Can source reduction of mosquito larval habitat reduce malaria transmission in Tigray, Ethiopia?

Authors


Authors
Mekonnen Yohannes and Mituku Haile, Mekelle University, Mekelle, Ethiopia. E-mail: fisah@freemail.et, mekelle.university@telecom.net.et
Tedros A. Ghebreyesus, Karen H. Witten and Asefaw Getachew, Tigray Regional Health Bureau, Mekelle, Ethiopia. E-mail: tedrosag@hotmail.com, witten\"@telecom.net.et, asefaw2000@yahoo.co.uk
Peter Byass, School of Community Health Sciences, University of Nottingham, Nottingham, UK; and Department of Epidemiology and Public Health, University of Umeå, Umeå, Sweden. E-mail: PeterByass@aol.com
Steve W. Lindsay (corresponding author), School of Biological and Biomedical Sciences, Durham University, South Road, Durham DH1 3LE, UK. Tel.: 0044 (0) 191 334 1349; Fax: 0044 (0) 191 334 1289. E-mail: s.w.lindsay@durham.ac.uk

Summary

The development of irrigation schemes by dam construction has led to an increased risk of malaria in Tigray, Ethiopia. We carried out a pilot study near a microdam to assess whether environmental management could reduce malaria transmission by Anopheles arabiensis, the main vector in Ethiopia. The study took place in Deba village, close to a dam; Maisheru village, situated 3–4 km away from the dam, acted as a control. Baseline entomological and clinical data were collected in both villages during the first 12 months. Source reduction, involving filling, draining and shading of potential mosquito-breeding habitats was carried out by the community of Deba in the second year and routine surveillance continued in both villages during the second year. Anopheles arabiensis was highly anthropophilic (Human Blood Index = 0.73), biting early in the night before people went to bed. The major breeding habitats associated with the dam were areas of seepage at the dam base (28%), leaking irrigation canals (16%), pools that formed along the bed of streams from the dam (13%), and man-made pools (12%). In the pre-intervention year, 5.9–7.2 times more adult vectors were found in the dam village compared with the control village. There was a 3.1% higher prevalence of an enlarged spleen in children under 10 years in the dam village than in the control village during the pre-intervention period, but no statistically significant difference was found in the incidence of falciparum malaria between the two villages during the same period. Source reduction was associated with a 49% (95% CI = 46.6–50.0) relative reduction in An. arabiensis adults in the dam village compared with the pre-intervention period. There were very few cases of malaria during the intervention period in both villages making it impossible to judge whether malaria incidence had been reduced. These preliminary findings suggest that in areas of low intensity transmission community-led larval control may be a cheap and effective method of controlling malaria. Further, large-scale studies are needed to confirm these findings.

Introduction

The construction of dams and development of irrigation schemes will provide many poor African farmers with greater food security, an improved diet and increased income. Nonetheless, there is concern that the introduction or expansion of irrigation systems in malaria-endemic areas may lead to a risk of malaria transmission, by creating more breeding habitat for vectors and extending the length of the transmission season (Hunter et al. 1982; Lindsay et al. 1991; Hunter et al. 1993; Ijumba & Lindsay 2001). This may be of enormous public health importance in areas of fringe transmission, such as the Ethiopian Highlands, where people have little or no immunity to malaria. In the Tigray region of northern Ethiopia, an extensive community-led microdam-based irrigation scheme is in progress, managed by the Tigray Sustainable Agricultural and Environmental Rehabilitation Commission (SAERT; SAERT 1994; WIC 2002). To date about 60 have been constructed ranging in size from 50 000 to 4000 000 m3 (unpublished data), and most are situated near human settlements. Water harvesting will continue in the region as a source of water for drinking, irrigation and power generation. Alarmed by the increasing danger of the continued presence of standing water on malaria transmission, we conducted an extensive study of the impact of these dams on malaria and found that malaria incidence in young children was sevenfold higher in communities near dams than those further away (Ghebreyesus et al. 1999).

Current control activities in Tigray, as in other parts of Ethiopia, mainly involve institution based diagnosis and treatment of cases, selective indoor residual spraying with DDT (and malathion in areas where DDT resistance is common) and presumptive treatment by community health workers (CHWs) at the village level (Ghebreyesus et al. 2000). Although there is a policy for CHWs to mobilize the community for environmental management during the main transmission season, this strategy has never been systematically evaluated. Apart from the general lack of resources with the available control measures, resistance of parasites to treatment and the vectors to DDT is a growing problem. Based on results of chloroquine-efficacy studies, the Ministry of Health in Ethiopia has recently changed the first-line treatment for malaria from chloroquine to Fansidar (sulphadoxine pyrimethamine; MOVBDCU 1999). Nevertheless, considering the history of drug resistance in other parts of the world, coupled with its wider use at a community level, development of resistance to this drug seems inevitable. There is also emphasis on community-wide use of insecticide-treated bed nets although this control strategy is still in its infancy in Ethiopia. The effective use and sustainability of bed nets is also uncertain in a society where malaria-related treatment and control was given free of charge and where there is a strong belief that the government should be the ultimate provider.

An integrated approach to malaria control that relies heavily on community involvement is one that may have the brightest future. The foundations of any such approach should be source reduction to reduce the level of malaria transmission in an area. This strategy needs to be based on a sound understanding of the local ecology and behaviour of the vectors. Source reduction through modifying larval habitats has been a neglected area of vector research and control although it was one of the main methods of malaria control and eradication efforts in many countries including the US, Israel and Italy (WHO 1982; Rafatjah 1988; Kitron & Spielman 1989; Ault 1994). Appropriate management of larval habitats in places where there are few vectors may help to suppress malaria transmission, as a proportional reduction in transmission is likely to lead to a similar reduction in clinical episodes of malaria in such areas (Macdonald 1956). We considered that environmental management may be effective in the highlands of Tigray an area of fringe malaria transmission where large community-based irrigation schemes are in progress. Thus, the main objective of this study was to assess the impact of dams on malaria transmission and test whether source reduction through filling, draining and shading of potential breeding sites could reduce malaria transmission.

Materials and methods

Study area

The study was undertaken near Meskebet microdam (14° 16′–14° 19′ N and 38° 11′–38° 14′ E), 330 km northwest of Mekelle in Tigray, northern Ethiopia (Figure 1). This is a highland area with raised plateaus intersected by broad canyons. The predominant crops grown in the area were maize (Zea mays), teff (Eragrostis abyssinica) and finger millet (Eleusine coracan). There are few trees and most vegetation consists of open savannah. Rainfall (900–1200 mm annually) occurs mainly from June to October, with a distinct peak in July. The rainy season is immediately followed by the coldest months of November through January. March–May is the hottest period of the year.

Figure 1.

Study area indicating the study villages (solid squares), dam (dark grey), major streams, automatic weather stations (solid circles) and the surrounding plateau (grey) ≥1800 m. Inset shows the position of the study site in Ethiopia.

Both study villages were in extremely similar topographical positions (Figure 1). Deba village is situated 0.1 km from Meskebet dam at an altitude of 1750–1790 m, while Maisheru village is 3–4 km away in a parallel valley at a similar altitude (1750–1790 m). Deba had a population of 372 individuals, while Maisheru's was 1237. Most of the villagers were farmers, who lived in ‘hidumos’: stone and mud-walled houses, with flat roofs of wattle and earth. These houses are typical of rural dwellings in Tigray and had open eaves between the wall and the roof. There were wide openings around the doors and often there were small windows in the house. The average number of people per household was 5.1 in both villages; most kept cattle, goats, sheep and pack animals, which were corralled in open stone-walled enclosures attached to their houses.

Meskebet dam, completed at the end of 1996, has a 4.9 km shoreline and a 570 m long dam embankment. It was made of compacted earth, and holds about 2.7 million m3 of water. The dam was designed to provide irrigation for 65 ha of land downstream although only 15 ha was under cultivation during the study period. The nearest house was about 100 m from the dam. There was an extensive area of seepage at the bottom of the dam embankment, irrigation canals leaked, irrigation waters overflowed, and puddles were common in the fields and in the irrigated valley below the dam. This was exacerbated because of substantial termite tunnelling under the fields, resulting in subterranean movement of water from the irrigation canals and production of numerous small pools at unwanted places and considerable wastage of water. The main, 2 km long, irrigation canal followed the contour at the edge of the village, within 70–200 m. This canal was not lined, except the first 200 m from the outlet where it leaked at several places, water logging the surrounding fields. Moreover, in some places, the main canal was inclined upwards, against the flow of water, causing water to overflow and leak from the sides. Bad construction of feeder canals also resulted in leakage.

Preparation of fields for irrigation and seedling plots started in December and the fields were irrigated using water from the dam between January and June after sowing and transplanting. The dam, irrigation canals, major waterways, and the position of each household were mapped using a global positioning system (Garmin 38, USA) after the selective availability was switched off in May 2000. Each household was numbered and demographic information recorded on a continuous basis. Both villages were sprayed with DDT in September 1999, during the pre-intervention phase because of a perceived threat of epidemic malaria in the area. Community Health Workers and enumerators treated clinical cases with Fansidar.

Study design

The study was conducted in two phases: a pre-intervention phase from March 1999 to February 2000 and an intervention phase from the second half of February to December 2000.

Clinical surveys

To monitor the incidence of malaria in the community blood samples were taken from all children under 10 years old in each village both actively and passively during the dry and wet seasons as described elsewhere (Ghebreyesus et al. 1999). Furthermore, a prevalence survey was carried out at the beginning of the rainy season (June 1999) and all children palpated for enlarged spleens during the pre-intervention phase (November 1999) only.

Entomological surveys

Larval survey.  The different types of water bodies in each village were identified at the beginning of the study and larval surveys made twice monthly. At each survey, the different types of breeding sites were sampled using standard dippers (350 ml). Up to 10 dips were made in each type of water body encountered at each survey and the presence or absence of mosquito larvae recorded. The location of each water body was recorded using a GPS. From each collection site larvae were transferred into separate vials, killed by gently heating and preserved in 70% alcohol. The larvae were then counted and identified (Verrone 1962b). Third and fourth instars were used for species identification of anophelines.

Adult collection.  Adult mosquitoes were sampled indoors using CDC light traps (Model 512; J.W. Hock Co., USA) and human-landing collections (WHO 1975). Collection by light traps was carried out in 30 houses each month. In order to obtain a representative sample of the mosquito population for the whole village, 10 houses were randomly selected from the edge of the village near the dam, 10 from the middle and 10 from the far side of the village. In the control village, the houses were roughly divided into three zones based on proximity to the nearest permanent breeding habitat on the valley floor and 10 houses randomly selected from each zone. Each day collections were made in six houses in each village, two from each zone. The same houses were used throughout the study. Light traps were operated between 18:00 and 07:00 hours the following morning. Each light trap was placed in a bedroom, near a wall, with the bulb about 45 cm above the head of a person sleeping under an untreated bed net. Light traps not working at collection were excluded from the analysis. Monthly knockdown collections were similarly made in the same 30 houses to supplement the number of specimens collected for blood meal analysis. Houses were sprayed with an aerosol of pyrethroids [Mobil; Mobil Africa Sales Inc., Belgium: composition (% weight): Tetramethrin 0.20; Phenothrin 0.12; Allethrin 0.25; Solvents, Propellants and essential oils 99.43]. To collect house-leaving mosquitoes, window exit traps were set at 18:00 hours and collected between 06:00 and 07:00 hours the following morning. In order to avoid any repellent effect of the insecticide, light trap collections were made in each house 5 days after space spraying. Human-landing catches were carried out in March, May, September, October and November during both years. Human baits collected mosquitoes landing on their exposed legs using aspirators and torchlight. Collections were made indoors and outdoors from 18:00 to 07:00 hours. Each month, catches were made in eight houses on the edge of each village nearest to the breeding habitat. Two houses were used each night, in each village, for four consecutive nights. At each house one collector was stationed indoors and the other outdoors 5–10 m from the house. Mosquitoes were collected at hourly intervals through the night and processed the following morning. Concurrent indoor and outdoor hourly light trap catches were also made in nearby houses for comparison with night-biting collection.

Processing of adult mosquitoes.  Adult anophelines were counted and identified to species based on morphological descriptions (Verrone 1962a; Gillies & Coetzee 1987). Samples of adult Anopheles gambiae s.l. were identified by polymerase chain reaction (PCR; Scott et al. 1993). The abdomen of freshly fed An. gambiae s.l. mosquitoes obtained indoors were squashed onto filter paper, dried and then tested in duplicate for traces of human and bovine (cow, sheep and goat combined) blood using a direct elisa technique adapted from Service and colleagues (Service et al. 1986). Head–thorax portions of samples of An. gambiae s.l. mosquitoes were also placed in the wells of microtiter plates, stored in a container with desiccant, and tested by the VecTestTM Malaria Sporozoite Antigen Panel Assay (Medical Analysis Systems Camarillo, CA, USA) for the presence of Plasmodium falciparum and P. vivax circumsporozoite proteins (CS; Pf, Pv210 and Pv247; Ryan et al. 2001).

Source reduction

Regular bi-monthly community-led larval intervention activities were carried out in the dam village from February to December 2000. Larval control was focused on water seepage areas and leaking canals, as these were the predominant breeding sites of An. gambiae s.l. Thus, the swamp created by seepage at the base of the dam embankment was drained in February 2000 and the drains regularly cleaned thereafter. Papyrus and other reeds were allowed to grow in the seepage area and planted in the gully and in areas that were difficult to drain. Entry of people and livestock into the seepage area and the gully below was prevented and the crossing points of cattle and humans along the riverbed were filled with rocks and gravel to prevent the creation of breeding sites in hoof and footprints. The streams in the gully were straightened by digging so that pooling did not occur at the sides. Unlined irrigation canals that leaked because of termite tunnelling were compacted. Simple drainage ditches were constructed to intercept the leaking water from the canals that had created water logging in the fields and these and the irrigation canals were regularly repaired. Sediment and vegetation were also removed from the irrigation and drainage canals to encourage water flow and prevent pooling. Puddles and abandoned water pools dug in the river or streambed for drinking purposes were filled with earth. Puddles and pools that were difficult to drain or fill were covered with the remains of reeds, grass and weeds so that mosquitoes could not breed in them because of mechanical obstruction and decomposition of the plant material. For mobilizing the community we used the existing administrative hierarchy and different local committees in the community. At the individual level, the inhabitants dug drainage canals across their fields and cleared vegetation from the feeder canals. The inhabitants of the control village learnt about the effective source reduction in the intervention village and carried out limited source reduction activities in their village at the end of the rainy season.

Data analysis

Incidence of malaria was based on rates of cases per 100 child months at risk as described elsewhere (Ghebreyesus et al. 1999). The enlarged spleen rates between the villages were compared using Fisher Exact Test. The presence of Anopheles arabiensis positive larval habitats was compared graphically, both for the pre- and post-intervention surveys. For each village, the geometric mean (GM) number of female An. arabiensis collected per catching effort was calculated for mosquitoes collected indoors and outdoors, both before and after intervention period.

The relative abundance of mosquitoes by light traps were compared between the two villages and seasons using Wilcoxon's Signed Ranks Test (W) for the pre-intervention year. The efficiency of light trap catches compared with human landing catches was determined by dividing the daily light trap catches (ln LTC + 1) by the matched human landing catches (ln HLC + 1). The biting cycle of An. arabiensis was also determined in both villages by hourly human landing and light trap collections as the percentage of mosquitoes caught each hour.

The relative reduction of adult mosquitoes by light trap collection during the intervention year was determined for the dam village. To correct for variation in mosquito numbers between the years, the GM of mosquitoes in the control village before (c1) and after (c2) intervention was used to adjust the mosquito numbers in the intervention village before (i1) and after (i2) intervention. The relative reduction (R) in the intervention was expressed as a percentage and calculated as follows (Curtis 1990):

image

Comparisons of the mosquito numbers in each village were made between the pre-intervention and the intervention year using the Mann–Whitney U-test.

Ethical clearance.  Approval for this study was obtained from the Tigray Regional Health Bureau and the Ethiopian Science and Technology Commission and local consent obtained from the local Baito through community meetings.

Results

Parasite and spleen surveys

The incidence rates in children <10 years old were too small to make any valid statistical comparisons about the impact of the interventions. The malaria prevalence rate, at the beginning of the rainy season (June 1999) during the pre-intervention period, however, was 2.3% (2/86) in the dam village and 0.3% (1/322) in the control village, 7.7-times greater than near the dam. All parasites found in the children's blood were P. falciparum. The spleen rate in under 10-year-old children was significantly higher [3.1% (3/93), Fisher exact test, P < 0.05] in the dam village (0%, 0/413).

Larval habitats

The percentage of positive pools encountered both during the pre-intervention and intervention phases are shown in Figure 2 for the dam village and in Figure 3 for the control village. In the dam village, a total of 61 positive pools were recorded during the pre-intervention year. Of these 68.8% were from four types of breeding sites (Table 1, Figure 2). No mosquito larvae were found in the dam itself. Mosquito larvae were found year round in the dam village but in the control village were restricted to rain pools in the wet season and the following 4–5 months to a swamp and riverbed pools associated with the rains (Table 1, Figure 3).

Figure 2.

Percentage of larval positive pools encountered for each type of breeding site in the dam village during the pre-intervention and intervention year surveys. n, number of surveys.

Figure 3.

Percentage of larval positive pools encountered for each type of breeding site in the control village during the pre-intervention and intervention year surveys. n, number of surveys.

Table 1.  A summary of aquatic forms collected during the pre-intervention and intervention years in the dam and control villages
ParameterIntervention villageControl village
Pre-interventionIntervention yearPre-interventionIntervention year
  1. * Significant difference pre-intervention vs intervention year in dam village (P = 0.003).

No. anopheline larvae1465441444625
% of Anopheles arabiensis third and fourth instars94 (673/720)73 (119/163)96 (201/209)83 (250/302)
% of water bodies with larvae*20 (61/312)10 (27/260)33 (24/72)38 (23/60)
Preferred breeding habitats (%)Seepage at dam base (28)Streambed pools (30)Riverbed pools (79)Riverbed pools (70)
Leaking canals (16)Seepage at dam base (20)Swamp (21)Rainfed pools (17)
Streambed pools (13)Man-made pools (11) Swamp (13)
Man-made pools (12)Leaking canals (11)  

In the dam village, a general reduction in the number of positive pools was observed in most types of breeding sites during the intervention year (mean no. positive sites = 19.4, 95% CI = 7.8–31.0, Mann–Whitney U = 19, z = −2.70882, P < 0.001) compared with the pre-intervention year (35.9, 95% CI = 23.8–48. However, in this village, there was no reduction in larval positive sites found in riverbed pools and a few positive breeding sites occurred in irrigation canals during the intervention year. In the control village, no reduction was evident during the intervention year. Similarly, a general reduction in the number of third and fourth instars was observed in the dam village in most types of breeding sites during the intervention year (mean no. of positive dips = 16.3, 95% CI = 7.3–25.3, Mann–Whitney U = 21, z = −2.5304, P < 0.001) compared with the pre-intervention year (66.3, 95% CI = 31.4–101.2). In the control village, no significant reduction was evident during the intervention year (30.2, 95% CI = 15.7–44.7, Mann–Whitney U = 37, z = −1.5195, P = 0.13) compared with the pre-intervention year (20.2, 95% CI = 6.8–33.6).

Adult collections

A total of 10 241 female mosquitoes were caught indoors by all methods (light traps, human landing and space sprays) in both villages during the study period. Light trap and human landing catches were used to get independent measurements of seasonal changes in malaria transmission intensity and space sprays for estimating the proportion of blood-fed mosquitoes feeding on people. A summary of adult mosquitoes collected during the study is shown in Tables 2 and 3. Of those caught before intervention, 82.1% (5469/6660) were An. arabiensis, 2.5% (164/6660) other anophelines and 15.4% (1027/6660) culicines. The composition during the intervention period was different from the previous year, when 26% (933/3581) were An. arabiensis, 41.9% (1500/3581) other anophelines and 32.1% (1148/3581) culicines. Overall, An. arabiensis constituted 70.3% (5559/7913) in the dam village and 36.2% (843/2328) in the control village during the whole study period. All 90 An. gambiae s.l. tested by PCR were An. arabiensis. Seventy-three per cent (142/194) of the An. arabiensis tested had fed on humans and 27% (52/194) on bovines. Of the human positive blood meals, 45.1% (64/142) were mixed or positive for bovine blood also. The ratio of humans to cattle was similar in both villages – 1:2.9 in the dam village and 1:2.2 in the control village. Of 2650 head–thorax specimens of An. arabiensis tested in groups of 10 for P. falciparum and P. vivax circumsporozoite proteins, only one positive specimen of P. falciparum was detected in the dam village giving a sporozoite rate of 0.04%.

Table 2.  A summary of adult mosquitoes collected during the pre-intervention and intervention years in the dam and control villages
 Intervention villageControl village
Pre-intervention (n = 12)Post-intervention (n = 10)Pre-intervention (n = 12)Post-intervention (n = 10)
No. mosquitoes5468244511921136
Anopheles arabiensis (%)4784 (87.5)775 (31.7)685 (57.5)158 (13.9)
Other anophelines (%)98 (1.8)979 (40)66 (5.5)521 (45.9)
Culicines (%)586 (10.7)691 (28.3)441 (37)457 (40.2)
Table 3.  The geometric mean of Anopheles arabiensis collected in light traps during the pre-intervention and intervention periods in the dam and control village and the percentage reduction in vector abundance (in GM/trap/night) in the dam village during the intervention period (March 1999–February 2000 vs. March–December 2000)
MonthPre-interventionPost-intervention
Dam villageControl villageDam villageControl villageReduction (%)
Dry season (March–May)3.100.430.490.1035.6
Wet season (June–October)5.660.960.850.3153.8
Mean4.010.630.660.2049

Anopheles arabiensis populations in the study villages showed an early biting peak (19:00–20:00 hours) followed by a general decline through the night (Figure 4). Hourly light trap collections were positively correlated with hourly human landing catches. This association was slightly stronger with indoor collections (r2 = 0.80, F = 43.59, P < 0.001, n = 64) than outdoor ones (r2 = 0.68, F = 23.72, P < 0.001). In general, more An. arabiensis were caught with light traps indoors than human baits and light traps were particularly efficient at catching mosquitoes when few were active. Overall, light traps were 1.5-times (95% CI = 1.2–1.8) more efficient than human landing catches indoors, but were much less efficient with outdoor An. arabiensis populations (mean ratio = 0.65, 95% CI = 0.41–0.89). Because of this and as light trap collections were routinely carried out monthly, we have used light trap data to evaluate the community-led larval intervention study and to compare the pre-intervention seasonal and between village differences in mosquito abundance.

Figure 4.

Biting cycle of Anopheles arabiensis (indoors and outdoors) assessed by light traps set near sleepers under untreated bed nets in the dam and control villages (pooled data 1999. 2000).

Abundance of Anopheles arabiensis

The seasonal abundance of An. arabiensis collected in light traps during the study is shown in Figure 5. Overall, during the pre-intervention period, there were 6.3-times more An. arabiensis collected in traps in the dam village (GM = 4.4; 95% CI = 3.8–5.1; z = −12.641, n = 354, P < 0.001) than in the control village (GM = 0.6; 95% CI = 0.51–0.76). There were two peaks in mosquito numbers in the dam village; one in May during the dry season and one in September at the end of the wet season. In contrast mosquitoes were only caught from the control village during the rainy season. The sharp drop in adult mosquitoes after September (Figure 5) was attributed to residual DDT spraying in both villages.

Figure 5.

Seasonal abundance (geometric mean) of Anopheles arabiensis collected using light traps in pre-intervention and intervention period in the dam and remote villages.

During the intervention period, the dry season peak in the dam village was absent and only a slight increase in mosquito density was apparent at the end of the rainy season. Mosquito numbers in the dam village during the intervention year was comparable with that found in the control village during the pre-intervention year. During the intervention period, there was a large increase in abundance of other anophelines, both in the dam and in the control village notably Anopheles demeilloni.

Collections of An. arabiensis caught during the pre-intervention and intervention periods are summarized in Table 3. There were fewer An. arabiensis collected/trap-night in the dam village during the intervention (GM = 0.7, 95% CI = 0.5–0.8, Mann–Whitney U = 25404.5, z = −12.823, P < 0.001) than during the pre-intervention period (GM = 4.0, 95% CI = 3.4–4.6). Overall, there was a 83% reduction in An. arabiensis abundance during the intervention year compared with the pre-intervention year, and a 49% relative reduction if one adjusted for the drop in mosquito numbers in the control village during the intervention year. The relative reduction of vectors was greatest during the wet months of the year from June to October (Table 3). On a month-to-month basis, the reduction was highest during the dry month of May, when a second peak was observed during the pre-intervention year (Figure 5).

There was also a significant reduction in An. arabiensis abundance in the control village in the intervention year, although this village was the experimental control (pre-intervention, GM = 0.63, 95% CI = 0.51–0.76, intervention, GM = 0.2, CI = 0.15–0.26, Mann–Whitney U = 41 285, z = −6.07, P < 0.001). It was for this reason that the GM of mosquitoes in the control village before and after intervention was used to adjust the mosquito number in the dam village before and after intervention.

Community participation

Most villagers participated in the larval intervention activities. Overall, approximately 5000 man-hours were used to perform these activities during the 10-month intervention period. Drainage of the seepage area and construction of a drainage canal at the start of the intervention took the largest share of the man-hours (about 2420). Afterwards bi-monthly intervention activities were regularly undertaken. Farmers with irrigated plots also dug drainage canals in their fields. Together with the agricultural development agents the community enforced a law prohibiting livestock from entering boggy areas near the dam. These activities were greatly facilitated by the strong social system of organization in the region, with agricultural and development agents, as well as various committees, locally elected community leaders (members of the militia and local judges), serving the community without incentive. Mass mobilization of communities for various activities, including for soil and water conservation, for construction of roads and microdams, has been a common practice in Tigray for some time (Ghebreyesus et al. 2000). Unfortunately, during the study period the local situation changed drastically as war broke out nearby and community input was much lower than expected. As the area was at the frontline, the majority of the population, including the local administration, was directly and indirectly involved in the war effort. As a result, their participation was less than anticipated.

At the end of the study, the community and the local administration were briefed about the outcome of the pre- and post-intervention results. This was followed by on site evaluation of the larval intervention study and impact of dams by community members together with representatives of various government and non-governmental institutions of the region, including the Bureaus of Health, Agriculture, SAERT (a commission responsible for constructing dams), the local administration and others.

Discussion

The clinical and entomological surveys showed a higher potential for malaria transmission in the dam village. This study supports the findings of a much larger study, which observed a sevenfold increase in malaria incidence in villages close to dams compared with neighbouring villages without a dam (Ghebreyesus et al. 1999). Our study indicates that the increased malaria associated with dams results from a rise in mosquito numbers caused by more breeding habitats in fields irrigated with water from the dams or from water seeping from the foot of the dam. In this pilot study, we found 5.9–7.2 times more adult An. arabiensis in the dam village than the control village during the pre-intervention period. Importantly a second peak of mosquito numbers was observed in the dam village during irrigation during the dry season, indicating that transmission may no longer be restricted to the rainy season as is typical for the region. Seepage water from the base of the dam, leaking irrigation canals, waterlogged fields and pools along streambeds were the main source of An. arabiensis throughout the year. Although we found no aquatic stages of mosquitoes in the dam, mainly because of wave action and the presence of fish feeding on the larvae, dams can be important sources of adult mosquitoes if the water level drops and water collects in puddles around the edge of the dam (Hunter et al. 1982; Jewsbury & Imevbore 1988; Birley 1991). Shoreline puddling is more likely with smaller dams as they have a disproportionately greater shoreline relative to the surface area, than large dams and natural lakes (Baxter 1977; Jewsbury & Imevbore 1988).

Anopheles arabiensis collected indoors were highly anthropophilic in the study area (HBI = 73%), although we cannot exclude the possibility that outdoor resting vectors differ in their feeding preference. Most biting occurred early in the night, between 19:00 and 20:00 hours, which differs from the typical pattern of biting by An. gambiae s.l. where biting is greatest after midnight (Haddow 1942; Haddow et al. 1947; Gillies & De Meillon 1968; Surtees 1970; Chandler et al. 1975; Dukeen & Omer 1986; Braack et al. 1994; Githeko et al. 1996; Maxwell et al. 1998). The early biting activity observed in the study area may have important consequences for malaria control in Tigray since large numbers of insecticide-treated nets are being distributed in the region. The efficacy of the nets is likely to be compromised as approximately 70% of bites will occur before 20:00 hours, before people, including the children, go to sleep under their bed nets. A similar pattern of biting has also been reported in villages adjacent to Lake Zwai, in Ethiopia (Abose et al. 1998), suggesting that this behaviour may be widespread in the country. Interestingly, nearly 40 years ago in the Lake Zwai area most An. gambiae s.l. fed readily after 23:00 hours and little early evening feeding was recorded (Rishikesh 1966), suggesting that this biting early in the night has evolved since then. The early evening biting peak seen with An. arabiensis is likely to be a consequence of the long-term use of residual application of insecticides, notably DDT, selecting for this behaviour as similar shifts in the biting cycle has also been shown following DDT indoor spraying or the use of insecticide-treated bednets (Taylor 1975; Ismail et al. 1978; Charlwood & Graves 1987; Magesa et al. 1991; Mathenge et al. 2001).

Most importantly, this study provides evidence that environmental management was able to reduce the abundance of malaria vectors with minimum participation of the community and by utilizing local resources. It should be appreciated that the intervention took place close to a major military conflict and during peaceful times even better control would be expected.

Among the intervention activities involved, draining of the swampy area below the dam embankment, propagation and maintenance of swampy plants (reeds (Carex spp.) and papyrus (Cyperus papyrus) in areas near the dam where seepage occurred and the construction of simple drainage canals were the most effective methods for reducing larvae. Planting reeds and papyrus reduced An. arabiensis proliferation by shading the breeding sites, making them unsuitable for adult female mosquitoes to lay their eggs and helping to dry out the land. Prohibiting the entry of people and livestock in boggy areas and cultivated fields, together with filling the crossing points of cattle and humans along the riverbed with rocks and gravel, prevented the destruction of the plants and the creation of breeding sites in hoof- and footprints. Construction of simple drainage ditches intercepted water from the leaking canals mainly because of termite tunnelling and prevented waterlogging in nearby fields. These methods were effective because they targeted seepage water, dam-related streambed pools and leaking canals, which constituted the largest sources of vector breeding throughout the year.

The efficacy of source reduction was demonstrated by a significant reduction in the number of positive dips and mean number of larvae collected in the intervention year compared with pre-intervention levels. The nearly 1:1 ratio of immature to mature An. arabiensis larvae in the pre-intervention phase was also greatly reduced during the intervention phase (ratio 1:0.6), indicating the reduced longevity of larvae. This study shows that good water management could prevent the production of high numbers of adult vectors in villages located near dams. This was well-illustrated in a study conducted in Mwea rice irrigation scheme in Kenya, where intermittent irrigation was tested for controlling the breeding of Anopheles mosquitoes (Mutero et al. 2000). The ratio between the fourth and first stage larvae of An. arabiensis was only 0.08 in the intermittently irrigated plots compared with the ratio of 0.3–0.7 in the continuously flooded control subplots, indicating very low survival rates in the former because of the short life-span of the pools.

The effectiveness of the larval intervention activity was also reflected in the dramatic decrease in adult An. arabiensis populations in the dam village; a 83% reduction was achieved compared with the pre-intervention year and a 49% relative reduction was obtained if one adjusted for the finding that there were fewer mosquitoes in the control site during the intervention year. This control effort also reduced the length of the transmission season, with the intervention village showing a similar seasonality to that in the control village. This was unlikely to be a consequence of residual application of DDT in September in the pre-intervention year causing a reduction in transmission in the following year. This is because the residual effect of DDT is limited to a few months (6 months at most) and neither village was sprayed during the intervention period. The reduction in adult mosquitoes seen in the second year of the study may also have been due partly to the lower rainfall in the second year than the first (dam village, first year = 962 mm vs. second year = 709 mm; control village, 968 mm vs. 772 mm).

Source reduction has been used successfully in different parts of the world especially before the advent of DDT (Russel et al. 1963; WHO 1982; Stevens 1984; Sharma et al. 1986; Sharma 1987; Kitron & Spielman 1989; Ault 1994). Many of these control activities were combined with improving malaria surveillance and chemotherapy, public health education and motivating communities to help themselves. Although reports on deliberate propagation and maintenance of reeds and papyrus in swampy areas to control the breeding of An. gambiae s.l. are scarce, their ability to reduce mosquito production has long been recognized (Hancock 1934; Steyn 1946; Goma 1960). The drainage of swamps containing reeds or papyrus and other swamp grasses to increase the land available for cultivation in southwestern Ugandan highland areas has been blamed for increased vector breeding (Hancock 1934; Steyn 1946) and for establishing the transmission of malaria in these areas (Mouchet et al. 1998; Lindblade et al. 2000). Reclamation of swamps in other parts of Africa has also been associated with increased breeding sites of malaria vectors such as An. gambiae (De Meillon 1947; Gillies & De Meillon 1968). Source reduction for malaria control may not be effective in all settings and there is a need to define precisely where it will be most beneficial.

Community led environmental management measures involving massive planting of eucalyptus trees in swampy areas reduced breeding sites of An. culicifacies in India. The trees not only dried up the puddles where the local vector bred but also created a source of income for local people (Sharma et al. 1986; Sharma 1987; Dua et al. 1988). However, in our study area eucalyptus were readily destroyed by termites so this potential intervention is not an option here. Deliberate pollution of breeding sites with plant materials against anopheline mosquito breeding has also been identified (De Meillon 1947) although polluted sites are not environmentally friendly and may increase culicine numbers. In this study, reed mats and associated grasses, as well as weeds and other plant remains, that were abundant in the area because of the presence of the dam, were used to cover only those breeding sites difficult to drain.

Environmental management is often a cost-effective strategy for vector control and can be integrated relatively easily with other intervention measures used for the control of malaria that requires community participation (WHO 1982; Rafatjah 1988; Ault 1994). For example, in India, Sharma et al. (1986) have shown that environmental control measures would cost about 15% of the expenditure being incurred in areas under malathion spraying. Nevertheless, despite its cost effectiveness and environment friendly nature, progress has been slow in the wider use of these measures. This was due in part to the unwillingness or reluctance of programme officials to utilize such control activities that require longer preparation, perseverance and are relatively slow to produce results (Rafatjah 1988) when the quick deployment and sharp effect of insecticides spares them extra work burden and from taking responsibilities for outbreaks. However, especially in vast countries such as Ethiopia, budgetary, logistic and other constraints make it difficult to effectively coordinate, supervise and execute the measures. Nevertheless, increasing problems with the effectiveness of insecticides, their high cost and toxic risks associated with some of them have emphasized the need for alternative or supportive control measures such as environmental management (WHO 1982; Rafatjah 1988). Shrinking budgets available for public health also necessitate health workers and communities alike to be more self-reliant in terms of identifying and implementing disease control and preventive measures. It is therefore highly important to prepare communities for this and let them participate in the process of looking for the best option within the available means. Disease control through environmental management thus empowers community members to be more self-reliant under economically restrictive conditions. We feel that the responsibility for mitigating possible impacts of small dams not only rests with health authorities but should also be taken seriously by other government departments, such as the Bureaus of Agriculture and Water Resources Development, as well as NGOs involved in water development and management. In this respect involving development and home agents, who live and work at community level, in the control activities is crucial.

Ancillary