Globally, the number of countries in malaria pre-elimination and elimination phases is increasing (1). For malaria control and elimination purposes, the allocation of national resources is commonly guided by regional case counts. However, trends in passively detected cases are dependent on health system coverage and treatment-seeking behavior, and thus may not represent the true malaria burden in all settings. Serological endpoints, typically the presence of antibody responses in resident populations, represent a population's exposure history and can identify areas with residual malaria transmission more accurately than the prevalence of infection at low transmission (2).

Until recently, sero-surveillance has primarily focused on antibodies representing cumulative exposure, such as Plasmodium falciparum apical membrane antigen 1 (AMA-1) and the 19 kDa fragment of merozoite surface protein 1 (MSP-1₁₉) (2). Age-specific increases in seroprevalence to these antigens, estimated as seroconversion rates (SCR), have been shown to be strongly correlated with entomological inoculation rates (EIR), the gold standard metric for transmission intensity, and with parasite prevalence (2). Antibodies to these antigens persist in the blood with repeated exposure. For MSP-1₁₉, model estimates suggested the time to sero-reversion is 23 (3) to 50 years or more (4), while limited data from observational studies suggest half-lives of long-lived antibody secreting cells to be 2 (5) to 16 (6) years. Although MSP-1₁₉ and AMA-1 antibody half-lives might be faster in children (7, 8), this may be due to insufficient repeated exposure in children at low transmission. Antibodies with shorter half-lives [i.e., those indicating incidence in the past year (9)] may be able to detect if and where changes in malaria transmission intensity take place more rapidly and accurately as compared to antibodies with long half-lives. Several potential candidates have recently been optimized for use in multiplex bead assays (MBA) (10).

As part of Haiti's aim to eliminate malaria (11), large-scale cross-sectional surveys were performed to assess if and where residual transmission, potentially undetected via routine surveillance, is occurring. Here, we assessed antibody responses to 23 P. falciparum recombinant proteins and peptides in 29,481 participants residing in two areas with different levels of transmission intensity. Our aim was to select antibodies associated with cumulative and current or recent exposure to malaria for the Haitian context in order to simplify analyses and interpretation of collected survey data that can be used to inform program decisions.

This study used multiplex serological data to select antigen-specific antibody responses associated with cumulative and current or recent malaria exposure in Haiti. IgG responses from almost 30,000 participants were analyzed from three surveys across two locations with differing levels of malaria transmission intensity. Results showed that IgG responses to antigens AMA-1 and MSP-1₁₉ best reflected cumulative exposure over time, whilst those to Etramp 5 ag 1 showed the highest accuracy in predicting current or recent infection when compared to hsRDT status. For the latter, survey data was used from the area experiencing the highest malaria transmission (Grand'Anse).

Continuous antibody responses to AMA-1 and MSP-1₁₉ were confirmed as the optimum representation of cumulative malaria exposure as they showed the strongest correlation with age. These antigens have been used extensively in sero-surveillance for P. falciparum transmission (2). Seroconversion rates to AMA-1 and/or MSP-1₁₉ have shown a strong correlation with household altitude (42), EIR (2, 42), child mortality (43, 44), the effect of malaria control interventions (43, 45), and patterns in malaria transmission over time (44). In addition, GLURP-R2 was moderately correlated with age. GLURP-R2 has been associated with protection from (symptomatic) malaria (46, 47) as well as cumulative exposure (40, 48). This study shows strong indications that seropositivity to AMA-1, MSP-1₁₉, and to a lesser extent, GLURP-R2, represent cumulative exposure in this Haitian population. Some previous studies have found that AMA-1 and MSP-1₁₉ can be good markers of active and recent infection in children (7, 8, 41), though at higher transmission, even in children, repeated exposure and subsequent antibody boosting limits their use as recent exposure markers (9). Given that we have confirmed the appropriateness of AMA-1 and MSP-1₁₉ as a cumulative marker of exposure in Haiti, there is an added benefit of enabling comparisons of results across malaria-endemic settings due to their extensive use in sero-surveillance to date (2, 40–42).

Next, continuous antibody responses were used to validate markers of current or recent P. falciparum exposure according to hsRDT results. Etramp 5 ag 1 and GLURP-R0 were found to be most accurate in predicting hsRDT status in the EAG-Grand'Anse survey. As no further improvement was seen by combining seropositivity to GLURP-R0 and Etramp 5 ag 1, seropositivity to Etramp 5 ag 1 alone was selected as representative of current or recent exposure to malaria. For children – in addition to Etramp 5 ag 1 and GLURP-R0 – SBP-1, Etramp 4 ag 2, HSP40 ag 1, and EBA-175 RIII-V showed high accuracy in predicting hsRDT status. Intuitively, antibody responses in young children are more likely to represent a recent infection as they are unlikely to have developed a robust memory response following multiple infections (49), especially in areas of low transmission. Here we showed that all antibodies tested, except those to GLURP-R0, had higher accuracy in predicting hsRDT status in children compared to individuals aged 16 or older. Ideally, an IgG antibody response representing recent exposure does so across all ages as it minimizes the antigen panel needed. IgM antibody responses have regularly been suggested as a measure of recent exposure following their rapid acquisition and decline in viral infections, however, recent data showed IgM responses to malaria persisted over time and were not more short-lived than IgG responses to merozoite antigens (50).

Age-specific patterns of recent and cumulative exposure were consistent with expected exposure histories in the two transmission settings in Haiti: (1) recent exposure was higher in EAG-Grand'Anse compared to EAG-Artibonite which corresponds with the 2017 malaria incidence estimated at 18.1 and 0.6 per 1,000 inhabitants in Grand'Anse and Artibonite, respectively (source: National Malaria Control Program, PNCM, Haiti); (2) recent exposure was higher in healthcare-seeking populations compared to (likely asymptomatic) children attending school in Grand'Anse; (3) recent exposure was similar in children visiting a health facility or attending school in Artibonite, indicating that in this low transmission setting, healthcare-seeking is likely not driven by malaria disease; (4) cumulative and recent exposure was lowest in the HH-Artibonite survey representing an asymptomatic, community-based population across all ages.

The gold standard for identification of antibodies with short (or longer) half-lives, is assessment of antibody acquisition and decay rates in longitudinal studies following naturally acquired malaria infections. However, such studies are costly, time and labor intensive, and thus rare. Moreover, no existing studies have used the same combination of the age range of the study population, antigen panel assessed and assay conditions (9, 25, 40). Here we used hsRDT data from cross-sectional surveys to inform the selection of current or recent exposure markers. It was assumed that the lack in specificity in using antibody levels to predict hsRDT status represented recent malaria infections that had become hsRDT negative. Alternatively, these could be malaria infections below the limit of detection of the hsRDT. The lack in sensitivity might represent infections that are too recently acquired for an antibody response to have developed. Alternatively, certain individuals may not produce detectable antibodies to a specific antigen following infection. The overlap in identified recent exposure markers with previous longitudinal studies [Etramp 5 in Helb et al. (9) and GLURP-R0 in Kerkhof et al. (25)] is promising for selection and standardization of antigen panels for sero-surveillance (51). However, Kerkhof et al. also identified GLURP-R2 and MSP-1₁₉ as recent exposure markers (25)—but these were better markers of cumulative exposure in our and other (42–45, 48) studies—indicating the need for caution and validation of these targets across populations and differing (histories of) transmission intensities. In settings with lower transmission, such as in Cambodia in the Kerkhof study (25), antibody levels are more likely to be reflective of recent exposure as they are not boosted following the lack of repeated exposure. This was also seen in our study, as nearly all antigen-specific antibodies analyzed accurately predicted hsRDT status in the Artibonite surveys where hsRDT prevalence was <1%.

There are different statistical techniques available to analyze multiplex antibody data. Boosted regression tree analysis has been used for P. vivax and P. knowlesi (52, 53). These studies showed that combined immune responses are more likely to be reflective of (recent) exposure though increments are small. It should be noted that the selection of antibodies for recent and cumulative exposure as described in this study is not meant to be exclusive. Remaining antibody responses may still be of interest for research-based studies interested in responses to, and patterns in, antibodies to specific targets. The antigens identified here were selected to enable rapid turnaround of analyses to directly inform malaria control programs in Haiti.

Here we described the selection of antibodies associated with cumulative and recent malaria exposure in Haiti. We used antibody responses (IgG) to 23 P. falciparum recombinant proteins and peptides from 29,481 participants across three surveys. In the absence of a gold standard (i.e., longitudinal data), we used age and hsRDT status to make this selection. Seropositivity to AMA-1 and/or MSP-1₁₉ was selected to represent cumulative malaria exposure, while seropositivity to Etramp 5 ag 1 was selected to represent recent exposure to malaria. The identification of these antimalarial antibodies will simplify analyses and interpretation where results are needed to inform program decisions in Haiti.

The datasets generated for this study are available on request to the corresponding author.

The EAG-Artibonite and EAG-Grand'Anse surveys were approved by the London School of Hygiene and Tropical Medicine (LSHTM) Research Ethics Committee (10393), Tulane Institutional Review Board (794709) and the National Bioethics Committee in Haiti (1516-30). The Center for Global Health Associate Director of Science reviewed and approved the protocols; Centers for Disease Control and Prevention (CDC) investigators were not considered to be engaged in human subjects research (protocol 2016-135a). The HH-Artibonite survey was approved by the CDC Institutional Review Board (6821), LSHTM Research Ethics Committee (10466) and the National Bioethics Committee in Haiti (1516-29 and 1617-31).

For the EAG-Artibonite and EAG-Grand'Anse surveys, consent was collected from school directors following community meetings to inform parents and allow opt-out. At health facilities, adult participants provided consent directly, and consent for children (<18 years) was provided by a parent or guardian and children above 6 years gave assent to participate. Individuals aged 16 or 17 who were married, head of household, or a parent were considered ‘mature minors’ and consented directly. Thumbprint consent or assent (countersigned by a witness) was used for illiterate participants. For the HH-Artibonite survey, verbal consent for overall permission to conduct the survey was obtained from the head of household and/or primary caretaker and documented in the electronic data collection instrument. Individual-level verbal consent was sought from each present individual for the blood sample collection, and assent from persons aged 7–17 years.

JL, MC, TPE, TD, RA, GS, CD, KH, and ER designed the surveys. TPE, TD, RA, VJ, GS, KH, and MC performed and supported field data collection. KT, JGB, SS and ER provided antigen constructs. JP, IR, TE, and GM performed laboratory data collection. LH, AE, JB, KT, CD, and ER supported laboratory data collection. LH, GS, KT, ER, and CD analyzed and interpreted the data. LH wrote the first draft of the manuscript with support from GS and CD. All authors read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.