Herpes simplex virus type 1 epidemiology in the Middle East and North Africa: systematic review, meta-analyses, and meta-regressions

This study aimed at characterizing herpes simplex virus type 1 (HSV-1) epidemiology in the Middle East and North Africa (MENA). HSV-1 records were systematically reviewed. Findings were reported following the PRISMA guidelines. Random-effects meta-analyses were implemented to estimate pooled mean HSV-1 seroprevalence. Random-effects meta-regressions were conducted to identify predictors of higher seroprevalence. Thirty-nine overall seroprevalence measures yielding 85 stratified measures were identified and included in the analyses. Pooled mean seroprevalence was 65.2% (95% CI: 53.6–76.1%) in children, and 91.5% (95% CI: 89.4–93.5%) in adults. By age group, seroprevalence was lowest at 60.5% (95% CI: 48.1–72.3%) in <10 years old, followed by 85.6% (95% CI: 80.5–90.1%) in 10–19 years old, 90.7% (95% CI: 84.7–95.5%) in 20–29 years old, and 94.3% (95% CI: 89.5–97.9%) in ≥30 years old. Age was the strongest predictor of seroprevalence explaining 44.3% of the variation. Assay type, sex, population type, year of data collection, year of publication, sample size, and sampling method were not significantly associated with seroprevalence. The a priori considered factors explained 48.6% of the variation in seroprevalence. HSV-1 seroprevalence persists at high levels in MENA with most infections acquired in childhood. There is no evidence for declines in seroprevalence despite improving socio-economic conditions.

Herpes simplex virus type 1 (HSV-1) is a widespread and incurable infection 1,2 . Although this infection is usually asymptomatic 3 , the virus is shed frequently and subclinically 4,5 . Clinically-apparent HSV-1 infection most often manifests as orolabial herpes lesions 6,7 , but the virus causes a diverse spectrum of diseases including neonatal herpes, corneal blindness, herpetic whitlow, meningitis, encephalitis, and genital herpes 7,8 . The infection's clinical manifestations depend on the virus' initial acquisition portal 6,7 -oral-to-oral transmission leads to an oral infection 6,7 , and oral-to-genital transmission (through oral sex) leads to a genital infection 6,9,10 .
HSV-1 is endemic globally as indicated by the high HSV-1 antibody prevalence (seroprevalence) across regions 2,11,12 . Although HSV-1 is typically acquired in childhood 8 , changes in hygiene and socio-economic conditions appear to have reduced exposure during childhood in Western 11,[13][14][15][16][17][18][19][20] and Asian countries 21 . A large fraction of youth in these countries reach sexual debut with no protective antibodies against HSV-1 infection, and thus at risk of acquiring the infection genitally 6,22 . A growing evidence indicates that HSV-1 is overtaking HSV-2 as the leading cause of first episode genital herpes in Western 6,22-26 and (apparently) Asian countries 21 . The extent to which such a transition in HSV-1 epidemiology is occurring in other global regions remains unknown.

study selection and inclusion and exclusion criteria. Search results were imported into Endnote,
where duplicate records were removed. Titles and abstracts of remaining records were screened independently by SC, MH, and HC, for relevance. Full texts of records deemed relevant or potentially relevant were retrieved for further screening. Bibliographies of relevant records and reviews were also screened for possible missing publications.
The inclusion criteria included any record reporting an HSV-1 seroprevalence measure, based on primary data and type-specific diagnostic assay such as glycoprotein-G-based enzyme-linked immunosorbent assays (ELISA).
The inclusion criteria also included any record reporting a proportion of HSV-1 viral detection in clinically-diagnosed GUD or in clinically-diagnosed genital herpes. The minimum sample size of included studies was 10, regardless of the outcome measure.
The exclusion criteria included case reports, case series, reviews, editorials, letters to editors, commentaries, qualitative studies, and animal studies. HSV-1 seroprevalence measures reported in <3 months-old infants were excluded since they may reflect maternal antibodies.
In this work, a "record" refers to a document (a publication) reporting an outcome measure of interest, while a "study" refers to the details pertaining to a specific outcome measure. Accordingly, one record may contribute multiple studies, and multiple records of the same study are considered as duplicates and only included once.
Data extraction and data synthesis. The extracted information included: author(s), publication title, year(s) of data collection, publication year, country of origin, country of survey, city, study site, study design, study sampling procedure, study population and its characteristics (e.g., sex and age), sample size, HSV-1 outcome measures, and diagnostic assay. Data were double extracted from relevant records by SC, MH, and HC.
Extracted outcome measures were based on their stratification in the original record. Stratifications of seroprevalence measures were considered using a pre-defined sequential order that prioritizes first population type, followed by age bracket, and then age group. Age bracket included children (<15 years of age) and adults (≥15 years of age). Age groups included <10, 10-19, 20-29, and ≥30 years of age-a stratification informed by the actual available data of age-strata.
The extracted seroprevalence data were synthesized by population type according to the following definitions: 1. Healthy general populations encompassing groups of presumably healthy persons (for example, pregnant women or blood donors) and outpatients attending a healthcare facility for an inconsequential health condition. 2. Clinical populations encompassing any population with a serious clinical condition, or with a condition potentially related to a clinical manifestation of HSV-1 infection. 3. Other populations encompassing populations not fitting the above definitions, or populations with an unclear risk of having acquired HSV-1, such as sex workers and mixed health-status populations.

Meta-analyses.
Random-effects meta-analyses were conducted to estimate the pooled mean HSV-1 seroprevalence in MENA by population type, age bracket, and age group. Pooled means were calculated using DerSimonian-Laird random-effects models 31  Cochran's Q statistic was calculated to test for heterogeneity in the pooled seroprevalence measures 33,34 . I 2 measure was calculated to assess the magnitude of between-study variation that is due to true variation in seroprevalence across studies rather than chance 33 . Prediction interval was estimated to characterize the heterogeneity in the seroprevalence measures 33 .
Sensitivity analyses were conducted using generalized linear mixed models (GLMM) 35 . The results were used to confirm the pooled mean HSV-1 seroprevalence estimates generated based on the Freeman-Tukey type arcsine square-root transformation, given a recently-identified potential pathology in this transformation 35 .
Meta-analyses were performed in R version 3.4.1 36 using the meta package 37 .

Meta-regressions.
Univariable and multivariable random-effects meta-regression analyses, using log-transformed proportions, were conducted to identify associations and predictors of higher HSV-1 seroprevalence and sources of between-study heterogeneity. Associations were described using relative risks (RRs), 95% confidence intervals (CIs), and p-values. Potential predictors were specified a priori and included: age bracket, age group, assay type, country's income, population type, sample size (<100 versus ≥100), sampling method (probability-based sampling versus non-probability-based sampling), year of data collection, and year of publication. Factors with p-value < 0.1 in univariable analysis were eligible for inclusion in the multivariable model. Factors with p-value < 0.05 in the multivariable analysis were considered as statistically significant predictors.
Assay type consisted of five assay types for which data were available: ELISA, enzyme immunoassay (EIA), immunofluorescence assay (IFA), neutralizing antibody assay (Nab), and western blot. Of note, different assays used different cut-off points. For example, for HerpeSelect® 1 ELISA, sera with optical density index values ≥ 1.10 were considered seropositive and <0.90 seronegative, with the rest deemed equivocal 38,39 . Meanwhile, for Euroimmun Anti-HSV-1 ELISA, sera with optical density index values ≥ 1.10 were considered seropositive and <0.80 seronegative, with the rest deemed equivocal 39,40 .
Country's income was determined based on the World Bank classification 41 for the countries for which HSV-1 seroprevalence data were available: lower-middle-income countries (Egypt, Jordan, Morocco, Pakistan, Palestine, Sudan, Syria, and Yemen), upper-middle-income countries (Algeria, Iran, Iraq, and Lebanon), high-income countries (Qatar and Saudi Arabia), and mixed for samples including specimens from different countries.
Missing values in the year of data collection variable were imputed using the median of the values generated (for studies with data) for the difference between the year of data collection and the year of publication.
Meta-regressions were conducted in Stata/SE version 13 42 using the package metareg 43 .
Quality assessment. There are documented issues with the sensitivity and specificity of HSV-1 diagnostic methods 44,45 . Therefore, an expert advisor, Professor Rhoda Ashley Morrow from the University of Washington, was consulted and assessed the quality of each diagnostic method in each identified relevant study. Only studies with sufficiently reliable and valid assays were included. Further quality assessment of included studies was conducted as informed by the Cochrane approach for risk of bias (ROB) 29 and precision assessment. Studies' assessment into low versus high ROB was based on two quality domains: sampling methodology (probability-based versus non-probability-based sampling), and response rate (≥80% versus <80%). For instance, if probability-based sampling was used in a given study, the study was classified with a low ROB for that domain. Studies with missing information for any of the domains were classified as having unclear ROB for that specific domain.
Studies were considered as having high (versus low) precision if the number of HSV-1 tested individuals was at least 100 participants. For an HSV-1 seroprevalence of 80% and a sample size of 100, the 95% CI is 70.8-87.3%-a reasonable 95% CI estimate for an HSV-1 seroprevalence measure.

Results
search results and scope of evidence. Figure 1 shows the process of study selection based on the PRISMA guidelines 30 . A total of 1,552 citations were retrieved (269 through PubMed, 537 through Embase, and 746 through national and regional databases). After duplicates' removal and titles' and abstracts' screening, 130 records were identified as relevant or potentially relevant. Three additional records were identified through screening the bibliography of a previously published review for Iran 46 .
After full text screening, 15 records reporting an HSV-1 seroprevalence in 14 out of the 23 MENA countries were deemed relevant. Thirty-nine HSV-1 seroprevalence measures were extracted yielding 85 stratified measures. No HSV-1 seroprevalence measures (fulfilling the inclusion criteria) were identified among clinical children populations.
Although we searched for records that reported the proportion of GUD or genital herpes attributable to HSV-1, no such records were identified. Table 1 summarizes the included HSV-1 seroprevalence measures.

HSV-1 seroprevalence overview.
Studies included were published starting the year 1986, with the majority being cross-sectional in design and based on convenience sampling methods.
Evidence of heterogeneity in seroprevalence was present in nearly all meta-analyses (p < 0.0001; Table 2). The I² measure indicated that most variation was attributed to true variability in seroprevalence across studies. The prediction intervals confirmed the considerable variation in seroprevalence across studies.
Forest plots of meta-analyses can be found in Supplementary Fig. S1.
predictors of seroprevalence and sources of between-study heterogeneity. Table 3 summarizes the results of the univariable and multivariable meta-regression models. In the univariable analyses, age bracket, age group, country's income, population type, and sampling method had a p-value < 0.1 and were included in the multivariable analyses. Age bracket alone explained 44.3% of the variation in seroprevalence, followed by age group at 28.7%. Each of assay type, sample size, sex, year of data collection, and year of publication was not significantly associated with HSV-1 seroprevalence.
To account for the fact that age bracket and age group both measure age, two final multivariable models were conducted. The first model included age bracket, country's income, population type, and sampling method. This model explained 48.6% of seroprevalence variation. HSV-1 seroprevalence in adults was 1.3-fold (95% CI: 1.2-1.5) higher than in children. Seroprevalence in upper-middle-income countries and high-income countries was, in both, 0.9-fold (95% CI: 0.8-1.0) lower than in lower-middle-income countries. No association with population type and sampling method was found.
The second model included age group, assay type, country's income, population type, and sampling method. The model explained 40.2% of seroprevalence variation, with similar results for country's income, population  Quality assessment. Out of 32 records that included seroprevalence measures, only 15 were included in the systematic review with the remaining 17 being excluded due to potential issues in the validity of the diagnostic method, such as potential cross-reactivity with HSV-2 antibodies (Fig. 1). Of the studies included, 64.1% had high precision, 7.7% had low ROB for the sampling methodology domain, and 48.7% had low ROB for the response rate domain. These results, in context of the meta-regression models results, with different factors including sample size, sampling method, and assay type not being predictors of HSV-1 seroprevalence, suggest that overall the studies had reasonable quality.

Author, year
Year(s) of data collection The detailed quality assessment of included studies can be found in Supplementary Table S3.

Discussion
HSV-1 epidemiology in MENA was investigated through a comprehensive systematic review and meta-analytics of existing evidence. HSV-1 seroprevalence was found at high level, suggesting considerable HSV-1-related morbidity that is yet to be quantified and tackled. Sixty-five percent of children and 90% of adults were found seropositive-seroprevalence increased rapidly with age at younger ages, and was consistent with most infections being acquired in childhood.
Remarkably, about half of the observed variation in seroprevalence was explained by factors set a priori and examined in this study. Age alone explained 44.3% of the variation. Despite improvements in socio-economic conditions and earlier speculation that seroprevalence levels may have been declining in MENA 47 , we did not find evidence for a declining trend over the last two decades. We also did not find evidence for variation in seroprevalence by sex, population type (healthy versus clinical), or study characteristics including assay type, sampling method, and sample size.
Though there was no evidence for recent declines in seroprevalence, youth had considerably lower HSV-1 seroprevalence than older subjects. As much as one-third of youth in MENA may be reaching sexual debut uninfected and thus potentially at risk of sexual acquisition, in context of recent evidence from Western countries and Asia reporting an increase in incidence of genital herpes attributed to HSV-1 rather than HSV-2 6,21-26 . We did not, nonetheless, identify any evidence for a potential role for HSV-1 sexual transmission in MENA. Despite the extensive search in multiple international, regional, and national databases, we failed to identify a single study that assessed the etiological role of HSV-1 in GUD or genital herpes in this region.
A comparison of the findings of the present study with that of a recent systematic review of HSV-1 in Asia 21 demonstrates key insights about what may be general (or somewhat general) patterns in the global epidemiology of HSV-1 infection. Age in both systematic reviews was by far the strongest predictor of HSV-1 seroprevalence. Remarkably, in both MENA and Asia, seroprevalence in children was assessed at about 60%, and seroprevalence among adults was about 30% higher than that in children. Country's income was also a predictor  of seroprevalence with higher income associated with lower seroprevalence, attesting to an apparently global association between HSV-1 infection and socio-economic status 11,48 . However, the association of HSV-1 and socio-economic status differed between the two regions. In MENA, lower-middle-income countries had the highest seroprevalence, whereas in Asia, upper-middle-income countries had the highest seroprevalence. Possibly, the rapid modernization of Asia compared to MENA may contribute to explaining this difference. In both MENA and Asia, sex, population type, assay type, sample size, and sampling method were not associated with HSV-1 seroprevalence. This suggests that HSV-1 is a truly general population infection that permeates all strata of society-there is no difficulty in sampling a representative sample provided the age distribution is representative. In both regions also, no evidence for a temporal trend in seroprevalence was identified despite the evidence for temporal declines in seroprevalence in Western countries 11,[13][14][15][16][17][18][19][20] . Notably in MENA, half of the variation in seroprevalence (49%) was explained by the a priori considered factors, but only 26% of the variation was explained in Asia by the same considered factors 21 . The latter finding may relate to a higher population heterogeneity across countries in Asia than in MENA.
Our review and meta-analytics are limited by the quantity, quality, and representativeness of included studies. No data were identified for nine (mostly non-populous) of the 23 MENA countries, thereby potentially affecting the generalizability of the analyses to all of MENA. The number of seroprevalence measures varied from each study to another-only one (large) study, for example, contributed 29% of all stratified seroprevalence measures 47 . The majority of studies used convenience sampling (as opposed to probability-based sampling) of opportunistic   Sample size denotes the sample size of the study population found in the original publication. Abbreviations: ARR = Adjusted relative risk, CI = Confidence interval, EIA = Enzyme immunoassay, ELISA = Enzymelinked immunosorbent type-specific assay, HIC = High-income country, IFA = Immunofluorescence assay, LMIC = Lower-middle-income country, Nab = Neutralizing antibody assay, RR = Relative risk, UMIC = Uppermiddle-income country.
Scientific RepoRts | (2019) 9:1136 | https://doi.org/10.1038/s41598-018-37833-8 populations such as blood donors or outpatients ( Table 1). The latter, though, may not have been a limitation in context of the findings of the meta-regression analyses (Table 3). Studies used different diagnostic methods, and such methods may differ in sensitivity and specificity 44,45 . Presence of HSV-2 antibodies may also affect diagnostic methods differentially, particularly the classic "relative-reactivity" methods such as IFA and Nab [49][50][51] . This limitation, however, may not have affected our results, as HSV-2 infection has a low seroprevalence in MENA 52 , and earlier work suggests minimal impact of this limitation on specifically HSV-1 seroprevalence (as opposed to HSV-2 seroprevalence) [49][50][51] . The meta-regression analyses found no variations in HSV-1 seroprevalence across assay types (Table 3).
There was extensive heterogeneity in HSV-1 seroprevalence measures, but half of this heterogeneity was subsequently explained by only two factors, age and country's income (Table 3). Lastly, no study of HSV-1 viral detection in GUD or in genital herpes in MENA was identified, thus limiting our ability to assess the epidemiological role of HSV-1 sexual transmission. In spite of these limitations, our study is the first to draw a comprehensive synthesis and analytics of HSV-1 seroprevalence for the MENA region, and to highlight opportunities for related research and public health response.

Conclusions
HSV-1 seroprevalence in MENA indicated that 65% of children and 90% of adults had been exposed to this infection, by inference, most often during childhood. Age and country's income were the strongest predictors of HSV-1 seroprevalence and explained half of seroprevalence variation. No evidence was found for a temporal trend in seroprevalence over the last two decades despite improvements in socio-economic conditions. With no identified study of HSV-1 viral detection in GUD or in genital herpes, the role of HSV-1 sexual transmission in MENA remains unknown. This lack of data calls for at least basic or opportunistic GUD/genital herpes etiological surveillance. The totality of the findings highlights the timeliness of accelerating HSV-1 vaccine development to control one of the most endemic infections worldwide.