Prevention of malaria in Afghanistan through social marketing of insecticide-treated nets: evaluation of coverage and effectiveness by cross-sectional surveys and passive surveillance


Mark Rowland, London School of Hygiene & Tropical Medicine, Keppel Street, London WC1E 7HT, UK. Fax: 00 44 7299 4720; E-mail:


Malaria is often a major health problem in countries undergoing war or conflict owing to breakdown of health systems, displacement of vulnerable populations, and the increased risk of epidemics. After 23 years of conflict, malaria has become prevalent in many rural areas of Afghanistan. From 1993 to the present, a network of non-governmental organizations, co-ordinated by HealthNet International, has operated a programme of bednet sales and re-treatment in lowland areas. To examine whether a strategy based on insecticide-treated nets (ITN) is a viable public health solution to malaria, communities were given the opportunity to buy nets and then monitored to determine population coverage and disease control impact. This was carried out using two contrasting methods: cross-sectional surveys and passive surveillance from clinics using a case–control design. Nets were purchased by 59% of families. Cross-sectional surveys demonstrated a 59% reduction in the risk of Plasmodium falciparum infection among ITN users compared with non-users (OR 0.41; 95% CI 0.25–0.66). The passive surveillance method showed a comparable reduction in the risk of symptomatic P. falciparum malaria among ITN users (OR 0.31; 95% CI 0.21–0.47). The cross-sectional method showed a 50% reduction in risk of P. vivax infection in ITN users compared with non-users (OR 0.50; 95% CI 0.17–1.49) but this effect was not statistically significant. The passive surveillance method showed a 25% reduction in the risk of symptomatic P. vivax malaria (OR 0.75; 95% CI 0.66–0.85). ITN appeared to be less effective against P. vivax because of relapsing infections; hence an effect took more than one season to become apparent. Passive surveillance was cheaper to perform and gave results consistent with cross-sectional surveys. Untreated nets provided some protection. Data on socioeconomic status, a potential confounding factor, was not collected. However, at the time of net sales, there was no difference in malaria prevalence between buyers and non-buyers. The abundance of Anopheles stephensi, the main vector, did not appear to be affected by ITN. ITN constitute one of the few feasible options for protection against malaria in chronic emergencies.


Controlled trials have demonstrated that insecticide-treated nets (ITN) can prevent malaria morbidity and mortality under a variety of epidemiological conditions (Choi et al. 1995; Lengeler 1998). This has stimulated many governments, UN agencies and non-governmental organizations (NGOs) to embark upon programmes of net distribution and treatment (Chavasse et al. 1999). While the utility of ITN is no longer in doubt, it is far from certain whether delivery and sales through the less controlled conditions that exist within development projects, or through the private sector, can approach the same coverage and impact as carefully executed trials (D'Alessandro et al. 1995; Lengeler & Snow 1996). The degree of uncertainty is still greater when the country being targeted is undergoing civil conflict or just emerging from conflict. Such places are increasingly the subject of internationally aided ITN distribution projects, yet no evidence has been gathered to indicate that such distributions have any impact on malaria (Rowland & Nosten 2001). As ITN projects continue to grow in size and number in both politically stable and unstable environments, methods for monitoring effectiveness and coverage become paramount to monitor progress and as a means for improving strategy.

Until the last decade, the use of bednets in Afghanistan was uncommon (Rowland et al. 1996). But with the country racked by 20 years of war, the public health system in disarray, and malaria becoming a major national problem, a strategy based on ITN for personal protection seemed the only feasible solution (Rowland et al. 1997). An efficacy trial in the secure environment of an Afghan refugee camp just over the border in Pakistan demonstrated that Afghans were ready to adopt the habit and stood to gain up to 70% protection against Plasmodium falciparum and 56% protection against P. vivax infection (Rowland et al. 1996). In 1992, HealthNet International (HNI) initiated social marketing of ITN through a network of district clinics run by several international and local NGOs in the eastern region of the country.

Within 2 years, the programme's initial objective of seeding demand across the region had been achieved, and the sales revenue generated was contributing towards restocking of nets. Popularity and progress was such that the network began to consider whether a malaria control strategy based on sale of ITN could achieve a coverage and epidemiological impact significant enough to replace the old strategy of indoor residual spraying with DDT that existed before the war. Groups of villages were given the opportunity to buy nets over a 2-year period. To measure impact, two methods were compared: the first involved repeated cross-sectional malaria parasite surveys among net-using and non-using families, the second involved passive surveillance of malaria cases at district clinics. A variant on the first method has recently been applied in Tanzania to evaluate the effectiveness of community-based ITN social marketing (Abdulla et al. 2001). The second method has been applied in Afghan refugee camps to evaluate the long-term effectiveness of ITN (Rowland et al. 1997).


Study area

The study was conducted in the rural districts of Behsud and Chaprahar in Nangarhar province, eastern Afghanistan (Figure 1). By the early 1990s the focus of the war had shifted to the centre and north of the country and the east had become secure. Most villages in Nangarhar had been partly destroyed during fighting but by 1995, the first year of the study, many inhabitants had returned from refugee camps in Pakistan, their homes rebuilt, the area de-mined, irrigation systems repaired, and agriculture stepped up. Agriculture was the main livelihood. The main crops grown were wheat from March to May, rice from June to November, and vegetables. Some farmers grew opium poppy.

Figure 1.

Map of eastern Afghanistan and bednet trial area.

Nangarhar marks the western edge of the monsoon rains, which fall in July and August. Rivers are also snow-fed. The main malaria vectors in eastern Afghanistan are Anopheles stephensi, A. fluviatilis, A. culicifacies, and A. pulcherrimus (Rowland et al. 2002). In Behsud and Chaprahar, these species breed in or around rice fields from June onwards. Malaria is both seasonal and unstable in Afghanistan, and outbreaks often occur in areas lacking clinical services. P. vivax accounts for 85% and P. falciparum for 15% of malaria infections. Incidence of P. vivax reaches a peak in August, P. falciparum in October – November, with transmission of both species continuing until the end of the year (Rowland et al. 2002). Chloroquine resistant P. falciparum is widespread (Rab et al. 2001), and is partly responsible for its resurgence in the last decade (Shah et al. 1997).

The interventions

The sequence of activities is summarized in the flow chart (Figure 2). Criteria for selecting the two study areas were presence of extensive rice culture and thus vector breeding, reputable NGO clinics able to diagnose malaria using microscopy, and little or no previous targeting with nets. Nets were introduced in a stepped manner in six villages from Behsud and in six villages from Chaprahar districts. Three of six villages from each district were randomly selected to receive nets in June 1995 and the other three received nets in June 1996. The purchase price per net was 8000 afghanis (US$2.9) in 1995, increasing to 35 000 afghanis (US$3.3) in 1996 as a result of inflation. This constituted a 25% subsidy on the manufacturer's price. A village was given a day's advance notice of the arrival of a mobile sales team, with the news being broadcast from the mosques' minarets. On arrival, the team gave a public health education session about malaria and the use of ITN using illustrated silk screen flip charts, registered the families who wanted to buy, and treated the nets in full public view. This mode of implementation was the standard approach used by HNI during day-to-day sales operations. The 12 villages were randomized with eight being offered ITN (polyester nets treated with either permethrin 500 mg/m2 or lambdacyhalothrin 25 mg/m2) and four being offered untreated nets (UTN). Treatment was free at the time of sales, but re-treatment was offered for US$0.05 per net in both 1996 and 1997.

Figure 2.

Key activities flow chart.

Cross-sectional parasite surveys were conducted in June and November of 1995 and 1996. Villages were census-surveyed and 25 households from each village were randomly selected from the enumeration lists. Blood smears were taken from all household members present on the day of survey, and each individual was asked whether they had been using a bednet recently (during the last fortnight). Individuals were classified as bednet users or non-users according to their response. About 15% of all households were sampled at each survey. Households were randomly selected anew from the enumeration list before each new survey. Individuals positive for malaria parasites were treated with chloroquine (25 mg per kg body weight in divided doses).

Passive surveillance of malaria was carried out from July 1995 to December 1997 through two clinics in Behsud and two clinics in Chaprahar. Microscopy was performed routinely on all patients presenting with a history of fever suspected of being malaria. Patients were asked whether they had been using a bednet recently. Positive cases were treated with chloroquine. Ideally, each patient would have been questioned about socio-economic status, in order to adjust for this potential confounder later on. But this would have placed an extra workload on the clinic staff and difficult to monitor. To raise diagnosis of malaria to a consistent standard, all microscopists received refresher training at the start of the study. Their performance was monitored monthly by taking samples of slides for re-examination and quality control.

For entomological surveillance, 15 sentinel houses were randomly selected from the enumeration list of each village, and collections of mosquitoes were made once a month by space-spray pyrethroid aerosol in a sleeping room and animal shed of each house. Collections were identified to species and scored. The same houses were monitored throughout.

Statistical analysis

We grouped the villages according to intervention received: (1) ITN from June 1995 onwards; (2) UTN from June 1995 onwards; (3) ITN from June 1996 onwards, and (4) UTN from 1996 onwards. We compared the odds of parasitaemia in bednet users vs. non-users within these groups. For the analysis of passive data, individuals presenting with symptoms of malaria and a smear positive for malaria parasites were classified as cases, distinguishing between falciparum and vivax. Patients presenting with possible symptoms of malaria at the same health centres during the same period, but with a negative smear, were classified as controls. We fitted logistic regression models adjusting for the effect of age, sex, village and district, and estimated the odds ratio (OR) of bednet use among cases compared with controls.

With entomological data, the monthly totals of each village were log transformed and analysed using Poisson regression.


Net sales and coverage

In total, 1680 households and 14 538 people from 12 villages were involved in the study. Population characteristics of the four village groups are shown in Table 1.

Table 1.  Baseline characteristics
CharacteristicsGroup 1Group 2Group 3Group 4
  • Group 1: Insecticide-treated nets (ITN) from June 1995 onwards.

  • Group 2: Untreated nets (UTN) from June 1995 onwards.

  • Group 3: ITN from June 1996 onwards.

  • Group 4: UTN from June 1996 onwards.

  • *

     Significantly different compared to group 1 (P < 0.01).

Number of villages4242
Number of households514440378348
Mean household/village (range)128 (77–176)110 (46–248)189 (153–225)174 (43–305)
Total population4852382431582580
Mean population/village (range)1213 (644–1999)956 (293–1486)1579 (1288–1990)1290 (313–2268)
Mean household size9.
Age distribution (%)
 0–4 years549 (21)448 (22)268 (21)287 (21)
 5–9 years623 (23)504 (25)360 (28)363 (26)
 10–14 years298 (11)271 (13)173 (13)164 (12)
 14+ years1197 (45)817 (40)497 (38)576 (41)
 Mean age18.317.316.917.3
 Male:Female ratio46:5448:5250:5047:53
Baseline parasite prevalence
 Sample size2272172810791286
 Positive for P. falciparum (%)17 (0.7)16 (0.8)15 (1.2)5 (0.4)
 Positive for P. vivax (%)170 (6.9)184 (9.5)*128 (10.5)*97 (7.0)

Twelve per cent of families (209/1680) had bought nets prior to the study. Within the six villages offered nets in year 1, 56% (530/954) of families took up the offer (range between villages 46%−83%) (Table 2). The following year 108 families bought nets a second time, but only 36 families were first time buyers, indicating a rapid penetration and saturation of the households willing or able to buy. Among the six villages offered nets in year 2 for the first time, a similar proportion took up the offer (54%, range 35%−72%). Seventy-one per cent (95% CI: 61%−79%) of families that took up the offer had purchased sufficient nets to cover their entire family. Re-treatment rates among village groups buying ITN were 64% (193/302) in year 2 and 47% (154/329) in year 3, despite only a nominal charge being levied (Table 2).

Table 2.  Uptake of nets and re-treatment in the four village-intervention groups
IndicatorsGroup 1
n (%)
Group 2
n (%)
Group 3
n (%)
Group 4
n (%)
Number of families buying nets in 1995302 (59%)228 (52%)00
Number of nets sold in 1995133980600
Average number of persons per net in net owning families in 19952.12.5
Number of families buying nets for the first time in 199625 (5%)11 (2.5%)255 (68%)135 (39%)
Number of nets sold in 1996250142730395
Average number of persons per net in net owning families in 19962.
Number of families re-treating nets in 1996193 (64%)
Number of nets re-treated in 1996816 (61%)
Number of families re-treating nets in 1997154 (47%)138 (54%)
Number of nets re-treated in 1997388 (24%)326 (45%)

Cross-sectional surveys

The prevalence of malaria at admission in June 1995 is shown in Table 1. The commoner species in mid summer was P. vivax. Both species were more prevalent in the 10–14-year age group and less prevalent in adults. Malaria was three to four times more prevalent in Behsud than in Chaprahar, and approximately 1.5 times more prevalent in village groups 2 and 3 than in groups 1 and 4. At the time of the second survey, nets had been in use for 5 months in village groups 1 and 2. In neither group was a significant effect of nets observed against either species of malaria (Table 3).

Table 3.  Cross-sectional study: comparison of parasitaemia between net users and non-users in the four village-intervention groups
 Parasite negative inP. falciparum positives (% positive) inP. vivax positives (% positive) inOR of P. falciparum
in net users vs.
OR of P. vivax
in net users vs.
Intervention groupsUsersNon-usersUsersNon-usersUsersNon-users
  • † 

    Odds ratios adjusted for the effect of age, sex and district.

  • ‡ 

    Treated net was in use for 5 months.

  • § 

    Untreated net was in use for 5 months.

  • ¶ 

    Treated net was in use for 12 months.

  • †† 

    Untreated net was in use for 12 months.

  • ‡‡ 

    Treated net was in use for < 1 month.

  • §§ 

    Untreated net was in use for < 1 month.

  • ¶¶ 

    Treated net was in use for 17 months.

  • ††† 

    Untreated net was in use for 17 months.

  • ‡‡‡ 

    Treated net was in use for 5 months.

  • §§§ 

    Untreated net was in use for 5 months.

  • * P < 0.05; **P < 0.01.

Survey 2
 Group 1 (ITN)70749912 1.714 (2.7)29 (3.9)22 (4.2)0.59 (0.27–1.30)0.95 (0.54–1.99)
 Group 2 (UTN)§2072615 (2.4)8 (3.0)13 (5.9)10 (3.7)0.78 (0.25–2.45)1.64 (0.69–3.91)
Survey 3
 Group 1 (ITN)11473661 (0.1)1 (0.3)17 (1.5)16 (4.2)0.34 (0.02–5.6)0.29 (0.14–0.59)**
 Group 2 (UTN)††4622060 (0)0 (0)32 (6.5)5 (2.4)2.79 (1.06–7.38)*
 Group 3 (ITN)‡‡7844155 (0.6)1 (0.2)52 (6.2)21 (4.8)2.74 (0.30–60.1)1.24 (0.74–2.13)
 Group 4 (UTN)§§2063081 (0.5)0 (0)6 (2.8)9 (2.8)1.01 (0.35–2.90)
Survey 4
 Group 1 (ITN)¶¶5918126 (1.0)20 (2.4)4 (0.7)11 (1.3)0.41 (0.17–1.00)*0.50 (0.17–1.49)
 Group 2 (UTN)†††1434224 (2.7)19 (4.3)5 (3.4)10 (2.3)0.65 (0.23–1.86)1.54 (0.54–4.39)
 Group 3 (ITN)‡‡‡36675710 (2.7)42 (5.3)7 (1.9)36 (4.5)0.49 (0.25–0.98)*0.40 (0.18–0.89)*
 Group 4 (UTN)§§§1723520 (0)18 (4.9)5 (2.8)8 (.2.2)0.00 (0.00–0.44)**1.27 (0.43–3.77)

At the time of the third survey, a significant effect of ITN was becoming evident against P. vivax in group 1, prevalence being almost threefold less among net users than among non-users. The vivax cases presenting at this time would have been transmitted in the previous summer–autumn when nets were already in use. Within groups 3 and 4, where nets had been purchased only a week before, there was no difference in P. vivax prevalence between net-using and non-using groups. This was an important finding because it indicated that within these villages there appeared to be no inherent difference in malaria risk between families that chose to buy nets and those that chose not to. Hence any subsequent difference in malaria prevalence arising between users and non-users in later surveys would probably be an effect of net use rather than unknown socioeconomic or behavioural factors.

At the time of the fourth survey, the prevalence of P. falciparum was significantly less in net users than in non-users in three of four groups, and even in the fourth group the trend was in the same direction as the others. Overall, there was a 59% reduction in the odds of P. falciparum parasitaemia in net users compared with non-users after adjusting for other predictors (OR 0.41, 95% CI 0.25–0.66). The OR for P. vivax in net users vs. non-users in group 1 had increased from 0.29 to 0.50 during the fourth survey. A significant effect against P. vivax was evident when groups 1 and 3, which had been using ITN, were combined (OR 0.38, 95% CI 0.20–0.74), but not in groups using UTN (Table 3).

Overall, the prevalence of P. vivax reduced from 4.4% (52/1143) in the second survey to 2.6% (21/1313) in the fourth survey among net users while it remained relatively stable from 3.4% (61/1811) to 3.2% (78/2527) among non-users. The prevalence of P. falciparum remained stable from 1.9% (25/1143) to 2.2% (25/1313) among net users while it increased from 3.8% (68/1811) to 4.7% (78/2527) among non-users.

Passive surveillance

Figure 3 shows the relationship between the monthly variation in bednet use, mean temperature, and mosquito density during 1995 and 1996. During the winter months net users comprised less than 5% of clinic attenders. Peak attendance by users was in August 1996 (36.0%). Net use rates correlated both with monthly changes in mosquito abundance (r = 0.78) and with ambient temperature (r = 0.87). Despite the hot, stuffy summer nights, net owners were presumably making use of nets to protect themselves against the greater discomfort caused by mosquito biting.

Figure 3.

Relationship between net usage, mosquito density and temperature.

The passive surveillance study confirmed that the differences in the distribution of age and sex between net users and non-users were negligible. Net use grew from 24% (1429/6023) in 1995 to 34% (1470/4300) in 1997. The odds of malaria infection was greater for the 5–14-year age groups than for the 0–4-year and >15-year age groups, and greater for females than for males. The odds of infection were lower for village group 1 that had used ITN for 3 years, and was lower for year 3 than for years 2 and 1 (Table 4). There was a negative association between use of bednets and malaria (OR 0.34 for P. falciparum and 0.89 for P. vivax). There was a significant protective effect of nets against P. falciparum in all village groups except for group 4 that had received UTN only in year 2 (Table 5). A significant protective effect against P. vivax was only apparent in the village group 1 that had used ITN since 1995.

Table 4.  Case–control study based on passive surveillance at district clinics
PredictorsControlsCases P. falciparum (% positive)Cases P. vivax (% positive) ORP. falciparum ORP. vivax
  • All the predictors shown in the table were included in the logistic model.

  • *

     P < 0.05;

  • **

     P < 0.01;

  • ***

     P < 0.001.

Age group
 0–4425971 (1.6)682 (14)1.01.0
 5–93400159 (4.5)812 (19)2.7 (2.0–3.6)***1.6 (1.4–1.7)***
 10–14209490 (4.1)450 (18)2.4 (1.8–3.4)***1.4 (1.2–1.6)***
 15+7963233 (2.8)1186 (13)1.8 (1.4–2.4)**1.0 (0.9–1.1)
 Male10351273 (2.6)1634 (14)1.01.0
 Female7429280 (3.6)1506 (17)1.4 (1.2–1.7)***1.2 (1.1–1.3)
 Behsud13804343 (2.4)1614 (11)1.01.0
 Chaprahar3988210 (5.0)1526 (28)0.49 (0.41–0.59)***0.33 (0.30–0.35)***
Village group
 112074212 (1.7)1507 (11)1.01.0
 23325202 (5.8)863 (21)3.0 (2.5–3.7)***1.9 (1.7–2.0)***
 3109493 (7.8)304 (22)4.2 (3.2–5.4)***2.0 (1.7–2.3)***
 4121140 (3.2)456 (27)1.5 (1.0–2.1)*2.6 (2.3–3.0)***
 19954175336 (7.4)1519 (27)1.01.0
 19963869150 (3.7)751 (16)0.67 (0.54–0.83)***0.69 (0.63–0.77)***
 1997429328 (0.6)241 (5)0.10 (0.07–0.15)***0.18 (0.15–0.21)***
 Non-users11263479 (4.1)2335 (17)1.01.0
 Users453755 (1.2)723 (14)0.34 (0.26–0.46)***0.89 (0.81–0.97)**
Table 5.  Passive surveillance case–control study: effect of use of bednets on clinical malaria in the four village-intervention groups
 ControlsP. falciparum casesP. vivax casesOR of net use in P. falciparum casesOR of net use in P. vivax cases
Intervention groupsUsersNon-usersUsersNon-usersUsersNon-users
  • † 

    Adjusted for the effect of age, sex, district and study period.

  • ‡ 

    Most nets were in use for 30 months from July 1995 to December 1997.

  • § 

    Nets were in use for 18 months from July 1996 to December 1997.

  • * P < 0.01; **P < 0.001.

Group 1 (ITN)346768102616939510420.31 (0.21–0.47)**0.75 (0.66–0.85)**
Group 2 (UTN)267813984882160.33 (0.16–0.67)*1.22 (0.92–1.64)
Group 3 (ITN)§6612499161841926610.34 (0.21–0.58)**1.11 (0.92–1.33)
Group 4 (UTN)§1271073436454091.15 (0.40–3.31)1.01 (0.70–1.45)

Individual protective effectiveness [IPE = (1 − OR)%] of ITN against P. falciparum infection from the cross-sectional surveys and the passive surveillance case–control study were 59% and 69%, respectively, and against P. vivax 50% and 25%, respectively. Community protective effect (CPE) is defined as the product of IPE × net coverage (Lengeler & Snow 1996). Net coverage estimates obtained from the cross-sectional survey and passive surveillance were 57% and 34%, respectively. Thus the CPE of ITN was 20.1% against P. falciparum and 17.0% against P. vivax from cross-sectional surveys and 23.5% against P. falciparum and 8.5% against P. vivax from passive surveillance.


Mosquitoes were more numerous in animal sheds than in sleeping rooms (Table 6). Mosquitoes were 3.4 times more abundant in animal sheds in 1996 than in 1995 (P < 0.001), but were not significantly more abundant in sleeping rooms in 1996 than in 1995. There was no significant effect on mosquito density of ITN over UTN or of ITN over no-nets in sleeping rooms or animal sheds in 1995. However, in 1996, there was a significant effect of ITN over UTN in sleeping rooms (rate ratio 0.55, 95% CI: 0.33–0.92, P = 0.03). This effect of ITN was not apparent in animal rooms (P = 0.45) and thus ITN did not appear to reduce mosquito densities overall.

Table 6.  Mosquito density per room: Anopheles stephensi geometric means (95% confidence interval)
Village groupBedroomAnimal shed
10.6 (0.5–0.8)0.8 (0.7–1.0)1.3 (1.0–1.6)4.5 (3.7–5.3)
20.7 (0.5–1.0)1.2 (0.9–1.5)1.5 (1.1–2.0)7.4 (6.0–9.1)
30.6 (0.4–0.7)0.6 (0.5–0.7)1.5 (1.2–1.9)4.3 (3.6–5.2)
41.3 (0.9–1.6)1.4 (1.0–1.8)3.5 (2.7–4.6)6.7 (5.0–8.9)


Repeated cross-sectional surveys and case–control methods based on passive surveillance demonstrated that bednets were less protective against P. vivax than against P. falciparum. An effect against P. vivax infection was not evident in every cross-sectional survey, and may take a few seasons to become apparent probably because of relapses from earlier infections masking the effect. An effect of nets against P. falciparum infection was more immediately apparent. UTN conferred some protection against malaria, and this conforms with recent evidence from Africa (Abdulla et al. 2001; Armstrong-Schellenberg et al. 2001; Clarke et al. 2001).

Cross-sectional surveys were a more expensive method of monitoring. Passive surveillance is more efficient because the only additional task is to ask questions about net usage in addition to the routine clinical history. Thus, passive surveillance would be preferable if the two methods were able to give similar results.

There were two important differences between the two study designs. Whereas cross-sectional surveys directly assessed a representative sample of the target population, passive surveillance monitored a more restricted source population, i.e. one that had fever and which chose to seek treatment at one of the clinics involved in the study. Secondly, whereas passive surveillance measured symptomatic parasitaemia, cross-sectional surveys measured both symptomatic and asymptomatic parasitaemia. Despite this, the fact that the IPE of ITN derived from the final survey and from the passive surveillance case–control study were similar (59% and 69%, respectively) seems to indicate that the clinic attender study population was representative of the target population.

If there were any reason why net users should be more or less prone to episodes of fever, not related to malaria, than non-users, selection bias would be introduced. It is conceivable that a family bednet used by several people would provide a better environment for the spread of infectious agents such as respiratory viruses. If this were so, net users might experience more episodes of fever unrelated to malaria, and this would lead to an overestimation of the protective effect of bednets. To our knowledge this has never been studied.

The study population was divided into cases and controls purely on the basis of presence or absence of malaria parasites. Any false positives or false negatives from the tests would bias the interpretation of the protective effect of bednets. If people were pre-treated with chloroquine before attending the clinic for diagnosis, some cases would be misclassified as controls. Selection bias would be introduced if the rate of pre-treatment with chloroquine was different between net users and non-users. Webster et al. (unpublished) showed that people from Behsud who used a bednet were less likely to pre-treat with chloroquine before attending clinic for malaria tests (OR = 0.26, P < 0.001). This pattern of behaviour underestimated the effectiveness of bednets.

The assessment of protective effect is liable to confounding if potential predictors of malaria risk such as economic status or access to education differ between net users and non-users. Socioeconomic status was not assessed, and this was a weakness in our study designs. Recent evidence shows that buyers of nets in Afghanistan are economically better off (Howard & Rowland, unpublished). Yet, a difference in wealth does not necessarily mean that net buyers and non-buyers will differ in their risk of contracting malaria. Parasite rate among buyers was similar to that among non-buyers at the time of buying. Yet, by the end of the year, prevalence was significantly lower among those who had been using ITN during the course of the year.

The use of ITN can, in regions where vectors are anthropophilic (e.g. East Africa), reduce survival of vectors which would be shown by a decreased sporozoite rate and a decreased vector density (Magesa et al. 1991). This should have a beneficial effect on non-users as well as on ITN users (Curtis et al. 1998). In Afghanistan, where the vectors are zoophilic, mosquito density did not appear to be reduced by ITN, and sporozoite rates were too low to show the possibility of an indirect protective effect on non-users (Rowland et al. 2002). Had there been such a reduction, the effectiveness of bednets would have been underestimated using the case–control methodology.

The local vector species are capable of transmitting both species of malaria parasites (Rowland et al. 2002). It may therefore be assumed that nets were equally protective against P. falciparum and P. vivax transmission. Because an effect against P. falciparum is immediate, the effect may be measured through either cross-sectional methods or passive data collection using a case–control design. But until it is possible to distinguish new infections from relapsed infections, neither the cross-sectional nor passive surveillance methods are adequate to show the real effect on P. vivax. The effect is only demonstrable as a trend over time.

As Behsud and Chaprahar are considered to be typical of rural Afghanistan, the study indicates that social marketing of ITN is a viable malaria control strategy for the country. Under the present uncertain conditions, it is more appropriate to help people to protect themselves than to return to the pre-war, centralized programmes of DDT spraying that are too easily disrupted by crisis. For families that are unwilling or unable to buy nets, an alternative strategy must be found. Achieving high coverage seems more a question of affordability than of desirability. In a recent study, 90% coverage was achieved when the price was set at US$1 per net (T. Freeman, unpublished). This suggests that a dual marketing strategy may offer the best prospect: After the first wave of sales has swept over rural districts, it should be possible to target the most endemic foci with highly subsidized nets. Alternatively, the second phase might target the most at risk by offering a highly subsidized net to anyone who is diagnosed with malaria at a public clinic. After the defeat of the Taliban and the promise of the West to help rebuild Afghanistan, what better use of aid than to provide the inhabitants with a form of protection against malaria they so clearly want. With the rural population at risk amounting to just a few million (Anon 1998), and with double nets costing about US$4 each (even less with cost-recovery), the aim seems eminently achievable with the resources promised.


We would like to express our appreciation of microscopy trainers Fazle Rahim, Aminullah and Sakhi Jan, entomologists Hamid Rahman, Mushtaq Faran and M. Kamal, and logistician Noorullah. Clive Davies kindly advised us on the analysis of entomological data. HealthNet International's malaria control and research programme is supported by the European Commission (DG1), the United Nations High Commissioner for Refugees, and WHO/UNDP/World Bank Special Programme for Research and Training in Tropical Diseases (project no. 960662). M.R. and D.C. are supported by the UK Department for International Development and the Gates Foundation. However, none of these donors can accept responsibility for any information provided or views expressed.