The epidemiology of hepatitis C virus in Pakistan: systematic review and meta-analyses

To characterize hepatitis C virus (HCV) epidemiology in Pakistan and estimate the pooled mean HCV antibody prevalence in different risk populations, we systematically reviewed all available records of HCV incidence and/or prevalence from 1989 to 2016, as informed by the Cochrane Collaboration Handbook. This systematic review was reported following the PRISMA guidelines. Populations were classified into six categories based on the risk of exposure to HCV infection. Meta-analyses were performed using DerSimonian and Laird random-effects models with inverse variance weighting. The search identified one HCV incidence study and 341 prevalence measures/strata. Meta-analyses estimated the pooled mean HCV prevalence at 6.2% among the general population, 34.5% among high-risk clinical populations, 12.8% among populations at intermediate risk, 16.9% among special clinical populations, 55.9% among populations with liver-related conditions and 53.6% among people who inject drugs. Most reported risk factors in analytical epidemiologic studies related to healthcare procedures. Pakistan is enduring an HCV epidemic of historical proportions—one in every 20 Pakistanis is infected. HCV plays a major role in liver disease burden in this country, and HCV prevalence is high in all-risk populations. Most transmission appears to be driven by healthcare procedures. HCV treatment and prevention must become a national priority.


Introduction
Hepatitis C virus (HCV) is a blood-borne pathogen and a significant global health concern [1]. Following the acquisition 2018 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited.

Inclusion and exclusion criteria
The inclusion and exclusion criteria used in this study were adapted from our previous systematic reviews [11][12][13][14][15][16][17]. Briefly, any article reporting HCV antibody incidence and/or antibody prevalence, based on primary data, qualified for inclusion in this review. An article was excluded if it was a case report, case series, editorial, letter to editor(s), commentary, review, referred to HCV as non-A non-B hepatitis, contained duplicate information, reported HCV prevalence based on self-reporting, and if the study population was Pakistani nationals residing outside Pakistan. [26].
In this work, for clarity, a 'report' refers to a document (article, conference abstract, country-level report and others) including one or several outcome measures of those included in our systematic review, while a 'study' refers to any one specific single outcome measure. One report may contribute multiple studies (say several prevalence measures in different populations), and multiple reports of the same outcome measure (say same prevalence measure in the same specific sample) were identified as duplicates and deemed as one study.

Data extraction and data synthesis
Data from relevant reports were extracted by Z.A.K., of which 20% were double extracted by S.P.K. to ensure consistency. Nature of extracted data followed our previous systematic reviews [11][12][13][14][15][16][17]. rounded to one decimal place except for measures below 0.1%, which were rounded to two decimal places.
Risk factors that were found to be significantly associated with HCV infection through multivariable regression analyses were extracted. HCV ribonucleic acid (RNA) prevalence among HCV antibodypositive individuals (that is HCV viraemic rate [20]) was extracted whenever available in reports including an HCV prevalence.
The extracted data were synthesized by risk population in six distinct categories defined according to the risk of exposure to HCV infection as follows: 1. General population (populations at low risk): these included blood donors, pregnant women, children, refugees, household-based survey participants and national army recruits, among others. 2. High-risk clinical populations: these included populations exposed to frequent medical injections and/or blood transfusions, such as haemodialysis, thalassaemia, haemophilia and multi-transfused patients, among others. 3. Populations at intermediate risk: these included populations whose risk of exposure is higher than the general population but lower than populations at high risk, such as healthcare workers (HCWs), household contacts of HCV-infected patients, patients with diabetes and prisoners, among others. 4. Special clinical populations: these included clinical populations whose risk of exposure to HCV infection is difficult to ascertain, such as patients with non-liver-related malignancies, dermatological manifestations and rheumatological disorders, among others. 5. Populations with liver-related conditions: these included patients with liver-related conditions of an epidemiological significance to HCV infection such as patients with chronic liver disease, acute viral hepatitis, hepatocellular carcinoma and liver cirrhosis, among others. 6. People who inject drugs (PWID).

Quantitative analysis
The quantitative analysis approach was similar to that in our previous HCV systematic reviews [11][12][13][14][15][16][17]. HCV prevalence measures were presented by risk population in reports with a sample size greater than or equal to 50 in tables 1-3; electronic supplementary material, S2-S4. If no explicit HCV prevalence measure was reported, it was calculated based on the sample size and number of events reported, if available. HCV prevalence for the total sample was replaced with stratified measures, whenever the sample size was greater than or equal to 25 participants for each stratum. Stratified data were included using a pre-defined order that prioritizes stratifications by population followed by sex, year, region and age. Meta-analyses were conducted for studies/strata with a minimum sample size of 25 participants. Only one final stratification per study was included in the meta-analyses.
The variance of the prevalence measures was stabilized using the Freeman-Tukey type arcsine squareroot transformation [138]. Estimates for HCV prevalence were weighted by the inverse variance and pooled using a DerSimonian-Laird random-effects model. This model accounts for sampling variation (random chance) and expected heterogeneity in effect size across studies [139]. Heterogeneity was assessed and characterized using several statistical measures.
With a recently identified potential issue with the Freeman-Tukey type arcsine square-root transformation [140], we conducted sensitivity analyses by performing meta-analyses using the generalized linear mixed models (GLMM) method to confirm validity of our results.
Meta-analysis of RNA HCV prevalence measures among HCV antibody-positive individuals (that is HCV viraemic rate) was also conducted to estimate the pooled mean of this prevalence measure.
A sensitivity analysis was further performed to examine whether the advent of more specific and sensitive diagnostic tools (third or fourth generation assays) could have affected the prevalence estimates in the general population. Meta-analyses were performed on the general population prior to and after 2005, since after this year the vast majority of studies were likely to have been conducted using third of fourth generation assays. The results of the meta-analyses were assessed to determine whether the estimated pooled mean HCV prevalence was significantly different prior to 2005.
Meta-analyses were conducted in R v. 3.1.2. [141], using the package meta [142]. Agboatwalla [29] 1990-1991 Kakepoto [30] 1989-1994 Karachi and Hyderabad      The table reports only studies whose sample size is greater than or equal to 50 participants. For space considerations, the table shows the overall HCV measure of each study rather than stratifications within population subgroups. b The decimal places of the prevalence figures are as reported in the original report, but prevalence figures with more than one decimal place were rounded to one decimal place, with the exception of those below 0.1%.   The table reports only studies whose sample size is greater than or equal to 50 participants. For space considerations, the table shows the overall HCV measure of each study rather than stratifications within population subgroups. b The decimal places of the prevalence figures are as reported in the original report, but prevalence figures with more than one decimal were rounded to one decimal place, with the exception of those below 0.1%. Prev, prevalence; CC, case-control; CS, cross-sectional; Conv, convenience; SRS, simple random sampling.   The table reports only studies whose sample size is greater than or equal to 50 participants. For space considerations, the table shows the overall HCV measure of each study rather than stratifications within population subgroups. b The decimal places of the prevalence figures are as reported in the original report, but prevalence figures with more than one decimal were rounded to one decimal place, with the exception of those below 0.1%. Prev, prevalence; CS, cross-sectional; Conv, convenience; RDS, respondent-driven sampling; SRS, simple random sampling.

Quality assessment
The quality of HCV prevalence measures was assessed for each study as informed by the risk of bias (ROB) Cochrane approach [143], as well as by examining the precision of each reported measure. The ROB assessment was based on three domains: type of HCV ascertainment (biological assays versus unclear), the sampling methodology (probability-based versus convenience sampling) and the response rate (greater than or equal to 80% versus less than or equal to 80% of the target sample size). Studies were considered as having high precision if the number of HCV tested individuals was at least 100 participants, as informed by previous studies [11][12][13][14][15][16][17]. Figure 1 describes the process of study selection, adapted from the PRISMA flow diagram [26]. A total of 1375 citations were identified: 480 through PubMed and 895 through Embase. A total of 517 reports were identified as relevant or potentially relevant after removing duplicates and screening the titles and abstracts. Out of these, 285 reports were excluded for various reasons as summarized in figure 1. An additional report was identified through screening of articles' references, and 11 HCV prevalence measures/strata were obtained from the Pakistan National Survey [10]. Finally, 233 eligible reports were included in this systematic review, yielding one incidence study and 248 prevalence measures. The 248 prevalence measures contributed 341 prevalence measures/strata. Though no language restrictions were imposed, all identified studies were in English.

HCV incidence overview
Our search identified one HCV incidence study, which reported seroconversion risk. This study included (as its baseline) HCV-negative HCWs who reported a needle stick injury from documented HCV-positive patients. After six weeks follow-up, investigators reported a seroconversion risk of 4.8% [144].

General population
Among the general population (table 1), our search identified 148 prevalence measures/strata, ranging from 0.4 to 44.0%, with a median of 5.3%. Among blood donors (number of studies; n = 57), HCV prevalence ranged from 0.4 to 20.8%, with a median of 3.5%. Among pregnant women (n = 12), HCV prevalence ranged from 0.7 to 20.7%, with a median of 6.0%. Among outpatients (n = 9), HCV prevalence ranged from 4.4 to 51.0%, with a median of 9.0%. Among other general populations (n = 65), HCV prevalence ranged from 0.4 to 35.9%, with a median of 6.8%.

High-risk clinical populations
Among high-risk clinical populations (table 2), our search identified 21 prevalence measures/strata, ranging from 7.8 to 68.0%, with a median of 34.5%. Among thalassaemia patients (n = 12), HCV prevalence ranged from 7.7 to 60.0%, with a median of 42.2%. Among haemodialysis patients (n = 7), HCV prevalence ranged from 16.4 to 68.0%, with a median of 28.0%. Only one study was conducted for each of haemophilia patients (prevalence of 51.4%) and multi-transfused patients (prevalence of 54.2%).

Special clinical populations
Among special clinical populations (electronic supplementary material, table S3), our search identified 18 prevalence measures/strata, ranging from 1.0 to 81.0%, with a median of 15.5%. Among patients with skin disorders (n = 4), HCV prevalence ranged from 3.0 to 23.4%, with a median of 7.7%. Among patients with urological conditions (n = 4), HCV prevalence ranged from 1.0 to 25.9%, with a median of 9.6%.

Populations with liver-related conditions
Among populations with liver-related conditions (electronic supplementary material, table S4), our search identified 73 prevalence measures/strata, ranging from 3.0 to 100.0%, with a median of 63.5%. Among chronic liver disease patients (n = 20), HCV prevalence ranged from 4.9 to 78.4%, with a median of 41.1%. Among cirrhosis patients (n = 21), HCV prevalence ranged from 28.0 to 100.0%, with a median of 68.0%. Among hepatocellular carcinoma patients (n = 18), HCV prevalence ranged from 33.3 to 92.0%, with a median of 70.1%. Among acute viral hepatitis patients (n = 6), HCV prevalence ranged from 6.4 to 57.1%, with a median of 20.9%.

People who inject drugs
Among PWID (table 3), our search identified 15 prevalence measures/strata, ranging from 8.0 to 94.3%, with a median of 44.7%.

Overview of HCV RNA prevalence among HCV antibody-positive individuals
Our search identified a total of 12 HCV RNA prevalence measures among HCV antibody-positive individuals (HCV viraemic rate). The details of these measures can be found in the electronic supplementary material, table S6. HCV viraemic rate ranged from 44.4 to 98.0%, with a median of 74.2%.

Pooled mean HCV prevalence estimates
Pooled mean estimates for HCV prevalence for the six risk populations are summarized in table 4. The pooled mean prevalence for the general population (populations at low risk) was estimated at 6.2% (95% CI: 5.7-6.7%). Meanwhile, the pooled mean HCV prevalence was estimated at 34 Of note, the GLMM meta-analyses produced similar pooled mean estimates for all risk populations. For example, the pooled mean HCV prevalence for special clinical populations, that showed the largest difference between the fixed effects result and the random-effects result, was 13.1% (95% CI: 6.9-31.3) using the GLMM method versus 16.9% (95% CI: 6.2-31.3%) using the Freeman-Tukey type arcsine square-root transformation method.
Statistically significant heterogeneity in effect size (that is HCV prevalence) was observed in all metaanalyses (Cochrane's Q-statistic's p-value was always less than 0.0001; table 4). Most of the variation across pooled studies was due to true difference in effect size rather than chance (I 2 > 93.7%). The prediction intervals were generally very wide. The totality of these heterogeneity measures indicates high heterogeneity in HCV prevalence measures in each risk population category.
Injecting drug-use-related risk factors were also commonly reported, including history of injecting drug use [ source of needles or syringes [135] and 'jerking' (drawing blood into a syringe while injecting) [131]. Sexual risk factors were also reported, including sex work (females and males), and sex for drugs [146].

Quality assessment of HCV incidence and prevalence measures
Findings of the quality assessment are summarized in the electronic supplementary material, table S5. Only one study was identified for HCV incidence [144] (not shown in the electronic supplementary material, table S5), in which there were greater than or equal to 100 participants, and was therefore classified as having high precision. As it was based on convenience sampling, it had high ROB for this domain. Meanwhile, it had low ROB in HCV ascertainment and in the response rate domains. The majority of HCV prevalence studies (86.7%) was based on samples with greater than or equal to 100 participants, and were therefore classified as having high precision. Most studies (67.7%) reported specific details about HCV ascertainment, but nearly 70% did not report the assay generation. When information was provided, 94.2% of studies reported use of third or fourth generation assays.
A sensitivity analysis was performed to assess whether HCV prevalence in the general population differed prior to and after 2005, because the vast majority of studies after this year were likely to have been conducted using third or fourth generation assays. The confidence intervals of the estimated pooled mean HCV prevalence prior to and after 2010 overlapped, indicating HCV prevalence was not significantly different between these two time durations.
The majority of HCV prevalence studies (92.3%) used convenience, non-probability-based sampling approach. Nearly half of studies had low ROB in the response rate domain and 48.8% had missing information-only 1.6% of studies had high ROB in this domain.
To summarize, 78.6% of studies had low ROB based on at least one domain, and 41.1% had low ROB based on at least two domains. Furthermore, 1.2% of studies had high ROB based on two domains, and no study had high ROB based on three domains. The totality of the quality assessment measures indicates reasonable study quality.

Discussion
We presented a systematic review and synthesis of HCV incidence and prevalence in Pakistan. Our results affirm that Pakistan has one of the highest HCV infection levels in both MENA [11][12][13][14][15][16][17] and worldwide [149][150][151]. HCV prevalence in the population at large is at about 5%-one in every 20 Pakistanis has been already exposed to HCV infection. HCV prevalence was also found to be high in all risk populations, testifying to the scale of the epidemic in this country. Our results further supported a major role for HCV infection in liver disease burden in Pakistan-over half of the populations with liver-related conditions were found HCV antibody-positive.
Our results collectively indicate a major role for healthcare in HCV transmission. High HCV prevalence was observed in the populations exposed to healthcare in one form or another. In high-risk clinical populations, the pooled mean HCV prevalence was high at 34.5% (95% CI: 27.0-42.3%) (table 4), with HCV prevalence ranging across studies from 7.8 to 68.0% (table 2)-much higher than that found in the general population. In special clinical populations, the pooled mean HCV prevalence was also high at 16.9% (95% CI: 6.2-31.3%) (table 4), with HCV prevalence ranging across studies from 1.0 to 81.0% (electronic supplementary material, table S3). In all identified reports on hospitalized populations, HCV prevalence ranged from 2.5 to 71.0%, with a median of 13.2% (electronic supplementary material, table S2).
Our assessment of HCV risk factors further indicates that HCV transmission appears to be primarily driven by healthcare-related exposures, such as therapeutic injections, intravenous infusions and poor sterilization of medical equipment [42,51,54,71,87]. Injecting drug use and other community-based exposures appear also to play a role, but their relative (as opposed to absolute) role is probably small compared with healthcare procedures [152]. These findings demonstrate the urgency of addressing the HCV epidemic in Pakistan, one of the world's largest, and where 10% of the global number of chronically infected people are living [6,21].
The apparent major role for healthcare in HCV transmission distinguishes Pakistan from most other countries. Though healthcare plays a role in both developing and developed countries [11][12][13][14][15][16][17][153][154][155], healthcare practices appear to have driven HCV prevalence to atypically high levels in this country, a pattern seen only in a limited number of countries globally, such as Egypt [13,22,23] and former Soviet republics [156]. This role for healthcare is not only manifested in the high HCV prevalence in the different clinical populations (table 4) [42,51,54,71,87], but also in the outcomes of viral hepatitis surveillance [157]. For example, the recently established viral hepatitis surveillance system in Pakistan indicated that healthcare-related exposures appear to be behind most newly reported HCV viral hepatitis cases [157]. Importantly, the surveillance demonstrated also that HCV accounted for over half of reported viral hepatitis cases [157], highlighting the special role of HCV infection in viral hepatitis disease burden in this country.
Of healthcare exposures, unnecessary therapeutic injections and reuse of syringes and needles were highlighted often as key factors [157,158]. Pakistan has one of the highest rates of therapeutic injections worldwide [159,160]-with widespread perception that injectable medications are more effective than oral medications [161][162][163]. Financial incentives appear also to sustain this preference for injectable medications, as healthcare providers can charge more for medications when they are administered by injections [163]. Though Pakistan has attempted to enhance provision and use of disposable injections and passed regulations for the management of disposable medical devices [164], implementation has been challenging in a country where the private sector accounts for 70% of healthcare services [157,162,165]. It bears notice that despite a possible key role for therapeutic injections, the totality of the evidence synthesized in the present study suggests that HCV healthcare exposures occur through multiple and diverse healthcare procedures.
The regional context of Pakistan and drug trafficking routes [166] support a conducive environment for injecting drug use. Our results indicated a high HCV prevalence among PWID (table 4), and evidence for injecting drug use as a mode of HCV exposure [54,87,131,157]. However, with an estimate of only 104 804 active PWID in Pakistan [167][168][169], the relative contribution of injecting drug use to HCV incidence is probably substantially smaller than that of healthcare, although the exact quantitative contribution remains uncertain.
Our results highlight the urgent and immediate need for expansion of HCV treatment and prevention programmes in Pakistan. High HCV prevalence was observed among all risk populations (table 4), with about one in every 20 Pakistanis being infected. Furthermore, three-quarters of all HCV antibodypositive individuals in Pakistan, per the meta-analysis of HCV viraemic rate (Section: Pooled mean HCV prevalence estimates), are chronically infected with HCV and can transmit the infection further. In spite of heavily discounted prices for DAAs in Pakistan [170], treatment scale-up has been limited, with only 311 000 chronic infections treated since 2013 [171]. To reach the WHO global target of reducing incidence by 80% by 2030, a recent modelling study indicated that the annual number of treatments must reach 490 000 and be sustained at this level for at least a decade [24]. To address the alarmingly high burden of HCV and achieve WHO global targets by 2030, Pakistan has recently developed the first National Hepatitis Strategic Framework, emphasizing the scale-up of interventions in healthcare settings and of HCV screening and treatment as well as harm reduction services [172].
Our study has identified key gaps and weaknesses in HCV epidemiological evidence in Pakistan. Despite the large epidemic, only one (now outdated) nationally representative and probability-based population-based survey was conducted in this country [10]. Repeating and enhancing this survey is critical to assess trends in prevalence and risk factors, as well as potential changes in the epidemiology. Such surveys have played an instrumental role in elucidating our knowledge of HCV transmission and in informing HCV response in other countries, such as in Egypt [173][174][175][176][177][178][179] and the USA [180].
Despite the major role for healthcare, a relatively small number of studies have been conducted among clinical populations, or investigated healthcare-related exposures. This is to be contrasted, for example, with Iran where a large number of studies investigated the role of healthcare-despite the relatively small role of this mode of exposure in this country [16]. Hardly any analytical cohort studies have been conducted in Pakistan despite the large epidemic, in contrast to Egypt [13,22], another MENA country with a large HCV epidemic [23]. Despite some suggestive evidence for community-based exposures [152], such as visiting roadside barbers [161], this mode of exposures remains to be clarified with concrete analytical studies. Though HCV vertical transmission appears to account for a quarter of HCV infections among children under 5 years of age in Pakistan [181], only one study appears to have investigated this mode of exposure in this country [32].
Our study is limited by the quantity and quality of reviewed studies, as well as their representativeness of the different risk populations-most studies used convenience sampling as opposed to probability-based population-based sampling. Only PubMed and Embase databases were searched, but other HCV data may exist in unpublished (grey literature) form, or are published in non-indexed journals. There was extensive heterogeneity in HCV prevalence measures in each risk population-possibly because of variability within the specific studied subpopulation, geographical location, sex and age-group representation in the sample, sampling technique and participant recruitment, year of study and study quality. Despite these limitations, the main strength of our study is that we identified a large number of studies that covered different risk populations, and that facilitated a comprehensive synthesis of evidence and identification of gaps and weaknesses that preclude a satisfactory understanding of HCV epidemiology in Pakistan.

Conclusion
Pakistan is enduring an HCV epidemic of historical proportions-one in every 20 Pakistanis has been already infected with this infection playing a major role in liver disease burden in this country. HCV prevalence is high in all risk populations with most transmission apparently driven by healthcare procedures. Though our knowledge of the specific modes of exposure that drive transmission is improving, our understanding is still hampered by key gaps and weaknesses in available evidence. Conduct of repeated and comprehensive nationally representative and probability-based populationbased surveys is critical to assess HCV prevalence and trends, identify risk factors and modes of exposure, examine the spatial variability in prevalence, and assess HCV knowledge and attitudes.
HCV treatment and prevention must become a national priority in Pakistan. Although Pakistan has made efforts to increase coverage of safe injection and blood screening and to improve infection control [164, [182][183][184], commitment to prevention in all segments of the healthcare system, including the private sector, should be secured for this country to accomplish the HCV elimination target by 2030. Major expansion of infection control in healthcare facilities, and of harm reduction services for PWID, are warranted, as well as adoption of the WHO guidelines for the use of safety-engineered syringes [185,186]. Funding. This publication was made possible by NPRP grant no. 9-040-3-008 from the Qatar National Research Fund (a member of Qatar Foundation). The findings achieved herein are solely the responsibility of the authors. The authors are also grateful for infrastructure support provided by the Biostatistics, Epidemiology, and Biomathematics Research Core at Weill Cornell Medicine in Qatar.