The baseline analysis presented here of a population-based, longitudinal malaria cohort study in two contrasting ecological villages in the Peruvian Amazon indicates substantial prevalence of malaria parasitaemia in the region and is an important benchmark for ongoing longitudinal studies. The riverine rural site, Lupuna (LUP), is home to a fairly stable population (85% of individuals >10 years having lived in the village 10 years or more). Here, the combined analysis of survey findings and routine reported malaria incidence suggests that malaria transmission has long occurred with primarily seasonal patterns. In contrast, Cahuide (CAH), a rural site with road-driven deforestation and a more transient population (about 60%), has experienced a notable and concerning increase of cases since May 2012 that compelled the regional government to carry out intensive outbreak control measures until August 2012 (1 month before the survey). Low parasite rates were observed in both sites during the survey, with high predominance of asymptomatic and sub-patent infections over symptomatic and patent infections. Moreover, the site-specific analysis of malaria prevalence and reported incidence suggested a high degree of heterogeneity at the micro-geographic level and over time for both P. falciparum and P. vivax.

Community surveys for determining malaria parasite carriage are commonly used to estimate malaria transmission in endemic areas [32, 33]. Because cross-sectional designs only include collecting data at one time point related to a specific outcome of interest (malaria), they are unable to capture the variations in malaria transmission over time (i.e., seasonal, epidemic or sporadic patterns in malaria transmission) [32]. Therefore, a better interpretation of the obtained prevalence rates in LUP and CAH requires understanding the appropriate context (riverine, road-associated) in Amazonia, and analysis of past trends in malaria incidence in relation with important events and factors influencing malaria transmission. Such analysis will be deepened by the longitudinally acquired data over the course of our ICEMR project (e.g., epidemiological, entomological, and ecological data), and will underlie the implementation of future control and elimination strategies.

The richness of these datasets is important to consider with regard to how future analyses will guide anti-malaria interventions. In riverine, rural LUP, malaria prevalence by PCR may indicate the seasonal-time dependent parasite carriage level in an endemic area (i.e., low seasonal transmission during the baseline survey) with no major recent changes in the population socio-demographics and in local control measures. In contrast, in road/deforestation and riverine rural CAH, prevalence of parasite carriers (with confirmed gametocytaemia by LM in some of them) reflects the parasite reservoir that enables persistence of malaria transmission by both P. vivax and P. falciparum. This was observed here to have occurred despite the intensive response from RHDL-MoH to a severe malaria outbreak from May to August 2012. These interpretations are further supported by a recently published entomological study of our Amazonian ICEMR that found that the highest monthly entomological inoculation rates (EIRs) are associated with seasonal transmission in both sites, with an important factor being seasonally high river levels that enable mosquito breeding [28]. Indeed, the entomological evaluation confirmed the presence of infected An. darlingi (by both P. vivax and P. falciparum) and higher human biting rates (HBRs) in LUP between February and June (maximum EIR in April) in successive years 2011 and 2012; while for CAH, the highest monthly EIR (2.52) with also confirmed P. vivax- and P. falciparum-infected An. darlingi occurred during the 2012 malaria outbreak. Interestingly, despite decreased HBR levels leading to a much lower estimated EIR in CAH in October 2012 (period of baseline survey), the entomological study found a high infection rate of collected mosquitoes (1.47%) which indicated onward malaria transmission in the area [28], and the potential new increase of cases if ecological conditions were favourable. This is exactly what happened in November–December 2012 [13] following the early onset of heavy rains generating an abrupt height increase of level rivers since October 2012, and consequently damaging and flooding to villages located along the Amazonas River and its tributaries [9].

As previously described elsewhere in the Peruvian Amazon [16–21], the findings presented here from both LUP and CAH consistently demonstrated that a large majority of malaria infections—by both P. vivax and P. falciparum—were asymptomatic (64–85%) and sub-microscopic (75–77%), despite differences in malaria prevalence and variations in malaria transmission over time. A limitation of such cross-sectional surveys is that they might not correctly estimate the proportion of infections that later might become symptomatic or have higher parasite density levels (as detectable by LM) [24, 34]. Few longitudinal studies in Amazonia, in particular Peru and Brazil, have provided data regarding such evolution [20, 22]. A 6-month cohort study in peri-Iquitos villages combining PCD, ACD and weekly monitoring of symptoms found that an important proportion of malaria infections which were asymptomatic at the time of detection (either by LM or PCR), developed malaria symptoms (reported and/or measured fever) after 1 week of follow-up, resulting in only 35–40% of asymptomatic infections among all confirmed cases in study villages [20, 34]. Another study conducted in western Brazilian Amazonia in 2004–2005 [22], including cross-sectional surveys, PCD and ACD over a 14-month period, found that among 93 asymptomatic infections confirmed by PCR during surveys, only 10 (10.7%) developed symptoms consistent with malarial disease over the subsequent 2 months of follow-up. Large prospective population cohorts with long and rigorous follow-up, as those recently implemented by the Amazonian ICEMR [26, 27], offer the opportunity to better delineate clinical and parasitological evolution of extant and/or newly acquired infections that were only detected by PCR. Population-based longitudinal study design is required to measure accurately the natural history of true and persistent asymptomatic and sub-microscopic carriers, and assess their contribution in maintaining malaria transmission in the Amazon region.

Considering the site-specific socio-demographics, household characteristics and ecological conditions in LUP and CAH, risk factor and spatial analyses of malaria infections were performed separately for each site. The low number of infections in LUP detected during the survey in low transmission season would primarily explain the difficulty in identifying malaria risk factors and spatial clusters of high transmission in the site for both P. vivax and P. falciparum. Household overcrowding, the only factor found to be significantly associated with P. falciparum infection in LUP, has also been reported in several studies conducted in areas where transmission is exclusively due to P. falciparum, or co-endemic with P. vivax [35, 36]. The effect of household overcrowding on malaria risk may be explained by the increase in human-released chemo-attractants for mosquitoes in reduced spaces [37], leading to higher human–vector contact rates. Despite the lack of association between P. falciparum infection and other proxy variables for socio-economic status in LUP, it is important to note that overcrowding can also reflect poor economic household conditions. The complex association between poverty and malaria is well known, which likely operates in both directions [38, 39]: poor households are less able to afford prevention and control measures and the higher burden of malaria may push these same individuals deeper into poverty.

Regardless of how the relationship of species-specific malaria with potential risk factors in CAH might be explained, it must be considered that our parasitological baseline survey probably captured the incipient onset of a second local outbreak of P. vivax and P. falciparum malaria. This outbreak likely included both: parasite carriers that remained despite the response to a first severe outbreak occurred between May and August 2012, and new infections by either similar or different parasite populations than those of the first outbreak. The increased risk for P. vivax in males and individuals aged >15 years together with the confirmed infection of mosquitoes in the site [28] suggests that transmission during the second outbreak may have been primarily determined by new infections (mostly asymptomatic infections with low parasitaemia) rather than by hypnozoite-triggered relapses. This explanation is further supported by entomological findings in the same site indicating that An. darlingi mosquitoes bite mainly outdoors [28], and by other studies in the Peruvian Amazon suggesting that the higher malaria risk of adult males is associated with increased human–vector contact rates following outdoor- and forest-related occupational activities (even though the latter activities were not resulted directly associated with the infection in the present study) [14, 40].

The differences in malaria risk between species with respect to the availability of electricity at household can be explained by the different connotations that this public service can have. While the lack of electricity may reflect poor socio-economic status associated with increased P. vivax infections and consequently suggest that transmission is not entirely outdoors; the electricity availability as risk factor for P. falciparum may be related with an increase of evening outdoor activities (e.g., watching television, chatting, resting, etc.) of household members [41]. On the other hand, the development of host immunity with the intensity of exposure to P. falciparum species in CAH (site that reports P. falciparum cases every year) may explain the significant relationship of P. falciparum infections with younger age groups [42–44], as well as, the finding that all P. falciparum infections had low parasite densities non-detectable by LM [16]. The results presented here from CAH differ from previous reports showing that prevalence and/or incidence of P. vivax decreased significantly faster with age than that of P. falciparum, suggesting that acquiring clinical immunity is different species-specific, as has been seen in other co-endemic countries such as Papua New Guinea [45], Thailand [46], Vanuatu [47], Ethiopia [48], and Brazil [42, 49]. The data presented here regarding LM−, PCR+ malaria infection do need to be interpreted with caution, considering the limitation of cross-sectional malaria epidemiology studies because risk for transmission is highly dynamic and multifactorial. The longitudinal cohort data from both sites, once available, will allow for more precise understanding of the complex risk factors for incident malaria infection. The rich data sets will allow for more sophisticated analysis, for example, using complex mixed-effects regression models to account for repeated measures, clustering within household/village, and hierarchical levels of risk factors. Further, the identification and characterization of parasite populations will provide insights into complex malaria transmission dynamics in diverse scenarios, for example those characterized by changing seasonal endemic and/or epidemic transmission due to climate change, or those influenced by human mobility patterns due to social policies.

The high heterogeneity in malaria transmission in the Peruvian Amazon creates opportunities for targeted interventions [14]. However the identification of hotspots of malaria transmission and their characterization are not always straightforward. In our study, SaTScan analysis detected clusters with the highest malaria prevalence for both P. vivax and P. falciparum only in the CAH site. Since the cluster located on the south of CAH for P. vivax was found to be associated with an increased presence of male individuals, age groups older than 15 years and the lack of electricity at households, an adjusted spatial analysis for these covariates was performed resulting in the disappearance of the cluster. The observation indicating that the aggregation of P. vivax infections was mainly determined by the aggregation of groups with high-risk than by specific ecological conditions within the site has also been reported in other villages of the Peruvian Amazon [14]. Local mosquito and human mobility patterns may explain the absence of clustering of specific environmental factors along the Iquitos-Nauta road. Indeed, the presence of several breeding and resting sites near households along the road which enlarge after rains or floods from Itaya River could result into a wider dispersal of An. darlingi, and the high mobility of infected individuals through the unique and highly accessible road could further increase the dispersal of parasites. Conversely, the cluster for P. falciparum remained in the closest area to the river even after being adjusted for age and electricity availability (variables associated with P. falciparum infection). Its highest prevalence may be explained by the unprecedented floods from Itaya River in 2012 resulting in large and multiple breeding sites in the area, but also by the presence of a fluvial port to which individuals (resident and non-residents) arrive after being exposed to malaria due to travel-related economical activities in other endemic riverine villages.