3.1. Study selection and Study characteristics
The initial database search yielded 280 potential articles. After reviewing the titles and abstracts against our inclusion criteria, we excluded 203 articles that did not meet the criteria. We then downloaded the full articles of the remaining 77 articles for further analysis. Of these 77 articles, 51 articles were excluded for various reasons: 24 included samples from an adult population, 18 included isolates from samples with no diarrhea, and 9 included isolates from animals. From the remaining 26 articles, we further excluded 6 articles for the following reasons: 2 were inaccessible, 1 didn't report co-trimoxazole sensitivity, 1 was a case-control study, 1 did not use disk-diffusion method , and 1 did not report resistance at all. Ultimately, we included 20 papers in our meta-analysis (Figure 1). Except for 4 articles [20–23], all remaining articles tested the sensitivity of co-trimoxazole in more than one bacterial species. The studies were conducted between 2003 to 2019and were primarily carried out in hospitals and/or health centers. The majority of the studies (n=7) were conducted in the Amhara region followed by SNNP (n= 5) (Table 2).
Table 1 Prevalence of co-trimoxazole resistance against enteropathogens
Bacteria
|
No. of studies
|
No. of Isolates
|
Resistant
|
Prevalence % [95%CI]
|
Weight
|
I2
|
Q
|
Shigella
|
13
|
214
|
111
|
51.4 [0.38-0.64]
|
35%
|
67.9%
|
37.44
|
Salmonella
|
15
|
334
|
196
|
45.5 [0.31-0.59]
|
37.4%
|
74.5%
|
54.97
|
E. coli
|
4
|
340
|
240
|
62.5 [0.38-0.83]
|
14.3%
|
93.4%
|
45.24
|
Campylobacter
|
4
|
152
|
74
|
46.5 [0.28-0.64]
|
13.6%
|
80.0%
|
15.01
|
Overall
|
36
|
1040
|
621
|
50.3 [0.41- 0.58]
|
|
81.2%
|
185.8
|
CI: Confidence Interval, I2 Higgin's test, Q test of heterogeneity
Our results show that co-trimoxazole sensitivity was tested in 15 instances for Salmonella and 13 instances for Shigella, while 4 studies each reported sensitivity test for E. coli and Campylobacter species (Table 1). Twelve authors tested E. coli and Salmonella sensitivity simultaneously (Table 2). Figure 2 shows that the percentage of co-trimoxazole resistance is higher in Shigella than Salmonella for all of the authors except Mamuye et al. [24] and Gebreegziabhier et al. [5]. The majority of the Shigella and Salmonella isolates showed some degree of resistance, except for the study reported by Mulatu et al. [14], who reported 100% susceptiblity to co-trimoxazole [Fig 2]. Beyene et al [25] reported that Shigella isolates were completely resistant to co-trimoxazole, while the lowest Shigella and Salmonella resistance was reported by Zelelie et al [8]. All Salmonella isolates tested by Hayamo et al [26] were found to be sensitive to co-trimoxazole.
Table 2.Characterization of the studies included in the meta-analysis
Study
|
Study Year
|
Location [Region]
|
Sample
Size
|
Isolates
|
Setting
|
Disk µg used
|
Reference
|
Abebe et al
|
2017
|
Hosana, SNNP
|
204
|
Shigella, Salmonella
|
General Hospital
|
25
|
[36]
|
Admasu et al
|
2012
|
Bahir Dar, Amhara
|
422
|
Shigella, Salmonella
|
Pediatric Clinic
|
25
|
[37]
|
Adugna et al
|
2012
|
Amhara
|
422
|
E. coli
|
|
25
|
[20]
|
Alemu et al
|
2018
|
Gondar, Amhara
|
272
|
Shigella
|
1 Specialized Hospital, 1 health center, and 1 pediatric clinic
|
25
|
[21]
|
Ameya et al
|
2017
|
Arba Minch, SNNP
|
167
|
Shigella, Salmonella
|
General Hospital,2 Health Centers
|
NS
|
[38]
|
Ayenew et al
|
2016
|
Addis Ababa
|
290
|
Shigella, Salmonella
|
1 Hospital, 2 Health Centers
|
25
|
[39]
|
Beyene et al
|
2012
|
Jimma, Oromia
|
260
|
Shigella, Salmonella
|
Health center
|
25
|
[25]
|
Beyene et al
|
2006
|
Addis Ababa, Jimma
|
1225
|
Shigella, Salmonella
|
2 Hospitals
|
25
|
[40]
|
Mulatu et al
|
2003
|
Jimma, Oromia
|
430
|
Shigella, Salmonella, Campylobacter
|
1 Hospital
|
25
|
[14]
|
Demissie et al
|
2014
|
Gondar, Amhara
|
372
|
Shigella, Salmonella
|
1 Hospital, 8 Health Centers
|
30
|
[33]
|
Gebreegziabher et al
|
2011
|
Mekelle, Tigray
|
260
|
Shigella, Salmonella
|
1 hopsital, 1 health center
|
25
|
[5]
|
Mulatu et al
|
2011
|
Hawasa, SNNP
|
158
|
Shigella, Salmonella, Campylobacter
|
1 hospital, 1 health center
|
30
|
[14]
|
Hayamo et al
|
2019
|
Hawasa, SNNP
|
263
|
Shigella, Salmonella
|
Health center
|
25
|
[26]
|
Lengerh et al
|
2011
|
Gondar, Amhara
|
285
|
Campylobacter
|
|
25
|
[22]
|
Mamuye et al
|
2012
|
Addis Ababa
|
253
|
Shigella, Salmonella, E.coli
|
1 Hospital, 2 Health Centers
|
25
|
[24]
|
Mekonen et al
|
2014
|
Dire Dawa
|
196
|
Shigella, Salmonella, E.coli
|
1 Hospital
|
25
|
[34]
|
Mengistu et al
|
2011
|
Butajira, SNNP
|
382
|
Shigella, Salmonella
|
Health center
|
25
|
[16]
|
Tosisa et al
|
2014
|
Ambo, Oromia
|
239
|
Shigella, Salmonella
|
1 Hospital, 2 Health Centers
|
25
|
[35]
|
Yemane et al
|
2011
|
Bahir Dar, Amhara
|
384
|
Salmonella
|
2 Clinics
|
25
|
[23]
|
Zelelie et al
|
2015
|
Debre Birhan, Amhara
|
163
|
Shigella, Salmonella, E.coli
|
1 hospital, 1 health center
|
NS
|
[8]
|
Figure 2. Percentage of co-trimoxazole resistance in Shigella and Salmonella species isolated from the stool of Ethiopian children.
3.2. Prevalence of co-trimoxazole resistance to bacterial enteropathogens
We included 20 articles, segmented into 36 sections based on the number and type of bacteria reported. We analyzed these 36 studies, which reported a total of 1040 bacterial stool isolations, and 621 events of co-trimoxazole resistance. We calculated the proportion of resistant isolates using a random-effects model, after double arcsine transformation to achieve normal distribution. Overall, E. coli showed the highest resistance at 62.3 % (n=240), followed by Shigella at 51.4% (n=111) , Campylobacter at 46% (n=74) and Salmonella at 45% (n=196). We found the overall resistance of enteropathogens to co-trimoxazole to be 50.3 % (Table 1).
3.3. Heterogeneity
The results of heterogeneity test using Cochrane's Q showed significant heterogeneity between the studies [Q= 185.84, degree of freedom (df)=35). We used Higgin’s I2 to calculate the overall between-study variance and found it to be 81.2% (74.6%-86%). Studies that tested E. coli sensitivity showed the highest I2 at 93.4%, while the lowest I2 was among studies that tested Shigella at 67.9%. We calculated the standard deviation of heterogeneity using Tau2 with the DerSimonian-Laird estimator. We found that Tau2 was 0.038 with 95 CI [0.021-0.082]. We used the dmetar() library in R-Studio to find outliers. We removed eight outlier studies (n=8) with significant effects, assuming a random-effects model. We then carried out the final analysis with (k=28) studies. After removing the outliers, we repeated the test and found that Cochrane’s Q was no longer significant (Q=32.4, df=27), and I2 decreased from 81.2% to 16.7%. However, the prevalence of co-trimoxazole resistance only slightly increased from 50.3% to 52.15% (95%CI: 0.4716 - 0.5712).
Figure 3. Forest plot for co-trimoxazole resistance in bacterial enteropathogens in diarrheic children.
3.4. Publication bias
We assessed publication bias using the funnel plot and Egger's regression test. The funnel plot shows an asymmetric distribution of studies. There are fewer studies on the right side of the funnel plot than the left side. The number of outliers is higher on the left side. The plot shows more studies with higher standard errors on the left side than on the right side. We used Eggers’s test to determine publication bias using a statistical method with the “linreg” method in the metabias function in R-studio. Egger's test of linear regression of the intercept showed that there was a significant publication bias (t=-4.04 and p-value=0.0003).
Figure 4. Funnel plot for assessing publication bias.
3.5 Subgroup analysis of co-trimoxazole resistance
3.5.1 Bacterial species: The subgroup analysis of the bacteria showed that the proportion was higher in E. coli at 0.6247 (0.38-0.83), followed by Shigella at 0.5142 (0.38-0.64), Campylobacter at 0.4649 (0.28-0.64) and Salmonella at 0.4547 (0.31 - 0.59). The test for subgroup differences using a random effects model was not significant (df=3, Q=1.6).
3.5.2. Age group: We classified the studies into two groups based on the ages of the sample population as either below 5 years (k=24) or below 14 years (k=12). The results of the subgroup difference were not statistically significant, however, the prevalence of antimicrobial resistance was higher in the below-14 subgroup.
3.5.3. Study Setting: We classified the studies into hospital [k=28] or non-hospital k=8] based on where they were carried out. The subgroup analysis showed that prevalence of antimicrobial resistance was higher in non-hospital settings at 0.5814 (0.4046-0.7501) than in the hospital setting at 0.4789 (0.3790-0.5796). The test for subgroup difference was not significant assuming a random effects model.
3.5.4. Region: We included region as a factor in the sub-group analysis based on where the studies were conducted. There was a significant difference (Q=34.81 df=6) in the proportion of resistance among groups based on their geographical location. The highest proportion of resistance (80.5%) was reported in two studies conducted in Addis Ababa, Tikur Anbessa University Hospital and Jimma University Hospital by Beyene et al [40]. The lowest proportion reported was 30.5% with a 95%CI (0.2026-0.4165) in the region of SNNP.
3.5.5. Sample size: We classified the studies based on sample size as above or below 250 to see the effect of sample size on the proportion of resistance. Using a random-effects model, significant differences were found with a Cochrane’s Q=4.73. Resistance was higher in studies with a sample size greater than 250 at 58.5% with a 95%CI (0.5070-0.6628), while in studies below sample size below 250 it was lower.