Bayesian Network Analysis of Lymphatic Filariasis Serology from Myanmar Shows Benefit of Adding Antibody Testing to Post-MDA Surveillance

Abstract

The elimination of lymphatic filariasis (LF) is achieved through repeated mass drug administration (MDA) of anti-filarial medications, which interrupts transmission and prevents new infections. Accurate transmission assessments are critical to deciding when to stop MDA. Current methods for evaluating transmission may be insufficiently sensitive, resulting in post-MDA resurgence. We, therefore, evaluated potential diagnostic testing scenarios for post-MDA surveillance. Data were used from two surveys (a household cluster and a cohort) conducted in an area of Mandalay Region, Myanmar, with ongoing transmission following several rounds of MDA. First, age- and sex-adjusted seroprevalence were estimated for the area using the household survey. Next, three Bayesian networks were built from the combined datasets to compare antigens by immunochromatic testing (ICT) and/or Og4C3 enzyme-linked immunosorbent assay (ELISA) and antibody (Ab) detection methods (Wb123 or Bm14 Ab ELISA). The networks were checked for validity and then used to compare diagnostic testing scenarios. The adjusted prevalence from the household survey for antigen, Wb123 Ab and Bm14 Ab were 4.4% (95% CI 2.6–7.3%), 8.7% (5.96–12.5%) and 20.8% (16.0–26.6%), respectively. For the three networks, the True Skill Statistic and Area Under the Receiver Operating Characteristic Curve for antigen, Wb123 and Bm14 Ab were 0.79, 0.68 and 0.55; and 0.97, 0.92 and 0.80, respectively. In the Bayesian network analysis, a positive case was defined as testing positive to one or more infection markers. A missed result was therefore the probability of a positive case having a negative test result to an alternate marker. The probability of a positive case prior to any testing scenario was 17.4%, 16.8% and 26.6% for antigen, Wb123 Ab and Bm14 Ab, respectively. In the antigen-only testing scenario, the probability of a missed positive LF result was 5.2% for Wb123 and 15.6% for Bm14 Ab. The combination of antigen plus Bm14 Ab testing reduced the probability of missing a positive LF case as measured by Wb123 Ab to 0.88%. The combination of antigen plus Wb123 Ab was less successful and yielded an 11.5% probability of a missed positive result by Bm14 Ab testing. Across scenarios, there was a greater discordance between Bm14 and both antigen and Wb123 Ab in the 1–10 age group compared to older ages. These findings suggest that the addition of Bm14 Ab improves the sensitivity of LF testing for current or past infection. The combination of antigen plus Bm14 Ab should therefore be considered for inclusion in post-MDA surveillance to improve the sensitivity of transmission surveys and prevent the premature cessation of MDA.

Spatial predictive risk mapping of lymphatic filariasis residual hotspots in American Samoa using demographic and environmental factors

“…bancrofti transmission but may be used to provide information about levels of historical exposure and infection. This finding also concurred with a recent study indicating that the combined use of Ag and Bm14 Ab provides a more sensitive marker of current or previous infection than each indicator alone [ 56 ]. Additional longitudinal studies are required to help monitor how the stage of the infection and magnitude of the immunological responses determine the spatial patterns of antifilarial Abs.…”

Being Bayesian in the 2020s: opportunities and challenges in the practice of modern applied Bayesian statistics

“…Using Bayes theorem [ 135 ], the model calculates the probability of a given outcome for any defined scenario. Bayesian networks are widely used in a range of decision support settings including public health [ 136 , 137 ], environmental conservation [ 138 ] and natural resource management [ 139 ].

Figure 5 An example Bayesian network with a single, dependent child node (vaccine effectiveness) and two parent nodes (vaccine doses and variant).

…”

“…Using Bayes theorem [135], the model calculates the probability of a given outcome for any defined scenario. Bayesian networks are widely used in a range of decision support settings including public health [136,137], environmental conservation [138] and natural resource management [139].…”

Progress and challenges for the application of machine learning for neglected tropical diseases

“…A recent study by Dickson et al (2022) investigated the potential of diagnostic testing scenarios surveillance after MDA campaigns against LF in detecting transmission and prevalence of the disease using a Bayesian network framework. The effectiveness of several infection markers in detecting signs of transmission was assessed by using a Bayesian network framework using antigen-and antibody-based data (Wb123 Ab and Bm14 Ab).…”