Cost-effectiveness of vector control strategies for supplementing mass drug administration for eliminating lymphatic filariasis in India

Setting

The State of Tamil Nadu had a population of 72 million persons in 2011 [29]. Lymphatic filariasis was endemic in at least 20 of its 33 districts [30]. The State Health Department funds health services in all districts. In 2008-09, the public system contained 42 health unit districts (HUDs) which supervised 1,421 primary health care (PHC) units in Tamil Nadu, with each PHC serving about 30,000 residents [31]. This cluster randomized trial collected data from 2010 to 2013 in Kallakurichi Health Unit District [32], which had a population of 3.5 million residents and 36 PHC units. The CRME, with headquarters in Madurai, Tamil Nadu, opened a field station in the Tirukoilur block of the Kallakurichi Health Unit of Villupurum [22]. The entomological team selected 36 villages as study sites based on their high risk of LF (Figure 1). With a combined population of 43,000 people, they were under three PHC units in the Tirukoilur block: Ariyur, Edaiyur and T.Kunnathur. Altogether, these three PHCs provided health care for about 150,000 people.

• Download figure
• Open in new tab

Figure 1:

Map of study villages by treatment arm in lymphatic filariasis vector control (LF-VC) project funded by the Bill & Melinda Gates Foundation (BMGF)

The CRME randomly assigned 12 villages to each of the three study conditions. The allocation was balanced exactly by LF microfilaremia baseline prevalence (MfP) and ecological circumstances for mosquitoes: 3 villages with MfP<1%, 3 villages with MfP 1-2%, and 6 villages with MfP>2%; and agreed closely in distribution by condition within each PHC block. The World Health Organization [33] defined the eligibility of communities for MDA treatment as the prevalence of filarial parasites (MfP) or antigen of more than 1%. While the population of villages varied [mean (M) = 1914, standard deviation (SD) =897], all villages included in this study were eligible for treatment (M = 85%, SD = 9%) with the WHO’s recommended two-drug regimen of DEC and albendazole [34].

Kallakurichi Health Unit District lies in northern Tamil Nadu, 40 to 80 kilometers inland from Puducherry on the Indian east coast. The general climate in the region is hot and dry, with moderate rainfall from September to December [35]. The majority of villagers are landless laborers who earn their living from agriculture and livestock [28]. For community management and administration, village committees are led by elected village presidents. Government health personnel contact with villagers relies on the village health nurse, who manages a PHC satellite unit called a sub-center, that serves approximately 5,000 inhabitants in two to three villages [36].

Prior to the study, no systematic VC operations had been routinely implemented in the study area. Villages have abandoned wells, since tap water standpipes have been installed. Sanitation involves pit latrines. The majority of houses lack effective barriers to insects, allowing endophilic mosquitoes easy access into the interior of dwellings and thus are able to bite residents inside. Due to the region’s hot and humid weather, villagers mostly prefer not to use bed nets for the prevention of mosquito bites, because they reduce ventilation. Traditionally, water used for showering and cleaning utensils is disposed of in soakaway pits and wet ditches, which readily serve as breeding sites for Culex mosquitoes.

Study design

The State Health Department (SHD) of Tamil Nadu implements MDA almost annually with tablets of DEC, procured from local manufacturers in India [4], plus albendazole donated by the pharmaceutical company GlaxoSmithKline [37], following procedures and dosage regimens recommended by the Global Program to Eliminate Lymphatic Filariasis [38]. In the study area, the SHD organized the local program to eliminate lymphatic filariasis through five management levels: the Tamil Nadu SHD, Kallakurichi Health Unit District, primary health centers (PHC), sub-centers, and village delivery units of filariasis prevention assistants.

Drugs for MDA were transported by the SHD to its HUDs, then distributed to their block level PHCs. Village health nurses of sub-centers carried drugs from the block level PHCs to their villages and supervised the most peripheral workers of the delivery chain, the filariasis prevention assistants, who were volunteers. Recruited from village communities, filariasis prevention assistants directly delivered drugs to the households and observed residents swallowing their tablets. In each annual MDA campaign, one day is scheduled for filariasis prevention assistants to deliver to households, and two days are scheduled for tracing and providing treatment of those missed on the preceding day.

Multiple levels of management are required to coordinate delivery of the MDA tablets to residents in the 36 villages investigated. Upper health management levels were involved in more extensive distribution efforts to villages across the State that were not part of this study’s population, although the cost shares of higher-level centers were pro-rated based on their total population coverages. Table 1 shows how a village is integrated into the combined State efforts of all health management levels and associated responsibilities.

The VC intervention delivery system for this CRME research study was less hierarchical than that of the National MDA program administered by the state. Major activities of added VC interventions were carried out solely by the CRME team, in close collaboration with Ariyur block PHC staff members, and other field-level government staff for some part of the monitoring and evaluation efforts. CRME recruited volunteers from villages, trained and motivated them for ‘Vector Control through Community’ [27] to facilitate VC activities by community members. While the two higher administrative levels (SHD and HUD) did not perform VC operations during the study, they would manage these activities should the VC programs be adopted nationwide. To make comparisons of interventions more realistic for future purposes, we simulated costs of senior management level efforts in addition to actual efforts of CRME and the contributions from its associated government management centers.

Description of treatment conditions

The MDA program (intended annually but subject to dates of convenience for administrative and logistical reasons) administered anti-filarial DEC and anthelminthic albendazole to target at-risk communities. DEC, by immune mechanisms still largely undefined, causes rapid clearance and destruction of the filarial larvae (microfilariae) circulating in the blood as well as the slower, progressive death of the adult filarial worms. Albendazole is a broad-spectrum anthelminthic drug that principally affects the adult worms, thereby inhibiting the production of microfilariae and suppressing transmission of LF.

Two interventions were applied for integrated control of the LF vector Culex quinquefasciatus. Against the developing stages of Culex, a floating layer of expanded polystyrene beads (EPB) was applied at rates of 350–400 g/m² water surface area to the main breeding sites (cesspits and wells) to deter mosquito oviposition and asphyxiate larvae and pupae by keeping them submerged [39]. Against the adult Culex mosquitoes, pyrethroid impregnated curtains (PIC) were installed to mosquito landing sites around houses (doors, eaves, and windows) to kill adult insects by contact with insecticidal pyrethroid [40].

Polystyrene resin granules (Shri Sakthi Insulations, Chennai, TN) were boiled to produce the expanded lightweight, ∼2mm diameter EPB [41]. To facilitate the placement and retention of EPB in mosquito breeding sites, it was necessary to modify cesspit frames and unused wells to prevent water overflowing, which would flush away the EPBs. Therefore, cesspits were lined with brick walls, and unused wells were cleaned and covered by a custom-made metal mesh screen in a wooden frame.

The PIC intervention required measurement of the doors, windows and eaves for tailoring curtains to fit these spaces. Curtain materials were made of white woven ventilated polypropylene (VIRGO Polymer India Ltd., Kanchipuram, TN https://virgopolymer.com/fibc/ventilated-fibc) impregnated with deltamethrin insecticide (2.5% SC, Chemet Wets & Flows Pvt. Ltd., Ahmedabad, Gujarat) by soaking and drying in the shade to achieve approximately 25mg ai/m² dosage, before cutting, sewing borders and installation by hanging with nails. Installed curtains had to be re-treated by spraying every six months to sustain insecticidal efficacy (validated by CRME bioassays). CRME implemented both interventions in their assigned villages from March to June 2010.

Collection of cost data

For both MDA and VC programs, personnel costs were calculated using time allocation surveys, salary and wage records, reports, and narrative interviews with paid and volunteer workers. Reports were used to allocate SHD staff time, while time allocation surveys were completed by managers and some community leaders for all levels below the SHD. The salary of each full-time worker and the opportunity cost of each volunteer were collected from administrative records at the SHD, HUD, three BLPHCs, the CRME headquarters and its field station. Narrative interviews clarified implementation procedures and tasks of health workers and community volunteers. Validated costs of major activities of MDA and VC for each management level of Tamil Nadu’s health system were entered into a customized Excel spreadsheet (Microsoft Corp., Redmond, Washington) cost analysis tool developed to combine the costs of MDA and VC activities from different management levels.

Both MDA and VC involved considerable efforts by community members. Current MDA programs engaged full community mobilization efforts once a year, while VC programs required full community mobilization in the initial year and regular maintenance by villagers in subsequent years. Community mobilization efforts for introducing VC were underway in the study villages at the time of the research team’s visit to India. For a narrative understanding, Brandeis investigators with CRME interpreters interviewed participants at every level from the SHD to households.

Non-personnel items consisted mainly of transportation costs (allocated costs of vehicles and fuel) for all activities, drugs for MDA, polystyrene granules, cooking pots and fire for EPB production, construction materials (bricks, cement, tools) for improving cesspits, insecticide, curtains and tools for cutting and sewing them for PIC. We collected information on non-personnel costs of MDA from administrative records from the PHC, interviews with supervisors, and vector control program reports [42]. To estimate economic costs of donated drugs, we asked informants for local retail prices and validated these against the International Drug Price Indicator Guide, 2009 [43]. For VC interventions, CRME documents and officials provided recorded expenditures on non-personnel items. Authors communicated regularly with CRME Field Unit staff (Supporting information S1 Figure and Supporting information S2 Figure).

Analysis of cost data

WHO and its international partners such as the United States Agency for International Development (USAID) adopted a 3-stage “roll-out package” [44] against NTDs, assuming that integration and scale-up of various programs to full-national scale is feasible [45]. Our analysis followed similar assumptions that various approaches to eliminating a single NTD can be integrated and scaled up by the national ministries.

In our current LF study, one arm applied only one intervention (regular MDA). The other two arms used combined interventions: VCS combined MDA with EPB and VCI combined MDA, EPB and PIC. We estimated the cost of each intervention separately and assumed that, for a combined program, we could add the components to estimate the final costs of the integrated approaches. Our analysis of unit costs of each intervention consists of three major steps: (1) estimating total recurrent costs (personnel and non-personnel) by management level, (2) annualizing capital costs, and (3) calculating final unit costs per beneficiary.

Microsoft Excel was used to create two cost analysis tools: mass drug administration cost analysis tool (MDA-CAT) and VC cost analysis tool (VC-CAT), which were used to enter the data and perform cost analyses. Raw data from our data collection instruments provided estimates of aggregate costs of personnel activities and non-personnel inputs. We categorized these costs by each management level, i.e., the State Health Department, intermediary management levels, and household groups. As higher levels served populations of varying sizes beyond the study’s 36 villages (Table 1), we converted the costs at each administrative level to their corresponding per capita amounts, including the administrative costs of hiring personnel and procuring materials for VC and MDA. To compare yearly costs across the implementation strategies, we amortized capital assets over their useful lives.

In contrast to the State-run annual MDA programs, the VC activities initiated specifically for this study were managed by CRME and funded with donor support. Community mobilization was most intense in the first year; project managers estimated that only 15% of the staff’s initial year efforts were necessary in subsequent years when communities were presumed to be self-sufficient for continuing their VC efforts. While MDA required the procurement and dissemination of drugs once a year, the lifetime of installed VC materials varied. For example, EPB required yearly production and reapplication, whereas PIC curtains were expected to last for three years, while insecticide spray on curtains needed to be reapplied every six-months.

Our analysis computed per capita costs based on the number of residents benefiting from each service. For MDA interventions, the combination of drugs served one person. For PICs, the curtain benefited one household (average of five persons) while one cesspit modification served 27 households (135 persons). Most cost data on personnel activities and non-personnel inputs were initially collected in Indian rupees. The project was planned, and many costs were incurred in 2010. We therefore converted Indian rupees to US dollars based on the market exchange rate of 45.8 rupees per US dollar in early 2010 [46] and expressed all costs in 2010 US$.

Analysis of impact

The CRME entomological team collected baseline data of parasitological prevalence and entomological status before the intervention (February 2010) and continued to collect yearly data on the post-intervention progress for three years. We collected data on microfilaria prevalence (MfP) tested by microscopic examination of night-time collection of blood smear samples and AgP using the immunochromatography card test (ICT, BinaxNOW® Filariasis, Alere Inc., Portland, Maine) for MfP and AgP, commonly used indicators to monitor progress towards LF elimination [47–49].

To evaluate the impact of the various control strategies, we first evaluated the impact of the reference strategy, MDA. As MDA is the standard treatment in India, there was on control area without MDA. Instead, we selected as the comparator the historical situation in 2010, as implementation of and compliance with MDA had traditionally been low [50]. We calculated the cumulative improvement from 2010 to 2013 in the share of villages reaching the benchmarks towards elimination of MfP and AgP as public health problems -- MfP<1% and AgP<2%. Dividing by 3 (for 3 years), we then calculated the average annual improvement.

To evaluate the incremental improvement of VC over MDA, we performed logistic regression analyses with these shares of villages controlled based on MfP<1% or AgP<2% as the dependent variable using Stata version 15 (StataCorp, College Station, TX). The independent variables were the final year of intervention (year), the intervention (MDA alone, as the baseline level, VCS, and VCI), and interaction terms. These interaction terms on MfP and AgP (intervention level x year) are the main variables of interest, indicating the differential effectiveness of each combination of interventions after controlling for baseline conditions and general trends. Finally, to contextualize our findings we examined India’s national prevalence of LF from 1990 through 2019, the latest data available.

Cost-effectiveness analysis

For assessing the cost-effectiveness of VC, we converted the average annual improvements from VC by arm into disability adjusted life years (DALYs) averted per 1,000 population. As the first step, we needed the annual DALY burden per person affected by LF. The one study of which we were aware reported an annual burden per case of 0.11 [51]. However, this relatively high value apparently reflected only symptomatic persons, rather than the broader population of persons infected. To derive an estimate for this broader population, we relied on the Global Burden of Disease (GBD) study. Using the most recent year available (2019), we divided India’s rate of years lost to disability (58.96/100,000) due to LF by its prevalence (2.68%) [52], deriving a burden of 0.022 DALYs per person affected. Finally, we divided the incremental cost per 1,000 population by the incremental DALYs averted per 1,000 population to get the incremental cost-effectiveness ratios (ICER) by arm.

Economic cost per person treated

Table 2 summarizes lifespans of the inputs and their target beneficiaries. Table 3 presents the ingredients costs (unit costs and quantities) of each component per 1,000 population at the village level. Table 4 presents the corresponding analysis at a higher level (the block level). Table 5 expands the cost analysis to all levels of the health care system. Table 5 shows that approximately 25% of the economic costs for the MDA program ($110.94 compared to $533.55) were borne by the communities served. From the community perspective, PIC was the costliest intervention, EPB was intermediate, and MDA was least costly. VCI, which combines EPB and PCI, costs $490.93 (i.e., $129.46 plus $361.47) per 1,000 households. This is 4.5 times as much as lowest cost intervention (MDA).

Proportions of inputs and activities

Figure 2 presents the breakdown of costs by activity and intervention. Among the three interventions, PIC is the costliest. Among activities within each intervention, procurement of the materials and delivery of the service (termed procurement and delivery) was the costliest category. The non-personnel costs of VC were much higher than those of the MDA program. As EPB and PIC are supplements to MDA, the arms with these interventions cost more than MDA alone.

• Download figure
• Open in new tab

Figure 2:

Cost per person per year by intervention and activity (US$)

Progress towards LF elimination in the study area

Descriptive analyses showed a decline in microfilaria prevalence rates (MfP) for village groups within each of the treatment conditions from 2010 through the end of CRME‘s collection of entomological data in February 2013. Under MDA-alone, MfP fell from its baseline levels (± standard error of the mean, using the approximation of a normally distributed variable) by 84 (±25) percentage points. Where MDA was augmented by VC, villages with VCS reduced MfP by 75 (±20) and villages with VCI reduced MfP by 90 (±22) percentage points, respectively. All reductions compared to baseline levels were statistically significant [27]. Similarly, the “disease-free” share of population (i.e., living in villages with MfP below 1%) increased from 17% to 83% (66 percentage point improvement) in villages with MDA alone, from 8% to 67% (59 percentage point improvement) in VCS villages, and from 17% to 92% in VCI villages (75 percentage point improvement).

The percentages of villages with elimination are shown in Figure 3 for MfP and Figure 4 for AgP. Figure 5 shows the average AgP by study area and year. Similar to the results for average prevalence, all arms showed substantial improvement over time. However, the ranking of arms varied by year and by indicator, suggesting that no single arm had dramatically superior results.

• Download figure
• Open in new tab

Figure 3:

Share of villages achieving micro-filaria prevalence (MfP) target by intervention arm

• Download figure
• Open in new tab

Figure 4:

Share of villages achieving antigen prevalence (AgP) target by intervention arm

• Download figure
• Open in new tab

Figure 5:

Average antigen prevalence (AgP) by intervention arm

Using the AgP measure for each intervention, the relative reductions from 2010 to 2013 were similar to those for MfP: 70 percentage points for MDA-alone, 89 percentage points for VCS, and 79 percentage points for VCI.

Elimination results and context

Table 6 summarizes the results of the logistic regressions on the effectiveness of the two LF measures. The variable Year2013 is highly significant, indicating marked progress towards elimination in all arms. However, none of the coefficients for interactions between Year2013 and intervention were statistically significant. Furthermore, for each VC arm (VCS and VCI), the coefficients for MfP and AgP were of opposite sign. This lack of consistency and absence of statistical significance indicates that the differences between MDA alone and MDA supplemented by VC initiatives were consistent with chance. That is, at the time and place of the randomized trial, neither VCS nor VCI offered any statistically significant incremental benefit.

As context, Figure 6 presents both the global and India national prevalence of LF from 1990 through 2019 based on the GBD study [52]. Over this period of almost three decades, the prevalence of LF declined by three quarters both globally and in India. Yet the prevalence rate in India remained about triple the global average. The latest (2019) prevalence rates of 2.7% in India and 0.9% globally translate to 37.2 and 71.9 million people affected in India and globally respectively. Of all people affected globally, 52% are in India, indicating that LF remained a serious public health problem in India. By dividing the DALY burden by the prevalence of LF in the GBD study, we found that each person-year of LF carried a 0.022 DALY burden. In other words, every 45 persons with LF constitute the loss in value equivalent to one year of good health (i.e., 1/0.022 = 45).

• Download figure
• Open in new tab

Figure 6.

Prevalence of lymphatic filariasis by year globally and in India, 1990-2019

Incremental cost-effectiveness ratio and summary results

Table 7 derives the ICER of MDA, the one intervention that was significantly effective. MDA significantly improved (lowered) both MfP and AfP compared to initial values 3 years earlier. The ICER for MDA is $112 per DALY. This ICER is substantially lower than India’s 2010 threshold of $1,350.6 per capita GDP in 2010 US dollars from the World Bank [53].