Sexual transmission of Hepatitis C Virus infection in a heterosexual population: A systematic review

Background: Hepatitis C virus (HCV) infection is an important cause of liver disease worldwide. Identification of risk factors can guide screening and prevention. Sexual transmission in monogamous heterosexual relationships is rare but it is uncertain which sexual behaviours are linked to HCV transmission. This review aimed to determine risk factors for sexual HCV transmission in heterosexuals in low HCV prevalence countries (PROSPERO registration CRD42016051099). Methods: We searched Medline, Embase, Science Citation Index-Expanded, Social Sciences Citation index, Conference proceedings (Web of Science), CINAHL, Scopus, LILACS, PubMed, and grey literature (04/11/2016). We included studies published in/after the year 2000 that examined sexual risk factors for HCV infection, other than interspousal transmission, in heterosexual adults (≥18 years). We excluded prisoners, people who inject drugs (PWIDs), people co-infected with HIV or from high prevalence countries. Two reviewers completed study selection, data extraction, risk of bias and quality of evidence assessment (GRADE) independently. Meta-analysis could not be conducted. Results: Eight studies were included, examining seven factors (multiple sex partners, receiving/providing sex commercially, PWID partner, and unprotected vaginal, oral, anal sex). None were significant, except the evidence for the factor having a PWID partner was conflicting. Conclusions: We are uncertain about the results due to the very low quality of evidence (GRADE). A more liberal approach to review inclusion criteria might be useful in further identifying factors associated with an increased risk of sexual transmission of HCV infection in a heterosexual population. However, caution should be applied to avoid the impact of confounders on the findings.


Introduction
Hepatitis C virus (HCV) infection was first identified in 1989 1 . Chronic HCV infection is an important cause of chronic liver disease and liver related death worldwide 2 with an estimated 130 to 150 million persons having chronic HCV infection globally 3 . The prevalence of HCV infection varies across countries and areas 4 . Of those infected with HCV, many are asymptomatic with approximately 20% to 30% developing acute symptoms 5,6 and 55% to 85% developing chronic infection 7 . Of those who develop chronic infection, 15% to 30% will develop liver cirrhosis within 20 years 3 . Annually, approximately 3% to 4% of patients with liver cirrhosis develop hepatocellular carcinoma 8 .
Identification of risk factors associated with HCV transmission is essential in guiding screening and prevention strategies to improve health outcomes and maximise cost-effectiveness. There is currently no effective vaccination for HCV, putting even more emphasis on infection prevention 9 . Since the introduction of routine screening of blood in the early 1990s, transfusion-related HCV infection is rare. Literature on risk factors for HCV infection indicate that Injecting Drug Use (IDU) is now the main mode of transmission 10 . Other reported risk factors include occupational exposure, tattooing or having blood transfusions, vertical transmission and sex with an infected partner 10 .
The role of sexual transmission in HCV transmission is not fully understood and an increasing number of studies examine this question. Some studies found HCV RNA in semen 11,12 , but other studies have contradicted these findings 13,14 . Tohme and Holmberg 15 conducted a systematic review concerning the risk of HCV sexual transmission. They found that having multiple sexual partners might increase risk of HCV infection, although this finding may be confounded by IDU. Moreover, HIV co-infected individuals and Men who have Sex with Men (MSM) were clearly more at risk. Sexual transmission in people in monogamous heterosexual relationships on the other hand is rare; however, it is uncertain what specific sexual behaviours in heterosexuals do increase the risk of HCV transmission. Since their review was published in 2010, additional studies have emerged and an updated review was required to inform guidance on screening for HCV infection. We also aimed to increase generalisability to the context of low HCV prevalence countries because in high endemic countries there may be other risk factors in the population that make it difficult to identify sexual transmission at the source. Moreover, we focussed on heterosexuals but excluded high risk populations such as people who inject drugs (PWID), prisoners and HIV co-infected people, to address the question of when HCV screening is warranted in a more general population. Subsequently, the aim of this systematic review was to determine what factors, if any, are associated with an increased risk of sexual transmission of HCV infection in a heterosexual population in low HCV prevalence countries. This review was registered in PROSPERO (CRD42016051099). More details on the search terms can be found in Figure 1.

Selection criteria
The selection criteria were set in Population, Exposure, Outcome, Study design (PEOS) format. The population of interest included heterosexual adults (≥18 years), excluding those with HIV co-infection, PWIDs, homo-or bisexuals, or prisoners, because these populations are at high risk of HCV infection. In addition, this review excluded studies conducted in high HCV prevalence countries, because the aim of this review was to provide guidance for assessing the need of screening specific populations in the setting of low HCV prevalence countries 16 . The list of high HCV prevalence countries was obtained from the Health Protection Surveillance Centre (Ireland), Infectious disease assessment for migrants 17 , but the HCV prevalence of Nigeria was changed to high (>3%) following the publication of the epidemiological report on hepatitis C and B by the European Centre for Disease Prevention and Control in August 2016 16 .
The exposure was any sexual behaviour factor including (but not limited to) having multiple sex partners, overlapping (more than one sexual relationship at the same time), changing sexual partners frequently, unprotected sex outside of monogamous relationship (sex acts without the use of a condom), exchange of sex for drugs (but not PWID) or money, being a commercial sex worker, sex with commercial sex workers, sex under the influence of drugs or alcohol, anal sex, having another Sexual Transmitted Infection (STI) (excluding HIV), having a high risk partner (defined as any of the above). The exposure could be self-reported or based on an objective measure e.g. the number of occasions condoms used/not used. The outcome HCV infection had to be determined by antibody/antigen or PCR RNA test, excluding self-reported HCV status.
We included cohort studies, case-control studies and crosssectional studies, but excluded case studies, case series and reviews. We only included studies published in or after the year 2000 because of the variability in the quality of HCV serological testing in earlier studies.

Study selection
Records identified in the search were screened independently by title/abstract and then by full-text by at least two reviewers (FW, VS, PF, GG, LM, SS). Conflicts were resolved by FW, VS and PF through discussion, and if necessary by involving another reviewer (DD).
Data extraction and risk of bias assessment Risk of bias (ROB) in the included studies was assessed by two independent reviewers (FW, JCJ) using the Quality In Prognosis Studies (QUIPS) tool 18 . The ROB criteria for each QUIPS domain specific to this review, including appropriate methods for HCV infection measurement and important confounders should have been adjusted for, were documents a priori and agreed by all authors. Conflicts were resolved through discussion. A data extraction form was developed and reviewed by all authors. Data extraction was conducted by two reviewers independently (VS, GG, LM and FW). Any conflicts were resolved by a third reviewer (FW).

Data analysis
When only raw data (proportions) were available, we calculated the unadjusted Odds Ratios (OR) and 95% Confidence Intervals (CI) using the natural log scale 19 . We planned to conduct metaanalyses in Revman 20 and to assess statistical heterogeneity (I 2 ≥ 50%, T 2 > 0, or the p-value > 0.10 for the Chi square test) 21 . However, it was not appropriate to carry out meta-analysis due to clinical and methodological heterogeneity, and findings are summarised narratively and presented in evidence tables.
The quality of evidence was assessed by two independent reviewers (VS and FW) using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach for prognostic factor research 22 . This review was reported according to the Meta-analysis of observational studies in epidemiology (MOOSE) guidelines (completed checklist in Supplementary File 1) 23 .

Results
A total of 10460 records were identified through the database searches. We did not identify any records through the grey literature searches. Two further duplicates were detected and 10458 records were screened by title and abstract, of which 274 were selected as potentially eligible and assessed by full-text. A total of eight studies were finally included.
We contacted the authors to obtain more information to assess eligibility for an additional 11 records in abstract format and these are awaiting classification. Full details of the search results and selection process are presented in Figure 2.

Characteristics of included studies
The 8 included studies were published between the year 2000 and 2015, in Mexico (n=2), USA (n=2), Vietnam (n=1), Scotland (n=1), Gambia (n=1), and Brazil (n=1). A total of 14036 participants were included in the eight studies and seven potential sexual risk factors were assessed across the studies. These were multiple sex partners, receiving or providing sex commercially, having a PWID (People who inject drugs) partner, and unprotected vaginal, oral or anal sex.
Only three factors were examined in more than one study. The majority of studies included sample populations from specific groups: one study examined risk factors in a sample of blood donors 24 , one involved pregnant women attending antenatal services 25 , one in homeless people 26 , one in non-PWIDs 27 , one in nurses 28 , and one in a sample of different risk groups 29 . Full details of the characteristics of the included studies are presented in Table 1.

Characteristics of excluded studies
A total of 266 studies were excluded at full-text selection.
Reasons for exclusion included participants under the age of 18 in the sample (n=38), study examined the prevalence of HCV but did not assess risk factors (n=31), and study publication not available in English (n=24), which may have introduced language bias. Many studies examined risk factors other than sexual factors of interest (n=25). Twenty-eight were excluded because of their study design; 24 were narrative reviews, three were systematic reviews, and one study was a case report. Sixty-seven studies included PWIDs, homosexuals and/or (ex-) prisoners in their sample and did not report findings separately for these groups.

Risk of bias of included studies.
Details of the judgment of ROB of each domain for each study are provided in Table 2. Study participation was judged as moderate (n=2) or high (n=6) ROB for all studies. Participants of included studies were specific groups that might have influenced the findings, such as healthy blood donor (n=1), nurses (n=1), pregnant women (n=1), homeless people (n=1). One study had a low recruitment rate 30 , but other studies did not describe their recruitment rate.
Seven of the 8 included studies were cross-sectional studies, hence there was low attrition bias. Only Neaigus et al. 27 followed up patients and had a low retention rate of only 62.2%, leading to high risk of attrition bias. Six of the 8 studies were judged as low ROB for the domain risk factor measurement as the risk factor was measured using a structure questionnaire. Two studies were of moderate ROB because they did not provide a clear description of risk factor measurement.
All studies appropriately assessed the outcome HCV infection, most commonly using the ELISA kit, and were thus judged as low ROB. Four studies did not adjust for confounders and were therefore judged as high ROB for this domain. Nyamathi et al. 26 adjusted for confounders but did not include some important confounders such as a history of tattooing and was judged as moderate ROB. The remaining three studies were judged low ROB. Four studies did not conduct multivariable analysis to adjust for confounders and hence were rated as high ROB for statistical analysis. Nyamathi et al. 26 adjusted for some confounders (gender, age, age started living on their own, daily alcohol use) but not all important ones (e.g. tattooing/body piercing) were included in the model; hence this study was rated as moderate ROB.

Risk factors
A total of seven potential risk factors were examined in the eight included studies. Evidence for all factors was of very low quality and full details of the GRADE profiles by factor are provided in Table 3.
Six studies examined having had multiple sex partners as a potential risk factor for HCV infection (Table 4). Different cutoffs and analysis methods did not allow us to pool data in metaanalysis. Only one study adjusted for confounders and did not find having more than three sex partners in the last six months to be a significant risk factor 26 . Similarly, the other five studies did not find a positive association. However, evidence should be interpreted with caution since the quality of evidence (GRADE) is very low.
Evidence of being a commercial sex worker as a risk factor for HCV infection was limited (measured in 2 studies only 27,31 ) and of very low quality (GRADE). Dunford et al. 31 found that 8.7% of commercial sex workers were HCV positive, and of those 40.2% (n=87) were non-PWIDs, but did not report data on the comparison group (non-PWID CSWs negative for HCV). Neaigus et al. 27 found a positive effect but it was not statistically significant (unadjusted OR 2.0 (0.6-6.7); p=0.3; n=277).
A history of having sex with a commercial sex worker was not associated with HCV infection (unadjusted OR 1.9 (0.5-8.0), p=0.4; Male: HR 4.1 (0.91-18.0); Female: not reported; n=277; one study), but the quality of evidence (GRADE) was very low due to only one study examining this factor with significant risk of bias 27 .
a Downgraded by one level because seven of the eight studies had at least one domain high ROB or two domains moderate ROB; b Downgraded by one level because some studies have wide confidence intervals and no power calculations provided; c Downgraded two levels since the study has more than 1 domain high ROB or two moderate; d Downgraded by one level because only one or two study(ies) has(ve) provided an effect estimate; e Downgraded by one level because the confidence intervals do not overlap; f Downgraded by one level because of a study has more than 1 domain high ROB or two moderate; g For one of the two studies the effect could not be estimated and was subsequently classified as no effect; h GRADE levels of evidence: + very low, ++ low, +++ moderate, ++++ high quality; i Phase 1 of investigation (Hayden 2008), conducted only univariate analysis for the factor of interest; j Phase 1 of investigation (Hayden 2008), multivariable analysis but no specific hypotheses tested; k Explanation of symbols: 'v' not downgraded/upgraded, 'x' downgraded/upgraded by one level, 'xx' downgraded/upgraded by two levels. There was conflicting evidence regarding the role of having a partner who is a PWID as a risk factor for HCV infection and quality of evidence was very low due to significant ROB and a limited number of studies examining this factor. Goldberg et al. 25 found that having a PWID partner was a significant risk factor (unadjusted OR 56.6 (18.5 -173.60), p<0.0001; n=3498), but Neaigus et al. 27  were not associated with HCV infection (very low quality evidence) 27 .

Discussion
Seven potential sexual risk factors for HCV transmission in a heterosexual population were examined in eight studies, including multiple sex partners, receiving or providing sex commercially, having a PWID partner, and unprotected vaginal, oral or anal sex. None of these factors were statistically significant risk factors in the included studies; however, we are uncertain about these results due to the very low quality of evidence (GRADE). Moreover, these results might have been affected by a potential lack of statistical power and none of the studies provided sample/power calculations. Goldberg et al. 25 did find that having a sex partner who is a PWID was associated with an increased risk of HCV infection resulting in conflicting evidence because the other study examining this factor did not find a significant association 27 . Goldberg et al. 25 was conducted in Scotland, which might provide more geographically relevant information as this is a similar context to Ireland and other European countries. However, being a partner of a PWID could also expose people to non-sexual HCV transmission and may have confounded this finding.
Evidence for all factors examined in the included studies was of very low quality, mainly due to a lack of replication and high ROB resulting from a lack of adjusting for confounders and selective samples. Incomplete or non-standardised measurement of sex practices in some of the included studies could also have impacted on the findings of this review. Moreover, only seven factors were examined in the included studies and other factors such as use of condom, sex during menses, rough sex, presence of other STIs etc., were not measured and assessed. Even though we excluded studies that examined PWIDs, subjects may not disclose being a PWID, particularly since data on factors was generally obtained through a self-reported questionnaire.
The strengths of this review lie in its comprehensive search, its double independent study selection, ROB assessment, data extraction and GRADE quality assessment.
This review adhered to the a priori selection criteria set and excluded any study that included PWIDs, prisoners and/or homo-or bisexuals in their sample in the absence of subgroup analyses for these groups. This approach limited the number of included studies. A more liberal approach to review study inclusion criteria (i.e. including the 67 studies that partly included these groups) might be useful in further addressing the objective of this review. However, caution should be had when doing so to avoid the impact of confounders on the findings and we would recommend conducting subgroup analyses in such case.

Data availability
All data underlying the results are available as part of the article and no additional source data are required.
Author contributions FW and DD designed the protocol with input from all authors. VL constructed and conducted the search. All authors reviewed the search. FW, VS, PF, SS, LM, GG conducted study selection and data extraction. FW and JCJ conducted the ROB assessment. FW and VS conducted the GRADE quality assessment. FW drafted the manuscript. All authors were involved in the data interpretation and reviewed the manuscript.

Competing interests
The National University of Ireland Galway received payment for the conduct of this research from the National Clinical Effectiveness Committee of the Department of Health (Ireland) who commissioned this work.
VL's institution received payment from the National University of Ireland Galway to conduct the search for this review.

Grant information
This work was supported by the Clinical Effectiveness Unit of the Department of Health (Ireland).
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.