Results

Demographic and clinical characteristics of patients

From January 2015 to December 2018, 3903 patients with a respiratory syndrome corresponding to the SARI (31.5%) or ILI (68.5%) case definitions were included in the study. The demographic and clinical characteristics for enrolled patients are given in Table 1. Slightly more nasopharyngeal samples were collected from male (51.9%) than female patients (48.1%; female: male ratio of 0.9:1). ILI cases were mainly recruited at 4 out of 5 sentinel centers. SARI cases were mainly recruited (56.1%) at the Bangui Paediatric Complex, a hospital that is specialized in pediatric care, while between 3.8 to 16.5% of SARI cases were recruited in the other 4 centers. There were little differences in the age and sex distribution across the 4 years of the study, but there was some variability between study sites. Children aged between 0 and 6 months had an increased risk of presenting with more severe symptoms necessitating hospitalization compared to older children (p<0.001), while females had a lower risk of presenting with SARI compared to male (OR=0.84, p=0.014; Table S1).

RSV detection and association with age and sex

In total, 8.0% (312/3903) of all patients were tested positive by RT-PCR for the presence of RSV (Table 2). Among these cases, 155 (49.7%) belonged to the RSV-A sub-group and 40 (12.8%) to the RSV-B sub-group, while 117 (37.5%) could not be typed. Of the 155 RSV-A samples, 52/155 (33.5%) were from ILI cases and 103/155 (66.4%) were from SARI cases, with an overrepresentation of SARI cases in 2017. Of the 40 RSV-B, 13/40 (32.5%) were ILI patients compared with 27/40 (67.5%) SARI cases.
Among the RSV-positive cases, increased RSV detection in males (9.5%) compared to females (6.4%) was observed (OR= 1.53, p<0.001; Table 3). Children aged 0-6 months (13.4%) had an increased risk to be RSV positive compared to all other age groups (4.0-7.5%; p<0.001; Table 3).

Disease severity and RSV related deaths

RSV positive cases were more frequently detected among hospitalized patients (13.3% vs 5.5%, OR=2.62, p<0.001; Table 3). RSV was associated with dyspnea (p<0.001), wheezing (p<0.001), chest indrawing (p<0.001) and inability to feed (p=0.002) but not with rhinorrhoea (p=0.993), diarrhoea (p=0.983), vomiting (p=0.107), lethargy (p=0.816) or convulsion (p=0.752; Table 3). In hospitalized children, RSV was significantly more frequently detected in children with a diagnosis of bronchiolitis or bronchopneumonia compared to pneumonia (Table 3), RSV had a significant influence on oxygen saturation levels (RSV neg, n=294, median SaO2=96% vs RSV pos, n=64, median SaO2=92%; p=0.049; SaO2 levels not determined, n=874) and was associated with increased duration of hospitalisation (p<0.001).
In total 58/3903 (1.5%) of the patients enrolled died, among which 8 (13.8%; two females, six males) had a RSV infection, but no association between RSV infection and death was observed (p=0.162). For these 8 patients all younger than 24 months, delay between symptom onset and hospitalisation ranged from 0-7 days while the delay until hospitalisation and death ranged from 0-9 days. Co-morbidities were reported for 4 patients and included malnutrition, malaria and congenital heart disease. When measured (5/8), oxygen saturation levels were <95%. Bacterial co-infections with Staphylococcus aureus , Haemophilus influenzae , Moraxella catarrhalis ,Streptococcus pneumoniae and/or co-infections with influenza A virus were detected in 5 patients (Table S2).

Seasonal circulation of RSV

RSV detection rates varied across years and ranged from 6.4% in 2015 to 10.6% in 2017 (Table 2). Prevalence in 2017 was significantly higher compared to 2015 (p=0.002), 2016 (p=0.002) and 2018 (p=0.031).
During the study period, RSV detections started in March-April and lasted until December (Fig. 1 and Fig. 2). Sporadic detections were recorded between March-April to August and RSV circulation peaked in September (2015), October (2016) or November (2018). In addition to a significantly higher number of cases reported in 2017, the peak of RSV incidence occurred in June, 3 to 5 months earlier than during the other 3 years (Fig. 2). RSV was also significantly more frequently detected in the rainy season compared to the dry season (255/1756, 12.7% vs 57/1892, 3.0%; p<0.001).
Among the RSV strains that could be typed, high RSV-A circulation was observed in 2015 (30/32, 93.7%), 2017 (70/76, 92.1%) and 2018 (40/40, 100%). RSV-B was detected in 2015-2017 and was predominant in 2016, representing more than half of the RSV-positive cases (32/47, 68.1%; Fig. 3; Table 2).

Phylogenetic analysis

Glycoprotein gene sequencing was attempted for positive samples successfully typed as RSV-A or RSV-B and resulted in 160 partial G gene sequences. Based on phylogenetic analyses 17 genotype RSV-A NA1, 120 genotype ON1 and 23 RSV-B genotype BA9 sequences were identified (Fig. 3). RSV-A NA1 genotype was almost exclusively detected in 2015 (15/17, 88.2%) while RSV-A genotype ON1 replaced NA1 as of 2016 (Fig. 3).
Genetic distances between strains of the same genotype ranged from 0-1.4% for RSV-A NA1, 0-6.3% for RSV-A ON1 and 0-6.4% for RSV-B BA9. Five unique NA1 sequences (5/17, 29.4%) were observed and the same strain was found in 2015 to 2017 (Fig. S1). Twenty-six unique nucleotide sequences clustered within ON1 genotype (26/120, 21.7%). The same strains were found over one or two years on only four occasions, while a large cluster of identical strains (n=79) was found in 2017-2018 which might suggest local transmission. Ten unique BA9 sequences were identified (10/23, 43.5%; Fig. S2). Four unique strains were interspersed within the BA9 genotype, while the other 6 unique strains (corresponding to 18 sequences in total) formed a separate cluster, which suggested local transmission. Most CAF sequences represented novel, previously unpublished strains.

Genetic diversity in the second hypervariable region of the G gene

A total of six amino-acid substitutions (D245N, N268S, N281Y, L282P, E292K and K305I) were identified in the study sequences of the NA1 genotype compared to the prototype strain AB470478 (Fig. S3). A total of 29 amino-acid substitutions were identified in the studied region of the ON1 strains. Eleven mutations (G284S, E286G, L289F, H290Y, E295K, Y297H, P300L, V303A, Y304H, S307P/F) were observed in the duplication region and six (E308K, L310P, S311L, T319N/I, T320A) upstream from the duplication region, six downstream from the duplication region (I243S, T245I, L248I/F, G254R, T259I) and six in the conserved region (E262K, L265H, Y273H, L274P, Y280H, S283F; Fig. S4). A total of 22 amino-acid substitutions were observed in the studied BA9 sequences including five upstream from the duplication region (P216L, K218T, L219P, L223P, K233I), two in the conserved region (S247P, T254I), four in the duplication region (T270I, V271A, L272P, D273N) and 11 downstream from the duplication region (I281T, S285F, H287Y, T290N, E292K, S297P, T302A, E305K, P306S, T312N, Q313 stop codon; Fig. S5).