Abstract:
Diarrhoea is a major cause of morbidity and mortality in low and middle income countries. Estimates of enteric viral diarrhoea are usually obtained from enzyme immunoassays and PCR-based diagnostic methods. However, the performance of these diagnostic methods can differ due to factors such as their principles, disease prevalence, viral diversity and evolution events. The study was designed as a repeated cross-sectional study to compare the diagnostic performance of an enzyme immunoassay, two PCR-based methods and agnostic Next Generation Sequencing (NGS) used in investigating viral diarrhoeal aetiology during the pre-(2013) and post-rotavirus (2016) vaccine introduction periods in coastal Kenya. The study compared an enzyme immunoassay (EIA) kit and two nucleic acid-based methods for rotavirus group A (RVA) detection in hospitalized children pre-post rotavirus vaccine introduction in Kenya. A total of 489 fecal specimens, 237 from 2013 (48.5%, pre-vaccine) and 252 from 2016 (51.5%, post-vaccine), were analyzed. Although the specificity of the 3 methods was ≥ 96%, the sensitivity of EIA, 68.9% (95% C.I: 53.4-81.8%), differed significantly from that of real-time PCR, 93.3% (95% C.I: 81.7%-98.6%), and taqman array card, 97.8% (95% C.I: 88.2-99.9%), in 2016 (p-value <0.05) unlike in 2013. The study also compared real-time PCR with taqman array card (TAC) in the detection of adenovirus, astrovirus, norovirus genogroup II and sapovirus. A total of 494 fecal specimens, 242 (49%) from 2013 and 252 (51%) from 2016, were analyzed. TAC detected more positives than real-time RT-PCR in all the four viruses. However, the difference was only statistically significant in adenovirus and astrovirus detection (p-value < 0.05). Agnostic NGS was applied to 69 samples that had tested positive for RVA by EIA in 2013 to determine viral coinfections. Detection of the coinfections varied by the bioinformatics workflows i.e. classification of unassembled reads using Kraken detected an enteric virus coinfection in 32 (46.4%) samples compared to Metaphlan which detected a coinfection in 21 (30.4%) samples and classification of contigs using Kraken detected a coinfection in 16 (23.2%) samples. In conclusion, RVA-EIA showed a significant decrease in sensitivity (p-value < 0.05) post-vaccine introduction compared to real-time PCR and TAC. Further, TAC detected significantly more adenovirus and astrovirus positives than real-time RT-PCR. The difference in the diagnostic performance of the methods impacted on the calculated prevalence estimates. Finally, the study also showed the potential of agnostic sequencing in the detection of viral coinfections in clinical samples.
