Introduction

Antimicrobial resistance (AMR) is a global health crisis, threatening to reverse decades of progress in the treatment of bacterial infections1. While AMR poses a universal challenge, its impact is particularly devastating in fragile healthcare systems, such as those in resource-limited settings across Africa and Asia. In these regions, limited access to diagnostics, unregulated inappropriate antibiotic use, and inadequate infection control measures create ideal conditions for the unchecked spread of resistant pathogens2,3.

Recent estimates indicate that AMR was responsible for over 1.2 million deaths worldwide in 2019, with the highest per capita mortality recorded in Africa, particularly in sub-Saharan Africa (SSA)3. Cephalosporin resistance is particularly concerning, occurring through reduced drug penetration, penicillin-binding protein modifications, or β-lactamase-mediated inactivation, compromising efficacy and promoting resistance spread4. Third-generation cephalosporins (3GCs) includes broad-spectrum β-lactam antibiotics such as ceftriaxone, cefotaxime, and ceftazidime, used from Gram-negative infections; 3GCs are among the most widely used antibiotics in SSA, frequently used for first-line treatment across diverse patient populations, including adults, children, and pregnant women5. Recognizing the urgency of the issue, the WHO has placed third-generation cephalosporin-resistant (3GC-R) Enterobacterales in the highest critical priority group, alongside carbapenem-resistant Enterobacterales and Acinetobacter baumannii6. Despite this classification, the broad-spectrum activity of 3GCs means they remain a cornerstone of empirical treatment in many low-resource settings, where alternative antibiotic classes are often unavailable in cases of resistance7,8. This reliance on 3GCs, coupled with rising resistance rates, significantly increases the risk of clinical treatment failure, prolonged hospital stays, poorer health outcomes, and mortality9.

In Africa Region, and particularly in SSA, reports of resistant pathogens are becoming increasingly prevalent10,11,12. However, surveillance systems in the region remain fragmented, often failing to capture the true extent of the problem. As highlighted by Okolie et al., existing monitoring systems lack uniformity and comprehensive coverage, especially in rural and low-resource areas13. This gap not only hampers efforts to track resistance trends but also restricts the development of targeted interventions.

To enhance estimates of the burden of AMR and effectively prioritize resource allocation for AMR mitigation strategies, we conduct a systematic review and meta-analysis, providing a thorough evaluation of 3GC-resistance in SSA, including its prevalence, pathogen-specific patterns, temporal trends, and regional disparities. By addressing this critical knowledge gap, our work aims to offer insights that can inform future surveillance and intervention strategies.

Methods

Protocol registration

This study was conducted following the recommendations in the Cochrane handbook for systematic literature reviews14. This systematic review and meta-analysis was reported following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, updated version to 202015. The protocol was registered in PROSPERO (CRD42025635495).

Information sources and search strategies

The research question for this systematic review was: “What is the prevalence of 3GC-R pathogens in SSA?” We searched Medline (via Ovid), Scopus and Web of Science from database inception to 16th August 2024 using a Boolean search strategy. The search for individual studies in these bibliographic databases was supplemented by a manual search of reference lists included in identified articles. The search strategies, detailed by each database, are reported in Supplementary Data 1.

Eligibility criteria

Inclusion criteria comprised the following: (1) observational studies (retrospective, cross-sectional and cohort/longitudinal studies); (2) studies that reported the prevalence of any 3GC-R pathogens among Gram-negative clinical samples; (3) studies carried out in SSA; (4) studies published in all indexed language filters. In the cohort/longitudinal studies, data about prevalence were extracted. Publications were excluded if they: (1) reported data on bacteria not causing a clinical infection; (2) were case reports or case series; (3) included only Salmonella spp. or Shigella spp.; (4) were produced in animals. Infections due to Salmonella and Shigella were excluded in order to reduce the heterogeneity of the study, as 3GC are less commonly used as first-line treatment for these infections.

Study selection

The management of potentially eligible references, at title/abstract level, was carried out using the Rayyan website (https://www.rayyan.ai/) by two researchers. Disagreements were resolved by a third senior author (FDG). The study selection process involved, first, a screening based on titles and abstracts, then a screening of studies retrieved from this first step based on the full-text manuscripts.

Data collection and data items

We extracted the following information: data about the manuscript (first author, title, year, DOI); general data about the study and demographics (country where the study was conducted, if the study was conducted on multiple countries, sample size, percentage of females, age, if the study was conducted in a university hospital and in a rural or urban setting); characteristics about pathogens, such as setting, type of infection, name of pathogen, total number of pathogen isolates and number of 3GCR pathogen isolates. These data were collected using REDCap16. The data extraction was carried out independently by one author (GG), with another senior author (NV) checking the quality of data extraction.

Risk of bias evaluation

The Hoy et al. framework was used to assess the study quality/risk of bias17,18. It consists of 9 domains that assess both external validity (sampling, response rate, and representativeness) and internal validity (measurement methods, case definition, and bias in data collection). The external validity domains encompass questions 1 to 4, whereas the internal validity domains include questions 5 to 9. Each domain is rated as having a low or high risk of bias, and an overall score helps classify studies as low, moderate, or high risk of bias.

Statistics and reproducibility

The cumulative prevalence and 95% confidence intervals (CIs) were estimated using a meta-analysis, under a random-effects model19. We calculated the pooled estimate after Freeman-Tukey Double Arcsine Transformation to stabilize the variances and to better account for works reporting a prevalence of “zero” cases20.

Heterogeneity between estimates was assessed using the I2 statistic. In case of an I2 over 50%, a series of sensitivity analyses (Year, categorized as intervals in five years; University hospital, yes vs. no; Rural/urban, yes, no, mixed; setting; type of infection, and geographical Area) were performed. Given that the year of sample collection was often not reported, the year of publication was used as a proxy for temporal classification, stratifying studies into four periods. Moreover, we performed two meta-regression analyses, using, as moderators, mean age and the percentage of females. Publication bias was assessed by visually inspecting funnel plots and using Egger bias test, with a p < 0.05 indicative of possible publication bias21. In case of publication bias, trim-and-fill analysis was performed22. All analyses were performed using “metaprop”, a command available in STATA 14.0.

Results

Literature search

As shown in Fig. 1, after removing duplicates, among 1382 articles initially screened, we evaluated the full text of 688 manuscripts: for 13 articles, full texts were not available, despite contact with the first/corresponding author via e-mail. Therefore, we examined 675 full-texts and, of them, 200 were included. The main reasons for exclusion were publications reporting data about bacteria not causing infections (n = 169) or case reports/case series (n = 60). The full list of the 200 manuscripts included is reported in Supplementary References.

Fig. 1: PRISMA flow-chart.
Fig. 1: PRISMA flow-chart.
Full size image

Flow diagram illustrating the identification, screening, eligibility assessment, and inclusion of studies in the systematic review.

Descriptive results and risk of bias

As shown in Supplementary Data 2, the 200 articles included a total of 355,136 sub-Saharan patients affected by a possible 3GC-R infection. The most represented country was Ethiopia (42 citations, 19.1% of all included studies) and most (50.5%) studies were conducted after 2019. When talking about the type of hospital involved in the diagnosis and treatment of these patients, university hospitals were involved in 51.4% of the studies, while exact information about rural and urban areas was missing in 117 studies (55.7%). Infections were mainly community-acquired (21.4%), but this kind of information was missing in 119 studies. Finally, urinary tract infections were the most common sources of infections among those listed (Supplementary Data 2). Data were present for A. baumannii, Citrobacter spp., E. coli, Enterobacter spp., K. pneumoniae, P. aeruginosa, and Proteus spp., whilst information about other pathogens were missing.

Meta-analysis

The main findings of our study are presented in Table 1 and visually depicted as a forest plot in Fig. 2.

Fig. 2: Prevalence of ESBL pathogen in Sub-Saharan Africa.
Fig. 2: Prevalence of ESBL pathogen in Sub-Saharan Africa.
Full size image

Forest plot of pathogen prevalence with 95% confidence intervals. Pooled prevalence estimates of major Gram-negative pathogens, with 95% confidence intervals and study/participant counts.

Table 1 Main results of the meta-analysis
Full size table

For infections caused by E. coli, data from 180 studies involving 44,073 patients revealed an overall weighted prevalence of 3GCR of 42.02% (95% CI: 37.05–47.06%). Although publication bias was detected, the trim-and-fill analysis did not change the results.

Infections caused by K. pneumoniae were analyzed using data from 163 studies and 17,134 patients, showing an overall 3GCR prevalence of 57.68% (95% CI: 52.41–62.88%). Despite high heterogeneity and publication bias, the trim-and-fill analysis confirmed the robustness of these findings.

For A. baumannii infections, data from 28 studies involving 693 patients indicated an overall 3GCR prevalence of 47.59% (95% CI: 33.26–62.09%), with substantial heterogeneity (I² = 90.68%). While publication bias was absent, trimming 25 studies from the left of the mean in the trim-and-fill analysis suggested a slightly higher 3GCR prevalence of 50.5%.

In the case of Citrobacter spp., data from 14 studies including 403 patients showed a 3GCR prevalence of 54.27% (95% CI: 34.74–73.24%), also marked by high heterogeneity. After trimming 16 studies from the left of the mean, the trim-and-fill analysis indicated lower prevalence rates (Table 1).

Data from 38 studies, comprising 938 patients, revealed a 3GCR prevalence of 54.21% (95% CI: 40.81–67.36%) for Enterobacter spp.

Finally, the overall 3GCR prevalence was found to be 39.13% (95% CI: 27.98–50.82%) for P. aeruginosa and 26.01% for Proteus spp. (Table 1, Fig. 2).

In summary, across the 200 included studies, the weighted prevalence of 3GCR infections was approximately 50% (prevalence = 45.34%; 95% CI: 36.68–54.39%), with significant heterogeneity observed (Table 1).

Sensitivity and meta-regression analyses

To explain heterogeneity, several sensitivity analyses were run, as reported in Tables 2 and 3. When considering the year of publication, we found that the prevalence of 3GCR pathogens was higher in 2015–2020 interval period and after 2020, compared to before 2009. Also, a significantly higher prevalence of 3GCR-E. coli and Proteus spp. was found in university hospitals compared to the others, whilst no differences were found for the other pathogens examined. Similarly, except for A. baumannii, the prevalence of 3GCR infections was higher in mixed and rural areas, compared to urban ones (Table 3). When analyzing the setting/ward, it appeared that, regardless from the infection, the prevalence rate was significantly higher in general surgery and lower among outpatients, with surgical sites being the most common location of infections. Finally, as graphically reported in Fig. 3, the prevalence of 3GCR infections was higher in East and West Africa compared to Central Africa.

Fig. 3: Graphical representation of prevalence of 3GCR pathogens among the 4 regions of SSA.
Fig. 3: Graphical representation of prevalence of 3GCR pathogens among the 4 regions of SSA.
Full size image

Map of SSA showing pooled prevalence estimates and 95% confidence intervals for key Gram-negative pathogens by subregion (West, Central, East, and Southern Africa).

Table 2 Prevalence of third-generation cephalosporin resistance in E. coli, K. pneumoniae, Proteus spp., and Enterobacter spp
Full size table
Table 3 Prevalence of third-generation cephalosporin resistance in P. aeruginosa and A. baumannii by year, setting, infection type, and region
Full size table

Finally, we ran several meta-regression analyses using as moderators the mean age of the sample size included and the percentage of females. However, these two moderators were not significantly associated with any of the prevalence rates analyzed.

Quality assessment

The 200 included studies were analyzed to evaluate the overall risk of bias. The mean risk of bias score across all studies was 2.60, with a standard deviation (SD) of 1.60. The scores ranged from 0 to 8, with a median value of 2, indicating that the majority of studies exhibited a relatively low risk of bias (Supplementary Data 3).

In terms of categorical distribution, 131 (65%) studies fell within the low-risk category (scores 0–3). A moderate risk of bias (scores 4–6) was observed in 64 manuscripts (32%), while only 5 manuscripts (3%) were classified as having a high risk of bias (scores 7–9). The risk of bias could not explain the high heterogeneity found in our results since, the overall prevalence in low-risk bias studies was 51.43%; 95%CI: 44.41–58.43; moderate risk: prevalence: 39.60%; 95%CI: 33.10–46.28; high risk: prevalence: 47.67%; 95%CI: 42.84–52.52. In all the three groups the I2 was 99% and the p for interaction was 0.06, indicating not statistically significant differences by risk of bias.

These findings suggest that, overall, the assessed manuscripts maintain a predominantly low to moderate risk of bias, with a minimal proportion of manuscripts classified as high risk. This distribution highlights a generally acceptable methodological quality across the analyzed dataset (Table 1).

Discussion

Our systematic review and meta-analysis of 200 studies and 355,136 sub-Saharan patients reveal a significant and escalating burden of 3GCR among Gram-negative pathogens in SSA, with an overall resistance prevalence of approximately 45%. Resistance rates vary notably by pathogen, region, and clinical setting, with a marked increase observed after 2015. These findings underscore the urgent need for targeted interventions to strengthen antimicrobial stewardship and surveillance in SSA23,24. Temporal trends reveal a concerning increase in resistance rates over the past decade; in fact, as shown in Table 2 and Table 3, there is a significant increase in resistance across different study periods: before 2009, 2010–2014, 2015–2020, and studies conducted after 2020 until the end of the search strategy. This trend highlights a rapid acceleration of resistance, which is observed in Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa, Acinetobacter baumannii, Proteus spp., and Enterobacter spp. This confirms the spread of AMR in SSA and the high risk of AMR-related mortality, as reported by numerous scientific studies25,26, however, this could also be attributed to an improvement in surveillance systems, such as GLASS27,28. Our findings are consistent with the alarming data reported by the GBD AMR Collaborators on sepsis-related mortality caused by third-generation cephalosporin-resistant pathogens in SSA. These pathogens remain among the leading contributors to AMR-related deaths in the region, with a persistently increasing trend that underscores the urgent need for targeted interventions and enhanced antimicrobial stewardship3. Notably, E. coli and K. pneumoniae are identified by GBD-AMR as both the pathogens with some of the highest prevalence of third-generation cephalosporin resistance and among the leading causes of AMR-attributable mortality worldwide. Our meta-analysis supports these prevalence patterns, suggesting a potential correlation between high resistance prevalence and mortality risk, although prevalence data alone cannot fully capture patient-level outcomes or treatment failure. Moreover, according to GBD-AMR, mortality vary widely across SSA, with higher rates in parts of West Africa.

Our review reflects similar geographical disparities in resistance prevalence, with East and West Africa showing the highest levels compared to Central Africa. This alignment between resistance prevalence patterns and mortality estimates supports the hypothesis that regions with a higher burden of antimicrobial resistance may also face disproportionately higher mortality, underscoring the need for targeted surveillance and intervention efforts. This geographical variation may also be attributed to multiple factors, including differences in healthcare infrastructure, antibiotic prescribing and acquisition practices, and the availability of AMR surveillance systems29. The relatively more developed healthcare systems in East and West Africa may contribute to higher reported resistance due to better diagnostic capacity, as well as the increased availability of newer antibiotics in the absence of adequate infection control policies27. However, these regions may also experience greater selective pressure from higher antibiotic consumption, both in human and veterinary medicine29.

Additionally, weaker antibiotic regulation, higher urbanization rates, climate and environmental changes, increased population mobility and greater exposure to international travel and trade could further contribute to the observed differences. The presence of large urban centers may facilitate the transmission of resistant pathogens, while economic disparities and inconsistent access to high-quality antibiotics might also play a role in shaping resistance patterns30. Furthermore, a higher burden of counterfeit or substandard antibiotics in certain regions may lead to suboptimal treatment and selection of resistant strains31. Differences in agricultural and livestock antibiotic use could also be relevant, as unregulated antimicrobial use in food production may drive resistance in zoonotic pathogens32. Lastly, variations in AMR surveillance coverage across regions could result in underreporting in some areas, particularly in Central Africa, where healthcare infrastructure and laboratory capacity remain more limited.

Variations were also observed between rural and urban areas, with a higher prevalence of 3GC-R pathogens in rural regions. This trend may be explained by several factors. First, rural areas often have reduced diagnostic capabilities, leading to a greater reliance on empirical treatment. This, in turn, increases the likelihood of indiscriminate or unregulated antibiotic use, which fosters the selection of resistant strains33. Moreover, healthcare infrastructure in rural regions is often inadequate, with limited access to trained specialists and a restricted availability of second-line or alternative antibiotics. This lack of resources may force healthcare providers to prescribe suboptimal or outdated antibiotic regimens, further compounding the issue. In addition, weaker surveillance and stewardship programs in rural areas may result in a delayed response to emerging resistance patterns, allowing resistant strains to become entrenched in the community34. Second, poverty levels tend to be higher in rural settings, potentially contributing to the discontinuation of antibiotic therapy due to financial constraints, especially in the absence of free healthcare services. Interrupting treatment courses can promote the survival of partially resistant bacterial populations, accelerating the spread of resistance. Taken together, these factors highlight that rural healthcare remains the most neglected and vulnerable segment of an already fragile health system in SSA, where access to appropriate medical care is often severely limited35, and where interventional antibiotic policy plan should be prioritized and targeted.

Pathogen-specific analysis revealed that K. pneumoniae and Enterobacter spp. exhibited the highest prevalence of resistance, with both pathogens showing significant impacts in critical care settings, such as neonatal intensive care units and surgical wards. These findings are particularly alarming given the vulnerability of neonates and surgical patients to multidrug-resistant infections.

E. coli, although slightly less resistant, remains the dominant pathogen in community-acquired infections such as urinary tract infections, emphasizing the need for surveillance at the community level. The lower resistance rates observed for Proteus spp. and A. baumannii warrant further investigation to determine if these are reflective of true epidemiological differences or limitations in data availability. Although A. baumannii is recognized by the WHO as one of the pathogens of concern4, classified within the critical priority group, the low prevalence of A. baumannii may be attributed to the limited bed availability in ICUs and the limited use of frequently colonized medical devices36, such as invasive ventilation and central venous catheters.

Clinical settings played a pivotal role in the variability of resistance prevalence. The higher resistance in studies reporting results from mixed wards, including intensive care and general surgery units, indicates the critical role of hospital-acquired infections, while also highlighting the need for robust antimicrobial stewardship in the AMR crisis. Data from individual African countries demonstrate prevalence rates of 3GC-R pathogens that are even higher in specific settings, with rates reaching up to 98% in cases of healthcare-associated infections (HAIs)37.

The lack of harmonized and comprehensive AMR surveillance systems in SSA remains a significant challenge38. Fragmented surveillance limits the ability to capture the true scale of resistance and impedes the development of effective intervention strategies. Investments in laboratory infrastructure and diagnostic capacity, as well as regionally coordinated surveillance networks, are essential to address this critical gap. A meta-analysis focusing on HAIs in SSA has highlighted how limitations related to AMR and infection control measures have resulted in a significantly higher prevalence of HAIs compared to other regions, such as Asia and Europe39. Notably, certain countries, including Ethiopia, have reported even higher peaks, further confirming the previously mentioned findings40.

As previously examined with GBD-AMR, studies consistently associate 3GCR infections with markedly elevated mortality, increased treatment failure rates, and prolonged hospitalizations41. Concurrently, healthcare facilities in resource-limited regions may exhibit deficiencies in infection prevention and control protocols, while antimicrobial stewardship and diagnostic capabilities may be constrained27. This circumstance may promote and perpetuate the transfer of resistant strains. Connecting elevated prevalence rates to these outcomes underscores the necessity of prioritizing AMR surveillance, enhancing antimicrobial stewardship initiatives, and augmenting access to effective antimicrobials, especially in high-burden areas of SSA. The strength of this study lies in its comprehensive aggregation of available evidence and synthesis of data from 200 studies, encompassing more than 350,000 patients across SSA. However, several limitations should be considered when interpreting the findings. First, data extraction was performed by a single reviewer and subsequently verified by a senior colleague, rather than through dual independent extraction. While this approach helped ensure a high level of consistency and accuracy, the absence of dual independent extraction remains a methodological limitation, as duplicate screening and extraction are considered best practice to further reduce the risk of human error. This is particularly relevant in the context of antimicrobial resistance research, where the complexity of resistance definitions and heterogeneous stratification variables (e.g., infection source, clinical setting, and geographic classification) can increase the potential for misclassification or omission. However, the combination of rigorous piloting of the extraction template, structured verification by a second reviewer with extensive infectious diseases expertise, and a transparent reporting framework mitigates these risks and strengthens confidence in the integrity of the extracted data. Despite our stratified and sensitivity analyses, heterogeneity remained extremely high across subgroups (I² consistently ~99%), indicating that much of the observed variability cannot be explained by study-level factors such as setting, pathogen, or risk of bias. This limitation reduces the interpretability of pooled prevalence estimates, which should therefore be considered as an indication of the overall magnitude and pervasiveness of resistance in the region rather than as precise prevalence values. Moreover, pooled prevalence estimates cannot readily be linked to clinical outcomes such as transmission risk, treatment failure, or mortality. As such, while our findings underscore the alarming scale of third-generation cephalosporin resistance in Sub-Saharan Africa, they also highlight the urgent need for harmonized, high-quality surveillance systems that can better clarify the relationship between prevalence data and patient outcomes. A further limitation is the classification of study settings as rural, urban, or mixed, which was based solely on the descriptions provided by the original articles. In most cases, formal definitions or national census criteria were not reported; this lack of standardized epidemiologic variables, together with incomplete reporting of infection settings (hospital vs. community), limits the reliability, comparability, and public health interpretation of the pooled AMR prevalence estimates. This review has two other methodological limitations: first, we did not perform a formal search of the gray literature, which may have resulted in the omission of relevant unpublished or non-indexed studies. Second, publications in African national or local languages were not captured, as such articles are largely absent from major bibliographic databases, reflecting a structural gap in global indexing that may disproportionately underrepresent regionally published evidence. Despite these limitations, this systematic review and meta-analysis provides critical insights into the epidemiological dynamics of cephalosporin resistance in SSA, serving as a foundation for targeted policy interventions, improved surveillance, and resource allocation to address the growing AMR crisis in the region.

Beyond these methodological considerations, our findings raise major public health concerns, as cephalosporins are among the most widely used antibiotics in SSA. They very often represent the first-line treatment for post-surgical infections and pneumonia and are commonly used as prophylaxis for cesarean sections. Their widespread use is also driven by their broad-spectrum activity, making them a preferred empirical choice in many clinical settings. However, the increasing resistance to 3GC antibiotics could compromise a large proportion of current treatment strategies, leaving clinicians with fewer effective options. This alarming scenario aligns with previous global mortality estimates, which highlight SSA as the region with the highest per capita deaths attributable to AMR11,42.

Conclusions

In conclusion, without urgent and coordinated multisectoral action, the rising prevalence of cephalosporin-resistant pathogens in SSA risks undermining current treatment strategies, leading to increasingly poor clinical outcomes, higher morbidity, and preventable mortality. Addressing this challenge requires strengthening AMR surveillance, promoting robust antimicrobial stewardship programs, and ensuring equitable access to affordable diagnostics and second-line therapies.

Moreover, a concerted global effort involving international health organizations, major donors, and academic partnerships is critical. These stakeholders must prioritize AMR mitigation by investing in surveillance, healthcare infrastructure, and targeted training programs. Without a unified, multisectoral approach, we risk a public health crisis where once-effective treatments fail, rendering infections increasingly untreatable, especially in SSA, and wider spread of drug-resistant infections. Urgent, collaborative action is needed to prevent further escalation and protect the future of infectious disease management in SSA and beyond.