFGE; and rRNA-targeted RT-qPCR and multiplex PCR for toxin gene profiling. Reported methods of testing for C. difficile toxins included: toxin A testing by Uniquick (in 2004) [43]; PCR assessment for toxin A, B and CDT genes [44]; TOX A/B QUIK CHEK® [45]; and multiplex PCR [46]. Rapid detection methods evaluated and reported in Japan include detection of GDH and toxin A/B in feces (using Immunocard® and TOX A/B, respectively) [47]; and detection of GDH and toxin A/B simultaneously using C DIFF QUIK CHEK COMPLETE® [41]. Although mass spectrometry methods have also been reported, these were not considered suitable for typing C. difficile [48].

Table 2 Summary of studies describing C. difficile strains, test methods and assay for binary toxin in Japan
Full size table

Our literature review (Table 2) suggests that most C. difficile isolates associated with infection in Japan produce both toxins A and B. Available studies suggest a low prevalence of binary toxin-positive (CDT+) strains—between 0 and 6.8% reported across studies [19, 44, 46, 49,50,51]. A small study from 2015 found that outbreak and non-outbreak isolates were predominantly smz/ysmz (by slpA typing) and that of five binary toxin-positive strains, one was ribotype 027 and one 078 [51]. Individual papers appear to support the conclusion that ribotypes 027 and 078 are rare in Japan [44, 49, 50], and that smz/018, yok, 002, 014 and 369 (trf by slpA typing) are common [3, 42,43,44, 46, 49,50,51]. A recently published substudy of a global RCT that isolated 54 strains of toxigenic C. difficile (by EIA/PCR assays) from the stool cultures of 93 hospitalized patients at 35 sites in Japan appeared to corroborate this finding [52]. The most common PCR ribotypes were 052 (28% of isolates), 018 (19% of isolates), 002 (15% of isolates) and 369 (9% of isolates), and 052 was considered to be an ‘established’ strain, as it was widely distributed across Japan. Ribotypes 027 and 078 were not isolated in the substudy [52].

Recurrence

Twelve publications reported on CDI recurrence (Table 3) [18, 21, 24,25,26, 31, 39,40,41,42, 44], with rates in studies that included specific definitions of recurrence ranging from 3.3% in 30 HSCT patients [24] to 27.3% in a cohort of 11 liver transplant recipients with CDI [26]. The time period over which recurrence was defined or assessed varied widely from 14 to 365 days after the initial CDI episode [18, 21, 24, 25, 31, 44]. One retrospective chart review of 242 liver transplant recipients, 11 of whom developed CDI, reported recurrence in 2 of 8 patients who received vancomycin [26]; no analysis of risk was made in relation to the choice of treatment for CDI. A retrospective cohort study based on chart reviews of hospital-onset CDI cases at four teaching hospitals found that neither the severity of CDI nor adherence to clinical practice guidelines affected the risk of CDI recurrence [21]. In a retrospective study, multivariate analysis identified malignant disease (p = 0.03) and ICU hospitalization (p = 0.049) as risk factors for CDI recurrence within 8 weeks of the previous CDI episode [20]. A systematic review and meta-analysis of published studies reported that use of PPIs was significantly associated with recurrent CDI (pooled OR 1.73, 95% CI 1.39–2.15; p = 0.02) (Table 3) [40].

Table 3 Incidence of, and risk factors associated with, C. difficile infection recurrence in Japanese patient populations
Full size table

Healthcare Utilization

A large retrospective chart review of 22,863 patients by Honda et al. reported a median LOS among 126 CDI cases of 41.5 days (17.5 days before and 18 days post-CDI diagnosis) [18]. The sparsity of available LOS data did not allow any analysis of trends for CDI-related LOS over the years examined. The adjusted attributable LOS and costs related to CDI were reported from one large database study of 143,652 hospitalized patients [29]. There were 409 cases of CDI with the infection contributing to a LOS increase of 12.4 days (95% CI 9.7–15.0; p < 0.001) and a cost increase of US$6576 (95% CI 3753–9398; p < 0.001) compared with control patients who did not develop CDI. However, of note, this study included patients who had undergone specific surgical procedures and had CDI identified using diagnostic codes rather than by diagnostic tests [29], which could potentially lead to inaccurate identification of CDI.

Cases of CDI may require patient transfer to the ICU. The study by Honda et al. found that 9.5% of 126 CDI cases needed ICU admission [18]. There were 65 (51.6%) patients who were classified as having severe CDI, using the severity assessment score developed by Zar et al. [53]; three patients (2.4%) underwent CDI-related colectomy/diverting loop ileostomy because of critical illness or failure of medical therapy. Another retrospective chart review based on stool samples from in- and outpatients at a single hospital identified 76 patients with hospital-onset CDI, with three (3.9%) cases requiring ICU care [20].

Mortality

All-cause mortality within 30 days ranged from 3.4 to 15.1% [18, 20, 21, 29, 39]. Honda et al. reported that mortality was associated with an increased Zar severity score [18]. Of note, the Zar severity criteria will score a patient 1 point (2 points are defined as severe CDI) based only on age 60 years or older [53], and the median age of patients in the Honda et al. study was 78 years [18]. A multicenter retrospective cohort study of 160 hospitalized patients with CDI reported 30-day all-cause mortality of 13%, and found no significant difference between the mortality rate among patients with severe and non-severe CDI, or among those whose treatment adhered and did not adhere to clinical practice guidelines [21]. At 90 days, all-cause mortality among CDI cases was reported as 14.5% in one retrospective chart review [20]. Hosokawa et al. [25] concluded that, among patients who underwent unrelated cord blood transplantation, overall survival at 2 years was no different for those who developed CDI than for those who did not (42 vs. 46%, respectively; p = 0.77). One database study demonstrated that inpatient mortality was significantly higher in CDI patients than in those without CDI (3.4 vs. 1.6%; p = 0.008) [29], although the results of this study should be viewed with caution owing to the reliance on recorded diagnoses of CDI from administrative databases (which are less well validated than those in prospective cohorts or registries, for example), the inclusion of only patients undergoing GI surgery, and the loss of many patients from the propensity score-matched analysis. The use of vancomycin was associated with reduced mortality (OR 0.43; 95% CI 0.25–0.75) in a multicenter, case–control and cohort study of 1026 CDI patients [39].

Discussion

This systematic literature review identified 55 papers providing insights into the rates of CDI, patient groups affected and impact of CDI in Japan. Most of the studies were retrospective data reviews, and many focused on patients with suspected CDI, and the rates and prevalence of CDI in those groups. Fewer studies reported overall rates of hospital- and community-acquired CDI. Nevertheless, the current literature suggests that hospital-onset CDI in Japan occurs at an incidence of 0.8–4.7 cases/10,000 patient-days; lower than that reported in Europe for winter 2012–2013 (country range 0.7–28.7/10,000 patient bed-days), and similar to the US for hospital-onset CDI in 2013 (6.0/10,000 patient-days for laboratory-identified CDI and 4.4/10,000 patient-days for traditional surveillance-detected CDI) [54, 55]. However, direct comparisons between studies are difficult owing to differences in design, population size and detection methods used. The prevalence of CDI in Japan (0.3–5.5/1000 patient admissions) was lower than that in a US Veterans Health Administration (VHA) report from 2014, which recorded that from October 2010 to June 2012, CDI prevalence at admission was between 5.3 and 6.9/1000 admissions, in settings where 50% of hospitals used nucleic acid amplification tests [56]. An analysis of US VHA data for February 2012 to January 2014 reported a pooled CDI admission prevalence rate of 0.38/100 admissions [57].

CDI in Japan appears to have similar epidemiology to that in South Korea, although distinct from regions of South Asia [3]. There are a number of factors that contribute to the unique epidemiology of CDI in Japan. First, if testing for CDI is not conducted, this makes data-gathering on both the rate of testing and test results difficult. While there is a national surveillance program for infectious diseases in Japan, there is no national C. difficile screening program. Methods of testing are also an important consideration. Indeed, datasets from CDI surveillance programs highlight that country-to-country variations in CDI incidence largely reflect differences in surveillance methods and how rigorously CDI is investigated in testing strategies [54]. This literature review highlights that testing, typing and laboratory methods for assessment and diagnosis of CDI have evolved in Japan in recent years, with new methods continuing to be evaluated against established methods. However, some authors have noted that many reports of CDI epidemiology in Japan have relied on insensitive testing methods, such as EIA toxin tests [16]. While more sensitive nucleic acid amplification tests are recommended in clinical practice guidelines for the detection of C. difficile [7], these are not yet subject to reimbursement in Japan and are therefore not widely used, thus reliance on EIA remains common in clinical practice. It could be argued that a low detection frequency, as seen for CDI toxin tests in stool samples, may in fact reflect a low disease burden, as demonstrated by Shimizu et al., who reported that patients with positive EIA toxin tests in stools had more severe CDI than those patients with negative stool toxin tests who then had positive cultures [41]. It is recognized that CDI is under-diagnosed in many regions and countries because of a combination of absence of clinical suspicion and suboptimal laboratory diagnostics [54]. This may also be the case in Japan, although no papers in this search specifically focused on reporting under-diagnosed or missed CDI cases.

Increased LOS is both a risk factor for, and an outcome of, CDI [29]. A notable factor in Japan is the general tendency for a longer LOS compared with many other countries, which may have an impact on, and thus affect, CDI risk and rates [58]. Incidence of CDI is typically presented as per patient-days (i.e. patient-bed days or inpatient days) and it is important to understand the differences in overall LOS in different geographic regions when comparing incidence data. National statistics from Japan show that in 2014 mean LOS was 31.9 days for all patients and 41.7 days in those ≥ 65 years [59], whereas the comparable 2014 statistics from the USA were 4.6 days for all patients (not restricted to those with CDI), 5.2 days for patients aged 65–74 years and 5.3 days for those aged 75 years and over [60]. This suggests that if follow up was similar between nations, roughly eight times more patients would be included in the US study compared with a Japanese study to yield 10,000 patient-days as the denominator. Given the lengthy LOS in Japan, as in studies from other countries, rates of, and risks for, CDI in Japanese patients are increased by factors such as older age, presence of comorbidities, certain clinical/surgical procedures and exposure to drug therapies, including antibiotics [1]. We found that CDI recurrence similarly depended on a variety of factors, including clinical circumstances and settings, ranging from 3.3% in HSCT patients [24] to 27.3% in liver transplant patients [26]. We found very few data on risk factors for CDI recurrence, such as the choice of treatment for initial CDI. One study identified malignant disease and ICU hospitalization as risk factors for CDI recurrence [20], although the broad CIs cited for both (Table 3) suggest that analysis of risk factors in a larger population is required. As previously established [1], one systematic review and meta-analysis identified in our review revealed a strong association between PPI use and the risk for initial and recurrent CDI in both adults and children [40].

Most CDI cases in Japan are caused by strains of C. difficile that produce both toxins A and B. Recent reviews of CDI in Asia, which are supported by our literature review, showed that ribotype smz/18 (or 018) predominates in Japan [3] and that ribotype trf/369 is gaining prominence in the country [61]. Several of the studies we reviewed confirmed a low prevalence of ribotypes 027 and 078 in Japan. It has been postulated that the low prevalence of these highly virulent strains that cause sporadic outbreaks in Western countries may account for the low general incidence of CDI in Japan [21]. Further, we found that recent papers reporting on CDI strain and ribotype increasingly included a binary toxin assessment, although CDT+ C. difficile was rare in Japan [42, 46].

There are a number of limitations to this review. Most reports on CDI epidemiology in Japan included here were based on hospital cohort data, and were often (in 32 of 55 studies) single-site reports concerning either inpatients or patients discharged from hospital to the community. We did not find papers reporting on CDI rates in long-term care facilities, although, given the prolonged LOS reported in studies in our review, some hospitals may ‘replace’ long-term care facilities in some instances. There was a large degree of heterogeneity in relation to the definition of initial and recurrent CDI in the studies included in our review. This hampered our ability to evaluate the epidemiological data in a consistent manner and reduced the potential to draw robust conclusions. Although the papers reviewed spanned more than a 10-year period, there were no data describing local infection changes over time. However, we did find sporadic snapshots of CDI in certain patient subgroups and cohorts. Our review was further limited by a lack of information relating to the rate of CDI testing and no data on community-associated CDI.

Future studies and surveillance are necessary to gather data on the numbers and types of patients affected by, or at risk of, CDI in Japan. Furthermore, data on the clinical impact of CDI are important for management and resource planning, and to ensure optimal patient care and outcomes.

Conclusion

The current literature offers some insights into the evolving epidemiology of CDI in Japan, yet highlights a number of unresolved questions. Notably, heterogeneity in the CDI definitions in the studies we reviewed limited our ability to draw robust conclusions. With the availability of newer diagnostic tools and release of clinical practice guidelines for CDI, there is a need to undertake more comprehensive and coordinated studies and surveillance of both CDI cases and C. difficile isolates to map current trends, review the impact of infection control measures, increase knowledge of risk factors, and more fully understand the extent and impact of hospital-onset CDI in Japanese patient populations today.