The presented surveys revealed surprisingly low levels of microscopically detectable malaria parasite prevalence at all study sites. Although direct comparison to previous studies is considerably limited by differences in study design (e.g. age groups surveyed) and the fact that surveys were carried out at different months of the year (shortly after the end of the transmission season versus peak of transmission season) and in different villages, the parasite prevalence values reported here are amongst the lowest published, remaining below the level reported from surveys carried out at the end of the non-transmission season. The data presented are thus in line with the decline in disease burden documented previously for the Gambia . The renewed efforts to eradicate malaria , require surveillance systems that strive to identify the last parasite carriers. The usefulness of PCR, RDT and serology was, therefore, explored as techniques for surveillance.
Not surprisingly, the RDT, having a sensitivity threshold of around 500 parasites/μl, which is designed to detect clinical malaria, proved unsuitable to identify parasite carriers in this low transmission setting, where the geometric mean parasite density, amongst those with microscopically detectable parasitaemia, was as low as 104 parasites/μl (CI 95%: 76 - 143).
Thick film slide microscopy, still considered as the diagnostic gold standard for malaria, missed out 66% of parasite carriers identified by nested PCR that has a detection limit of ~1 to 10 parasites/μl [13, 14]. This is in line with a recently published meta-analysis on studies comparing PCR and microscopy for malaria diagnosis in population surveys . This paper further documents, that the proportion of infections missed by microscopy may increase up to 88% where PCR prevalence declines to 10%.
Clearly, the considerable number of individuals with low density parasitaemia missed by microscopy constitutes a major challenge for a malaria surveillance system: given that gametocyte prevalence amongst parasite carriers may be as high as 90% when assessed by Pfs25 QT-NASBA , and that even artemisinin-based combination therapy fails to completely abrogate post-treatment infectiousness [40, 41], the accurate identification of individuals carrying parasites below the threshold for slide microscopy is necessary to determine the remaining potential for malaria transmission in a community. Here, highly sensitive techniques such as PCR will become indispensable.
The decline in burden of disease documented previously , and the reduced parasite prevalence described here occurred after renewed efforts to combat malaria in the Gambia in 2004, such as more efficacious first-line treatment , campaigns to retreat bed nets and ITN distribution in the Western Health Region. However, the temporal association between the observed decline and major control interventions does not necessarily prove a causal relationship. In fact, a longitudinal study from Kenya documented a substantial decline in malaria morbidity and mortality even prior to large scale implementation of control measures .
To assess whether interventions had a sustained impact, surveillance should therefore be able to capture long and mid-term trends of transmission too, which could be achieved by age profiles of anti -malarial antibody prevalence [17, 19]. A successful intervention to control malaria might be expected to result in a distinct "step" in the curve describing the relationship between sero-prevalence and age [18, 43], with significantly lower levels of antibodies in individuals born since the introduction of the control measures. To explore whether such a step in the sero-prevalence profile of MSP-119 antibodies was evident, both models with a fixed ⌊  (force of infection), and "step models" where ⌊ was allowed to vary at one point  were fitted. For all sites, the models using a single ⌊ gave the best fit, suggesting the decline in transmission is likely to have occurred gradually over several years, and the EIR reduction from 4 to 1 over the last 20 observed here for Farafenni supports this. Further examination on the nature of the gradual decline in transmission using a steady reduction in ⌊ did not provide outcomes with a significantly better fit. Further work is needed to define if the existing data are sufficiently robust to allow fitting a variable force of infection and how the decline in the force of infection is best parameterized.
Considering that the Gambian study sites were outside the areas targeted for ITN distribution in 2004, this is plausible and is supported by Multiple Indicator Cluster Survey (MICS) data available for 2000  and 2005/6  in which questionnaire-based data indicated that the proportion of children < 5 sleeping under an ITN has increased from 13.3% to 44.8% in an area prioritized for ITN distribution (Brikama), whereas it only increased from 14.7% to 16.8% in the Basse area, where parts of this survey were carried out. In areas where malaria transmission is highly seasonal, serology offers the advantage of being considerably less dependent of the timing of the survey than direct assessment of parasite burden by PCR or microscopy that can only provide a snapshot.
Regardless of the method used (slide microscopy, PCR or serology), significant variability of parasite (or sero)-prevalence amongst villages of the same site were observed, whereas the variability between sites did not differ significantly. This remained true even after adjusting for the potentially confounding effects of bed net use and ethnicity, suggesting that other factors may contribute to the variability.
The difference in the distance to malaria breeding sites that has been identified previously as a risk factor for malaria in the Gambia , as well as the considerable local variation in both malaria vector density  and vector persistence  may account for some of the remaining inter-village variability. Serology was equally good at detecting the inter-village variability observed by PCR or microscopy. Considering that sero-prevalence is a cumulative measure of exposure over time and less susceptible to transient fluctuations, the inter-village differences described here may indeed reflect a relatively stable pattern. Where complete data are available (Basse, Caio), this notion is further supported by the distribution of spleen rates in 1-15 year old children (Figure 3D).
In view of the planned malaria elimination efforts that will require adequate surveillance, the spatial heterogeneity of malaria transmission observed locally also constitutes a considerable challenge as it renders the identification of a sentinel village that ought to be representative for surveillance purposes of a larger area difficult. Alternative approaches, such as surveys carried out in schools and/or at health facilities which combine PCR/serology with geographic information (GIS) will become invaluable to identify remaining foci of transmission.
Compared to previous reports, a shift of peak parasite prevalence (both by PCR and slide reading) towards the group of 11-15 years old was observed. This is in sharp contrast to a survey carried out in rural Gambia in the dry season of 1950 where over 95% of children < 5 years and 89% of children aged 6-10 years were found to be parasitaemic . Still in October 1988, parasitaemia rates documented for a village in the Farafenni area remained constantly above 60% until the age of 15 years .
While age prevalence profiles of malaria parasitaemia may shift to older age groups with decreasing transmission regardless of its cause, the most obvious reason in our study area probably is the comparably high level of bed net usage in children < 5 years in the Gambia, currently being amongst the countries with the highest coverage of ITNs in children < 5 . As expected, overall bed net-usage had a protective effect with regard to microscopic parasite prevalence. Especially children up to 10 years not using bed nets had significantly higher parasite prevalence. However, no such difference was observed when parasite prevalence was measured by PCR, confirming that bed nets may reduce but not entirely abrogate exposure to malaria. The remaining low level exposure resulting in submicroscopic infections among bed net users may also explain that, with the exception of those < 5 years old, anti-malarial antibodies develop at similar rates regardless whether or not a bed net is used. The notion that anti-malarial immunity can develop despite bed net-usage is consistent with results from longitudinal studies demonstrating that introduction of ITNs  or insecticide-treated curtains  does not increase child mortality in older children. However, the age shift for peak parasite prevalence also means that acquisition of protective immunity may occur later, which finds its expression in an increased mean age of paediatric malaria admissions  - an observation that points out the need to review current guidelines on the management of childhood illness for areas of low malaria prevalence, as these guidelines focus on children below five years .