Introduction

Lymphatic filariasis (LF), commonly known as elephantiasis, is a mosquito-borne neglected tropical disease (NTD) caused by infection with the filarial nematode species Wuchereria bancrofti, Brugia malayi or B. timori [1]. The parasites are transmitted from person to person by mosquitoes belonging to the genera Culex, Anopheles, Mansonia and Aedes. Clinically, LF infection may be asymptomatic, but can also lead to acute and chronic conditions, including lymphoedema (swelling of the tissues, primarily the legs) and hydrocele (scrotal swelling). In 2000, the World Health Organization (WHO) launched the Global Programme to Eliminate LF (GPELF) as a public health problem [2,3]. The GPELF has two principal aims: (i) to interrupt LF transmission, and (ii) to manage morbidity and prevent disability. To interrupt transmission of LF, the GPELF recommends annual community-wide mass drug administration (MDA) of two-drug treatment regimen (albendazole plus either ivermectin or diethylcarbamazine) to entire at-risk populations for 4 to 6 years at adequate levels of coverage. Modelling studies have estimated adequate treatment coverage to be at least 65% of the total population in endemic areas [4,5].

It is estimated that the GPELF resulted in substantial health and economic benefits during the period 2000–2014 [6]. Overall, an estimated 36 million clinical cases and 175 million disability-adjusted life years (DALYs) were potentially averted during the period. In economic terms, an estimated US$100.5 billion was potentially saved through medical expenses, potential income loss, and costs to the health system.

Approximately 4 million people residing in six counties (23 sub-counties) of the coastal region of Kenya, including Kilifi, Kwale, Lamu, Tana River, Mombasa, and Taita Taveta are considered to be at-risk of LF infection. Kenya started an MDA programme with diethylcarbamazine (DEC) plus albendazole in 2002 in the then Kilifi district and expanded to include Malindi and Kwale districts in 2003. The geographic coverage was progressively expanded over the years to include other districts in the coastal region except Mombasa city and its peri-urban areas. However, the programme faced financial and operational challenges which led to MDA being missed in some years and complete cessation after the 2011 MDA. Nonetheless, the non-consecutive rounds of MDA led to considerable reduction in the prevalence of LF [7]. The MDA programme was restarted in 2015 and baseline population surveys confirmed that CFA prevalence was still greater than the 2% antigenemia threshold in several areas [8], thus justifying the resumption of MDA campaigns.

Findings of clinical trials conducted in Papua New Guinea have shown that co-administration of the three drugs used to treat LF (ivermectin, DEC, and albendazole—IDA) results in improved efficacy for clearing W. bancrofti microfilariae (Mf), which is the stage of the filarial parasite that is ingested by mosquitoes, and sterilization of adult filarial worms [9,10]. In 2017, following a thorough review process of data from safety and efficacy trials of the triple-drug regimen, the WHO recommended it as an alternative MDA regimen for the acceleration of LF elimination [11]. The guidelines recommend the use of annual IDA in specific settings where onchocerciasis is not co-endemic with LF including places where MDA has not yet started, fewer than 4 effective MDA rounds have been administered, and where MDA results have been sub-optimal. Following the WHO’s formal approval and release of the alternative treatment guidelines, Merck and the MECTIZAN Donation Program (MDP) promised to work with the WHO, national LF programs and other stakeholders to develop a mechanism to enable for the deployment of any required additional donation [12]. The announcement was a major enabler for the adoption of the IDA regimen by endemic countries. Computer simulation modelling suggests that the triple-drug regimen has the potential to dramatically accelerate the elimination of LF if epidemiological treatment coverage is high and systematic non-compliance is low [13].

Sentinel site surveys conducted throughout the coastal region in 2015 and 2016 identified Lamu East and Jomvu sub-counties as the sites with the highest prevalence of LF (6.3% and 6.7% antigenemia, respectively) [8]. Lamu county had received one round of MDA in 2011 and additional two rounds in 2016 and 2017 using the two-drug regimen of DEC and albendazole, and all the three MDA rounds had achieved effective coverage (i.e., ≥65% of the total population was treated). Jomvu, however, had received two rounds of MDA (in 2016 and 2017), but none exceeded the 65% treatment coverage threshold. Due to the relatively high prevalence of LF, the intermittent history of MDA and the sub-optimal coverage, both Lamu county and Jomvu sub-county were selected to receive two annual rounds of IDA in order to accelerate LF elimination.

In evaluation units where two-drug MDA is given for 5 to 6 consecutive years and the LF infection reduced to the recommended thresholds, the WHO recommends post-MDA surveillance using transmission assessment surveys (TAS) among six- and seven-year-old children [14]. Critical cut-off values of numbers of circulating filarial antigen (CFA) positive children are employed to determine whether transmission has been interrupted in the evaluation units. While the TAS has proven to be an effective tool for basing MDA stopping decisions under the standard two-drug regimens, it is unclear whether the target age group and epidemiologic target (<2% antigenemia in areas with W. bancrofti infections) are appropriate when IDA is used. A follow-on study to the clinical trial in Papua New Guinea found that a single dose of IDA sterilizes adult filarial worms for at least 5 years, but may fail to clear CFA (15). The study suggests that a better biomarker or a different surveillance strategy will be needed to assess the effect of the three-drug regimen on LF populations. In this paper, we report on the NPELF experience delivering MDA with IDA in three implementation units in coastal Kenya. Because IDA was anticipated to result in an accelerated interruption of transmission, an operational research study was nested within the work of the national program with the goal of establishing the target population(s), infection indicator(s), sampling strategies, and thresholds required to determine when it is safe to stop MDA with IDA. Consequently, in this paper we present both the experience of the NPELF in delivering IDA as well as the baseline and impact assessment survey results from the closely related operational research study in two evaluation units following IDA implementation.

Methods and study design

Results

In Kenya, the MDA implementation unit is the sub-county (equivalent of district in most other African countries). Although Lamu County was considered as a single EU for the operational research study, the two sub-counties in this area were considered separate implementation units (IUs) during IDA-MDA. Table 1 shows the epidemiological treatment coverage (reported coverage) in the three IUs in 2018 and 2019. An independent treatment coverage evaluation survey undertaken soon after the IDA-MDA to validate the epidemiological treatment confirmed that high treatment coverage was indeed achieved (Table 1).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Coverage of two rounds of mass drug administration (MDA) with triple-drug therapy (ivermectin, diethylcarbamazine and albendazole—IDA) in Lamu and Jomvu implementation units, 2018 and 2019.

https://doi.org/10.1371/journal.pntd.0011942.t001

Overall, a total of 6,559 individuals were enrolled for the baseline survey out of whom 3,095 and 3,464 were from Lamu and Jomvu, respectively (Table 2). Out of 10,694 individuals enrolled during the IDA-MDA impact assessment survey, 5,235 and 5,459 were from Lamu and Jomvu, respectively.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Baseline and IDA-MDA impact assessment survey characteristics in each evaluation unit (EU).

https://doi.org/10.1371/journal.pntd.0011942.t002

Table 3 summarizes the prevalence of CFA in the two EUs, while comparing children and adult populations and the random and purposive clusters. In general, the overall CFA prevalence in both types of clusters was relatively low at the baseline survey, with Jomvu having 1.39% (95%CI: 0.91,2.11) and Lamu having 0.48% (95%CI: 0.21, 1.13). Of 15 CFA positive individuals identified during the baseline survey in Lamu, 3 were children between the ages of 5 and 9 years. In contrast, of the 48 CFA positive cases identified during the baseline survey in Jomvu, 11 were children aged 5–9 years. Of the 63 CFA positive cases during the baseline survey, one individual (a 45 years old male in Jomvu) was found to have three (3) Mf in his night blood sample. All the CFA positive individuals in the impact assessment survey provided a blood sample for Mf testing but none was found to have microfilaremia.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Prevalence of CFA during baseline and impact assessment surveys, and relative reduction (RR %) stratified by cluster type.

https://doi.org/10.1371/journal.pntd.0011942.t003

While there was an apparent reduction in the overall CFA prevalence following the administration of two rounds of MDA with IDA, the reduction was not consistent across the two EUs and study populations. Despite Lamu having very low numbers of CFA positive cases in the baseline survey, there was observed statistically significant decrease in prevalence between the baseline and impact assessment surveys, RR = 52.71% (Wald test: Z = -1.97, p = 0.049). This significant decline in Lamu was largely attributed to the random clusters, RR = 80.29% (Wald test: Z = -2.56, p = 0.011). In Jomvu, where CFA prevalence was relatively higher at baseline survey, the overall reduction in CFA prevalence was significant in the purposive villages, RR = 90.06% (Wald test: Z = -3.64, p < 0.001), but not in the random sites, RR = 39.70% (Wald test: Z = -1.97, p = 0.116). The reduction in CFA prevalence among the adult population after the two rounds of IDA-MDA in Jomvu was significant in both random, RR = 50.80% (Wald test: Z = -2.36, p = 0.018) and purposive, RR = 83.89% (Wald test: Z = -2.56, p = 0.010) clusters. Restricting the analysis to children in Jomvu, the reduction was only significant among the purposive villages, RR = 100.00% (Wald test: Z = -7.29, p < 0.001), but not in the randomly selected villages, RR = 52.70% (Wald test: Z = -0.91, p = 0.365). The 3 CFA positive children detected during the impact assessment survey in 2021 came from 3 different random villages in Jomvu. Two of these villages had overall CFA prevalence above 2% (2.68% and 3.60%) whereas the prevalence in other one village was below 2% (0.67%).

In general, there was a greater number of CFA positive clusters in Jomvu compared with Lamu during both surveys (Table 4, and Figs 1 and 2). A total of 11 out of 35 clusters in Jomvu had CFA prevalence of 2% or above during the baseline survey, compared with 4 out of 39 clusters during the impact assessment survey. In contrast, Lamu had 2 out of 35 clusters with CFA prevalence of 2% or above during the baseline survey and 1 out of 36 clusters during the impact assessment survey.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. A map showing filarial antigenemia rates by FTS test at baseline in 2018 and impact assessment in 2021 (after 2 rounds of IDA-MDA) among study clusters in Jomvu and Lamu evaluation units.

The inset map of Kenya shows the location of the two study areas. The map was created using ArcGIS Desktop version 10.2.2 software (Environmental Systems Research Institute Inc., Redlands, CA, USA). The base layer of the map was obtained from Environmental Systems Research, Inc. (ESRI) (https://www.esri.com/en-us/arcgis/products/arcgis-online/resources).

https://doi.org/10.1371/journal.pntd.0011942.g001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Village-level prevalence of filarial antigenemia by FTS test at baseline in 2021 and impact assessment in 2021 (after 2 rounds of IDA-MDA), stratified by evaluation unit. Points represent point prevalence (%) with 95% confidence interval (CI) error bars.

https://doi.org/10.1371/journal.pntd.0011942.g002

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. LF positivity among the study clusters (villages) in the two EUs during the baseline and impact assessment surveys stratified by point prevalence categories.

https://doi.org/10.1371/journal.pntd.0011942.t004

At the impact assessment survey, none of the 5–9-year-old children tested for CFA were positive in Lamu, while 3 were found to be positive in Jomvu. Compared with the WHO recommended transmission assessment survey (TAS), the CFA positive child cases observed in both EUs were below the threshold (Table 5).

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. IDA-MDA impact survey results in 5–9 years children, compared with transmission assessment survey (TAS) thresholds.

https://doi.org/10.1371/journal.pntd.0011942.t005

Discussion

Compliance to the triple-drug regimen MDA in Lamu county and Jomvu sub-county was high due to excellent social mobilization by the National Programme for Elimination of LF (NPELF) and implementing partners. The various integrated strategies that were employed during MDA with IDA in the two areas in coastal region of Kenya showcased how a comprehensive approach can be employed to achieve high treatment coverage. Through partnerships, tailored communication, robust adverse event management, community engagement, innovative delivery mechanisms, effective CDD selection, and the involvement of LF champions, the program demonstrated the potential for achieving high treatment coverage. The treatment data demonstrated that the 80% treatment coverage target recommended by WHO could be realized. Indeed, the NPELF and partners were so successful in their social mobilization efforts that people from neighbouring sub-counties not meriting IDA-MDA travelled to Jomvu to receive the drugs (as is evident from the MDA coverage exceeding 100% in Jomvu). The results imply that NTD endemic communities will support the introduction of innovative treatments when the benefits of this change are clearly explained to them.

The results of the IDA-MDA impact assessment survey offered strong evidence that it was safe to stop the IDA regimen in the two EUs because transmission appears to have been interrupted. As a result of this survey, the Kenyan program stopped IDA-MDA in the two evaluation units and has since moved on to a period of surveillance. This study generated valuable data on the population, infection indicators, and sampling strategy that are effective for determining when it is safe to stop IDA-MDA. Adults had the greatest prevalence of CFA compared to children, suggesting that adults may be a preferred age group for IDA-MDA stopping decisions because they constitute a high-risk group for persistent infection. Despite the elevated CFA prevalence in the adult population, there was a significant reduction after the two rounds of annual IDA-MDA in the random clusters in both EUs. This reduction in CFA prevalence may be due to the low intensity of infections at baseline, and it would be interesting to determine if these results could be replicated in higher microfilarial intensity settings. A larger and more consistent reduction in CFA was seen in children across both the purposive and random sites in both EUs.

Importantly, measuring CFA alone may not be an adequate way for monitoring and evaluation of LF elimination programs using IDA-MDA [18]. When IDA-MDA is used to accelerate the timeline to elimination (as opposed to the standard two-drug regimen which is delivered for 5+ years), there is not a clear interpretation for CFA results alone. Although the presence of filarial antigenemia in untreated individuals may be considered indicative of active W. bancrofti infection, its epidemiological significance in LF populations that have been given the required rounds of mass treatment is not clear. Indeed, evidence from initial IDA safety and efficacy clinical studies showed persistence of CFA for five years after a single dose treatment with IDA even with sustained clearance of Mf, which suggests sterilization of adult filarial worms albeit failure to clear filarial antigenemia [9,15]. Consequently, it is important that a more definitive indicator, such as the presence of Mf, be used to confirm whether CFA positive individuals are contributing to ongoing transmission. Because patent infection can take up to years to develop from the time of exposure, testing of adults for microfilaremia might be more useful for public health monitoring. A CFA-positive, Mf-negative child may indicate the presence of immature worms that will later produce Mf. Unfortunately, very low microfilaremia prevalence at baseline (one positive individual) and apparent absence of Mf during the impact assessment does not allow for any direct conclusions regarding the suitability of microfilaremia as an indicator in the present study.

The development of new and more sensitive diagnostics tools, particularly for use during post-MDA with IDA surveillance, is urgently required to detect recrudescence. The use of entomological sampling and molecular detection of filarial DNA in mosquito vectors, known as molecular xenomonitoring is also considered a surveillance tool for LF that may have particular utility in low prevalence settings, and thus could supplement the standard diagnostics [19–21]. However, although molecular xenomonitoring can also be used to detect the presence of the parasite in mosquito vectors and is a sensitive indicator of microfilaraemia, it is not a direct measure of infectivity or current rates of parasite transmission [14]. Another challenge with molecular xenomonitoring is that it is not yet known how to interpret a signal of infection in mosquitoes (e.g., what is the threshold for action?) and the approach is laborious, expensive and requires diagnostic equipment and techniques that aren’t readily available to many NTD programs.

Interestingly, the comparison of randomly vs. purposively sampled sites leads to different conclusions in the two EUs. In Jomvu, the randomly sampled sites had a greater prevalence of CFA among adults compared to the purposively sampled sites at impact assessment (1.10% vs. 0.22%). Conversely, in Lamu the randomly sampled sites had a lower prevalence than the purposively sampled sites during the impact assessment (0.10% vs. 1.84%). While some of this difference may simply be random variation due to such low overall positive counts, it may also be driven by the differences between the two EUs. Jomvu is peri-urban with a lot of human movement, making it harder for the national program to know which sites were the highest risk areas for purposive selection. Lamu, in contrast, is less densely populated and has fewer known high-risk areas. It is important for future WHO guidance to recognize these two potential scenarios and develop a sampling strategy that balances purposive sampling with random; both strategies should be part of an overall MDA stopping decision to maximize the chance that areas with ongoing transmission are detected.

Prior to this study, the default recommendation for LF programs to determine whether IDA-MDA can be stopped is to follow the TAS methodology [11]. In Kenya, due to the low baseline prevalence and successful IDA implementation, the conclusion that it is safe to stop treatment would have been drawn from either the TAS in children or the impact assessment in adults. Nonetheless, the stark difference in prevalence between children and adults in both evaluation units highlights the risk of basing an IDA stopping decision on children. Furthermore, even in settings where the pre-IDA prevalence of antigenemia is greater than was observed in Kenya, the prevalence of antigenemia in children may remain above the target threshold after only two rounds of IDA, due to the persistence of an antigen signal; whereas measuring microfilaremia in adults is more likely to provide an accurate assessment of the potential for ongoing transmission in the population.

There is growing appreciation that basing MDA stopping decisions on the mean EU prevalence is not enough. As LF transmission approaches interruption, the remaining infection becomes increasingly focal and it is important to critically examine cluster-level results when determining whether a stop treatment decision is appropriate. The cluster data presented in Figs 1 and 2, as well as Table 4, present reassurance that post-IDA-MDA, there do not appear to be any remaining foci of infection that merit programmatic action. Nonetheless, the fact that four clusters in Jomvu and one in Lamu had a cluster-level FTS prevalence above 2% is an indication that the NPELF should remain vigilant of these sites during post-treatment surveillance to ensure that the lingering signal is a sign of waning infection and not recrudescence. Based on these findings, we suggest that future IDA impact survey guidance include consideration of not only the mean EU prevalence, but also provide guidance as to when cluster-level results merit targeted action.

In an effort to make impact assessments and post-treatment surveillance of NTDs cost-effective in low-resource settings, the use of model-based geostatistics (MBG) has been proposed. MBG leverages spatial correlation within the data and enables one to design a survey that requires sampling fewer clusters and a smaller total sample size while achieving the same predictive qualities as conventional impact assessment [22,23]. A reduction in sample size through MBG would be particularly beneficial in the context of IDA impact assessments because sampling of adults and collecting night blood is particularly onerous for survey teams. It is therefore important to consider a follow-on operational research study as part of the second IDA-MDA impact surveillance in the two EUs in order to provide additional evidence on the application of MBG-designed surveys.

This study has several limitations worth highlighting. The coast region of Kenya may be considered a low prevalence setting due to the impact of several rounds of the two-drug regimen MDA with DEC and albendazole and the widespread bednet usage. Lamu had received one round of MDA in 2011 (66% epidemiological treatment coverage) before the NPELF stopped mass treatments. Thereafter, Lamu received 3 rounds of consecutive MDA rounds in 2015, 2016 and 2017 and all were above the 65% epidemiological treatment threshold. Lamu was included primarily due to the historical persistence of LF infection in Lamu East particularly in Ndau island. With regard to Jomvu, two rounds of MDA with DEC and albendazole had been given in 2016 and 2017 but both were below the recommended 65% threshold. The low CFA prevalence and single Mf-positive adult at baseline made it challenging to detect any spatial or temporal patterns in prevalence between the two study time points. The missed opportunity to assess the impact of IDA-MDA on other diseases, particularly soil-transmitted helminthiasis (STH), scabies, and head lice may be considered a study limitation. Ivermectin is known to act strongly against a wide variety of insects, nematodes, and acarine parasites, including lice and scabies [24,25]. A study to assess the impact of IDA-MDA for LF and STH control in Timor-Leste reported significant reduction in the prevalence of scabies, impetigo, and STH (Trichuris trichiura and Ascaris lumbricoides) among schoolchildren [26].

In this study, local and international partners provided significant financial and human resources which bolstered the national LF elimination program’s efforts to deliver IDA-MDA with high treatment coverage. Considering that the health systems in low- and middle-income countries (LMICs) are often faced by limited resources and capacity, the need for the significant investment might be considered a limitation. A systematic review of factors that facilitate the implementation of high-quality MDA for LF in sub-Saharan Africa identified partnerships and collaborations as essential for sustained and successful implementation of MDA for LF [27]. Thus, endemic countries that wish to accelerate the elimination of LF as a public health problem through IDA-MDA would need to establish strategic local and international partnerships, among other programmatic requirements.

A noteworthy strength of the study, however, is that the design of the cross-sectional surveys provided for the inclusion of both children and adult populations in both the baseline and impact assessment. The random sampling in Jomvu at baseline revealed pockets of ongoing transmission hitherto unknown to the NPELF. The results of the study also indicate that purposive sites have a critical role in LF surveillance, particularly in identifying areas with highest risk of LF transmission (putative hot spots). In Lamu for example, Ndau island is historically known to be an area with relatively high LF transmission [28], which was also confirmed during a previous survey [8]. Laboratory serologic and molecular (PCR) results are still forthcoming and when reported will shine even greater light on the impact of MDA with IDA–both spatially and across the age groups.

In conclusion, this study shows that the two rounds of IDA-MDA reduced filarial antigenemia below the recommended thresholds in the two EUs. There are other sub-Saharan Africa countries that are endemic for LF but nonendemic for onchocerciasis or loiasis (e.g., Zanzibar, Comoros, Sao Tome and Principe, Eritrea, Madagascar, Zambia, and Zimbabwe), where IDA could be considered for the acceleration of their LF elimination programs [29]. The study also demonstrated that programmatic delivery of MDA with IDA is a feasible and efficient strategy for endemic countries looking to accelerate their LF elimination programs as they move towards post-treatment surveillance. The operational research study highlights the utility of sampling adults in the context of IDA-MDA and shed important insight on the complementary role that random and purposive sampling can play in detecting pockets of ongoing transmission. The data collected at both baseline and impact assessment will be useful to guide global policy on monitoring and evaluation of LF programs that implement IDA-MDA. Indeed, since this study was conducted, interim guidance for IDA-MDA impact surveys (IIS) has come out recommending the testing of adults aged 20 years and above for CFA in 30 randomly selected clusters, with follow-up night time blood samples collected from the CFA positive individuals and examined for microfilaremia. While this IIS approach leverages random selection of clusters (with probability proportionate to estimated size sampling), it is preceded by an epidemiological monitoring survey conducted in two purposively selected high-risk sites. Further work is needed to understand the potential roles for MBG, xenomonitoring and serology to support LF surveillance; this and any follow-on studies in Kenya are well-suited to contribute to this important effort.

Supporting information

Acknowledgments

We express our gratitude to the county departments of health in Lamu and Mombasa for their excellent cooperation and support of the MDA with IDA and the associated operational research study. We would also like to thank the fieldwork teams for assistance with data collection. The local residents and operational research participants are acknowledged for their voluntary participation and commitment in this study. The partners, including Evidence Action, Africa Institute for Health and Development, Vector-Borne and Neglected Tropical Diseases Unit, Kenya Medical Research Institute, and the NTD Support Center provided excellent technical support.