The epidemiology of hepatitis C virus in Iran: Systematic review and meta-analyses

The aim of this study was to characterize hepatitis C virus (HCV) epidemiology in Iran and estimate the pooled mean HCV antibody prevalence in different risk populations. We systematically reviewed and synthesized reports of HCV incidence and/or prevalence, as informed by the Cochrane Collaboration Handbook, and reported our findings following the PRISMA guidelines. DerSimonian-Laird random effects meta-analyses were implemented to estimate HCV prevalence in various risk populations. We identified five HCV incidence and 472 HCV prevalence measures. Our meta-analyses estimated HCV prevalence at 0.3% among the general population, 6.2% among intermediate risk populations, 32.1% among high risk populations, and 4.6% among special clinical populations. Our meta-analyses for subpopulations estimated HCV prevalence at 52.2% among people who inject drugs (PWID), 20.0% among populations at high risk of healthcare-related exposures, and 7.5% among populations with liver-related conditions. Genotype 1 was the most frequent circulating strain at 58.2%, followed by genotype 3 at 39.0%. HCV prevalence in the general population was lower than that found in other Middle East and North Africa countries and globally. However, HCV prevalence was high in PWID and populations at high risk of healthcare-related exposures. Ongoing transmission appears to be driven by drug injection and specific healthcare procedures.

Quantitative assessment. The quantitative analyses were conducted following an analysis plan similar to that in our previous HCV systematic reviews 7,9,[21][22][23][24][25]27 . HCV prevalence data in reports comprising at least 50 participants were stratified by risk and summarized using reported prevalence measures. Meta-analyses of HCV prevalence measures were conducted by risk category for studies consisting of a minimum of 25 participants. Stratified measures were used in place of HCV prevalence for the total sample only if the sample size requirement was met for each stratum.
A pre-defined sequential order was followed when considering stratifications. Nationality was prioritized, followed by sex, year, region, and age. One stratification was included per study to avoid double-counting.
The variance of the prevalence measures was stabilized using the Freeman-Tukey type arcsine square-root transformation of the corresponding proportions 37 . Estimates for HCV prevalence were weighted using the inverse variance method and then pooled using a DerSimonian-Laird random effects model. This model accounts for sampling variation and expected heterogeneity in effect size across studies 38 . Heterogeneity was assessed using several measures. The forest plots were visually inspected and Cochran's Q test was conducted, where a p-value < 0.10 was considered significant 38,39 . The I² and its confidence intervals were calculated 38 . The prediction intervals were also calculated to estimate the distribution of true effects around the estimated mean 38,40 .
Univariable and multivariable random-effects meta-regressions, based on established methodology 31 , were conducted to determine population-level associations with HCV prevalence and sources of between-study heterogeneity. Variables entered into the univariable model included risk population, sample size (<100 or ≥100), study site, sampling methodology (probability-based or nonprobability-based), publication year, and median year of data collection. Variables were included into the final multivariable model if the p-value was <0. 10. Variables with a p-value < 0.05 in the final multivariable meta-regression were considered significant. The majority of HCV prevalence measures in the general population were among blood donors, a population that mainly includes healthy adults. Therefore, we performed a sensitivity analysis to ascertain the impact of excluding blood donors on our pooled mean estimate for HCV prevalence among the general population (Fig. S5).
Descriptive analyses of HCV genotypes and subtypes were also performed. Individuals with mixed HCV genotypes contributed to the quantification of each identified genotype separately. Meta-analyses of genotype proportions were also performed to estimate the pooled mean proportions for each genotype. The diversity of HCV genotypes was assessed using the Shannon Diversity Index 41 .
Meta-analyses were performed on R version 3.1.2 42 , using the package meta 43 . Meta-regressions were performed on STATA 13, using the metan command 44 . Qualitative analysis. Similar to our previous HCV systematic reviews 7,9,[21][22][23][24][25]27 , the quality of each incidence or prevalence measure was determined by assessing sources of bias that may affect the reported measure. The Cochrane approach was used to infer the risk of bias (ROB) 31 , and the precision of the reported measures was also evaluated. Studies were categorized into low or high ROB based on three quality domains: type of HCV ascertainment (biological assay or otherwise), rigor of sampling methodology (probability-based or nonprobability-based), and response rate (≥80% of the target sample size was reached or otherwise).
Studies with missing information for any of the three domains were categorized as unclear ROB for that specific domain. Studies where HCV measures were obtained from individuals presenting voluntarily to facilities where routine blood screening is conducted, or retrieved from patients' medical records, were considered as having low ROB on the response rate domain. HCV prevalence measures obtained from country-level routine reporting, with limited description of the methodology used to be able to conduct ROB assessment, were categorized as of unknown quality.
Studies where HCV measures were obtained from a sample size of at least 100 individuals were considered as having high precision. For an HCV prevalence of 1% and a sample size of 100, the 95% confidence interval (CI) is 0-5%; a reasonable CI for an HCV prevalence estimate.

Results
Search results. Figure 1 describes the selection process by which studies were included in this systematic review, adapted from the PRISMA flow diagram 32 . We identified a total of 3,696 citations: 443 from PubMed, 772 from Embase, 1,885 from SID, 242 from IMEMR WHO, and 354 from the abstract archive of the IAS. After exclusion of duplicates and screening of titles and abstracts, 844 unique reports remained, for which the full-texts were retrieved for full-text screening. After full-text screening, 515 reports were excluded for reasons specified in Fig. 1. An additional 10 records were identified through screening references of full-text articles and reviews. One country-level report was retrieved and included from the MENA HIV/AIDS Epidemiology Synthesis Project database 33,34 . In total, 340 eligible reports were included in this systematic review. This yielded five HCV incidence measures and 472 HCV prevalence measures.
All 3,696 citations underwent an independent secondary screening for HCV genotype studies (Fig. S2). After title and abstract screening and exclusion of duplicates, the full-texts of 144 reports were screened. In total, 44 reports were found eligible for inclusion in this secondary systematic review, yielding 66 HCV genotype measures. HCV incidence overview. We identified five incidence measures through our search (Table 1), three of which were conducted in Tehran. The highest sero-conversion risks were observed in thalassemia patients and hemodialysis patients, of 6.8% and 4.3%, respectively 45,46 . In special clinical populations, HCV incidence was measured in renal transplant patients and impaired glucose tolerance patients. The HCV sero-conversion risks were 2.1% and 0.71%, respectively 47,48 . In female drug users on methadone treatment (where the route of drug use was not specified) the sero-conversion risk was 2.5% 49 . No studies reported incidence rate, nor provided sufficient information for incidence rate to be calculated. HCV prevalence overview. General population. A total of 122 HCV prevalence measures were identified in the general population ( obtained from blood donors (n = 72) where HCV prevalence ranged from 0.0% to 3.1%, with a median of 0.3%. In pregnant women (n = 6), HCV prevalence ranged from 0.0% to 0.8%, with a median of 0.3%. In other general populations (n = 44), HCV prevalence ranged from 0.0% to 2.4%, with a median of 0.5%.
Populations at high risk. A total of 208 HCV prevalence measures were identified in populations at high risk (Table 3), ranging from 0.0% to 90.0%, with a median of 26.3%. The majority were conducted on high risk clinical populations (n = 127). In hemophilia patients (n = 25), HCV prevalence ranged from 6.0% to 90.0%, with a median of 54.0%. In thalassemia patients (n = 58), HCV prevalence ranged from 0.0% to 68.9%, with a median of 16.6%. In hemodialysis patients (n = 41), HCV prevalence ranged from 0.0% to 31.4%, with a median of 8.3%. In HIV positive patients (n = 25), HCV prevalence ranged from 3.9% to 89.3%, with a median of 67.7%. Among PWID (n = 56), HCV prevalence ranged from 11.3% to 88.9%, with a median of 51.4%.

Populations at intermediate risk.
A total of 70 HCV prevalence measures were identified in intermediate risk populations (Table S2), ranging from 0.0% to 48.0%, with a median of 3.3%. In prisoners (n = 15), HCV prevalence ranged from 0.7% to 37.9%, with a median of 4.1%. In homeless people (n = 10), HCV prevalence ranged from 0.0% to 48.0%, with a median of 3.0%. Half of these studies were conducted on homeless children, among which HCV prevalence ranged from 0.0% to 3.5%, with a median of 1.0%. In household contacts of HCV index patients (n = 5), HCV prevalence ranged from 0.0% to 3.3%, with a median of 2.2%. In healthcare workers (n = 11), HCV prevalence ranged from 0.0% to 37.0%, with a median of 0.0%. In drug users (where the route of drug use was not specified (n = 13), HCV prevalence ranged from 3.4% to 36.1%, with a median of 14.5%.
Special clinical populations. A total of 72 HCV prevalence measures were identified in special clinical populations (Table S3), ranging from 0.0% to 69.1%, with a median of 3.2%. In hepatitis B virus patients, prevalence ranged from 0.0% to 18.0%, with a median of 10.3%. In viral hepatitis patients (n = 9), HCV prevalence ranged from 0.0% to 34.9%, with a median of 6.1%. In patients with liver cirrhosis (n = 5), HCV prevalence ranged from 1.7% to 14.9%, with a median of 7.3%.
Pooled mean HCV prevalence estimates. Table 4 shows the results of our meta-analyses for HCV prevalence. The estimated national population-level HCV prevalence, based on the pooled HCV prevalence in the general population, was 0.3% (95% CI: 0.2-0.4%). There was significant evidence of heterogeneity (p < 0.0001). I 2 was estimated at 99.8% (95% CI: 99.8-99.8%), indicating that almost all observed variation is attributed to true variation in HCV prevalence rather than sampling error. The prediction interval was 0.0-1.5%. The pooled mean HCV prevalence for populations at high risk was 32.1% (96% CI: 28.1-36.2%). There was significant evidence of heterogeneity (p < 0.0001), with an I 2 of 99.0% (95% CI: 99.0-99.1%). The prediction interval was 0.0-88.5%. For the subpopulations of PWID and populations at high risk of healthcare-related exposures, the pooled means were 52.2% and 20.0%, respectively.
The forest plots for the HCV prevalence meta-analyses can be found in Figs S3 and S4.
HCV RNA prevalence. Our search identified a total of 55 HCV RNA measures. The details of each of these measures can be found in Table S4. These were reported either among HCV antibody positive individuals, or as a proportion of the entire sample. HCV RNA prevalence among HCV antibody positive individuals ranged from 0% to 89.3%, with a median of 61.9%. HCV RNA prevalence as a proportion of the entire sample ranged from 0% to 60.0%, with a median of 8.6%. HCV RNA prevalence as a proportion of the entire sample was high in several populations at high risk of healthcare-related exposures. HCV genotypes. HCV genotype data was identified in 66 studies including a total of 24,029 HCV RNA positive individuals. Of these, 895 individuals had an undetermined genotype and were therefore excluded from further analysis. The vast majority of individuals were infected by a single genotype, with only 2.9% being infected by multiple genotypes. The proportion of infections for each HCV genotype was highest in genotype 1 (58.2%), followed by genotype 3 (39.0%), genotypes 2 (1.7%), and genotype 4 (1.0%). The pooled mean proportion for genotype 1 was 56.3% (95% CI: 52.9-59.6%), genotype 3 was 38.8% (95% CI: 35.7-41.9%), genotype 2 was 0.4% (95% CI: 0.0-1.0%), and genotype 4 was 0.0% (95% CI: 0.0-0.1%).

Risk factors for HCV infection.
Genotype 1 was more common among populations at high risk of healthcare-related exposures than genotype 3. Meanwhile, genotype 3 was more common among PWID than genotype 1. Within genotype 1, subtype 1a and subtype 1b were isolated (where subtype information was available) from 79.5% and 20.5% of individuals, respectively. Quality assessment. The results of the quality assessment are summarized in Table 5. The majority of HCV incidence measures (60%) were based on samples with >100 participants, and therefore were classified as having high precision. Incidence studies were based on convenience sampling from clinical facilities, and 60% had a response rate >80%. All incidence measures were based on biological assays. The majority of HCV prevalence measures (77.4%) were based on samples with >100 participants, and therefore were classified as having high precision. Of the 403 prevalence measures, ROB assessment was possible for 402 measures.
All HCV prevalence measures were based on biological assays. In 25.0% of measures, information on the exact biological assay was missing. Approximately one third of the samples underwent secondary confirmatory testing, with the majority using the more sensitive and specific recombinant immunoblot assay (RIBA). Among studies where information was available on assay generation, the majority (71.2%) used the more recent, sensitive, and specific 3rd generation Enzyme-linked immunosorbent assay (ELISA) tests, and 26.9% used 2 nd generation ELISA. The majority of samples (82.6%) were drawn using non-probability based sampling. Response rate was high in 92.0% of studies.
In summary, HCV prevalence measures were of reasonable quality. All studies had a low ROB in at least one quality domain, 92.3% had a low ROB in at least two of the three quality domains, and 11.7% had a low ROB in   all three quality domains. Only 2.5% of studies had a high ROB in two of the three quality domains, and no study had a high ROB in all three quality domains.

Meta-regressions and sources of heterogeneity.
The results of our meta-regression models can be found in Table 6. The univariable meta-regression analyses identified population, study site, sample size, and year of data collection as significant predictors (with p < 0.1), and therefore eligible for inclusion in the final multivariable meta-regression model. Sampling methodology used (probability-based or nonprobability-based) was not associated with HCV prevalence (p > 0.1). In the final multivariable meta-regression analysis, all variables remained statistically significant (p < 0.05) with the exception of healthcare setting and unspecified study site. The final multivariable model explained 71.7% of the variability in HCV prevalence. Of note, the model indicated a statistically significant declining trend in HCV prevalence in Iran-year of data collection had an AOR of 0.93 (95% CI: 0.91-0.96).

Discussion
We presented a comprehensive systematic review and synthesis of HCV epidemiology in Iran. The pooled mean HCV prevalence in the general population was estimated at only 0.3%, on the lower side of the levels observed in other MENA countries 7,9,[21][22][23][24][25]27 and globally [75][76][77] . Despite this low prevalence in the general population, high prevalence was found among PWID and populations at high risk of healthcare-related exposures. These findings suggest that most ongoing HCV transmission in Iran is driven by injecting drug use and specific healthcare-related exposures. Genotypes 1 and 3 were the most frequently circulating strains. Of note, HCV prevalence in Iran is on a declining trend (Table 6). Our estimate for the general population is slightly lower than an estimate provided for the whole adult population as part of a global estimation using a different methodology-0.3% in our study versus 0.5% in Gower et al. 78 . In all three quality domains 47 11.7

High risk of bias
In at least one quality domain 350 87.1 In at least two quality domains 10 2.5 In all three quality domains 0 0 Total studies where risk of bias assessment was possible 402 99.8 Total studies 403 100 Table 5. Summary of precision and risk of bias (ROB) assessment for the hepatitis C virus (HCV) incidence and prevalence measures extracted from eligible reports. a Studies with missing information for any of the domains were classified as having unclear risk of bias for that specific domain. b Studies extracted through country-level routine reporting with limited description of the sample (not permitting the conduct of risk of bias assessment) were classified as being of unknown quality. The difference may be explained by the fact that our estimate is strictly for the general (normally healthy) population. Moreover, our estimate is a pooled estimate of 122 studies as opposed to Gower's et al. estimate which was based on five studies 78 . Inclusion of blood donor studies in our estimation did not explain the difference-our sensitivity analysis showed that estimated HCV prevalence in the general population was invariable with exclusion of blood donors (Fig. S5). Iran has one of the highest population proportion of current PWID in the adult population (0.43%) in MENA, with an estimate of 185,000 current PWID 16,79 . Our synthesis indicated that injecting drug use was one of the most commonly reported risk factors for HCV infection, and that the pooled mean HCV prevalence among PWID was 52.2% (Table 4). These results suggest that injecting drug use is a main driver, if not the main driver, of HCV incidence in this country ( Table 6). The regional context of Iran and the drug trafficking routes 21,80,81 , support an environment of active injection and a major role for PWID in HCV transmission. In this regard, HCV epidemiology in Iran appears to resemble that in developed countries, such as in the United States of America (USA), where most HCV incidence is attributed to drug injection 17,82,83 . Of note, we identified high HCV prevalence even among drug users where the route of drug use was not specified or excluded drug injection. This may suggest under-reporting of drug injection among those who report just drug use, or past drug injection among them before shifting to other forms of drug use.
Having said so, the estimated low HCV prevalence in the general population of only 0.3% apparently contradicts with a large PWID population in Iran. In the USA, it is estimated that the population proportion of current PWID is 0.3% 84 , and that of lifetime PWID is 2.6% 84 , compared to 0.43% for current PWID in Iran 16 . HCV prevalence among PWID in the USA is just over 50% 85 , therefore comparable with the pooled estimate of 52.2% for PWID in Iran (Table 4). HCV prevalence in the wider adult population in the USA is estimated at 1.0% 86 , much higher than the pooled estimate for HCV prevalence in the general population in Iran (0.3%). This discrepancy may be explained by an over-estimated current PWID population in Iran, very recent trend of drug injection with relatively small lifetime PWID population, or that the estimated HCV prevalence in the general population considerably underestimates the actual HCV prevalence in the whole adult population in Iran.
Our synthesis suggests that prisons have been a major setting for HCV transmission in Iran (Table 6). With nearly 60% of prisoners being incarcerated for drug-related offences 81 , high reported injecting risk behaviors in prisons 16,28 , and the high HCV prevalence among prisoners (Tables 3 and S2), prisons should be a main focus of HCV prevention and treatment efforts. Iran has made major and internationally-recognized strides in establishing harm reduction services for PWID including in prisons 16,33,[87][88][89][90] , but further scale-up of these services in all prisons is warranted.
High HCV prevalence was found in populations at high risk of healthcare-related exposures such as hemodialysis, hemophilia, and thalassemia patients, though with geographical variation (Tables 3 and 6). This finding, along with the higher HCV prevalence generally among clinical populations (Table S3), suggests that healthcare is also a main driver of HCV transmission, though less so than in most other MENA countries 7  of healthcare and application of stringent protocols for infection control appear also to vary by setting within Iran. Overall, however, Iran seems to have made major progress in reducing HCV exposures through healthcare, which may explain the declining trend in HCV prevalence (Table 6) [91][92][93] . For example, HCV prevalence among hemodialysis patients was reported in one study to have declined from 14.4% in 1999 to 4.5% in 2006 94 . HCV genotype 1 was the dominant circulating strain in Iran (56% of infections), followed by genotype 3 (39% of infections). This shows similarity to the pattern observed in multiple countries globally 95 . Nevertheless, this genotype distribution differs substantially from that found in most other MENA countries 29 . Several recent studies have also indicated an increasing presence of genotype 3 96,97 . This shift may be due to the fact that injecting drug use is a major driver of HCV incidence 29,98 (Table 6), or the fact that this is a sub-regional pattern-genotype 3 is the main circulating strain in neighboring Pakistan 29 .
Our meta-analyses confirmed high heterogeneity in estimated effect sizes (Table 4). This was expected, due to differences between studies in variables such as risk population, study site, sampling methodology, sample size, and year of data collection, among others. Our meta-regressions identified several sources of heterogeneity in HCV prevalence studies in Iran. As expected, large differences in HCV prevalence by risk population were observed (Table 6). A small-study effect was also observed, with small studies reporting higher HCV prevalence. Importantly, a time trend was also observed with a declining HCV prevalence with time.
Our study is limited by the quality of available studies, as well as their representativeness of the different risk populations. High heterogeneity in prevalence measures were identified in all meta-analyses for all risk populations (Table 4). Meta-regression analyses were performed to identify the sources of heterogeneity, and while the final multivariable regression model accounted for 71.7% of observed heterogeneity, there are variables that we are unable to assess, such as "hidden" selection bias in recruitment.
Another limitation is the absence of reporting of the specific used biological assay in 25.0% of studies. The majority of included studies were based on convenience sampling. Although this is presumed a limitation, the meta-regression analyses did not identify sampling methodology as a statistically significant source of heterogeneity in HCV prevalence (p = 0.114; Table 6).
Despite these limitations, the main strength of our study is that we identified a very large number of studies, in fact the largest of all MENA countries 7,9,[21][22][23][24][25]27 , that covered all risk populations and that allowed us to have such a comprehensive synthesis of HCV epidemiology.

Conclusions
HCV prevalence in the wider population in Iran appears to be considerably below 1%-on the lower range compared to HCV prevalence in other MENA countries and globally. However, high HCV prevalence was found among PWID and populations at high risk of healthcare-related exposures. Most ongoing HCV transmission appears to be driven by injecting drug use and specific healthcare-related exposures. Genotypes 1 and 3 were the most frequently circulating strains.
There are still gaps in our understanding of HCV epidemiology in this country. Conduct of a nationally-representative population-based survey is strongly recommended to provide a better estimate of HCV prevalence in the whole population, delineate the spatial variability in prevalence, identify specific modes of exposure, and assess HCV knowledge and attitudes, as has been recently conducted in Egypt 10,99-103 and Pakistan 6,15,104 .
Our study informs planning of health service provision, development of policy guidelines, and implementation of HCV prevention and treatment programming to reduce HCV transmission and decrease the burden of its associated diseases. Our findings suggest the need of a targeted approach to HCV control based on settings of exposure. Iran has established internationally-celebrated harm reduction services for PWID 16,[87][88][89][90]105 , but these services need to be accessible to all PWID across the country, as well as in relevant settings, such as prisons. Further focus on infection control in healthcare facilities is also warranted, such as the adoption of the new WHO guidelines for the use of safety-engineered syringes 106,107 .