The success of MDA depends on its capacity to reduce the prevalence of pathogens below a critical threshold level where re-emergence is unlikely and control can be maintained [
52]. For malaria, this threshold prevalence is very low; the basic reproductive number (R
0), the number of secondary episodes arising from a single infectious human case in a fully susceptible population, can be well over 100 if malaria vectors are present [
53]. The chance of achieving this threshold is highest when parasites are cleared from the human population towards the end of the dry season when very few mosquitoes are present [
38]. The MDA intervention was completed three weeks before the increase in mosquito numbers following the seasonal rains and used a gametocytocidal drug combination of SP+AS+PQ that clears gametocytes in an average of 6.3 days; significantly quicker than after ACTs alone [
54]. Despite the demanding three-day dosing scheme and logistical challenges to exclude pregnant women from treatment with AS and PQ [
55,
56], our coverage was high compared to that of previous malaria MDAs. Of the total population living in intervention clusters, 94% received at least one dose and 93% received a complete dose of efficacious anti-malarial drugs in the month preceding the start of the transmission season or immediately upon arrival in the area. In The Gambia in 1999, a coverage rate of 73% was achieved after one and of 88% after two rounds of drug administration. The drug combination given was SP plus a single dose of AS, given on a single day [
36,
57]. Coverage of other reported MDA campaigns in Uganda (50%) [
34], Nicaragua (70%) [
58] and Garki, Nigeria (85%) [
59] were lower than that in the current setting. The successful malaria eradication campaign on Aneityum, Vanuatu, used 9 rounds of MDA in a population of 718 individuals with participation rates of 79-92% for each round [
33]. The maximum number of tablets that had to be taken by adults on Aneityum was 13 on a single day, compared to 7 per day on the first two days and 10 on the third day in the current study. Although this indicates that a complicated dosing scheme and high tablet number do not preclude high participation rates, the number of tablets was mentioned as burdensome by some participants (Shekalaghe, unpublished data) [
33,
60].
The intervention strategy that involved local village leaders and village representatives as part of the intervention teams is likely to have contributed to the high level of community participation [
60]. The stepwise consent procedure and repeated village meetings gave inhabitants of the study area sufficient time to consider participation and opportunities to raise questions. A study on community perceptions of MDA in The Gambia concluded that knowledge about malaria was related to a higher individual participation rate and that people who had discussed the study with other villagers and those who understood the necessity for a high participation rate for MDA to work were also more likely to participate [
57]. During community meetings the necessity for asymptomatic individuals to participate in order to achieve community benefits was strongly emphasized, although no attempt was made to quantify community understanding. The approach to give local leaders a central role in explaining the study purposes, drug administration and monitoring migration will also have influenced the response rate. Several opinion makers in The Gambia were antagonistic to the intervention and reduced participation rates [
57]; the current study received full participation of all 62
balozi's in the study area.
Despite its high coverage, no impact of MDA on the transmission in the area was observed. The study area was selected for its very low transmission intensity and marked seasonality with almost complete absence of vectors for several months per year [
40]. The site selection and trial design were based on the assumption of a sub-microscopic parasite carriage of ≥ 30% [
28], that would maintain malaria transmission in the area [
31] and would form a suitable primary outcome measure for the intervention. Contrary to these expectations, malaria had all but disappeared from the area prior to the MDA intervention during which no major additional malaria specific control efforts were undertaken. Considerable reductions in transmission intensity without evident underlying causes have been reported across northern Tanzania [
5,
61,
62]. Reported use of insecticide treated net (ITN) in this period increased from 25.1% (490/1952) in 2004 [
28] to 36.1% (879/2434) in 2007. If these numbers reflected a change in effective ITN use, this could have resulted in a substantial reduction in parasite prevalence in the study area of low endemicity [
63] although coverage remains too low to explain an almost complete elimination of malaria in the area [
64,
65]. Presumptive treatment with ACT could also have contributed to the decline in transmission intensity. Although treatment prescribing behaviour in the area was not quantified, presumptive treatment with artemether-lumefantrine, introduced as first-line treatment in 2007, was common practice and will have provided parasitaemic and non-parasitaemic individuals with a curative dose and effective prophylaxis for several weeks [
18,
66]. Rainfall was unreliable in the period preceding the MDA intervention; being 34% of the preceding 10-year average in 2004, 35% in 2005, 116% in 2006 and 66% in 2007. A single treatment campaign for trachoma with azithromycin was undertaken by a non-governmental organisation with an unknown coverage level. Azithromycin has some anti-malarial activity at doses given for trachoma [
67]. These components may all have contributed to the fall in transmission intensity observed. The findings from this study also illustrate the limited understanding on the importance of sub-microscopic parasite densities for maintaining stable malaria transmission. In 2005, a large-scale survey was conducted in the study area and detected
P. falciparum in 32.5% of the general population by PCR compared to 1.9% by microscopy [
28]. The large proportion of sub-microscopic parasite carriers is typical for areas of low endemicity; in areas where the microscopic parasite prevalence is below 10%, microscopy on average detects just under a quarter of all parasite carriers [
30]. Although low-density infections can be highly efficient in eliciting protective immunity in experimental infections [
68], their dynamics are poorly understood in endemic settings. Harris and colleagues observed that in a low endemic area on the Solomon islands, parasites were carried at microscopically detectable densities by 2.2% of the population of whom > 85% were afebrile. The majority of parasite carriers in all age groups had parasites at densities below 500 parasites/μL and parasite prevalence increased almost four-fold when PCR was used for detection [
69]. Low parasitaemia and frequent asymptomatic parasite carriage was also reported for an area of very low endemicity in Vanuatu where parasite prevalence by microscopy was 2-3% and > 80% of the parasite carriers harboured parasites asymptomatically and at a densities below 500 parasites/μL or below the microscopic threshold for detection [
70]. It is currently unknown how individuals with limited cumulative exposure to malaria manage to control malaria at low parasite densities or how long sub-microscopic infections persist. The average duration of malaria infections has been estimated at 210 days in area of intense malaria transmission [
71] where most infections are microscopically detectable [
30]. The duration and gametocytaemic period of sub-microscopic infections, that dominate in low endemic areas, are currently unknown. These factors are crucial in determining the relative contribution of sub-microscopic infections, that are sufficient to infect mosquitoes [
26,
29], to malaria transmission in a population.