Abstract
Despite decades of interventions, 240 million people have schistosomiasis. Infections cannot be directly observed, and egg-based Kato-Katz thick smears lack sensitivity, affected treatment efficacy and reinfection rate estimates. The point-of-care circulating cathodic antigen (referred to from here as POC-CCA+) test is advocated as an improvement on the Kato-Katz method, but improved estimates are limited by ambiguities in the interpretation of trace results.
We collected repeated Kato-Katz egg counts from 210 school-aged children and scored POC-CCA tests according to the manufacturer's guidelines (referred to from here as POC-CCA+) and the externally developed G score. We used hidden Markov models parameterized with Kato-Katz; Kato-Katz and POC-CCA+; and Kato-Katz and G-Scores, inferring latent clearance and reinfection probabilities at four timepoints over six-months through a more formal statistical reconciliation of these diagnostics than previously conducted. Our approach required minimal but robust assumptions regarding trace interpretations.
Antigen-based models estimated higher infection prevalence across all timepoints compared with the Kato-Katz model, corresponding to lower clearance and higher reinfection estimates. Specifically, pre-treatment prevalence estimates were 85% (Kato-Katz; 95% CI: 79%-92%), 99% (POC-CCA+; 97%-100%) and 98% (G-Score; 95%-100%). Post-treatment, 93% (Kato-Katz; 88%-96%), 72% (POC-CCA+; 64%-79%) and 65% (G-Score; 57%-73%) of those infected were estimated to clear infection. Of those who cleared infection, 35% (Kato-Katz; 27%-42%), 51% (POC-CCA+; 41%-62%) and 44% (G-Score; 33%-55%) were estimated to have been reinfected by 9-weeks.
Treatment impact was shorter-lived than Kato-Katz-based estimates alone suggested, with lower clearance and rapid reinfection. At 3 weeks after treatment, longer-term clearance dynamics are captured. At 9 weeks after treatment, reinfection was captured, but failed clearance could not be distinguished from rapid reinfection. Therefore, frequent sampling is required to understand these important epidemiological dynamics.