Following the launching of GPELF in 2000 and the targeting of LF for global elimination, most endemic countries have established national control programmes and many are in the process of implementing annual MDAs with the recommended two-drug combinations . Regular monitoring of the effect of the programmes is essential to evaluate the progress, to make evidence based adjustments and to ultimately end the MDAs when specified programme stopping criteria have been met [13, 14]. Thorough assessments and analyses of the effect of MDAs on transmission and human infection in sentinel sites have been documented, especially from countries using the DEC/albendazole combination such as Egypt , Papua New Guinea [16, 17], American Samoa [18, 19], India  and Samoa . Many LF endemic countries in sub-Saharan Africa are co-endemic for onchocerciasis, and due to the potential risk of DEC induced side-effects in individuals with Onchocerca volvulus infections, use a combination of ivermectin and albendazole in their MDAs for LF control. The effect of this drug combination in a low-moderate LF endemic area of central Nigeria, West Africa, was recently documented [22, 23]. Here we present and analyse the effect of six rounds of MDA with ivermectin and albendazole in a highly endemic area of Tanzania, East Africa.
The coastal area of north-eastern Tanzania is well known to be endemic for LF, and has been a focus for comprehensive research on epidemiology and measures for control of LF in the past e.g. [24–30]. Prior to start-up of the NLFEP-activities in Tanga Region in 2004, no control had taken place in any of the sites used in the present study. In consistency with this, it was noted that pre-MDA indices of LF infection and transmission were very high and resembled those reported during earlier studies in nearby areas. High levels of clinical manifestations were also recorded in Kirare study community, and were moreover strikingly visible when moving around in the area.
The study design and methodology and the effects of MDAs in the early study period was reported previously for Kirare  and the 10 schools . Some changes in study design and methodology in the community study were necessitated in later study period. First, increasing study fatigue among inhabitants of Kirare resulted in decreasing survey coverage from one MDA to the next. This was not only due to the yearly blood sampling, but also to some political rivalry among groups of villagers. As a consequence, it was decided to leave out two of the village hamlets and instead focus more attention on the remaining two hamlets where most cooperation was met. To compensate for the smaller survey population in Kirare, two other communities were included, whereby also a larger geographical area was covered. Second, due to the reduced human infection burden after the first MDAs there was a need to adopt a new and more practical diagnostic approach. The primary community screening tool was therefore changed from mf-testing to CFA-testing by ICT cards , followed by examination of only CFA positive individuals for mf. As it was still essential to have an idea about the mf prevalence and intensity in the community (as indicators of availability of infection to the vectors), methods for calculating these indices based on the new screening procedure were developed. Third, the availability of a new commercial kit for examination of dried finger prick blood spots on filter paper eliminated the need for venous blood sampling in the later part of the study and made it practically feasible to examine larger populations for antibodies to Bm14. The changes in methods for measuring both CFA and antibodies to Bm14 are not likely to have had a major impact on the obtained results, but should still be kept in mind when analysing and interpreting the findings. In contrast to the opposition sometimes experienced from villagers during the human surveys, the trapping of mosquitoes in the 50 houses of Kirare generally proceeded smoothly throughout the study period.
In Kirare, the mf prevalence and community mf GMI generally decreased progressively with increasing number of MDAs. An exception to this was survey 5, where slight but statistically insignificant increases in these indices were seen, most likely due to the long interval between MDA 3 and 4 (21 months). An increase in mf and CFA was reported from Haiti after a single MDA with DEC/albendazole had been missed , and both of these observations therefore seem to emphasize the importance of keeping a relatively short spacing between the MDAs. Decreases in mf prevalence and community mf GMIs were also obvious between survey 7 and 8 in the two new communities. Mf GMIs calculated on the basis of mf positive individuals generally remained high throughout the study, suggesting that some mf positive individuals regularly either avoided or were left out from treatment during the MDAs (systematic non-compliance). These individuals comprise an important source of mf for the continued transmission of LF.
CFAs are primarily released by adult W. bancrofti, and their presence is a sensitive and specific indirect measure of adult worm infection . In the early study period, when CFAs were measured by ELISA, the decrease in CFA prevalence in Kirare was small. A more substantial decrease was seen in the GMIs, most likely because intensities were extremely high in the early period of the study. Thus, although intensities decreased they only reached the cut-off level for positivity in a few individuals . In the late study period, when CFAs were measured with ICT cards, prevalences were substantially lower, and a continuing and statistically significant decrease was seen between survey 7 and 8 in all three communities. The two tests for CFA detection (ELISA and ICT cards) have been shown to be well in agreement when measuring CFA status . Although the community CFA prevalence decreased considerably from the pre-MDA survey to survey 8, it was still quite high after 6 MDAs.
The specific antibody response to the filarial antigen Bm14 is used as an indirect marker of exposure to W. bancrofti transmission [21, 35]. In the early part of the study the Bm14 antibodies were measured with ELISA (by using antigen donated by Prof. G. Weil through the NIH/NIAID Filariasis Research Repository Centre, Smith College, USA) and all age-groups were examined, whereas in the later part of the study tests were performed with a commercially available test-kit (with antigen from the same source) and only included children aged 5-14 years. Although the different techniques used and age groups examined prevent exact comparisons, the very pronounced decrease in both prevalence and OD-value GMI between the pre-MDA survey and survey 8 in Kirare indicated that a major decrease in exposure to transmission had taken place in the human population, which is also in alignment with the findings from the entomological surveillance.
The vector species composition changed considerably during the study period from predominantly anopheline during the pre-MDA period, to almost exclusively culicine in the late study period. This remarkable shift has been documented and analysed in more detail elsewhere [36, 37]. With respect to the role of the MDAs, there is documented evidence that anopheles mosquitoes taking a blood meal on ivermectin treated individuals show increased mortality , and that MDAs for LF in this way may affect transmission of malaria . It is, however, difficult immediately to accept that the MDAs should be a primary cause for the shift, and a multitude of factors, including those related to environmental and climate change may be involved. Whatever the reason, there can be no doubt that the shift has important consequences for the epidemiology and control of LF, as the anopheline and culicine vector species differ in biology, biting habits and vectorial capacity. In addition to the shift in species composition there were marked seasonal fluctuations in vector abundance, with most vectors being present during and shortly after the two annual rainy seasons. Transmission was also most intense, and during the late period of the study only observed, in these seasons.
Transmission, assessed by dissection of vectors and recovery of L3s, decreased after each MDA, from high levels in the pre-MDA period to very low levels in the post MDA 6 period. It is likely that most of the decrease in transmission was due to the MDAs, which reduced the availability of microfilariae in the human population to the vectors. However, part of the decrease may also be due to the shift in vector species composition, whereby long-lived strongly anthropophilic anophelines (and thereby highly efficient vectors) were replaced by shorter lived and less strongly anthropophilic culicines (and thereby less efficient vectors) . In the later period of the study, only very few infective mosquitoes were identified despite large numbers of vectors being dissected. Collection and dissection of vectors is a time-consuming and expensive task, but it provides the most exact information about vector dynamics and transmission of LF and as such is an important research tool for detailed investigation of transmission . Newly developed PCR techniques may replace the laborious dissection activity in future assessments. However, the number of mosquitoes to be collected, and the efforts to be invested in collecting them, in order to get statistical reliable information about change in transmission intensity in the late period of control programmes may be enormous. In the present study, where mosquitoes were collected one night each week from 50 houses by use of light traps, too few were obtained to demonstrate a statistical change in transmission after post MDA 3 (although a decreasing trend was seen). Under these conditions, either much more effective mosquito collection methods are needed or indirect assessment methods such as measurement of CFA or antibodies to Bm14 in the human population (especially children) have to be employed.
The school based study was used as an alternative approach to assess the effect of MDAs on transmission in the area. The idea being that a reduced transmission will lead to reduced acquisition of infection in young children, and thereby a reduction in prevalence of CFA when examining children of the same age from year to year after start of MDA . Indeed, after the first three MDAs the CFA prevalence in the new intake of Standard 1 pupils decreased rapidly, and thus provided a strong indirect indication that children were less and less exposed to infection from year to year after start of MDA. Further decrease in CFA prevalence was noted in the last two surveys, but the effect levelled off and the prevalence was still 5.6% in survey 8. The majority of children examined during the last two surveys were aged one year or less when the first MDA was administered in 2004 (67.7% in survey 7 and 90.7% in survey 8) and among these the CFA prevalence (5.9% in survey 7 and 5.1% in survey 8) was comparable to the overall CFA prevalence for these surveys. Examination of the Standard 1 pupils from four of the schools for Bm14 antibody in survey 7 and 8 showed relatively high rates of positivity. However, the significant reduction in both Bm14 prevalence and OD-value GMI between these two surveys suggests that transmission was on the decrease.
Sustained high drug-intake coverage during MDAs for LF control is critical in order to reach the programme target within a reasonable time-frame, in particular in areas like the present with very high pre-MDA levels of infection [41, 42]. Treatment coverage and compliance are therefore important factors to consider when monitoring and assessing the impact of MDAs, but reliable information is difficult to get. In the present study we used three different approaches to assess treatment coverage, namely the official “reported coverage” from NLFEP for Tanga District and two “surveyed coverages” obtained from interviews carried out shortly after each MDA in the study communities (individuals aged 5 years and above) and the study schools (Standard 1 pupils), respectively. Apart from MDA 1, where all three approaches indicated about 80%, the surveyed coverages were considerably lower than the reported coverage, in particular for the Standard 1 pupils. The discrepancy in treatment coverage obtained by different approaches in this area, and the possible courses, was recently addressed in detail in a social science study , and only few comments related specifically to observations at the present study sites will be added. Firstly, there were frequent complaints from inhabitants in the study communities that they were not well informed by the programme about the drug distributions beforehand, neither regarding timing or purpose. Second, it was noted by the research team that the method for drug distribution varied between different MDAs and different sites, and obviously house-to-house distribution was more effective than distribution from a central village point. The very low coverages among the Standard 1 children, especially during the last three MDAs were noted to be related to: drug distribution during school holidays, some schools had no distribution point nearby, and parents refusing treatment of their children due to rumours of side effects. Finally, the activities of the research project probably created extra awareness about LF and its control in the study communities, which may have resulted in higher treatment coverages during MDAs than in other communities in the area. These important issues should be addressed in subsequent treatment rounds, in order to ensure sufficient treatment coverage, and thereby effectiveness, of the MDAs.