This article examines trends in malaria circulation in recent years in a hard-to-reach population, one of the last relevant pockets of transmission in a region approaching elimination.

The data analysed describe a significant reduction in malaria transmission, visible in both cross-sectional studies (PCR and serology results, number of recent malaria episodes reported by participants) and surveillance data. This reduction is linked to the considerable efforts made to strengthen malaria control in the study areas. Specific interventions targeting the population involved in AGSM took place during the period analysed, such as the activity in Suriname of a network of community health workers (malaria service deliverers [MSDs]) created in the mining sector in 2006 as part of the project “Looking for gold, finding malaria” [28, 29], and the international Malakit research project between 2018 and March 2020 [25]. The latter provided self-testing and self-treatment kits, together with health education activities, to gold miners working in mining sites located in French Guiana. This intervention had a significant impact on the management of malaria symptoms in the study population, and on malaria transmission more generally. The content of the kit in terms of treatment (artemether-lumefantrine and single low-dose primaquine) makes it particularly suitable for controlling the cycle of P. falciparum [24, 30].

During the period analyzed an increase in the proportion of cases due to P. vivax and a decrease in the number of areas with active transmission (higher spatial heterogeneity) were observed. This trend, as well as the appearance of epidemiological instability associated with outbreaks, is expected when malaria transmission decreases [31,32,33].

A comparison of survey data with surveillance data reveals some discrepancies in sectors with active transmission. There are several possible explanations for these differences. First, the cross-sectional surveys were not designed to describe a geographical stratification of transmission, so the numbers are insufficient to obtain accurate estimates. In addition, the cross-sectional survey data are based on the systematic analysis of a large number of asymptomatic participants using molecular biology methods, much more sensitive than those (RDTs and microscopy) used to confirm cases reported by the Surinamese surveillance system. It is also possible that some of the cases were captured by the French surveillance system; the lack of information in the French surveillance system on gold mining activities and on the geographical details of the gold mining sector meant that we were unable to use these data in this analysis. Overall, these results demonstrate the complementary nature of these two approaches to describing transmission, and the value of systematic surveys and molecular biology in regions of low to moderate transmission.

The use of serology complements the data derived from the detection of blood-stage parasites (RDT, microscopy and PCR). In our study, SEM 63–90 data provide a general description of the intensity of the recent exposure of the population to P. vivax: it has decreased but the prevalence of exposed gold-miners was still considerable.

However, in the perspective of using the SEM to select individuals for treatment in the context of malaria elimination interventions, questions may arise in this epidemiological context. In our study population, the circulation of P. vivax is still significant, and has been very high until recently. For some of the gold miners, exposure to malaria may have occurred in their previous places of residence in Brazil, even before joining the gold mining business. For older participants, and for those with a longer history of gold mining in the Guiana Shield and the Amazon region, the history of exposure to malaria may therefore be intense. This is reflected in the study data: age, time spent in gold mining and past history of malaria have a significant influence on the SEM 79–79 results. This can also be illustrated by the distribution of anti-RBP2b.F2 antibodies in our study population according to the reported recent and longstanding history of malaria.

The analysis of seropositivity was not explored in geographical detail because the mobility data available referred only to the last gold mining area frequented: in a context of high mobility (median duration of presence 6 months) these spatial data therefore did not have sufficient chronological depth.

Our study suggests a classification of participants according to the likelihood of recent P. vivax infection. This is seen as a proxy for the risk of being a carrier of the parasite (either asymptomatic blood stage or hypnozoite carrier) in order to identify people who might benefit from antimalarial drug administration in population-based interventions. This classification is undoubtedly questionable, but it attempts to take into account of all the data that may be available in cross-sectional surveys, which by design suffer from a lack of biological and clinical follow-up data. The proposed classification should therefore be validated by longitudinal data from comparable populations.

Nevertheless, there are a number of interesting points to be made about this classification. First, it illustrates how individuals identified through passive detection (symptomatic individuals) represent a very small tip of the iceberg of potential hypnozoite carriers. Even an active case detection strategy based on systematic screening using sophisticated molecular biology methods (which are often not widely available in the field) would only be able to capture a fraction of these cases. In this context, the cost-effectiveness of using these strategies to eliminate the human parasite reservoir is questionable.

In our sample, the use of SEM 79–79 would allow the selection of almost all PCR-positive individuals, and the majority of individuals reporting a recent history of malaria. The choice of these thresholds therefore appears to represent a significant gain in sensitivity when aiming to eliminate malaria, but at the cost of a larger proportion of people being wrongly treated compared with SEM 63–90.

Furthermore, in our sample it can be observed that the introduction of a simple epidemiological criterion, such as that used in the CUREMA project’s intervention TDA (collection of information on the recent history of malaria symptoms, associated with recent exposure to active transmission areas), makes it possible to identify almost all the participants with a positive PCR and a significant proportion of those with a positive SEM 79–79. On the other hand, this criterion is weighed down by a relatively high proportion of individuals who may have been wrongly treated. The TDA selection criterion chosen for the CUREMA project attempts to combine satisfactory sensitivity with ease of use in the field when interviewing participants. In fact, it uses broad geographical definitions (linked to the difficulty in the field of asking individuals about their complete itineraries in contexts of toponymic ambiguity) and, for the sake of simplicity, extends the search for exposure to places of active transmission in the previous year. It will be further evaluated in the cross-sectional post-intervention survey of the study.

The lack of longitudinal data means that the categories constructed to estimate the probability of P. vivax infection are compared with the criteria used to define them: this approach therefore has limitations that must be taken into account. Despite the limitations of this analysis, it seems to indicate that the use of SEM results with sufficient sensitivity would be an ideal compromise for selecting patients for treatment in these settings. Nevertheless, in field settings where it is not possible to use laboratory results and while serological RDTs are not available, taking into account clinical and epidemiological history is a cheap and simple selection method that is able to capture the majority of the highly probable carriers. Even if this selection is associated with a significant proportion of people still being treated unnecessarily, the expected risk and cost–benefit would be better than administering drugs to the whole high-risk population.

Furthermore, although modelling studies have been carried out, it remains to be investigated in real life what proportion of these asymptomatic carriers will go on to contribute to disease transmission and what the real benefit of systematic treatment would be [11, 16].

The significant association found in these data between the reported history of malaria symptoms and its biological markers (PCR or serology) seems to support the observation from qualitative studies that symptomatic forms of malaria seem to be relatively well recognized in the community studied, despite the many potential differential diagnoses. This seems to be related to the individual and collective experience of the disease and to the cultural background of this Amazonian population, which has historically been subject to the burden of malaria [34].