During the nine years of the study, the global incidence rate in children of Camopi remained very high with two epidemics in 2004 and 2006, reaching 800 and 1,000 per 1,000 person-years respectively. During the first year of life, the incidence was quite low presumably because the presence of maternal antibodies provides partial immunity. At one year of age, the incidence began to increase sharply to reach a maximum between two and three years old (almost 1,000 per 1,000 person-years). Immunity seems to begin to develop at this age, for both species, but appeared to be mainly protective against P.vivax relapses. Amerindian children have different activities as they grow and gain independence in their movements in the village. Moreover, around seven or eight years of age, boys begin to go to the forest for hunting and to the river for fishing whereas girls accompany their mother to slash. Hence, around five years of age, a different exposure may be responsible for the small increase of P. falciparum.
The survival analysis showed a strong association between environmental exposure and malaria transmission in children under seven years old. Indeed, the analysis permitted to identify several risk factors linked to the environmental characteristics of the surroundings of the house. Malaria transmission increased with distance from the central main hamlet of Camopi. Therefore, children living in isolated hamlets had an increased risk of malaria. However, a higher number of people in the same home (more than seven occupants) were independently associated with a higher risk of malaria. Human aggregation is likely to increase the probability for vectors near homes to be infected [4, 15].
Regarding other environmental factors, the proportion of cleared vegetation within 50 m around the houses was a protective factor for malaria, as previously described. Indeed, this characteristic of land cover is not favourable for the rest of adults and the maintenance of breeding sites. Proximity to the forest was associated with a higher risk of malaria. According to other authors, when houses are located not far from the forest, An. darlingi returns to the forest after feeding [16–19]. However, non-environmental factors may be partly responsible for the relation between increasing incidence and the distance from the main hamlet of Camopi. There was a significantly different risk of malaria according to the ethnic group in univariate analysis. A previous study in Camopi found a strong association of ethnicity with first malaria attack, even after adjusting for behavioural and environmental factors . These results could suggest that Emerillon children have higher genetic susceptibility than Wayampi children. However, this phenomenon was not visible in our multivariate analysis, which did not take into account P. vivax relapses and thus was then closer to the transmission phenomenon.
Furthermore, another variable is likely to better explain the malaria incidence while being correlated to ethnicity. This is the case of the variable "river" divided in three groups: the upstream Oyapock riverside where the Wayampi live, the downstream Oyapock where the mixed ethnic groups live and the Camopi riverside where the Emerillon live. The behaviour of the residents and the protection measures used also play a role in the incidence of the disease. Indeed, the use of topical repellents and domestic insecticides and interventions of the county mosquito control service were significantly associated with a lower risk of transmission in univariate analysis. In addition, children who used to go to sleep after 7:00 pm had a higher risk of transmission. It is likely that children going to sleep earlier are protected by mosquito net at dusk, when An. darlingi reaches its first peak of activity .
Climatic and hydrologic variations appeared to have an impact on malaria incidence at relatively short-term (lag 0 to lag 3) and at longer term (lag 9 to lag 12). Considering the short- term effect, some plausible explanations can be put forward. The mean minimum temperature was globally positively associated with malaria incidence one month later and the minimum temperature was positively associated with malaria incidence three months later. Moreover, the mean temperature was positively associated with a higher incidence that occurred two months later, and especially at the beginning of the rainy season. These observations could be explained by the fact that an increase temperature shortens the interval between egg-laying episodes and enhances the larval development. Moreover, a higher temperature is also likely to accelerate the sporogonic cycle of the Plasmodium. Thus, a high temperature may allow better survival of vector populations and therefore a higher transmission that could be responsible for an increasing incidence in the following months. Conversely, lower minimum temperatures may be responsible for a decreased incidence, slowing the sporogonic cycle of the parasite and decreasing vector survival. Regarding hydrological factors, incidence rates were positively associated with the maximum river level at the same month and one month earlier. Thus, particularly high water may create flooding on the river bed, leading to the creation of suitable larval breeding sites, particularly at the end of the year. This phenomenon has been observed along the Maroni River and in other countries of Central and South America [20–24]. Thus, an increase in the larval anopheline abundance may increase the malaria transmission related to the adult stage. This has been previously observed at a weekly temporal resolution in Camopi where vector abundance was positively correlated with the river level a few weeks earlier .
Regarding the long-term impact of climate on malaria incidence, it is difficult to grasp the meaning of these correlations and the statistical results have to be considered with caution. The significant results could be due to unidentified confounding factors or residual effects of seasonal factors that are not taken into account at a short term. This long-term hypothetical effect has been previously observed in Cacao, French Guiana, where meteorological conditions in a given year may affect malaria in the following year . Nevertheless, the biological impact of meteorological factors on vector populations over a long period can only have hypothetical explanations. Overall, annual climate seasonality was linked to malaria seasonality as observed by others [26, 27]. In French Guiana, a global and durable climate anomaly such as El Niño episode is likely to increase malaria .
Anopheles darlingi human biting rate was correlated to the malaria incidence rate in children one month later. This is consistent with what was previously found in correlating entomological data with malaria incidence in the general population . Nevertheless, none of the 148 specimens of this species collected from January 2003 to December 2006 in Camopi was found naturally infected with Plasmodium . This study focused on young children with the hypothesis of a nightly transmission due to the characteristics of An. darlingi . Other authors mentioned that An. darlingi has a 24 hours activity and can be found outside during the day in French Guiana . Other anopheline species could play a role in the transmission of malaria including during the morning [6, 29, 30] or in a sylvatic environment around the hamlets . Indeed, An. nuneztovari could play a role in the transmission, along the Camopi River where the incidence remains higher. Indeed, this species could be a secondary vector when present in sufficient numbers. This species, exophilic and aggressive on humans, may be collected on humans in large numbers in some Amazonian areas [20, 31, 32].
Given that it can be assumed that An. darlingi transmitted P. vivax and P. falciparum, it was intriguing to observe that P. vivax had a high transmission in May and June whereas P. falciparum had a much lower transmission during the last period. A tentative explanation is that the age composition of the An. darlingi population may depend on the seasons and the environment [9, 33]. Thus, long-lived females that would be good vectors for the two plasmodial species might be responsible for malaria transmission in December and January. On the other hand, females with a lower life expectancy would be poor vectors for P. falciparum which has a longer extrinsic cycle than P. vivax, and should be responsible for the transmission of P. vivax in May and June only. Unfortunately, the comparison of survival rates of females in these two periods could not be performed due to the low numbers of mosquitoes collected. Another explanation is that P. vivax relapses in the months following the first peak provide gametocytes to the emerging vector population thus amplifying the vivax transmission.
Even if the distribution of malaria is determined by climatic and other geographic factors which affect mosquito and Plasmodium reproduction at a given time, malaria is also influenced by environmental changes . Therefore, the impact of deforestation on malaria transmission that has been previously described [18, 34–36] should be investigated in this area where the settlement of Amerindian populations coupled with gold-mining activities constantly cause forest openings.