Hepatitis C seroprevalence and HIV co-infection in sub-Saharan Africa: a systematic review and meta-analysis

**Bhargavi V Rao1,2, MRCP, Nur Johari2, BSc, Phillipdu Cros1, MRCP, JaneyMessina4, PhD, NathanFord2,3*, PhD Graham S Cooke2*PhD

1 Manson Unit, Médecins Sans Frontières (MSF), London, UK

2 Division of Infectious Diseases, Imperial College London, UK

3 Médecins Sans Frontières (MSF), Geneva, Switzerland

4 Spatial Epidemiology and Ecology Group, Department of Zoology, University of Oxford, UK

*contributed equally to this work

Total word count: 2998

**Corresponding authors: nd

Address: Manson Unit, MSF-UK, 67-74 Saffron Hill, London EC1N 8QX, UK

Tel: +44 (0)2074046600

Fax: +44 (0)2074044466

Oxford

Abstract

Background

An estimated 150 million people worldwide have been infected with hepatitis C virus (HCV). HIV co-infection accelerates the progression of HCV andrepresents a major public health challenge. In a systematic review and meta-analysis, we aimed to determine the epidemiologyof HCV and the prevalence of co-infection with HIV in sub-Saharan Africa.

Methods

We included HCV seroprevalence datafrom2002to end of 2014 relating to the WHO defined regions of sub-Saharan Africa.We estimated pooled regional prevalence estimates using a DerSimonian-Laird random effects model. Data were further stratified by risk factor and HIV status.

Findings

From 213 studies, we identified 287 separate cohorts with a total of 1,198,167 individuals. The mean pooled HCV seroprevalence from all cohorts was 3% (95%CI: 2.9-3.1).The pooled HCV seroprevalence was 2.7% (95%CI: 2.5-2.8) across all 185 low risk cohorts; 3% (95%CI: 2.2-3.8) in antenatal groups,2% (95%CI: 1.9-2.1) among blood donors but 6.9% (95%CI: 6.1-7.5) in other general population cohorts. The pooled seroprevalence of HCV across allhigh-risk groups was 11.9% (95%CI: 7.1-16.7) and 10% (95%CI: 6.8-13.1) in patients with liver disease. 101cohorts included HIV-positive samples tested for HCV (42,648individuals) with pooled seroprevalence of 5.7% (95%CI: 4.9-6.6).

Interpretation

We found a high seroprevalence of HCV across populations of sub-Saharan Africa includingwithin HIV-positive adults, with evidence of regional variation in the general population. Monitoring antenatal HCV prevalence may be a helpful indicator of population trends in HCV infection. There are limitations to available data with a need for larger survey studies. Access to prevention and treatment needs to be improved for both mono-infected and co-infected individuals.

Funding

None.
Introduction

Hepatitis C virus (HCV) is estimated to have infected 130-150 million individuals worldwide1,2 and contributes significantly to the Global Burden of Disease3,4.There is no vaccine close to market for HCV,hence control of the epidemic rests on preventative measures - including the screening of blood products, opiate substitution, and needle exchange - and treatment to prevent progression to cirrhosis and hepatocellular carcinoma. Until recently, the complexity of HCV therapy was a barrier to widespread treatment; standard therapy involved injectable pegylated interferon and ribavirin tablets, in a long and complex regimen dependant on viral genotype and requiring frequent monitoring5,6. HCV management has been revolutionised by recently licensed directly acting antiviral agents (DAAs), including protease inhibitors, NS5B inhibitors (e.g. sofosbuvir7,8) and NS5A inhibitors (e.g. ledipasvir and daclatasvir), which allow shortened treatment with improved success rates without the use of interferon9,10. The new treatments have the potential to reduce substantially the clinical and operational barriers to treatment throughout the world, provided they are accessible and affordable6,11.

Accurate epidemiological information is essential to inform treatment and control priorities at national and regional levels. In particular, data are needed on the burden of co-infection with HIV given the poorer outcomes for HCV in this population12.

There is paucity of such data in many regions, particularly parts of sub-Saharan Africa, to some extentdue to the relative public health importance of HCV infection being only recently recognised3. A review of HCV epidemiology in sub-Saharan Africa published in 2002 estimated the overall prevalence of HCV to be 3%, with significant regional variation13. A further review of the association of HIV and hepatitis B and C showed a relative risk of HIV co-infection of 1.6 (95%CI: 1.05-2.45) 14. However, both analyses found little data on the prevalence of HIV-HCV co-infection, which is critical to the current policy debate surrounding expansion of HCV treatment.

In asystematic reviewand meta-analysis of studies published since 200213, we aimed to determine the seroprevalence of HCV and the prevalence of HCV/HIV co-infection across sub-Saharan Africa,addressing the availability and type of data included and categorising cohorts by risk.

Methods

Inclusion criteria

Countries included in the study were limited to countries in sub-Saharan Africa grouped according to WHO Africa regions. In order to directly compare results with earlier published analyses, six countries were excluded (Algeria, Cape Verde, Comoros, Mauritius, Sao Tome & Principe, and Seychelles). Three countries not listed in the WHO regional list – Sudan (and South Sudan after 2012), Djibouti, and Somalia were included in the study.

Studies published in English and French were included. Studies were notrejected on thebasis on sample size or design (including both retrospective and prospective studies), and included cross-sectional national surveys, as well as surveys from screening programmes, antenatal clinics (ANC), blood donations, hospitals, and other institutionssuch as prisons. Papers that did not state sample size or HCV seroprevalence or identify the HCV diagnostic assays used were excluded. No study was discountedon the basis of HCV diagnostic assay; this review includes studies using both non-confirmatory (ELISA, EIA, screening assays) and confirmatory (RIBA, Western blots, PCR) assays. The prevalence of detectable HCV RNA prevalence was included where available.

Search strategy

References were identified from Medline and EMBASE (via Ovid). We searched for all papers that may contain HCV seroprevalence data in different population groups, the search strategy is shown indetail in Supplementary Table 1 and includes terms to assess the availability of data on HIV/HCV co-infection. Papers were included if published between 1st January 2002 and 31st December 201413.

Data were extracted by one author (NJ) and verified in full by another (VBR) on the following: country, year of publication, year of collection; study type, study design, HCV assay used, study population, sample size, proportion male, age, HCV seroprevalence, PCR prevalence, genotype (including percentage of each genotype where available), subtypes, and HIV co-infection prevalence.

Cohort classification

Cohorts were separated by either time of collection (if this was defined by the paper) or by collection site i.e. ANC or Blood donor. We also separated out control cohorts from intervention cohorts in case control studies.If there was a separate HIV sub-analysis within a study, this was included as a separate cohort within the HIV pooled prevalence but not counted overall as a separate cohort. Cohort studies were initially categorised as either low risk or high risk using definitions comparable with previous work13. The low-risk category followed the definition from Madhava et al 13, namely according to the setting from which the samples were obtained and tested. The aggregated low risk group included (1) pregnant women at Antenatal Care (ANC), (2) blood donors and (3) other samples that were collected from thegeneral population - randomly selected communities whether rural or urban, students, and samples from inpatients or outpatients seeking care for non-hepatic illnesses or who had not had multiple blood transfusions . The high-risk group was divided into two groups: (1) patients suffering from known liver disease, whether acute or chronic; and (2) patients who had not been documented to have liver disease but who had high risk exposures such as multiple blood transfusions, haemodialysis, renal transplants, sickle-cell disease, orinjecting drug use. HIV-infected cohorts were grouped separately to assess the prevalence of HIV and HCV co-infection.

Data analysis

Point estimates and 95% confidence intervals (CI) were calculated for the proportion of people with HCV in each study. Prevalence of HCV by WHO region was estimated through pooling data from each study. Data were pooled using a DerSimonian-Laird random effects model 15which incorporates an estimate of between-study variance, allowing that the true effect size may vary from study to study.16 To assess between-study heterogeneity for the estimates of pooled prevalence by region, the 2 statistic was calculated. The variance of raw proportions was stabilised using a Freeman-Tukey type arcsine square-root transformation.17 Several methods of pooling proportions exist; the Freeman-Tukey method works well with both fixed-effects and random-effects meta-analysis.18,19Data were further stratified by risk group and HIV status. Analyses were conducted using STATA (v12,

Role of the funding source

There was no specific funding for this article.The corresponding author had full access to all the data and had final responsibility for the decision to submit for publication.

Results

Search results

213studies met eligibility criteria for inclusion identified from 33countries in sub-Saharan Africa (Supplementary Figure 1; for all papers and references seeSupplemental Table 6). These 213studies included data on287cohorts comprising1,198,167 individuals; 241cohorts had sample sizes greater than 100.

Most studies (177/213) were cross-sectional surveys; 15, were retrospective cohort studies and 19 were case-control studies.

Testing methodology

Of the 287 cohorts, 81(28%) used confirmatory assays to screen for the presence of anti-HCV antibodies and HCV RNA. Less than 20% of the cohorts (55of 287) described testing for HCV RNA (i.e. reported HCV prevalence based on PCR testing). 39 cohorts presented data on HCV genotyping.

Pooled seroprevalence

The estimated pooled HCV seroprevalence across all 287cohorts identified was 3% (95%CI: 2.9-3.1;2 0.06), varying widely between 7.8% (70cohorts; 95%CI: 6.6-9.0; 2 0.05) in the Central region to 4.5% (142cohorts; 95%CI: 4.1-4.9; 2 0.05) in West Africa and 0.8% (75 cohorts; 95%CI: 0.6-0.8; 2 0.02) in Southern and Eastern (SE) Africa. 142/287 (49%) cohorts were collected in the West Africa region, but the majority of individuals included were from SE Africa skewed by a single survey from the South African Blood Service (732,250 samples)20.

Low-risk groups

185 cohorts studies (including 1,151,337individuals) were identified as low-risk populations (Supplementary information Table 2) representing 30/33 countries. The pooled HCV seroprevalence among all low risk groups aggregated (defined above), across all regions of sub-Saharan Africawas2.7% (95%CI: 2.5-2.8), Figure 1). This was higher than the pooled seroprevalence among all blood donor cohorts of2% (95%CI: 1.9-2,1) but lower than the 6.9% (95%CI: 6.1-7.5) observed in other general population cohorts (1.8%, 95%CI: 1.6–1.8).

There was considerable regional variation among the low risk cohorts (Figure 1) with pooled HCV seroprevalence highest in the central African region (6.9%; 50cohorts; 95% CI:6-7.8) and lowest in SE Africa (0.7%; 35 cohorts; 95%CI: 0.7-0.78). West Africa contributed the largest number of cohorts (100) and most closely approaches the overall estimate.

The pooled prevalence in aggregated low risk populations was a little higher than among blood donors (overall 2%, 76 cohorts; 95%CI:1.9–2.1%), but this difference was most marked in the Central region with a seroprevalence amongst blood donors of 2.6% (19 cohorts; 95% CI: 1.7-3.6). The HCV pooled seroprevalence in the West Africa region among blood donors was 3.2% (95%CI: 2.8-3.6; 45 cohorts), and 0.4% (12 cohorts; 95%CI: 0.3-0.5) in SE Africa. The highest HCV seroprevalence estimate in a blood donor cohort included in this analysis was 14.6% from a study that sampled blood donated at teaching hospitals and privately owned blood banks in Nigeria (n=624)21.

Among ANC populations,overall pooled HCV seroprevalence was 3% (21 cohorts; 95%CI: 2.2–3.8%), similar to the aggregated low risk cohort pooled estimate(Figure 1;).However, regional sub-analysis, despite low cohort numbers, showed the ANC pooled HCV seroprevalence in Central Africa was again markedly lower (1.71; 95% CI: 1.3-2.2) than the overall Central region estimate in low risk populations. Pooled ANC seroprevalence in the West Africa (4.3%; 95%CI: 3-5.6) was higher than theCentral Africa regionSE Africa (pooled HCV seroprevalence of 0.4% (95%CI: 0-0.8).

Finally a sub-analysis of the other 88 cohorts that were included in the low risk category, namely samples from the “general population” as well as inpatients and outpatients, not deemed to be a high risk of HCV, showed an overall pooled seroprevalence of 6.9% (95% CI: 6.1-7.7), i.e. higher than the estimates for ANC and Blood Donor populations. Regional variation was marked with a pooled seroprevalence estimate in Central Africa of 12.1% (26 cohorts; 9.4-14.8) compared with 2.4% (21 cohorts; 96% CI: 1.8-3.1) in SE region and 5.7% (41 cohorts; 95% CI: 4.7-6.8) in West Africa. In all regions, pooled estimates in these “other” low risk cohorts was higher than in ANC and Blood donor populations..

High-risk groups

41cohorts across 15countries were categorised as highrisk for HCV infection (20 from individuals with known liver disease, 21from individuals with high-risk exposures, e.g. multiple transfusions, injecting drug use, haemodialysis, healthcare workers, and prisoners), see Figure 2. The studies are detailed in Supplementary information table 3.

Pooled HCV seroprevalence estimates across all 21high-risk exposure cohorts were higher than those for low risk cohorts at 11.9% (95%CI: 7.1-16.7), with similar values across all regions (Figure 2). The estimates for high-risk cohorts in the Central region (8.4; 95% CI: 5.1-11.8) was lower than the pooled seroprevalence estimate amongst “other” low risk cohorts for the same region, although cohort numbers are low. The highest HCV seroprevalence estimate identified through the literature review was recorded in Kenya in a sample population of 145 injecting drug users(46.2% HCV infected)22.

In contrast, analysis of the 20liver disease cohorts(Figure 2 ; Supplementary information table 3) found an overall pooled HCV seroprevalence estimate of 10% (95%CI: 6.8-13.1). Although similar levels of infection were estimated in West and SE regions, in the Central region the mean pooled seroprevalence of HCV among liver disease cohorts sampled was 2.5% (4 cohorts; 95%CI: 1.0-4.0). This is lower than estimates in this region for the aggregated low risk cohorts and in line withestimates for blood donor and ANC populations in theCentral region.

Our resultssuggest that whilst high-risk exposures may result in a similar level of HCV infection across the regions of Africa, the proportion of liver disease caused by HCV varies significantly.

HIV co-infection prevalence

Seroprevalence of HCV/HIV co-infection was reported in 101cohorts from 27 countries; some of which were recruited as HIV positive cohorts (61) and some identified as a sub-group within the primary analysis of each study. (A total of 42,648 HIV-positive individuals were included(Supplementary informationtable 3). 74% of cohorts (75/101) included more than 50 individuals.

The overall pooled seroprevalence of HCV co-infection among HIV-infected individuals was estimated at 5.7% (95%CI: 4.9 - .6.6, Figure 3;). The majority of co-prevalence studies were from countries in the West Africa region (42cohorts), with a pooled HCV seroprevalence of 6.7% (95%CI: 4.8-8.5and SE Africa (43cohorts) with apooled HCV seroprevalence of 4.6% (95%CI: 3.7-5.4). Estimates of co-infection in HIV-positive cohorts in all regions lie between those for blood donor/ANC cohorts and high-risk cohorts.

Some countries, e.g. Senegal, Rwanda, Djibouti, and Gambia report relatively low HIV prevalence and low levels of co-infection (Figure 4). Some areas of high HIV prevalence, e.g. Malawi, Zimbabwe, and Zambia in southern Africa are estimated to have low HCV co-prevalence levels. The highest levels of co-prevalence are seen in East and South-East Africa (Tanzania, Mozambique, and Kenya) and Central Africa (Cameroon, Burundi, and Angola) of which only Mozambique reports HIV prevalence above 10%. Considering only countries with either a sample size greater than 500 HIV-positive patients or more than four HIV-positive cohorts, co-infection with HCV was estimated to be greater than mean seroprevalence of 10% in Burundi, Cameroon, Kenya, Mozambique, Tanzania, and Burkina Faso, and less than 5% in Uganda, Malawi, South Africa, and Côte d’Ivoire.

40cohorts with 3422HCV-infected individuals were included, with pooled HIV prevalence of 15.9% (95%CI: 12.5-19.2). Regional breakdown revealed the highest rates of HIV co-infection in those with known HCV were in SE Africa (33.4%,14 cohorts; 95%CI: 17.5-49.4) by contrast with West Africa (10%, 18 cohorts; 95%CI: 6.1-14) and Central Africa (5.9%, 8 cohorts; 95%CI: 2.5-9.3).

Confirmatory and molecular testing

Only 52/185low risk cohorts employed confirmatory assays to screen for the presence of anti-HCV antibodies. The mean estimates for pooled HCV seroprevalence overall and in each of the regions werenot significantly different to those found from all general population cohorts when restricted to those with confirmatory assays.

Analysis of the 35low risk cohorts that also reported PCR testing revealed the pooled mean proportion of positive serology samples that were PCR positive was 49.5% (95%CI: 40.3.-58.9).(Figure 5) Data for the genotype distribution of HCV was reported for only 39cohorts in 19 countries.

Discussion

We have estimated the pooled seroprevalence of HCV infection in sub-Saharan Africa to be 3% (95%CI: 2.9-3.1)when all samples were analysed collectively, and2.7% (95%CI: 2.5-2.8) when limited to‘low risk”samples in line with previous definitions. We identified regional variation among groups considered “low risk”, with the highest pooled seroprevalence calculated within the Central Africa region. In all regions, the estimates of HCV seroprevalence were lower among blood donor cohorts and ANC groups thanwithin aggregated low risk cohorts. However the regional and overall pooled seroprevalences in the “other” low risk groups were higher. These cohorts included individuals in healthcare settings as part of case control studies or convenience sampling. Such settings are known to be associated with transmission of HCV due to potential nosocomial transmission through fomites or non-exposure prone procedures including venepuncture and cannulation14,24-26 and may explain why observed prevalences are higher than in population surveys.

The overall estimates for HCV seroprevalence are very similar to those described previously, despite the estimates being based on different studies and different time periods [13] {Madhava, 2002 #396}. This gives more confidence on the robustness of these estimates and suggests no major change in HCV prevalence between the periods of study. However, larger population surveys, repeated over time, are required to monitor these trends. Given that population surveys are time consuming and expensive, other complementary methods are required to monitor changes in prevalence. The observation that HCV prevalences in antenatal cohorts are similar to those in the overall population suggests antenatal surveillance may represent a useful population for monitoring, particularly as there are many existing programmes testing for HIV in this group.