Fig 1.
Comparison of the model-predicted and observed pre-treatment association between microfilaraemia (Mf) and circulating filarial antigenaemia (CFA) prevalence at community-level.
Observed data are from villages in the annual and biannual treatment arms in Côte d’Ivoire (triangles) and Liberia (squares), with CFA prevalences assessed using either the filarial test strip (FTS, black) or the immunochromatographic card test (ICT, white). FTS- and ICT-based observations made in the same community are connected via black dotted lines. Model predictions (dots) are shown for settings without (dark blue) and with (light blue) an external force of infection.
Fig 2.
Observed and simulated prevalence of microfilaraemia (Mf) and circulating filarial antigenaemia (CFA, measured by filarial test strip) for Côte d’Ivoire, at baseline (2014, before the first treatment round) and in 2015, 2016, and 2017 (i.e. 11 months after the first, second and third annual MDA rounds).
Observations are shown as coloured markers with 95% confidence intervals. Model predictions are shown as small dots. Simulation results from runs matched to specific villages at baseline are shown in the colour of that village, and all other runs are shown in grey. A run was considered a match if the predicted Mf-CFA prevalence combination at baseline fell within the ellipse drawn around the observed MF-CFA prevalence combination based on the 95% confidence intervals. For both the model and the observed data, crude prevalence estimates are presented in the figures (i.e. not age-standardized). The MDA coverage was assumed to be 65% of the total population per round in the simulation runs. See Table B in S1 Supplement for details about the simulated scenario and see Fig C in S1 Supplement for a similar figure for Liberia.
Fig 3.
Receiver-operator characteristic (ROC) curves for predicting the eventual occurrence of elimination of transmission under different MDA scenarios, based on the simulated microfilaraemia (Mf) or circulating filarial antigenaemia (CFA) prevalence 1 year after the last treatment round.
Different lines show the predictive performance of the Mf and CFA prevalence assessed in all people aged 5 or 15 years and above (Mf: red and grey; CFA: green and pink) and the CFA prevalence in all 6-7-year-old children (blue). Sensitivity is the percentage of simulation runs ending in elimination that are correctly identified based on Mf or CFA prevalence below a range of thresholds (see legend). Similarly, 100%-specificity is the percentage of simulation runs that is falsely classified as having achieved elimination, out of all runs that did not result in elimination, resulting in premature stopping of MDA. The optimal situation is in the upper left corner of the panels (100% sensitivity and 100% specificity). Results are based on 2000 simulations per scenario, with baseline Mf prevalence varying between 20% and 40%.
Fig 4.
Probability of achieving elimination within 50 years after the last round of MDA, if the 1-year post MDA prevalence was below a given threshold (i.e. the positive predictive value of using this threshold).
Different lines show the predictive performance of the Mf and CFA prevalence in a sample of 200 individuals taken from the 5+ or 15+ population (MF: red and grey; CFA: green and pink) and the CFA prevalence in a sample of 200 6-7-year-old children (blue). Results are based on 2000 simulations per scenario, with baseline Mf prevalence varying between 20% and 40%. The PPV curves based on CFA prevalence in the 5+ or 15+ population are somewhat erratic at low prevalences in the panel for settings that had 6 rounds of MDA with 60% coverage. This is explained by the fact that only few simulation runs ended with such low CFA prevalences in this treatment scenario (i.e. the denominator for calculating the PPV). If one of these few runs ended in resurgence, it causes relatively large fluctuations in the estimated PPV.