Systematic review of the rotavirus infection burden in the WHO-EMRO region

ABSTRACT Rotavirus gastroenteritis imposes a heavy burden on low- and middle-income countries. The World Health Organization defines the Eastern Mediterranean region (WHO-EMRO) as a diverse area in terms of socioeconomic status and health indicators. Rotavirus vaccination has been introduced, at least partially, in 19 out of the 22 EM countries; however, vaccine coverage remains low, and data on rotavirus disease burden is scarce. Available data on rotavirus prevalence, seasonality, vaccination status, and genotype evolution was systematically compiled following a literature review that identified 165 relevant WHO-EMRO epidemiology studies published between 1990 and 2017. Although the infectious agents responsible for acute gastroenteritis vary over time, rotavirus remained the leading cause of acute gastroenteritis in children, as seen in 76.3% of reviewed publications. Younger children (<2 years old) were at higher risk and thus increased vaccination coverage and surveillance systems are required to reduce the rotavirus gastroenteritis burden in WHO-EMRO countries.


Introduction
Rotavirus (RV) is the leading cause of diarrheal morbidity and mortality in young children worldwide and causes severe acute gastroenteritis (AGE) requiring hospitalization, and if dehydration is not treated in time leads to mortality especially in the developing world. 1 By the age of five years, nearly every child will have had an episode of RV gastroenteritis (RVGE), one in five of them will visit a clinic, one in 65 will be hospitalized, and approximately 1 out of 293 will eventually have a fatal outcome. 2,3 According to a report, more than 90% of RVGE deaths in 2013 occurred in 72 low-income and lowmiddle-income countries. 4 Implementation of an RV vaccination in national immunization programs (NIPs) reduces the RV disease burden substantially. 5 Countries need to have adequate knowledge and information on local burden, trends, and age distribution of disease to help decision-makers consider the introduction of an RV vaccine as part of their immunization programs. 6 The Eastern Mediterranean region, as defined by the World Health Organization (WHO-EMRO region, Figure 1), is a geographically and socioeconomic diverse area with varying health indicators. At present, the region has a population of over 656 million (81.4 million (12.4% of the total population)is under 5 years of age). 7 RV-associated mortality and morbidity also vary considerably in the region: the annual morbidity rates among children under five years of age ranged from 0 to 112/100,000 with an average mortality rate of 39/10,000 per year. 5,8 Lowincome countries (e.g. Afghanistan, Pakistan, Sudan, Yemen, and Somalia) had a higher mortality rate due to RVGE compared with countries where the per capita income was high (e.g. Saudi Arabia and Kuwait). However, the overall hospital and health center visits due to RVGE among children under five years of age was similar in both high-and low-income WHO-EMRO countries. 5,8,9 An RV vaccine has been introduced as part of the NIPs in 15 out of the 22 countries (albeit partially in Pakistan and Palestine) in the WHO-EMRO region. Vaccine coverage is predominantly suboptimal, with the exception of Saudi Arabia, with 97% vaccine coverage, and burden of disease data are still scarce. 1,6,10 Furthermore, a lack of surveillance systems prevents decision makers from understanding the magnitude of the problem and the need for RVGE prevention through vaccination. 11 The purpose of this literature review is to assess the burden of RVGE in the pediatric population and to summarise the current status of recommendations for RV vaccination across the WHO-EMRO region. Moreover, this review will also be useful as a baseline for post-vaccination surveillance for both AGE and RVGE.
Details on study subjects, dates, settings and calculated RV prevalence from the studies across the region is shown in Tables S1-S3. Table S1 presents data from the North African countries (Egypt, Morocco, Tunisia, Somalia, Sudan, and Libya), Table S2 includes the Middle Eastern countries (Saudi Arabia, Lebanon, Bahrain, Oman, Palestine, Qatar, UAE, Yemen, Jordan, Iraq, and Kuwait) and Table S3 shows data from the Asian countries (Iran, Pakistan, and Afghanistan). A total of 113,959 cases of confirmed RVGE are reported in this review, with the highest number from Pakistan (n = 13,546) and the lowest from Somalia (n = 213).  Figure 3) 66 There were differences in RV isolation rates among multiple studies conducted within the same country, e.g. Pakistan (8.2% to 65.9%), Saudi Arabia (3.9% to 65.5%) and Egypt (range: 11% to 76.9%). However, in other countries, prevalence appeared to vary less, as seen by a narrower distribution in Figure 3 for certain countries including Morocco (17.2% to 44%), Tunisia (6.1% to 33.9%), or Jordan (26.6% to 49.5%) There were also important differences in inpatient and outpatient RV prevalence. For example, in Egypt, in the 2011-2012 period, and using the same technique, the RV prevalence was 29.9% for outpatients and 43.9% for inpatients; 29 in Tunisia, this prevalence was 18.5% and 24.9%, 166 and in Bahrain 6% and 27.4%, respectively. 13 Changes in the detection of RVGE over time In Jordan, the RVGE detection rates among AGE cases were fairly constant over a ten-year period (25% in 1995, 35.2% in 2002, 24.7% in 2003, and 25.8% in 2004). 76 On the other hand, the proportion of RVGE cases appears to have increased over time in Pakistan; as the RVGE rates were ranging from 8.2% to 29% in the earlier years (1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996), 105,107,116 while in the more recent years (2007-2014) the reported rates were ranging from 23.8 to 65.9. [110][111][112][113][114][115]118,119 This can be due to the difference of sensibility of EIA used during these two different periods, or to the choice and characteristics of patients investigated.

Prevalence of RVGE
A reduction in RVGE occurrence is clearly evident after the introduction of RV vaccination in some countries. For instance, there was a 41.5% reduction in RV detection rate in the post-vaccination period compared to the prevaccination period in Morocco. 96 In Yemen there was a steady decline in RV detection following vaccine introduction; the detection rates during the pre-vaccination period was 43.8% and 37.4% in 2009 and 2010, respectively, but declined during the post-vaccination period to 17.9% in 2013 and to 10.5% in 2014. 174 In the same study, a 75.9% reduction in hospitalization due to RV diarrhea was reported over five years (2009: 43.8%; 2014: 10.5%).
Banajeh et al. 173 also showed that in Yemen, after vaccination had started, the prevalence of RVGE decreased by 48% and the hospitalization rate due to RV diarrhea declined from 42.9% during the pre-vaccination period (2007/2011) to 18.5% in the post-vaccination period (2013/2014). A similar pattern was observed in Saudi Arabia, where the RVGE prevalence was reduced from 12% to 46% to 9.2% over a three-year period following the introduction of the RV vaccine in 2013. 141

Concomitant enteropathogens
About one fourth (38 of 165) of the reviewed publications concomitantly reported isolation of pathogens other than RV.  The overall incidence was reported for studies detailing prevalence over several years, except when the study was comparing pre-and post-vaccination periods. In this case, the latest value was retained, as more actual.

RVGE seasonality patterns
A seasonality pattern of RVGE was reported in 95 of the 165 (57.6%) reviewed studies. Most studies (66/95, 69.5%) reported a peak of RVGE in the cold season from November to April. However, some studies also showed exceptional RVGE peaks occurring in late summer. 22,111,154,172 Studies from Iran, Bahrain, and Saudi Arabia did not observe any seasonal variation of RVGE. 13,39,132 Detection techniques used A total of 126 out of 147 studies that described RVGE prevalence also reported the isolation technique used for RV detection: 97 studies (77%) used enzyme immunoassay (EIA), 12 (9.5%) used latex agglutination (LA), 25 (19.8%) used polymerase chain reaction (PCR), 4 (3.2%) used polyacrylamide gel electrophoresis (PAGE) techniques, and a comparison with electron microscopy (EM) was made in 2 studies. Some of the studies also compared detection rates based on which techniques had been used. In Egypt, the positivity of RV detection varied between 34% and 39.5% according to LA and EIA techniques, respectively. 16 In Jordan, the sensitivity of EM and EIA RV detection techniques were 18.9% vs. 39.6%, respectively. 72 In Egypt, the sensitivity was higher when PCR (76.9%) was used as compared to EIA (67.7%). 30

Nosocomial RV infection
Nosocomial infection due to RV is also common. A study in Iraq showed that 32.4% of hospitalized children had nosocomial infections and 18.5% of them were due to RV. 41 The nosocomial infection rate was 33.3% in Morocco, 92 and in three different studies from Iran, the reported rates were 26.3%, 18.5% and 30%. 38,41,54 Disease severity and mortality rates In Morocco, 1 child per 32 RV-infected children was hospitalized and 1 child per 389 RV-infected children died due to RVGE in a group of children under 5 years of age. 91 Another Moroccan study reported that 30% of hospitalizations for diarrhea were due to RVGE, with a mortality rate of 40-50%. 90 In Pakistan, 1 child per 40 infected children

RV genotype and evolution of genotypes post-vaccine introduction
The genotype distribution of RV isolates in different studies was also reviewed. We evaluated the genotype combinations, where both G and P capsid protein combinations are reported. Studies conducted in 13 WHO-EMRO region countries (Bahrain, Egypt, Iran, Iraq, Jordan, Libya, Lebanon, Morocco, Oman, Pakistan, Saudi Arabia, Tunisia, and Yemen) reported RV genotype distribution (Table 3). Studies from Afghanistan, Palestine, Qatar, Somalia, and Sudan did not report on genotype distribution; Kuwait and UAE reported only G-or P-type, and were therefore not included in this evaluation. Overall, the genotype distribution was diverse across the region. G1P [8] was the most prevalent genotype combination in 7 of the 13 countries (i.e. Iran, Saudi Arabia, Jordan, Bahrain, Morocco, Iraq) for which data from the past 15 years are available (Table 3). 15,34,37,45,48,57,69,78,79,90,[93][94][95][96]98,[100][101][102][130][131][132]138,139,141 G1P [8] and G2P [4] were the most common genotypes in Yemen and Egypt. 19,23,24,27,29,32,83,84,171,172 In Lebanon, the genotype distribution varied over time and geographical location: the G4P [8] type disappeared from the northern part of the country in two out of the three studied seasons, while the G2P[4] genotype did not. 83 In Oman, the distribution pattern of the G2P [4] and G1P [8] RV genotypes also varied with time. [100][101][102] In Tunisia, both G1P [8] and G3P [8] were common, followed by G4P [8]. 148,158,159,[162][163][164]166 RV genotype G9 was first detected in 2004 in Tunisia and from 2010, there was an increase in the prevalence of related genotypes (10.3-37.5%). 165,166 In Pakistan, the distribution of genotype combinations was heterogeneous [110][111][112]117,118 and G12P [6] was found to be one of the most prevalent RV genotype in 2014. 115 The introduction of RV vaccines had a substantial effect on G1P [8] detection. In Saudi Arabia, a decrease was noted in G1P [8] detection from 51% to 37.1% after the vaccine was introduced, but G2P [4] increased from 21.6% to 33.3%. 141 In Morocco, a periodic decrease in the detection of G1P [8] genotype was reported: 57% one-year post-vaccination and 54.7% two years post-vaccination. 96 However, G2P4 was detected in 15% of samples in the pre-vaccination period and in 54% and 33% of samples, respectively, in the second and third years following the vaccine introduction. An increase was also seen for G9P [8] (7.8% in the pre-vaccination period, 16.6% and 67% in the second and third year of the post-vaccination period, respectively).
In Yemen, the G1P[8] detection rate increased from 45.5% to 87.5% and the G2P[4] detection rate decreased from 76.5% to 0.2% after vaccine implementation. 173 In another study in Yemen, the G1P[8] detection rate increased from 15% in the pre-vaccination period to 31% in the post-vaccination period and became the predominant genotype, followed by G9P [8] (27.5%). 174 In Iran, there was also an emergence of new RV genotypes over time, with G12P [8] being detected for the first time in 2016. 57 In Egypt, genotype G6P [14] was first reported in 2011. 26 Like in other parts of the world, mixed or non-typeable RV strains have also been detected in the WHO-EMRO region. In Iran, mixed genotype combinations, including G1P [4] and G1P [8], were detected in two studies, and the rate of mixed genotypes varied from 2.7% to 33.3%. 47,49 Non-typed (Nt) strain detection in Iran ranged from 5% to 40.9%. 34,68 In Pakistan, Egypt and Tunisia, the detection rates for mixed RV genotypes were 6.7%, 7% and 5.9%, respectively. 19,113,148

Vaccination status
The demographic characteristics and the type of RV vaccine used in the WHO-EMRO countries are shown in Table 4 (Table 4).

Discussion
This review evaluated the RV infection status in the WHO-EMRO region, identified reductions in RV prevalence and mortality following implementation of RV vaccination and described existing RV surveillance systems in the region.
In the WHO-EMRO region, the observed RVGE burden varied between countries. In general, a low RVGE prevalence (<30%) was observed in Egypt, Saudi Arabia, and Tunisia, and a high prevalence (>30%) was reported in Iran, Lebanon, Morocco, Jordan, and Oman. 8 The annual proportion of RVGE among reported episodes of AGE in the under-fiveyears-of-age population in this region was 42%, and recent studies have estimated that about 65,000 children die each year from RV infections in the 22 WHO-EMRO countries. 9 This is likely to be an underestimation of the true prevalence, due to inadequate surveillance systems and lack of routine RV testing. Although a majority (81.8%) of EMRO countries have implemented RV vaccination, seven countries (Egypt, Iran, Lebanon, Oman, Somalia, Syria, and Tunisia) have not yet implemented RV vaccination in their NIPs. Furthermore, no RV vaccine is available in three (13.6%) EMRO countries (Iran, Somalia, and Syria).
RV was the most common AGE agent in the WHO-EMRO region in 76.3% of the reviewed studies. Our review also revealed that there is a wide variation (>20% difference between studies) in RV isolation rates in different studies conducted within the same countries; for example, in Iran, 40 Pakistan, 105,114 Saudi Arabia 132,135 and Morocco, 89,95 the difference of reported isolation rates was wide. However, in countries such as Tunisia 156,163 and Jordan, 74,78 the rate differences were smaller. Differences in study design, patient selection, and case definition could contribute to explain differences in observed incidences across studies.
This review also confirmed that as in Western Europe, 179 G1P [8] and G2P [4] are the predominant genotypes circulating in the WHO-EMRO region and that available RV vaccines can provide protection against them. It is apparent that the circulating RV strain distribution regularly changes, as well as the dominant genotype, even in the absence of vaccines. Interestingly, G12P [8], a recently emerging serotype previously detected in Europe, Asia, and the Americas, has also been reported in some studies in Saudi Arabia. 8  Unusual genotypes such as G1P [6], G2P [6], G3P [9], G4P [6], G9P [6] and G9P [8] have been detected in different countries, 118,132,148,157 including the recent emergence of genotype G12P [6] in Pakistan 115 and of genotype G12P [8] in Iran. 60 Data of genotyping diversity from the WHO-EMRO region confirm that these findings are not specific to a particular geographic location but a natural phenomenon worldwide.
Mixed genotype combinations were detected in Iran, 49,60 Tunisia, 149,150 and Pakistan. 112 These results are similar to recent findings in India, Indonesia, and Vietnam and are likely due to reassortments. 10 The proportion of non-typeable and partially typeable genotype combinations also varied widely across countries. 6,48,49,100 Differences in laboratory techniques, as well as different geographic settings, can explain these findings. With the use of more appropriate genotyping primers and advanced molecular techniques, the proportion of non-typeable strains has decreased in recent studies.
A shift in genotype predominance and circulation before and after vaccine introduction is not unusual and cannot be attributed to available vaccines. Countries such as Australia and Brazil that have used different vaccines have also faced the re-emergence of G2 strains after vaccine introduction. 180,181 These data simply reflect the normal fluctuation in RV genotype frequency. In order to substantiate any vaccine pressure on genotype shift, continued surveillance over time, detailed phylogenic analysis, and full RV genome sequencing are required. Although these measures are currently not in place in the WHO-EMRO region, policy measures aimed at implementing these measures would greatly enhance our understanding of changes in RV strain circulation in the region.
In general, RV infections peak in the cooler winter months, and this pattern is similar for all WHO-EMRO countries except Gulf region countries. Studies from Saudi Arabia and neighboring countries showed no distinct peak, as RV disease was observed throughout the year with occasional peaks irrespective of season. 132 Several countries that have implemented routine childhood vaccination against RV have documented a dramatic impact on severe diarrhea and RVGE requiring hospitalization. [182][183][184][185] In the WHO-EMRO region, a small number of studies also show the effectiveness of vaccination on RVGE hospitalization, as well as RV detection in AGE cases. The results of these studies reflect that vaccination can markedly reduce RVGE disease burden, especially hospitalization due to diarrhea. 99,141,173,174 The WHO recommends sentinel surveillance before and after the introduction of vaccines to monitor vaccine impact. As financial resources may be limited in this region, regional governments should consider sharing their data with local laboratories or existing surveillance networks to help achieve this goal. Documentation of the burden of disease and of circulating strains is essential for decision-making on vaccine introduction and vaccine effectiveness assessments. The burden of RVGE and the circulation of different RV strains should be monitored in a surveillance system and through standardized methods, which would enable comparisons between countries.
We observed that no single standard technique was used to isolate and genotype RV from stool samples. This is likely to have led to a variation of estimates, not only in different countries but also across studies conducted within a country. Thus, we decided to present descriptive results instead of direct comparisons.
The various studies reported here included patients from different age strata in both in-and outpatient settings, which may potentially lead to misclassifications. It is easily conceivable that patients admitted as inpatients were suffering from more severe disease and were therefore more likely to yield a pathogen from their stool sample than outpatient stool samples originating from patients with less severe infections. This was evident from studies in Saudi Arabia, 127 where the RV detection rate was 5.9% in outpatient settings compared with 34.6% in inpatient settings; in Iran 66 the rates were 20.9% and 79.1% in outpatient and inpatient settings, respectively.
Variations in the estimates between and within countries could also be due to different study populations, study time periods, RV isolation and detection techniques, and the definition of RVGE used as study inclusion criteria. In some countries such as Afghanistan, Bahrain, Iraq, Kuwait, Lebanon, Palestine, Qatar, Somalia, and Sudan, the number of studies and subjects enrolled was not adequate for accurately estimating the burden and strain distribution of RVGE.
Although our review is one of the first to explore the prevalence of RV in the WHO-EMRO region, it has several limitations. Firstly, there is also a possible bias in identifying a denominator for the estimates. In some studies, RV rates were detected over total stool samples collected, while in others, the RV proportion was reported among the cases where at least one pathogen was detected. Secondly, our search was limited to published studies with abstracts in English and French, which may have resulted in the exclusion of studies published only in local languages. Thirdly, gaps in reporting from several WHO-EMRO countries may be due to political instability and not a lack of RV infections, and political instability may also prevent the successful implementation of vaccination programs and longitudinal RV vaccine studies.
In conclusion, this review of 165 studies from WHO-EMRO countries revealed that RV was strongly associated with young age in children and remained the dominant etiology of diarrhea requiring hospitalization, even in countries that had recently introduced RV vaccination. The collated data in this study will serve as useful baseline RV epidemiological data for the planning and implementation of vaccination programs in the region. Although the burden of RVGE is likely to decrease following the successful introduction of RV vaccination programs, as exemplified by Yemen, 171 ongoing surveillance is needed to determine the residual burden of RV, the genotype distribution of RV, and to monitor the evolving etiology of diarrhea in these countries. Data compiled in this report highlights that nearly all serious infections due to diarrhea occurred in children under 2 years of age, which stresses the need for protection early in life, and advocates for RV vaccine schedule completion by 12 months of age.

Methods
The inclusion criteria for the search was any publication in English or French in a country of the WHO-EMRO region on RV diarrhea in children less than five years of age during the period of January 1990 to December 5, 2017. We used the keywords "rotavirus" and "[country name]" (for each individual country out of the 22 included in the WHO-EMRO) to search in The National Library of Medicine's PubMed and grey literature (Figure 2). We also included publications in local medical journals and in local languages for which abstracts in English or in French were available. We further reviewed the cross-referenced articles from the retrieved ones.
A total of 426 published articles were identified, 387 from PubMed and 39 from local journals. Articles studying animals or environment (97), reporting the clinical aspects of infection studies (116), studying vaccine implementation or costeffectiveness (32), or using non-standard detection techniques (16) were excluded, and 165 articles were included for complete full-text review and qualitative assessment ( Figure 2).
Authors, year of publication, type of article, patient type and numbers were reported for each publication. The prevalence of RV infection mentioned in the articles was the main focus of this study, and we also retained articles mentioning RVGE mortality, or RV genotyping. In addition, and where available, data on nosocomial infections, seasonality, techniques used for the detection, RV genotype distribution, variation over time in the proportion of RVGE, and evolution of genotype distribution, presence of other enteric co-pathogens and the vaccination status of the country was also abstracted. Each datum was abstracted by one author, and crossed-check by another author. Next, we calculated the RVGE prevalence as the number of reported RVGE cases over the total number of samples included in each study, to use a standard measure of the RVGE prevalence. When possible (i.e. when three studies or more were retrieved for one country), the mean and median prevalence was given for each country. Data were stratified by country.

Trademark statement
Rotarix is a trademark of the GSK group of companies. Rotateq is a trademark of Merck and Co Inc.