Outdoor malaria transmission in forested villages of Cambodia

Within the Greater Mekong Region, progress in malaria control has been substantial over the last ten years. However malaria remains an important public health concern in some provinces of each country [1]. In 2010, Cambodia reported 49,356 confirmed malaria cases [2]. It is estimated that 2.1 million people (15% of the population) in Cambodia are at risk of malaria, of which approximately half a million live in forest and forest-fringe areas with high malaria transmission [3]. Despite active and uncontrolled deforestation, about 61% of the total Cambodian land area was estimated to be covered with forest in 2002 [4], of which more than 80% is located in malaria-endemic areas [5]. Most of these forested areas are located in provinces bordering Vietnam, Laos and Thailand. People living in villages at the edge of the forest or having forest activities are at high risk of malaria because of the presence of the highly efficient forest malaria vectors Anopheles dirus s.s. and Anopheles minimus s.s. [6–9]. Because of the complexity of interactions that may involve vector populations in and outside the forest, it is not easy to predict the impact of deforestation on malaria transmission in this context [5].

Large-scale vector control interventions occurred during the last years in Cambodia, particularly based on the free distribution of long-lasting insecticidal nets (LLINs) [2]. This has contributed to a substantial decrease in malaria cases. However these LLINs only protect people when they are sleeping inside the houses. The main vectors An. dirus s.s. and An. minimus s.s. are exophagic and exophilic, jeopardizing the impact of LLINs [10]. Several studies in Vietnam, and Thailand show that Anopheles dirus s.l. and Anopheles minimus s.l. are outdoor and early biters [11–13]. In Vietnam, a higher vector abundance and malaria transmission was observed in forest camps as compared to the nearby village, and a higher risk was observed of being bitten by a Plasmodium-infected mosquito during the early evening as compared to the rest of the night [11]. Also, in western Cambodia, this phenomenon of early and outdoor biting has been reported [14], meaning that additional vector control measures are necessary. Additionally, human risk behaviour favouring exposure to malaria vectors, ie staying outside during the night, presents a great challenge. Besides the major forest vectors An. dirus s.s. and An. minimus s.s., a large number of other anopheline species occur in the vicinity of human dwellings. Transmission by ‘secondary’ vectors that have outdoor or early biting behaviour might become more important than transmission by primary vectors in the context of high coverage of insecticide-treated nets (ITNs) [15]. As secondary vectors are often less anthropophilic, and might be more exophagic and early biting, the planning of vector control should take into account their behaviour. Moreover, as pointed out in [11], secondary vectors might be better vectors of Plasmodium vivax as compared to Plasmodium falciparum, as the extrinsic incubation period of P. vivax is shorter. In Vietnam, Anopheles sawadwongporni, a very early biting secondary vector, was found positive for P. vivax[11].

Recently, Cambodia has declared its intention to eliminate malaria by 2025 [16]. It is therefore important to study to which extent the malaria vectors in Cambodia are outdoor-biting (exophagic) and early biting, and to assess the importance of secondary vectors in this context. However, facing the decrease of malaria transmission as a result of the control programme, entomological surveys are not sensitive enough to estimate changes in transmission intensity. In this context, serology is being proposed as an additional tool, as proxy for malaria transmission [17] for measuring the force of malaria infection [18].

In the framework of a larger study on the force of malaria infection in the forested environment in Cambodia the results of an entomological survey designed to have a better understanding of early and outdoor malaria transmission by primary and secondary vectors in forested areas in the eastern and western part of Cambodia, are presented here. Vector abundance, malaria transmission, and early biting rates in forest camps were compared to the situation in the nearby village. Results obtained from the entomological survey are analysed in relation to the serological data obtained in the same localities during the same time frame [19]. The study was performed in 2005 and will be used as a reference to assess the entomological situation in a fast-changing environment.

Vector control has always been one of the cornerstones of malaria control [27]. However, for vector control and the evaluation of its efficacy it is important to know the behaviour of the targeted vectors. As forest malaria is one of the major challenges in the elimination of malaria in Southeast Asia [7], the present study evaluated the vector density, early biting activity and malaria transmission of outdoor biting malaria vectors in two forested regions in Cambodia.

While all vector complexes occurred in all study villages, most of the variability in mosquito densities and early biting rate was explained by geographical factors, which was mainly at village level, and to a lesser extent at district and regional level. Cook et al. [19] also observed that malaria transmission as measured by serological markers can be linked to a certain village, and unpublished incidence data based on rapid diagnostic tests show a high within village correlation of the malaria incidence in 2010 and 2011 in Cambodia (Somony Heng, in preparation). This means that some villages are more prone to malaria transmission, while some are more prone to higher (or lower) vector densities, although, as discussed below, vector density does not linearly relate to malaria transmission. Variability in mosquito density also occurred between regions and surveys. In the east, higher An. dirus s.l. and An. maculatus s.l. densities were observed in survey 1 as compared to survey 2, whereas in the west, higher densities were observed in survey 2. However, in the east, the amount of rain is almost double compared to the west, with the main rain peak occurring in July and August. In the west, the rain peak occurs in September and October. This difference in rain pattern might explain these general differences observed in vector densities between both regions. The surveys were carried out at approximately the same moment in time in the east and the west. Therefore, survey 1 fell just after the rain peak in the east, and before the rain peak in the west (Figure 2), whereas survey 2 was at the start of the dry season in both regions. The survey effect on the mosquito densities should thus be interpreted with caution. For the An. maculatus complex this difference between east and west could also be related to a different proportion of An. maculatus s.s. versus An. sawadwongporni. However, a previous study in Thailand has shown that these species have a similar association to rainfall [28]. Despite the lower total amount of rain, the highest densities of An. dirus s.l., An. minimus s.l./An. aconitus, and An. maculatus s.l. were observed in villages in the west (Figure 3), showing that not only the amount of rain, but also other ecological factors are important determinants for mosquito density.

Strikingly, no difference was observed in vector densities between the village sites and the corresponding forest camps for An. dirus s.l., which is a forest-breeding species [7, 29]. This was not only due to the close distance between villages and related forest plots, as in villages where the forest plots were further away (PX and VY), still comparable densities of malaria vectors were observed; in villages where the forest plot was very close (BY, BZ, OZ), higher densities of An. dirus s.l. were observed in the forest camps (7.08 BMN) as compared to the nearby villages (4.23 BMN), but only in the first survey. The latter three villages are situated in areas with the highest degree of conserved forest as compared to the other villages in the study (Table 1), confirming that higher degrees of conserved forest sustain higher An. dirus s.l. densities in the forest [29]. Nevertheless, present results contrast with previous studies reporting much higher An. dirus s.l. densities and malaria transmission in forest plots as compared to nearby villages, including a study in Pailin district [30] at a time when forest was still conserved, and studies in Central Vietnam (Khanh Hoa province [31], Ninh Thuan province [11]), and in south-western Vietnam (Binh Phuoc province [32]). In the early 2000s, the bed-net use in Vietnamese villages was higher [11] than the one observed during the current study performed in 2005 in Cambodia, which could explain the lower vector densities in the villages in Vietnam.

In the present study, in general, no difference was observed in the densities of the other vectors An. minimus s.l./An. aconitus, An. barbirostris s.l. and An. maculatus s.l. between forest plots and villages. These malaria vector complexes do not require the forest for breeding sites; An. minimus s.l. is more associated with mosaic vegetation and crop [9], and An. maculatus s.l. is a widespread species [9] preferring open or only partially shaded breeding sites, similar to An. barbirostris s.l. and An. aconitus[33].

Although the densities of the malaria vectors differed between villages, all vector complexes were found in all villages and forest plots, regardless if the villages were surrounded by deforested areas or by forest (evergreen or deciduous), or plantations. The forest in the region where this study took place was in most cases scattered and fragmented. It has been reported that An. dirus s.l. is able to adapt to peripheral areas where natural forests are replaced with orchards, and tea, coffee, and rubber plantations [29]. It is not yet clear however if these vectors have the ability to totally adapt to these new habitats in case of total deforestation. They probably still need patches of forest to survive during the dry season [7, 29], from which they can spread to other breeding areas, such as plantations, in the wet season. The fact that in the present study these vectors remain present, even in very deforested areas, suggests that even a small patch of fragmented forest would be enough, or that An. dirus s.l. is able to overcome larger distances than the 1.5 up to 3 km in [29] commonly thought. Therefore vector control measures should focus on these important vectors [34], even in very deforested areas, and should take into account their early- and outdoor-biting behaviour. Moreover, areas with high forest fragmentation will be more accessible to people [5], meaning that the vector-host contact could also increase with increasing forest fragmentation.

Although densities were almost similar between villages and nearby forest plots, 11 (65%) out of 17 infective bites were found in the forest plots of BY, BZ and OZ, which are the villages with the highest degree of conserved forest, being in line with previous observations of forest-related transmission [11, 29, 35]. However, as shown by the current entomological data and the seroprevalence data [19], a considerable part of the transmission also occurs in the villages. This is in contrast to previous observations in Cambodia and Vietnam, where transmission is mostly limited to forest [11, 31, 35] and the way to the forest [11].

In the study regions, people combine living in the villages with a second house near to their farm plot in the forest. In 2005, a great proportion of people sleeping in the forest slept unprotected (17 to 31%) particularly in Borkeo, the district with the highest transmission rate. It is clear that both human behaviour (sleeping unprotected in the forest, outdoor activities in the early evening in the village, sleeping unprotected in the village) [36] and vector behaviour (biting outdoors and early) provide ample opportunity for malaria transmission.

The data from the current study show that LLINs provide useful although only partial protection against malaria: 71% of the infectious bites occurred after 22.00, when people are expected to be sleeping under a net. All infective bites before 22.00 (29%) occurred in the forest plots, underlining the importance of additional protective methods in such environments (eg long-lasting insecticidal hammocks [14], topical or spatial repellents). In the western part of the country malaria transmission, mainly P. vivax, occurred later in the night and was only observed inside the villages and not in the forest plot. It has been observed that the use of LLINs can alter the biting behaviour of the vector species to earlier biting [37], either through selective pressure of the LLINs resulting in, for example, species replacement, or through adaptation of the vector species through phenotypic plasticity [15]. In the Ninh Thuan province in Vietnam, where LLIN use was reported to be much higher (85% in the villages and 53% in the forest) than in the Cambodian study region in 2005, a higher proportion of vector bites occurred before sleeping time [11] (61% in the villages, and 45% in the forest plots versus 37 and 38%, respectively for the present study). Interestingly, in some villages located in O’Chum and Pursat districts, a higher proportion of early biting An. dirus s.l. was observed in the villages (39%) as compared to the forest plots (26%). In these districts, people reported a high use of bed nets (treated and untreated) in the village, but not in the forest. An increasing trend of early biting due to increasing LLIN use cannot be excluded. The current study took place in 2005, when LLIN coverage and use was much lower than presently observed. The Cambodia Malaria Survey shows a national increase in LLIN use from 29 and 25% in 2004 and 2007, respectively, until more than 50% in 2010, both in the east and the west [38]. This coverage has increased in 2011 and 2012 after massive distribution campaigns of LLINs. Therefore, the present survey could serve as a baseline for future studies on the effect of vector control measures in Cambodia.

Besides the primary vectors An. dirus s.s. and An. mini-mus s.s., other vectors can also transmit malaria. The current study is the first to find An. barbirostris s.s. positive for P. malariae in Cambodia. An. barbirostris s.l. is a confirmed vector of P. falciparum in Timor, based on salivary gland infection [39], and has been found positive in CSP-ELISA in Indonesia [40], Sri Lanka [41] and Thailand [42]. These ELISA results however were not confirmed by PCR or heating the ELISA lysates, and as such they might consist of false positive reactions as reported earlier [21]. In a malaria-endemic region in Thailand, An. barbirostris s.l. is highly suspected for maintaining malaria transmission in the absence of the main vectors [43], but, despite the observed high densities (up to 14 B/M/N in the O’Chum district), the importance of An. barbirostris s.l. as a secondary vector in Cambodia is not known. In Thailand [33] and in Vietnam [11, 44], also members of the An. maculatus group have been found to carry Plasmodium sporozoites and are considered important primary or secondary vectors. In the current study however, none of more than 3,500 An. maculatus s.l. tested positive in the CSP-ELISA. Moreover, this study has only focused on the secondary vectors An. barbirostris s.l. and An. maculatus s.l.. Given the diversity of anophelines collected, more potential secondary vectors may be present, although not infected at the time of the surveys. Even if these secondary vector species on their own cannot maintain malaria transmission, co-occurrence of several secondary vector species could constitute a vector population which is capable of maintaining malaria transmission [45]. For example, in Vietnam, a combination of secondary vectors was shown to maintain transmission, though at a low rate [44].

Within the primary and secondary vectors, there is a large variation in anthropophily, exophily and early biting activity between regions in Southeast Asia [10, 12]. In the current study it has been confirmed that, as in other parts of Southeast Asia [12, 13, 28] the studied vectors in Cambodia bite early and outdoors. The combination of domestic animals present in all study villages, and the zoophilic behaviour of these primary [12, 46] and secondary [10, 28] vectors, reduces the impact of vector control measures such as ITNs and indoor residual spraying and alternative vector control methods should be explored [47].

The results of the serological evaluations of antibody responses to P. falciparum and P. vivax in relation to risk factors is discussed in detail elsewhere [19]. The current study has focused on a possible relation between the entomological and parasitological or serological data only. This study clearly shows that vector densities are not, or even negatively, correlated with the seroconver-sion rate. Therefore vector densities cannot be used as proxy for malaria transmission. As expected, the strongest predictor of the P. falciparum and P. vivax seroconversion rate in survey 2 was the P. falciparum and P. vivax EIR of survey 1. Therefore, this study confirms previous findings of correlation between entomological transmission and serological surveys [18]. 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 [17, 18]. However an entomological survey using human landing as the collection method is the only way to identify vector species involved in malaria transmission and their biting behaviour. As in the present study, the analysis was carried out on village level, it shows that the EIR is still an important tool in documenting trends in malaria transmission at local level, even in areas with low malaria transmission intensity, such as Cambodia.

The data presented in the current study, based on outdoor human landing collections, clearly emphasizes the importance of outdoor malaria transmission in the forest as well as in the village. Although vector species are present in all sampled forested villages (12), their densities vary according to villages, rather than region, without major differences between the sites of collections (inside the village or in the nearby forest plot). The consequences of forest fragmentation and deforestation on malaria transmission in Southeast Asia are difficult to predict as a wide diversity of forest, near-forest and non-forest malaria vectors occurred. A suboptimal habitat in the forest fragments might result in lower densities and survival of the main forest vectors, with a reduction of malaria transmission as a consequence. However, higher accessibility of forest fragments can result in a higher contact rate between man and vector. Additionally, in the long run, deforested areas or areas with fragmented forest might be invaded by other efficient malaria vectors which are now considered secondary vectors. Therefore, although the capacity of secondary vectors of Cambodia in maintaining malaria transmission in the absence of the primary vectors is not known, their behaviour should also be taken into account when applying vector control measures. Because of outdoor and early biting by primary and secondary vectors in Cambodia, a behavioural trait that can be selected in vectors by a higher use of LLINs, additional measures should be explored. Personal protection using LLIHs, or topical and spatial repellents can have added value in tackling residual malaria transmission.