Mathematical modelling of lymphatic filariasis elimination programmes in India: required duration of mass drug administration and post-treatment level of infection indicators

Abstract

Background: India has made great progress towards the elimination of lymphatic filariasis. By 2015, most endemic districts had completed at least five annual rounds of mass drug administration (MDA). The next challenge is to determine when MDA can be stopped. We performed a simulation study with the individual-based model LYMFASIM to help clarify this.

Methods: We used a model-variant for Indian settings. We considered different hypotheses on detectability of antigenaemia (Ag) in relation to underlying adult worm burden, choosing the most likely hypothesis by comparing the model predicted association between community-level microfilaraemia (Mf) and antigenaemia (Ag) prevalence levels to observed data (collated from literature). Next, we estimated how long MDA must be continued in order to achieve elimination in different transmission settings and what Mf and Ag prevalence may still remain 1 year after the last required MDA round. The robustness of key-outcomes was assessed in a sensitivity analysis.

Results: Our model matched observed data qualitatively well when we assumed an Ag detection rate of 50 % for single worm infections, which increases with the number of adult worms (modelled by relating detection to the presence of female worms). The required duration of annual MDA increased with higher baseline endemicity and lower coverage (varying between 2 and 12 rounds), while the remaining residual infection 1 year after the last required treatment declined with transmission intensity. For low and high transmission settings, the median residual infection levels were 1.0 % and 0.4 % (Mf prevalence in the 5+ population), and 3.5 % and 2.0 % (Ag prevalence in 6-7 year-old children).

Conclusion: To achieve elimination in high transmission settings, MDA must be continued longer and infection levels must be reduced to lower levels than in low-endemic communities. Although our simulations were for Indian settings, qualitatively similar patterns are also expected in other areas. This should be taken into account in decision algorithms to define whether MDA can be interrupted. Transmission assessment surveys should ideally be targeted to communities with the highest pre-control transmission levels, to minimize the risk of programme failure.

Keywords: Culex quinquefasciatus; Diethylcarbamazine and albendazole; Elimination; India; Individual-based model; Lymphatic filariasis; MDA duration; Mass treatment; Post-MDA residual infection; Prevalence; Wuchereria bancrofti.

Fig. 1

Comparison of the model-predicted association between Mf and Ag prevalence at community level to observed data from literature from Asian settings (black squares) and other regions including Africa, Oceania and the Americas (open black circles). The scale of the horizontal axis is restricted based on the observed values from Asian settings. Coloured dots show the model-predicted Mf and Ag prevalence, which were obtained by varying the average monthly biting rate between 1500–4000 bites per adult person per month. The model predicted Ag prevalence is shown for three different hypotheses on the mechanistic association between the presence of adult worms and detectability of antigenaemia. Hypothesis 1: antigenaemia is detectable in the presence of at least one male or female worm (blue). Hypothesis 2: the Ag detection rate is 50 % for single worm infections, but increases with the number of adult worms, simulated by assuming that antigenaemia is only detectable in the presence of at least one female worm or worm pair (red). Hypothesis 3: antigenaemia is detectable in the presence of at least one male + female worm pair (green). The darker and lighter colours show the association if Mf prevalence is measured in 40 and 60 μl blood, respectively. Simulated prevalence was for the whole population aged 5 years and above (triangles) or was standardized to give the expected prevalence in a study sample in which children under 10 and elderly individuals (squares) are underrepresented. With these provisions, the model captures the entire range of observed Mf prevalence levels in Asian settings

Fig. 2

Observed and model predicted age-specific Mf and Ag prevalence post-MDA. Empirical data are from two primary health centres in Thanjavur district, India, where 8 rounds MDA took place Thanjavur (MDA with DEC alone was given in 1997, 1999, 2000, and 2004; MDA with the combination DEC + ALB was given in 2001, 2002, 2003, and 2007). The model predictions show expected post-MDA age-prevalence patterns for a setting with low baseline endemicity (assumed mbr = 1600), with MDA rounds scheduled as in Thanjavur. a Visual qualitative comparison of model predictions to age-specific Mf prevalence data, under different assumptions for the achieved coverage per treatment round; b Visual qualitative comparison of model predictions to age-specific Ag prevalence data, under different hypotheses for the association between presence of worms and antigenaemia

Fig. 3

Age patterns of Mf (a) and Ag (b) prevalence of infection prior to MDA in the four simulated endemic settings. Antigenaemia is assumed to be detectable if at least one male or female worm is present in the host, but the detection rate increases with the number of adult worms (hypothesis 2). The model-predicted pattern of Mf-prevalence for Pondicherry (solid red line) matched well to the observed pattern (dots) from 1981. The predicted Mf prevalence prior to MDA at community level (8.5 %) for Pondicherry was within the range of the observed prevalence (8.6 %; 95 % CI: 7.9–9.4 %), as was the prevalence (5.3 %) in 6–7 year-old children (4.5 %; 95 % CI: 2.3–6.6 %). The model clearly mirrors the observed decline in prevalence in higher age groups (above 30 years)

Fig. 4

Predicted Mf and Ag prevalence for the population aged 5 years and above, prior to MDA (a & b) and 1 year after the required treatment duration (c & d). Antigenaemia is assumed to be detectable if at least one male or female worm is present in the host, but the detection rate increases with the number of adult worms (hypothesis 2). The boxes show the 25th and 75th percentiles of the distribution of the prevalence values and the horizontal line across the box is the median prevalence. The whiskers extend to 1.5 times the height of the box (i.e. the interquartile range, IQR) or, if no case/row has a value in that range, to the minimum or maximum values. If the data are distributed normally, approximately 95 % of the data are expected to lie between the inner fences. Values more than three IQR’s from the end of a box are labelled as extreme, denoted with an asterisk (*). Values more than 1.5 IQR’s but less than 3 IQR’s from the end of the box are labelled as outliers (o). The boxes combine information from the ~99 % runs ending in elimination and the ~1 % runs that did not achieve the target. The red dots indicate the prevalence levels for the few runs that did not result in elimination

Fig. 5

Predicted Mf and Ag prevalence for 6–7 year-old children, prior to MDA (a & b) and 1 year after the required treatment duration (c & d). Antigenaemia is assumed to be detectable if at least one male or female worm is present in the host, but the detection rate increases with the number of adult worms (hypothesis 2). See legend to Fig. 4 for additional information regarding the interpretation of the boxplots

Fig. 6

Sensitivity analysis: impact of modified assumptions on the residual Mf (a) and Ag (b) prevalence that is expected if MDA is continued long enough to achieve elimination with ≥ 99 % probability. See legend to Fig. 4 for additional information regarding the interpretation of the boxplots