Acquisition of multidrug-resistant Enterobacterales during international travel: a systematic review of clinical and microbiological characteristics and meta-analyses of risk factors

International tourism increased from 25 million tourist arrivals in 1950 to over 1.3 billion in 2017. These travelers can be exposed to (multi) resistant microorganisms, may become colonized, and bring them back home. This systematic review aims to identify the carriage rates of multidrug-resistant Enterobacterales (MDR-E) among returning travelers, to identify microbiological methods used, and to identify the leading risk factors for acquiring MDR-E during international travel. Articles related to our research question were identified through a literature search in multiple databases (until June 18, 2019) - Embase, Medline Ovid, Cochrane, Scopus, Cinahl, Web of Science, and Google Scholar. Out of 3211 potentially relevant articles, we included 22 studies in the systematic review, and 12 studies in 7 random-effects meta-analyses. Highest carriage rates of MDR-E were observed after travel to Southern Asia (median 71%), followed by travel to Northern Africa (median 42%). Carbapenemase-producing Enterobacterales (CPE) were identified in 5 out of 22 studies, from a few patients. However, in only eight out of 22 studies (36.4%) the initial laboratory method targeted detection of the presence of CPE in the original samples. The risk factor with the highest pooled odds ratio (OR) for MDR-E was travel to Southern Asia (pooled OR = 14.16, 95% confidence interval [CI] = 5.50 to 36.45), followed by antibiotic use during travel (pooled OR = 2.78, 95% CI = 1.76 to 4.39). Risk of acquiring MDR-E while travelling increases depending on travel destination and if antibiotics are used during travel. This information is useful for the development of guidelines for healthcare facilities with low MDR-E prevalence rates to prevent admission of carriers without appropriate measures. The impact of such guidelines should be assessed.


Introduction
Multidrug resistance, defined as acquired nonsusceptibility to at least one agent in three or more antimicrobial categories, of clinically important bacteria is recognized as a major threat for human health worldwide. However, remarkable geographical differences in prevalence and trends exist [1,2]. Multidrug-resistant (MDR) Enterobacterales (MDR-E) that produce extended-spectrum beta-lactamases (ESBL) and/or carbapenemases are of most concern, since these bacteria are able to colonize the human gut and may cause a variety of infections that subsequently require more complicated treatments [3,4]. Fecal colonization rates with ESBL-producing Enterobacterales (ESBL-E) are estimated to be 14% among healthy individuals worldwide, with an annual increase from 1990 to 2015 of around 5% [5]. These rates are higher in the Mediterranean, the West Pacific, Africa and South-East Asia, and lower in Northern Europe and North America [5]. The carriage rate of carbapenemase-producing Enterobacterales (CPE) in healthy individuals is estimated to be low or absent, although only few studies have included healthy people in a community setting. In East-London, 200 community stool samples were screened and no CPE was identified [6]. CPE are mostly seen in people with exposure to healthcare [7]. In some countries, however, CPE are widespread in the environment [8,9].
When people travel from low-prevalence areas to areas with a higher prevalence of ESBL-E or CPE in the community, such strains may become part of their gut flora and then carried to the travelers' home country. This risk is on the rise, since international tourism has increased from 25 million tourist arrivals in 1950, to 1326 million tourist arrivals in 2017 with an expected annual growth of 3.3% [10,11]. This means that by 2030, 1.8 billion tourist arrivals will be reported. Additionally, between 2015 and 2016, travel to Oceania, Africa and South-East Asia increased the most, by 9.4, 8.1 and 7.8% respectively [10].
Tängdén et al. first reported on travel and acquisition of ESBL-E in 2010 [12]. Since then, numerous reports and several systematic reviews have been published on the relationship between fecal colonization with MDR-E and international travel [13,14]. However, for healthcare settings, especially those with a low prevalence of MDR-E, it is still unclear how to translate this knowledge into policies or guidelines for infection control and patient care. In addition, it is unclear how travel clinics or general practitioners can use the existing information for pre-travel advice. This review adds to the existing literature by performing an extensive systematic review to describe carriage rates of MDR-E among returning travelers, to describe microbiological methods used, and to perform a meta-analysis in order to identify the leading risk factors for acquiring MDR-E during international travel.

Methods
This systematic review and meta-analyses followed the guidelines presented in the PRISMA statement (see Additional file 1) [15]. Moreover, this study is an update and extension of the study published by Hassing et al. (Prospero registration number CRD42015024973), whose database search was conducted on August 17, 2015 [16].

Study selection
Articles related to our research question were identified through a literature search in multiple databases (until June 18, 2019) ─ Embase, Medline Ovid, Cochrane, Scopus, Cinahl, Web of Science, and Google Scholar (see Additional file 2). The search was not limited by language, date of publication, country of publication or study design.
We used the following inclusion criteria during the study selection: (i) related to foreign travel, (ii) reports on systematic and selective screening for the carriage of ESBL-E and/or CPE among travelers without signs of infections when performing the screening, and (iii) report on fecal Enterobacterales carriage. We excluded studies related to nonhuman infections, hospital studies, studies about symptomatic patients (e.g., travelers' diarrhea [TD]), conference abstracts, letters to the editor, commentaries, weekly reports, and editorials. First, titles and abstracts of all retrieved citations were screened independently by KM and AFV. After this screening, KM, AFV and BB performed a second screening based on the full-text. Disagreements were resolved by discussion. Reference lists of reviews and systematic reviews on the same subject, which were identified during the literature search, were screened to identify additional studies that had been missed by our search strategy.

Data extraction
We designed a data extraction form and pilot-tested it on two randomly selected articles, and redefined it according to the outcomes. The following data were extracted by AFV and BB: first author, journal, year published, country, study design, study period, where were the participants recruited (e.g. travel agency, vaccination clinic), total number of participants, mean age, percentage female, mean duration of travel, sample method, microorganism(s) studied, co-travelers or household members included, percentage of carriage before and after travel, acquisition rate, acquisition rate to household members, acquisition rates for each United Nations geographical region, laboratory methods (e.g. species determination, phenotypic approaches, molecular approaches), risk factors and protective factors identified in multivariable models; and corresponding odds ratio, 95% confidence interval and P-value. The completed data extraction form was sent to the corresponding author of the original manuscript to verify the extracted data, and to gain additional information if relevant. In case we did not receive any response after the given deadline (i.e. 2 weeks), a reminder was sent. If no response was received and crucial information was missing, the study was excluded.

Data analysis
Carriage rates of multidrug-resistant Enterobacterales by travel destination The following geographical classification was used, based on the United Nations geographical regions: (i) Southern Asia, (ii) Asia except Southern Asia, (iii) Northern Africa, (iv) Sub-Saharan Africa, (v) South and Central America, (vi) North America, (vii) Europe, (viii) Oceania (see Additional file 3). Carriage rates immediately after return were grouped into the following 5 categories, and for a visual presentation of the results a color was added from green to dark red: (i) 0-20%, low, green; (ii) 21-40%, moderate, yellow; (iii) 41-60%, high, orange; (iv) 61-80%, very high, red; (v) 81-100%, extremely high, dark red.

Meta-analysis
All risk factors extracted from the articles for which an odds ratio (OR) with 95% confidence interval (95%CI) was reported were grouped into eight categories: (i) diarrhea during travel, (ii) antibiotic use during travel, (iii) travel to Southern Asia, (iv) behavior during travel, (v) food consumption during travel, (vi) length of stay, (vii) sex, and (viii) age.
The meta-analyses for each category were performed using StatsDirect statistical software (Altrincham, United Kingdom) including the random-effects model of DerSimonian and Laird [17]. We used a random-effects model to limit the influence of heterogeneity. A P-value of < 0.05 was considered statistically significant. Publication bias was examined visually with use of funnel plots, and assessed with the Egger and Begg-Mazumdar indicators [18,19].

Study quality
The methodological quality was assessed for all included studies using the strengthening the reporting of observational studies in epidemiology (STROBE) guideline [20]. Studies with a score ≤ 15 out of 33 points were considered to be of relatively low methodological quality, studies receiving a quality score of 16-19 points were considered as of moderate quality, and studies with ≥20 points were considered to have a relatively high study quality. Study quality was not considered an exclusion criterion.

Results
The literature search identified 3211 non-duplicate articles, of which 811 articles were potentially relevant for this current study (Fig. 1). Titles and abstracts of these 811 articles were screened, which resulted in the exclusion of 795 articles (98%). One additional study was included after searching the reference lists of reviews of interest. The remaining 17 articles underwent a second screening based on the full-text, after which 6 articles were excluded (35.3%) (Fig. 1). The remaining 11 studies were added to the 11 previously identified by Hassing et al., and used in his review. Hence, in total 22 articles were included in this systematic review [16]. Of the 22 included articles, 12 were included in random-effects meta-analyses ( Fig. 1).
For the 22 included studies, the corresponding author was contacted to provide feedback on extracted data or to request additional information. The corresponding authors of all studies (100%) responded to our request to provide feedback on the extracted data. For three studies, we requested additional information on the multivariable analysis since crucial information was missing. Unfortunately, additional information was not received and these studies were therefore excluded from the random-effects meta-analyses.

Study characteristics
All 22 included studies were prospective cohort studies. The studies were conducted in Western Europe (n = 17, 77.3%), North America (n = 2, 9.1%), Japan (n = 2, 9.1%), and Australia (n = 1, 4.5%). The characteristics of the studies are shown in Table 1. Fifteen studies investigated travelers visiting a travel or vaccination clinic, one study investigated hospital staff and contacts, one study investigated healthcare students, one study investigated business travelers, one study investigated Hajj pilgrims, and two studies did not state the type of study population. The studies by van [36,37].
The median study size was 160 participants, ranging from 18 to 2001 participants, and the median age of participants ranged from 25 years to 66 years, with participants included from 0 to 84 years old. The median proportion of women was 61%, ranging from 26 to 78%; one study did not report the gender of the participants. Most studies had participants with a similar median duration of travel, ranging from median 14 to 21 days. However, the study investigating healthcare students had a median duration of travel of 45 days (13-365) [28] and in the study investigating business travelers, participants travelled for at least 6 months [33]. Sample collection in the included studies was via stool sample (n = 14, 63.6%), rectal swab (n = 6, 27.3%), a rectal swab or a stool sample (n = 1, 4.5%), or a rectal or perianal swab (n = 1, 4.5%). Four studies also investigated co-travelers in addition to the study population [12,29,36,37]. Three studies did not report on identification of microorganisms, and one study investigated a variety of antimicrobial-resistant bacteria: ESBL-E, CPE and colistin-resistant Gram-negative bacteria. Between the latter two groups, Enterobacter cloacae and Escherichia coli were the most frequently identified Enterobacterales, respectively. In all other studies, ESBL-positive E. coli was the most dominant MDR-E identified in post-travel samples.
Only one study clearly defined TD as more than 3 loose/liquid stools per 24 h or more frequently than normal for an individual [29] and one study referred to the World Health Organization (WHO) definition [42]. The other included studies just asked in questionnaires if the participant had experienced TD yes or no.

Microbiological methods
Enrichment was used in 12 out of 22 included studies (54.5%), all with a different composition, and 21 out of 22 included studies (95.5%) used selective agar plates, with 13 studies using ChromID ESBL ( Table 2). The method most often used for antimicrobial susceptibility testing was the VITEK2 system (bioMérieux, 10 out of 22, 45.5%), and phenotypic confirmation of ESBLproduction was most often performed by disk-diffusion (12 out of 22, 54.5%) ( Table 2).  In eight out of 22 studies (36.4%) the initial laboratory method targeted detection of the presence of CPE in the original samples, i.e. a selective agar plate method without a pre-enrichment with a broth containing second or third generation cephalosporins (Table 2) [26, 29-31, 35, 36, 39, 42]. In only 4 of these eight studies, a CPE was found [26,31,36,42]. In 13 out of the 22 studies (59.1%), screening for CPE was carried out in isolates or directly on the specimen by a variety of phenotypic and genotypic methods ( Table 2). In other studies, carbapenem susceptibility testing was partly or not performed, or isolates with reduced susceptibility to carbapenems were not further analyzed. Therefore, in those studies CPE could have remained unidentified. Table 3 shows the rates of travelers acquiring CPE, or MDR-E (i.e. without CPE) during international travel. It was not possible to report ESBL-E separate from MDR-E, since multiple studies, when reporting prevalence rates, combined ESBL-E and for example AmpCproducing Enterobacterales. Highest carriage rates of MDR-E were observed after travel to Southern Asia (Table 3), with proportions ranging from 29 to 88%, with as median 71%. Second was travel to Northern Africa, with proportions ranging from 31 to 100%, with as median 42%. CPE carriage was only identified in 5 studies [26,31,34,36,42], and mainly found in E. coli. Carbapenemases identified were IMI-2, NDM, NDM-1, NDM-1/ 2, NDM-7, OXA-48, OXA-181 and OXA-244.

Household members
Acquisition of MDR-E from the traveler to a nontravelling household member was described by 3 studies [25,36,37]. In the study by Paltansing et al., 1 out of 11 (9.1%) household members carried the same ESBLproducing E. coli as the traveler [25]. In the study by Arcilla et al., 13 out of 168 (7.7%) household members carried a microorganism with the same ESBL group as the traveler [37]. When the authors estimated the transmission rate after introduction into a household using a Markov model, the probability of transmission was 12% (95% CI = 5 to 18%) [37]. In the study by van Hattem et al., acquisition of a bla OXA-244 -positive E. coli from a traveler to a household member was highly suspected. Three months after travel a bla OXA-244 -positive E. coli with a similar AFLP pattern was isolated from a fecal sample from a spouse and travel companion. All other fecal specimens from this household member were CPE negative, which suggested post-travel acquisition of the same bacterium [36].

Protective factors and risk factors
We identified 12 studies describing protective and/or risk factors for acquiring MDR-E during travel, obtained from multivariable analyses (see Additional file 4). The highest OR was reported for the risk factor TD (OR = 31.00, 95% CI = 2.70 to 358.10). Examples of identified protective factors were handwashing with soap before meals, a beach holiday, tap water consumption, and travel to various countries compared to traveling to Asia (see Additional file 4).
All identified risk factors, protective factors, and factors identified as non-significant in multivariable models were grouped into 8 categories. For these 8 categories, 7 meta-analyses were performed. For the category length of stay, three factors were identified. However, for one factor the confidence interval was missing. Therefore, for the category length of stay no meta-analysis was performed. The factor antibiotic use during travel was most frequently described (n = 13 times identified).

Summary of evidence
We identified that when travelling to Southern Asia (i.e. Afghanistan, Bangladesh, Bhutan, India, Iran, Maldives, Nepal, Pakistan and Sri Lanka) people are at highest risk of acquiring and carrying a MDR-E upon return. Though, acquisition of MDR-E when visiting Northern Africa or Asia except Southern Asia was also high (Table 3). Additionally, we showed that acquiring a CPE while travelling is still rare, which is supported by the findings of Jans et al. [45]. However, it should be emphasized that in most studies a culture method was used that was not specifically targeting CPE. Especially CPE with OXA-48-like carbapenemases may be missed [36]. The risk factors for acquiring MDR-E in order of those with the highest to those with the lowest pooled OR are: (i) travel to Southern Asia; (ii) antibiotic use during travel; and (iii) TD. Older age, sex, food consumption during travel and behavior during travel were found to be non-significant (Fig. 2). With this systematic review, we aimed to provide aggregated data on acquisition of MDR-E and risk factors for MDR-E acquisition during international travel, which can be useful for the development of guidelines and policies in areas with a low prevalence of MDR-E. Travel to Asia, especially to India, is a known high risk for acquiring MDR-E. MDR-E are highly prevalent in this area because of the overuse of antibiotics, the lack of (clean) toilets and the lack of clean water. Hereby, bacteria can become resistant and are easily spread between people and to the environment. TD is associated with contaminated food or water, and is related to the lack of hygiene and sanitation [46]. Bacteria are responsible for the majority of cases [46]. TD in combination with antibiotic use does not only increase the risk of acquiring MDR-E, but also selects for antibiotic resistant Fig. 2 Forest plots of random-effects meta-analyses of risk factors for acquiring multidrug-resistant Enterobacterales during international travel (a to g appear from top to bottom). (a) Experienced diarrhea while travelling (i.e. TD); (b) antibiotic use during travel; c) travelled to Southern Asia; (d) behavior during travel (e.g. brought disposable gloves, consumed bottled water); (e) food consumption during travel (e.g. ice cream and pastry consumption, meals at street food stalls); (f) male gender; (g) older age bacteria [46][47][48][49]. Our results in combination with the studies by Kantele et al. highlight the need to avoid antibiotic use in mild to moderate TD [47][48][49]. Because most diarrheal episodes are self-limiting, it is only important to avoid dehydration [46]. Additionally, only one study used a clear definition of TD [29]. As described by Lääveri et al., the impact of the definition of TD is substantial on the results and conclusions [50].
Interestingly, food consumptiona known risk factor for TD and thus acquiring MDR-Ewas identified as non-significant. It may be that, because of all warnings and available guides, people are aware of the risks and stopped eating food from street vendors, raw food, and stopped drinking tap water, milk from open containers and fountain drinks [51]. Alternatively, it is also possible that recall bias played a role in the questionnaires' outcomes, and food consumption was rarely identified as a risk factor because people unknowingly eat risky food. In our opinion, this is more likely, as it may be difficult for travelers to determine if establishments adhere to food safety standards [52,53].

Towards a guideline
Although in our opinion the aggregated data do support the implementation of additional recommendations that can be given by travel clinics and general practitioners to people before travelling, there are still a number of knowledge gaps that need be filled before national and international guidelines on infection control (screening and/or isolation) and patient care (adjustment of empiric treatment) for healthcare facilities can be developed. First, the proportion of people with recent travel history to a foreign country with increased risk of MDR-E acquisition among patients admitted to hospitals is currently unknown. Second, it is unknown whether the strains that are carried by travelers do spread in hospitals, although it is known that in general ESBL-E and CPE can be transmitted between patients and into the hospital environment, especially when contact precautions are not taken, which can lead to outbreaks [54]. The fact that not only strains, but also resistance genes on mobile genetic elements such as plasmids can spread, makes this knowledge gap even more difficult to resolve. The cost-effectiveness of a program that would include screening and subsequent isolation of recent travelers can therefore not be estimated with the currently available data, nor can the overall impact of such a program on healthcare workers, laboratories and patients. The threshold of a carriage rate after travel that warrants screening and/or isolation is also an unresolved issue, but is likely to be dependent on the local carriage rates. For example, when travelling to Sub-Saharan Africa, 17% of travelers acquire ESBL-E. For the Netherlands, a country with a carriage rate in the community of 5.3 to 9.9%, 17% can be considered as high [54]. However, for example in countries with higher community carriage rates, other approaches may be more applicable. Such policies would also require systematic surveillance of carriage rates amongst travelers, or of local carriage rates. The burden of disease of travel-related MDR-E is also unknown. Follow-up data on infections in travelers is scarce, as is data on phylogenetic groups (PG) of E. coli and virulence factors in general. The limited available data suggest that infections are rare, clones may belong to low-virulent sequence types, and the PG varies between studies [12,25,26,30,35,40,55]. Third, most studies were performed in Europe and included travelers who visited a travel clinic. Therefore, just a few studies included travelers who visited Europe. In addition, few travelers to North America or Oceania were included in the studies in this review, possibly due to travelers not seeing a travel clinic when visiting these continents. Travelers visiting friends and relatives abroad are also underrepresented, since they usually do not seek health advice in a travel clinic before travelling. Additional studies are needed to assess the risks of these groups of travelers.

Strengths and limitations
A strength of our study is that it is an extensive literature search. In addition to the systematic review by Hassing et al., we performed meta-analyses to identify the main risk factors, looked more into detail to the laboratory methods and subsequent possibility to identify CPE and identified knowledge gaps. In addition, we performed an in-depth analysis about carriage rates (e.g. carriage rates of travelers, acquisition to household members and persistence of carriage).
This study has some limitations. First, the heterogeneity of included studies. We included studies performed in different countries and studying different types and groups of travelers. Additionally, the prevalence of MDR-E in each country is different and this was not incorporated in the risk factor analysis. To limit the influence of heterogeneity, in the meta-analysis we used a random effects model. Second, publication bias was present in several meta-analyses. Despite our extensive search for all available evidence, small studies with no effect are simply not performed and/or published. However, because of our extensive search, we think that the influence of publication bias to our results and conclusions is limited. Third, seven studies were included with a low methodological quality, of which 2 were included in the meta-analyses. These were relatively small studies, which did not have a big influence in weight on the pooled estimate. Therefore, we consider the influence of studies with a low methodological quality as limited.

Conclusion
This systematic review shows that travel to South Asia, together with antibiotic use and TD, are leading risk factors for acquiring MDR-E. It is advisable for travelers to contact a travel clinic in their home country before travel to be informed about TD and antibiotic use, and to limit self-prescribing of antibiotics and buying antibiotics over the counter during travel when suffering from TD. Acquisition of CPE during travel is still rare, but possibly underreported. The information in this review is useful for the development of guidelines for healthcare facilities with low MDR-E prevalence rates to prevent admission of potential carriers of MDR-E without appropriate measures. However, we identified a number of knowledge gaps that should be filled in before guidelines for healthcare facilities can be developed and implemented, since the impact of the measures cannot be estimated yet.

COVID-19
Currently, we are in the midst of the COVID-19 pandemic. Governments are discouraging or forbidding travel of any kind, and calling on everyone to stay at home as much as possible. Additionally, several countries have implemented a full lockdown or shelter-inplace measures. Healthcare systems are severely affected. Furthermore, these measures have a significant impact on domestic and international travel. This also affects the spread of MDR-E: we expect a decreased transmission rate during this period due to the decrease in (inter)national travel. However, an increased use of antibiotics has also been observed. We expect that this, combined with overcrowding and a shortage of personal protective equipment in hospitals, will lead to an increase of local spread of MDR-E, and consequently, we expect an increased local prevalence of MDR-E in lowand-middle income countries and in Southern European countries. If in the second half of 2020 international travel is resumed due to relaxing of COVID-19 measures, we will see the results of this local spread. We expect that the proportion of travelers who acquire a resistant microorganism after international travel will increase after COVID-19. Future surveys will provide more insight in the effect of the COVID-19 pandemic on travel-related spread of MDR-E.