Available evidence of antibiotic resistance from extended-spectrum β-lactamase-producing Enterobacteriaceae in paediatric patients in 20 countries: a systematic review and meta-analysis

Abstract Objective To make a systematic review of risk factors, outcomes and prevalence of extended-spectrum β-lactamase-associated infection in children and young adults in South-East Asia and the Western Pacific. Methods Up to June 2018 we searched online databases for published studies of infection with extended-spectrum β-lactamase-producing Enterobacteriaceae in individuals aged 0–21 years. We included case–control, cohort, cross-sectional and observational studies reporting patients positive and negative for these organisms. For the meta-analysis we used random-effects modelling of risk factors and outcomes for infection, and meta-regression for analysis of subgroups. We mapped the prevalence of these infections in 20 countries and areas using available surveillance data. Findings Of 6665 articles scanned, we included 40 studies from 11 countries and areas in the meta-analysis. The pooled studies included 2411 samples testing positive and 2874 negative. A higher risk of infection with extended-spectrum β-lactamase-producing bacteria was associated with previous hospital care, notably intensive care unit stays (pooled odds ratio, OR: 6.5; 95% confidence interval, CI: 3.04 to 13.73); antibiotic exposure (OR: 4.8; 95% CI: 2.25 to 10.27); and certain co-existing conditions. Empirical antibiotic therapy was protective against infection (OR: 0.29; 95% CI: 0.11 to 0.79). Infected patients had longer hospital stays (26 days; 95% CI: 12.81 to 38.89) and higher risk of death (OR: 3.2; 95% CI: 1.82 to 5.80). The population prevalence of infection was high in these regions and surveillance data for children were scarce. Conclusion Antibiotic stewardship policies to prevent infection and encourage appropriate treatment are needed in South-East Asia and the Western Pacific.


Introduction
Antimicrobial resistance occurs when bacteria are no longer susceptible to the drugs used for treatment. 1 Increasingly, there are fewer antimicrobial drugs available to effectively treat common as well as life-threatening infections. Annual deaths from untreatable infections may rise from estimated 700 000 in 2015 to 10 million by 2050 if antimicrobial resistance is not controlled. 2 Common procedures such as surgery or cancer chemotherapy may become too dangerous to perform without effective antibiotics.
Extended-spectrum β-lactamases are enzymes that cause resistance to some of the most commonly used antibiotics, 3 including all penicillins, cephalosporins and monobactams. 3 Fortunately these enzymes have yet to confer resistance to carbapenems, making these drugs valuable for serious extended-spectrum β-lactamase-producing bacterial infections. 4 However, there have been recent outbreaks of extended-spectrum β-lactamase-producing Klebsiella spp. with carbapenem resistance, resulting in extremely high rates of mortality. 5,6 Within the already limited selection of antibiotics available to treat these infections, fewer are approved for use in children. 7 Children are particularly vulnerable to bacterial infections compared with young adults, due to their immature immune systems. 8,9 The World Health Organization (WHO) South-East Asia and Western Pacific Regions have over 4.3 billion of the world's population of 7.7 billion, including two of the most populous countries with heavy consumption of antibiotics: China and India. 10 Research suggests these regions have high antimicrobial resistance rates to extended-spectrum β-lactamaseproducing bacteria in the paediatric population. 11 Poor-quality antibiotics and unsupervised use are common across the Regions. The available studies provide an overall impression of the prevalence of antibiotic resistance in the Regions, but better evidence is needed about the risk factors and outcomes for children with these infections. We therefore aimed to make a systematic review and meta-analysis of the risk factors and outcomes of infection with extended-spectrum β-lactamaseproducing Enterobacteriaceae in children and young adults in the South-East Asia and the Western Pacific. We also mapped the prevalence of extended-spectrum β-lactamase-associated infections in countries and areas of the Regions using the available surveillance data.

Meta-analysis
We conducted the meta-analysis in accordance with the Cochrane handbook for systematic reviews of interventions. 12 All procedures followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. 13 The study was registered in the PROSPERO International prospective register of systematic reviews (CRD42017069701).

Search strategy
We made a comprehensive search, without language limitation, of online databases for articles published from 1 January 1940 to 30 June 2018 (Box 1). Two researchers independently conducted the search and screened the titles, abstracts and full texts of the papers. We used a standardized, piloted data collection form to determine whether papers were appropriate for inclusion. The researchers applied the Newcastle-Ottawa scale to assess risk of bias in non-randomized studies. 14 Studies scoring ≥ 5 and ≤ 8 were designated low risk of bias, ≥ 3 and ≤ 4 as moderate and ≤ 2 as high. We incorporated the quality assessment results into our sensitivity analysis using the Meta-analyses Of Observational Studies in Epidemiology checklist. Discrepancies at any stage of the analysis were resolved by consensus of the researchers.

Selection criteria
We included cohort, case-control and observational or cross-sectional studies. We defined the target population as children aged from birth to 21 years, according the American Academy of Paediatrics guidelines. 15 We included studies that were conducted in the WHO South-East Asia and Western Pacific Regions and that recorded both positive and negative results of testing for extended-spectrum β-lactamaseproducing bacteria.

Outcome measures
The principal outcome measure was patients' infection status, defined by whether specimens obtained tested positive or negative for infection with extended-spectrum β-lactamaseproducing bacteria. We analysed infection status by risk sub-groups: medical history in the 3 months before the infection (hospital stay, intensive care unit admission, surgery), exposure to invasive life support, antibiotic therapy and co-morbidities or underlying conditions. Other outcomes recorded were: hospital length of stay, mortality, persistent bacteraemia, antibiotic residence and duration of fever after antibiotic therapy.

Data synthesis and analysis
For the meta-analysis we pooled the data on number of isolates (four studies) or patients with isolates (37 studies) using a Mantel-Haenszel random-effects model to determine the risk of infection with extended-spectrum β-lactamaseproducing bacteria. 16 We calculated pooled odds ratio (OR) and 95% confidence intervals (CI) for dichotomous outcomes and weighted mean difference and 95% CI for continuous outcomes. All tests were two-tailed and P < 0.05 was considered statistically significant. If studies provided median and interquartile range, we made estimates of the mean and standardized deviation (SD). 17 We assessed the heterogeneity of the studies using the I 2 statistic, which eval-uates the consistency of study results. The cut-off for defining heterogeneity was I 2 > 50%. 18 Our sensitivity analyses were based on sample size on the overall summary estimates. 19 We evaluated whether this restricted analysis affected the magnitude, direction and statistical significance of the overall summary estimate. Additional sensitivity analyses assessed the different types of study designs, settings and risk of bias.
We carried out meta-regression to explore each potential factor contributing to heterogeneity between studies, such as study location, design, duration and setting, and patients' age and diagnosis, for all included studies reporting mortality and persistent bacteraemia. We used funnel plots with Egger regression test to assess publication bias (P < 0.1).
All statistical analyses were performed with R software, version 3.4.0 (R Foundation for Statistical Computing, Vienna, Austria), using the Meta and Metafor meta-analysis packages.

Prevalence study
We obtained data on the prevalence of extended-spectrum β-lactamaseassociated infection from the same studies included in the meta-analysis. We also made a search for other data sources in the published and grey literature (Box 1). We included data on children (ages 0-21 years), where available, and all age groups, if data for these ages were unavailable. We calculated percentage prevalence by the number of people or isolates testing positive for extended-spectrum β-lactamases out of the total population or isolates tested. For case-control studies, the overall prevalence rate was extracted instead. Numbers of cases and samples were extracted if stated by the source. Where population maps were provided in the source material, the average of the range were extracted as the prevalence in the country. We pooled the prevalence data by calculating the mean of the extracted data from all sources for each country.

Systematic reviews
Antibiotic resistance in Asia Pacific children Yanhong Jessika Hu et al.
screening titles and abstracts. After assessing the full text of 577 studies, we excluded 537. Screening of reference lists and conference abstracts yielded no additional studies. In total, 40 studies were included in the meta-analysis (Fig. 1). Three studies were reported in Chinese language, one study each in Korean and French, and the remaining 35 were in English.
Overall, the 40 studies reported 46 960 bacterial isolates from 17 829 children providing samples. We pooled data from 2411 samples testing positive and 2874 testing negative for extendedspectrum β-lactamase-producing bacteria over the study period up to June 2018. The most common method of detection of bacterial phenotypes was agar disk diffusion in 32 studies. The study designs were 11 retrospective cohort studies, 14 prospective cohort studies, six observational studies, two cross-sectional studies and seven case-control studies. We found studies from 11 different countries and areas: Taiwan, China; India; Indonesia; Japan; Malaysia; Republic of Korea; Singapore; Sri Lanka; Thailand; and Viet Nam. In 15 studies the focus was specifically on neonates (< 28 days old), 15 studies were of age groups 0-21 years (excluding neonates), seven studies were of age 0-5 years (excluding neonates) and three studies did not specify the ages (Table 1; available at: http://www.who.int/bulletin/volumes/96/7/18-225698).

Risk factors
The risk of infection with extendedspectrum β-lactamase-producing bacteria was significantly higher for patients whose medical history included intensive care unit admission (OR: 6
The duration of fever was 0.61 days longer in patients with extendedspectrum β-lactamase-producing bacteria than patients without (95% CI: 0.18 to 0.72; I 2 : 92%; seven studies; Fig. 2). Pooling five studies we found that persistent bacteraemia was four times higher in patients positive for extended-spectrum β-lactamaseproducing bacteria (95% CI: 2.66 to 6.14; I 2 : 0%; Fig. 3). Results from eight studies showed that the mean difference in length of hospital stay was 25.9 days (95% CI: 12.81 to 38.89; I 2 : 100%) for patients with extendedspectrum β-lactamase-associated infection than those without such infection (Fig. 4). Subgroup analysis showed that the mean length of hospital stay associated with infection was 29 days longer for patients who had recently  been admitted to an intensive care unit care than the patients not receiving this care. Similar results were seen for invasive life support; the mean length of stay after central venous catheterization was 33 days longer than without catheterization. 59 Eleven studies reported a pooled number of 188 deaths among 565 patients with extended-spectrum β-lactamase-associated infections compared with 86 deaths in 745 patients without these infections (OR: 3.2; 95% CI: 1.82 to 5.80; I 2 : 49%; Fig. 5). When analysed by subgroups, the risk of death for patients who had previously been admitted to the intensive care unit or exposed to central venous catheterization were not significant. However, the risk of death was higher among patients with sepsis (OR: 4.9 95% CI: 2.11 to 11.39; I 2 : 38%) than those without sepsis (OR: 2.3 95% CI: 1.19 to 4.26; I 2 : 35%;). 59 We also looked at the ORs for neonates and non-neonates but the differences not statistically significant between these groups. 59

Validity tests
None of the factors we analysed by meta-regression were contributors to between-study heterogeneity. In the Newcastle-Ottawa analysis of risk of bias, we found that 60% (24 out of 40) of studies scored high on risk of

Systematic reviews
Antibiotic resistance in Asia Pacific children Yanhong Jessika Hu et al. Table 3.

Systematic reviews
Antibiotic resistance in Asia Pacific children Yanhong Jessika Hu et al.
bias and 40% were low risk (Table 4; Table 5). Only four studies had clear statements about comparability and 10 about representativeness. The results from Egger's regression test revealed that publication bias was significant (P < 0.001). Sensitivity analysis excluding small studies with samples less than 10 revealed that the funnel plots were consistently asymmetric (P < 0.001; available from the corresponding author). The sensitivity analysis showed that the data were not consistent with from the overall estimated ORs and similar trends were observed. This evaluation showed that a more restricted analysis of the data did not affect the magnitude, direction and the overall summary estimate.

Prevalence study
The overall pooled prevalence of extended-spectrum β-lactamase in the studies included the meta-analysis combined with surveillance reports was 25.3%. The pooled prevalence from the studies in the meta-analysis was 39% among the 31 studies conducted in hospital settings and 31% in the seven studies conducted in community settings (two studies were in both hospital and the community).
Using data from other sources, we mapped population surveillance data from a total of 21 countries and areas in the South-East Asia and the Western Pacific Regions ( Table 6). The pooled data from all available surveillance resources that included adults and chil-dren showed that India had the highest pooled prevalence (90.0%) and Australia the lowest (3.6%; numerators and denominators unavailable). The pooled data specifically for children, where available from surveillance resources and published data, showed similar results (Fig. 6).

Discussion
This study revealed that the average combined prevalence of infection with extended-spectrum β-lactamaseproducing bacteria among children in South-East Asia and the Western Pacific is high. Risk factors for infection included recent intensive care unit admission, hospitalization, surgery or antibiotic  exposure, and co-existing bacteraemia, nosocomial infections, lower respiratory tract infections, sepsis or recurrent urinary tract infections. Infection was associated with higher mortality, higher morbidity and longer hospitalization.

Fig. 5. Mortality among children and young adults with and without extended-spectrum β-lactamase-associated infection in South-East Asia and Western Pacific countries
The prevalence of infection we found in South-East Asia and the Western Pacific countries are similar to those reported from other surveillance systems worldwide, 64 although many locations do not report data for children. A review of worldwide trends in extended-spectrum β-lactamase-associated infection reported higher prevalence in Asia, Latin America and the Middle East (from 28 to 40%) compared with other, higher-income areas (from 8 to 12%). 64 As many of the studies we found were hospital-based our results support the need for resources and policies for control of nosocomial infection. A recently published modelling study showed that antibiotic use in hospital is a major driver for antimicrobial resistance in human infection compared with animal and environmental antibiotic exposures. 65 Although infection control and hygiene may be sub-optimal in the countries we studied, infection control is easier to manage within health-care institutions than in other unstructured systems such as animal husbandry and the environment. Without proper control of antimicrobial resistance in hospitals, patients can disseminate antibiotic residues and resistance genes to the community and environment. This still highlights the importance of hospital-based stewardship for controlling antibiotic use and how this stewardship can reduce the risk of developing multidrug resistant organisms. 66 At the same time, the rising community prevalence of extended-spectrum β-lactamaseassociated infection provides evidence for expanding prevention to other settings. 3 Our meta-analysis showed that recent medical care, including intensive care unit stays, hospitalization, surgery and antibiotic therapy, was associated with increased risk of infection. These results suggest that children may acquire such infections during health care, especially when undergoing invasive procedures. Specifically, children who had exposure to third-generation cephalosporins, carbapenems and fluoroquinolones had three to four times greater risk for extended-spectrum β-lactamase-associated infection, which Notes: We applied the Newcastle-Ottawa scale to assess risk of bias in non-randomized studies. 14 Only studies scoring ≥ 5 and ≤ 8 were designated low risk of bias, ≥ 3 and ≤ 4 as moderate and ≤ 2 as high. We made Mantel-Haenszel radom-effects Table 5. Notes: We applied the Newcastle-Ottawa scale to assess risk of bias in non-randomized studies. 14 Only studies scoring ≥ 5 and ≤ 8 were designated low risk of bias, ≥ 3 and ≤ 4 as moderate and ≤ 2 as high.

Systematic reviews
Antibiotic resistance in Asia Pacific children Yanhong Jessika Hu et al.
is similar to previous reports. 26,[67][68][69][70][71] As these antibiotics are primarily used for treating severe infections, their use may be a marker for disease severity rather than a direct contributor to developing resistance. Nevertheless, if excessive fluoroquinolone use does contribute to emergence of resistant bacteria this adds another reason to avoid the unnecessary use of these broad-spectrum antibiotics in children.
Coexisting illnesses, including bacteraemia, nosocomial infection, lower respiratory tract infections, sepsis and recurrent urinary tract infections, were associated with increased risk of infection. These co-morbidities could be risk factors for use of invasive treatments such as a central venous catheterization, mechanical ventilation, intravenous nutrition or increased risk of interactions with health-care settings. In a twocentre case-control study of risk factors for infection with extended-spectrum β-lactamase-producers in children, multivariable analysis identified sepsis and neurological illnesses as potential risk factors, which supports our findings. 72 Previously published studies among both young adults and children found that prolonged hospital stay or prolonged use of invasive medical devices were associated with infection by, or being colonized with, extended-spectrum β-lactamase-producing bacteria, 26,69,71 which is consistent with our findings.
Recent surgery and antibiotic prophylaxis were associated with extendedspectrum β-lactamase infection in our  study. Others have shown that surgical antibiotic prophylaxis increases the risk for antimicrobial resistance and acquisition of infection. 73 One study from Switzerland found that half of all surgical ward prescriptions (680 out of 1270) were inappropriate. 74 Antibiotic stewardship programmes have been shown to improve surgical antibiotic prophylaxis and treatment of surgical site infections. 75 Our study found that initiation of appropriate empirical antibiotics was protective against extended-spectrum β-lactamase-associated infection, indicating the importance of thoughtful selection of antibiotics. The details of this finding warrant further study. The risk is especially high for critically ill patients requiring surgery or intensive care and who need antibiotics urgently before susceptibility has been established but who are also at increased risk for drugresistant infections. Therefore, antibiotic stewardship programmes and guidelines in health-care facilities fill an important function. Furthermore, as studies in Asia have shown a high prevalence of easy access to unsupervised antibiotics within the community, more attention is needed to improving appropriate antibiotic use through training, education, policy and regulation outside of hospitals. 76 Children infected with extendedspectrum β-lactamase-producers had significantly longer length of hospital stays (26 days) and required more intensive care unit days (29 days) than those without such infection. This leads to higher health-care costs, 77 in addition to the costs to society in terms of family and community pressures and lost productivity. At the same time, prolonging intensive care unit and hospital stays increases the risk of further acquisition and transmission of drug resistance.
Mortality and persistent bacteraemia were three to four times higher for patients infected with extendedspectrum β-lactamase-associated infections than those without. This adds to the economic and social burden of these infections. Based on our metaregression, the study location, study design, patient's diagnosis, sex or intensive care unit stay did not influence mortality. This implies that worse outcomes may be directly attributable to the presence of extended-spectrum β-lactamase-associated infection. The severity of the diseases associated with these infections might also contribute to mortality risk, as the patients diagnosed with sepsis had higher risk of mortality than those without sepsis. However, we were unable to determine for each study whether other factors may have influenced outcomes because comprehensive information was not available.
One of the strengths of our study was the comprehensive data collection strategy, which provided a high sample size and study power. Second, two different tools were used to assess for bias, which, together with risk factor and outcomes sensitivity analysis, strengthened the study's validity and reliability. Third, we assessed previous antibiotic history with different antibiotic categories, providing a detailed insight into the link between antibiotic use and resistance. Fourth, we also conducted meta-regression to determine if other factors might have influenced treatment outcomes. This established association between patients' mortality, length of stay and extended-spectrum β-lactamase infections.
There were several limitations to this study. The distribution of studies between locations was not uniform. Of the 48 Member States and areas in the South-East Asia and the Western Pacific Regions, we were able to find and extract data for the meta-analysis from 12 countries. For prevalence estimates we added surveillance data from 10 other countries and areas but we found data on 0-21-years-olds for only three countries with available paediatric data, which might underestimate the real situation among children. Moreover, although we made subgroup analyses, most of the pooled prevalence from selected studies were from hospital settings. Most of the surveillance sources reported only prevalence, without denominators and numerators. Nevertheless, the study provides a rough indication of the extent of extended-spectrum β-lactamaseassociated infection and highlights the need for establishment of surveillance systems in these Regions. We can expect that within large Regions, rates of infection are unlikely to be homogenous, particularly where there are large urban and rural disparities. Among 40 studies, only seven were community based. This might have underestimated antibiotic resistance in the community. With the rising concern for community-acquired infections and reports of increased rates of faecal colonization with extended-spectrum β-lactamase-producing bacteria in healthy children, risk factors might not only arise from hospital influences but also from community exposure and international travel. [78][79][80] Because of limited information in the articles, we are unable to determine whether longer hospitalization increased the risk of infections or vice versa. Both situations are likely and further studies are needed to clarify the associations.
Another limitation we faced was the lack of laboratory standardization for the identification of the extendedspectrum β-lactamase-producer phenotypes. Quality and standardization may vary between laboratories, although most followed Clinical and Laboratory Standards Institute guidelines. Sensitivity analyses found that use of different laboratory guidelines or test methods or the study year did not affect our results. All studies used phenotypic methods, as opposed to the gold standard through genotyping, with the majority using agar double-disk diffusion test, while a few studies used the Vitek® system (bio-Mérieux, Marcy l'Etoile, France). Thus, detection rates could be underestimated.
We hope this study will provide important information for policy-makers who need to allocate resources to improve surveillance, monitor treatment outcomes, improve infection control in intensive care unit and surgery wards and develop policies for the use of empirical and prophylactic antibiotics. Knowledge of resistance rates can guide treatment recommendations. Countries without established antibiotic stewardship programmes should prioritize these activities, along with public education programmes. With very high burden of neonatal sepsis 0.42 million (39%) of the total 1.09 million deaths related with sepsis in these Regions, 81 scaling up strategies to prevent infection and encourage appropriate treatment for this vulnerable group is needed. More studies are also needed to measure the impact of antimicrobial resistance in children. ■