The incidence and drug resistance of Clostridium difficile infection in Mainland China: a systematic review and meta-analysis

It has been widely reported that the incidence and severity of Clostridium difficile infection (CDI) have increased dramatically in North America and Europe. However, little is known about CDI in Mainland China. In this study, we aimed to investigate the incidence of CDI and the main epidemic and drug-resistant strains of C. difficile in Mainland China through meta-analysis of related studies published after the year 2010. A total of 51 eligible studies were included. The pooled incidence of toxigenic C. difficile among patients with diarrhoea was 14% (95% CI = 12–16%). In Mainland China, ST-37 and ST-3 were the most prevalent strains; fortunately, hypervirulent strains, such as ST-1 (BI/NAP1/027) and ST-11 (RT 078), have only occurred sporadically to date. The rates of C. difficile resistance to ciprofloxacin (98.3%; 95% CI = 96.9–99.7%), clindamycin (81.7%; 95% CI = 76.1–87.3%) and erythromycin (80.2%; 95% CI = 73.5–86.9%) are higher than in other counties; however, none of the C. difficile isolates reported in Mainland China were resistant to metronidazole (n/N = 0/960), vancomycin (n/N = 0/960), tigecycline (n/N = 0/41) or piperacillin/tazobactam(n/N = 0/288).


Results
Search results. We reviewed 7 electronic databases and identified 1,440 articles published from 2010 to 2016 ( Fig. 1). After initial evaluation of the titles and abstracts, 1,351 articles were excluded because of their irrelevance and duplication. The full text of the remaining articles was reviewed. Among the 89 articles, 38 were excluded again for specific reasons: 8 were reviews, 4 studies did not include data from 2010-2016, 1 study had fewer than 5 samples, 12 studies used the same samples, 5 studies didn't originate from Mainland China, 2 studies were repeated in different languages, and incomplete information was provided in 6 studies. Finally, 51 studies were included in this systematic review and meta-analysis. Of these, 39 articles focused on Topic 1, 16 focused on Topic 2 and 10 focused on Topic 3. Characteristics of the eligible studies. All   articles concerned Zhejiang; and 1 article each originated from Anhui, Gansu, Henan, and Ningxia. A total of 15,313 samples were collected to detect the existence of C. difficile and the associated toxins. Some 16 of the included studies were related to Topic 2, and multi-locus sequence typing (MLST) and PCR ribotyping (RT) were used to detect the molecular type of C. difficile. Ten articles were relevant to Topic 3. Antibiotic sensitivity testing (AST) were carried out with 3 methods differently; 6 used the agar dilution method; 3 used the E-test; and 1 study combined the agar dilution, disk diffusion and E-test methods.
The molecular epidemiology of C. difficile. We next reviewed the molecular classification methods, including MLST and RT, for C. difficile in the published articles. As shown in Table 2  The antibiotic resistance of C. difficile. A total of 24 drugs were reported in the included studies.
However, 14 drugs were investigated by several studies (Table 4), while the others were only examined in single studies. The results shown in Table 4     resistant to fidaxomicin, amoxicillin/clavulanic acid, ampicillin, piperacillin, or teicoplanin, and nearly all of the strains were sensitive to chloramphenicol and imipenem according to the published articles [23][24][25][26][27][28] . The strains reported in one article were resistant to ceftriaxone and cefoxitin 25,29 .

Discussion
Because only a limited number of studies have investigated C. difficile in Mainland China, some provinces were only investigated in one study and others were not included in the analyses. Although the data were not sufficient to provide an accurate conclusion, we still obtained valuable information through meta-analysis of related published data. In this study, we found that the incidence of toxigenic C. difficile in Mainland China was 14% (95% CI = 12-16%) (n/N = 2,132/15,313). Fortunately, highly virulent strains were rarely reported, although CDI outbreaks have been reported more than 100 times in North America and Europe since 2003 30 . From 2000 to 2009, the confirmed CDI cases were more than doubled from 139,000 to 336,600 in US hospitals, and the number of patients with a primary confirmed CDI cases tripled from 33,000 to 111,000 31 . Moreover, the incidence, mortality, and medical care costs of CDIs have reached historic highs in America 31 . However, Asian countries seem to have a low incidence of CDI. One survey conducted in Japan found only a small number of patients with serious CDI and recurrent CDI. Meanwhile, regarding the annual number of CDI patients at all of the hospitals that participated in this survey, the largest group of hospitals reported treating "1 to 5 patients a year, " representing 17.8% of the sample, and the second largest group of hospitals reported treated "no patients a year, " representing 13.1% 32 . In addition, the CDI incidence in Korea was estimated to be 5.06 per 100,000 population in 2011, whereas the estimated incidence in America per 100,000 population ranged from 30 to 120 cases of community-associated infection and from 50 to 160 cases of health care-associated infection across the Emerging Infections Program (EIP) sites in the same year 33    China 35 . Therefore, we re-analysed the articles concerning Ningxia and Henan. The CDI study in Henan only included 60 samples, and the authors detected 9 cases of C. difficile, but only 2 produced toxins. The investigation in Ningxia included 233 diarrhoea patients; of these, 162 cases were due to non-infectious diarrhoea and 71 cases were caused by pathogenic bacteria, where 9 C. difficile strains were assayed with the VIDAS CDAB kit. On the other hand, Anhui, Hebei, Hubei and Sichuan reported a much higher positive rate, which attracted our attention. These studies were reviewed, which showed that some risk factors were associated with these results. First, the diarrhoea patients included in these studies represented special groups; some diarrhoea patients came from the intensive care unit (ICU), some suffered from hematologic diseases, and some patients had used antibiotics for a long time. Moreover, some diarrhoea patients were diagnosed with antibiotic-associated diarrhoea (AAD). Thus, these results may increase the positive rates of CDI to a high level. In addition, 2,570 patients from 9 articles were diagnosed with AAD or confirmed to have used antibiotics, and this feature undoubtedly increased the rate of CDI in our survey. Notably, studies from Guangdong province included one-third of all the pooled nationwide cases approximately, which may have impacted the overall results of this study. Thus, reanalysis was necessary. We deleted the data from Guangdong and recalculated the mean incidence of CDI based on the data from the other 12 provinces. Finally, the recalculated incidence of CDI was 14% (95% CI = 11-16%) (n/N = 1,343/10,024). The same phenomenon was found from the studies of Zhejiang province. Therefore, we recalculated without the data from Zhejiang and got the same result of 14% (95% CI = 12-17%) (n/N = 1,732/11,583). Therefore, these changes did not produce a significant difference in the overall rate of CDI, in other words, the data from Guangdong and Zhejiang had a small impact on the nationwide average value of CDI incidence despite their big data weight. Based on the risk factors mentioned above, some researchers are worried about the incidence of CDI in Mainland China due to the growing elderly population and the well-recognized problem of over-prescribing broad-spectrum antibiotics in Mainland China 35 . The lifestyles of Chinese people, particularly those of young Chinese, have changed tremendously over the last 20 years, partially because of the rapid growth of the Chinese economy. These lifestyle changes, which were not as common in Mainland China in the past, may represent a major contributor to the increasing emergence of CDI 19 .  39,40 . However, the epidemic strains in Mainland China are not the highly virulent type of C. difficile, which may be the reason why CDI has not had an outbreak in our country in recent years. Additionally, the number of reports about toxin A-negative and toxin B-positive strains has gradually increased in recent years 41 . Many studies in Mainland China have found the same situation 42 . In Asia, the numbers of toxin A-negative and toxin B-positive strains are significantly higher than that in European and American countries, which may also explain why the prevalent strain in Asia is RT 017 26,43 and not RT 027. Nevertheless, rational tests and multi-centre or national-level surveillance for CDI, particularly for RT 027, should be introduced to provide essential data and guide future clinical practice 44 . RT 027 C. difficile isolates were detected in Beijing in 2012 and 2013, following the first discovery in Hong Kong 39,44 . Thus, it is time to stop neglecting CDI in Mainland China, and this problem should receive greater awareness.
CDI shows a close relationship with AAD, as described above. A number of international studies reported on the wide use of broad-spectrum antimicrobials, particularly the third generation of cephalosporin, the fluoroquinolone drugs and the clindamycin drugs, which subsequently increased the resistance to these drugs. Meanwhile, the balance of intestinal flora is destroyed following antibiotic treatment, which serves as a risk factor for C. difficile-associated diarrhoea (CDAD) [45][46][47][48] . The resistance rates to ciprofloxacin and cefoxitin were almost 100% in this study, which is consistent with these previous reports. According to the study by Pituch H et al., the rate of resistance to moxifloxacin was 38.5-40.1%, which is similar to that identified in this study (39.0%; 95% CI = 27.9-50.1%) 49 . Additionally, Ilchmann, C. and Oka, K. et al. found that the rates of resistance to clindamycin and erythromycin were more than 80% 50,51 , which is also similar to the results of our study ((81.7%; 95% CI = 76.1-87.3%); (80.2%; 95% CI = 73.5-86.9%), respectively). In recent years, a foreign survey reported the appearance of metronidazole-and/or vancomycin-resistant strains 52 . For example, Martin H et al. discovered that the rate of drug resistance to metronidazole and vancomycin was 1.8% and 0.4% respectively 53 . However, the articles included in this study did not identify metronidazole-and/or vancomycin-resistant strains. Therefore, the resistance profiles of C. difficile to these two drugs differ between Mainland China and other countries. In Mainland China, only heteroresistance to metronidazole was reported in Shanghai 54 , and this type of resistance only appears in the first generation of cultured strains and will disappear after several generations 55 . However, the epidemic ST-37 strain in Mainland China presented the highest resistance rate to many antibiotics, particularly clindamycin, tetracycline, and moxifloxacin, and all ST-37 isolates exhibited multi-drug resistance 56 . Some studies also indicated that the recurrence rate has increased in recent years following the use of metronidazole and/or vancomycin 57 . For example, the recurrence rates following metronidazole and vancomycin were reported at 25% and 60%, respectively 46 .
We realize that there are additional limitations of this study. First, we included few articles about C. difficile in Mainland China, particularly those reporting on the incidence and antibiotic resistance. Second, there is no standard test for detecting toxigenic C. difficile. Enzyme immunoassays and PCR were widely used to detect toxigenic C. difficile, even though the positive detection rate of toxigenic C. difficile was not statistically significant 58 . Third, the methods used to detect antibiotic resistance differed between studies. The agar dilution method and E-test were most commonly adopted in these articles. Moreover, the selection criteria for Topic 1 Scientific RepoRts | 6:37865 | DOI: 10.1038/srep37865 have disadvantages as well, as these criteria do not separate diarrhoeal patients into high-risk groups and low-risk groups, such as AAD and non-AAD. Moreover, studies published in local journals, which were not indexed in the electronic databases, might have been missed in this meta-analysis. Last, significant heterogeneity was observed between studies. For example, the incidence of toxigenic C. difficile among diarrhoea patients varied in different subgroups, such as outpatients and inpatients, ordinary inpatients and ICU patients. However, we did not obtain sufficient information about these subgroups for further analysis. Therefore, the prevalence and antibiotic resistance of C. difficile in Mainland China requires further study.
Despite the limitations listed above, there are some noteworthy findings of this study. First, the incidence of toxigenic C. difficile among diarrhoea patients in Mainland China was 14% (95% CI = 12-16%) (n/N = 2,132/15,313), and this rate may be slightly higher than the actual status because some patients included in Topic 1 had additional risk factors. Second, although RT 014/020 were the most common RT strains in Western Australia 8 , RT 017 (ST-37) was the major epidemic strain in Mainland China. The existence of RT 027 (ST-1) and RT 078  in Mainland China was only reported in a few studies, whereas these are the dominant strains in outbreaks in Europe and North America. Finally, the rates of resistance to erythromycin and clindamycin were greater than 80%, and the rates of resistance to ciprofloxacin and cefoxitin almost reached 100%. These results in Mainland China are consistent with other reports on C. difficile drug resistance in other countries. Most importantly, Mainland China has not detected the metronidazole-and vancomycin-resistant strains of C. difficile to date. Together, the information provided in this study will help guide rational drug use in the clinic and enhance the awareness of C. difficile infection and the epidemiological characteristics of these infections in Mainland China.

Methods
Search strategy. We searched the NCBI PubMed, Web of Science, Science Director, OVID, China National Knowledge Infrastructure (CNKI), Wanfang (Chinese) and Weipu (Chinese) databases to identify research studies that described C. difficile infections in Mainland China. The following keywords were used in the searches: "C. difficile", "prevalence", "incidence", "molecular epidemiology", "in Mainland China" and "antibiotic resistance. " The search was limited to publications from 2010 to 2016. The same strategies were used for each database. We placed no language restrictions on the searches or search results.

Selection criteria. Two investigators (Chenjie Tang and Lunbiao Cui) independently reviewed the poten-
tially appropriate studies to determine whether they met the predetermined eligibility criteria. Disagreements between the reviewers were resolved by consulting the other authors. Studies obtained from the literature search were checked by titles and abstracts. The titles and abstracts of the potential references were carefully scanned to exclude irrelevant articles. The remaining articles were evaluated to identify research that contained the relevant information, and the full texts were then reviewed in depth. In addition, the inclusion and exclusion criteria were established by the investigators prior to the literature review, and we also estimated and scored the relevance and quality of the references according to JBI (Joanna Briggs Institute). The inclusion criteria were as follows: Topic 1: C. difficile and its toxins were detected in stool samples from diarrhoea patients; Topic 2: C. difficile was detected by MLST or RT methods; and Topic 3: AST was performed on C. difficile. Moreover, studies were only included if they constituted original works and if the samples in these studies came from Mainland China. Only studies published within the last 6 years were included. Studies were excluded if they met the following conditions: (1) review or case report; (2) the data were not from 2010-2016; (3) studies with less than 5 samples; (4) samples from different studies were repeated; (5) the samples did not come from Mainland China; (6) the studies were repeated in different languages; (7) the information was not complete; and (8) the samples did not come from the diarrhoea patients included in Topic 1. Figure 1 shows the flowchart of the procedure used to select the articles. Data extraction. All data were extracted by two independent investigators. Disagreements in data extraction were resolved by reaching a consensus in accordance with the original study and the other authors' opinions. The following relevant data were extracted in three predefined tables. Details about the data extraction can be found in the Supplementary Tables 1, 2 and 3.

Statistical analysis.
Microsoft Excel (version 12.0) and Stata (version 12.0) were used in this meta-analysis.
Based on the possibility of significant heterogeneity, we used the Q-statistic (p < 0.05 was considered to indicate statistically significant heterogeneity) to estimate the heterogeneity between studies. A random effect model (REM) or fixed effect model (FEM) was chosen for the meta-analysis according to the p value. We calculated the incidence of toxigenic C. difficile among diarrhoea patients in the provinces of Mainland China, as well as the molecular epidemiology and antibiotic resistance using 95% confidence intervals (CIs) and a suitable model. Data manipulation and statistical analyses were performed using Stata 12.0(StataCorp. 2011. Stata Statistical Software: Release 12. College Station, TX: StataCorp LP.).