Study area and population
On the French Guianese side, the border region of Oyapock includes three French Guianese’s municipalities: Camopi, STG and Ouanary, with approximately 1,700, 4,000 and 100 residents respectively in 2017 according to the STG health center census. The region marks the north-eastern border between French Guiana and Brazil (Figure 1). On the Brazilian side, the frontier region area includes the Brazilian municipality of Oiapoque, which holds 25,514 inhabitants according to the 2015 estimation by the Brazilian National Institute of Statistics, IBGE), and lies along the Oyapock river (Figure 1). To the north and east of the Oiapoque municipality, there are a large number of small villages. Amerindian territories (Uaçá, Galibi, Juminã) are predominant. This border region consists of a vast remote territory covered by the Amazonian rainforest associated with highly variable population densities. There is also a great diversity of populations, including Amerindians (mainly Wayãpi, Teko, Palikur, Karipuna, Galibi-Marworno and Galibi tribes), Creoles, Saramaka and migrants from other Brazilian states who migrate to the area to work in the gold mining sector (mainly illegal gold mining or border supply zones) . Therefore, daily transborder exchanges exist among these populations . The climate is equatorial, with four alternating seasons: a long rainy season from April to June, a dry season from July to December, a short, rainy season from January to February and a short dry season in March. The mean annual temperature is 25.9°C and the annual rainfall is around 3405 mm . The cities of STG and Oiapoque are persistent, low malaria endemicity areas .
Malaria case definition
Malaria diagnosis was performed in health centers using the Rapid Diagnosis Test (RDT) - SD Bioline® Malaria Ag Pf/Pan in French Guiana, and thick and thin smears or RDTs were used in Brazil. A malaria case was defined as a patient with RDT or microscopy-positive results. Data included passive monitoring of cases in border health centers in French Guiana and Brazil, but also active case detection around positive cases in the city of Oiapoque.
Before treatment with primaquine, a glucose-6-phosphate dehydrogenase (G6PD) deficiency test (Roche diagnostics®, instrumentation Cobas 6000) was conducted at least two weeks after the malaria infection in French Guiana .
In French Guiana, P. vivax relapse was defined as having a medical history of malaria within a period of 7 to 90 days since the last malaria diagnosis. This interval was considered adequate in length to distinguish follow-up (0-7 days), relapse (7-90 days) and new infection (>90 days) [7, 14].
In Brazil, the absence of a unique patient identification code did not permit the use of the same relapse identification method. However, malaria attacks related to patient follow-up, treatment failures and potential relapses were identified during the medical consultation and denoted as treatment verification slide (Lâmina de Verificação de Cura, LVC) in the database. A malaria attack was considered as an LVC for P. vivax if the patient was positive for P. vivax and received a treatment against P. vivax malaria during the last 60 days. It is worth noting that P. vivax relapses are expected to be less likely to occur in Brazil due to a systematic primaquine administration (except for specific cases such as pregnancy), which did not exist on the French Guiana side.
As listed below, two investigations were performed according to two different spatial scales and type of information details:
- a local study on the French Guiana border, which enabled a thorough investigation of the malaria cases treated at the STG health center and of the entomological situation in the most affected neighborhood;
- a regional and cross-border study, which enabled exploration of the regional spatial-temporal epidemic dynamic.
Local investigations on the French Guiana side
Data from medical records of the STG health center allowed a retrospective analysis of epidemics. Malaria cases diagnosed between January 1, 2017 and January 31, 2018 were included, and the following variables analyzed: age, gender, outcome and location of acute P. vivax malaria cases and relapses treated in the only STG health center . Census data were retrieved from the health center in order to calculate the incidence and the attack ratio by neighborhood of STG (around 4000 inhabitants). Factors associated with risk of attack/relapse were identified by univariate analysis. Chi-square test was used for nominal data, Student’s t-test was applied when the test statistic followed a normal distribution and Mann-Whitney U test was used for skewed data.
The entomological investigation focused on Trois-Palétuviers, a neighborhood of STG where the greatest incidence of malaria cases occurred during this period, according to the local health center team. Mosquitoes were collected monthly from August to November 2017 over 2 to 3 consecutive nights per month. Two octenol-baited Mosquito Magnet® traps were used to collect anopheline species and were supplemented with BG-Sentinel and CDC light traps . The collections were performed periodically from 18:00 to 07:00. Intra-domiciliary aspirations were done inside four houses between 19:00 and 20:00 in August. Mosquito species were identified morphologically by entomologists of Pasteur Institute of French Guiana according to identification keys specifically adapted to the Anopheline species present in the region .
The infectious status of the Anopheles spp. captured from August to October was investigated. The head and thorax of 1,218 females were dissected and placed in an agitator with grinding beads. The DNA of each sample was extracted using a Magjet Genomic DNA kit (Thermo Scientific, K2722), then 10 female DNA samples were pooled for polymerase chain reaction (PCR) detection. The presence of P. falciparum, P. vivax and P. malariae parasites was investigated by nested PCR using the Snounou et al. method . Individual confirmation was performed for each positive pool.
The meteorological influence on malaria cases and vectorial abundance was investigated, as the development of the vector An. darlingi is correlated to monthly rainfalls in the region . The average air temperatures and daily precipitation were obtained from the STG meteorological station (Météo France n°97308001) .
Regional analysis of the cross-border epidemic dynamic
French Guiana data came from border area health centers (Delocalized Centers for Prevention and Care CDPS) in French Guiana.
Brazilian data came from the Brazilian information system dedicated to epidemiological surveillance (Sistema de Informações de Vigilância Epidemiológica da Malária, SIVEP-Malária).
International cooperation, research and technological development works have been conducted for several years and led to the establishment of an operational system for the harmonization and visualization of the data from the two databases mentioned above . Such system relies on expert knowledge, meets international standards for information representation and uses dedicated and robust harmonisation techniques and tools (notably the Extract, Transform and Load - ETL - approach). It enables the monitoring, in time and space, of the epidemiological situation of the cross-border malaria between French Guiana and Brazil.
Epidemic profile clustering
In order to describe the spatial-temporal epidemic dynamics in the study area, the epidemic profiles of the cross-border area localities were defined and clustered, and the resulting clusters were represented and interpreted both temporally and spatially. To achieve this, first we selected localities presenting a significant number of cases, by considering each country individually. The method included:
- ranking all localities according to their total number of cases during the study period;
- selecting all localities that contributed up to 90% of the total number of cases. This was done by first ranking the localities in descending order of the number of cases and calculating the cumulative number of cases as a percentage of the total. This permitted to consider epidemic profiles only for localities with a significant number of cases, while ensuring the representativeness of the dataset. Time-series of malaria cases were obtained by aggregating the daily case counts on a weekly basis. Locality epidemic profiles were defined by the normalized cumulated numbers of malaria cases. This method highlighted the curves dynamics and facilitated their interpretation.
Next, Ward’s hierarchical clustering method using Euclidean distance was applied to the epidemic profiles.
The French Guianese database was anonymized and declared to the Commission Nationale Informatique et Libertés (CNIL) (authorization N°1939018). Brazil’s surveillance registries database was anonymized prior to being sent. The cross-border malaria information system (Saldanha et al., submitted) was also approved by the CNIL (N°2135463). In Brazil, all the actions carried out in brazilian malaria registries are authorized as part of the Fiocruz public health activities, as per the Brazilian “free access” law 12.527 of November 18, 2011 and in compliance with law 13.709 of August 14, 2018.