Antimicrobial resistance in Enterobacterales infections among children in sub-Saharan Africa: a systematic review and meta-analysis

Summary

Background

The burden of antimicrobial resistance (AMR) has been estimated to be the highest in sub-Saharan Africa (SSA). The current study estimated the proportion of drug-resistant Enterobacterales causing infections in SSA children.

Methods

We searched MEDLINE/PubMed, Embase and the Cochrane Library to identify retrospective and prospective studies published from 01/01/2005 to 01/06/2022 reporting AMR of Enterobacterales causing infections in sub-Saharan children (0–18 years old). Studies were excluded if they had unclear documentation of antimicrobial susceptibility testing methods or fewer than ten observations per bacteria. Data extraction and quality appraisal were conducted by two authors independently. The primary outcome was the proportion of Enterobacterales resistant to antibiotics commonly used in paediatrics. Proportions were combined across studies using mixed-effects logistic regression models per bacteria and per antibiotic. Between-study heterogeneity was assessed using the I² statistic. The protocol was registered with PROSPERO (CRD42021260157).

Findings

After screening 1111 records, 122 relevant studies were included, providing data on more than 30,000 blood, urine and stool isolates. Escherichia coli and Klebsiella spp. were the predominant species, both presenting high proportions of resistance to third-generation cephalosporins, especially in blood cultures: 40.6% (95% CI: 27.7%–55%; I²: 85.7%, number of isolates (n): 1032) and 84.9% (72.8%–92.2%; I²: 94.1%, n: 2067), respectively. High proportions of resistance to other commonly used antibiotics were also observed. E. coli had high proportions of resistance, especially for ampicillin (92.5%; 95% CI: 76.4%–97.9%; I²: 89.8%, n: 888) and gentamicin (42.7%; 95% CI: 30%–56.5%; I²: 71.9%, n: 968). Gentamicin-resistant Klebsiella spp. were also frequently reported (77.6%; 95% CI: 65.5%–86.3%; I²: 91.6%, n: 1886).

Interpretation

High proportions of resistance to antibiotics commonly used for empirical treatment of infectious syndromes were found for Enterobacterales in sub-Saharan children. There is a critical need to better identify local patterns of AMR to inform and update clinical guidelines for better treatment outcomes.

Funding

No funding was received.

Research in context.

Evidence before this study

It has been estimated that sub-Saharan Africa (SSA) has the world’s highest incidence of deaths due to antimicrobial resistance (AMR). However, in most areas in SSA, access to diagnostic and microbiological testing is limited and empirical data on the burden of AMR are scarce, particularly for children. The eleven previous meta-analyses on drug-resistant infections in SSA presented alarming rates of AMR but included relatively few studies, some of which were outdated. Most did not focus on children. A previous literature review, published in The Lancet Infectious Diseases in 2018, identified 18 studies on AMR conducted among sub-Saharan children. Several articles have been published since then, therefore an updated systematic review expands the current knowledge base. We searched MEDLINE/PubMed, Embase and the Cochrane Library on the 01/06/2022 to identify retrospective and prospective studies published from 01/01/2005 to 01/06/2022 reporting AMR of Enterobacterales causing infections in sub-Saharan children (0–18 years old).

Added value of this study

We included 122 studies providing data on more than 30,000 bacterial isolates from blood, stool or urine samples obtained from children with a clinical infection in SSA. To the best of our knowledge, this is the most comprehensive review on this subject to date, which provides important new insights in the threat of AMR in SSA.

Compared to previous reviews, our meta-analysis found a higher proportion of Enterobacterales resistant to third-generation cephalosporins in children in this region. In the reported blood cultures, 40.6% of Escherichia coli and 84.9% of Klebsiella spp. were resistant to these antibiotics. At the same time, these isolates were reported to have high proportions of resistance to other commonly used antibiotics like amoxicillin, gentamicin, ciprofloxacin or cotrimoxazole. There is, therefore, a significant risk of administering inappropriate empirical treatments to children with infections in SSA.

Implications of all the available evidence

This review supports the growing view that there is an urgent need to strengthen local-level capacities in microbiological analysis in SSA to support antimicrobial stewardship measures. Coupling this with other measures, such as strengthening infection prevention and control in healthcare services, could dramatically improve patients’ outcomes in the region.

Introduction

Despite recent improvements, sub-Saharan Africa (SSA) has the highest child mortality rates globally, with 74 deaths per 1000 live births.¹ Infectious diseases, including pneumonia, malaria, diarrhoeal diseases and sepsis, remain the major cause of death among children under five.² Bloodstream infections are responsible for 6% of newborn deaths and 14% of childhood deaths worldwide.^3,4 Enterobacterales are among the most common bacteria isolated from blood cultures in low- and middle-income countries. In SSA, Klebsiella spp. predominate in the blood isolates of newborns and Salmonella spp. predominate among older children.^3,5

Concurrently, antimicrobial resistance (AMR) has become a major global health problem. It has been estimated that SSA has the world’s highest risk of death due to AMR. In western SSA, the death rate attributed to AMR exceeds 100 per 100,000 individuals.⁶ Enterobacterales are species with a high potential for resistance acquisition.⁶ Thus, they may be a major contributor to AMR in children in SSA.

In this context, treating childhood infections in SSA effectively has become increasingly challenging. Access to microbiological testing is limited, and treatment decisions are mainly based on empirical recommendations that may not be adapted to current AMR patterns.

Data are essential to measure the extent of this problem. Unfortunately, they remain scarce in SSA, mainly due to limited access to microbiology laboratories and the lack of AMR surveillance network.⁷ Few meta-analyses on the prevalence of AMR in SSA have been conducted, and most are outdated, did not focus on children and could only include a small numbers of studies. However, they all showed that Enterobacterales displayed very high rates of resistance to ampicillin, gentamicin and third-generation cephalosporins (3GCs), which represent the first- and second-line treatments for sepsis in children.^3,8,9

The present study summarised the current publicly available evidence on the prevalence of Enterobacterales’ AMR among children with infections in SSA.

Methods

The study protocol is available online (PROSPERO registry CRD42021260157) and was informed by the PRISMA guidelines for systematic reviews,¹⁰ Our primary outcome was the proportion of Enterobacterales resistant to 3GCs in clinical samples of children with an infection in SSA, secondary outcomes included resistance to alternative antibiotics (penicillin, other cephalosporins, carbapenem, quinolones, aminoglycosides and cotrimoxazole).

As it is a meta-analysis, an ethics approval was not required.

Search strategy and selection criteria

We used a combination of MeSH and free terms (Appendix pp 3–5) to search MEDLINE/PubMed, Embase and the Cochrane Library. No language restrictions were applied. To be included, studies had to be published between 1 January 2005 and 1 June 2022 and present AMR data on Enterobacterales in clinical samples from paediatric patients (0–18 years old) in SSA. We chose these dates to gather as much information as possible and to capture the spread of ESBL in more recent years. We included studies conducted in both inpatient and outpatient settings. Studies from South Africa were excluded because the availability of AMR data from this country is greater than for other SSA countries, and we did not want South Africa data to have an excessive weight in our results as we are primarily interested in less studied and less well equipped SSA countries.⁸

We excluded case reports, systematic reviews and meta-analyses, studies not reporting on Enterobacterales or on the prevalence of AMR in paediatric patients, studies that did not clearly document the antimicrobial susceptibility testing methods used. We also excluded studies reporting fewer than ten observations per bacteria as small studies may disrupt models with random effects in meta-analyses of trials.¹¹ As this review focused on Enterobacterales causing infections, we excluded studies reporting solely on screening isolates.

Titles and abstracts were screened, followed by full-text analysis of selected manuscripts. Five investigators (MK, NW, BO, GC or AG) selected eligible studies, independently and in duplicate. Disagreements were resolved via consensus by a third investigator (NW or AG). The investigators discussed the inclusion criteria in detail and in depth before the screening process and held weekly meetings to discuss any doubts about their application to standardise their approach as much as possible. We contacted the authors of papers if relevant data was missing. We used Covidence systematic review software to manage article selection (Veritas Health Innovation, Melbourne, Australia. Available at www.covidence.org).

Data extraction

Two investigators independently entered the following data into Excel® spreadsheets: year of publication and data collection, country, sample size, participants’ ages, population characteristics (e.g. HIV+, sickle cell disease, neonates, malnutrition, etc.), study design, types of clinical samples, and bacterial susceptibilities to reported antibiotics. We recorded data on the prevalence of antibiotic non-susceptibility for the following Enterobacterales: E. coli, Klebsiella spp., Enterobacter spp., Salmonella spp., Shigella spp., Salmonella typhi and Enterobacterales species that were not classified into subspecies. The Proteus, Citrobacter, Providencia, Morganella and Serratia genera were grouped into a category named ‘other Enterobacterales’. This review focused on specimens from blood, urine and stool (only Salmonella and Shigella spp. were considered in stools), but data from other sample types were also extracted.

Definitions

Children have been defined as individuals aged 0–18 years old, with neonates as a subgroup aged 0–1 month old.

For clarity and ease of reading throughout this paper, Enterobacterales considered resistant or intermediately resistant to a given antibiotic were group into one group that we called « resistant ».

Extended-spectrum cephalosporin-resistant (ESCR) Enterobacterales have been defined as phenotypically non-susceptible to either ceftriaxone, ceftazidime or cefotaxime.¹² Carbapenem-resistant Enterobacterales (CRE) have been defined as phenotypically non-susceptible to either imipenem, meropenem or ertapenem. Susceptibility profiles were accepted if they were based on phenotypic (disk diffusion, E-test MIC determination and double-disk synergy test for ESBL) or genotypic methods.

Sample types other than blood, urine, stool and cerebrospinal fluid, were grouped as ‘other samples’ (mainly throat, middle ear and wound swabs). ‘All samples’ refers to all of the reported categories.

Risk of bias

Each study’s quality and risk of bias were assessed using the Newcastle–Ottawa assessment scale (NOS) for cohort studies. The NOS is a risk of bias assessment tool for observational studies that is recommended by the Cochrane Collaboration.¹³ We adapted the Newcastle–Ottawa Quality Assessment Form for Cohort Studies to assess the risk of bias in the included studies.¹³ Five reviewers (MK, NW, BO, GC or AG) independently and in duplicate assessed risks of bias, and any disagreements were resolved by consensus (Appendix pp 6–7).

Statistical analysis

Due to the expected between-study heterogeneity, proportions of resistance were combined across studies by systematically using models with random effects. For this purpose, we used mixed-effects logistic regression models per bacteria and per antibiotic, with a random intercept.¹⁴ The approach’s advantage is that it does not require a continuity correction when some studies report zero events. Between-study heterogeneity was assessed using the I² statistic.¹⁵ Planned univariable subgroup analyses (neonates versus children, geographical regions, year of publication and sample types) investigated potential sources of heterogeneity by introducing them as variables in the models. Leave-one-out sensitivity analyses were conducted to identify influential studies. The level of significance was set at 5%, and statistical tests were two-sided. Statistical analyses were carried out using R software, v4.0.2 (R Core Team, 2020).¹⁶

Role of the funding source

There was no funding source for this study.

All authors confirm that they had full access to all the data in the study and accept responsibility for the decision to submit for publication.

Results

Overview of the studies included

The initial search yielded 1111 non-duplicated records. After reviewing titles and abstracts, 525 titles were selected for full-text screening, of which 122 fulfilled the inclusion criteria. The most common reasons for exclusion were studies not reporting on Enterobacterales (N: 95) or not reporting a separate proportion of AMR for paediatric patients (N: 88) (Fig. 1).17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35^,36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72^,73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102^,103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138 Five authors were contacted because relevant data were missing, and they provided the missing information from two studies, allowing their inclusion.^61,107

Fig. 1.

Study selection process based on the PRISMA flowchart. Abbreviations: N = number of studies.

Characteristics of the included studies

The 122 included studies reported AMR data from 24 African countries, of which most came from eastern African countries (56/122, 45.9%) (Fig. 2). There were 54 (54/122, 44.3%) prospective studies and 35 (35/122, 28.7%) cross-sectional studies. The remaining studies used observational and retrospective case–control and cohort designs. Of these 122 studies, 47 (38.5%) reported data from blood cultures, 40 (32.8%) reported results from stools, and 16 (13%) reported results from urine. Seventeen studies (13.9%) focused on neonates, the remainder focused either on older children either on a mixed population including neonates and children. A number of studies selected specific risk groups, including six (4.9%) on malnourished children, two on children with HIV and one on children with sickle cell disease (Appendix pp 8–15). No information was reported to distinguish community-acquired from healthcare-acquired, or healthcare-associated infections. Over 160,000 samples were collected across all the studies, yielding more than 30,000 Enterobacterales isolates.

Fig. 2.

Study location. Abbreviations: N = number of studies by region; Country (number of studies by country). The total number of countries does not correspond to the total number of studies included because two studies examined multiple countries in different regions.

Proportion of resistant Enterobacterales in blood, urine and stool samples

Blood samples

Of 1032 E. coli isolates from positive blood samples, we found a pooled ESCR proportion of 40.6% (95% CI: 27.7%–55%; I²: 85.7%, number of studies (N): 19). Of 2067 Klebsiella spp. isolates, pooled ESCR proportion was 84.9% (95% CI: 72.8%–92.2%; I²: 94.1%, N: 26). By contrast, a pooled proportion of only 3.3% (95% CI: 1.1%–9.9%; I²: 82.7%, N: 13) of Salmonella spp. were ESCR on 2196 positive blood cultures. Of the 750 Salmonella typhi isolated in blood culture, none was ESCR (Table 1 and Appendix pp 16–17).

Table 1.

Non-susceptibility of Enterobacterales in blood and urine samples (all populations included).

Blood cultures	Escherichia coli				Klebsiella spp.				Salmonella typhi				Salmonella spp.
	% (CI)	I2	n	N	% (CI)	I2	n	N	% (CI)	I2	n	N	% (CI)	I2	n	N
ESCR	40.6 (27.7–55)	85.7	1032	19	84.9 (72.8–92.2)	94.1	2067	26	0 (0–100)	0	750	7	3.3 (1.1–9.9)	82.7	2196	13
CRE	1.9 (0.5–7.2)	0	106	5	2 (0.5–7.5)	66.7	897	11	0 (0–23.2)	..	14	1	0 (0–N/A)	0	881	2
Ampicillin	92.5 (76.4–97.9)	89.8	888	13	..	..	..	..	63.8 (27.2–89.3)	92.5	700	6	77.9 (58.1–89.9)	94.8	2372	13
Amoxicillin and clavulanic acid	51.8 (30.4–72.6)	87.7	517	11	82.2 (62.2–92.8)	93.5	1217	18	..	..	..	0	61.6 (42.7–77.5)	84.5	341	6
Piperacillin and tazobactam	35.9 (22.5–51.9)	0	39	2	16.8 (2.8–58.7)	96.1	481	3	..	..	..	0	..	..	..	0
Cefuroxime	43.8 (29.7–59)	49.8	80	5	68.1 (22.7–93.9)	85.8	592	7	..	..	..	0	42.5 (25.6–61.4)	79.9	187	3
Ceftriaxone	45.4 (27.9–64.1)	89.7	893	13	85 (66.4–94.2)	93.2	1397	15	0 (0–100)	00	738	6	2 (0.5–7.9)	84.2	2097	10
Cefepime	43.5 (23.2–65.5)	..	23	1	88.3 (83.2–92.1)	0	206	2	..	..	..	0	..	..	..	0
Gentamicin	42.7 (30–56.5)	71.9	968	17	77.6 (65.5–86.3)	91.6	1886	22	52.9 (27.8–77)	Na	17	1	22.5 (15.4–31.6)	0	262	5
Amikacin	10.4 (0.1–90.6)	0	91	5	24.7 (9.8–49.7)	92.1	782	11	..	..	..	0	..	..	..	0
Co-trimoxazole	80.1 (67.8–88.5)	90.7	839	14	89.9 (79.7–95.3)	92.9	1653	14	84.1 (71.9–91.7)	90.1	818	8	80.8 (71.7–87.5)	89.7	2342	13
Ciprofloxacin	25.3 (16.3–37.1)	79.1	951	17	35.4 (22.7–50.6)	92.8	1862	20	0.5 (0–24.7)	84.7	796	7	0.7 (0–9.7)	75.7	1492	12
Imipenem	0 (0–14.8)	..	23	1	1.1 (0.1–9)	0	454	6	0	..	0	0	0 (0–6.3)	..	57	1
Meropenem	0.9 (0.1–6.4)	0	106	5	2.7 (0.7–9.6)	73	751	8	0 (0–23.2)	..	14	1	0 (0-NaN)	0	881	2
Urine samples
ESCR	41.2 (30.1–53.3)	80.7	512	13	55.5 (28.1–79.9)	76.9	78	4	..	..	..	..	..	..	..	..
CRE	2.6 (0.8–7.7)	0.0	116	5	4.9 (0.7–27.5)	0.0	37	2	..	..	..	..	..	..	..	..
Ampicillin	95.9 (82.5–99.2)	37.8	253	8	..	..	..	..	..	..	..	..	..	..	..	..
Amoxicillin and clavulanate	71.6 (56.7–82.9)	80.9	285	10	75.5 (65.9–83.2)	15.3	94	5	..	..	..	..	..	..	..	..
Ceftriaxone	27.4 (18.8–38.1)	62.1	251	7	35.3 (11.3–70.2)	74.6	61	3	..	..	..	..	..	..	..	..
Gentamicin	40.7 (26.7–56.4)	71.2	250	9	64.3 (43.7–80.7)	70.3	74	4	..	..	..	..	..	..	..	..
Amikacin	3.1 (0.4–19.1)	0.0	32	2	11.8 (1.5–36.4)	..	17	1	..	..	..	..	..	..	..	..
Co-trimoxazole	68.1 (53.8–79.6)	77.2	298	10	85.5 (75.1–92.0)	0.0	78	4	..	..	..	..	..	..	..	..
Ciprofloxacin	19.2 (12.6–28.3)	54.4	336	11	23.2 (14.5–34.9)	31.7	37	5	..	..	..	..	..	..	..	..

‘n’ represents the number of isolates and ‘N’ represents the number of studies assessing for resistance to specific antibiotics.

CI = confidence interval. ESCR = extended-spectrum cephalosporin-resistant. CRE = carbapenem non-susceptible Enterobacterales. N/A = not applicable.

Carbapenem resistance was uncommon, with a pooled proportion of CRE in blood of 1.9% (95% CI: 0.5%–7.2%, N: 5) on 106 E. coli isolates and 2% (95% CI: 0.5%–7.5%, N: 11) on 897 Klebsiella spp isolates (Table 1 and Appendix p 18-19 for forest plots).

E. coli isolates were generally resistant to ampicillin (pooled resistance proportion 92.5%; 95% CI: 76.4%–97.9%, N: 13) and frequently also to ciprofloxacin (25.3%; 95% CI: 16.3%–37.1%, N: 17), cotrimoxazole (80.1%; 95% CI: 67.8%–88.5%, N: 14) and to gentamicin (42.7%; 95% CI: 30%–56.5%, N: 17). The pooled proportion of E. coli resistance to amikacin was lower, at 10.4% (95% CI: 16.3%–37.1%, N: 5).

We observed even higher pooled resistance proportions for Klebsiella spp. in blood samples, with isolates showing resistance pooled proportions of 77.6% to gentamicin on 1886 isolates (95% CI: 65.5%–86.3; N: 22), 89.9% to cotrimoxazole on 1653 isolates (95% CI: 79.7%–95.3%, N: 14), 35.4% to ciprofloxacin on 1862 isolates (95% CI: 22.7%–50.6%, N: 20) and 23.8% to amikacin on 782 isolates (95% CI: 10.3%–45.8%, N: 11) (Table 1 and Appendix pp 20–22 for forest plots).

Salmonella spp. and S. typhi found in blood cultures were generally resistant to cotrimoxazole but remained highly sensitive to ciprofloxacin (Table 1).

Urine samples

Pooled estimates ESCR proportions were 41.2% (95% CI: 30.1%–53.3%, N: 13) on 221 urine samples positive for E. coli and 55.5% (95% CI: 28.1%–79.9%, N: 4) on 78 urine samples positive for Klebsiella spp. (Table 1).

Stool samples

In stool samples, the susceptibility of Shigella spp. and Salmonella spp. to 3GCs was usually conserved, with a low pooled proportions of ESCR of 9% on 384 isolates (95% CI: 2.9%–25%, N: 12) and 11.8% on 425 isolates (95% CI: 9.0%–15.2%, N: 4), respectively. As in blood cultures, Salmonella spp. were generally resistant to cotrimoxazole, but remained highly sensitive to ciprofloxacin (Table 2).

Table 2.

Non-susceptibility of Enterobacterales in stool samples (all populations included).

	Shigella spp.				Salmonella spp.
	% (CI)	I2 (%)	n	N	% (CI)	I2	n	N
ESCR	9.0 (2.9–25.0)	25.7	384	12	11.8 (9.0–15.2)	11.8	425	4
CRE	0.0 (0.0–2.8)	..	128	1	0.0 (0.0–1.0)	..	360	1
Ampicillin	85.4 (69.1–93.9)	39.4	320	15	68.7 (34.2–90.3)	89.9	154	5
Amoxicillin and clavulanic acid	36.2 (13.0–68.3)	87.8	293	6	14.2 (7.1–26.4)	73.0	415	2
Ceftriaxone	12.0 (4.4–28.7)	44.4	225	10	3.4 (0.3–26.8)	0.0	48	2
Gentamicin	11.5 (4.2–28.1)	65.7	402	13	6.3 (2.8–13.4)	14.6	489	5
Amikacin	7.8 (0.2–81.3)	0	180	3	6.7 (0.5–52.9)	96.6	377	2
Co-trimoxazole	74.7 (62.1–84.1)	62.1	506	16	47.8 (17.3–80.1)	91.6	552	6
Ciprofloxacin	6.7 (1.5–24.5)	56.5	420	14	3.3 (0.4–20.3)	0.0	448	5

‘n’ represents the number of isolates and ‘N’ represents the number of studies assessing for resistance to specific antibiotics.

CI = confidence interval. ESCR = extended-spectrum cephalosporin-resistant. CRE = carbapenem non-susceptible Enterobacterales.

The proportion of Enterobacterales resistant to the various antibiotics calculated for all the samples together can be found in the Appendix on page 23.

Neonates

Sixteen studies presented data specific for neonates from 2118 positive blood cultures. Resistance to 3GCs in blood cultures was slightly but not significantly higher among neonates than among other children for E. coli and Klebsiella spp., with pooled resistance proportions of 48.6% (63/129; 95% CI: 36.3%–61.0%; I²: 46.9%) and 85.8% (579/728; 95% CI: 67.6%–94.6%; I²: 91.9%), respectively. However, we observed a significant higher proportion of E. coli resistant to gentamicin (56.0%; 95% CI: 32.0%–77.5%, N: 3) and Klebsiella spp resistant to amikacin in neonates (42.6%; 95% CI: 17.7%–71.9%, N: 6) compared to children (Table 3).

Table 3.

Non-susceptibility of Enterobacterales in blood cultures from neonates.

	Pooled prevalence (95% CI)	I2 (%)	n	N
Escherichia coli
ESCR	48.6 (36.3–61.0)	46.9	129	8
CRE	2.6 (0.4–16.5)	0.0	38	2
Ampicillin	93.8 (65.0–99.2)	0.2	136	7
Amoxicillin	97.2 (85.5–99.9)	..a	36	1
Ceftriaxone	50.0 (30.4–69.6)	73.6	115	6
Gentamicin	56.0 (32.0–77.5)	29.5	178	9
Amikacin	2.4 (0.0–66.6)	0.0	46	3
Ciprofloxacin	17.7 (7.2–37.3)	37.1	142	8
Meropenem	2.6 (0.4–16.5)	0.0	38	2
Klebsiella spp.
ESCR	85.8 (67.6–94.6)	91.9	728	14
CRE	2.3 (0.5–9.7)	77.7	463	6
Ceftriaxone	78.7 (41.9–95.0)	93.5	433	8
Gentamicin	77.1 (53.8–90.7)	92.9	688	12
Amikacin	42.6 (17.7–71.9)	93.2	370	6
Ciprofloxacin	27.6 (12.1–51.5)	92.1	655	11
Meropenem	4.5 (1.1–16.4)	83.3	336	4

‘n’ represents the number of isolates and ‘N’ represents the number of studies assessing for resistance to specific antibiotics.

CI = confidence interval. ESCR = extended-spectrum cephalosporin-resistant. CRE = carbapenem non-susceptible Enterobacterales.

^aHeterogeneity was not assessed because a single study reported this prevalence.

Heterogeneity

Across all the samples and specifically in blood cultures, the heterogeneity in proportion of ESCR Enterobacterales and Enterobacterales resistant to aminoglycosides were very high (Table 4 and Appendix pp 16–17 and 20–22). The following potential sources of heterogeneity were investigated.

Table 4.

Sources of heterogeneity for ESCR proportion in blood cultures.

	Escherichia coli			Klebsiella spp.
	N	Pooled prev. (95% CI)	I2 (%)	N	Pooled prev. (95% CI)	I2 (%)
Geographic regiona
West	8	43.2 (22.4–66.8)	86.5	9	78.5 (42.8–94.7)	94.9
Central	0	..	..	1	89.5 (66.3–97.4)	..
East	11	42.2 (27.1–59.0)	77.0	12	85.6 (71.3–93.4)	92.9
South	1	22.3 (18.0–27.2)	..	5	91.5 (69.8–98.1)	94.0
p-value		0.0101			0.7785
Age range
Neonates	8	48.6 (36.3–61.0)	46.9	14	85.8 (67.6–94.6)	91.9
Children over 1 monthb	11	36.5 (19.1–58.3)	86.6	12	83.8 (65.3–93.4)	95.8
p-value		0.3394			0.8352
Periods
2004–2010	5	37.4 (7.6–81.2)	89.1	6	73.5 (32.7–94.0)	82.8
2011–2015	5	38.9 (23.1–57.4)	78.5	7	79.6 (48.8–94.1)	93.2
2016–2022	9	45.4 (30.5–61.2)	79.0	13	90.5 (79.7–95.8)	91.7
p-value		0.8491			0.3427

Prev. = prevalence. CI = confidence interval. N = number of included studies.

^aRecording the WHO African region. South Africa was excluded.

^bIncluding those listed as of paediatric age or children.

Age groups (neonates versus children)

In blood cultures, no significant difference was found between neonates and children in the prevalence of ESCR E. coli (p-value 0.34) and ESCR Klebsiella spp. (p-value 0.83). However, proportion of E coli resistant to gentamicin was significantly higher among neonates (p-value 0.03). The factors affecting the heterogeneity of E. coli resistant to amikacin could not be investigated due to the small number of studies (N: 5) reporting this outcome.

Significant higher proportion of Klebsiella spp. resistant to amikacin was found among neonates (p-value 0.01) but no significant difference was found for Klebsiella spp. resistant to gentamicin (p-value 0.97) (Table 4 and Appendix pp 24–30).

Regions

Reported resistance proportions among blood cultures revealed no significant differences between regions for ESCR Klebsiella spp, for Klebsiella spp. resistant to gentamicin and amikacin, or for E. coli resistant to gentamicin. The only significant inter-regional difference was a lower prevalence of ESCR E. coli in southern SSA compared to eastern and western regions. However, as only one study for the southern region was included, this result should be treated cautiously (Table 4, Appendix pp 24–30).

Period

We noticed a non-significant trend towards an overall increase in ESCR E. coli and ESCR Klebsiella spp. in blood cultures from 2004 to 2022 (Table 4 and Appendix pp 24–27). Although no significant differences in E. coli resistant to gentamicin were found over time, we found significant increases in Klebsiella spp. resistant to gentamicin and amikacin when comparing 2004–2010, 2011–2015 and 2016–2022 (p-value < 0.01) (Appendix pp 28–30).

Quality of studies and risk of bias assessment

Because their respective studies included subgroups of children, such as neonates, 33.6% (57/122) of the populations examined were not representative of SSA’s general paediatric population. In about 7.3% (9/122) of studies, descriptions of participants were considered insufficient because children’s precise ages were not given. In 31.1% of studies (38/122), the samples were not sterile (stool, urine, smear) with the risk of interpreting colonising bacteria as pathogenic (exposure bias) (Appendix pp 6–7).

Discussion

We identified 122 studies from SSA providing data on AMR from more than 30,000 Enterobacterales isolates obtained from children and neonates with bacterial infections. In blood samples, we found high proportions of ESCR E. coli and Klebsiella spp., of 40.6% and 84.9% of isolates, respectively. Most E. coli isolates were resistant to ampicillin. Around half of E. coli and most Klebsiella spp. isolates were resistant to gentamicin. In addition, 25.3% of E. coli and 35.4% of Klebsiella spp. pathogens were resistant to ciprofloxacin, and 80.1% of E. coli and most Klebsiella spp., Salmonella spp. and S. typhi isolates were resistant to cotrimoxazole. E. coli isolates were usually susceptible to amikacin, but 24.7% of Klebsiella spp. were resistant to that antibiotic. In contrast to ESCR, resistance to carbapenem was low. We did not find any significant differences in patterns of C3G resistance between neonates and children. However, we observed a significant higher proportion of E. coli resistant to gentamicin and Klebsiella spp resistant to amikacin in neonates compared to children. In urine samples, around half of E. coli and Klebsiella spp. isolates were resistant to cephalosporins. Finally, Salmonella spp. in stool samples were generally resistant to ampicillin and cotrimoxazole. These findings indicate that Enterobacterales had high proportions of resistance to several of the antibiotics that are usually recommended in clinical guidelines used in SSA and in WHO guidelines.⁹

We found higher overall proportions of ESCR E. coli and Klebsiella spp. isolates than previous studies (Table 5).^3,8,139, 140, 141, 142, 143, 144, 145, 146 The most recent meta-analysis by the World Health Organization (WHO), which included children and adults in Africa, reported a prevalence of cephalosporin resistance of 10.2% in E. coli isolates and 3.3% in Klebsiella spp. isolates, but this was based on very small numbers of samples (123 and 57, respectively). On the other hand, a recent systematic review by Wen et al., focusing on neonates and including studies from 2010 to 2021 found ESCR rates of 88% among Klebsiella spp. These results are in line with our findings.¹⁴³

Table 5.

Meta-analyses reporting on the 3GC resistance of E. coli and Klebsiella spp. in Africa.

Authors, year	Population	Inclusion period	Number of studies included	Countries	Sample types	% E. coli resistance	% Klebsiella spp. resistance
Williams et al., 2016	Children	2005–2016	18	Sub-Saharan Africa	Blood mainly	16% (95% CI: 12%–34%) N = 387	Range 33%–50% (NA) N = 110
Le Doare et al., 2015	Children	2002–2013	10	Africa	Blood	0 (IQR: 0%–10.9%) N = 338	30% (IQR: 0%–64.0%) N = 921
Le Doare et al., 2015	Neonates	2002–2013	3	Africa	Blood	0 (IQR: 0%–50%) N = 7	50% (IQR: 0%–86.5%) N = 109
Droz et al., 2019	Children	1990–2019	11	Africa	Blood	21.2% (95% CI: NA)	NA
Bernabe et al., 2017	Children and adults	1990–2012	120 (112 for BC)	West Africa	Blooda	11.9% (95% CI: 4.3%–22%) N = 720	24.2% (95% CI: 8.1%–44.8%) N = 476
Tadesse et al., 2017	Children and adults	2013–2016	144	Africa	Blood, urine, pus	31.5% (IQR: 6.9%–47.7%) N = 2800	47.3% (IQR: 25%–62.8%) N = 1547
Okomo et al., 2019	Neonates	2008–2018	151	Sub-Saharan Africa	Blood, CSF	Around 33% (“one third”) (95% CI: NA) N = NA	49% (95% CI: 47%–83%) N = NA
Wen et al., 2021	Neonates	2010–2021	88	Africab	Blood	47% (95% CI: 28%–68%) N = 474	88% (95% CI: 72%–96%) N = 1209
WHO Regional Office for Africa	Children and adults	2016–2020	167	WHO African region	Blood, urine	10.2% (95% CI: NA) N = NA	3.3% (95% CI: NA) N = NA
Haindongo et al., 2022	Children and adults	2008–2019	27	Africa	Blood	32% (95% CI: 9%–69%) N = 416	NA
Lester et al., 2020	Children and adults	1990–2019	40	Sub-Saharan Africa	Blood	18.4% (95% CI:10.5%–35.2%); N = NA	54.4% (95% CI: 24.3%–81.2%) N = NA

BC = blood culture, CSF = cerebrospinal fluid, IQR = interquartile range, N = number of isolates, NA = Not available, WHO = World Health Organization, CI = confidence interval.

^aWe only report blood culture results, but Bernabe et al. also reported results for urine, cerebrospinal fluid, stools, blood, sputum, bronchoalveolar liquid and pleural fluid.

^bWen et al. also collected data from Asia and the Middle East, but we only report their data from Africa.

The high proportion of ESCR Klebsiella spp. were found also in neonates. This is an alarming finding given that a recent review identified Klebsiella spp. as the leading cause of neonatal sepsis in Africa.¹⁴⁷ Indeed, many sub-Saharan countries have reported nosocomial outbreaks of resistant Klebsiella spp. infections in neonatal units.148, 149, 150

We observed a high level of heterogeneity in the proportion of patients with Klebsiella spp. resistant to amikacin. It should be noted that this result was strongly influenced by data from two studies focusing on neonates and involving only one hospital.^79,110 These high proportions could have been the result of ongoing outbreaks of Klebsiella spp. resistant to amikacin or of a diagnostic bias as blood cultures might be ‘reserved’ for use in the most severe cases or for multi-treated children.

The overuse of antibiotics is a key cause of AMR.¹⁵¹ Saleem et al. reviewed current patterns of antibiotic consumption in Africa’s hospitals. Rates of antimicrobial prescription were high across most of the countries included in our study. Hospitals in Nigeria saw the highest antimicrobial utilization rates at 59.6–97.6% of in-patients surveyed.¹⁵² Moreover, the frequent use of over-the-counter antibiotics in SSA is a major barrier to antibiotic stewardship.¹⁵³ The worldwide dissemination of successful ESBL-producing Enterobacterales clones over the last decade could also partly explain the high rates of resistance that we found.154, 155, 156, 157

There was a lot of heterogeneity in the results among the reviewed studies. Multiple factors may be at play, such as different access and exposure to antibiotics across different settings, use of antibiotics in local food industries, administration of first-line antibiotics prior to blood cultures, difference in microbiological methods used by the hospital laboratories, and bias due to patient selection for culturing and variable culturing rates in different settings.

Four recent studies focusing on blood cultures were published after our literature search and could not be included in this review. They assessed neonates with sepsis in Ethiopia^158,159 and Zambia¹⁶⁰ and children under 15 years old (including neonates) suffering of suspected bacterial infections in Rwanda.¹⁶¹ The three studies on neonates found a majority of Klebsiella spp in positive blood cultures, which showed ESCR proportions between 95 and 100%.158, 159, 160 In the Rwandan paediatric study, most of the positive blood cultures concerned neonates and also showed a majority of Klebsiella spp, with an ESCR proportion of 100%.¹⁶¹ These recent studies illustrate once again the very high proportion of Klebsiella spp ESCR in children in Sub-Saharan Africa. They showed an even higher proportion of resistance than those found in our meta-analysis suggesting a possible worsening over time.

There are some limitations that should be considered when interpreting our results. The ability to perform a culture depends on available resources, including trained personnel who can gather sterile samples and process them, access to a microbiology laboratory and the affordability of the test (often charged to the patient). Moreover, the indications for performing cultures differ in each healthcare centre, leading to multiple biases in patient selection in the studies included in this systematic review. In addition, bacteriology laboratories are not widely available in SSA, usually only in large hospitals. The sickest children, or those who have not responded to first-line treatments, will be referred to these tertiary centres, inducing a supplementary selection bias. In some studies, the decision to perform blood cultures may have been driven by suspicions of an infectious outbreak, particularly in neonatal units. All of these reasons may have led to an over-representation of multi-resistant Enterobacterales. Results from stool and urine samples must also be interpreted with caution, as these might indicate colonization rather than infection, as definitions of ‘symptomatic patients’ varied across studies and were often unclear. In addition, most of the included studies were from the western and eastern regions of SSA and we excluded South Africa limiting the representativeness of these findings.

Since we could not distinguish community-acquired from hospital-acquired infections, as this information was not provided, it is difficult to indicate whether the presented results are representative of AMR in the community or hospital settings.

Despite these limitations, and to the best of our knowledge, this review is the most comprehensive synthesis of AMR Enterobacterales in children from SSA, including data from a large number of isolates collected in 24 African countries, applying robust statistical methods to generate pooled estimates.

Our findings suggest that antibiotics such as ampicillin, gentamicin and ceftriaxone, which are the currently recommended treatment for sepsis in children,⁹ may frequently be ineffective. Amikacin (the current WHO-recommended second-line treatment for neonatal sepsis in combination with cloxacillin) may represent a better alternative for bloodstream infections and could help to save carbapenems.

These data, and the limitations inherent in our meta-analysis, showed the crucial importance of conducting prospective studies on drug resistance among children in SSA in the community and in primary and secondary healthcare structures.

Local microbiological data should inform antibiotic usage recommendations, but these are scarce in SSA.¹⁶² At the individual patient level, this information is essential to informing therapeutic decisions and avoiding any unnecessary escalation to second-line regimens. At the hospital level, this information is crucial to identifying potential sources of nosocomial infections and informing infection prevention and control measures. At the regional and national levels, microbiological data is key to monitoring the spread of AMR and informing the development of effective regional or national clinical guidelines.

Although microbiology laboratories are costly and require highly trained personnel, recent initiatives could help to scale up microbiology capacities in SSA. For example, Doctors Without Borders and their partners are currently testing the feasibility of using a transportable, stand-alone clinical bacteriology laboratory that can be operated by relatively inexperienced technicians in remote settings.¹⁶³

In conclusion, even if more research is needed to address some of this review’s limitations and to have a better picture of AMR at the community and hospital levels, the present review suggests that the situation concerning AMR in SSA is alarming. Adapting antibiotic recommendations at local level, based on local patterns of AMR, are urgently needed to provide effective treatments to patients. This will require a dramatic scaling up of SSA’s microbiology capacity. The present review also suggests that some antibiotics remain highly effective and their efficacy must be protected. Antimicrobial stewardship measures, which have been extensively described in the literature, must be expanded and strengthened at the community level, in the food industry, and in all healthcare services. Relevant policy-makers and key stakeholders should take ownership of the importance of addressing AMR. They should prioritise the development of policies addressing its determinants and secure funding to implement the measures mentioned above.

Contributors

Every author contributed to the study. MK, GC, AM, CC, AGL, and NW participated in the conceptualization; MK, BO, GC, MR, AGL, and NW in the data curation; AM and CC in the formal analysis; MK, BO, GC, AM, MR; AGL, and NW in the investigation; MK GC, AM, CC, AGL, and NW in the methodology; AGL and NW in the project administration; AM in providing analysis tools; JD, KN, MdK, AGL and NW in supervision; MK, BO, GC, AM, DA, MR, JD, KN, MdK, CC, AGL and NW in validation of the data; MK, BO; AGL, and NW in visualization and data presentation; MK, BO, JD, CC, and NM in writing the original draft; and MK, BO, GC, AM, DA, JD, KN, MdK, CC, AGL, and NW in writing, review and editing the final draft. All authors have full access to all the data in the study and had final responsibility for the decision to submit for publication. AGL and MR directly accessed and verified the underlying data reported in the manuscript.

Data sharing statement

The protocol study is registered and available in Prospero, ID: CRD42021260157. All the data collected will be shared following publication upon request without end date, to anyone who wishes to access the data for any purpose. Requests should be directed to noemie.wagner@hcuge.ch.

Declaration of interests

We declare no competing interests.