Lymphatic filariasis (LF) is a mosquito-transmitted parasitic disease that is ranked by the World Health Organization (WHO) as a leading cause of permanent disability worldwide. LF affects an estimated 120 million people in 81 countries, with over 1 billion, or one-fifth of the world's population, at risk for filarial infection due to their exposure to infective larvae through the mosquito vector [1]. LF causes debilitating genital disease (hydrocele) in an estimated 25 million men and lymphedema or elephantiasis of the leg in 15 million people, mostly women [2].

The Global Programme to Eliminate Lymphatic Filariasis (GPELF) was established in 2000 with the goal of eliminating LF as a public health problem worldwide by 2020. The programme is centered on annual mass drug administrations (MDAs) which are based on the community-wide distribution of albendazole plus either diethylcarbamazine (DEC) or ivermectin to all those at risk in an endemic community. These drug combinations suppress the parasite in the blood, thereby reducing the transmission potential of the parasite, and also kill a broad spectrum of intestinal worms [3]. The WHO currently recommends mass treatment in settings where the prevalence of antigenemia is ≥1%. [4] For endemic countries, it is thought that 4–6 rounds of MDA, with at least 60–70% compliance, are adequate to interrupt transmission.

GPELF has scaled up impressively since it began. By the end of 2007, 48 countries had implemented elimination programs and approximately 570 million people had been treated. [1] In Haiti, the national program began in 2001 after a demonstration project in the Leogane commune started in 2000. The program conducted an initial LF infection survey by testing schoolchildren for the presence of filarial antigen in each of the 133 communes in the country. The results of this survey were compiled into an infection map [5]. Treatment of the hyperendemic communes (>10% antigenemia in children) has been the highest priority of the National Programme for Elimination of Lymphatic Filariasis (NPELF). Annual MDA in Haiti was gradually scaled up between 2000–2005 in all hyperendemic communes, except those areas where political instability in 2004–2005 threatened the safety of distribution staff (e.g. Port au Prince and Gonaives). Because of an interruption in funding there was no MDA in Haiti in 2006, but MDA resumed in 2007.

The current benchmark for success, as defined by the WHO, is a microfilaria prevalence of less than 1% in a community. Microfilaremia is indicative of the presence of motile larvae of the parasite (microfilariae) in the blood stream. If microfilaremia in a community is less than 1%, it is thought that transmission of the parasite cannot be sustained and there should be few or no incident infections. In order to measure the effect of the MDAs on transmission, and thus their effectiveness, surveys were conducted at sentinel sites around Leogane commune [6]. In 2005, the microfilaremia in Leogane was below 1%, reaching the WHO benchmark for successful interruption of transmission [7]. In these settings, current WHO guidelines recommend conducting cluster surveys of young children (2–4 years) to confirm the interruption of transmission and determine whether MDA can be halted [8]. In the absence of transmission and after at least 5 rounds of MDA, children under 5 years of age are expected to be antigen and antibody-negative. In August 2007 a survey was conducted of children 2–4 years of age, residing in Leogane. Blood samples for antigen and antibody testing were obtained. The results of the survey showed 14% antigen prevalence by ICT test and 19% antibody prevalence in children ages 2–4 years (n = 304) (Donovan et. al, unpublished data). These findings indicate that six rounds of MDA have not been sufficient to interrupt transmission in Leogane.

Although the WHO recommendations do not include monitoring of mosquito infection, information regarding the persistence of infection in the mosquito population is useful in determining whether transmission is ongoing. Persistence of filarial DNA in the mosquito population could be an indication that transmission has not been interrupted. The current study was designed to confirm continuing transmission of LF using blood and mosquito diagnostic tools and to analyze factors that may be contributing to the continued transmission of LF in Leogane, including population migration and systematic noncompliance.

In Haiti 7 rounds of MDA have been delivered to the commune of Leogane, with a missed round of MDA in 2006. By 2005 MF prevalence was below the <1% threshold that the WHO recommends for stopping MDA. However, subsequent studies in children suggested that transmission was ongoing despite the apparent success of the MDAs. The aim of this study was to determine the transmission status of LF in Leogane and to examine possible factors contributing to transmission in the area.

The levels of microfilaremia and antigenemia found in the communities in this study (with the exception of Corail Lemaire) indicate that transmission is still ongoing in Leogane. Only Corail Lemaire had MF prevalence below the <1% WHO threshold for stopping MDA. Antigen prevalence by both the ICT and Og4C3 test in children 3–5 years of age was above 10%. If transmission had been interrupted there should be little to no infection in this age group as they were born after the MDAs started. In addition to the high prevalence of microfilaremia and antigenemia in most of the communities surveyed, there was a high rate of infection in mosquitoes collected from those communities. The presence of a significant number of infected mosquitoes further supports the conclusion that transmission is ongoing. Although the MDA rounds have succeeded in lowering the overall prevalence of LF infection in Leogane, they have not succeeded in interrupting transmission of the parasite. Initial projections for the LF elimination predicted that 5–6 rounds of MDA would be sufficient to interrupt the transmission cycle of the parasite [19]. Since transmission was persistent in Haiti after more than the recommended rounds of MDA, migration, knowledge of LF, and noncompliance with MDA were considered as potential factors for ongoing transmission.

Of the factors examined, only systematic noncompliance was statistically associated with infection status. Noncompliance has been previously reported in Haiti. After 3 rounds of MDA in Leogane, 18.6% of adults surveyed had not participated in MDA [20]. Talbot et al found levels of noncompliance around 25% after 4 years of MDA [21]. The overall noncompliance rate in this study was very similar (24.2%). This noncompliance rate yields a compliance rate of around 76%, which is consistent with the WHO recommendation of >60–70% compliance for interruption of transmission. In Leogane, this level of compliance with MDA has not been adequate to interrupt transmission of LF.

These results suggest that a consistent proportion of the population has not been mobilized to participate in MDA. Social mobilization strategies employed by the program clearly are not reaching this segment of the population. It is important to acknowledge the limitations of questionnaire-based coverage surveys. It is possible that people do not remember taking the drug in previous years. Nonetheless, infection levels were significantly higher in noncompliant individuals as compared to compliant individuals, demonstrating a biologic correlate of the nonparticipation of the respondents in MDA. High rates of noncompliance maintain a reservoir of infection, which drives LF transmission.

Systematic noncompliance has been examined as a factor in MDA success in a number of different national elimination programs. In Egypt the noncompliance rate was very low (7.4%) and consequently, the program was able to successfully reduce infection levels to a point where transmission was probably interrupted [22]. Although residual infection rates were the highest in noncompliant individuals, the authors also observed a trend towards reduced infection rates in those individuals, indicating the possibility of a “herd treatment effect” as transmission levels decline [22]. These results are very encouraging for the Global Elimination Programme but they may not be transferable to Leogane where the baseline prevalence rates of MF and antigenemia were considerably higher than in Egypt. Also, the mode of drug distribution is different. In Egypt, health workers go door-to-door to distribute drugs, whereas in Haiti, community distribution points are used. A much higher compliance rate was seen with the door-to-door distribution in Egypt and that, combined with the lower baseline infection rates, led to the impressive results that were observed.

Other countries have experienced noncompliance rates more similar to that of Haiti. The compliance rates in India have been consistently low. In the southern state of Tamil Nadu compliance rates ranged between 46% and 64% [23],[24]. A 2005 study in Tamil Nadu found that only 30% of the study cohort complied with all six rounds of MDA and a study in Orissa state found 83% of the population had received the drug but only 49.5% consumed them [25],[26]. The picture is similar in other areas of the world. Recent studies in the Philippines and the Colombo district of Sri Lanka found noncompliance rates of 30% and 28% [27],[28]. However, it is important to distinguish between noncompliance for a single or several MDA rounds and systematic noncompliance (never taking a drug for LF). Sporadic noncompliance is problematic as it may reflect faults in the distribution system or education message of the MDA. However, if there is only sporadic noncompliance, all individuals in a community will have been treated at some point in time and the development of the “herd treatment effect” referred to by El-Setouhy et al will likely lead to an overall reduction in infection prevalence and interruption of transmission. Systematically noncompliant individuals continue to provide a reservoir of MF and perpetuate the transmission cycle. This is illustrated in our study with the high rates of systematic noncompliance, filarial infection and mosquito infection and the significant association between systematic noncompliance and filarial infection. Systematic noncompliance not only reflects possible weak points in the MDA program but threatens to undermine the program's goal of eliminating LF by interrupting transmission.

With similar rates of noncompliance in various countries, is there an underlying determinant or determinants that influence compliance with MDA? In the Philippines and India the perceived benefit of the MDA was associated with compliance. Knowledge of LF was found to be linked to MDA compliance in the Philippines and in Haiti [27],[29]. A KAP survey conducted in Haiti in 2000 as well as follow-up survey in 2004 found that women were more likely to be noncompliant [29]. This observation was also made in India [23]. In Haiti, the difference in compliance between the sexes was most likely due to the initial exclusion of women of child-bearing age from albendazole treatment during the first two rounds of MDA. This policy was reversed in 2002 and no discrepancy in compliance status was seen between males and females in our study. Other factors previously found to be associated with compliance status were ability to swallow pills and the perceived status of the interviewee in the community [21]. In the current study, the majority of noncompliant individuals cited “Don't know” as the reason for not participating in the MDA. The second highest response was fear of side effects (Table 4). While the side effects have reduced significantly in Leogane over the course of the MDAs [30], there seems to be residual concern based on the anticipation of side effects to the drugs. Equally as important are the “Don't know” respondents. Is this a proxy response for “Don't care” or is there some other determinant of compliance status that is not being captured in this or previous surveys? Averted cases of lymphedema and hydrocoele are hard to quantify and since there is little direct evidence that the MDA provides clinical relief for those chronic conditions, community members may not perceive any benefit to participating in MDA. Given that the age group with the highest noncompliance was 3–5 years of age, the de-worming effects of albendazole do not seem to be a major driver for parents to have their children participate in MDA. This could be due either to lack of exposure to messages regarding the de-worming properties of albendazole or a lack of understanding of these messages. Renewed health education efforts could provide incentive to participate by emphasizing the prevention of future lymphedema and hydrocele cases and the benefits of de-worming from albendazole. While health education messages were highly publicized at the beginning of the MDA cycle, the de-worming effects of albendazole were not emphasized. This added benefit could induce more individuals to participate, especially mothers and their children.

The variability of infection levels between communities (Table 2) emphasizes that infection is focal in nature, a reflection of poorly understood differences in mosquito habitat and density as well as host factors. There is previous evidence for spatial variation in LF infection. Ramzy et al reported non-uniform infection in LF [31]. A study done in 2001 in Papua New Guinea found micro-spatial heterogeneity in LF infection [32]. In Haiti, a 2003 study found spatial clustering of antigen positivity and IgG1 positivity [33]. The authors of that study concluded that the transmission dynamics of LF in Leogane may vary over as small a distance as tens of meters.

Using any of several definitions of infection status (e.g. antigen by ICT card or antibody positivity) we found evidence for clustering of filarial infection by the nearest neighbor analysis (Table 6). In contrast, noncompliance was not found to be spatially clustered. Why noncompliance would be statistically related to infection but not show the same spatial relationship as infection is unclear. There may be a spatial relationship to noncompliance that was not captured by this study, perhaps because of our sampling design, and further investigation of this relationship is warranted, specifically to determine the effect of a noncompliant individual on the infection status of his/her neighbors.

Although noncompliance was the only factor that was significantly related to infection in our study, it may not account for all of the transmission seen in Leogane. Initial infection prevalence, vector density, biting density and topography all play a role in infection. In a 2008 paper that used statistical methods to define the Risk of Infection Index (RII) based on community microfilariae load (CMFL) and vector density per man-hour (MHD), the authors concluded that transmission may continue in areas where MF prevalence is low but vector density is high [34]. In such situations it may be cost effective to use vector control, in contrast to areas of higher MF prevalence and low vector density where MDA may be the more cost effective tool [34]. The mosquito data collected in this study provides information about infection rates but not about vector and/or biting density. It was also not possible to make any conclusions about the vector-parasite relationship from this study. Investigations examining the effect of increases MF loads on vector survival have reported conflicting results and the exact relationship between W. bancrofti and its Culex vector is not clear [35]. Further studies are needed to examine vector issues in Haiti in order to determine the impact of vector ecologies on local transmission, and the nature of the parasite-vector relationship. This additional knowledge would aid in determining if vector control would be a cost effective measure to interrupt transmission there.

Population migration was not found to be a significant contributor to transmission in this study; however, this could be a consequence of the timing of the study. It is thought that the instability in Haiti between 2004 and 2005 led to a migration of individuals out of affected areas such as Port-au-Prince to less affected areas such as Leogane. Port-au-Prince has never undergone MDA, and these individuals could have represented a reservoir of infection. Since this study was undertaken several years after the unrest many of those individuals could have returned to their place of previous residence. If this scenario was true then these individuals would not be captured in the study. Thus, population migration may still be a factor in the ongoing transmission of LF in Leogane but was not reflected in this analysis.

An added factor contributing to transmission of LF is the missed round of MDA in Haiti in 2006. A study conducted in September 2007 by Won et al argued that this missed round of MDA was responsible for a rebound in infection to levels that were present in 2003 [36]. This rebound in infection rates underscores the importance of maintaining the MDA schedule and ensuring compliance in the population.

The LF elimination program in Leogane has been ongoing for eight years. The program has been successful in reducing MF and antigen rates from baseline levels over the course of seven rounds of MDA. Despite these achievements, and despite reaching the WHO benchmarks for success, there is ongoing transmission of LF in Leogane commune. It appears that one of the main contributors to this transmission is individuals who have never participated in the MDA. They may provide a pool of infection by which the mosquitoes become infected and the parasite is transmitted to others. The reasons for noncompliance are not completely clear. Also, noncompliance is one of many factors that could play a part in transmission. Other factors include vector density, a missed round of MDA, and the heterogeneity of transmission. New tools and approaches are needed in this environment in addition to the further studies recommended above. Increased health education and awareness campaigns may improve compliance. The addition of vector control methods such as insecticide treated bed nets could provide the extra push needed to stop transmission. It is also possible that DEC-fortified salt represents a programmatic alternative that should be re-visited. The incorporation of new tools should be investigated and implemented so that the program in Haiti can proceed toward elimination. The situation in Haiti is not dissimilar to that found in other parts of the world. Haiti can be used as a model for LF elimination in areas of high infection prevalence and high vector pressure. Understanding obstacles and solutions from the program in Haiti could be helpful for elimination programs in other countries.