Malaria prevention in highland Kenya: indoor residual house-spraying vs. insecticide-treated bednets
Dr Helen Guyatt Kenya Medical Research Institute/Wellcome Trust Collaborative Programme, PO Box 43640, Nairobi, Kenya. Fax: +254 2 711673; E-mail: firstname.lastname@example.org
This study compares the effectiveness and cost-effectiveness of indoor residual house-spraying (IRS) and insecticide-treated bednets (ITNs) against infection with Plasmodium falciparum as part of malaria control in the highlands of western Kenya. Homesteads operationally targeted for IRS and ITNs during a district-based emergency response undertaken by an international relief agency were selected at random for evaluation. Five hundred and ninety homesteads were selected (200 with no vector control, 200 with IRS and 190 with ITNs). In July 2000, residents in these homesteads were randomly sampled according to three age-groups: 6 months–4 years, 5–15 years, and > 15 years for the presence of P. falciparum antigen (Pf HRP-2) using the rapid whole blood immunochromatographic test (ICT). The prevalence of P. falciparum infection amongst household members not protected by either IRS or ITN was 13%. Sleeping under a treated bednet reduced the risk of infection by 63% (58–68%) and sleeping in a room sprayed with insecticide reduced the risk by 75% (73–76%). The economic cost per infection case prevented by IRS was US$ 9 compared to US$ 29 for ITNs. This study suggests that IRS may be both more effective and cheaper than ITNs in communities subjected to low, seasonal risks of infection and as such should be considered as part of the control armamentarium for malaria prevention.
Research over the last 20 years has proved that insecticide treated bednets (ITNs) could substantially reduce the burden caused by Plasmodium falciparum in Africa (Choi et al. 1995; Lengeler 1998; Armstrong-Schellenberg et al. 2001). ITNs are now widely promoted as a means of preventing man–vector contact in the control of malaria and form the cornerstone of disease prevention within the recently launched Roll Back Malaria initiative (Roll Back Malaria 2000). The role and cost-effectiveness of indoor residual-house spraying (IRS) for malaria control has received far less attention. A recent comparison of historical trials of IRS in Africa against contemporary evidence of the effects of ITNs highlighted the need to re-visit the comparative advantage of ITNs over IRS under a variety of endemic settings (Curtis & Mnzava 2000). Despite the widespread use and evaluation of ITNs and IRS under the epidemiological conditions of Asia (Verle et al. 1999; Kamolatanakul et al. 2001), there have been only two simultaneous evaluations of ITNs and IRS in Africa. In a high transmission area of Tanzania their effectiveness was equivalent but ITNs were cheaper (Curtis et al. 1999), whereas in the low transmission area of KwaZulu-Natal in South Africa, ITNs were more effective, but more expensive (Goodman et al. 2001). In this paper, we report upon an investigation of the cost-effectiveness of IRS and ITNs mounted as part of an emergency relief operation in an epidemic-prone area of western Kenya.
Study area and population
The ecology and climate of the western highlands of Kenya support conditions that favour low intensity, stable transmission of P. falciparum by two principal vectors, Anopheles gambiae ss and An. funestus (Adungo 1992). Transmission is acutely seasonal following the climatic variation in the area (Hay et al. 2002). The greatest concentration of clinical disease occurs between June and August with the extent of the burden varying considerably between years. The districts of Gucha and Kisii are located in the western highlands at altitudes between 1400 and 2200 m above sea level. The area is densely populated with a combined population of 957 250 in 1999 occupying 1450 km2 (CBS 2001). The population predominantly comprises rural farmers growing combinations of subsistence crops for home consumption and tea for cash income.
In July 1999, Kisii and Gucha districts suffered a malaria epidemic, prompting much political concern and nongovernmental, donor-supported investment in improving case management and vector control. In order to pre-empt a similar epidemic in 2000, one of the emergency relief organizations deployed to the area (Merlin) expanded a programme of vector control throughout the two districts.
In November 1999, Merlin established 32 organized community groups (OCGs) throughout Kisii and Gucha districts to act as conduits for the distribution of bednets and insecticide net treatments. OCGs were trained in net treatment, community awareness raising and managing local funds. The groups were supplied with nets and all the materials necessary for net treatment including K-Othrine 1% SC (deltamethrin; Aventis, Nairobi, Kenya). A total of 15 000 nets were distributed amongst the groups. The first batch of treated nets at the OCGs (160 nets in Kisii and 125 nets in Gucha) were sold at a highly subsidized price of Kshs. 50 (US$ 0.64) to pregnant women and women with children < 5 years from `poor' homesteads identified by the local government civil administration. All subsequent nets and treatments were sold at a combined price of Kshs. 350 (US$ 4.47). The ITNs were targeted at communities in the immediate vicinity of the OCGs.
In February 2000, local health staff identified high-risk communities likely to benefit most from aggressive preventative measures before the seasonal rise in malaria incidence in June. Priority areas were selected based on a number of criteria including whether communities were near swamps, underwent brick-making activities, had a history of high malaria morbidity and mortality or were remote from health care facilities. Between February and May 2000, four mobile teams, composed of Merlin and Ministry of Health staff, were trained by staff from the local supplier, Zeneca, and subsequently undertook IRS using Icon 10% WP (lambda-cyhalothrin, 100 g/kg ai; Zeneca, Nairobi, Kenya). Community volunteers supported the mobile teams using locally available spray pumps. This service was provided free of charge to homesteads. Delays in the supply of Icon resulted in some priority areas not being sprayed.
In July 2000 a homestead survey was undertaken in 5 of the 12 sublocations (clusters) in Nyamache Division (one of seven divisions in Gucha District). A strictly randomized study design was not possible because interventions were targeted to particular communities according to operational rather than experimental priorities. Instead, the operational framework allowed for random selection of homesteads from five neighbouring clusters in receipt of either ITNs or IRS. IRS had been carried out in three clusters (11 928 people) in May 2000. In the other two clusters (9180 people), there had been no IRS, but ITNs had been sold by a single OCG, from December 1999. In the two ITN clusters, a list was prepared in consultation with registers maintained by the OCG and the local administration staff on homesteads with one or more ITNs (ITN homesteads) and without ITNs (Control homesteads). In the three clusters targeted with IRS, a list of homesteads receiving IRS (IRS homesteads) was provided by local administrative staff. Approximately 200 homesteads were randomly selected in each of the three study arms (Control, ITN and IRS). Only IRS homesteads that did not have ITNs were selected. Only 190 ITN homesteads were identified. A postsurvey census of all homesteads in the five clusters established the incidence of hospital admissions during 1999.
Information on a series of basic demographic, travel, social and wealth indicators was collected from the selected homesteads, together with details on the use of ITNs and the coverage with IRS by each member of the homestead. Residents were classified into three age groups (6 months–4 years, 5–15 years, and over 15 years), and one individual from each age group was chosen at random from each homestead. From these individuals, a finger prick blood sample was taken for the determination of P. falciparum antigen (Pf HRP-2) using the rapid whole blood immunochromatographic test (ICT Diagnostics, Amrad, Australia). Hospital admissions since the beginning of the year were recorded. Given that IRS was initiated in May 2000 and ITNs had been in use since December 1999, only the results for control and IRS homesteads for the period January–April 2000 are presented as preintervention morbidity statistics. The study was approved by the National Ethical Committee.
A retrospective cost analysis of the vector control activities was undertaken in consultation with the Merlin team. Costs were structured according to cash expenditures by Merlin (and Zeneca for the IRS training) during the activities and opportunity costs of using existing Ministry of Health staff, Government Civil Administration staff and members of the community. Details of the cost analysis are given elsewhere (Guyatt et al. 2002). The economic costs included opportunity costs of using existing personnel and volunteers, and assumed the costs for capital items (bednets, spray pumps and megaphones) and training were annualized over their life expectancy and discounted at 3%. The life-expectancy of a bednet was assumed to be 5 years, and training was assumed to last 2 years before a refresher course would be required (Guyatt et al. 2002).
All data were entered twice into relational databases (Access 2000, Microsoft, Seattle, WA, USA) and verified for data entry errors. Analysis was performed using SPSS version 9.0 for Windows (Chicago, IL, USA) and EPI INFO 6.0 (CDC and WHO 1994). Significant differences in the median values of homestead variables (e.g. number of residents) in ITN or IRS homesteads compared with the control homesteads were tested using a two-tailed Mann–Witney U-test and differences in proportions (e.g. ownership of a radio) using Chi square. Differences in the rate of hospital admissions were assessed using a normal test to compare two incidences (Kirkwood 1988).
Significant differences between homesteads in the prevalence of malaria infection were tested using a Chi- squared test. The protective efficacy of both IRS and ITNs was calculated from 1 – risk ratio (prevalence of infection with intervention divided by prevalence of infection without intervention). Confidence intervals were calculated according to Miettinen using the formula given by Kirkwood (1988): risk ratio (1 ± 1.96/χ). A weighted risk ratio was estimated for the three age-groups according to the Mantel–Haenszel technique with Greenland–Robins confidence intervals using EPI INFO 6.0. The difference in the prevalence of infection in individuals examined in the control homesteads and the prevalence in the intervention homesteads (ITN or IRS) was used to assess `effectiveness', infection cases prevented. This was combined with the unit cost (cost per person protected) to provide the cost per infection case prevented.
The homesteads randomly selected from the three groups [control (no intervention), ITNs and IRS] were comparable with respect to demographic composition and wealth indicators (Table 1). Almost all residents examined were of the Wakisii ethnic group (99.7%) and had not travelled outside the study area in the last 2 months (99.4%). The health status of the study arms was also comparable, with no significant difference in the rate of hospital admissions between the study homesteads during the first 4 months of 2000 (control vs. IRS) or the study clusters during 1999 (control and ITNs vs. IRS) (Table 1).
The demographic, wealth and health indicators for homesteads in the three study groups
An effect of IRS and ITNs was observed first through differences in the proportion of homesteads with any of the three individuals examined showing a positive result to the malaria test. In 28% of the 200 homesteads with no intervention at least one of the three residents examined tested positive for parasite antigen, compared with 20% of 190 ITN homesteads (P=0.08). The prevalence of any malaria infection in 200 IRS homesteads was only 8%. This was significantly different from both control homesteads (P < 0.001) and ITN homesteads (P=0.0016).
The main comparison concerned the prevalence of malaria infection in all individuals examined. The prevalence of P. falciparum infection in the 600 individuals in the 200 homesteads not covered by ITNs or IRS was 12.7% (Table 2). In the 190 homesteads that possessed an ITN, approximately half of the residents examined for parasite antigen slept under a treated bed net. The prevalence of infection in those sleeping under a net (n=274) was 4.7%, suggesting that the risk of malaria infection was reduced by 63% compared to control homesteads (95% CI: 58–68%). In the 200 homesteads receiving IRS, 94% of the 600 individuals examined slept in a sprayed room. Sleeping in a sprayed room reduced the risk of infection by 75% (95% CI: 73–76%) (see Table 2). Stratification by age-group demonstrated that the effects were consistent across all ages, except for sleeping under a net for 5–15 year olds. The reduced protective efficacy in this group resulted in a slightly lower age-adjusted protective efficacy of ITNs of 59% (Table 2). There was no significant difference in the prevalence of infection between individuals sleeping under a net and those sleeping in a sprayed room (P=0.364).
The effect on Plasmodium falciparum
infection of sleeping in a sprayed room (IRS) or under an insecticide-treated bednet (ITN)
The cost per person protected for ITNs and IRS (Table 3) is based on a more detailed economic analysis summarized in Guyatt et al. (2002). For both the financial and economic analysis, IRS was consistently cheaper than ITNs, even when a provider's perspective of the financial costs is taken (which are net of costs recovered from the sale of ITNs). Taking the best-case cost scenario for ITNs (economic analysis), the annual cost per person protected with ITNs remains more than twice the amount for IRS (US$ 2.34 compared with US$ 0.88) (Guyatt et al. 2002). For ITNs, nearly two-thirds of the costs were attributed to the OCGs (62%), with the remaining 38% distributed equally between nets and insecticide. In contrast, three-quarters of the costs for IRS were consumed by the price of the insecticide. Opportunity costs for using existing personnel constituted < 5% of the total costs. Incorporating the estimated reductions in prevalence of infection from sleeping under an ITN (8%) or sleeping in a sprayed room (9.5%), the cost per infection case prevented in this population in July 2000 was estimated as US$ 29 for ITNs and US$ 9 for IRS.
The cost-coverage and cost-effectiveness ratios for ITNs and IRS1
The epidemiology of clinical malaria in the highlands of Eastern Africa is the subject of continued debate (Bodker et al. 2000; Shanks et al. 2000). The western highlands of Kenya are best described as low intensity transmission areas subject to acute within and between year variations in clinical burden. During the 1950s and 1960s control efforts effectively contained malaria in these highland areas through a combination of IRS, mass drug administration and chemoprophylaxis (Strangeways-Dixon 1950; Roberts 1964). This was followed by a period of quiescence for malaria until the late 1980s and early 1990s when a series of malaria `epidemics' were again reported in these regions (Some 1994; Malakooti et al. 1998; Cox et al. 1999).
The detection and containment of epidemics of malaria in Africa has re-emerged as an international priority as part of the Roll Back Malaria campaign (WHO 2001). Nevertheless, recommendations to national governments on best practices for early warning and detection systems or how to effectively manage epidemics are not based upon any empirical evidence of precision, applicability, or most importantly, cost-effectiveness (Najera et al. 1998). Without this kind of evidence, Ministries of Health in the subregion most affected by epidemics have faced considerable problems in defining responses to the public health crises caused by malaria epidemics.
Once the mainstay of control in some parts of Africa, IRS has been somewhat over-shadowed by the unmistakable successes of trials of ITNs. The role of IRS under specialized epidemiological conditions of Africa has been neglected by researchers. During the 1950s, Roberts demonstrated that three annual cycles of Dieldrin spraying following 2 years of mass drug administration with pyrimethamine in highland Nandi District could reduce the prevalence of infection from 23 to < 2% (Roberts 1964). Similarly in the highland lake Bunyonyi area of Kigezi in Uganda, two spraying cycles of DDT combined with mass drug administration (chloroquine and pyrimethamine) resulted in a more than 99% reduction in infection levels (De Zulueta et al. 1964). Since the late 1970s there have been few systematic efforts to prevent or contain epidemics through the use of IRS in the Kenyan highlands.
This study of the operational impact of two vector control activities upon the risks of exposure to P. falciparum infection in a semi-immune population suggests a comparative advantage, both in terms of cost and efficacy, of a single round of IRS over ITNs. IRS was shown to reduce the risk of malaria infection by 75% at a cost of less than US$ 0.90 per person protected. This proved to be less than half as expensive as providing ITNs and three times more cost-effective in reducing malaria infection. An important qualifier in interpreting the efficacy results is that these were obtained through observation in an operational setting rather than under conditions of a randomized control trial. However, if anything, both the costs and efficacy of IRS presented in this study may represent a worse-case scenario (highest costs and lowest efficacy). The costs cover the spraying of all rooms in the homestead (as this was the procedure under operational conditions) and the homesteads sprayed were identified as high priority, though there is no quantitative data to support that they were at higher risk from malaria. IRS was also well received by the homesteads, with all homesteads wanting to have their homes sprayed again the following year and 96% expressing a willingness to pay for this.
Excluding southern Africa, more than 54 million people in sub-Saharan Africa (12% of the total population) are estimated to reside in areas at risk of epidemic malaria (Snow et al. 1999). In Kenya, nearly 7 million people (23% of the total population) reside in the 15 districts defined as highland, `epidemic-prone' areas. There is an urgent need to examine the potential role of IRS in controlling malaria in these often densely populated and economically important areas of Africa.
We thank the people of Nyamache division who participated in the study; Una MacAskill (Programme Manager, Merlin) and the local Ministry of Health and Government Civil Administration staff in Gucha for their support. The paper is published with the permission of the Director, KEMRI. Financial support was provided by the Wellcome Trust as part of H.L.G.'s Research Career Development Fellowship (#055100). R.W.S. is in receipt of a Wellcome Trust Senior Research Fellowship (#033340).