Alternative approaches for monitoring and evaluation of lymphatic filariasis following mass drug treatment with ivermectin, diethylcarbamazine and albendazole in East New Britain Province, Papua New Guinea

Abstract

Background: WHO recommends two annual rounds of mass drug administration (MDA) with ivermectin, diethylcarbamazine, and albendazole (IDA) for lymphatic filariasis (LF) elimination in treatment naïve areas that are not co-endemic for onchocerciasis such as Papua New Guinea (PNG). Whether two rounds of MDA are necessary or sufficient and the optimal sampling strategies and endpoints for stopping MDA remain undefined.

Methods and findings: Two cross-sectional studies were conducted at baseline (N = 49 clusters or villages) and 12 months after mass drug administration (MDA) with IDA (N = 47 villages) to assess lymphatic filariasis (LF) by circulating filarial antigenemia (CFA) and microfilariae (Mf). Before MDA, children aged 6-9 years (N~50) and those ≥ 10 years (N~50) in each village were randomly sampled. Before MDA, the population mean prevalence of LF in East New Britain Province (ENBP), Papua New Guinea, was estimated using population proportionate sampling (PPS, N = 30) to be 59/2,561 (2.3%) CFA positive and 14/2,561 (0.6%) Mf positive. No children were Mf positive. However, LF infection was highly heterogeneous; 8 villages (26.7%) had a CFA prevalence >2%, and 7 villages (23.3%) had an Mf prevalence >1%. To identify sentinel villages with LF in areas under-sampled by PPS, 19 additional villages suspected to have LF were sampled, with 15 (79%) having >2% CFA prevalence and 7 (38%) >1% Mf (range 1-22%). Twenty-four villages were evaluated before and after MDA in age-matched adults ( ≥ 18 years). Treatment reduced CFA prevalence by 34% and Mf prevalence by 90%. Post-MDA model-based geostatistics efficiently selected an additional 23 villages, of which 20 (87%) had a CFA prevalence > 2%. None of these villages had >1% Mf. Post-MDA, two of four districts had no villages with >1% Mf.

Conclusions: Model-based geostatistics was more effective than PPS in sampling high-risk LF sites in a heterogeneous area. Low LF prevalence and partial reduction of CFA limit children's effectiveness as sentinels. A single round of high-coverage MDA with IDA achieved elimination targets in low-prevalence villages in PNG. Higher-prevalence areas will need additional MDA rounds, which could be targeted to smaller evaluation units to cut costs.

Trial registration: Clinicaltrials.gov NCT04124250.

Fig 1. Map of East New Britain Province.

The maps were made using R package, leaflet. Map data copyright from OpenStreetMap contributors and data is available under the Open Database License at https://www.openstreetmap.org/copyright. Basemap of PNG map (insert) from “CartoDB Basemaps” (https://github.com/CartoDB/CartoDB-basemaps) designed by Stamen and Paul Norman for CartoDB Inc. Basemap of East New Britain province from Esri World Street Map (sources: Esri, DeLorme, HERE, USGS, Intermap, iPC, NRCAN, Esri Japan, METI, Esri China (Hong Kong), Esri (Thailand), MapmyIndia, Tomtom) was obtained from https://leaflet-extras.github.io/leaflet-providers/preview/#filter=Esri.WorldStreetMap). Both maps are licensed under CC-BY 4.0.

Fig 2. LF infection parameters in villages according to different sampling strategies at baseline.

Thirty villages used a population proportion random sample. ENBP Provincial Health Authority purposefully selected five additional villages with prior evidence of LF. Fourteen more villages were chosen to adequately sample rural areas in Pomio (N = 9) and Gazelle (N = 5) districts. The dashed line on the left panel and the red line on the right panel represent WHO-specified thresholds of 2% CFA and 1% MF rates, suggesting ongoing LF transmission in villages above these thresholds.

Fig 3. Circulating filarial antigen (CFA) and microfilaremia (MF) prevalence by age groups pre-MDA for 49 villages.

Fig 4. Spatial distribution of lymphatic filariasis circulating antigen positivity (CFA, left panel) and microfilaremia rates (Mf, right panel) throughout ENBP before mass drug distribution.

Small solid dots indicated communities without any detectable LF infection. The colors and circle size represent LF prevalence for CFA or Mf. Mapping: The map was made using R package, leaflet. Map data copyright from OpenStreetMap contributors and data is available under the Open Database License at https://www.openstreetmap.org/copyright. Basemap from “CartoDB Basemaps” (https://github.com/CartoDB/CartoDB-basemaps) designed by Stamen and Paul Norman for CartoDB Inc., licensed under CC-BY 4.0.

Fig 5. Geostatistical probability map of LF prevalence.

The dark red area has a low probability that CFA prevalence is <2%; thus, these areas are more likely to have LF. The map resolution is 1 km. White regions have insufficient data to assess LF risk. Risk mapping was generated based on covariates of elevation, rainfall, temperature, distance from the sea, socioeconomic status, and LF prevalence using baseline survey villages (black dots). Mapping: Map was developed using R package, PrevMap. No basemap was used in the creation of this image.

Fig 6. Impact of treatment on CFA and Mf prevalence in 24 villages before and after MDA.

Prevalence estimates are for ≥18 years old with N = 816 before MDA and N = 2015 12 months post-MDA. Dashed lines represent means.

Fig 7. The Geographical distribution and prevalence of villages were sampled following MDA.

The left panel shows CFA prevalence, and the right panel shows Mf prevalence. Small white dots are villages without detectable CFA or Mf in participants. The color gradient of the maps corresponds to the different CFA prevalence (left panel). The size of green circles correlates with Mf prevalence, with the smallest green circles being 1% Mf. Mapping: The map was made using R package, leaflet. Map data copyright from OpenStreetMap contributors and data is available under the Open Database License at https://www.openstreetmap.org/copyright. Basemap from “CartoDB Basemaps” (https://github.com/CartoDB/CartoDB-basemaps) designed by Stamen and Paul Norman for CartoDB Inc., licensed under CC-BY 4.0.