Seroprevalence of seven arboviruses of public health importance in sub-Saharan Africa: a systematic review and meta-analysis

Abstract

Background

The arboviruses continue to be a threat to public health and socioeconomic development in sub-Saharan Africa (SSA). Seroprevalence surveys can be used as a population surveillance strategy for arboviruses in the absence of treatment and vaccines for most arboviruses, guiding the public health interventions. The objective of this study was to analyse the seroprevalence of arboviruses in SSA through a systematic review and meta-analysis.

Methods

We searched PubMed/MEDLINE, Web of Science, Embase, Scopus and ScienceDirect databases for articles published between 2000 and 2022 reporting the seroprevalence of immunoglobulin G (IgG) antibodies to seven arboviruses in various human populations residing in SSA. The included studies were assessed using the checklist for assessing the risk of bias in prevalence studies, and the data were extracted using a standard form. A random effects model was used to estimate pooled seroprevalences. The potential sources of heterogeneity were explored through subgroup analyses and meta-regression. The protocol had been previously registered on International Prospective Register of Systematic Reviews with the identifier: CRD42022377946.

Results

A total of 165 studies from 27 countries, comprising 186 332 participants, were included. Of these, 141 were low-risk and 24 were moderate-risk. The pooled IgG seroprevalence was 23.7% (17.9–30.0%) for Chikungunya virus, 22.7% (17.5–28.4%) for dengue virus, 22.6% (14.1–32.5%) for West Nile virus, 16.4% (7.1–28.5%) for yellow fever virus, 13.1% (6.4–21.7%) for Zika virus, 9.2% (6.5–12.3%) for Rift Valley fever virus and 6.0% (3.1–9.7) for Crimean–Congo haemorrhagic fever virus. Subgroup and meta-regression analyses showed that seroprevalence differed considerably between countries, study populations, specific age categories, sample sizes and laboratory methods.

Conclusion

This SRMA provides information on the significant circulation of various arboviruses in SSA, which is essential for the adoption and planning of vaccines. These findings suggest the need to invest in surveillance and research activities on arbovirus in SSA countries to increase our understanding of their epidemiology to prevent and respond to future epidemics.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Arboviruses represent a global public health concern, but the burden of these infections remains underestimated in sub-Saharan Africa (SSA) due to the limited capacity of countries to conduct effective and sustainable surveillance of arboviruses and vectors.

WHAT THIS STUDY ADDS

• The pooled immunoglobulin G (IgG) seroprevalence of arboviruses in SSA was high for dengue virus, Chikungunya virus, West Nile virus and yellow fever virus, but low for Zika virus, Rift Valley fever virus and Crimean–Congo haemorrhagic fever virus.

• The pooled IgG seroprevalences of arboviruses were higher in blood donors, patients presenting with fever and non-febrile persons in the general population.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• These results are essential for the adoption, planning and deployment of vaccines available against arbovirus in SSA.

• These results will contribute to strengthen advocacy efforts with political and global public health decision-makers for sustained investment in research and epidemic preparedness capacities in SSA.

Introduction

Arthropod-borne viruses pose a significant threat to global public health because of their capacity to cause epidemics and potentially fatal diseases.¹ Dengue is the most widespread arbovirus and causes the highest number of arboviral disease cases. About half of the world’s population is now at risk of dengue with an estimated 100–400 million infections occurring each year.² On 31 March 2022, the WHO launched the Global Arbovirus Initiative with the aim of raising global awareness about the potential risks of arbovirus epidemics and pandemics.^{3 4}

Numerous factors have contributed in the emergence and geographical expansion of arboviruses and their vectors. Environmental factors, such as climate change characterised by rising temperatures, have created a more favourable environment for the proliferation and spread of arboviruses and their vectors. Additionally, habitat modification, including deforestation, has disrupted the natural balance, leading to increased contact between humans and arbovirus-carrying vectors. Social factors, such as enhanced international mobility, have facilitated the introduction and spread of arboviruses across different regions. Furthermore, community-level factors, like increased population density, inadequate water management and poor housing conditions, have contributed to the global spread of competent vectors and arboviruses.⁵,⁸

In Africa, the burden of arbovirus infections remains unclear due to various factors. The similarity of the common clinical symptoms of arbovirus infections with those of malaria and other tropical febrile illnesses makes it challenging to accurately diagnose and distinguish these diseases. Moreover, the limited capacity of laboratories and the low level of knowledge among healthcare workers about arboviruses to hinder the prompt detection and confirmation of arbovirus infections hampers the timely reporting of cases and the implementation of preventive measures. Insufficient surveillance and limited case reporting further hinder the accurate estimation of arbovirus infections.⁹ Consequently, these infections are often underestimated, leading to a gradual increase in the risk of epidemics. Recently, several epidemics linked to dengue, Rift Valley fever, chikungunya, Zika, Crimean–Congo haemorrhagic fever and yellow fever have been reported in sub-Saharan Africa (SSA).⁹,¹¹ In this context, seroprevalence investigations are crucial as an adjunct to conventional symptom-based and laboratory-based surveillance to monitor the circulation of arboviruses and anticipate the risk of epidemics.^{12 13}

While seroprevalence studies have been conducted in multiple SSA countries, they are generally limited in terms of geographical scope; thus, local seroprevalence estimates may not accurately reflect national prevalence. Moreover, factors such as temperature, humidity and enzootic interactions, which exhibit regional variations, can also influence the density of vectors and arbovirus infections.¹⁴ Analysing changes in seroprevalence data is essential for the development, optimisation and prioritisation of effective and efficient surveillance and control programmes in the respective regions, as well as the deployment of vaccines. Additionally, public health authorities can gain valuable insights to enhance their emergency response systems and effectively manage potential epidemics associated with arboviruses.^{15 16}

Previous systematic reviews and meta-analyses (SRMA) on the seroprevalence of arboviruses have often focused on specific population groups (such as blood donors, children, pregnant women, livestock and farm workers) or specific arboviruses like dengue (DENV), chikungunya (CHIKV) and zika (ZIKV).^{114 17},²⁵ Limited SRMAs have estimated the seroprevalence of yellow fever virus (YFV)²⁶ and Rift Valley fever virus (RVFV) in SSA.^{15 27} Notably, no study has evaluated the overall prevalence of Congo haemorrhagic fever virus (CCHFV) and West Nile virus (WNV) in human populations in SSA. Given the growing threat of these arboviruses and the substantial increase in relevant literature on the subject, it is crucial to obtain up-to-date data on the seroprevalence of seven arboviruses (DENV, YFV, CHIKV, ZIKV, RVFV, CCHFV and WNV) of public health significance in SSA. The aim of this SRMA was to estimate the overall seroprevalence of these arboviruses in SSA.

Methods

Search strategy and selection criteria

In this systematic review and meta-analysis, we assessed studies conducted in SSA and tested IgG antibodies against arboviruses as defined above. The eligible studies were seroprevalence studies focusing on different human populations, such as symptomatic patients, non-febrile persons in the general population, children, adults and specific subgroups like pregnant women, blood donors and individuals belonging to certain professional categories (livestock and farm workers). However, case reports, conference abstracts and review articles were excluded from this review as they offer limited information on the studies’ characteristics and do not permit a critical analysis of their quality. Similarly, studies involving non-human subjects, as well as SRMA, were not considered. The PubMed/MEDLINE scientific databases, Web of Science, Embase, Scopus and ScienceDirect were used to search studies conducted in SSA countries and published between 2000 and 2022 in English and French. The search equations were developed using Medical Subject Headings (MeSH) and free terms using Boolean operators: [‘seroprevalence’ OR ‘seroepidemiology’ OR ‘seroepidemiologic studies’ OR ‘serosurvey’ OR ‘risk factors’ OR ‘associated factors’] AND [‘arbovirus’ OR ‘arboviral’ OR ‘arboviruses’ OR ‘arthropod-borne virus’ OR ‘dengue’ OR ‘chikungunya’ OR ‘yellow fever’ OR ‘zika’ OR ‘West Nile virus’ OR ‘Rift Valley fever virus’ OR ‘Crimean-Congo haemorrhagic fever’ OR ‘mosquito-borne disease’] AND [‘Sub-saharan Africa’ OR ‘Africa’]. The search was further complemented by a review of non-indexed African databases and the reference list of selected studies. The studies obtained from various databases were imported into the Zotero reference manager, where duplicates were removed. Screening of titles, abstracts and full text was done by STB and reviewed by NK and AT independently. This review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines for the identification of studies.²⁸

Quality assessment

The methodological quality of each selected study was evaluated by two teams of two reviewers (LK, AM, SS and CGH) independently using the risk-of-bias checklist for prevalence studies adapted from Hoy et al.²⁹ The checklist consists of 10 items divided into two categories. The first four items pertain to the external validity of the study, focusing on selection bias and non-response bias. The remaining six items assess internal validity, with items 5 to 9 specifically addressing measurement bias and the final item examining analysis bias. Each item was assigned a risk-of-bias score, with a score of 0 indicating low bias and a score of 1 indicating high bias. The overall risk of bias for each study was then determined by considering the total score, with a score between 0 and 3 indicating low bias, a score between 4 and 6 indicating moderate bias and a score between 7 and 10 indicating high bias (online supplemental table 1). Any disagreements between reviewers were resolved through discussion or by involving a third reviewer as an arbitration.

Data extraction

A standard form was used to extract and enter the data into an Excel spreadsheet. The data included the name of the first author, the year of publication and the country where the study was conducted; the specific region within Africa as defined by the WHO African, the type of arbovirus, the setting (hospital vs community), the context (epidemic or non-epidemic period), the design, the included population, the methodology of sampling, the age group included (either adults if participants were 15 years or older, children if they were younger than 15 or both), the laboratory technique used for serological analysis, the total number of subjects tested (for analytical purposes, the sample size was categorised into four groups: <200; 200–499; 500 to 999; >1000) and the number of subjects who tested positive for IgG antibodies (online supplemental table S2). When a study reported results for two or more arboviruses, different study populations or different countries, the data were collected and recorded separately in the database.

Data analysis

The data were analysed using the ‘meta’ and ‘metafor’ packages of the R statistical software (version 4.3.1, The R Foundation for statistics computing, Vienna, Austria). Descriptive statistics and a narrative summary were generated to provide an overview of the selected studies’ characteristics. The random effects model was used to estimate grouped seroprevalences along with their 95% CIs for each arbovirus studied. These results were presented using forest plots. To stabilise the variance of the proportions,³⁰ a double arc-sine transformation was applied. Furthermore, the restricted maximum likelihood (REML) estimation method was used to estimate the Tau. The heterogeneity between studies was evaluated using the I² statistic and the Cochrane Q statistic test. Heterogeneity exceeding 50% was considered to indicate a high level of variation among studies. Sub-group analyses and meta-regression were conducted to explore potential sources of heterogeneity. The robustness of the findings was assessed through Jackknife sensitivity analysis, which involved individually omitting each study to determine its influence on the overall prevalence. A study that was deemed to have substantial significant impact if its removal resulted in a substantially different overall effect. Small-studies effects and publication bias were evaluated using the funnel plot, and its significance was evaluated using the Egger statistical test and the Trim-and-Fill methods. A p value <0.05 was considered statistically significant for pooled estimates of seroprevalence for any arboviruses and its subtypes, and p value <0.01 for subgroup interactions to compensate the effects of multiples testing. The protocol of this study had previously registered on International prospective register of systematic reviews with the identifier: CRD42022377946.

Patient and public involvement

Patients and the public were not specifically involved in the design, conduct, reporting or dissemination plans of our research.

Results

Selection of studies

Figure 1 shows the PRISMA diagram describing the study selection process. A total of 1086 studies were identified from electronic databases, with 476 studies in PubMed, 295 in Scopus, 206 in Web of Science, 77 in ScienceDirect and 32 in Embase. Additionally, 153 studies were identified from African databases (African Journal Online) and the bibliographic references of the selected studies. After removing duplicates, 490 studies were excluded: 331 based on titles and abstracts, 131 studies conducted exclusively on non-human subjects, 19 systematic reviews/meta-analyses and 9 abstracts of papers for which the full text was not accessible, even after contacting the corresponding author via email. The full-text review of the remaining studies resulted in the inclusion of 165 studies in this SRMA (figure 1) (online supplemental text S1).

Figure 1. Flowchart illustrating the study selection process.

Characteristics of studies

The selected studies were conducted in 27 SSA countries, of which 88 studies were carried out in East Africa (Comoros, Djibouti, Ethiopia, Kenya, Madagascar, Rwanda, Sudan, Tanzania, Uganda), 44 studies in West Africa (Benin, Burkina Faso, Ghana, Guinea, Mali, Niger, Nigeria, Senegal, Sierra Leone), 20 studies in Central Africa (Cameroon, Central African Republic, Democratic Republic of Congo, Gabon, Republic of Congo) and 13 studies in Southern African countries (Mozambique, Namibia, South Africa, Zambia). Included studies were published between 2002 and 2022, with the majority published between 2016 and 2022 (no. of studies=109) (figure 2A). The DENV was the most reported arbovirus in the included studies (no. of studies=76), followed by CHIKV (no. of studies=56), RVFV (no. of studies=47), WNV (no. of studies=27), ZIKV (no. of studies=25), CCHFV (no. of studies=17) and YFV (no. of studies=16) (figure 2B). Most studies were carried out in non-febrile persons in the general population (no. of studies=117), followed by studies in patients with fever (no. of studies=99), blood donors (no. of studies=22), pregnant women (no. of studies=16) and livestock/farm workers (no. of studies=10) (figure 2C).

Figure 2. Description of variables. (A) Distribution of studies included by year (2000–2022); (B) Distribution of studies included by arbovirus; (C) Distribution of studies included by study population.

Out of the studies included, 87 were conducted exclusively in the community and 78 in hospitals. The majority of these studies were cross-sectional in design (no. of studies=150). There were also 12 retrospective and three descriptive cohort studies. The age groups targeted varied across studies. More than two-thirds of the studies (no. of studies=115) included participants of all age groups (children and adults). However, there were 39 studies and 11 studies included exclusively adults and children (under 15 years of age), respectively. The most commonly used sampling method was exhaustive recruitment (no. of studies=102), followed by random sampling (no. of studies=58) and voluntary sampling (no. of studies=3). Various laboratory techniques were used to test for arbovirus seropositivity in the studies, including enzyme-linked immunosorbent assay (ELISA) (no. of studies=142), rapid diagnostic tests (no. of studies=13), indirect immunofluorescence tests (no. of studies=7) and Luminex MAGPIX (no. of studies=3). Out of the 165 studies included, 141 were assessed to have a low risk of bias, and 24 were assessed as having a moderate risk of bias. None of the studies were found to have a high risk of bias (table 1). Of the 186 332 participants included in this SRMA, 42 186 (no. of studies=106) were males and 45 404 were females (no. of studies=122), including 7966 pregnant women. The mean age of the participants was 29.5 years (no. of studies=64).

Table 1. Characteristics of the included studies.

Characteristics	No. of studies=165 (%)
Year of publication
2000–2010	16 (9.7)
2011–2015	40 (24.2)
2016–2022	109 (66.1)
African region
Central Africa	20 (12.1%)
East Africa	88 (53.3%)
South Africa	13 (7.9%)
West Africa	44 (26.7%)
Study setting
Community	87 (52.7%)
Hospital	78 (47.3%)
Study type
Cross-sectional	150 (90.9%)
Descriptive cohort	3 (1.8%)
Retrospective	12 (7.3%)
Age category
All age groups	115 (69.7%)
Adult	39 (23.6%)
Child	11 (6.7%)
Sampling methods
Exhaustive	102 (61.8%)
Random	58 (35.2%)
Volunteer	5 (3.0%)
Laboratory method
ELISA	142 (86.1%)
Indirect immunofluorescence test	7 (4.2%)
Luminex MAGPIX	3 (1.8%)
Rapid test	13 (7.9%)
Bias risk
Low	141 (85.5%)
Moderate	24 (14.5%)

ELISA, enzyme-linked immunosorbent assay.

Pooled IgG seroprevalence of arboviruses

According to arbovirus type, the pooled IgG seroprevalence was 23.7% (17.9%–30.0%; no. of subjects=31 377; cases=7185; no. of studies=56) for CHIKV, 22.7% (17.5%–28.4%; no. of subjects=59 987; cases=14 243; no. of studies=76) for DENV, 22.6% (14.1%–32.5%; no. of subjects=16 899; cases=2646; no. of studies=27) for WNV, 16.4% (7.1%–28.5%; no. of subjects=11 282; cases=1994; no. of studies=16) for YFV, 13.1% (6.4%–21.7%; no. of subjects=12 870; cases=1845; no. of studies=25) for ZIKV, 9.2% (6.5%–12.3%; no. of subjects=40 318; cases=2887; no. of studies=47) for RVFV and 6.0% (3.1%–9.7%; no. of subjects=13 599; cases=859; no. of studies=17) for CCHFV (figure 3; online supplemental figures S1–S7).

Figure 3. Forest plot showing the pooled IgG seroprevalence for arboviruses.

Subgroup analysis and heterogeneity by arbovirus type

Chikungunya virus

The pooled seroprevalence of anti-CHIKV IgG varied not only between countries but also within the same country across different studies. The highest seroprevalence was reported in Rwanda (63.0%, in 2015 among blood donors), followed by Gabon (61.2%; in 2021 among non-febrile persons in the general population), Nigeria (44.9%, data from five studies) and Cameroon (42.4%; data from four studies). In contrast, Djibouti had the lowest seroprevalence at 0.77%, in a study of the general population (figure 4A, online supplemental table S3). Sub-group analyses showed that Central, Southern and West Africa had higher seroprevalence rates (37.8%, 33.5% and 31.2% respectively) compared with East Africa (16.5%). Furthermore, the seroprevalence was significantly higher during the period from 2000 to 2010 (54.3%) compared with the periods from 2011 to 2015 (14.4%) and 2016 to 2022 (24.2%). When considering different populations, pregnant women had the highest seroprevalence at 36.0%, followed by blood donors at 35.3%, non-febrile persons in the general population at 30.3%, and livestock and farm workers at 2.7% (online supplemental table 1). Interestingly, the seroprevalence decreased as the sample size increased, ranging from 35.4% in studies with fewer than 200 people to 14.1% in studies with 1000 or more participants (online supplemental figure S8). The multivariate meta-regression analysis indicated that only 37.5% of the variation in seroprevalence between studies could be explained by factors such as the publication period, study population, age category sample size, laboratory method and the risk of bias. This analysis yielded a significant test of moderators (p<0.0001).

Figure 4. Geographical distribution of pooled immunoglobulin G seroprevalence of different arboviruses in Sub-Saharan. Africa.

Dengue virus

The distribution of pooled dengue seroprevalence can be seen in figure 4B. Studies conducted in Comoros, Sierra Leone, Mali, Ghana and Ethiopia reported the highest seroprevalences, ranging from 42% to 75%. On the other hand, Uganda, Madagascar, Tanzania, Namibia, Gabon and Zambia reported lower IgG seroprevalence rates of less than 10% (figure 4B, online supplemental table S4). Furthermore, the seroprevalence of dengue varied depending on the population groups studied. It was 30% among blood donors, while the seroprevalence rates among non-febrile persons in the general population, patients with fever and pregnant women were 22% each. The subgroup analyses showed that the pooled seroprevalence of dengue fever was higher in the studies conducted in the West Africa region, reaching 34.5%. In contrast, the studies conducted in the East and Central Africa region exhibited a seroprevalence of 18%, while the South Africa region reported a seroprevalence of 12.5% (p<0.0237). Likewise, studies involving a population of fewer than 200 individuals showed a higher seroprevalence of dengue was 43.4%, compared with studies involving 200–499 (18%), 500–999 people (18%) and 1000 people or more (19.8%) (p<0.0017). Moreover, the seroprevalence of dengue varied depending on the laboratory technique employed, with studies using the ELISA technique reporting a seroprevalence of 24.1% compared with studies using other techniques (rapid tests, indirect immunofluorescence test and Luminex), which reported a seroprevalence of 17.7%. However, this difference was not statistically significant (p<0.34). This meta-analysis of dengue revealed significant heterogeneity among the studies (I²=99.6%, p<0.001). The multivariate meta-regression model, including country, sample size, study population, laboratory technique and risk of bias, showed that this heterogeneity was significantly associated with the countries in which the studies were conducted (online supplemental table 1).

West Nile virus

This seroprevalence varied across countries, ranging from 0.22% in Djibouti to 70.3% in Sudan. Nigeria (60.5%), Sierra Leone (53.7%), Ethiopia (40.6%), Cameroon (34.2%) and Gabon (26.3%) also exhibited high seroprevalences (figure 4C, online supplemental table S5). The subgroup meta-analysis revealed higher seroprevalence in West Africa (37.9%) and lower seroprevalence in Southern Africa (10.3%) (online supplemental table 1). Over time, pooled seroprevalence increased from 19.6% to 27.5% during the publication periods of 2000–2010 and 2016–2022, respectively (online supplemental figure S9). Among the different populations studied, patients with fever had the higher pooled seroprevalence (33.2%), followed by pregnant women (24.7%), non-febrile persons in the general population (18.1%) and blood donors (14.4%). Studies exclusively focusing on children (<15 years) had a lower seroprevalence (6.1%) compared with those including adults (16.5%) or all age groups (33.3%) (online supplemental table 1). The meta-regression model including country, study population, sample size, diagnostic method and study quality explained 73.6% of the observed heterogeneity (test of moderators: p<0.0001).

Yellow fever virus

The highest seroprevalences were observed in Gabon (60.7%), Sudan (48.4%) and Cameroon (43.0%). Conversely, Djibouti, Comoros, Tanzania and Uganda had the lowest seroprevalence rates at 0.4%, 0.5%, 0.8% and 3.3%, respectively (figure 4D, online supplemental table S6). Analysis of subgroups showed a decreasing trend in the seroprevalence of anti-YFV IgG over time. It was higher (45.7%) in the period from 2000 to 2010, compared with 10.6% in the period from 2011 to 2015, and 14.3% in the period from 2016 to 2022. The seroprevalence of anti-YFV IgG antibodies varied across different populations, ranging from 3.3% in blood donors to 42% in pregnant women, based on a study in Kenya in 2011. While most of the studies were carried out in non-febrile person in the general population, the seroprevalence of anti-YFV IgG antibodies in this population was 15.5% compared with 19.6% in febrile patients (online supplemental table 1). The heterogeneity of seroprevalence was influenced by the country where the studies were conducted, the sample size and the age category (R²= 96.54%; test of significant moderators with p<0.0001).

Zika virus

The highest seroprevalence were observed in Sudan (62.7%), Burkina Faso (45.7%), Ethiopia (42.2%) and Gabon (40.3%), while the lowest were found in Madagascar (0.9%), Rwanda (1.4%), Kenya (1.7%) and Ghana (1.9%) (figure 4E, online supplemental table S7). When considering the different regions of Africa, the seroprevalence of anti-ZIKV IgG was 18.0% in East Africa, 12.4% in West Africa and 10.7% in South Africa. A study conducted during the period 2000–2010 reported a much higher seroprevalence (37.9%) compared with the pooled seroprevalence observed during the period 2016–2022 (12.3%). The seroprevalence varied across different population, ranging from 18.2% in patients with fever to 16.8% in non-febrile persons in the general population, 12.0% in blood donors and 2.0% in pregnant women. Studies using diagnostic methods other than ELISA, such as rapid diagnostic tests and Luminex MAGPIX, reported a higher (25.3% compared with 11.7%). Furthermore, studies with a sample size of fewer than 200 people had a higher seroprevalence (29.4%), while studies with a sample size of 1000 or more persons had a lower seroprevalence (5.8%) (online supplemental table 1). There was significant heterogeneity between studies I (I² = 99%; p<0.0001), with more than half of this heterogeneity (R² = 55.5%) being attributed to factors such as the country of study, study population, sample size, laboratory method used, and study quality. The test associated with the significance of the covariates showed a significant result with p<0.0001.

RVFV

The highest seroprevalence was observed in Sudan (81.9%), while the lowest rate was recorded in Djibouti (0.9%). In countries where of the majority of the studies were conducted, such as Kenya (no. of studies=17) and Tanzania (no. of studies=10), the seroprevalence rates were 9.6% and 6.7%, respectively (figure 4F, online supplemental table S8). It was also observed that the seroprevalence rates decreased over the years, dropping from 16.3% in studies published between 2000 and 2010 to 7.7% in studies published between 2016 and 2022. Furthermore, high seroprevalence rates were observed in specific populations, such as patients with fever (16.8%) and livestock and farm workers (9.9%) (online supplemental table 1). The analysis indicated also that the seroprevalence decreased as the sample size increased (online supplemental figure S10), with studies involving a sample size of less than 200 people showing higher seroprevalence rates compared with studies with a sample size of 1000 or more (online supplemental table 1). Additionally, the meta-regression analysis revealed that approximately 44.5% of observed heterogeneity between studies could be attributed to factors such as country, age category and sample size. These covariates were found to be statistically significant with a p<0.0001.

Crimean-Congo haemorrhagic fever virus

The highest seroprevalence rates were observed in Kenya (13.1%) and Uganda (12.5%), while the lowest was found in Madagascar (0.7%) among slaughterhouse workers (figure 4G, online supplemental table S9).

Based on the study populations, the pooled seroprevalence of IgG antibodies was 6.9% in patients with fever, 6.8% in non-febrile persons in the general population, 6.0% in blood donors and 2.9% in livestock and farm workers (online supplemental table 1). The analysis showed a high level of heterogeneity in the pooled seroprevalence (I² = 98%, p<0.0001), which could not be explained by the meta-regression. The test of moderators did not show a significant effect of study characteristics.

Sensitivity analysis and publication bias

In order to evaluate the reliability of the pooled IgG seroprevalence, a sensitivity analysis was conducted for each arbovirus by systematically removing each study one by one. The results showed that none of the studies had a significant impact on the estimates of overall seroprevalence (online supplemental figures S11–S24). Quantitative assessment (Egger’s test, p<0.05) and visual inspection of funnel plots revealed evidence of publication bias for all arboviruses (online supplemental figure S25).

Discussion

Although the arbovirus is increasingly widespread in SSA, particularly in tropical and subtropical regions, there is still a lack of clarity regarding its extent and distribution due to limited surveillance and inadequate diagnostic tests.^{31 32} This lack of knowledge about the level of circulation and spatial extent of these viruses can be overcome by determining the seroprevalence of arbovirus on a continental scale, which will enable the identification of the risk of future epidemics and the appropriate areas for the deployment of control measures.³³ To address this issue, an extensive and up-to-date meta-analysis was conducted in different human populations.

The pooled IgG seroprevalence of arboviruses in SSA was high for DENV, CHIKV, WNV and YFV, but low for ZIKV, RVFV and CCHFV. While these arboviruses are endemic in SSA, these estimates provide strong evidence that they are actively transmitted, although they are probably underestimated. Our estimates differ from those reported in previous meta-analyses conducted in Africa, which indicated pooled seroprevalences of 25% for DENV IgG,³⁴ 33% for CHIKV²¹ and 19% for YFV.²⁶ However, low grouped IgG seroprevalences have been reported in other meta-analyses for DENV (13%),²¹ CHIKV (16.4%),²⁵ ZIKV (4%)²¹ and RVFV (7.8%; 5,9 %).^{15 27} These variations in seroprevalence can potentially be attributed to the differing number of studies included in the meta-analysis. Our meta-analysis, in comparison with previous meta-analyses, included a greater number of studies, thereby precision of the accuracy of the observed estimation.

The seroprevalence of arbovirus varied significantly across different countries, suggesting that various factors contribute to this variability. These factors include differences in study period, heterogeneity of study populations, sample size and study design. In addition, the use of different diagnostic methods to estimate arbovirus seroprevalence can result in a wide range of variations. Most studies have used ELISA as the diagnostic method, which can yield inconsistent results depending on the antibody detection kits, positivity threshold and reference samples used. To address this issue, international standardisation of reagents and protocols is necessary to ensure quality control and facilitate result comparison between laboratories.³⁵ Additionally, other factors, such as ecological environment, climate, vector abundance and diagnostic and surveillance capabilities of arboviruses, may significantly contribute to the observed variability. Furthermore, the disparity in the distribution of seroprevalence studies on arbovirus between countries could also explain the variability.

Most studies have been conducted in East African countries, where some of these arboviruses, such as CHIKV, RVFV, ZIKV and WNV, were initially isolated.³⁶ Moreover, these results reveal the lack of serological evidence of arbovirus infection in certain countries. Several factors could explain this observation, including challenges in conducting epidemiological surveillance, lack of funding and/or research interest in arboviruses.³⁷ Moreover, except for yellow fever, for which vaccination is required for international travellers in many African countries, arboviruses are insufficiently taken into account by the epidemiological surveillance system. Awareness of the risk of epidemics linked to these viruses is low, both among the authorities and among healthcare workers in health facilities. Hence, there is an urgent need for more seroprevalence studies in these countries to understand the real burden of arbovirus in different populations.

The pooled IgG seroprevalences of arboviruses in different populations were found to vary. Elevated seroprevalences were observed in blood donors, patients with fever and non-febrile persons in the general population, and high seroprevalences were observed. A global meta-analysis of blood donors conducted by Giménez-Richarte et al reported overall IgG seroprevalences of 22.0% for DENV, 37.8% for CHIKV and 4.0% for ZIKV in Africa.¹⁹ The transmission of arboviruses through blood transfusion is a reality and presents a significant challenge for the blood supply system.³⁸ Although the existence of IgG antibodies does not necessarily indicate contagious blood, there is still a considerable risk of arbovirus transmission during blood transfusions due to short viremia periods and asymptomatic donors.³⁹

Asymptomatic patients have consistently posed challenges in controlling and preventing arboviruses, as they can disseminate the virus without being under surveillance.²² Several studies have reported cases of arbovirus transmission through blood transfusion, including DENV,⁴⁰ WNV,⁴¹ YFV⁴² and ZIKV.^{43 44} Therefore, it is crucial to carry out in-depth studies on arboviruses in blood donors to determine the necessary interventions for ensuring the safety of blood transfusions.²²

Other meta-analysis conducted in Africa reported pooled seroprevalence of anti-DENV IgG of 15.6% (CI 95%: 9.9–22.2) in non-febrile persons and 24.8% (CI 95%: 13.8–37.8) in patients with fever.²⁴ Additionally, a meta-analysis of general population studies found a pooled seroprevalence of anti-CHIKV IgG of 31% (CI 95%: 21–41).⁴⁵ Moreover, our meta-analysis revealed a lower seroprevalence of DENV in children. A recent meta-analysis on DENV seroprevalence in children in SSA reported a pooled prevalence of 15.8% (CI 95%: 10.2–22.3).²⁰ This observed difference in seroprevalence may be primarily attributed to the age range used to define the paediatric population. Our review included studies that defined a child as an individual under 15 years of age, while the comparative meta-analysis included children aged between 1 month and 19 years. Second, the difference could be due to the combined seroprevalence data. Our meta-analysis combined only DENV IgG results regardless of the diagnostic method, whereas the previous meta-analysis combined IgG/IgM serology and PCR results.

The results of this meta-analysis support the findings of previous meta-analyses and provide evidence of the active circulation of arboviruses in SSA, but the specific epidemiological patterns and methods of this circulation still need to be documented. To the best of our knowledge, our meta-analysis represents the first attempt at estimating the pooled seroprevalence of CCHFV and WNV IgG in human populations in SSA. Previous meta-analysis on the seroprevalence of anti-CCHFV IgG have been conducted globally, but they have included very few or no studies conducted in SSA.^{46 47} In addition, this study estimated the pooled seroprevalence of these arboviruses in different populations, including non-febrile persons in the general population and specific population subgroups such as pregnant women, blood donors and livestock and farm workers. In comparison with previous studies, this meta-analysis included the largest number of studies published over a period of two decades, increasing the statistical power and providing a more accurate grouped effect.

Furthermore, the findings also raise questions about the capacity of countries to perform effective and sustainable surveillance to prevent, detect and respond to current and future epidemics. In recent years, there has been a considerable increase in the frequency of epidemics caused by arboviruses in SSA countries, resulting in high mortality and morbidity rates.⁹⁴⁸,⁵⁰ The information provided from this meta-analysis could therefore help public health authorities strategically manage the limited resources available for the control and prevention of arboviruses, such as vaccines. Additionally, the transmission of these viruses primarily depends, with few exceptions, on the presence of competent mosquito vectors,⁵¹ with Africa known to have a high density and diversity of such vectors.⁵²,⁵⁹ In the absence of available vaccines in Africa (except for yellow fever), efforts to prevent the global spread of these diseases in the near future will be more effective if they focus on vector control and surveillance.⁶⁰ Regional and sub-regional collaboration could help in establishing protocols for surveillance, laboratory diagnosis and epidemic preparedness, as isolated measures will not yield the same results and could even be ineffective.^{61 62} Furthermore, future research should adopt a holistic and comprehensive approach, considering all aspects of arbovirus transmission, including vectors, reservoirs and hosts.³⁷

However, this meta-analysis study has several limitations. First, it only included peer-reviewed literature from selected databases, which means that non-published data may not have been included in the analysis. Second, there was significant heterogeneity between the studies. In an attempt to address this limitation, a subgroup analysis and meta-regression were conducted. However, most of the data subgroups still exhibited significant heterogeneity that could not be explained by the variables examined in the meta-regression models. This suggests that there are other characteristics of the studies that could have influenced this heterogeneity, but we were unable to investigate them in this meta-analysis. Additionally, it is worth noting that meta-analyses of proportions generally have a higher level of heterogeneity compared with comparative studies, and it is not uncommon for the I² statistic to exceed 90% in these types of analyses. The I² statistic represents the percentage of variability not attributed to sampling error, but its value is highly dependent on the precision of the included studies. As the number of studies increases and the sampling error decreases, the I² statistic tends to reach 100% due to the larger sample size. Therefore, relying solely on the I² statistic for the reliability of a meta-analysis of proportions is not advisable.^{63 64} Third, the asymmetry of the funnel plots suggests the existence of a publication bias. However, it is important to interpret these results with caution as traditional methods for assessing publication bias may struggle to identify it in meta-analyses of proportions, since bias in non-comparative studies can arise for reasons unrelated to statistical significance.⁶⁵ Consequently, it is unlikely that significance levels will influence publication decisions for these types of studies. Authors who report both low and high proportions are equally likely to have their work published, provided that the quality of the study adheres to rigorous publication standards.⁶⁵,⁶⁷ Lastly, it should be noted that serological cross-reactions between antibody-based tests for flaviviruses can lead to overestimation or underestimation of seroprevalence in the absence of confirmatory tests, which are often difficult to perform or unavailable. However, given the critical role of serology in the diagnosis and surveillance of arbovirus infections, it is imperative to establish standards for the development of more specific serological tests that can differentiate between exposures to different arboviruses.⁶⁸

Conclusion

This meta-analysis has drawn attention to the active circulation of several arboviruses in SSA, as well as the level of exposure of populations. The results of our study have indicated a variability in the seroprevalence of the arboviruses examined across different studies and countries, with a tendency towards higher prevalence in blood donors, patients with fever and non-febrile persons in the general population. In addition, these findings provide information for effective public health policy formulation and can guide the prevention strategies. While much of the focus and investment in vector-borne diseases is currently directed towards malaria and tropical neglected diseases, the results of this review should reignite discussions regarding the necessity of investing in surveillance and research activities on arboviruses in SSA in order to prevent future public health emergencies. Additionally, the implementation of measures to control the arbovirus vectors, combined with the enhancement of laboratory and healthcare worker capabilities, as well as community engagement, holds potential to prevent epidemic outbreaks. Further research on these arboviruses in SSA countries could prove valuable in enhancing our understanding of their epidemiology.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.