A worldwide systematic review and meta-analysis of bacteria related to antibiotic-associated diarrhea in hospitalized patients

Introduction Antibiotic-associated diarrhea (AAD) is a major hospital problem and a common adverse effect of antibiotic treatment. The aim of this study was to investigate the prevalence of the most important bacteria that cause AAD in hospitalized patients. Materials and methods PubMed, Web of Science and Scopus databases were searched using multiple relevant keywords and screening carried out based on inclusion/exclusion criteria from March 2001 to October 2021. The random-effects model was used to conduct the meta-analysis. Results Of the 7,377 identified articles, 56 met the inclusion criteria. Pooling all studies, the prevalence of Clostridioides (Clostridium) difficile, Clostridium perfringens, Klebsiella oxytoca, and Staphylococcus aureus as AAD-related bacteria among hospitalized patients were 19.6%, 14.9%, 27%, and 5.2%, respectively. The prevalence of all four bacteria was higher in Europe compared to other continents. The highest resistance of C. difficile was estimated to ciprofloxacin and the lowest resistances were reported to chloramphenicol, vancomycin, and metronidazole. There was no or little data on antibiotic resistance of other bacteria. Conclusions The results of this study emphasize the need for a surveillance program, as well as timely public and hospital health measures in order to control and treat AAD infections.

Introduction among clinical isolates of Klebsiella species is increasing. AAHC has also been reported after antibiotic therapy with quinolones and cephalosporins [21].
Staphylococcus aureus is also a less-known agent for AAD, often referred to as large-scale enterocolitis of watery diarrhea. From 1955 to 1970, S. aureus was suspected as the cause of AAD [28,29]. Increase in the prevalence of C. difficile in recent years has led to a lack of recognition of S. aureus as a cause of nosocomial infections and AAD [30]. In contrast to food poisoning, S. aureus AAD is a gastrointestinal infection that often occurs following antibioticinduced dysbiosis of the gut microbiota [7]. Evidence demonstrates that enterotoxin-producing strains of methicillin-resistant S. aureus may cause diarrhea associated with nosocomial antibiotics [31]. AAD-associated methicillin-resistant S. aureus strains have been reported in the blood of some patients, causing colitis to be a probable source of septicemia [32].
In early 1983, Holmberg et al. reported 18 patients with diarrhea due to multidrug-resistance Salmonella newport strains that were resistant to ampicillin, carbenicillin, and tetracycline. The source of infection was a hamburger eaten from an infected beef from cattle fed with subtherapeutic doses of chlortetracycline. Twelve of these patients had taken penicillin derivatives for medical complications other than diarrhea. According to reports, seems likely that these patients had an asymptomatic infection with drug-resistant S. newport before taking antibiotics for C. difficile and its toxins. Given the rarity of the disease, this is the only report in which Salmonella has been identified as an AAD agent [33].
In spite of the diagnosis importance of AAD-causing bacteria, it seems that less attention has been paid to these bacteria, in particular to C. perfringens, S. aureus and K. oxytoca. Also, a few number comprehensive reviews have been published in this field. Therefore, here we systematically reviewed all published articles on bacteria related to AAD in hospitalized patients from March 2001 to October 2021.

Search strategy
Three literature databases, PubMed, Web of Science, and Scopus were used to systematically identify studies of bacterial AAD. All studies that have been published from March 2001 to October 2021 were covered. The search terms were (Antibiotic associated diarrhea OR AAD) AND (diarrhea) OR (diarrhoea) AND (Antibiotic) AND ("Clostridioides difficile" OR Clostridium difficile OR C. difficile) AND (Clostridioides difficile associated diarrhea OR Clostridium difficile associated diarrhea) AND, ("Clostridium perfringens" OR C. perfringens) AND (Clostridium perfringens antibiotic-associated diarrhea), ("Staphylococcus aureus" OR S. aureus) AND (Staphylococcus aureus antibiotic-associated diarrhea), ("Klebsiella oxytoca" OR K. oxytoca) AND (Klebsiella oxytoca antibiotic-associated diarrhea). Manual searches were performed in the reference list of retrieved articles to identify more relevant papers. Duplicates were removed using EndNote X7 (Thomson Reuters, New York, NY, USA). The PRISMA guidelines were followed to perform the study [34].

Ethical statement
This systematic review and meta-analysis study was carried out with the Code of Ethics Committee No. IR.KUMS.REC.1398.017 adopted by Kermanshah University of Medical Sciences.

Inclusion/Exclusion criteria and data extraction
Any cross-sectional study was included in the analysis and clinical trials, case reports, narrative and systematic reviews and meta-analysis papers were excluded. The cross sectional articles were included if they met all the following criteria: hospital related studies, frequency or prevalence of C. difficile, C. perfringens, K. oxytoca, and S. aureus among AAD, published study based on English language only, reporting laboratory-confirmed bacterial AAD. Laboratory test included: Culture, Polymerase Chain Reaction (PCR), Enzyme immunoassay (EIA), reversed passive latex agglutination test (RPLA) and enzyme-linked immunofluorescent assay (ELFA). The exclusion criteria were: 1) insufficient information on bacterial AAD, 2) articles with similar titles published in different journals. The following information were extracted from included studies: observed study, publication year, sampling year, study country, sample size, antibiotics used, history of antibiotic, prevalence /frequency of AAD, diagnostic test, and antibiotic resistance. The antibiotics that had been used in studies were applied for a randomeffects model subgroup analysis. The Critical Appraisal tools of Joanna Briggs Institute (JBI) were used to perform the quality assessment (risk of bias) of each study [35].

Bacterial prevalence in AAD patients
In each study the number of bacterial isolates has been found by culturing, toxin detection (by immunoassay or molecular methods), or both approaches. Therefore, the pooled prevalence of C. difficile, C. perfringens, K. oxytoca, and S. aureus in AAD patients was assessed based on the detection method (culturing or toxin detection) or regardless the method (total). For all metaanalyses on bacteria prevalence, the random-effects model was applied and the proportion of bacteria cases over sample size was used as effect size. In the tables and text, the proportions were multiplied by 100 to present the results by percentages.

Limitations of study
Some limitations of this meta-analysis must be considered here. First, the distribution of studies on AAD related to bacteria is not uniform in continents and countries, and there was no published data from many countries. Second, the data concerning the age and sex of the patients as well as the exact antibiotics used or caused AAD were missing in many papers and could not be addressed or analyzed. Finally, just like many systematic review and meta-analysis papers, potential bias should be considered as a limitation.

Statistical analysis
All analyses were performed using of version 2.2.064 of Comprehensive Meta-Analysis software. The prevalence of C. difficile, C. perfringens, K. oxytoca, and S. aureus in AAD patients, and the proportion of antibiotic resistance of each bacterium were presented with 95% confidence intervals (CIs). The random-effects model was used to conduct the meta-analysis. Based on the sampling years and the region of each study, several subgroup analyses were performed to assess the source of heterogeneity. Cochrane Q and I 2 statistics were used to measure between studies heterogeneity. Considering the potential asymmetrical data distribution, Egger's linear regression test was used to evaluate any publication bias. The p-value <0.05 was accepted as the statistically significance threshold.

Study selection
A total of 7,377 articles were found. After applying screening and eligibility approaches, finally, 56 full-text articles were included on basis of our criteria (Fig 1). Continental distribution of the studies was as follow: 26 studies from Asia, 15 studies from Europe, 10 studies from South America, three studies from North America and two studies from Africa. The characteristics of the final included studies are represented in Table 1.    (Continued )

Bacterial prevalence in AAD patients
C. difficile prevalence in AAD patients. Regardless the detection method, 52 studies were applied for meta-analysis, in which the pooled prevalence of C. difficile in AAD patients was 19.6% (CI 95%: 15.1-25.1). Considering culturing method, 24 studies were applied for meta-analysis, in which the pooled prevalence of C. difficile in AAD patients was 17.4% (CI 95%: 12.6-23.7). Based on toxin detection methods, 42 studies were applied for meta-analysis, in which the pooled prevalence of C. difficile in AAD patients was 17.6% (CI 95%: 12.7-23.9) (Fig 2).
In all three meta-analyses on C. difficile prevalence in AAD patients the confidence intervals of summary effect did not include zero and the null hypothesis was rejected, meaning that there was a positive prevalence of C. difficile in AAD patients. Also, in all analyses, the Q-values were much more than the number of studies minus 1 (degrees of freedom) indicating a significant heterogeneity between studies. The I 2 statistics showed that a range of 95.5 to 97.4% of the variances in the observed effects is because of variances in the true effects (Fig 2).
C. perfringens prevalence in AAD patients. Eleven studies were used for meta-analysis regardless the detection method, in which the pooled prevalence of C. perfringens in AAD patients was 14.9% (CI 95%: 10.6-20.6). Five studies were applied for meta-analysis based on culturing method in which the pooled prevalence of C. perfringens in AAD patients was 17.9% (CI 95%: 11.3-27.1). Considering toxin detection methods, eight studies were applied for meta-analysis, in which the pooled prevalence of C. perfringens in AAD patients was 10.5% (CI 95%: 6.1-17.5) (Fig 3).  The confidence intervals of summary effect in all analyses did not include zero, hence rejecting the null hypothesis and showed that there was a positive prevalence of C. perfringens in AAD patients. The Q-values in all analyses were much more than degrees of freedom showing a significant heterogeneity between studies. The I 2 statistics showed that a range of 92.0 to 92.9% of the variances in the observed effects is because of variances in the true effects (Fig 3).
K. oxytoca prevalence in AAD patients. The pooled prevalence of K. oxytoca in AAD patients was 27.0% (CI 95%: 8.2-60.3) in four studies included, regardless the detection method. Based on culturing method, three studies were applied for meta-analysis, in which the pooled prevalence of K. oxytoca in AAD patients was 20.2% (CI 95%: 4.3-59.1). Regarding toxin detection methods, two studies were used for meta-analysis, in which the pooled prevalence of K. oxytoca in AAD patients was 27.2% (CI 95%: 4.7-74.1) (Fig 4).
Similar to other bacteria, there was a positive prevalence of K. oxytoca in AAD patients since the confidence intervals of summary effect did not include zero in all three meta-analyses. Also, the Q-values indicated a significant heterogeneity between studies. The I 2 statistics showed that a range of 89.5 to 94.9% of the variances in the observed effects is because of variances in the true effects (Fig 4).
S. aureus prevalence in AAD patients. In another meta-analysis using three studies and regardless the detection method, the pooled prevalence of S. aureus in AAD patients was 5.2% (CI 95%: 0.4-43.1). Considering culturing method, three studies were applied for meta-analysis, in which the pooled prevalence of S. aureus in AAD patients was 5.2% (CI 95%: 0.4-43.1). Regarding toxin detection methods, one study was applied for meta-analysis, in which the pooled prevalence of S. aureus in AAD patients was 1.1% (CI 95%: 0.5-2.2) (Fig 5).
The null hypothesis was rejected in all meta-analyses on S. aureus prevalence in AAD patients meaning that there was a positive prevalence of S. aureus in AAD patients. Also, in analyses regardless methods and that based on culturing, the Q-values were more than the degrees of freedom indicating a significant heterogeneity between studies. Only one study included in analysis based on toxin detection, thus the heterogeneity test was not applicable for it. Based on I2 statistics, 97.3% of the variances in the observed effects based on culturing and regardless the detection method is because of variances in the true effects (Fig 5). , and D5 (2016�). Based on these times, 38, 7, 4, and 3 studies were used for subgroup analysis on the prevalence of C. difficile, C. perfringens, K. oxytoca, and S. aureus in AAD patients, respectively. Subgroup analyses were done through the randomeffects model on the total prevalence of bacteria (regardless the detection methods). For C. perfringens, K. oxytoca, and S. aureus the studies were divided into only one group and other groups included no or only one study. Therefore the statistical comparison of the bacterial prevalence between time subgroups was not accurately applicable for these three bacteria. The prevalence of C. difficile was decreased after 2006 onward, although there was not significant heterogeneity between subgroups (Q-value: 4.808, p-value: 0.308) ( Table 2).

Subgroup analysis of bacterial prevalence in AAD patients based on the region.
To subgroup analysis of the bacterial prevalence in AAD patients based on the region, the studies were divided into five groups as follow: Africa, Asia, Europe, North America, and South America. Based on these regions, 52, 11, 4, and 3 studies were used for subgroup analysis on the prevalence of C. difficile, C. perfringens, K. oxytoca, and S. aureus in AAD patients, respectively. The prevalence of all four bacteria was higher in Europe compared to other continents, although there was not significant heterogeneity between subgroups ( Table 3).

Publication bias
To assess the publication bias, the prevalence of C. difficile in AAD patients regardless the detection methods was applied. The Egger's test showed a significant publication bias in the reports of C. difficile prevalence in AAD patients (p-value = 0.03).

Discussion
The studies on bacteria associated with AAD are limited. Therefore, here we collected all published data on bacteria associated with AAD including C. difficile, C. perfringens, K.oxytoca, and S. aureus.

PLOS ONE
C. difficile is known to be the most important cause of AAD in the world. Using various databases, we found 52 articles about the C. difficile AAD data. The pooled prevalence of C. difficile among hospitalized patients with AAD was 19.6%. These prevalence is similar to that was previously published in a systematic review and meta-analysis by Nasiri et al. (20%) [36], but it was a little different from study by Curcio et al. about C. difficile AAD in developing countries (15%) [8]. The prevalence of C. difficile AAD varies in different continents so that the highest prevalence of C. difficile AAD was in Europe (32.5%) and the lowest frequency was in Africa (13.5%).This difference can be attributed to various factors including large population migrations, and appropriate program monitoring about C. difficile AAD in Europe than other countries.
The prevalence of C. difficile has decreased after 2006 onward, which could be due to increasing in world health state, more proper prescription, and increasing in general awareness of the adverse effects of antibiotic overuse. However, inappropriate use of antibiotics is constantly continued nowadays, although the emergence of drug-resistant bacteria cause real concerns in the world. In order to prevent the spread of resistant isolates and bacterial infections, continuous monitoring of how the antibiotic resistance of bacteria appears is essential.
Based on the meta-analysis, the percentages of antibiotics resistance of C. difficile AAD was high for CIP and low for CHL, VAN and MTZ that were in concordant with Nasiri [36] and Ackermann [37] studies. Although, the frequency of resistance to first line antibiotics to treat AAD (MTZ and VAN) is still low, there is a concern to increase this rate in the future due to overuse and inappropriate use of antibiotics.
Accurate and on-time diagnosis of AAD-related bacteria assists in controlling of C. difficile transmission in communities and medical centers, as well as reducing the prevalence of AAD. There are several methods for identifying C. difficile AAD or their toxins, among them culture and ELISA are used mostly. These techniques were also the most frequent methods used in our included studies. More studies are needed to compare different techniques of detection of AAD-causing bacteria.
Due to the limited or scattered information obtained from the included studies, variables such as age, sex, and the exact prevalence of antibiotics used before AAD incidence could not be analyzed in our systematic review.
C. perfringens is another bacterium that can cause disease through hospital transmission. This bacterium is a part of the gut flora in healthy humans, however its colonizing and overgrown can cause severe AAD if allowed a longer period of growth [38]. Our study seems to be the first systematic review related to C. perfringens AAD, in which the mean frequency of C. perfringens among AAD hospitalized patients was 14.9%.
Distribution of the studies in patients with C. perfringens AAD according to continent was as follows: 9.4% in Asia, 19.5% in Europe, and 12% in Africa. No study of C. perfringens AAD was found in other parts of the world, which may be due to the limited number of published articles or the lack of its correct laboratory diagnostic methods.
Like C. difficile, culture and ELISA were the most diagnostic methods used to detect C. perfringens AAD or their toxins. These techniques are able to detect C. perfringens correctly if used appropriately by experienced technicians, although other methods may also be useful in detecting of this bacterium. Only three studies have been reported the prevalence of S. aureus in AAD patients, in them the pooled prevalence of S. aureus was low (5.2%). This prevalence is not very reliable due to the low number of published studies in this field. Likewise, routine diagnosis of S. aureus in AAD cases does not seem to be justified. Therefore, more studies are needed to find the true prevalence of S. aureus in AAD patients.
K. oxytoca, known to cause AAHC, is a distinct form of AAD. This pathogen acts as a pathobion in the human intestinal microbiota of dysbiotic and causes AAHC [39]. Our study was the first systematic review in K. oxytoca AAD. Using different databases, only four articles were finally included. The pooled prevalence was 27% and Europe (35%) and Asia (14%) were the highest and the lowest prevalence continents, respectively. Due to the limited information presented in these articles, we could not provide accurate information on K. oxytoca frequency in AAD patients.
We detected a significant heterogeneity between studies, showing that the bacterial prevalence in AAD patients is significantly different in various countries. This difference could be attributed to the quality of the studies, the sample sizes, the efficiency of diagnosis methods, or the true different distribution of bacterial causing AAD in different parts of the world.

Conclusion
Limited studies have been reported on the most important bacteria related to AAD in different countries of the world, which may be due to lack of proper laboratory diagnostic tests. The analysis of the studies indicated that K. oxytoca, C. difficile and C. perfringens are the most prevalent among hospitalized patients with AAD in the world. The prevalence of all four bacteria was higher in Europe compared to other continents. The highest resistance of C. difficile was estimated to ciprofloxacin and the lowest resistances were reported to chloramphenicol, vancomycin, and metronidazole. There was a little data on antibiotic resistance of other bacteria. Therefore, the results of this study emphasize the need for a surveillance program, as well as timely public and hospital health measures in order to control and treat AAD infections.