Antimicrobial resistance in Clostridioides (Clostridium) difficile derived from humans: a systematic review and meta-analysis

Background Clostridioides (Clostridium) difficile is an important pathogen of healthcare- associated diarrhea, however, an increase in the occurrence of C. difficile infection (CDI) outside hospital settings has been reported. The accumulation of antimicrobial resistance in C. difficile can increase the risk of CDI development and/or its spread. The limited number of antimicrobials for the treatment of CDI is matter of some concern. Objectives In order to summarize the data on antimicrobial resistance to C. difficile derived from humans, a systematic review and meta-analysis were performed. Methods We searched five bibliographic databases: (MEDLINE [PubMed], Scopus, Embase, Cochrane Library and Web of Science) for studies that focused on antimicrobial susceptibility testing in C. difficile and were published between 1992 and 2019. The weighted pooled resistance (WPR) for each antimicrobial agent was calculated using a random- effects model. Results A total of 111 studies were included. The WPR for metronidazole and vancomycin was 1.0% (95% CI 0–3%) and 1% (95% CI 0–2%) for the breakpoint > 2 mg/L and 0% (95% CI 0%) for breakpoint ≥32 μg/ml. Rifampin and tigecycline had a WPRs of 37.0% (95% CI 18–58%) and 1% (95% CI 0–3%), respectively. The WPRs for the other antimicrobials were as follows: ciprofloxacin 95% (95% CI 85–100%), moxifloxacin 32% (95% CI 25–40%), clindamycin 59% (95% CI 53–65%), amoxicillin/clavulanate 0% (0–0%), piperacillin/tazobactam 0% (0–0%) and ceftriaxone 47% (95% CI 29–65%). Tetracycline had a WPR 20% (95% CI 14–27%) and meropenem showed 0% (95% CI 0–1%); resistance to fidaxomicin was reported in one isolate (0.08%). Conclusion Resistance to metronidazole, vancomycin, fidaxomicin, meropenem and piperacillin/tazobactam is reported rarely. From the alternative CDI drug treatments, tigecycline had a lower resistance rate than rifampin. The high-risk antimicrobials for CDI development showed a high level of resistance, the highest was seen in the second generation of fluoroquinolones and clindamycin; amoxicillin/clavulanate showed almost no resistance. Tetracycline resistance was present in one fifth of human clinical C. difficile isolates.


Introduction
Clostridium difficile, recently reclassified as Clostridioides difficile [1], is an important pathogen of healthcare-associated diarrhea [2]. Recently, however, an increase in the occurrence of CDI outside hospital settings has been reported [3,4].
Previous antibiotic use was recognized as one of the risk factors for developing CDI through an alteration of gut microbiota. The accumulation of antimicrobial resistance mechanisms may provide an advantage to C. difficile as it is not affected by antimicrobials present in the gut [5].
An antibiotic stewardship intervention, that limited the use of the fluoroquinolones, clindamycin, amoxicillin/clavulanate, and cephalosporins, was shown to be effective in reducing the occurrence of multidrug-resistant epidemic ribotypes, e.g. 001 and 027 [6].
Currently, three antimicrobial agents, metronidazole, vancomycin and fidaxomicin are recommended for the treatment of CDI [7][8][9] and several new anti-CDI drugs are being tested in clinical trials [9]. The new data suggest tigecycline is effective in treating patients with a severe course of CDI [10], and rifaximin might be beneficial in preventing a CDI relapse [11].
In addition to humans, C. difficile has been cultured from livestock, food and the environment [12]. Tetracycline is one of the most commonly used antimicrobials in agriculture providing antimicrobial selective pressure in this sphere. This is supported by observations of a high prevalence of the tetracycline resistance gene tetM in livestock-associated C. difficile ribotype 078 isolates [13]. Moreover, the zoonotic transmission of C. difficile between farm animals and humans has been demonstrated [14].
Carbapenems are antimicrobials used for the treatment of infections caused by multidrug-resistant gram-negative pathogens. However, some carbapenem resistance mechanisms are transferable to other bacterial species [15]. Hence, the monitoring of carbapenem resistance in C. difficile is justified.
We aimed to review the data on the resistance of antimicrobials to C. difficile that have been recommended for CDI treatment; alternative drugs for CDI treatment; high-risk antimicrobials associated with CDI development; agriculture-related antimicrobials; and antimicrobials reserved for the treatment of multidrug pathogens.

Search strategy and study selection
Five bibliographic databases, including international databases (MEDLINE [PubMed], Scopus, Embase, Cochrane Library and Web of Science) were searched for relevant articles (Until October 2019) using the following keywords: ("Clostridium difficile" OR "Clostridioides difficile" OR C. difficile) AND ("Antimicrobial-Drug Resistance" OR "drug resistance" OR "antibiotic resistance" OR "aminoglycosides" OR "beta-lactams" OR "cephalosporins" OR "clindamycin" OR "tetracyclines" OR "fluoroquinolones" OR "macrolides" OR "vancomycin" OR "metronidazole" OR "fidaxomicin" OR "carbapenems") in the Title/Abstract/Keywords fields. No limitation was used while searching the databases, but for the study to be included in our analysis, the available abstract had to be written in English. The recorded hits were merged, and any duplicates were removed using EndNote X7 (Thomson Reuters, New York, NY, USA).

Selection criteria and data extraction
All selected studies were reviewed by three authors independently: Ebrahim Kouhsari, Behnam Ahmadzadeh and Abbas Maleki. Studies were excluded if they met the following conditions: (1) C. difficile antibiotic resistance was not presented; (2) resistance rates were not clearly reported; (3) no human clinical C. difficile strain was tested; (4) it was a meta-analysis and systematic review or a review article or not an original research article; (5) a duplicated report using the same database; (6) a conference abstract and article without the full text upon request from the author; (7) less than 5 isolates were tested. Any discrepancies and inconsistencies with the selection of an article were resolved through discussion, and a fourth author (Nourkhoda Sadeghifard) acted as arbiter.

Quality assessment
A quality evaluation of the included studies was performed independently (Behnam Ahmadzadeh, Ebrahim Kouhsari), using an adapted version of the tool proposed by the Newcastle-Ottawa assessment scale adapted for cross-sectional studies [16] (Supplementary Table 1). A score ranging from 0 to 8 points was attributed to each study (≥ 5 points: high quality, 4-3 points: Moderate quality, ≤ 2 points: low quality). A higher score indicated a higher study quality. A third reviewer (Leila Molaeipour) adjudicated in any cases where there was a disagreement.

Statistical analysis
Studies presenting raw data on antimicrobial resistance were included in the meta-analysis which was performed by computing the pooled prevalence of resistance for each antimicrobial agent using a random-effects model with Stata/SE software, v.14.1 (StataCorp, College Station, TX). The inconsistency across studies was examined by the forest plot as well as the I 2 statistic. Values of I 2 (25, 50 and 75%) were interpreted as the presence of low, medium or high heterogeneity, respectively and the random effects models were used [19]. Subgroup analyses were then employed by assuming continents, year, antimicrobial susceptibility testing, and the quality of studies as sources of variation. All statistical interpretations were reported on a 95% confidence interval (CI) basis.

Study outcomes
The main outcome of interest was the weighted pooled resistance rate (WPR) of strains resistant to specific antimicrobial agents according to the CLSI and/or EUCAST guidelines, respectively. A subgroup analysis was performed (1) for geographical regions (Asia, Europe, Africa, Oceania, South and North America); (2) publication date (1992-2014, and 2015-2019, 3) antimicrobial susceptibility testing method (agar dilution, Etest, and microbroth dilution); and (4) the quality of the studies (high quality, moderate quality, low quality). Subgroup analyses were not performed when the number of studies in the category was lower than five.

Search results
We evaluated six electronic databases and categorized 14,582 articles published up to October 2019 ( Fig. 1). From these, after an initial screening of the title and abstract, 11,204 articles were excluded, due to their irrelevance and duplication, but the full text of the remaining 335 articles was reviewed (Fig. 1). From the 335 articles, 224 were excluded again for the following reasons: review, not original research, conference abstract and article without full text (n = 162), no human clinical C. difficile strains (n = 24), no data for susceptibility testing or used disk diffusion method or no resistance data (n = 27), and data using the same isolates or low number of isolates (n = 11). Finally, 111 studies were included in this systematic review and meta-analysis (Supplementary Data 1). The studies included in the meta-analysis assessed antibiotic resistance to metronidazole, clindamycin, tetracycline, moxifloxacin and ciprofloxacin, meropenem, piperacillin/tazobactam, amoxicillin/ clavulanate, vancomycin, rifampin and tigecycline.

Characteristics of the included studies
The 111 included studies  were performed in 35 countries and investigated 19,733 C. difficile isolates. The majority of the studies originated in Asia (n = 42), followed by Europe (n = 37).
Epsilometer (E-test) strips were the most frequent antimicrobial susceptibility testing method used (n = 58), followed by agar dilution (n = 49). All studies had a cross-sectional design, and the mean Newcastle-Ottawa score was 4.5. The quality was high in 62 (55.8%) studies, medium in 46 (41.4%) studies, and low in 3 (2.7%) studies (Supplementary Data 1). Most of the studies (93.69%) included in the meta-analysis had determined the resistance to metronidazole.
The WPR rates for each antimicrobial are shown in Table 1 and  Table 2.

Resistance to metronidazole
The susceptibility to metronidazole was determined in 104 studies and included 19,645 C. difficile isolates.

Resistance to vancomycin
A susceptibility to vancomycin was determined in 94 studies where 15,515 C. difficile isolates were tested.
The resistance rates differ significantly when comparing the quality of studies (P = 0.01). In the low quality articles, the WPR was 6% (95% CI 2-11%) higher than in the moderate and high quality articles with a WPR of 2% (95% CI 0-2%). No statistical difference was found in the method used for AST (P = 0.47).

Resistance to moxifloxacin
A susceptibility to moxifloxacin was determined in 78 studies and from those studies 14,383 isolates were investigated.

Resistance to ciprofloxacin
The susceptibility to ciprofloxacin was determined in 28 studies investigating 4339 C. difficile isolates and used a breakpoint of 8 mg/L. From them, 3356 isolates were found to be resistant (77%); the WPR to ciprofloxacin was 95% (95% CI 85-100%) with a substantial heterogeneity (I 2 = 99.12%, P = 0.00).

Resistance to clindamycin
The susceptibility to clindamycin was determined in 64 studies investigating 19,645 C. difficile isolates and, using the CLSI breakpoint (8 mg/l), 6685 C. difficile isolates were reported to be resistant (34.0%).

Piperacillin/Tazobactam
The susceptibility to piperacillin/tazobactam was investigated in 17 studies applying the breakpoint of ≥128/4 mg/L mg/L and included 3041 C. difficile isolates. Eight isolates were found to be resistant (0.3%); the WPR to this antibiotic was 0% with (95% CI, 0-0%). No subgroup analyses were performed.

Fidaxomicin
The susceptibility to fidaxomicin was investigated in 1184 isolates from six studies. One isolate found to be resistant (0.08%) based on the breakpoint of ≥8 mg/L. The analyses were not performed because of the absence of a recommended breakpoint and the low number of studies.

Discussion
Due to the limited number of antimicrobials that can be used for the treatment [7][8][9] of CDI, it is important to obtain information about the resistance profiles of circulating C. difficile strains. Moreover, the accumulation of resistance mechanisms gives C. difficile an advantage since CDI can develop after the use of antimicrobials due to an alteration in gut microbiota [5].
Several methods can be used to determine the MIC in antimicrobial susceptibility testing. In our study, the Etest was the most used method followed by agar dilution. The agar dilution method is suitable for AST when there is high number of isolates since there is a need to prepare fresh testing plates for each experiment; however the commercially available Etest can be used independently for individual isolates.
Three antimicrobials are recommended for the treatment of CDI; metronidazole, vancomycin and fidaxomicin. For AST, there is still no MIC breakpoint available for fidaxomicin; for vancomycin and metronidazole two values exist but with a wide range: EUCAST 2 mg/L and CLSI 32 mg/L, The difference between the resistance rates, according to the breakpoint used, was also noted in our study. For metronidazole, the WPR was 1% (95% CI, 0-3%) using EUCAST but using CLSI, the WPR was 0% (95% CI, 0-0%). A similar pattern was also observed for vancomycin where using the EUCAST breakpoint, the WPR was higher (1% (95% CI 0-2%) than for the CLSI breakpoint 0% (95% CI, 0-0%).
Recently, a systematic review and meta-analysis [131] of metronidazole and vancomycin resistance in C. difficile showed higher WPRs than observed in our study; 1.9% (95% CI, 0.5-3.6%) for metronidazole and 2.1% (95% CI, 0-5.1%) for vancomycin. The analyses differed in the date of publication for data collection, (1982-2017) vs (1992-2019), and in the origin of the isolates since, in our analyses, the data on the C. difficile isolates of animal origin were not included.
The data on the susceptibility testing for metronidazole, vancomycin and moxifloxacin were included in the "enhanced level" of a CDI surveillance protocol published by the European Centre for Disease Prevention and Control (ECDC) [2]. Moxifloxacin, a fluoroquinolone, is not considered as a drug for CDI treatment but moxifloxacin resistance in C. difficile strains was shown to be an important marker for the spread of C. difficile in a healthcare setting [132]. Two representatives of fluoroquinolones were analysed in our study: ciprofloxacin and moxifloxacin. From all the antimicrobials in our study, ciprofloxacin showed the highest level of resistance (WPR 95%) and the resistance to moxifloxacin was 32 and 49% according to the CLSI and EUCAST breakpoints, respectively.
In addition to fluoroquinolones, clindamycin, amoxicillin/clavulanate and cephalosporins are indicated for limited use in hospital settings in order to reduce CDI rates [6]. From these four classes of antimicrobials, three classes exhibited high rates of resistance; however with amoxicillin/clavulanate, only 4 isolates out of 2803 isolates were investigated.
Rifaximin has been suggested as an alternative to existing CDI therapies, especially in CDI recurrences and their prevention [133,134]. Data on rifampin resistance, which correlate with rifaximin [135], showed a high level of resistance in investigated C. difficile isolates (787/ 1861) and suggest more risk to treatment failure due to C. difficile strain resistance compared to recommended CDI treatments.
The effectiveness of tigecycline use in the treatment of CDI was evaluated in several studies [10]. According to the reported resistance rates in our study, treatment failure is less likely with tigecycline than with rifaximin. However, recently, the emergence of mobile tigecyclineresistance genes, tet(X3) and tet(X4) that inactivate all tetracyclines, including tigecycline, was reported recently in gram-negative bacteria [136]. Moreover, Tet proteins have, in vitro, the potential to acquire mutations leading to an increased MICs for tigecycline [137]. From the available data, the tet classes of ribosomal protection genes are the most common molecular mechanism for tetracycline resistance in C. difficile [13]. The spread of newly detected tet(X) genes or mutations in present tet classes genes (e.g tetM or tetW) could increase the prevalence of resistance to tigecycline.

Conclusion
A resistance to metronidazole, vancomycin, fidaxomicin, meropenem and piperacillin-tazobactam is reported rarely. From alternative CDI treatment drugs, tigecycline had a lower resistance rate than rifampicin. The highrisk antimicrobials for CDI development showed a high level of resistance, the highest was seen in the second generation of fluoroquinolones and clindamycin; amoxicillin/clavulanate showed almost no resistance. Tetracycline resistance was present in one fifth of human clinical C. difficile isolates.

Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. We thank the Ilam University of Medical Sciences (A-10-2575-2) for financial support.
Availability of data and materials All data were included.

Consent for publication
Not applicable.