Rotavirus A infection in pre- and post-vaccine period: Risk factors, genotypes distribution by vaccination status and age of children in Nampula Province, Northern Mozambique (2015-2019)

Mozambique introduced the monovalent rotavirus vaccine (Rotarix®, GSK Biologicals, Rixensart, Belgium) in September 2015. Previous analysis, showed that Nampula province continues reporting a high frequency of Rotavirus A (RVA) infection and the emergence of G9P[6], G9P[4] and G3P[4] genotypes. This analysis aimed to determine the RVA frequency; risk factors; genotype distribution by vaccination status and age between pre- and post-vaccine periods in children under-five years old with diarrhea in Nampula. A cross-sectional, hospital-based surveillance study was conducted in the Hospital Central de Nampula in Mozambique. Socio-demographic and clinical data were collected to assess factors related to RVA infection in both periods. Stool specimens were screened to detect RVA by ELISA, and positive samples were genotyped. Between 2015 (pre-vaccine period) and 2016–2019 (post-vaccine period), 614 stool specimens were collected and tested for RVA in which 34.9% (67/192) were positive in pre-vaccine period and 21.8% (92/422) in post-vaccine (p = 0.001). In the post-vaccine period, age, year, and contact with different animal species (chicken, duck, or multiple animals) were associated with RVA infection. RVA infection was higher in children partially vaccinated (40.7%, 11/27) followed by the fully vaccinated (29.3%, 56/191) and the unvaccinated (15.3%, 21/137) (p = 0.002). G1P[8] and G9P[4] were common in vaccinated children less than 12 months. The present analysis showed that RVA infection reduced slightly in the post-vaccine period, with a high proportion of infection and genotype diversity in children, under 12 months of age, vaccinated. Further research on factors associated with RVA infection on vaccinated compared to unvaccinated children and vaccination optimization should be done.

LG was supported by Fundação para Ciência e Tecnologia, Portugal (UID/04413/2020, UIDB/00006/2020, UIDP/00006/2020. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Study design, site, population
A cross-sectional, hospital-based surveillance was conducted between the pre-vaccine (March -December 2015) and post-vaccine (January 2016 -December 2019) periods. The surveillance was conducted at the Hospital Central de Nampula (HCN) in Nampula province, located in the northern region of Mozambique. Nampula province is highly populated, and has enormous challenges related to Water, Sanitation and Hygiene (WASH), high levels of poverty, and lower vaccination coverage [19,20]. Children aged up to 59 months who presented to the HCN pediatric services with diarrhea, defined as the passage of three or more loose or liquid stools in the last 24 hours before seeking healthcare, were surveyed [21]. Children were not included in this analysis if the caregivers did not consent to their participation as well as children diagnosed with nosocomial diarrhea. Moreover, children considered overweight (+2 < Z-score � +5) or with outlier Z-scores values were excluded from the nutritional status analysis [22].

Data collection
Data was collected by interviewing the child's caregiver and by accessing the child's medical records. The main variables include age (in months), categorized as 0-11, 12-23, 24-59; sex; exclusive breastfeeding; the age of weaning in months, categorized by six months; animal contact and type of animal; source of drinking water; HIV status; admission type; number of diarrhea episodes in the last 24h; vomit; fever; weight; vaccination status, and nutritional status that was estimated through the weight-for-age Z-score (WAZ), specific to underweight, using WHO software Anthro version 3.2.2. Weight was measured by lying down children younger than two years, and standing for children with two years or older. Children were classified as well-nourished if the Z-score was between -2 � Z-score � +2 and as underweighted if the zscores was between -6 � Z-score < -2. Vaccination status of the children was confirmed with a valid vaccination card. Trained healthcare professionals of ViNaDia, completed all information on a structured data collection form.

Sample collection
ViNaDia methodology has been previously described elsewhere [16,17,23,24]. Briefly, a single stool specimen (10 ml) was collected from each child in a polyester container and stored at -20˚C until shipment, once a week, to the INS laboratory in Maputo for diagnosis.

Laboratory procedures
Collected samples were tested for RVA using the commercial enzyme-immuno-sorbent assay (ELISA) kit (Prospect, Oxoid Ltd, United Kingdom) according to the manufacturer's instructions. All positive samples were submitted to total RNA extraction using the QIAamp Viral RNA protocol (QIAGEN, Hilden, Germany) following the manufacturer's instructions.
The PCR product was analysed using 2% agarose gel electrophoresis, stained with ethidium bromide, and visualized under ultraviolet illumination.

Sample size
Data used in the present analysis are from a hospital-based surveillance program (ViNaDia), whose primary aim is to monitor diarrhea and RVA cases before and after Rotarix 1 vaccine introduction [17]. Since ViNaDia is an ongoing surveillance system, the sample size calculation was not estimated for the present analysis. However, we calculated the statistical power associated to the obtained sample sizes to compare two proportions. For children aged up to 59 months, statistical power was 92.1%. For the secondary analysis in children younger than 24 months, the sample power was 88.0% (higher than 80% minimum required).

Data management and statistical analysis
Data collected was double entered in two similar databases by two independent typists in Epi Info™3.5.1. (Centers for Disease Control and Prevention, Atlanta, 2008), followed by data comparison. IBM SPSS (International Business Machines Corporation Statistical Package for the Social Science, Armok, NY: IBM Corp, 2011, version 26.0) software was used to conduct the data analysis.
Overall RVA frequencies with the corresponding Wilson's 95% Confidence Interval (CI) were estimated, for each RVA vaccine period and year [30]. Also, RVA proportions by age (in months) groups for each year were calculated.
Descriptive statistics and chi-square test or Fisher's Exact test were used to identify sociodemographic and epidemiological variables associated with RVA infection in each vaccine period in children age up to 59 months and in children younger than 24 months (S1 Table). Due to the absence of RVA cases among children aged between 24-59 months in the post-vaccine period, a more detailed analysis was done to children under 24 months. Multiple logistic regression models were explored to obtain adjusted odds ratio (aOR), with 95% confidence intervals, for each vaccine period. The bivariate analysis of the socio-demographic and epidemiological characteristics for children under 24 months of age is shown in S1 Table. Variables with initial p < 0.05 were included in the multiple logistic regression model for children under 24 months of age. Goodness-of-fit of the multiple logistic regression models was obtained using Hosmer and Lemeshow test.
For children aged up to 59 months, clinical factors associated with RVA infection in each vaccine period and in general were assessed by cross-tabulation followed by the chi-square test or Fisher's Exact test if assumptions were not met. Cross-tabulation was performed between vaccination status and RVA infection; vaccination status and genotype distribution; age and genotype distribution. Parameter estimation was done using 95% confidence interval, and hypothesis test were based on a 5% level.

Ethical statement
The study protocol was approved by the National Bioethics Committee for Health in Mozambique (IRB00002657, reference 348/CNBS/13). Parents or legal guardians of the children signed or provided their fingerprint to the informed consent form, which described the study objectives and clarified that participation in the study was voluntary. Data confidentiality was ensured by storing the physical data collection and consent forms in a lockable cabinet with limited access.

Overall and yearly RVA frequencies by vaccination period
In total, 700 children were recruited by ViNaDia in Nampula hospital (221 from the pre-vaccine period and 479 from the post-vaccine period), of which 614 (87.7%) provided a stool sample for laboratory procedures, (192 from the pre-vaccine period and 422 from the post-vaccine period). Overall
Similarly, to the pre-vaccine period, in the post-vaccine period, the child's age and weightfor-age were also associated to RVA infection, p < 0.001 and p = 0.044 respectively. Several factors that were not associated with RVA infection in the pre-vaccine period were associated with RVA infection in the post-vaccine period. These included animal contact (27.1%; 75/277; p < 0.001) and weaning within the first six months of life (27.3%; 3/11; p = 0.004) ( Table 1).

Type of animal which children had contact in pre-and post-vaccine periods
The animal species that children had contact in pre-and post-vaccine periods show that none of the animals were associated with RVA infection in pre-vaccine period (Table 2). However, it was observed an association between chicken contact (37.2%; 29/78; p <0.001); duck (37.0%; 17/46; p = 0.008), or multiple animals (30.6%; 34/111; p = 0.009) contact with RVA infection in the post-vaccine period ( Table 2).
In the multivariable analysis for the pre-vaccine period, children from household's that drunk piped water had 2.49 times greater odds of being infected by RVA compared to the children who did not drink piped water (aOR = 2.49, 95% CI: 2.49 1.04-5.95; p < 0.040) ( Table 3).

Clinical factors related to RVA infection in children from Hospital Central de Nampula
Overall, the factors related to RVA infection were a greater number of diarrhea episodes in the last 24h before admission (p = 0.041); and the occurrence of vomiting (30.6%; 126/418; p < 0.001) ( Table 4). In the pre-vaccine period occurrence of fever was the only clinical factor related to RVA infection (48.4%; 15/31; p = 0.038) ( Table 4).
In the post-vaccine period, clinical factors related to infection by RVA, were the number of diarrhea episodes in the last 24h, infection was more common in children with four or more episodes (p < 0.01); and occurrence of vomit (27.1%, 72/266; p = 0.001) ( Table 4).

RVA genotypes distribution by age group in pre-and post-vaccine period in Hospital Central de Nampula
Sixty out of sixty-seven (89.6%) RVA positive samples from the pre-vaccine period had a sufficient amount for RT-PCR. G1P [8] genotype combination was the most common among children younger than 24 months of age (45/55) (Fig 3., S2 Table).
All RVA positive samples collected in the post-vaccine period were submitted to molecular characterization and were from children below 24 months of age. The G1P [8] was the most common with 28.3% (26/92). However, in children younger than 12 months of age, a higher frequency of emergent strains that were not detected in the pre-vaccine period were observed, including mixed infections. In the older group (12-23 months) G1P [8] 100% (22/22) was the

Monthly distribution of RVA infection in pre-and post-vaccine period
Since the introduction of the vaccine in the country, the number of diarrhea and RVA cases have declined, mostly since late 2017 at Hospital Central de Nampula. RVA cases occurred in both seasons, cold/dry and wet/rainy. The frequency of RVA during the cold/dry season (April to September) was 35.8% (108/302) and in the wet/rainy season (October to March) was 16.3% (51/312) (p < 0.001) (Fig 5).

Discussion
We investigated the epidemiology of RVA, and the distribution of genotypes by vaccination status and age among children with diarrhea attending Hospital Central de Nampula, in Nampula province northern region of the country. Our analysis showed that RVA infection  b Refers to strains that were non-typeable for G. https://doi.org/10.1371/journal.pone.0255720.t005

PLOS ONE
Rotavirus A infection in pre-and post-vaccine periods in Nampula Province, Mozambique declined following vaccine introduction by 13.1%. In comparison, two other studies in Maputo city, a southern region of the country, found a drop in RVA infection of 31.2% and 14.2% respectively [23,31]. The differences observed can be explained by the lower vaccine coverage in Nampula province (46%) compared to Maputo city (61%) (data not published from the Expanded Program on Immunization) in 2016, and the lower health indicators in Nampula province [32]. Despite the reduction in RVA morbidity during the post-vaccine period, an increase in RVA infection frequency was seen in 2017. This event may be explained by the lower RVA vaccine coverage [33], and the impact of increasing the surveillance after introducing of RVA vaccine in the country.
After the vaccine introduction, there was a reduction of RVA infection in different age groups, with a significant decrease among children from 24-59 months of age. However, the frequency of RVA infection remains high in children under 12 months of age. This finding differs from an analysis from Hospital Geral de Mavalane and Hospital Central de Maputo, which reported a substantial reduction of RVA infections among children under 12 months of age, with a shift to older age-groups [14,23].
Differences in the age infection shifting may be due to lower vaccination coverage in Nampula [32], reinforced by our data in which 38.6% of the children were not vaccinated even being eligible for vaccination in post-vaccine period. RVA is a zoonotic agent with a risk of interspecies transmission [34,35]. In Chókwè, a rural area of Mozambique, phylogenetic data of G4P [6] strain detected in children, clustered with porcine and human porcine-like strains [14]. Animal contact with duck, chicken or multiple animals points to a potential association of RVA interspecies transmission. The occurrence of RVA in chickens was reported in Nigeria [36] and Kenya [37]. Furthermore, a study conducted in Germany identified an avian VP4 gene, strain P [37] closely related to mammalian RVA [38]. These data highlight the need of expanding the RVA surveillance to domestic animals in Mozambique.
Mozambique has problems related to the availability of water and as a strategy to mitigate the scarcest; intermittently distribution of water between few hours during the day was adopted, which forces households to storage the water in recipients for long use. Our surveillance has limited information on the conditions that the piped water were stored, or in what extent would be safe for consumption after storage. Most of the Mozambican pipelines are old and damaged in some sections aligned with the fact that are not properly maintained [39,40], which can contribute for contamination of the piped water mainly in flood situation as the supply network are emerged [39,41,42]. A recent study conducted in the capital of the country (Maputo), which analyzed different sources of water (home bottled, piped and supply well water) observed higher microbiological contamination than recommended for human consumption in piped water [42]. Although the study was carried out in Maputo, the company that supplies water at the national level is the same. If so, this finding helps to explain the high frequency of RVA in children who consume piped water in Nampula.
The Demographic and Health Survey showed that only 43% of the children under six months are solely breastfed in Mozambique [19]. Surprisingly in this analysis, RVA infection was more common in children exclusively breastfeed. Similar results were reported in Kenya in children under 24 months, where a high proportion of positive children were exclusively breastfed than those with other types of food, with 16% and 10.3% respectively in the post-vaccine period [43]. One of the hypotheses behind this finding is related to the maternal antibodies which may interfere with the RVA vaccine efficacy in lower and middle-income countries, however, the causal link is not well described [44].
The fact that the proportion of RVA infected children with fever, vomiting, multiple diarrhea episodes, or visit requiring hospitalizations decreased in the post-vaccine period, may highlight the role of vaccination in reducing the severity of RVA infection [45,46].
More than one-third of the children who received only the first dose of vaccine had a high frequency of RVA infection, which raises the question of whether they are shedding the vaccine strain. Vaccine shedding was accessed and the minimum days between the symptoms onset and the vaccination day for dose one were 32 days and for dose two were 26 days (data not shown), which excludes the possibility of shedding [47]. This result indicates a need for further research in Nampula to understand which factors are behind the high frequency of infection in vaccinated children.
Different genotype combinations in the three groups of children; the unvaccinated, partially vaccinated, and fully vaccinated were observed. The combination G1P [8] was the most common in fully vaccinated children. The fact that vaccinated children hospitalized due to RVA infection, with the same genotype from the vaccine, may be related to the high prevalence of chronic malnutrition in this province (55%), the highest in the country [19]. It is known that malnutrition can impair the immune system and weaken the vaccine's immune response [48,49]. However, this hypothesis cannot be tested in this analysis due to study design, and we cannot rule out other factors as concomitant enteric infections, histo-blood group antigen, microbiota composition and mothers antibodies [50,51].
The G1P [8] strain was the most common genotype combination in all age groups before the introduction of the vaccine. A high genotype diversity after vaccine introduction was observed in children under 12 months of age. Our findings are not concordant with a study conducted in the southern region of the country in which the diversity of RVA genotypes strains were observed in all age groups during both periods [31]. Also, in Turkey, genotype diversity was common in children from 13 to 24 months of age [52].
Regardless of the vaccine period, RVA was more common in the dry/cold season (April-September) than in the wet/rainy season (October-March), as expected [23,31].
The study findings should be interpreted carefully as the following limitations were identified i) variables with non-response in the forms, even with continuous training of the staff, ii) unbalanced sample size between pre and post-periods and small sample size to some subgroups, iii) time period in the pre-vaccine was lower than the ideal minimal two years' period (March-December 2015). However, our sample size meets the minimum sample power (80%) of the main hypothesis tests. This study was carried out at a quaternary-level hospital, which probably included only children with severe disease, where the epidemiological distribution of the disease may be different from hospitals at other levels (tertiary, secondary, rural, post). Despite the limitations, we were able to describe the epidemiology of RVA infection at HCN and the risk factors related to infection in each period (pre-and post-vaccine periods). Some questions need to be addressed, such as: (i) the reason why vaccinated children are infected with RVA to the point of requiring hospitalization; (ii) the role of animals in the occurrence of new/uncommon RVA genotypes to optimize the vaccination.
In conclusion, RVA infection was high in Hospital Central de Nampula even after vaccine introduction, mostly in children under 12 months. During the pre-vaccine period, piped water was associated with RVA infection, however, additional studies are needed to investigate this association. In contrast, after vaccine introduction, risk factors for RVA infection were age (0-11 months) and contact with animal (chicken, duck and multiple animals). The infection was more common in vaccinated children, G9P [4] and G1G3P [8] were the most common genotypes in partially vaccinated, while in fully vaccinated G1P [8], G9P [6], and G9P [4] were the most common. Our results show a need for further research to understand the factors behind RVA infection in vaccinated children such as the vaccine immune response and factors that contribute to strains diversity through the whole genome sequence and phylogenetic analysis in Nampula.