WHO considers LF to be eliminated as a public health problem, if periodic transmission assessment surveys (TAS, with a predefined survey design) show that the average infection prevalence has been reduced and sustained below a critical threshold, expecting that transmission will eventually cease and the risk of resurgence is minimal ¹⁹ . In areas with bancroftian filariasis transmitted by Anopheles or Culex, the critical threshold has been set at 1% microfilaria (mf) prevalence in the community or 2% antigen prevalence in 6–7 year-old children; slightly lower values are used where Aedes is the main vector of bancroftian filariasis; for brugian filariasis, 2% antibody prevalence is used as critical threshold. Passing TAS does not necessarily mean that infection prevalence is below the threshold across the entire district; small foci with low-level residual transmission can be missed by TAS-like surveys, and additional effort is needed to detect microfoci ²⁰ . Uncertainty about the dynamics of, and association between, different infection indicators ²¹ makes it difficult to quantify the risk of resurgence associated with signals of residual transmission.

Models have been used to examine timelines to achieving elimination as a public health problem, usually defined as mf prevalence below 1%. Modelling suggests that achieving the 1% mf prevalence target is technically feasible with the standard WHO-recommended strategy of annual MDA with a two-drug combination (diethylcarbamazine + albendazole (DA) or ivermectin +albendazole (IA)). However, the required treatment duration strongly depends on baseline endemicity and achieved coverage (here defined as percentage treated out of the total population) and may often exceed the initially anticipated 5–6 years ⁹ . Poor coverage severely impedes elimination programs, especially when a large group of people is systematically not treated in repeated MDA rounds (also called systematic non-adherence or systematic non-compliance) ^{9, 22} . The risk of not achieving the 2030 targets is highest in areas with late-start MDA, high local baseline prevalence, and/or insufficient coverage.

Models were also used to explore to what extent elimination can be accelerated by using alternative strategies. Firstly, the required treatment duration can be minimized by optimizing the coverage and preventing systematic non-adherence ^{9, 23, 24} . This will enhance the impact per round and reduce the risk that residual transmission persists in an untreated population subgroup. This is particularly important for areas with a history of poor coverage. Secondly, treatment duration can likely be reduced by treating with more efficacious treatment regimens, such as a triple-drug combination of ivermectin, diethylcarbamazine and albendazole (IDA). This triple-drug combination was shown to be more efficacious than the standard two-drug regimens ²⁵ and our modelling suggested that the required treatment duration can be reduced by a third by using IDA instead of DA ^{9, 26} . However, using IDA is not a solution for poor coverage. Additionally, IDA is not safe in areas endemic for onchocerciasis or loiasis, and therefore cannot be used in large parts of Africa. The use of DEC-medicated salt can also be highly efficacious, but will require a completely different treatment delivery approach ²⁷ . Thirdly, models predicted that the required treatment duration can also be reduced considerably by treating biannually (i.e. twice per year) instead of annually if coverage remains the same, assuming that a) the second round reaches some people who were missed in the first round and b) people treated twice benefit from additional chemotherapeutic effects on worms and mf ^{9, 23, 24} . However, these predictions were not confirmed by recent community intervention trials and concerns exists on feasibility of biannual campaigns in low-resources settings. Lastly, models showed that complementary vector control (enhanced coverage of insecticide-treated bednets) has little impact on the required programme duration for reducing mf prevalence below 1%, but will help to reduce risk of resurgence ^{24, 26} . In 2017, WHO issued new guidelines on the use of alternative MDA regimens for LF elimination, informed by empirical data and modelling ²⁸ . They recommend the use of IDA in onchocerciasis and loiasis-free areas that have not started MDA or have not yet had four rounds with effective coverage (i.e. >65% of the total population), and for areas that failed to meet epidemiological thresholds for elimination as a public health after five or more treatment rounds with effective coverage. The use of biannual MDA is not recommended.

Detailed predictions of when elimination of LF as a public health problem can be achieved in African countries, under current or alternative strategies, have been published elsewhere ²⁴ . Accurate prediction is often difficult due to geospatial variation in and uncertainty about baseline endemicity and achieved coverage levels. Programmatic data on coverage are often unreliable due to different factors (e.g. not everyone who receives a tablet may also swallow, uncertainty about the overall population size makes it difficult to estimate coverage as percentage, health workers and/or programme managers at different levels may be incentivized to report inflated coverage figures). Data from sentinel sites can be used to validate and constrain models ²⁹ .

A key challenge will be ensuring high effective coverage. This can be done, as shown in various studies ^{30– 32} , but preventing systematic non-adherence remains important. Although models suggest that treating biannually can be very effective to accelerate elimination, there can be a reluctance to adopt biannual MDA due to costs and logistics, so it may not always be feasible to implement. Moreover, biannual MDA at a lower coverage could exaggerate the effects of systematic non-adherence, whereas increasing coverage will decrease heterogeneity ¹⁴ . Focusing resources on achieving high coverage for annual treatment may be more resource-effective than biannual MDA with lower coverage ²³ .

Loiasis co-endemicity presents a severe impediment for LF elimination programs, as both diethylcarbamazine and ivermectin can cause severe side effects in people highly infected with loiasis. The World Health Organization-recommended strategy for such areas is twice-yearly MDA with albendazole alone. Early modelling of twice-yearly albendazole, guided by limited empirical data, suggests that the required treatment duration under this strategy will be longer than for annual MDA with IA or DA ⁹ . Test (for loiasis)-and-not-treat (those with too high L. loa mf) could be an alternative strategy, if LoaScopes or other rapid diagnostics to test for loiasis become available ³³ . As only a small proportion of the population has to be excluded because of high L. loa mf density ³⁴ , this strategy will likely be successful in almost the same timespan as with standard MDA if adherence is equally good. However, this strategy may be relatively costly ³⁵ .

An important question is what measures are needed after the cessation of MDA to sustain the achievements. Field studies showed that low-level transmission can continue after passing TAS ³⁶ . Indeed, TAS can be passed with some residual infection remaining and, moreover, small foci with residual infection may be missed by TAS methodology (see above). Residual infection remaining after MDA cessation can lead to resurgence and reintroduction in areas that had been freed of LF, as was shown by a modelling framework for LF in American Samoa ¹⁷ , although these findings are not necessarily generalizable to other areas with different transmission conditions. Therefore, even after validating elimination of LF as a public health problem by passing the 3 ^rd TAS, some form of post-validation surveillance is required for early detection of possible resurgence. Quantitatively-informed guidance is needed for post-validation surveillance and for measuring elimination of transmission.

The risk of some residual infection remaining after stopping MDA will vary within treatment areas due to geospatial variation in baseline endemicity, transmission conditions (vector species, biting rate, heterogeneity in the exposure to vectors, etc.), or uptake of interventions. The risk of resurgence depends on the abundance of residual infections and the epidemiological setting. Theoretically, there is threshold prevalence below which the mating probability of any given adult worm is too low to sustain transmission, so that transmission will eventually cease to occur (elimination of transmission) even in the absence of further interventions. This breakpoint depends on specifics of the epidemiological setting, including vector species characteristics, vector abundance or local biting rate, heterogeneity in exposure to mosquito bites within the human population, density dependence in transmission processes, etc. For example, fewer vectors (e.g. through control) increase the threshold ^{10, 11, 37} , whereas assortative mixing will decrease the threshold ³⁸ . Modelling work has been conducted to assess breakpoint thresholds for mf prevalence, antigen prevalence and third-stage larvae (L3) prevalence in the vector population. Mf prevalence threshold values can be far below 1% and vary from site to site ^{11, 24, 29} , and are unmeasurable in the current TAS framework. L3 prevalence in mosquitoes could be the most sensitive indicator of transmission ¹¹ , and sequential sampling approaches based on infection in vectors could be more sensitive ²⁹ . Xenomonitoring gives a real-time indication of parasite presence and levels in communities ^{39– 41} . When prevalence is above the breakpoint, transmission can still die out stochastically. However, the risk of failure increases with increasing prevalence ⁴² .

Better understanding of spatial variations in transmission and uptake of interventions is critical for understanding which settings are at greatest risk of resurgence. Strengthening vector control during the endgame could reduce this risk and overcome site-to-site variation in timelines to elimination ¹¹ .

A recent systematic review found that the WHO recommended strategies for LF elimination are consistently cost-effective or cost-saving across a wide range of settings and assumptions ⁴³ . Model projections suggest that 175 million disability-adjusted life years (DALYs) were potentially averted by the first 15 years of the GPELF, saving a possible $100.5 billion (USD) over the lifetime of the benefited cohorts ⁴⁴ . Models suggest that the increased biannual treatment costs will be compensated for by shorter timescales ²³ . In poor coverage areas, enhancing coverage is the most cost-effective way to accelerate success ²³ .