A clinical and molecular epidemiological survey of hepatitis C in Blantyre, Malawi, suggests a historic mechanism of transmission

Abstract Hepatitis C virus (HCV) is a leading cause of liver disease worldwide. There are no previous representative community HCV prevalence studies from Southern Africa, and limited genotypic data. Epidemiological data are required to inform an effective public health response. We conducted a household census‐based random sampling serological survey, and a prospective hospital‐based study of patients with cirrhosis and hepatocellular carcinoma (HCC) in Blantyre, Malawi. We tested participants with an HCV antigen/antibody ELISA (Monolisa, Bio‐Rad), confirmed with PCR (GeneXpert, Cepheid) and used line immunoassay (Inno‐LIA, Fujiribio) for RNA‐negative participants. We did target‐enrichment whole‐genome HCV sequencing (NextSeq, Illumina). Among 96,386 censused individuals, we randomly selected 1661 people aged ≥16 years. Population‐standardized HCV RNA prevalence was 0.2% (95% CI 0.1–0.5). Among 236 patients with cirrhosis and HCC, HCV RNA prevalence was 1.9% and 5.0%, respectively. Mapping showed that HCV RNA+ patients were from peri‐urban areas surrounding Blantyre. Community and hospital HCV RNA+ participants were older than comparator HCV RNA‐negative populations (median 53 vs 30 years for community, p = 0.01 and 68 vs 40 years for cirrhosis/HCC, p < 0.001). Endemic HCV genotypes (n = 10) were 4v (50%), 4r (30%) and 4w (10%). In this first census‐based community serological study in Southern Africa, HCV was uncommon in the general population, was centred on peri‐urban regions and was attributable for <5% of liver disease. HCV infection was observed only among older people, suggesting a historic mechanism of transmission. Genotype 4r, which has been associated with treatment failure with ledipasvir and daclatasvir, is endemic.

prevalence and attributable fraction of hepatitis C to liver disease.
Third, we sequenced whole HCV genomes from viraemic patients to characterize the molecular epidemiology of HCV in a region with very limited prior genotypic data. Finally, we conducted a systematic review of existing epidemiological data on hepatitis C to characterize the existing evidence on HCV prevalence in the general population in Southern Africa.

| Census and serological survey
Individuals aged ≥16 years were selected using single-stage random sampling from a recent full-population demographic census of an urban population of 96,386 and invited to participate in a serosurvey conducted between 14 December 2016 and 12 April 2018 as part of a co-incident ongoing typhoid epidemiology study (Strategic Typhoid Alliance across Africa and Asia) in Ndirande, Blantyre. 7 If a randomly selected individual could not be located or did not consent to participation, another household member aged >16 years was requested to participate, or secondarily, a replacement was selected using further randomization from the census. Demographic, educational, marital, and occupational data were recorded. Venous EDTA samples were collected in participants' households by research nurses, stored in cool boxes and transported to the study laboratory from the field. Plasma samples were separated by centrifugation and stored at −80°C.

| Clinical evaluation of liver disease
We returned to households of serosurvey participants who tested positive for HCV Ag-Ab by enzyme immunoassay (ELISA) and invited them to participate in a clinical and virological evaluation. We Among 236 patients with cirrhosis and HCC, HCV RNA prevalence was 1.9% and 5.0%, respectively. Mapping showed that HCV RNA+ patients were from peri-urban areas surrounding Blantyre. Community and hospital HCV RNA+ participants were older than comparator HCV RNA-negative populations (median 53 vs 30 years for community, p = 0.01 and 68 vs 40 years for cirrhosis/HCC, p < 0.001). Endemic HCV genotypes (n = 10) were 4v (50%), 4r (30%) and 4w (10%). In this first census-based community serological study in Southern Africa, HCV was uncommon in the general population, was centred on peri-urban regions and was attributable for <5% of liver disease. HCV infection was observed only among older people, suggesting a historic mechanism of transmission. Genotype 4r, which has been associated with treatment failure with ledipasvir and daclatasvir, is endemic.

K E Y W O R D S
Africa, cirrhosis, epidemiology, hepatitis C, Malawi, South of the Sahara additionally randomly selected a sample of 65 unmatched community controls aged >16 years from the serosurvey population and negative for HCV Ag-Ab and HBsAg to measure liver stiffness normal range. HIV testing was offered to participants using rapid diagnostic tests in accordance with national guidelines using Determine HIV (Alere, South Africa) and confirmatory testing with Uni-Gold HIV (Trinity Biotech, Ireland). We assessed liver stiffness using transient elastography in the right mid-axillary intercostal space after fasting for >3 h (FibroScan 430 Mini, Echosens, France). 8 Reliability criteria were IQR/median <0.3 if median LSM >7.1 kPa. 9

| Sample size calculation
Based on previous estimates of anti-HCV prevalence from Malawi in the preceding 20 years from convenience samples ranging from 0.5% to 10%, 10 we estimated sample size using a conservative prior prevalence estimate of 10% with 1.5% precision with a 95% confidence interval, giving a sample size requirement of 1537 for the community serosurvey (Appendix 1). No previous data were available describing the normal population distribution of liver stiffness among healthy individuals in sub-Saharan Africa, nor among community patients with hepatitis C infection. Using estimates from a study of people with hepatitis C infection who inject drugs in Tanzania, 11 and a normal population reference from a systematic review of 26 cohort studies, 12 our calculated sample size requirement for detecting a difference in the predicted mean liver stiffness measurement between patients with hepatitis C and healthy population controls was 64 in each group (Appendix 1).

| Laboratory investigations
A combined hepatitis C virus (HCV) antigen and antibody ELISA (HCV Ag-Ab ULTRA V2, Bio-Rad, France) was used as an initial test, in accordance with the manufacturer's instructions. All positive and indeterminate samples (sample-to-cut-off ratio (S/CO) >0.9) were retested in duplicate. Positives were tested with HCV RNA PCR (GeneXpert HCV, Cepheid, South Africa). We tested all HCV Ag-Ab+/HCV RNA-patients with available samples using a line immunoassay (Inno-LIA HCV score, Fujirebio, Japan). We tested for HBsAg using Monolisa HBsAg-Ultra (Bio-Rad), and positive and indeterminate samples (S/CO >0.9) were repeated in duplicate.

| HCV sequencing and bioinformatic analysis
Hepatitis C virus whole-genome sequencing was performed for all HCV RNA-positive participants using a previously described method for target enrichment using probe capture, followed by next-generation sequencing using the NextSeq 550 platform (Illumina, USA) with 150-bp paired read length. 17 Tanoti (http:// www.bioin forma tics.cvr.ac.uk/tanoti.php) was used to align unmapped reads against HCV reference sequences from the International Committee on Taxonomy of Viruses (https://talk.ictvo nline.org/). 18 HCV GLUE was used to identify the subgenotype and resistance-associated mutations, and repeat alignment was made using the closest subtype reference sequence. 19 Sequence alignment was performed using MAFFT with L-INS-I against near-fulllength genotype 4 genomes in the HCV Glue reference repository.
Uncorrected p-distance from closest reference sequences was calculated using MEGA 10.1.6. JModeltest2 was used to identify the optimal phylogenetic model. 20

| Ethical review
Ethical permission to conduct this study was obtained from the National Health Sciences Research Committee of Malawi

| Systematic review of existing data
We systematically searched for previous epidemiological data of hepatitis C prevalence in the general population in Southern Africa (full details are described in Appendix 2). Searches were conducted in PubMed, with no language or date restriction, for synonyms of hepatitis C or diagnostic tests and the countries in the African Union Southern Region. We included studies which reported anti-HCV antibodies or HCV RNA prevalence among a sample of the general population using an epidemiologically representative sampling method.

| Statistical analysis
We estimated population-standardized HCV prevalence using poststratification iterative proportional fitting using 5-year age groups and sex from the census to adjust prevalence estimates to reflect the population age and sex distribution. 22 We calculated the populationattributable fraction (PAF) of HCV to liver disease by comparing the prevalence of HCV RNA among patients with cirrhosis and HCC to the community prevalence. Following logistic regression, a postestimation predictive margin of response was estimated for a modelled scenario with HCV RNA prevalence of zero. The ratio between the logit of the baseline likelihood and the zero exposure scenario was calculated, representing the population-unattributable fraction (PUF), and PAF was calculated using (1-PUF). PAF calculations were implemented using the punafcc package. 23 Association between HCV RNA-positive participants and explanatory variables were assessed by comparison with the HCV RNA-negative population using Fisher's exact, chi-squared or Wilcoxon rank-sum test as appropriate, and using a binomial logistic regression model. We selected variables for inclusion in the model based on a priori clinical relevance and to optimize model fit using the Bayesian information criteria. We modelled the geographic incidence of cases of cirrhosis and HCC per 10,000 population using data from the National Population Census 2018. 24 Analyses were conducted using Stata 16.1 (StataCorp, USA) and ArcGIS Pro 2.8 (ESRI, USA).

| Community serological survey
We randomly selected a total of 1661 participants from the community census aged ≥16 years for HCV testing, with median age

| Hospital cohort study
Over an 18-month period, we screened 708 patients with suspected liver disease and recruited 236 consecutive patients attending the hospital and meeting study definitions for cirrhosis HCV Ag-Ab ELISA sample-to-cut-off ratios (S/CO) were significantly higher with a mean S/CO of 13.4 (95% CI 11.9-14.9) relative to 3.8 (95% CI 0.8-6.9) among HCV RNA-individuals, p < 0.001 (Appendix 5). All HCV RNA+ individuals had a S/CO value ≥9.

| Risk factors for HCV infection
Hepatitis C virus RNA+ individuals from the community and hospital study were significantly older than the community HCV RNA-population ( Figure 2) and were less likely to have undergone secondary education, and were more likely to be divorced or separated, by univariable analysis (

| DISCUSS ION
In this first census-based community HCV prevalence using representative random sampling in Southern Africa, we found that in the general population, HCV was uncommon with a prevalence of 0.2%. Previous prevalence studies from the region have suffered from important biases that may limit sample representativeness.
These include recruitment without a random sampling framework or using opportunistic sampling methods. We consulted three systematic reviews of HCV prevalence and performed an additional systematic search and did not identify any previous epidemiologically representative community prevalence studies for the region. 2,5,27 Our sample used household recruitment, based on random sampling from a population census, and closely matched the national census age and sex distribution. Our estimated HCV prevalence rates were significantly lower than existing estimates for the wider region, 2,5 and our data suggest a concentrated HCV epidemic that is confined to older individuals and peri-urban areas. If these data are confirmed with other representative population surveys from other sites in Southern Africa, the viraemic prevalence may be significantly lower than current estimates of 0.7%, representing 500,000 among the population of Southern Africa of 76 million. 2 Analysis of the geographic distribution referred with liver disease to the regional centre showed that all but one HCV patient was from Blantyre City, relative to an urban distribution for less than half of remaining patients referred with liver disease of other causes. Our data suggest that current data that are largely convenience samples from urban populations may overestimate true prevalence, since the majority of individuals live in rural settings in this region.
In our study, the strongest specific risk factor associated with HCV infection was increasing age. All people diagnosed with HCV in both community and hospital cohorts were aged older than 45, and all HCV-associated liver disease patients were older than 55 years, and significantly older than HCV-negative control populations. This suggests that a historic risk factor for transmission, such as noso- Studies are required to evaluate the efficacy of empirical treatment regimens with locally prevalent genotypes, particularly describing the outcomes of treating 4r in Malawi.
In conclusion, in this representative census-based community sample in Southern Africa, HCV infection was uncommon and lower than regional estimates based on convenience samples. We did not identify any prior epidemiologically representative community prevalence data from the region. In our hospital study, HCV was attributable to 5% of HCC and 2% of cirrhosis cases. HCV infection was observed only among people older than 45 years, suggesting a historic mechanism of transmission. All sequences were genotype 4, and subtype 4r, which has been associated with NS5A inhibitor resistance, is endemic. Representative sampling from specific risk groups in the region is now required.