Detecting Schistosoma infections in endemic countries: a diagnostic accuracy study in rural Madagascar

Abstract

Background: Schistosoma haematobium and S. mansoni are endemic in Madagascar, but reliable diagnostic tools are often lacking, contributing to exacerbate transmission and morbidity. This study evaluated the diagnostic accuracy of three tests for schistosome infection in Malagasy adults from areas of medium to high endemicity.

Methods: This cross-sectional study enrolled adults from three primary health care centres in Madagascar. Urine and blood samples were tested for schistosome infection using polymerase chain reaction (PCR), up-converting reporter particle lateral flow for the circulating anodic antigen (UCP-LF CAA), and point-of-care circulating cathodic antigen (POC-CCA) tests. Bayesian latent class models were used to assess diagnostic accuracies and disease prevalence.

Results: Of 1339 participants, 461 were from S. haematobium and 878 from S. mansoni endemic areas. Test detection rates were 52% (POC-CCA), 60% (UCP-LF CAA), and 66% (PCR) in the S. haematobium area, and 54%, 55%, and 59% respectively in the S. mansoni area. For S. haematobium, PCR and UCP-LF CAA showed high sensitivity (Se, median 95.2% and 87.8%) but moderate specificity (Sp, 60.3% and 66.2%), while POC-CCA performed moderately (Se: 64.5%; Sp: 59.6%). For S. mansoni, PCR and POC-CCA demonstrated high diagnostic accuracy (Se > 90%, Sp > 80%), while UCP-LF CAA showed good sensitivity (79.9%) but moderate specificity (69.7%).

Conclusions: While population-level prevalence estimates were similar across tests, individual-level agreement was only low to moderate. Our findings suggest that optimal diagnostic strategies should be tailored to specific endemic settings, continued development of accurate diagnostics suitable for highly endemic settings remains a priority.

Keywords: Bayesian latent class models; Diagnostics; Imperfect reference standards; Madagascar; Prevalence; Schistosomiasis.

Fig. 1

Heuristic model. The model shows the assumed relationships between latent class (oval) and diagnostic test results (rectangles). All measures are worm-based. Abbreviations: DNA deoxyribonucleic acid, POC-CCA point-of care circulating cathodic antigen, PCR polymerase chain reaction, PZQ praziquantel, UCP-LF CAA up-converting reporter particle lateral flow circulating anodic antigen

Fig. 2

Participant recruitment and selection for diagnostic accuracy evaluation of schistosomiasis tests in Madagascar. From 1500 initially assessed participants, 1339 were eligible for analysis after excluding those with recent praziquantel (PZQ) treatment or missing test results. The final sample comprised 461 participants from S. haematobium and 878 from S. mansoni endemic areas. PZQ praziquantel

Fig. 3

Proportion of positive test results expressed as percentages by each of the three diagnostic tests among the assessed individuals in S. haematobium (left) and S. mansoni (right) endemic areas. Lighter shades indicate trace results for POC-CCA and species-specific results for PCR. Error bars represent 95% confidence intervals

Fig. 4

Distribution of PCR Ct values and UCP-LF CAA concentrations measured in pg/ml displayed by three POC-CCA categories (i.e., negative, traces, positive). Subfigures show (a) PCR Ct values by POC-CCA, S. haematobium endemic, b PCR Ct values by POC-CCA, S. mansoni endemic, c UCP-LF CAA concentrations by POC-CCA, S. haematobium endemic, d UCP-LF CAA concentrations by POC-CCA, S. mansoni endemic. The colour gradient indicates POC-CCA G score grading. Ct cycle threshold, PCR polymerase chain reaction, pg/ml picograms per millilitre, POC-CCA point-of-care circulating cathodic antigen, UCP-LF CAA up-converting reporter particle lateral flow circulating anodic antigen

Fig. 5

Relationship between diagnostic test results for S. haematobium and S. mansoni infections. Scatter plots show the relationship between PCR cycle threshold (Ct) values, UCP-LF-CAA concentrations (pg/ml), and POC-CCA grades. UCP-LF CAA antigen concentrations that were either imputed by a fixed value of 0.5 × cut-off of 2 pg/ml or the maximum values of 1000 were excluded from this visualisation to ease interpretation. Point sizes indicate the frequency of observations (< 20, 20–39, 40–59, ≥ 60 observations). POC-CCA grades range from 1 (negative) to 10 (strongly positive), represented by the color gradient. Data are shown separately for samples from S. haematobium endemic regions (left) and S. mansoni endemic regions (right). Note the logarithmic scale on the y-axis

Fig. 6

Diagnostic accuracy of the three tests under investigation based on Bayesian Latent Class Analysis in (a) S. haematobium and (b) S. mansoni endemic regions. CI conditional independence, CDP conditional dependence in positives, CDN conditional dependence in negatives, CDPN conditional dependence in positives and negatives, PCR polymerase chain reaction, POC-CCA point-of-care circulating cathodic antigen, UCP-LF-CAA up-converting reporter particle lateral flow circulating anodic antigen