Dengue in the Middle East and North Africa: A Systematic Review

Background Dengue virus (DENV) infection is widespread and its disease burden has increased in past decades. However, little is known about the epidemiology of dengue in the Middle East and North Africa (MENA). Methodology / Principal Findings Following Cochrane Collaboration guidelines and reporting our findings following PRISMA guidelines, we systematically reviewed available records across MENA describing dengue occurrence in humans (prevalence studies, incidence studies, and outbreak reports), occurrence of suitable vectors (Aedes aegypti and Aedes albopictus), and DENV vector infection rates. We identified 105 human prevalence measures in 13 of 24 MENA countries; 81 outbreaks reported from 9 countries from 1941–2015; and reports of Ae. aegypti and/or Ae. albopictus occurrence in 15 countries. The majority of seroprevalence studies were reported from the Red Sea region and Pakistan, with multiple studies indicating >20% DENV seroprevalence in general populations (median 25%, range 0–62%) in these subregions. Fifty percent of these studies were conducted prior to 1990. Multiple studies utilized assays susceptible to serologic cross-reactions and 5% of seroprevalence studies utilized viral neutralization testing. There was considerable heterogeneity in study design and outbreak reporting, as well as variability in subregional study coverage, study populations, and laboratory methods used for diagnosis. Conclusions / Significance DENV seroprevalence in the MENA is high among some populations in the Red Sea region and Pakistan, while recent outbreaks in these subregions suggest increasing incidence of DENV which may be driven by a variety of ecologic and social factors. However, there is insufficient study coverage to draw conclusions about Aedes or DENV presence in multiple MENA countries. These findings illustrate the epidemiology of DENV in the MENA while revealing priorities for DENV surveillance and Aedes control.


Introduction
Dengue virus (DENV) is a globally distributed flavivirus with nearly 400 million estimated annual infections and a growing geographic distribution and disease burden [1][2][3].DENV has a historic presence in the Middle East and North Africa (MENA), with outbreaks of dengue and dengue-like disease reported across much of the Eastern Mediterranean region in the 19 th and early 20 th centuries [4,5].Today, DENV may be resurging in the MENA [6,7], with recent outbreaks of unprecedented or previously unrecognized magnitude occurring in the Arabian Peninsula and Pakistan [8,9], and a 2015 outbreak in Egypt that occurred following a decadeslong absence of reported cases from that country [10].Still, despite increasing global concern about the threat of Aedes-transmitted arboviruses, the epidemiology of DENV in the MENA region remains largely uncharacterized.
Understanding the epidemiology of DENV in the MENA represents an ongoing challenge for multiple reasons [11].Inadequate human and vector surveillance, non-reporting of illness syndromes, and poor diagnostic capacity limit DENV detection in many countries, resulting in delays in outbreak recognition and sparse data with which to estimate disease burden and infection rates [12][13][14].Case series, outbreak reports, and national notification reports, which contribute much to the epidemiologic knowledge of DENV, may also contain bias in reflecting only those areas with sufficient capacity to detect and report DENV when it occurs [1].Moreover, clinical diagnosis of DENV infection in the absence of laboratory confirmation is often unreliable [12,[15][16][17][18]. Cross-sectional serologic surveys for DENV exposure have the potential to shed light on the broader population burden of DENV without these biases.However, serologic cross-reactions among antibody-based assays for flaviviruses can limit the reliability of such studies in the absence of confirmatory testing, though the latter is difficult to perform and often unavailable [19,20].
To further the knowledge of the epidemiology of DENV in the MENA, we undertook a comprehensive summary and appraisal of published DENV prevalence, incidence, vector infection rates, reported outbreaks, and Aedes occurrence reports in the MENA region.This report aims to enhance the understanding of the epidemiology of DENV in the MENA while informing priorities for future research.

Materials and Methods Objectives
The objective of this study was to characterize the epidemiology of DENV in the MENA region through a systematic review of human prevalence and incidence studies and infection rates in Aedes mosquitoes.We also aimed to summarize reported human outbreaks and Ae.aegypti and Ae.albopictus occurrence in the region.The original search was last updated on December 9, 2015.

Eligibility criteria
Table 1 displays the eligibility criteria.In brief, studies containing primary prevalence, incidence, and vector infection rates for DENV in the MENA region were considered eligible for the systematic review.Publication year was not considered an inclusion criterion, as we reasoned that the historic distribution of DENV could be useful in understanding its current epidemiology by depicting ecologically viable regions in which DENV transmission continues to occur or could re-emerge.For incidence studies, those that reported the number of acute infections or seroconversions over any time interval were eligible.Vector infection rate studies were included if they contained a measure of the estimated proportion of infected Ae. aegypti or Ae.albopictus at a given time and setting in the MENA region.

Outcomes
For the systematic review, the primary outcomes were DENV human prevalence, incidence, and vector infection rates in the MENA region.Secondary outcomes were reports of dengue outbreaks and vector occurrence.

Data sources and search strategy
We conducted a systematic search for DENV in the MENA following Cochrane Collaboration guidelines [21] and reported our findings using the Preferred Reporting Items for Systematic reviews and Meta-analyses (PRISMA) guidelines [22].The PRISMA checklist is found in S1 Fig and our search criteria in S2 Fig.Briefly, we searched PubMed, Embase, the World Health Organization (WHO) Index Medicus for the Eastern Mediterranean Region and WHO African Index Medicus without publication date or language restrictions, using text and MeSH/ Emtree terms exploded to include all subheadings.Our review covered the 23 countries included in the MENA definitions of the WHO/EMRO, World Bank, and the Joint United Nations Programme on HIV/AIDS (UNAIDS) for consistency with earlier regional analyses of various infectious diseases including HIV [23].

Study selection
For each search, titles and abstracts were imported into Endnote (Thompson Reuters, Philadelphia, PA, USA), duplicates were removed, and were screened by one author (JH) with potential eligibility determined by consensus with a second author (NC) when eligibility was unclear.Full texts of potentially relevant records were retrieved and assessed for eligibility, contacting the author of the report as necessary.Reference lists of all potentially eligible articles and reviews were also searched.In this study, 'report' refers to the document (paper, abstract, or public health record) containing an outcome measure of interest, while 'study' refers to the outcome measure(s) within that report.Hence, reports could contribute more than one study, though multiple reports of the same study were counted only once.

Data Extraction and Synthesis
Data were extracted by one of the authors (JH) using a pre-piloted data extraction form and entered into a database created in Microsoft Access.Data from reports in English were extracted from the full texts, while reports in French (n = 6), Turkish (n = 3), Dutch (n = 1), and German (n = 1) were extracted from the abstracts and full texts with the help of online language software and French, Turkish, and German language speakers [24].There were no records in other languages.Studies were compiled by country and organized by year, using separate tables for human prevalence, incidence, and vector infection rates.Prevalence studies were further stratified as follows: 1) general prevalence studies measuring the prevalence of anti-DENV antibodies among populations without acute infection (e.g.DENV exposure); and 2) acute DENV infection studies assessing the prevalence of laboratory-confirmed DENV infection in those with a) undifferentiated acute febrile illness (AFI) and b) suspected DENV infection (Table 2).These stratifications were made because of the different study aims and probabilities of having laboratory evidence of DENV infection in each of these populations.Finally, the geographic distribution of all included prevalence studies were mapped according to the first-level administrative division (e.g.state, province) in which each study was conducted (Tableau Software, Seattle, WA, USA).

Risk of bias assessment
In order to gain a better understanding of the quality of prevalence studies identified through the systematic review, the risk of bias (ROB) was assessed for each study based on the

General prevalence
Seroprevalence studies reporting anti-DENV IgG prevalence measures among individuals not suspected to have acute dengue infection, including community members, blood donors, military, students, and hospitalized patients and outpatients receiving care for non-febrile illnesses.

Acute DENV infection
Undifferentiated acute febrile illness (AFI): studies for which acute dengue infection is not differentiated by clinical grounds alone; IgG prevalence measures obtained during the acute phase of illness is these studies are presumed to reflect secondary infection.
Suspected dengue infection: studies in which defined or undefined clinical criteria for probable dengue infection is stated as an inclusion criterion in the study. doi:10.1371/journal.pntd.0005194.t002 Cochrane approach [25] and by evaluating the precision of the reported measures.The methodology for this assessment is similar to that which we have previously developed for reviews of HIV and hepatitis C prevalence in the MENA region [26][27][28].Each DENV prevalence measure was considered to have a low, high, or unclear ROB in three domains: sampling methodology, DENV infection ascertainment, and response rate.The latter was defined as the number of tested individuals divided by the number of persons invited to participate in the study [29].ROB was considered low if (1) sampling was probability-based (i.e. using some form of random selection), (2) DENV prevalence measures included viral neutralization testing (VNT) for general prevalence studies or biological assays (i.e.cell culture, PCR, and NS1 ELISA) for acute infection studies, and (3) response rate was !80%.Studies with missing information for any of the domains were classified as having unclear ROB for that specific domain.Sampling strategy was not evaluated for acute infection studies because these studies enrolled individuals presenting to a health facility with acute infection, hence, no population-based sampling is needed to capture this population.Studies were considered to have high precision if the number of individuals tested was !100.We considered this to be a reasonably sensitive cutoff for precision given the heterogeneous epidemiology of DENV across the region (e.g. a prevalence of 1% entails a 95% CI of 0-3%).

DENV outbreaks and Aedes distribution
To supplement the epidemiologic data gathered through the systematic search, reported outbreaks and Ae.aegypti or Ae.albopictus occurrence in the MENA were also sought from the articles retrieved through the search databases as well as through ProMED-MENA and Google Scholar.Given that the characteristics and definitions of dengue outbreaks in the literature are implicitly variable and that there is currently no consensus on how to define such events [30], we broadly included any outbreak report if the author of the report defined the event as an outbreak.Multiple reports of the same outbreak were recorded only once.We manually marked the location of reported outbreaks on the map as well, designating one mark per each firstlevel administrative division in which one or more outbreaks were identified.In a separate map, we mapped the country-level occurrence of Ae. aegypti and Ae.albopictus in order to further inform the existing or potential epidemiology of DENV in the MENA.

Search results
The selection process based on PRISMA guidelines is illustrated in Fig 1 [22].Briefly, the DENV search yielded 1,258 citations, 91 of which were ultimately eligible for inclusion in the study following the addition of 4 reports identified from the bibliographies of relevant reports and reviews.Four studies from the 1970-80s were excluded that contained DENV seroprevalence of 0-11% in wild and domestic animals in Pakistan, Tunisia, and Turkey, though these may have represented cross-reactions with other flaviviruses [31][32][33][34].

Characteristics of included studies
A total of 105 human prevalence studies for DENV were identified from eligible reports (Table 3).These studies covered 13 of 24 MENA countries and were conducted from 1962-2015.The geographic distribution of these studies is illustrated in Fig 2 , and Table 4 contains a frequency summary of these studies.Anti-DENV antibodies were detected in 12 of 13 countries in which studies were reported with a single 1973 study from Libya reporting 0% seroprevalence [35].The highest number of studies were reported from Pakistan (n = 32) and Sudan (n = 16), most of which targeted populations with acute DENV infection (undifferentiated AFI or suspected dengue infection).Among general population studies, IgG prevalence measures ranged from 0% to 61% and were reported from Djibouti (n = 4, 0-21%), Egypt (n = 4, 0-7%), Iran (n = 3, 0-7%), Kuwait (n = 3, 0-56%), Lebanon (n = 3, 0-61%), Pakistan          (n = 3, 9-28%), Saudi Arabia (n = 4, 0-33%), and Sudan (n = 5, 9-49%).ELISAs were the most commonly used diagnostic method for all study types and the majority studies from the MENA used in-house assays (Table 4).VNT results were reported in 3% (n = 3) of all studies while observed or potential serologic cross-reactions with other flaviviruses were present in multiple studies [36][37][38] (Tables 3 and 4).Three human incidence measures for DENV were identified (Table 5): the first reported an ELISA IgM incidence of 35 cases per 10,000 people living in urban homes in Port Sudan City, Sudan where DENV-carrying mosquitoes were identified over an 11-month period [39]; the second reported an ELISA IgM incidence of 94 cases per 10,000 people in a general population in Port Sudan, Sudan over a 17-week period in 2010 [40]; the third reported an ELISA IgM, NS1 antigen, or PCR incidence of 185 cases per 100,000 febrile children in an urban slum in Karachi, Pakistan from 1999-2001 [41].Three vector infection rate studies for Ae.aegypti and Ae.albopictus were identified from Pakistan and Yemen [42,43] (Table 6).

Risk of bias assessment results
The quality assessment for each study is found in S1 Table and a summary of the precision and risk of bias assessment is found in Table 4.In brief, most studies (!65%) contained high precision as defined by a sample size of !100 participants.A minority (36%) of general population seroprevalence studies utilized some form of random sampling, and response rates were either <80% or not reported in 86% of general population studies.VNT or a biologic confirmatory assay (i.e.cell culture, PCR, and NS1 ELISA) was performed in 5% and 36% of general population seroprevalence and acute DENV infection studies, respectively, entailing low ROB for the assays used.

Dengue outbreaks and Aedes occurrence
Reported outbreaks of DENV in the region were gathered through citations collected from the search databases (S2 Table ) and mapped along with the geographic distribution of prevalence studies in Fig 2 .For DENV, 81 outbreaks were reported from 9 countries in the region from 1941-2015, including sentinel reports of autochthonous transmission in Egypt (2010) and Yemen (1983).Reports contained variable descriptions of outbreaks including 'estimated', 'suspected', 'reported', and/or laboratory 'confirmed' cases (S2 Table ).The definition that qualified each event as an outbreak was unclear in most instances.Outbreaks of DENV serotypes 1-3 were reported from countries surrounding the Red Sea and DENV-4 was only reported from Pakistan [120,121].Although, in general, DENV serotypes were not reported consistently.Published reports of Ae. aegypti and Ae.albopictus occurrence are recorded in S3 Table and mapped by country in Fig 3 .Ae. aegypti occurrence was reported in 11 MENA countries and historically (i.e.prior to 1960) in Algeria, Libya, Morocco, Syria, and Tunisia.Ae. albopictus was reported in seven MENA countries, including Algeria, Palestine, and Syria, countries where Ae. aegypti is not currently reported.No published reports of Ae. aegypti or Ae.albopictus occurrence (or DENV outbreaks) were identified in seven MENA countries: Bahrain, Iran, Iraq, Jordan, Kuwait, Qatar, and United Arab Emirates.Since 2005, Ae. aegypti and/or Ae. albopictus occurrence has been documented in Afghanistan, Algeria, Lebanon, Oman, Palestine, Syria, and Turkey, though autochthonous transmission of DENV has not yet been reported from these countries.

Discussion
Our study offers an assessment of published prevalence, incidence, and outbreak reports pertaining to the epidemiology of dengue in the MENA region.Based on the study results, the MENA contains two apparent subregions known to harbor DENV: 1) Pakistan, and 2) the Red Sea countries (Djibouti, Egypt, Saudi Arabia, Somalia, Sudan, and Yemen).No seroprevalence or outbreak data was identified across broad areas of the MENA, however, including some Aedes endemic areas.There was also a paucity of reports estimating human incidence and vector infection rates.These findings suggest priorities for future research.However, they also

Dengue seroprevalence in the MENA
In our review, Pakistan reported the highest number of prevalence studies and the broadest study coverage among MENA countries.Multiple studies reported >20% prevalence in both general population and those with undifferentiated AFI [1,106,[122][123][124].DENV serotypes 1-4 are known to circulate in Pakistan, unlike other MENA countries [82].Pakistan also reported the largest number of confirmed cases among all DENV outbreaks in the MENA, with 21,580 cases reported during the 2011 DENV-2 outbreak [79,120] (S2 Table ).
In the Red Sea region, multiple general population and AFI population IgG seroprevalence measures exceeding 20% were published from in Djibouti, Saudi Arabia, Somalia, Sudan, and Yemen within the past decade (Table 3) along with multiple confirmed outbreaks of DENV serotypes 1-3 since the 1980s (S2 Table ).DENV-4 has not yet been identified in this subregion to our knowledge.Although reported outbreaks and cases often localize along the Red Sea coastline in these countries [1], seroprevalence studies suggest a broader distribution of DENV infections that are likely underdetected (Fig 2).This is illustrated by the sentinel report of a DENV-infected traveler returning from Yemen in 1983 [125], despite the first outbreaks of DENV in Yemen and Saudi Arabia not being reported until 1994 [126][127][128].Our search also identified no published prevalence studies or outbreaks in Egypt after 1969 until a dengue outbreak was reported in November 2015 [10].However, DENV transmission was suggested years prior by a report of two travelers diagnosed with dengue after returning from southern Egypt in 2011 [129] and the identification of Ae. aegypti in southern Egypt that same year [130].It is plausible that undetected DENV transmission had been occurring in Egypt prior to this outbreak.However, it is not clear whether this and other recent outbreaks represent increasing incidence, increasing detection, or both, amidst the heterogeneity in study coverage and reporting in the MENA.

Clinical and methodological diversity among studies
An important finding in our study was the clinical and methodological diversity among DENV prevalence studies.This diversity represents a challenge to synthesizing the epidemiologic literature for DENV in the MENA.Clinically, studies represented a diversity of human populations of different ages and demographics, in different years, and different locations and transmission contexts.Ninety-six percent of studies from Afghanistan, Pakistan, Saudi Arabia, Somalia, and Yemen were conducted during or prior to 1990.However, 53% of studies in other MENA countries were conducted prior to 1990, when study methods and DENV epidemiology may have been different.Methodologically, most studies utilized convenience samples without reporting response rates, entailing high risk of bias and uncertainty in the representativeness of reported measures (Table 4).These findings, along with the high variability in regional study coverage, precluded meta-analyses of the available data.

Flavivirus cross-reactivity
Serologic cross-reactions remain a challenge to seroepidemiologic studies for DENV and other flaviviruses.Viral neutralization tests, considered the gold standard serologic assay for DENV, were performed in only 5% of general population seroprevalence studies in our review.Compared to ELISAs, seroprevalence measures were 22-86% lower by secondary/confirmatory testing with immunofluorescence or VNT in our review [14,36,37,113].This illustrates the potential uncertainty surrounding the reliability of ELISAs in DENV serologic studies, particularly in areas where the prevalence of antigenically similar viruses is broad or unknown.West Nile virus (WNV), for example, is thought to be distributed across the MENA on account of its ubiquitous Culex spp.vector and migratory bird flyways [131][132][133].With up to 80% of WNV infections occurring subclinically, the potential for serologic cross-reactions with DENV antibody assays must be considered.Yellow fever vaccine-derived and natural antibodies may also cross-react with anti-DENV antibodies, especially relevant in YFV endemic regions such as Sudan [38,107,111].As the emergence of zika virus in the Western Hemisphere or the re-emergence of YFV has shown, serologic assays with low specificity are inadequate to tackle the epidemiologic challenges of emerging arboviral diseases [134].

Heterogeneity in dengue outbreak reports
Our review identified DENV outbreaks in over a third of MENA countries, with most outbreaks reported from Pakistan, Sudan, and Saudi Arabia (S2 Table ).Outbreaks varied widely across time and space in the MENA: reported cases varied from <10 to >100,000 over a span of months to years, reported from the village level to the level of the province and region.This presents a challenge to epidemiologic monitoring and policy planning for DENV, as use of different outbreak definitions results in differences in early detection and response [30].There is currently no consensus on how to define DENV outbreaks, and adopting a common definition for the MENA is challenging given the region's heterogeneous infection pressures, multiple DENV serotypes, and variable surveillance and detection capacity.At present, assessing whether a reported transmission event in the MENA significantly deviates from baseline transmission, and thus constitutes an outbreak, is often unclear.

Risk factors and research priorities
Our study did not identify confirmed DENV transmission in any of the MENA countries west of Egypt and east of Saudi Arabia until Pakistan (Fig 2).However, the paucity of published data in these sub-regions does not preclude the possibility of unrecognized transmission in some areas or the risk of emergence in others.Indeed, modeling studies suggest ecologic niches for Aedes along the coastal Mediterranean Basin of North Africa [1,123,135], and Ae.albopictus and/or Ae. aegypti has been recently reported in Algeria, Lebanon, Palestine, Syria, and Turkey [5,[136][137][138][139][140] (Fig 3 and S3 Table).In contrast, Ae. albopictus has been identified along the Mediterranean coast of Europe for decades along with local transmission of DENV and chikungunya since 2007 [141].Near the Pakistan border, serologic evidence suggests possible DENV transmission in Iran [14,51,53] and Afghanistan [44], though local transmission has not been confirmed to our knowledge [53].The presence of Aedes or DENV transmission in these areas should not be ruled out [53].
Several ecologic and social factors in the MENA may promote the spread of Aedes-borne viruses like DENV.Urbanization [142] may increase the risk of outbreaks and use of open water storage containers that promote Ae. aegypti breeding [1,39,43,55,95,111,123,135].Heavy rainfall has been implicated in DENV outbreaks in Sudan, Djibouti, and Yemen [12,47,143,144], which may become increasingly unpredictable through climate change [145].Armed conflicts and economic turmoil in Iraq, Syria, and Yemen may render these areas vulnerable to vector-borne diseases while diminishing surveillance and response [146].Inter-regional migration poses risk for imported DENV, as millions of migrants travel from DENV-endemic countries to the Arabian Peninsula [111,126,[146][147][148][149] and to Mecca, Saudi Arabia to attend Umra and Hajj [126].Intraregionally, heavy commerce in the Red Sea region likely drives DENV serotype mixing and spread [116,148], as evidenced by multiple DENV outbreaks occurring at port cities in Djibouti [37,45], Saudi Arabia [126], Sudan [39], and Yemen [116,148].Contiguous spread of DENV from Yemen to Oman [150], or from Pakistan to Iran or Afghanistan [14], may also pose risk.
A number of research priorities emerge concerning the epidemiology of DENV in the MENA.First, broader seroepidemiologic coverage in the region is needed.Such studies are efficient means of characterizing infection pressures in populations lacking surveillance and diagnostic capacity.Multiplexed diagnostics are increasingly available and are well-suited for concurrently exploring the distribution other undercharacterized arboviruses in the region (e.g.Alkhumra, Chikungunya, Crimean-Congo Hemorrhagic Fever, O'Nyong-nyong, Rift Valley Fever, Sandfly Fever virus complex, Usutu, and West Nile viruses).Second, serologic studies should include methods to minimize cross-reactions, particularly for flaviviruses [151].Third, seroepidemiologic studies should incorporate uniformity in study design and enrollment criteria to minimize confounding, such as standard case definitions for studies of 'suspected' dengue [126].Ideally this could include population-based sampling that provides baseline data to benchmark the regional impact of these pathogens over the coming years.Fourth, studies should incorporate vector surveillance and infection rates.Such studies are important for understanding transmission dynamics that inform vector control strategies and predict future transmission and disease risk [123,135,141].Guidelines and tools for calculating vector infection rates are available [141,152].Finally, attaining a meaningful definition of DENV outbreaks in the MENA countries will require a thorough assessment of baseline surveillance, control, and treatment capacities in endemic regions [30].

Study limitations
Our study was limited by its reliance on select databases of peer-reviewed literature screened by one investigator with the exclusion of grey literature which may have provided additional data.Reviewing other Aedes-transmitted pathogens or studies reporting Aedes distribution in the MENA may also have provided further insights regarding the potential geographic distribution of DENV.Due to the limitations in the content and distribution of studies, we did not perform a meta-analysis nor did we explore bias in overall outcome measures through a funnel plot or Egger test.Non-publication of studies with small or zero effect size or studies targeted to known dengue-endemic areas may have biased the distribution and quantity of DENV studies.The prevalence measures themselves may have been biased through serologic-cross reactions, targeting of older study populations (with higher seroprevalence), and lack of convalescent titers for acute DENV infection studies (possibly underestimating seroprevalence).

Conclusions
DENV seroprevalence in the MENA is high among some populations in the Red Sea region and Pakistan, while recent outbreaks in these subregions suggest increasing DENV incidence driven by ecologic and social factors.Published prevalence and incidence, vector occurrence, and vector infection rates are lacking in broad areas of the MENA and available studies contain methodological limitations.These findings illustrate the need to strengthen programs for surveillance, reporting, and control of DENV and Aedes in the MENA, both to define DENV and Aedes epidemiology and to mitigate the risk of emerging Aedes-transmitted pathogens in the future.

Fig 1 .
Fig 1. PRISMA flow diagram of article selection.Flow diagram for dengue prevalence, incidence, and vector infection rates in the Middle East and North Africa.doi:10.1371/journal.pntd.0005194.g001

Table 1 . Criteria for study inclusion or exclusion.
Reported Ae. aegypti or Ae.albopictus infection rates by any laboratory method Basic science research studies, infection rates in other mosquito species or non-MENA country doi:10.1371/journal.pntd.0005194.t001