Prevalence of Vancomycin resistant enterococci (VRE) in Ethiopia: a systematic review and meta-analysis

A comprehensive search was conducted at PubMed, EMBASE, Google scholar and African journals online (AJOL). To include unpublished studies (theses, dissertations); the repository of Addis Ababa University was searched. Reference lists of included studies were also sought. The database search was done following the PRISMA guideline/checklists [27] (Fig. 1). The PubMed was searched using MeSH terms and Boolean operators. The search string in PubMed was: ((((((((Enteroccoc*) OR Enterococcus faecalis) OR Enterococcus faecium) OR E. faecalis OR E. faecium AND Vancomycin resistan*) OR antibiotic resistan*) OR antimicrobial resistan*) OR drug resistan*) OR VRE) AND Ethiopia)))))))). Search results were combined in to EndNote X6 (Clarivate Analytics USA) and duplicates were removed. Studies published/reported up to June 30, 2019 and fulfilled the eligibility criteria (Table 1) were included.

The quality of included studies was assessed by the Joanna Briggs Institute (JBI) critical appraisal checklist for prevalence data [28] (additional file 1); which contains nine sections. The assessment was done independently by two authors (AM and TA). Studies were included in the analysis if consensus was reached among the two reviewers. The quality of the 20 included studies is given in (additional file 2).

After studies were identified based on the predefined eligibility criteria; author name with year of publication, study period, region of study, study design, sample size, study population, types of specimens, antimicrobial susceptibility testing (AST) methods, number of isolates (both the total and vancomycin resistant enterococci), types of isolated species and history of publication were extracted using Microsoft Excel 2013 data collection sheet especially designed for this study. Resistance profiles of enterococci to other antimicrobials were also extracted and the study level proportions were pooled. The data extraction was done independently by two authors (AM and TA).

Whenever studies were not reporting the prevalence of VRE, it was calculated by dividing the numbers of VRE isolates to the total numbers of tested enterococcal isolates and multiplying by 100. Studies reporting a zero number of VRE isolates were imputed to 0.5 as a continuity correction to be include in the meta-analysis [29]. Subgroup analyses were done by the study region, study period, publication history, AST and types of specimens used to isolate enterococci.

Acknowledging the presence of heterogeneity in observational studies conducted in diverse settings, the random effects model was used in determining the pooled prevalence of VRE as well as resistance to other antimicrobials. Heterogeneity was evaluated by the Cochran’s Q-test and I² statistics. Funnel plots were drawn to see the presence of publication bias and the Begg’s rank correlation and Egger’s regression tests were used to quantify the degree of publication bias. P-values <  0.05 in any of the Begg’s rank correlation and Egger’s regression tests were indicative of significant publication bias. In asymmetrical funnel plots, the Trim-and-Fill method was applied to include missing studies and estimate adjusted effect sizes. Sensitivity analysis in a leave-one-out approach was done to see the stability of the pooled prevalence of VRE and to explore the potential source of heterogeneity between studies. Data were analyzed using CMA version 2.0 for windows and used to generate forest and funnel plots.

The results of database search and process of study selection is shown in the flow chart below (Fig. 1). The search returned 1143 records; of which 62 studies were subjected for full text review for inclusion against the eligibility criteria. Finally, 42 studies were excluded and only 20 were included in our analysis.

All of the 20 studies included in this review were cross-sectional by design. Most of the studies were reported from Amhara region (n = 8) [30‐37] and Addis Ababa (n = 7) [38‐44]. The remaining studies were from Oromia (n = 4) [45‐48] and Southern nations (n = 1) [23]. Studies were not available from administrative regions of Tigray, Afar, Dire Dawa, Harari, Somali, Gambela and Benishangul-Gumuz. Nineteen studies were conducted in hospital settings. Among the 6017 study participants included, 831 enterococci were isolated and tested with a variety of antimicrobials; of which 71 isolates were resistant to vancomycin. Stool, urine, blood and swab specimens were used to isolate enterococci. The highest numbers of enterococcal and VRE isolates were identified from stool followed by multi-site specimens.

Seventeen studies used disc diffusion and three studies employed dilution/minimum inhibitory concentration (MIC) as antimicrobial susceptibility testing (AST) method to determine Vancomycin resistance. Resistance to antimicrobial agents by either methods was defined based on the performance standards for antimicrobial susceptibility testing guidelines prepared by Clinical and Laboratory Standards Institute (CLSI, various editions). The prevalence of VRE ranged from 1.8% in Jimma to 60% in Addis Ababa. Species level enterococci were reported by four studies [23, 39, 47, 48] and E. faecalis and E. faecium were the most frequently isolated species. Six of the included studies were unpublished and 14 were published between 2013 and 2019. Details of the characteristics of the included studies is summarized in (Table 2) below.

The pooled prevalence of VRE was estimated at 14.8% (95% CI; 8.7–24.3%; I² = 74.05%; P <  0.001) (Fig. 2). Significant heterogeneity (Q = 73.21; I² = 74.05%; P <  0.001) was observed in the estimation of overall pooled result. But, the sensitivity analysis revealed that no single study significantly influenced the heterogeneity and pooled prevalence of VRE. The pooled prevalence of VRE in the sensitivity analysis ranged from 13.2 to 16.7% which lies within the 95% CI bounds of the overall pooled estimate. The presence of publication bias was observed from the drawn asymmetric funnel plot (Fig. 3a). The Trim-and-Fill method was then applied to include the “missing” studies from the analysis. The asymmetric studies were trimmed to locate the unbiased effect and fills the plot by re-inserting the trimmed studies as well as their imputed counterparts. Accordingly, one study was missed and fall at the left side of the pooled estimate (Fig. 3b). In the Trim-and-Fill method, the adjusted estimate of VRE was 13.5% (95% CI; 7.8–22.2%); almost similar with the original pooled estimate. The Egger’s regression (intercept = 0.91; 95% CI; − 0.75 – 2.57; p = 0.263) and Begg’s rank test (p = 0.381) did not suggest significant publication bias.

The prevalence of VRE was computed by region, type of antimicrobial testing (AST) method, study period, types of specimen used to isolate enterococci, and publication history. The prevalence of VRE by region was 26.1% (95% CI: 10.7–50.9%; I² = 41.65%; P = 0.113) in Addis Ababa, 15.0% (95% CI: 6.9–29.6%; I² = 79.39%; P <  0.001) in Amhara, 9.0% (95% CI: 2.8–25.7%; I² = 71.49%; P = 0.015) in Oromia and 1.9% (95% CI: 0.1–23.1%) in Southern nations, nationalities and peoples region (SNNPR) (Table 3, Fig. 4). The prevalence of VRE pooled from studies conducted in the period before 2015 was 16.5% and that of the post-2015 was 16.3%; which indicates unchanged trend of VRE infections in Ethiopia. On the other hand, the pooled prevalence of VRE from studies which used disc diffusion to determine AST was 16.9% and it was 7.9% when AST was measured by dilution/minimum inhibitory concentrations (MICs). Relatively; high rates of VRE were isolated from urine (37.3%) and blood (22.0%) specimens. Use of multisite specimens did not increase the isolation rate of enterococci. Unpublished studies reported high rate of VRE than published studies (31.9% Vs. 11.3%; respectively) (Table 3).

The resistance profile of enterococci was also pooled for antimicrobials other than Vancomycin. Resistance rates were pooled if at least two studies reported on a specific bacterium-antibiotic combinations. High level of resistance was observed to all classes of tested antimicrobials except to Daptomycin and Linezolid. The pooled resistance rate of enterococci to Daptomycin was 3.2% (95% CI; 0.5–19.7%) and that of Linezolid was 9.9% (95% CI; 2.8–29.0%). The pooled resistance rate to other antimicrobials was 60.7% (95% CI; 39.2–78.3%) to Penicillin, 56.5% (95% CI; 49.6–63.2%) to Amoxicillin, 53.7% (95% CI; 35.8–70.7%) to Tetracycline, 55.1% (95% CI; 22.2–84.9%) to Doxycycline, and 49.6% (95% CI; 36.5–62.7%) to Erythromycin. Studies reporting resistance to three or more antimicrobials were also pooled to estimate the prevalence of multidrug resistant (MDR) enterococci in Ethiopia. Hence; the overall prevalence of MDR enterococci was 63.0% (95% CI; 48.6–75.4%; I² = 90.27%; P <  0.001) (Table 4).

A comprehensive search with clear inclusion and exclusion criteria was used, examined commonly used specimens and methods of susceptibility testing, and included unpublished studies retrieved from Addis Ababa University repository. The Trim-and-Fill method was applied to asymmetric funnel plots to produce adjusted estimates. There were a number of limitations in the depth and breadth of data. First; inability to report pooled estimates of VRE at species level due to the paucity of included studies reporting enterococci at species level. Second; the definition of VRE was not consistent across studies and different AST methods were combined limiting comparability and strength of this analysis. Third; data was not available from 54.5% of the regions, outside health care setting and non-human studies were excluded that may be difficult to generalize the pooled results. Fourth; combing resistance results from different patients across different regions might pool out the peaks of resistance in some settings. Lastly; the study protocol was not registered at PROSPERO.