Malaria Journal a Low-cost Repellent for Malaria Vectors in the Americas: Results of Two Field Trials in Guatemala and Peru

Background: The cost of mosquito repellents in Latin America has discouraged their wider use among the poor. To address this problem, a low-cost repellent was developed that reduces the level of expensive repellent actives by combining them with inexpensive fixatives that appear to slow repellent evaporation. The chosen actives were a mixture of para-menthane-diol (PMD) and lemongrass oil (LG).


Background
In 2005, the WHO reported that 41% of malaria cases occur outside of Africa [1]. This marked a significant increase over their 2001 estimate of 13.6% [2] and reflects a growing awareness of the malaria problem beyond Africa. Greater recognition of this should encourage increased research on malaria vector control in other areas of the globe, an important consideration, as vectors in these regions generally exhibit behavior patterns that make them less susceptible to control measures shown to be effective in Africa, such as insecticide treated bednets (ITNs) and indoor residual spraying (IRS).
These behaviors include tendencies to [i] outdoor resting e.g. Anopheles darlingi [3] and Anopheles dirus [4]; ii) outdoor feeding e.g. Anopheles minimus [4], An. darlingi [3], and Anopheles sinensis [4]; and iii) significant feeding activity during early evening e.g. Anopheles albimanus [5], Anopheles nuneztovari [5], Anopheles farauti No.2 [6] and An. darlingi [7]. A study for implementing ITNs in four Latin America countries showed that 25% of An. albimanus in Nicaragua, 28% of Anopheles punctimacula in Ecuador, 57% of An. albimanus in Peru, and 30% of An. nuneztovari, also in Peru, fed before 9:00 pm, when people are still active and often still outdoors [8]. More recently, a case-control study in Colombia [9] showed that ITNs provided only 50% reduction in malaria. The authors attributed this to bites received when people were not sleeping under their nets. Even in areas where ITNs are considered to be highly effective malaria control tools, it appears that their introduction may have caused behavioral shifts among malaria vectors, with outdoor and early evening feeding becoming more frequent among them.
In such conditions, ITNs may be usefully supplemented by an effective insect repellent [10,11]. A recent household randomized trial in Pakistan [12] confirmed that the widespread provision of a repellent soap incorporating DEET and permethrin can reduce the risk of malaria by >50%. Moreover, an unpublished clinical trial in the Bolivian Amazon [13], with a 30% para-menthane-diol (PMD) formulation, showed an 80% reduction in P. vivax in those using repellent and ITNs, compared to an ITN only group.
In countries burdened with malaria, nearly all families affected by the disease face severe economic pressure from loss of income and treatment costs. In fact, for the poorest families in Latin America these indirect costs may represent as much as 20% of annual household income [14,15]. In response to these conditions, a low-cost repellent lotion (LCR) was developed by S.T. Darling for distribution to malaria-endemic communities in the Americas.
This plant-based formulation (patent pending) incorporates two principal active ingredients: para-menthane-diol (PMD) and lemongrass oil (LG). PMD is derived from pine extractives or from lemon eucalyptus (Corymbia citriodora), a tree that is grown commercially in Latin America and China. PMD is a highly effective, broad-spectrum insect repellent [16][17][18] that has proven to lower malaria incidence where malaria transmission is maintained by An. darlingi [13]. Lemongrass oil, a distillate of Cymbopo-gon citratus or Cymbopogon flexuosus leaves, is traditionally used in many parts of the Americas to repel mosquitoes [19] and it is repellent to An. darlingi and other disease vectors (Moore et al., in preparation). To lower the cost of the repellent and maintain its efficacy, the PMD and LG actives were combined with some low-cost ingredients (fixatives) that extend the repellent effect by slowing the evaporation of volatile repellent actives [20]. The studies reported here were designed to measure whether such a repellent, containing lower amounts of actives than reported in previous studies, could prove its efficacy against malaria vectors and other mosquitoes at a muchreduced cost per application.

Study Site A
The field test was conducted near the Port of Champerico (091°55'W,14°18'N) on the Pacific coast of Guatemala, in June 2005. The estuaries and mangrove swamps that characterize the area are fed by rivers that descend from nearby volcanoes around Quezaltenango. The tests were conducted on a local finca -a farm where cattle are raised. The finca has a large permanent lagoon and swamp area that is a known An. albimanus breeding site. As An. albimanus feeds preferentially on cattle [21] it was ensured that livestock were enclosed 50 meters behind the volunteers, so that host-seeking mosquitoes emerging from the breeding site and attracted by the animal odours, would first encounter the collectors.

Study site B
The trial took place outside of Zungarococha, a small village near Iquitos, the capital of Loreto, Peru (3.8° S 73.2°W ). It was staged in late February 2006 in order to coincide with high An. darlingi populations, but while malaria transmission remained low. The region is lowland tropical forest with two tributaries of the Amazon River flowing through: the Nanay to the north, and the Itaya to the south. The volunteers collected mosquitoes in a sparsely wooded area that was situated between a group of houses and a lake, or cocha, fed by the river that is a permanent breeding site for An. darlingi.

Test Procedures. Study A and B
Both studies were controlled, double-blinded, Latinsquare designs that utilized human-landing catches from treated or untreated volunteers in order to measure repellency. All solutions were placed in unmarked containers labelled by code. On any one night, human volunteers had both lower legs treated with either the PMD/LG candidate repellents or a positive or negative control at a rate of 0.002 ml/cm 2 between the ankle and the knee. Volunteers' leg length and circumference were measured to calculate surface area, and the correct dose of treatment was measured using a micropipette. Repellent was then applied using a latex glove to minimize absorption of material by the hand of the volunteer. During the humanlanding catches, the volunteers wore shorts to the knee, work boots, and a loose bug jacket (ProBuy) to ensure that blood-seeking mosquitoes had access only to their lower legs. After midday, volunteers did not smoke, consume alcohol, or use soap when washing. This was intended to minimize variation in their headspace kairomones [22,23].
The designated locations within the field sites were 10 m from each other and a minimum of 20 m from alternate sources of kairomones such as houses and livestock. As insect repellents act over a distance of less than a meter, and the maximum distance of host attraction of a single human to mosquitoes is 10 m [24], this design minimizes the "relativity effect" wherein insects must choose between two hosts simultaneously. Mosquitoes were collected as soon as they landed on the exposed lower legs of the volunteers, but before probing of the skin commenced, using a mouth aspirator, flashlight, and collection vessel designed for this purpose. Collection vessels were changed each hour to provide hourly measures of repellence. Umbrellas were also provided to protect the volunteers from any rain showers that might wash away their repellent.

Test Procedures. Study A
The three repellents and two controls were applied to the five volunteers at 14.30 h, and human-landing catches were performed at the field site for one hour before and one hour after sunset (1730-1930 h), when the evening mosquito biting is at its peak as shown by preliminary human-landing catches. The times chosen allowed an assessment of the protection afforded by the repellent over five hours while exposing volunteers to bites for only two hours. This helped minimize exposure and risk to the collectors. The study was a balanced 5 × 5 Latin-square design that required each volunteer to test each treatment five times over a period of 25 nights. Every evening, each individual was allocated one of five treatments, and sat in one of five allocated positions. Consequently, the volunteers changed position every five days.

Test Procedures. Study B
The repellent and two controls were applied at 16.00 h, and man-landing collections were performed in the field, between 18.00 and 22.00 h, as this is the time of peak An. darlingi activity in the area. The times chosen allowed an assessment of the protection afforded by the repellent to be made for a period of six hours while exposing volunteers to bites for only four hours. This helped minimize exposure and risk to the collectors. The volunteers took a 15-minute break between 20.00 and 20.15 h. The study was a 3 × 3 balanced Latin-square design that required each volunteer to test each treatment three times over a period of nine nights, with volunteers changing positions every third night. Due to unforeseen circumstances, one of the volunteers was replaced after four nights, and this was factored into the statistical analysis.

Ethical Issues
All volunteers were experienced at conducting man-landing catches. A form outlining procedure was given to the volunteers to ensure that they had full understanding of the potential risks of a study of this kind. In addition, each was given a chloroquine (Guatemala) or mefloquine (Peru) prophylaxis in accordance with WHO guidelines. Full ethical approval was obtained from Universidad del Valle, Guatemala (Study A); and from London School of Hygiene and Tropical Medicine Ethics Board and Instituto Nacional de Salud, Peru (Study B).

Statistical Analysis
Mosquitoes were maintained overnight and killed by cooling prior to identification the following morning. Data were normalized after transformation with natural log (x+1), verified by Anderson-Darling Normality Test, and were analysed with General Linear Model (GLM) using SPSS 13 for Windows. The model measured the effect of position, individual, hour and treatment (as fixed factors) and day (as a random factor), on the transformed mosquito counts. Further post hoc testing of individual variables was performed using a Tukey's Honestly Significant Different test (Tukey's HSD).

Study A
In 25 nights, 6,140 mosquitoes were captured comprising 55.6% Psorophora varipes (Coquillett) and 24.8% Aedes ochlerotatus taeniorhynchus. The average number of mosquito landings on the negative control was 108 per person/hour and there was no significant difference in hourly numbers of landings in this treatment (F = 0.896, d.f. = 1, p = 0.345).
Each of the four repellents provided excellent protection from host-seeking mosquitoes, and the PMD/LG repellents provided >97% protection up to five hours after application, with T15 and T20 providing 99% protection. DEET (15%) provided 92% (Table 1). GLM analysis showed that there was a significant difference between the four repellents and the negative control, and between DEET and the three PMD based repellents; although there was no significant difference between the three PMD/LG repellents (Tukey's HSD, p < 0.0001) ( Table 1). Sources of error in the experimental design were also investigated. There was no significant difference between the collection positions within the field site (F = 2.149, d.f. = 4, p = 0.76), although individuals varied significantly in their mean attractiveness to mosquitoes/collection ability (F = 6.73, d.f. = 4, p < 0.0001).

Study B
In nine nights, 2,358 mosquitoes were captured, of which 86% were An. darlingi. The average number of landings on the negative control was 46.28 per person/hour. There was no significant difference in the hourly number of mosquitoes captured from the control (F = 1.167, d.f. = 3, p = 0.326), or An. darlingi (F = 1.667, d.f. = 3, p = 0.179) which indicates that the repellents' efficacy did not significantly decline during the six hours of the test.
The PMD/LG repellent significantly outperformed DEET, providing an average of 95% protection six hours after application (F = 128.8, d.f. = 2, p < 0.0001) ( Table 2). In contrast, 20% DEET provided an average of 64% protection over the duration of the trial. Sources of bias were investigated and there was no difference in the number of mosquitoes captured in the three positions within the field (F = 0.87, d.f. = 2, p = 0.422) or individual variation in attractiveness to mosquitoes/collection ability (F = 1.492, d.f. = 2, p = 0.230).

Discussion and Conclusion
In both field trials, the PMD/LG repellents with fixatives showed excellent efficacy against a broad range of mosquito species, with greater than expected longevity for a 15% PMD formulation. Also, in these trials the PMD/LG repellents showed greater efficacy than corresponding doses of DEET. This is an important point, as several other studies have shown PMD to have longevity similar to [17], or lower than [25], a corresponding dose of DEET. During a 2001 Bolivian field trial against An. darlingi where the biting pressure was 75 mosquito landings per person/ hour, a repellent containing 30 % PMD in ethanol provided 97 % protection, and 15% DEET provided 85% protection for four hours after application [17]. However, in the current Study A, the repellent containing half that concentration of PMD (T15) provided 99% protection for five hours after application with a biting pressure of 108 mosquitoes per person/hour, compared to 92% for 15% DEET. In Study B, a repellent with 16% PMD provided 95% protection for six hours after application, compared to 64% protection for 20% DEET.
It may be inferred from this that the addition of fixatives to the repellents tested in Guatemala and Peru slowed the release of repellent volatiles, thereby extending the repellents' duration and lowering its cost. Additionally, because PMD is the most costly ingredient in the repellent, the savings realized by a reduction in PMD content from 30% to 15% have reduced the cost of this disease prevention tool even more. There is much evidence that the indigenous poor are less likely to purchase ITNs and repellents, and more likely to rely upon cheaper, less effective methods of personal protection [26]. Therefore, mak- ing the repellent available at the lowest cost possible could enhance user acceptance.
The potential benefit of this portable protection to antimalaria campaigns in the Americas can be seen in the economics of the 1998 malaria epidemic in Peru. It has been estimated that the total cost of treating malaria that year was $190 per individual case [27]. That includes all associated state health expenses and the costs to the family for treatment, lost income, and death. Adjusted for inflation, that is approximately $250 in 2007. However, the estimated annual cost per person for the repellent intervention (coverage during the 7-month transmission season at $0.024/day) is $5.00. That is 2% of the estimated total cost of treatment today. Consequently, the PMD/LG repellent may be an excellent candidate for incorporation into existing vector control strategies.
If such economies could be achieved in Peru, while reducing the disease burden from malaria, this model might be applicable to other regions where early evening biting is problematic. Essentially, the model for "repelling" malaria proposes the following: where the crepuscular feeding behavior of the most significant malaria vectors is already established, or where it may be shifting to early evening biting or outdoor resting [28] (from selection due to IRS and/or ITN use), it may be possible to achieve a reduction of 60% in malaria cases [13]. This could be achieved by saturation of at-risk communities (to avoid diversion of malaria vectors to non-users) with high-efficacy LCRs that are affordable and aromatically attractive to the indigenous poor. Combined with ITNs, the reduction could be substantially greater.
A five-month Phase 3 study, currently under way in the Peruvian Amazon, is providing an opportunity to measure the parameters of this model on 16 population clus-ters. With 3,300 subjects divided into three cohorts (Repellent Only, Repellent + ITNs, and No Intervention), this community-wide study marks the first time that the effect of a Repellent Only intervention on malaria rates has been measured in a discrete population group in the tropics. If the LCR's demonstrated capacity to repel malaria vectors leads to a measurably significant reduction in the infection rate, the Puerta del Cielo Foundation will begin distributing repellents and treated bednets at cost to poor malaria-endemic communities throughout the Peruvian Amazon.
the study protocols in Peru, guided them through ethics committees there, and contributed to editing the manuscript.