Background
The annual malaria incidence and mortality have steadily declined in the Greater Mekong Subregion over the past 15 years [
1]. In particular, in the Kingdom of Cambodia malaria cases were reduced by more than 75 %, due to the improvements in malaria control by the National Malaria Control Programme, such as free distribution of long-lasting insecticidal nets (LLINs), performing better case management (i.e. diagnosis by RDT and treatment with ACT) [
2,
3]. Although malaria transmission in Cambodia nowadays is low [
4,
5], still 21 out of 25 provinces are endemic, of which the northeast region (Ratanakiri) accounts for more than 70 % of the malaria burden [
3,
6]. Moreover, a shift towards more heterogeneous malaria transmission has been observed. This results in areas that support malaria transmission, which are referred to as foci [
7,
8]. Within these foci, elevations in malaria transmission in small areas (sometimes <1 km
2 [
7]) or populations can be identified, which are respectively called hotspots or hotpops, presenting a higher risk of infection as compared to the rest of the focus [
7,
9]. Some studies have shown that stable hotspots with a permanent transmission of the parasite over consecutive dry seasons mainly consist of asymptomatic carriers [
10]. Consequently, in low endemic areas as Cambodia, in the wet high transmission season, when the malaria vector population expands, these remaining reservoirs tend to fuel malaria transmission to surrounding areas or populations [
7,
11]. Despite considerable malaria control efforts in Cambodia, this persisting transmission makes it impossible to eliminate malaria by current control strategies. Therefore, it is necessary to move forward from these strategies by creating a novel strategic plan based on the understanding of the hotspots/hotpops [
3,
12]. It is assumed that there will be no persistence of malaria transmission once these recurrent sources of infection are eliminated by targeted interventions to these hotspots and hotpops [
3,
8].
So far, hotspot identification has been carried out by various approaches, including microscopy, rapid diagnostic tests (RDTs), PCR detection and entomological surveys. Microscopy and RDTs cannot detect low-density infections [
13]. In addition, among the PCR-based assays, real-time PCR is a highly sensitive technique capable of detecting higher numbers of infected individuals, including asymptomatic infections. However, this technique has a couple of drawbacks when used in the field e.g. high costs and complexity of its applicability [
13]. Serological markers of malaria exposure, specifically antibodies (Abs) against
Plasmodium antigens (Ags), are appropriate to use when detecting stable hotspots of malaria transmission in low endemic areas [
7]. These Ab-responses increase by cumulative exposure and the longevity of the Abs depends on the Ag [
7]. Therefore, this method can provide an indication of past and recent malaria exposure that can be used to pick up temporal and spatial trends in malaria transmission [
7,
14]. Moreover, previous elimination programs have already observed that the absence of Ab-titers in the youngest age-groups could be used as proof of the cessation of malaria transmission [
14]. However, in Southeast Asia, primo infection by malaria parasites may be delayed to adolescence, due to behavioural and occupational activities [
15]. Serology has already been used to detect spatial trends by previous studies in high endemic settings [
8,
11,
16‐
20]. On the contrary, the serological value for detecting spatial clustering of malaria exposure in low endemic areas has not yet been completely confirmed [
7,
8].
The proposed study aims to further validate five
Plasmodium markers for their potential to detect recent infection [
21] by defining spatial patterns in malaria exposure over two different surveys in comparison with PCR prevalence and malaria incidence data. As in this study, community (= cluster) based data were used, the outcomes were defined as ‘malaria pockets’ [
22] referring to an area in between a hotspot (<1 km
2 [
7]) and a foci (>1 province) where the malaria exposure is higher than the surrounding areas.
Discussion
Methods that can identify stable areas of transmission over time are suggested to be most effective for assessing geographical variations in malaria exposure. Therefore, Ab-responses acquired with cumulative malaria exposure, measured over several seasons, were recommended for implementation in geographical clustering analyses [
7]. During the ‘90s geographical cluster analyses mainly relied on symptomatic cases with accurate details about the place of infection and/or residence [
20,
40], and were merely based on passive case detection (PCD) [
25]. It was not until the 21th century that due to a lack of information about the parasite reservoir in asymptomatic cases [
10], new studies arose focussing on the spatial distribution estimated with PCR-prevalence data of species-specific geographical areas of infections based on asymptomatic carriers [
25]. This approach was especially important in countries with a low endemicity where the majority of infected people are asymptomatic carriers [
20]. Another innovative approach is the application of serological markers in defining these geographical areas. Serology already proved its ability to improve predictions of low transmission risk [
26,
41].
Where most studies only focused on PfAMA1 and PfMSP1.19 [
8,
11,
16‐
19], the advantage of this study is the amount of additional Ags from both falciparum and vivax malaria investigated compared to most other studies. Only one previous geostatistical study has used several Ab markers (namely PvAMA1, PvMSP1.19, PfAMA1 and Pf.GLURP.R2). However, in contrast with the current study, the researchers considered an individual positive when it responded for any of the two Ags for each species, not taking into account the differences in biological activity (e.g. longevity) among these Ags [
20].
The previous study performed by Kerkhof et al. [
26] has led to identification of serological markers with a relatively short half-life that were most likely to be reflective for recent exposure, such as
P. falciparum Ags Pf.GLURP.R2, Pf.MSP1.19 and CSP. These serological markers could map the transmission risks with more precision and accuracy, as they provide the ability of distinguishing recent from past exposure [
26,
41]. The current study, presented here, explored whether or not the use of serological markers is comparable to the use of PCR prevalence (asymptomatic cases) and malaria incidence data (symptomatic cases) to investigate spatial patterns in malaria transmission.
In Ratanakiri, significant malaria pockets were observed for both
P. falciparum and
P. vivax Ags. The largest pockets were located around the most northerly site of the ‘Tonle San River’ for all Ags. In comparison with the PCR prevalence data,
P. falciparum exhibited similar pockets, whereas for
P. vivax differences were seen. The similar pockets found between the PCR prevalence rates and sero-reactivity are in line with a study performed by Bousema et al. [
16], that observed tight correlations as well.
When comparing the serologically based pockets with the incidence based pockets, the pockets neighbouring Vietnam were comparable, while the most northerly pockets at the ‘Tonle San River’ were slightly shifted to the West. There are malaria incidence based pockets found for
P. falciparum situated around the capital ‘Ban Lung’ of the Ratanakiri province. Different studies [
24,
42,
43] investigated the movement of individuals between villages, districts and countries. This might explain the malaria pockets seen around ‘Ban Lung’ raising the possibility that these individuals travel occasionally towards communities nearby the river or to remote areas. Overall, overlap was seen in the serological based pockets compared to the malaria incidence and parasite prevalence data.
That most pockets were perceived around the river confirms findings from other studies that also found more malaria pockets along open water bodies [
19,
20,
44,
45]. The same pattern was observed by Sluydts et al. [
25] who suggests that this is perhaps associated with increased movements of infected individuals and mosquito populations along the ‘Tonle San River’, and with the more remote location of these villages [
25].
The specificity of serological markers for P. falciparum and P. vivax was acceptable (between 72 and 95 %) to predict PCR and malaria incidence pockets. However, sensitivity was in general much lower, except in predicting P. falciparum PCR pockets (between 74 and 95 %). In comparison, sensitivity of P. falciparum and P. vivax incidence in predicting PCR pockets lies between 10 and 50 %.
When looking at the different serological markers, variable patterns were observed, going from malaria pockets that move between the east and west in November 2012 and 2013 (CSP, Pf.MSP1.19 and Pv.MSP1.19) to lasting pockets that became smaller (Pf.GLURP.R2) or remained similar in size (PvAMA1). These varying patterns require further investigation related to the differences in immunogenicity and persistence of the Ab-responses [
11,
46]. The only Ag that follows an expected altering pattern over time was Pf.GLURP.R2, which seems to correlate best with the PCR-prevalence and malaria incidence data. The latter is probably explained by the fact that this is a blood stage Ag with a short estimated Ab half-life [
26]. This might reflect recent exposure with observing pockets that decline over time, suggesting that this serological marker might have potential in evaluating targeted malaria control efforts.
Risk factors related to sero-reactivity were identified by univariate and multivariable analyses. Significant elevated risks for
P. falciparum malaria were seen for age, ethnicity and overnight stay at the plot hut. There were also differences observed in gender and sleeping in the forest, however, this is most probably negligible, as the IRR was very close to 1. Significant elevated risks for vivax markers were seen for age, whereas staying in plot huts showed to be a risk factor for Pv.MSP1.19 only. The
P. falciparum outcomes are in line with a previous study performed in the same area by Sluydts et al. [
25]. In this PCR prevalence based study that was performed on the baseline survey during the dry season, the most important risk factor detected was the overnight stay in the plot hut, based on both univariate and multivariable analyses. However, in the current serologically based study, it seems that age, concerning the older age groups, was the most important factor determining Ab-levels, compatible with cumulative exposure [
7,
17,
26]. When immunity is acquired these Abs can persist for several years. This is caused by the presence of long-lived plasma cells that with every new exposure rapidly produce Abs against these parasites [
47]. Therefore, when defining current exposure it is important to observe the Ab-levels in especially the younger age groups [
17,
47]. Differences between
P. falciparum and
P. vivax could be explained by that fact that
P. vivax shows relapse patterns that influence the serological outcomes, and that longer half-lives were observed for the
P. vivax Ags in a previous study [
20].
Overall, these outcomes confirm the findings of Sluydts et al. [
25], and are also in line with the findings of Incardona et al. [
48]. These researchers mentioned that entire families go together to the field and sleep in plot huts resulting in an increased exposure risk [
25,
48]. Although this is not related to the age differences, as the age composition was similar inside and outside the pockets. However, this can be explained by the immunological maturity-status where children that acquire a malaria infection have the ability to boost their IgG titers, followed by a rapid decay [
49]. These outcomes explain the population characteristics in the Greater Mekong Subregion, where ethnic minority groups, forest workers (of all ages) and migrants are known as the most widely recognized groups at risk [
25,
50].
This study contributes in the validation of serological markers to distinguish very recent from past exposure, as suggested by Sturrock et al. [
41]. This is especially the case for Pf.GLURP.R2, but also for Pf.MSP1.19, CSP and PvAMA1. By this means, combining more different Ags, covering the entire
Plasmodium life cycle and having a longevity ranging from very short (~1 months) to long (year round), might lead to other promising results. However, methods to acquire the exact Ab-persistence are still in its infancy [
20]. While PvAMA1 showed stable malaria pockets, Pf.GLURP.R2 suggest a decline in the remaining malaria pockets. The stability of pockets was also observed by Mosha et al. [
11] on
P. falciparum Ag AMA1 in the high endemic setting of Tanzania. Further development in quantifying exposure over different timescales, as well as the measurement of very recent exposure, serological approaches will provide a major contribution in estimating spatio-temporal patterns of risk [
41]. The use of serology could benefit future malaria control programmes, since the use of serological markers can more precisely identify variation in transmission in low endemic areas. It should be noted that more serological markers that are competent to estimate exposure over different time-scales are required, as at present Pf.GLURP.R2 is most informative [
41], as well as Pf.MSP1.19, CSP and PvAMA1 to a lesser extent. However, Pv.MSP1.19 should certainly not be ruled out, as it probably reflects transmission in the former past for which no PCR-prevalence data may be available.
Authors’ contributions
Sample collection of the survey was performed by the CNM team (National Centre for Parasitology, Entomology and Malaria Control), the ITM Antwerp team and the Institut Pasteur du Cambodge (IPC) team in Ratanakiri Province in Cambodia. Screening of the blood spot samples was performed at the IPC by KK and LW and LC. The study design and literature research has been performed at the ITM by KK, MP, LD, VS and MC. Data entry and quality control was performed at the ITM by KK, LD and VS. Statistical analyses were performed by KK, LD and VS. The first draft of the manuscript was written by KK. All authors read and approved the final manuscript.