Malaria elimination is once again a stated goal for a number of malaria-endemic countries [1]. In areas where transmission intensity is low, malaria control programmes are considering switching from sustained control to elimination. Monitoring of malaria transmission intensity (MTI), and targeting interventions to settlements experiencing higher malaria transmission will be vital in this endeavour [2]. Interventions in geographically isolated areas, such as islands, may be more successful due to better control of imported transmission, geographically defined areas to target, and high community participation [3, 4]. Sustained control efforts and monitoring in these areas can lead to the successful eradication of malaria, such as occurred in Mauritius in the 1960s [5].

However, areas of low transmission, or areas where transmission has been reduced substantially, pose considerable challenges for monitoring and evaluation. Exposure to infection can be markedly heterogeneous which can be an important determinant of the rate of progress, and thus the time required to achieve elimination [6]. When transmission is low, traditional measures of MTI such as the entomological inoculation rate (EIR) and parasite rates (PR) lack sensitivity because numbers of positive samples (mosquito and human) are low. Additionally, the low frequency of positive samples represents a formidable logistical challenge, and further, both entomological and parasitological measures are affected by seasonality [7–9]. An alternative measure for MTI is to calculate the prevalence of anti-malaria antibodies in the local population. Serological markers of transmission show greater sensitivity in low transmission areas, as seroprevalence reflects cumulative exposure and thus is less affected by seasonality due to the longer duration of specific antibody responses. Sample collection and analysis can be readily scaled up, and analysis is relatively simple, making this approach adaptable for resource poor settings [10].

Seroepidemiological studies have previously been used to assess malaria transmission intensity [11], reductions in transmission [12–14] and malaria eradication [5, 15, 16]. However, variation in source of antigen and the subjectivity of the detection methods led to this method falling out of favour [17]. The availability of specific recombinant malarial antigens, and the development of standardized, sensitive enzyme-linked immunosorbent assays (ELISA) mean that seroepidemiological study has once again become an attractive methodology, both for assessment of malaria transmission and for assessing changes in prevalence following the implementation of control programmes [10, 18–21].

The utility of serological markers as a methodology for describing malaria exposure was investigated as part of the evaluation of malaria elimination efforts in Vanuatu. Vanuatu is a South Pacific archipelago made up of over 80 islands, each with varying levels of malaria transmission. There have been several large-scale control programmes implemented in the country over the last 50 years and a reduction in reported cases has been recorded over the last decade (Ministry of Health, Vanuatu). In March 2008, the Vanuatu government revised the goal of their National Malaria Programme (NMP) from control to elimination, with the aim to eliminate malaria from Tafea, the most southerly province, by 2015 (Figure 1). The most southerly island of the province, Aneityum (Population c.800), has been subject to extensive malaria control programmes. There were no reported malaria cases on Aneityum between 1991 and 2002, when a P. vivax epidemic occurred [3]. Since 2002, no more cases have been reported. Tanna, the second largest of the islands in the province, (Population c.20,000) is characterized by low but continuing P. falciparum and P. vivax transmission. A malariometric study performed in 2008 revealed low level parasitaemia and mainly coastal transmission for both species [22].

This paper reports on the sero-epidemiological results from samples collected from Tanna and Aneityum in May 2009, with the aim of assessing the utility of specific serological tools for both ongoing monitoring of MTI, and assessment of changes in MTI, in the pre- and post-elimination era.

Due to current malaria control activities, ownership and reported use of bed nets was generally high, with usage close to 80% in Northern Tanna, Southern Tanna and Aneityum (Table 1). However in the Northern highlands in Tanna the reported usage of nets was lower, with only 38% (167/436) of survey respondents reporting to have used the nets the night before despite 98% (503/512) reporting they owned nets (Table 1). This is likely to be a result of the perceived low malaria risk in this area[23].

Antibody responses in the population

Seroprevalence for each settlement, by age group, is summarized in Table 2 and Figure 3. Overall, seroprevalence increased with age and was higher to AMA-1 antigens than to MSP-1₁₉ antigens. Seroprevalence was highest to PfSE. There was no difference in seroprevalence between males and females for any of the antigens tested (p > 0.1 for all antigens). Overall, seropositivity in Tanna was 9.4% and 12.4% and 16.6% for P. falciparum MSP-1₁₉, AMA-1 and PfSE respectively and 12.6% and 15.0% for P. vivax MSP-1₁₉ and AMA-1 respectively. In Aneityum, seropositivity was 6.6%, 5.8% and 11.8% for P. falciparum MSP-1₁₉, AMA-1 and PfSE respectively and 6.2% and 10.1% for P. vivax MSP-1₁₉ and AMA-1 respectively.

Location	Age group (years)	N Pf MSP	Pf MSP +ve	% Pf MSP +ve	N Pf AMA	Pf AMA +ve	% Pf AMA +ve	N Pf SE	Pf SE +ve	% PfSE +ve	N Pv MSP	Pv MSP +ve	% Pv MSP +ve	N Pv AMA	Pv AMA +ve	% Pv AMA+ve
	0-5	64	4	6.3 (1.6-12.3)	65	7	10.8 (3.0-18.5)	66	5	7.6 (1.0-14.1)	66	13	19.7 (9.8-29.5)	65	8	12.3 (4.1-20.5)
Northern Tanna	5-20	176	15	8.5 (4.4-12.7)	174	29	16.7 (11.1-22.3)	176	25	14.2 (9.0-19.4)	172	27	15.7 (10.2-21.2)	175	20	11.4 (6.7-16.2)
	20-80	121	44	36.4 (27.7-45.1)	119	43	36.1 (27.4-45.0)	120	68	56.7 (47.7-65.7)	119	37	31.1 (22.7-39.5)	121	41	33.9 (25.3-42.4)
	Total	361	63	17.5 (13.5-21.4)	358	79	22.1 (17.8-26.4)	362	98	27.1 (22.5-31.7)	357	77	21.6 (17.3-25.9)	361	69	19.1 (15.0-23.2)
	0-5	115	3	2.6 (0-5.6)	119	5	4.2 (0.5-7.9)	119	1	0.8 (0.0-2.5)	117	6	5.1 (1.1-9.2)	119	4	3.4 (0.1-6.6)
Northern Tanna Highlands	5-20	216	13	6.0 (2.8-9.2)	218	14	6.4 (3.1-9.7)	221	12	5.4 (2.4-8.4)	216	14	6.5 (3.2-9.8)	220	15	6.8 (3.5-10.2)
	20-80	173	21	11.6 (7.2-17.1)	174	35	20.1 (14.1-26.1)	176	66	37.9 (30.6-45.2)	174	12	6.9 (3.1-10.7)	173	22	12.7 (7.7-17.7)
	Total	504	37	7.3 (5.1-9.6)	511	54	10.6 (7.9-13.2)	514	79	15.4 (12.2-18.5)	507	32	6.3 (4.2-8.4)	512	41	8.0 (5.6-10.4)
	0-5	52	1	1.9 (0-5.8)	52	2	3.8 (0-9.3)	51	0	0.0	51	5	9.8 (1.4-18.3)	52	8	15.4 (5.2-25.5)
Southern Tanna	5-20	195	3	1.5 (0-3.3)	188	7	3.6 (1.0-6.2)	192	3	1.6 (0.0-3.3)	195	22	11.3 (6.8-15.8)	194	31	16.0 (10.8-21.2)
	20-80	123	12	9.8 (4.4-15.1)	122	11	9.0 (3.9-14.2)	121	26	21.5 (14.1-28.9)	123	19	15.5 (9.0-21.9)	122	37	30.3 (22.1-38.6)
	Total	370	16	4.3 (2.2-6.4)	349	20	5.4 (3.1-7.7)	364	29	8.0 (5.2-10.8)	369	46	12.5 (9.1-15.9)	368	76	20.7 (16.5-24.8)
	0-5	90	1	1.1 (0-3.3)	90	4	4.4 (0.1-8.8)	90	4	4.4 (0.1-8.7)	90	3	3.3 (0-7.1)	88	3	3.4 (0-7.3)
Aneityum	5-20	218	9	4.1 (1.5-6.8)	218	10	4.6 (1.8-7.4)	219	5	2.3 (0.3-4.3)	219	9	4.1 (1.5-6.8)	210	12	5.7 (2.6-8.9)
	20-80	208	24	11.5 (7.2-15.9)	208	16	7.7 (4.0-11.1)	208	52	25 (19.1-30.9)	208	20	9.6 (5.6-13.7)	207	36	17.4 (12.2-22.6)
	Total	516	34	6.6 (4.4-8.7)	516	30	5.8 (3.8-7.8)	517	61	11.8 (9.0-14.6)	517	32	6.2 (4.1-8.3)	505	51	10.1 (7.5-12.7)

Number of positive individuals and seroprevalence (95% CIs) to PfMSP-1₁₉, PfAMA, PfSE, PvMSP-1₁₉ and PvAMA for three age groups in three settlements on Tanna and two settlements combined on Aneityum

The seroprevalence data suggest that residents in Northern Tanna have experienced the highest exposure to P. falciparum malaria infection, with 22% (79/358) of individuals being seropositive to one or both of the specific P. falciparum antigens tested. Southern Tanna had the lowest specific P. falciparum seroprevalence (5%, 20/369). P. vivax seroprevalence was also highest in Northern Tanna (30%, 107/361), followed by Southern Tanna (25%, 92/369), with the lowest seroprevalence in the Northern Tanna Highlands (11%, 54/512).

Figure 3 demonstrates the increasing seroprevalence from PfMSP-1₁₉ to PfAMA with the highest seroprevalence to PfSE in each settlement. Interestingly, children under five consistently had lower seroprevalence to PfSE than to the specific antigens. This contrasts with adults who had the highest seroprevalence to PfSE in all settlements.

Seroprevalence to specific antigens in children between the ages of 1 and 10 years (inclusive) by village is shown in Figure 4. It suggests that exposure is not uniform across the settlements with focal areas becoming evident in North Tanna and Southern Tanna. Unfortunately numbers were too small to allow for micro-spatial analysis.

Changes in transmission intensity over time

Changes in transmission intensity were identified most readily for PfSE, with weak evidence in responses to the more specific PfMSP-1₁₉ seroprevalence (Figure 5). No clear peaks were identified for PfAMA or for the two P. vivax antigens. For PfMSP-1₁₉, the PLPs did not show strong evidence for a change in transmission, as demonstrated by the log likelihoods all being above the 95% confidence intervals (Figure 5) however, the MSP-1₁₉ PLPs generally peaked at the same point as the PfSE plots. The maximum log likelihoods indicate that a reduction in P. falciparum transmission intensity occurred approximately 30 years ago in Northern Tanna (PfSE PLP: 33 years (95% CI: 27-50) and the North Tanna Highlands (PfSE PLP: 28 years (22-32 years)), 23 years ago in Aneityum (PfSE PLP: 23 years (21-32 years)) and approximately 18 years ago in Southern Tanna (PfSE PLP: 18 years (9-28 years)). The resulting SCRs suggest a 3, 9 and 12 fold decrease in transmission for North Tanna, North Tanna Highlands and Southern Tanna respectively (Figure 6). Transmission on Aneityum appears to have decreased over seven-fold. Although the decrease in transmission is not reflected in the age seroprevalence data for the specific antigens, the SCRs are very low for all settlements.

Peaks were less clear in PLP for seropositivity to P. vivax antigens, suggesting that P. vivax transmission may not have undergone a distinct reduction. A PvSE antigen was not available at the time of processing; it is possible that evidence for a change in transmission would have been detected with this antigen, as was seen with the P. falciparum equivalent.

On Aneityum, it appeared that changes in age seroprevalence were more readily detected in population PfSE antibodies. The PfSE log likelihood indicated a change occurred approximately 23 years ago (18-38 years); the PfMSP-1₁₉ PLP indicated a change occurred between 25 and 31 years ago, although this was not significant (Figure 7). The PvAMA seroprevalence curve for Aneityum (Figure 7) showed some evidence for a change in force of infection, although this was not reflected in the fitted model.

The Vanuatu government aims to eliminate malaria from Tafea province (which includes Tanna & Aneityum) in the next six years. It is thought that malaria has already been eliminated from Aneityum [3] and that transmission is confined to specific locations across Tanna. In this study, the applicability of using serological methods to provide current estimates of transmission intensity was assessed and the utility of serology to provide evidence for the cessation of transmission in Aneityum was also investigated.

Seroepidemiological studies are especially useful in low transmission settings where the sensitivity of parasite prevalence surveys is limited by the scarcity of parasite positive individuals. This study was undertaken in the dry season, and as such only three RDT positives (0.2% of the population surveyed) were found in the population. In the previous wet season, surveys using PCR methods in the same areas on Tanna found 1% prevalence of P. falciparum and 2.2% prevalence of P. vivax infection in children between 2-11 years old [22]. Both parasite-based prevalence measures, which offer a snapshot of malaria prevalence and are influenced by temporal changes, are, not surprisingly, significantly lower than the antibody prevalence data (which offers a period prevalence measure) for Tanna in this study (24%, 299/1242 P. falciparum and 20%, 253/1242 P. vivax). This extra sensitivity allows for a more detailed examination of malaria transmission.

Modelling changes in transmission intensity can give an insight into the success of interventions in a particular area. Perhaps surprisingly, changes in transmission were most readily detected using PfSE. Schizont extract contains multiple proteins some of which will be antigenic and similar between parasite species, it is, therefore, not surprising that PfSE has the highest overall seroprevalence. However, whilst adults demonstrate high seroprevalence to PfSE, seroprevalence in children is lower than for the more specific antigens, MSP-1₁₉ and AMA-1. It is not clear why this should be; it may be the result of lower exposure in children resulting in lower levels of antibodies to the antigens in PfSE. This has been demonstrated in Kenya, where individuals were more likely to have antibodies to specific antigens rather than PfSE [30]. An alternative explanation is that the relative concentration of the specific antigens is lower in PfSE- this needs further investigation.

Statistical analysis of the seroprevalence profiles suggests that a significant drop in transmission occurred approximately 30 years ago. These seroprevalence curves are consistent with those described by Kaneko et al (submitted) using a P. falciparum lysate. However, analysis of country-wide case data suggests that a drop in malaria cases occurred in 1992 (17 years before this study), which was attributed to ITN distribution across the islands [24]. There are two (not mutually exclusive) potential explanations for this. Firstly, the drop in seroprevalence may indicate that the IRS campaign in place in the early 1980s successfully reduced exposure. At the time, this campaign was considered unsuccessful due to little reduction exhibited in national parasite rates. However, it is possible that it was effective in certain areas such as Northern Tanna and on Aneityum. An alternative explanation is that the reduction in seroprevalence is related to the later malaria control programme that was implemented nationwide in 1988 and in Aneityum in 1991, and which resulted in the elimination of malaria from Aneityum [3]. The observation that the step in serological data occurs approximately 10 years before this point in this scenario is possibly due to differential rates of loss of antibodies with age. Younger children who had low antibody levels when the 1988 interventions were implemented, may have lost these antibodies more rapidly than older individuals with a more established antibody response [31, 32]. Clearly more information on the rates of antibody loss is required to allow a full interpretation of these data, and further work is ongoing.

The observation that immune responses to the specific antigens do not reflect the changes in transmission is perplexing. Transmission is known to be very low across Tanna and it is possible that in these conditions models for assessing for an increase in seropositivity with age may be inappropriate. Several of the age stratified curves demonstrate little to no increase in seroprevalence with age. In areas where transmission has been consistently low for many years a loss in seropositivity in older individuals may result in uniform seropositivity across the population; masking any change-point in transmission. This is clearly not the case for the multi-antigenic PfSE- one might expect an increased chance that antibodies to this conglomeration of antigens have longer half lives, although this was not observed in a longitudinal study in Thailand [33]. In a situation where age seroprevalence reveals little information, it may be more pertinent to investigate antibody titres and to focus on children born since interventions have been in place. This work is ongoing.

It is important to understand the dynamics in the transmission of different species in areas where more than one species of Plasmodium are transmitted. Serological measures have allowed us to successfully distinguish between areas where P. falciparum and P. vivax malaria are transmitted. Current SCRs are low for both species but consistently higher for P. vivax in all areas. Seropositivity to P. vivax antigens was much higher in younger individuals, suggesting either that transmission of this species has been higher than P. falciparum transmission over the past 10 years, or that preferential infection of younger age groups, as has been reported previously [34], results in a different age seroprevalence curve than is found with P. falciparum seroprevalence. Additionally, morbidity data from previous studies in Vanuatu [35] and Papua New Guinea [36] show that clinical immunity to P. vivax builds up faster than for P. falciparum which could explain the seemingly high seroprevalence in younger children. The high proportion of P. vivax seropositive individuals in younger age groups in Northern Tanna made the data difficult to fit the conventional seroconversion model. A similar age seroprevalence profile would be expected in areas where there has been an epidemic as the result of a defined period of exposure in all ages of the population. However, the effect of possible presence of hypnozoites in P. vivax infections has not previously been studied and we do not know how infections that are the result of hypnozoites (i.e. relapses) influence serological profiles. Previous data has shown that P. falciparum is more successfully reduced by interventions than P. vivax, due to the limited effects interventions have on relapses, and due to differences in biting behaviour by mosquitoes infected with each species [37]. The profile likelihood plots do not indicate a clear change in transmission intensity based on seroprevalence of P. vivax antibodies. The drop in P. vivax transmission may not have been as pronounced or sudden as the P. falciparum reduction, in which case, a step in the seroprevalence curves would not be evident. P. vivax prevalence tends to be higher in areas of low transmission intensity [38], and an increase in P. vivax cases has been reported in areas where P. falciparum transmission appears to be on the decline [39, 40]. However, on Aneityum clear evidence of a change in transmission for P. vivax was seen in seroprevalence data by Kaneko et al (submitted); this data also indicated a change in transmission had occurred in the early 1980s. This study used a lysed P.vivax preparation as antigen which, as for the P. falciparum equivalent, may be more likely to detect changes in a polymorphic antibody response.

The P. vivax epidemic in 2002 may also have had an adverse effect on the seroprevalence curves to P. vivax antigens in Aneityum. There have been no reported cases since 2002, and whilst seropositivity on this island is mainly restricted to adults, low seropositivity in children under 7 suggests that there may still be residual transmission in the area. However, importantly, four of the 11 children seropositive to any antigen (falciparum or vivax) were classified as aged one. Birthdays and ages are not generally recorded in Aneityum, and it is possible these children were younger than one and, therefore, the antibodies detected could represent residual maternal antibodies. Alternatively, children may have been exposed to malaria whilst away from the island. Limited information was collected on travel history during the survey so this possible explanation cannot be confirmed. Monitoring of cases on Aneityum is very thorough, with everyone returning to the island tested for infection via thick and thin slide films. However, slide reading has limited sensitivity, and it is possible that some people may have arrived carrying infection. PCR methods detect parasites more sensitively and it may be pertinent to check for sub-patent infections amongst the inhabitants of Aneityum. Imported malaria can remain a problem as long as Anopheline mosquitoes are present on the island. This highlights the importance of continued monitoring and surveillance on islands where elimination has been achieved [41], and the recording of accurate travel history information.

An alternative explanation for the observed seropositivity in young children is the presence of antibodies cross-reacting with antigens from other infectious agents [42–44]. Detecting antibodies to specific recombinant antigens using ELISA, rather than the less specific immunofluoresence antibody test (IFAT) minimizes this danger. However, a few studies have suggested the cross-reactive potential of PfAMA with Toxoplasma antigens [45, 46], although this infection is reportedly not prevalent in Vanuatu.

Malaria transmission is notably heterogeneous, and this will become more pronounced as transmission decreases. For malaria elimination, areas where residual transmission is occurring need to be identified in order to target interventions to communities where they are most needed. In Vanuatu, interventions have been very successful at reducing MTI over the past 20 years. However, areas of higher transmission were identified on Tanna during the 2008 survey [22]. Serological analysis and monitoring of these areas can offer further insight into the dynamics of malaria transmission. The additional sensitivity afforded by the longevity of antibody responses also allows an examination of the success of control measures also in confirming the cessation of transmission. Serological measures are robust and high throughput and importantly, remain sensitive in very low transmission settings.