Background
Significant progress has been achieved in malaria control worldwide over the past decade, with a global reduction of incidence rates by 21% and a decrease in global mortality rates by 29% between 2010 and 2015. Endemic areas with low levels of malaria transmission are now engaging in elimination programmes aimed at interrupting transmission, while preventing re-introduction and resurgence [
1]. Therefore, the identification and elimination of all infections, including asymptomatic and sub-microscopic infections, is crucial and calls for intensified and new control strategies [
2‐
4]. However, in pre-elimination settings, malaria transmission tends to be highly heterogeneous in space and time due to a complex interaction of human, socio-cultural, environment, and biological factors that remains mostly unknown.
In Vietnam, malaria transmission occurs mainly in remote and forested areas in the central highlands and along international borders of Lao PDR (Laos) and Cambodia [
5,
6], with 12.5% of the Vietnamese population at risk (mainly ethnic minorities living on forest-related activities and migrant workers from non-endemic areas) [
5‐
11]. Malaria morbidity and mortality in Vietnam has been reduced by 93.4 and 97.9%, respectively, over the past 20 years [
12]. As a consequence, in October 2011, the Vietnamese Government officially launched the National Malaria Control and Elimination Programme (NMCP) with the aim of eliminating malaria from the country by 2030 [
13]. Major control strategies include national distribution of insecticide-treated bed nets (ITNs) and long-lasting insecticidal nets (LLINs) (supported by awareness campaigns), and the widespread use of artemisinin-based combination therapy (ACT) for case management [
11,
14]. In addition, indoor residual spraying (IRS) is used in hot spot areas, and areas without bed nets [
14].
Plasmodium falciparum has been the predominant species in Vietnam, accounting for approximately 70% of malaria infections between 2006 and 2010. However, in recent years control measures have had a stronger impact on
P. falciparum than
Plasmodium vivax, resulting in a near equal ratio of both species since 2014 [
13]. As transmission declines, the main challenges of malaria elimination efforts in Vietnam are the high prevalence of asymptomatic and sub-microscopic infections [
2,
6,
9,
15,
16] and the emergence and spread of ACT-resistant parasites in the greater Mekong sub-region and within Vietnam [
17‐
20]. Asymptomatically infected individuals do not seek treatment and generally harbour low parasite density infections undetectable with currently used routine diagnostics (light microscopy (LM) and rapid diagnostic tests [RDT]). Hence, parasites can persist in these individuals from one season to the next maintaining local transmission [
3,
9,
21].
In this context, monitoring malaria transmission and dynamics is important to identify areas with residual transmission and hotspots, identify population-level risk factors, and plan new interventions or evaluate the impact of those currently deployed [
2,
4,
22]. However, as malaria transmission continues to decrease and spatial heterogeneity increases, this becomes increasingly difficult and costly for national programmes [
2,
4,
15]. Tools currently available for monitoring transmission and measuring the impact of interventions, include the entomological inoculation rate (EIR), passive case detection, methods of detecting circulating parasites, e.g., by LM, RDTs or nucleic acid detection through polymerase chain reaction (PCR), or defining malaria exposure with serological methods [
2,
15]. Each method has inherent limitations, which become an increasing problem as transmission further decreases [
23,
24].
Serologic methods have received renewed interest as valuable tools for transmission measurement, especially in the elimination context, due to their capacity to integrate malaria exposure (i.e. infection) over time, to identify foci of recent transmission and to determine the presence or absence of recent transmission in specific populations, such as young children [
2,
15,
25‐
28]. Low levels of antibodies are produced for many months after antigen exposure [
26,
27], some persisting years after exposure [
25]. Analysis of age-specific seroprevalence rates enables differentiation of recent changes in transmission intensity from longer-term transmission trends, while the use of mathematical models of the annual rate of seroconversion can estimate the longevity of the antibody response [
29,
30]. However, there is a need to standardize protocols and antigens used in serology in order to increase comparability between studies [
2,
15]. This is especially relevant for
P. vivax, where besides few studies having used serology to estimate transmission, little variation in exposure throughout peak and dry seasons has been reported most likely due to relapsing infections [
31‐
34].
A pilot study was conducted in Central Vietnam aiming to characterize changes in the parasite reservoir in space and time throughout the dry season and into the wet season, in order to further target elimination strategies and surveillance. Earlier epidemiological studies, conducted in Ninh Thuan and Binh Thuan Provinces, characterized risk factors of disease while malaria transmission was still relatively high [
5‐
9,
16]. The current study was conducted in Quang Nam Province, which has had declining malaria incidence over the past 2 decades. Following a full census of the population in selected villages, a 1-year prospective cohort study with regular malaria screenings was conducted to detect low-density asymptomatic infections and exposure occurring during the study period.
Discussion
As Vietnam progresses to eliminate malaria by 2030, control efforts now focus on remaining malaria foci in the central highlands and along international borders. Among these, in Nam Tra My district (Quang Nam Province) malaria incidence has significantly dropped since 2005. In three villages of Trà Cang commune, few infections were detected by LM and qPCR during 6 malaria surveys in 2014–2015, despite rigorous quality control and repeated experiments. Simultaneously, the CHC reported very few cases of confirmed malaria. Prior to and during the study surveys, the NMCP conducted IRS (November 2013, April 2014 and 2015, and August 2015) and LLIN distribution in (July 2014) in order to put a halt to the high transmission observed in 2012 and 2013. Based on the parasitological data presented here and community health centre data, these interventions were successful at reducing the incidence of clinical disease and infections. However, serological data confirmed that 13.5% of the surveyed population was nevertheless exposed to P. falciparum and/or P. vivax parasites during the study period. Of the exposed individuals, 32.6% were seronegative at the start of the study confirming ongoing transmission in the area, mainly in Tu Nak and Xe Xua villages.
Consistent with parasitological observations and CHC data, a decrease in seroconversion rates indicates that overall transmission intensity for P. falciparum for all ages decreased during the study year. P. vivax seroconversion rates between S1 and S6 are hard to compare, as the age ranges of the rates are different, due to the high number of exposed P. vivax children in Tu Nak in S6. Higher P. vivax seroconversion rates at younger ages in S1 indicated a higher force of infection compared to P. falciparum. Seroreversion rates for both species between S1 and S6 are similar as expected due to its dependency mostly on antibody half-lives.
Risk factor analysis for seroprevalence and exposure to
P. falciparum and/or
P. vivax investigated individual and household characteristics, and identified structural or economic risk factors (e.g., house structure and livestock ownership) and activity/behaviour-related factors (e.g., occupation (farmers) and bed net use). Adults were at higher risk of seropositivity and recent exposure to malaria than children, which is most likely due to risk-behaviour (i.e, increased likelihood of exposure to infected mosquitoes), and the effect of life-time exposure. Bed-net use and number of bed nets in a household were not associated with seroprevalence, however, those reporting occasionally not sleeping under a bed net, were 5 times at higher odds of
P. falciparum (but not
P. vivax) recent exposure. While several risk factors for
P. vivax seroprevalence were identified, only village location was identified as a risk factor for
P. vivax exposure. The absence of association of bed-net use with
P. vivax exposure could be explained by the contribution of relapsing infections [
46], which are independent of mosquito transmission. With current methods it is not possible to distinguish between new infections and relapses, however if individuals that were
P. vivax exposed but seronegative at S1 were considered to have been exposed to a primary infection instead of relapse, the risk factors in this sub-group can be estimated. The ratio of bed nets in a household was significantly associated with protection of these individuals with assumed primary infections, as well as working in the field.
While older age is a strong predictor of
P. vivax seroprevalence, it was not strongly associated with exposure, and a high proportion of participants under the age of 20 were exposed in the western part of Tu Nak. Many of the exposed households in Tu Nak have children that go to school and sometimes stay there overnight (in Tu Nak, 70% of exposed households have children sleeping at school vs. 53% of all households), which suggest that initial transmission may have occurred at the school. Indeed, several children reported not sleeping under bed nets at school, despite availability of nets (1 net/4 children) and teachers trying to enforce sleeping under the nets (pers. comm. M. Bannister-Tyrell, qualitative study [
47]). In addition, many children from Xe Xua cluster reported sleeping at the school in Tu Nak prior to S6. Infections could have subsequently spread amongst others in the household and neighbouring households when the children were back at home.
In contrast to the current study, previous studies in Ninh Thuan and Binh Thuan Provinces in Central Vietnam, where parasite rates and clinical disease was higher than in the current study, showed that regular forest activity was one of the main risk factors for clinical disease, asymptomatic infection and/or seropositivity [
5,
6,
8]. Additional risk factors not captured by the present analysis may affect exposure and in order to identify these, a qualitative ethnographic study was performed in a proportion of the adult population of Trà Cang [
47]. The qualitative study offers an explanation why forest and field activities per se were not consistently associated with seropositivity and exposure, as transmission in the adult population seems maintained by evening outdoor activities that delay or disrupt sleeping in a permanent structure in which a bed net could be hung such as drinking and TV watching in the villages or evening fishing or logging in the field or forest [
47]. Finally, additional factors often associated with risk of exposure to malaria (such as differences in vector composition between villages, vector behaviour and proximity to vector-breeding sites), that might influence exposure, were not investigated in this study.
This study confirms that passive case detection and clinical surveillance will not be sufficient to guide malaria control/elimination programmes as transmission declines [
2,
15]. Clinical surveillance throughout the study period detected only 11 symptomatic malaria cases in the entire commune (including villages not included in the survey; the CHC serves roughly 3500 people). In addition, no symptomatic cases were identified within the study participants despite ongoing passive case detection at the CHC. Malaria screening surveys with LM and qPCR detected only four cases. On the contrary, serological evidence indicate that 13.5% of the population was exposed during the study period. More frequent surveys, more sensitive assays (e.g., increasing the volume of blood on filter paper or the amount of blood spots in the extraction [
48], or targeting other genes [
49]) and/or better coverage of the entire population (> 80% at each survey) might have increased the likelihood of detecting ongoing infections with qPCR. Notably, low attendance was not associated with exposure, therefore, it is unlikely that more infections were captured with increased sampling coverage. However, 8% of the population in the villages were not surveyed at all, of which 65% are farmers, and these people could have contributed to the asymptomatic reservoir.
Most studies using serology to estimate transmission levels have reduced continuous antibody data to a dichotomous seropositive vs seronegative classification, which can widely vary depending on the method used to determine the cut-point [
50,
51]. In this study, antibody quantitative data is exploited to another level by classifying three categories of seropositivity used to define recent exposure. This can be used in specific foci (or previous foci) to determine whether transmission has stopped or is still ongoing, elucidate population risk factors, investigate the impact of control measures or detect routes of transmission, which is difficult using merely seropositivity as the long time required for seroreversion, especially in the adult population that has been highly exposed in the past. Risk factors for recent exposure, especially of
P. vivax, were different than risk factors for seropositivity, which reflects the historical long-term exposure rather than recent events that are informative to guide control programmes. In addition, serology identified the western part of Tu Nak as a focus of ongoing
P. vivax transmission, while it was missed by clinical and parasitological surveillance.
In this study, antibodies (with relatively short half-lives) against two parasite antigens were chosen for each species in order to determine recent exposure [
52]. The near to 100% seroprevalence of
P. falciparum at older ages reflects lifelong exposure to relatively high past
P. falciparum transmission, and high antigenicity and/or long antibody half-life of the tested parasite antigens.
High antigenicity, high age-related seroprevalence and little variability of antibodies against PfAMA1 and PvAMA1 were found, suggesting that these antigens might be more suitable to investigate long-term changes in exposure. Conversely, antibodies against PfGLURP-R2 were a good indicator of recent exposure as PfGLURP-R2 showed low antigeniticy and low seroprevalence compared to PfAMA1, due to a short half-life (~ 6 months) [
52]. This is in agreement with the decrease in
P. falciparum transmission observed in the area. A similar proportion of
P. falciparum and
P. vivax exposed individuals was observed, despite a lower clinical incidence and decrease in seroprevalence of
P. falciparum. It has been shown that cross-reactive (boosting of) antibody responses against
P. vivax or
P. falciparum antigens could be generated by infections with either species [
53,
54]. For
P. vivax, PvMSP1
19 was a suitable marker to investigate recent exposure, in agreement with a previous study that described a rapid decline of this antibody levels within 2–4 months [
55].
Conclusions
Malaria control campaigns in this area could be improved by increasing awareness of bed net protection against malaria and by stressing the importance of bed net use to stop transmission in these villages. In addition, risk factors for exposure at schools should be further investigated, while control strategies to increase bed net use amongst children sleeping at the schools should be developed to prevent transmission. Specific or targeted interventions or campaigns could be implemented aimed at preventing malaria transmission in those with the least access to resources (such as people living in houses on the ground rather than stilts), such as awareness and bed net campaigns, active case detection (focal screening and treatment, FSAT), additional IRS, mass drug administration (MDA).
In Central Vietnam, previous studies have emphasized the high occurrence of asymptomatic and sub-microscopic infections among ethnic minorities at higher risk of exposure, but few asymptomatic infections were detected when transmission declined despite serological evidence of continued transmission. Overall, this study demonstrates the difficulties encountered to accurately study populations at risk as transmission decreases [
23]. This could be overcome by increasing the number of surveyed individuals, increasing the duration of sample collection or surveying a broader geographical area, as well as using a qPCR strategy with increased sensitivity. However, these alternatives come with additional efforts and costs, while potential challenges include changes in seasonality or heterogeneity in the micro-epidemiology [
23,
56,
57]. In this study it was shown that by using serological classification of recent exposure could be used to monitor malaria transmission and exposure in areas of declining malaria.
Authors’ contributions
ERV, NXX, AE, and ARU conceived of and designed the study. NXX, AE, TMH, NVV, and VKAD conducted field data collection and data entry. TTD and NXX oversaw data collection. NVH, ERV, NTHN, and JHK contributed to laboratory analysis. MT provided antigens used for serology. JHK, MBT and AE analysed the data. JHK wrote the manuscript. MBT, ERV, ARU, AE, NVH, NXX, AAL, AB, and MT critically revised the manuscript. All authors read and approved the final manuscript.