Malaria transmission in Rwanda varies widely. Traditionally, the central plateau (altitude 1,500 to 1,800 m) is considered as one of four distinct ecological zones with overall P. falciparum prevalence rates of 5% to 15% . While transmission in Rwanda is regarded to be stable with seasonal peaks in the valleys and unstable (and potentially epidemic-prone) at higher altitude , a linear correlation between altitude and transmission would be over-simplified: in a recent tabulation of the years 2001-2007, endemicity at 1,600-2,000 m above sea level ranged from hypo-to holoendemic and annual malaria incidences (presumed and confirmed) from 2.4 to 20.4 per 1000 capita . Likely, these figures have declined in recent years [4, 7, 8]. Data from the 2007-2008 Demographic and Health Survey (DHS)  indicate that in >4,600 children <5 years of age and sampled across Rwanda, 2.6% had malaria infection based on rapid Plasmodium lactate dehydrogenase tests. In the present study from the vicinity of Butare (altitude, 1,768 m), 11% and 16% of community children were infected with P. falciparum based on microscopy and PCR, respectively. Data from the 2010 Rwanda Health Management Information System indicate that 13.4% of patients attending health facilities in the Huye district had microscopically confirmed malaria. In the present study, this figure was 12.3%.
Given the scarcity of published community-based data, the reasons for the discrepancy in the prevalence of infection reported in the DHS and observed in the present study are difficult to appraise. Higher sensitivity of PCR as compared to microscopy or rapid test devices [20, 21] may partially be involved. Selection bias during recruitment at home, e.g., due to preferential presentation by the parents of children with (a recent history of) sickness cannot completely be excluded. However, recruitment teams were instructed to select children from households randomly and into pre-defined age strata. Also, most infections in the communities were asymptomatic. One limitation of the present study is its cross-sectional nature by which e.g., seasonal fluctuations are not reproduced. Geographical variation of infection between villages was evident (Figure 2) but attempts to relate this to e.g. altitude or proximity to a water stream, failed. Likely involved, most villages comprise homesteads scattered in the hills rather than agglomerated settlements . Other parameters in the present study, e.g. self-reported bed net use, socio-economic factors, anaemia, were largely in the reported range [4, 19, 22]. Thus, the present data are not representative for the central plateau, let alone Rwanda, but rather provide a detailed and up-to-date picture of P. falciparum infection in southern highland communities and in health facilities serving this population. In contrast, routine health facility based surveillance has clear limitations in providing complete or representative data on e.g., malaria in the community, also because patients lacking access or choosing alternatives are not registered . A low health insurance coverage (43%) in the communities may have deterred parents from seeking formal health care. Such, in turn, could have lead to an over-estimation of disease burden at the community level as compared to the end of the year when more have paid their fees. Nevertheless, only 20% of African children with suspicion of malaria are considered to come to the attention of any formal health system , a figure that might have improved in recent years . Community-based surveys, despite their local limitations, thus provide essential information, also for control campaign monitoring [10, 11].
In the communities, infection prevalence increased from 10% to almost 25% at four years of age, which was not accompanied by a decline in parasite density with age or increase in MOI. The additional age-dependent increase in malaria and trend for declining asymptomatic infections indicate that semi-immunity did not develop to the extent observed in highly endemic areas [25–27]. In line with this, P. falciparum infection including submicroscopic ones had an impact on Hb levels, which exceeds the one commonly seen in children in high-endemicity areas [27–29]. On the other hand, only a quarter of parasitaemic children had malaria, suggesting a majority of asymptomatic infections. Irrespective of the erratic nature of fever in the definition of malaria, the presence of (usually undetected) asymptomatic P. falciparum infections has important implications for malaria control in highland areas. In two sites in highland Kenya, both, high and low levels of asymptomatic P. falciparum infections have been observed among children and adults [30, 31]. Studies from Ghana and Sudan indicate that asymptomatic infections can persist for a year or longer [32, 33]; in highland Kenya the median duration in children aged 5-9 years was five months . Gametocyte carriage was not consistently assessed in the present study but appeared to be low. Nevertheless, individuals with low level, long-lasting, and asymptomatic infections form a major reservoir for transmission [21, 34, 35]. In situations of increased rainfall, higher temperatures, or changed land use such asymptomatic infections may give rise to epidemics which have increased in frequency and intensity in East Africa during the last two decades [36–39]. Targeted antimalarial treatment even of asymptomatic children may thus be a justifiable part of malaria control in highland areas. However, the differing findings on the level of asymptomatic parasite carriage in the present and the two Kenyan studies [30, 31] illustrate that results may not be readily extrapolated.
At the health facilities, roughly half of the children had respiratory tract infections or gastrointestinal problems. Every sixth child at the health centre had malaria and every fifth was P. falciparum infected. Irrespective of the better socio-economic status of the patients' families as compared to the communities (Table 2) this indicates that malaria is among the top three reasons to seek primary health care in this area. At eight percent prevalence, malaria was of lesser importance at the district hospital which receives referrals from several health centres and patients bypassing the referral system by self-paying. There, 38% of the patients were reportedly pre-treated including 6% with AL. The validity of (malaria) treatment histories frequently is questionable , and no data on the dose and duration of treatment were collected in the present study. Nevertheless, the finding that recent AL treatment was positively associated with current malaria is remarkable. This is suggestive of recurrence of parasitaemia following treatment. In fact, drug resistance markers associated with reappearing parasitaemia following AL treatment tended to be increased in these infections (Zeile et al , unpublished observations). Latest cure rates of AL in Rwanda from 2006 have been reported as 97% . Nevertheless, against the background of intense AL drug pressure in Rwanda in recent years, this finding underlines the necessity of the upcoming re-evaluation of the drug's efficacy in this country.
Intake of chloroquine was stated by none of the respondents but the drug was present in plasma in 1.4% and 15.4% of non-infected and infected children, respectively. With the assay applied, chloroquine intake can be detected for several weeks, depending on the dose; cross-reactivity with amodiaquine is negligible . Likely, the finding of an increased infection prevalence in chloroquine positive children reflects the combination of previous home-treatment and persisting or recrudescent parasites due to intense chloroquine resistance which is prevalent in Rwanda .
Among the age-adjusted risk factors for P. falciparum infection was a decreasing MUAC. This crude proxy parameter for malnutrition was, however, not associated with malaria itself. Chronic malnutrition affects every second child in Rwanda  and compromises anti-pathogen immunity . Lacking effect on malaria morbidity as observed in the present study corresponds with previous findings [29, 44] but contrasts with others [28, 45, 46]. Possibly, the differential effect on infection and malaria depends on the specific yet unknown type of malnutrition in the study area. Considering the modifiable nature of this risk factor and Rwanda's recently renewed commitment to fight malnutrition, more research into this field is needed.
At variable statistical significance, several parameters reflecting low socio-economic status were associated with increased risks of infection. Remarkably, possession of a radio or a bicycle were independently associated predictors of reduced P. falciparum prevalence. This may reflect increased access to malaria-related information, improved awareness and increased usage of curative services. Given the latter is true, this points to accomplishable ways of reducing P. falciparum infection in the area, i.e. health communication and education.
Lastly, although self-reported bed net usage was in the previously observed range [4, 19, 22], the detectable impact was modest and non-significant in multivariate analysis. This finding points to deficits in an established mean of malaria prevention the efficacy of which has been confirmed in many studies [4, 7, 10, 11]. The reasons may be diverse and rather involve caregivers' beliefs about causation and vulnerability as well as obstacles in translating knowledge into behaviour than insecticide resistance [47–49]. Nevertheless, these actual reasons need to be assessed at the community-level and subsequent campaigns should address potential obstacles to promote consistent and correct use.