Policy Research f 9c n i °RA WQRKING PAPERS Population, Health, and Nutritlon Population and Human Resources Department The World Bank April 1992 WPS 883 Malaria The Impact of Treated Bed-Nets on Childhood Mortality in the Gambia Pedro L. Alonso, Allan G. Hill, Patricia H. David, Greg Fegan, Joanna R. M. Armstrong, Andreas Francisco, K. Cham, and Brian M. Greenwood In rural Gambia, as in many parts of Sub-Saharan Africa, malaria remains a major cause of death for children below the age of five - indeed, the principal cause of death when vaccination cover- age rates are high and death rates from common infectious diseases of childhood are reduced. In recent years, concern has grown about thedevelopmentof drug-resistant strains of malaria - provoking renewed interest in vector control and the reduc- tion of man-vector transmission rates. Policy Research WodsingPapers disseminate the findings of work in progress and encourage the exchangeof ideas amongBank staffand all others interested in developmentissues. These papers, distibuted by theResearchAdvisoqy Staff,carry thenamesoftheauthors.reflect onlythcirviews,and should beused and cited accordingly.The rundings, interpretations, andconclusions aretheauthors'own.Theyshould not be attributed to the World Bank, its Board of Directors, its management, or any of its member counties. Policy Research Population, Health, and Nutritiorn WPS 883 This paper - a product of the Population, Health, and Nutrition Division, Population and Human Resources Departient - emanated from the Bank's support of Johns Hopkins University for a child survival workshop focused on PHR's Health Sector Priorities Review. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Otilia Nadora, room S6-065, extension 31091 (April 1992, 30 pages). The effectiveness of insecticide-treated materials death. This "frailty protection" effect was had been unclear, as earlier studies had based substantial but is largely unexplained - more their results on the effects on vectors rather than basic research is needed. on human morbidity and mortality rates from malaria. So in 1988 the UK Medical Research Also, not all children have to be sleeping in Council began a systematic trial of a combined bed-nets for the benefits of the treatment to be intervention for controlling malaria around the felt. Small rates of noncompliance need not small town of Farafenni, in central Gambia. Two invalidate the effectiveness of the intervention. interventions - bed-nets treated with Permethrin and chemoprophylaxis with The nets were dipped by village women, Maloprim (dapsone = pyrimethamine) - were supervised by the village health worker and the conducted in "primary-health-care" villages, traditional birth attendant, with the support of the with non-PHC villages serving as controls, women's association. It appears that the washing and dipping process can be undertaken success- The study showed that general and malaria- fully by local people with a minimum of supervi- specific mortality in young children was sharply sion, at a cost for the so'ution of a few US cents reduced by introducing Permethrin-treated bed- per net dipped. nets. The effects of using treated bed-nets were clear, because many children had been sleeping The extra reduction in mortality attributable in bed-nets before the intervention began without to the use of Maloprim as a prophylactic was the same strong effects. probably slight and difficult to detect in this study because of the strong effect of sleeping The treated bed-net intervention had the under a treated bed-net. additional effect of reducing other causes of The Policy Research Working Paper Seriesdisseminates thc findings of work under way in theBank. Anobjective of the series is to get these findings out quickly, even if presentations are less than fully polished. The findings, interpretations, and conclusions in these papers do not necessarily represent official Bank policy. Produced by the Policy Research Dissemination Center Malaria: The Impact of Treated Bed-Nets on Childhood Mortality in the Gambia by Pedro L. Alonso, Allan G. Hill, Patricia H. David, Greg Fegan, Joana R. M. Armstrong, Andreas Francisco, K. Cham, and Brian M. Greenwood Table of Contents The study 2 The interventions in the PHC villages 5 Mortality surveillance 5 Cause of death ascertainment 6 Results 7 Childhood mortality 7 The Impact of the bed-nets 10 General mortality 10 Cause-specific mortality 12 Theoretical explanation 14 The demographic impact of the bed-nets 15 Conclusion 17 Tables 20 Figures 26 MALARIA: THE IMPACT OF TREATED BED-NETS ON CHILDHOOD MORTALITY IN TRE GAMBIA Pedro L. Alonso, Allan G. Hill, Patricia H. David, Greg Fegan, Joanna R.M. Armstrong, Andreas Francisco, Kabir Cham, Brian M. Greenwood In rural Gambia, as in many other parts of sub-Saharan Africa, malaria remains one of the most important causes of death in children under five. 1 When vaccination coverage rates are high and deaths due to the common infectious diseases of childhood are reduced, malaria may be the principal cause of death. 2 Malaria is particularly important to control not only because of its direct effect on mortality and morbidity but also because it acts in conjunction with other diseases to increase both morbidity and mortality rates from other causes. 3 THE STUDY In recent years, there has been growing concern about the development of drug-resistant strains of malaria, provoking renewed interest in vector control and the reduction of man- vector transmission rates. b The efficacy of insecticide- X BM Greenwood, AK Bradley AM Greenwood et al. Mortality and morbidity from malaria in a rural area of The Gambia, Trans R Soc Trop Ned Xyg 1987 81:478-86. U and A Brinkmann, Malaria and health in Africa: present situation and epidemiological trends, Trop Med & Parasitol 1991, 42:204-13. 2 BM Greenwood, AM Greenwood, AK Bradley et al. Deaths in infancy and early childhood in a well-vaccinated, rural West African population. Ann Trop Paodiat 1987 7:91-9. 3 DCW Mabey, A Brown, BM Greenwood. Plasuiodium falciparum malaria and Salmonella infections in Gambian children. J Infect Dis 1987 46:181-202. 4 JA Rozendaal. Impregnated mosquito nets and curtains for self-protection and vector control. Trop Dis Bull 1989 86:Rl- R41; CF Curtis, JD Lines, P Carnevale et al. Impregnated bed nets and curtains against malaria mosquitoes. In: Appropriate methods 2 treated materials has been unclear since previous studies have based their results on the effects on the vectors rather than on the malaria morbidity and mortality rates in the human population. In 1988, therefore, the UK Medical Research Council began a systematic trial of a combined intervention for the control of malaria in central Gambia around the small town of Farafenni. On the North Bank, some earlier work had suggested that chemoprophylaxis with Maloprim (dapsone + pyrimethamine) and Permethrin-treated bed-nets had reduced malaria morbidity but the intervention was too small scale for much to be said about mortality. 5 For the new double-intervention trial with permethrin-treated bed-nets and Maloprim used in combination, a new site was chosen south the Gambia river and east of Farafenni town. The area included 73 villages in all, of which 17 of the larger villages, generally those with 400 or more people, had joined the government's Primary Health Care scheme. The 56 smaller villages are here referred to collectively as 'non-PHC' villages. Both the interventions, the treated bed-nets plus the chemoprophylaxis, were conducted in the PHC villages with the non-PHC villages serving as controls (see Table 1). of vector control, ed. CF Curk. i, CRC Press, Florida, 1990:5-46; WHO. The use of impregnated bednets and other materials for vector-borne disease control. (WHO/VBC 89.981). WHO, Geneva, 1987. For a recent general review, see SC Oaks et al, 1991, Malaria: obstacles and opportunities, Inatitute of Medicine, US National Academy Press. 5 RW Snow, SW Lindsay, RJ Hayes, BM Greenwood. Permethrin- treated bed-nets (mosquito nets) prevent malaria in Gambian children. Trans R Soc Trop Ned Hyg 1988: 82:838-42. 3 Table 1: Summary of activities in the PHC and non-PHr villages Period PHC treatment villages Non-PHC (17 villages: children= 2616) controls (56 villages: children= 1795) PRE-INTERVENTION PERIOD June 1988 to Census followed by routine collection of data on July 1989 births, deaths and migration with verbal autopsy questionnaires for all children dying under age 6 INTERVENTION PERIOD July 1989 to Treated bed- Treated bed- Mortality and June 1990 nets plus 1/4 nets and morbidity strength placebo for the surveillance Maloprim for other half of only half of all all children 6 children 6 months to 5 months to 5 years _ _ _ _ _ _ _ _ _ _ _ _ _ _ y e a r s_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ POST-INTERVENTION PERIOD May-June 1990 New census with up-dating of all records of births and deaths plus a single round retrospective demographic survey with a full birth history for all women of reproductive age. Verbal autopsy analysis. The total population of all the villages averaged 21,157 individuals, comprising mostly Mandinka-speakers with some Fula as well as other smaller ethnic groups. In this area of flat Sudan savanna, with mangrove swamps at the river's edge, the rainfall is close to 1000 mm. per year: in 1988, it was 1051 mm and in 1989 it was 887 mm. The rainy season from June to October is the most important season for the transmission of malaria by the AnoRheles gambiae complex and most malaria deaths are concentrated in September, October and November of each year. In this paper, only the effects of interventions on the mortality of young children are discussed. Full details of the conduct of the intervention, the dipping of the bed-nets and the 4 chemoprophylaxis is available elsewhere. 8 The effects of the bed-nets on the incidence of clinical malaria and the prevalence of malaria infection will be reported separately. The interventions in the PHC villages In July 1989, all bed-nets in the PHC villages were first washed and then dipped in a Permethrin solution designed to produce a dose of about 500mg per m2 of net. Altogether, 5380 bed-nets were treated in this way, 88% of all nets in use in the villages. In all, 92% of all children in the PHC villages slept in treated bed-nets during the intervention trial. Before the intervention began, it should be noted, 96% of children in the P}IC villages had been sleeping in untreated bed-nets. Even in the non-PHC villages, 77% of children had been sleeping in untreated bed-nets. Then, all children in the PHC villages in the target age group, 6 months to five years, were randomized into two groups for receipt of either 1/4 strength Maloprim (25mg dapsone + 3.13 mg pyrimethamine) or a placebo of similar colour and taste. Chemoprophylaxis was continued for 20 weeks, July to November 1989, the main period of malaria transmission. Detailed compliance records were kept and random urine tests were conducted for the presence of dapsone. Compliance in the treatment and placebo groups was the same (95%) and more than 75% of the children in the treatment and control groups took their tablets on 90% of the required occasions. Mortality surveillance Once the study area had been identified, a full census of the 73 villages was carried out, listing all usual residents of 6 p Alonso, SW Lindsay, JKM Armstrong, M Conteh, AG Hill, PH David, G Fegan, A de Francisco, AJ Hall FC Shenton, K Cham, BM Greenwood. Reduction of mortality in Gambian children by insecticide-treated bed-nets. Lancet 337: 1499-1502, 22 June 1991. 7 Alonso et al, forthcoming. 5 each compound and family by name, age and sex. From 1 July 1988, village reporters recorded all births, deaths and permanent migration movements occurring in their villages. This information was collected weekly and checked by MRC field assistants, men and women with secondary school education from the area or nearby. At the MRC field station near Farafenni, these data were regularly entered into a computer so that up- dated census files could be maintained and checked on a continuous basis. In addition, a single-round retrospective survey of the lifetime fertility and childhood mortality experience of all women of reproductive age in the study area was carried out early in 1990 for two reasons. First, information on the longer-term trends in childhood mortality were needed to be sure that childhood mortality had not been on diverging paths in the pre- intervention period in the PHC and non-PHC villages. There were some concerns that mortality might have been changing at a different rate in the larger PHC villages compared with the non- -HC villages. In addition, despite the care and attention which goes into the routine recording of vital events in the surveillance area, there are always a few events which slip through the net. Some of these can be picked up during the annual dry season re-enumeration of the population. Accordingly, a household form including the basic Brass questions on children ever-borne alive and surviving to interview was designed, as well as a simple birth history for all women of reproductive age, regardless of marital status. The childhood deaths from the birth histories were then matched with the MRC records to ensure completeness of coverage. Only a small number of discrepancies were found, largely due to errors in the retrospective dating of vital events in the single-round survey. Cause of death ascertainment All the parents or guardians of children who died in the study period (1 July 1988 until 30 June 1990) in both PHC and 6 non-PHC villages were interviewed by a senior field assistailt using a special questionnaire developed initially on the North Bank and at Basse for the ascertainment of the likely cause of death. This form was then examined independently by three physicians familiar with the area and with the symptoms of tne main causes of death in the area. Each then decided on the principal cause of death using whatever information wLs on the form or on attachments to it. These attachments often included the child health card, sometimes a record of treatment for a prior illness, and occasionally a report from a clinic or a hospital if the child had been taken there before death. In 43.4% of the 318 valid cases, all three physicians agreed on the nain cause of death for the children of all ages who had died during the period of the intervention. In 93.4% of all valid cases, two of the physicians were in agreement. On four main causes, acute respiratory infections, diarrhoea/malnutrition, acute gastro-enteritis and tetanus, the agreeiaent between the three physicians was over 50 per cent. For malaria, they agreed in 44% of cases. It turned out that the most difficult cases were neonates or children under six months; in almost a third of the cases of deaths under six months, a final cause of death could not be determined. Recognizing these difficulties and indeed the problems of post-mortem cause of death ascertainment without a hospital autopsy, the three physicians then conferred with all the evidence before them. Overall, a final diagnosis was agreed upon for 73% of the cases, 66% for children who died at 6 months of age or older. It is this final diagnosis agreed between the three physicians which is used here. RESULTS Childhood mortality The routinely collected surveillance data can be used quite readily to estimate childhood mortality for the two years, before and after the interventions. The only difficult feature is the estimation of the denominators, the children at risk, since there 7 is considerable movement in the area and it is impossih'3e to keep track of all short-term migratioai. In view of the seasonal nature of many ..f these movements, it was decided to astinate the mid-period population by taking the average number of children present in the area. The end of period populations had been updated to take account of losses or omissions picked up in the annual re-enumeration of the population. These results from period life table calculations are summarized in Table 2. Table 2: Childhood mortality for PHC and non-PHC villages south of Farafenni before and after the introduction of bed- nets Primary Health Care Non-Primary Health care villages villages IMRT 4q1 5qI IMR 4q | sqo 1988-89 0.110 0.157 0.249 0.093 0.100 0.184 Deaths 59 75 134 35 32 67 1989-90 0.072 0.045 0.114 0.116 0.103 0.207 Deaths 41 25 66 43 35 78 Note: The separation factors for the calculation of the probabilities of surviving or dying were taken from the data, since in almost all cages, dates of birth an," death were known in the form of day, month and year. IMR = infant mortality rate. The data in Table 2 have several interesting features. One is the level of childhood mortality indicated. Although we have come to expect values of around 100 per thousand for infant mortality, it is a surprise to see that this value is still the figure in The Gambia where there has been a good Primary Health Care system since 1983 and where immunization coverage is remarkably high. Each year from 1985 onwards, The Gambian government has carried out an EPI survey around December-January with the results shown in Table 3. The figures for the whole country in 1989 and 1990 are very impressive, especially so since 8 the coverage in the Central ar.d Eastern region3 was es high as in the Western region around Banjul. The mortality figures have to be interpreted in the context of this high immunization coverage, sincea some common causes of childhood mortality in West Africa, such as measles and tetanus, are almost absent from the cause of death data from Farafenni. A second interesting feature of the mortality data by age is the high rates for the age group 1-4 relative to the infant mortality rates. It was once thought that one explanation for the high later childhood mortality in West Atrica might be measles but this is clearly not the case since measles mortality has been drastically reduced by tht. immunization programme. Another factor must be responsible for the high ratio of the of mortality of 1-4 year-olds to the mortality of infants. These dramatic reductions in mortality of the 1-4 age group were not matched in the non-PHC villages for the same period. Some fall in infant mortality was recorded in tha PHC villages. This seems surprising but there is good evidence that more infant deaths than previously thought are related to malaria in this area. In addition, there appear to have been some complex interactions between malaria and other illness which complicate the identification of the effect of the bed-net intervention. Table 3: National immunization coverage rates estimated from the annual EPI coverage surveys (percentages) Antigen [ 1985 1986 J 1987 1989 1990 BCG 92 97 90 98 98 DPT (3) 68 72 77 83 87 Polio (3) 62 69 83 89 92 Measles 75 80 82 81 85 Yellow fever 70 73 81 87 85 Fully 52 ', 61 70 78 immuniz -d 9 To look further back in time at trends in childhood mortality in the two different sets of villages before the intervention began, mortality rates were calculated from the birth histories obtained from women of reproductive age in January-February 1990. In addition, it was possible to go back to the data tapes frxa the 1983 national population census and to produce tables of the proportions dead of children ever-borne alive by age of mother for exactly the same villages included in the intervention trial. The data from the birth histories are more reliable since it was possible to re-check some results with reference to the MRC surveillance data. The level and trends in under 5 mortality are shown on Figure 1. Much the same results are obtained from the indirect estimates from the 1983 census but omission of children ever-born as well as dead children by comparison with the surveillance data make these results difficult to interpret. The data on Figure 1 strongly suggest that whilst there had beer a slow improvement in child mortality before the intervention, the mortality in the two sets of villages chosen for the intervention was not very different despite the different sizes and different participation in the national primary health care system by the two types of villages. A clear demonstration of the role of malaria in the mortality decline can be seen in the graph of deaths by month for PHC and non-PHC villages (see footnote 6). THE IMPACT OF THE BED-NETS General mortality The data in Table 4 show that the probability of dying for 1-4 years olds fell sharply in 1989-90 to 29% of the 1988-89 level in the PHC villages. Mortality in this age range was unchanged in the non-PHC villages. These results can be expressed in a more detailed way using the rate ratios of the age-specific death rates from the life tables (Table 4). 10 Table 4: Age-specific death rates and rate ratios before and after the intervention. _______________ PRE-INTERVENTION Age PHC villages Non-PHC Ratios villages o 115.9 99.1 1.169 (59/509) (35/353) 1-4 44.5 27.7 1.609 (75/1685) (32/1156.5) POST-INTERVENTION 0 75.2 135.2 0.556 _ (41/545-5) (43/318) 1-4 11.8 28.4 f.416 (21/1775) (35/1232) The effect of the bed-nets on the mortality of the 1-4 year olds is particularly clear although the decline in infant mortality is also significant. There is no obvious explanation for the higher initial lpvels of 1-4 year old morta7ity in the PHC compared to the non-PHC villages in the pre-intervention period. There is a suggestion in the cause of death data from the two sets of villages that malaria and severe diarrhoea/malnutrition with gastro-enteritis may be more important causes in the larger PHC villages. Whatever the reasons for the initial differences, the bed-net intervention has reversed this relationship between the mortality in the two sets of villages. Although some of the numbers are very small, examination of the rate ratios for single years of age reveals that the decline in mortality in the PHC villages was especially marked for one year old children. In the PHC villages, deaths in this age category fell from 35 to 8 over the two years; for the non-?HC villages, the number of deaths of one year olds rose from 14 to 18. The results from the morbidity surveillance, including measurement of body temperatures and blood slides for those with fever, and the studies of the mosquito populations, will show that these mortality result,s are directly associated with a 11 proportionE7.e reduction in cases of mild and severe clini-al malaria and in overall transmission rates (Alonso et al, forthcoming). Cause-specific mortality Although the intervention was directed at children aged 6 months to 6 years, this section will examine the mortality of all children under 6. We have retained the data on the younger children both to keep the numbers as large as possible and because we have good evidence that the intervention had a substantial and surprising impact on childhood mortality in general (Table 4). In addition, by working with all the under fives we can more easily make comparisons with life table measures of childhood mortality from other populations. The main causes of death for all children dying before and after the intervention are shown in Figure 2. The 'don't know' category includes both those deaths for which the three physicians could not agree and those deaths where no single cause could be identified by any uf the physicians reading the post- mortem questionnaires. Before the intervention, in both the PHC and the non-PHC villages, malaria was thought to be the main cause of death in 28% of all deaths. If we exclude the deaths of those under 6 months, malaria deaths constituted 39% of all deaths in the pre-intervention year taking both PHC and non-PHC villages together. Turning to changes in cause-specific mortality by treatment group, we see from Figure 3 the relatively small changes recorded in the non-PHC villages. There were some changes in the numbers dying of acute respiratory infections, meningitis, acute gastro- enteritis and septicaemia but the most notable feature of Figure 3 is the relatively small change in the number of malaria deaths. Figure 4 presents the mortality data on cause of death for the PHC villages. The sudden fall in deaths due to malaria is obvious. The reduction in the numbers of malaria deaths in the PHC villages was not made up by increases in other categories. 12 On the contrary, it seems that the bed-nets intervention had an effect on some other significant causes of death, deaths attributed to acute respiratory infections in particular. There are problems in looking at these changes in specific causes both because of the uncertainty which surrounds the post-mortem cruse of death assignment and because of small numbers. In an attempt to examine some of these changes in a relatively simple way, the ratio of the numbers of deaths in the pre- and the post- intervention periods have been calculated for PHC and non-PHC villages separately. Overall, the ratio of all deaths of children under age 5 in PHC and non-PHC villages was 1.77 (158/89) largely because of differences in the size of the two study populations. Figure 5 shows these ratios as well as the average for all the causes combined. Taking into account the small numbers in some categories (see Appendix 1 for the basic data), there are some remarkable reductions in the ratios for the acute respiratory infections and the diarrhoea/malnutrition categories between the pre- and post-intervention periods. The change in deaths due to the acute respiratory infections is especially significant because of the number of deaths due to this czuse (in the pre-intervention period, 38 under fives died of this cause in the PHC and non-PHC villages combined). The effects of the Maloprim administered to half the children in the PHC villages were swamped by the more powerful effects of the treated bed-nets. There were only 11 deaths, 2 attributable to malaria, in the Maloprim group and 10 in the placebo group, 6 attributable to malaria. Although not detectable in the mortality data, the data on morbidity shows that even for those children not sleeping in treated bed-nets, there were reductions in the frequency of fevers and of positive blood slides. This observation points to the possibility of a more general community effect which may be very important when compliance rates for the intervention are lower than in this study. The effects on the mosquitoes are thought to operate both through the insecticide and through the 13 emulsion in which the insecticide is dissolved. Different emulsions may have different effects in deterring mosquitoes from even entering the sleeping areas. Work continues on this problem and on the development of waterproof emulsions which can resist washing. It should be noted that in both the PHC and the non-PHC villages, weekly blood slides were made for all children with a temperature of over 37.50. Those rith parasitaemia were referred to the Village Health Worker for treatment. Field workers were asked to follow up on these children and to arrange further treatment outside the village if necessary. Theoretical explanation The results indicate a strong direct effect of the treated bed-nets in an area where the use of bed-nets was well established before the intervention began. If such high levels of compliance can be achieved nationally, major reductions in both malaria-specific and overall childhood morta'lity can be anticipated. The effects described here are clearly much larger than can be anticipated in the national programme but the trial has brought to light some new aspects of the disease. One is the proportional reduction in the numbers of infected bites, in mild and severe cases of malaria as well as in malaria-related deaths. 8 Some mild cases still became savere cases but the additional effect of the Maloprim was to reduce the numbers of severe cases as well as deaths. The bed-nets and chemo-prophylaxis therefore have different effects on the demography of malaria since Maloprim is effective on just one part of the continuum from infected bite to malaria death whereas the effects of bed-nets were seen at every level in this transition. It seems that the reduction in malaria mortality is due to reduction in biting 8 See BM Greenwood, K Marsh and RW Snow, Why do some African children die of severe malaria? Parasitol Today 17(10): 277-81. 14 rates by mosquitoes and is not due to the reduction in other insect-borne diseases such as sleeping sickness and kala azar which in The Gambia are now quite rare. There is a possibility that elimination of some flies might have reduced enteric infections but the reduction in acute respiratory infections is greater than the reductions in the diarrhoeas and the severe gastro-enteritic infections. The 'frailty protection' effect resulting from t:he avoidance of malarial infections is more difficult to explain. As with measles infections, there may be considerable and lasting gains to be had from the avoidance of infection by the malarial parasite and avoidance of the consequences of this infection such as anaemia. The demographic impact of the bed-nets There are a variety of ways of estimating what might be referred to as the 'gross' and 'net' effects of the bed-net intervention on childhood mortality. Here we compare the mortality outcomes we might anticipate by re-analyzing the life tables constructed from the surveillance data using the new information on cause of death. Three different comparisons are made. One is to compare the mortality before the intervention with the mortality which would ensue if all the deaths whose main cause was judged to be malaria were eliminated. A second comparison is between this mortality in the absence of malaria with the observed mortality rates in the post-intervention period in the PHC villages. Finally, we can check our calculations by comparing these results with the data from the non-PHC villages. Table 5 contains the basic data for these comparisons. Using the data from the PHC villages first, we can see that the elimination of all malaria deaths would reduce infant mortality to 83% of its pre-intervention levels but the reduction in the mortality of 1-4 year-olds would be much greater: 56% of pre- intervention levels. Moving to the last line of the table to consider the same comparison for the non-PHC villages, similar 15 figures are obtained. Overall, the elimination of all deaths believed to be directly due to malaria would reduce the probability of dying before age 5 to about 70% of pre- intervention levels. Table 5: Comparison of changes in childhood mortality following the elimination of malaria deaths and the use of bed- nets PHC VILLAGES IMR 4qq sqo With malaria 109.6 156.7 249.1 No malaria 91.0 88.4 171.3 Bed-nets 72.4 44.6 113.8 Ratios No malaria/malaria 0.83 0.56 0.69 Bed-nets/malaria 0.66 0.28 0.46 NON-PHC VILLAGES 1988-9 with malaria 93.0 100.3 184.0 1989-90 with malaria 116.1 102.9 207.0 Mean 1988-90 104.5 101.6 195.5 No malaria (2 yr average) 102.2 66.0 161.4 Ratios No malaria/malaria 1988-90 0.98 0.65 0.74 Much more exciting is the comparison between the pre- intervention levels and the actual mortality outcomes as measured in the post-intervention year. As we see from the middle section of Table 5, the mortality of the 1-4 year olds was cut to 28% of pre-intervention levels and the probability of dying before age 5 fell to 46% of its pre-intervention level. Quite clearly, the 16 prevention of malaria has had a powerful synergistic sffect on some other causes of death such as acute respiratory infections. In The Gambia, the case for this effect is stronger than it might be elsewhere in Africa because some possibly confounding causes of deaths (measles in particular) have already been controlled by the time of the intervention. We refer to the effective immunization programme which makes it unlikely that the reductions in childhood mortality can be simply explained by the temporary absence of some other epidemic causes of death such as whooping cough or measles. Certainly, no exceptional outbreaks of other causes of death such as meningitis were recorded during the intervention trial. Conclusion This study ha, shown that general and malaria-specific mortality in children under 5 have both been sharply reduced by the introduction of Permethrin-treated bed-nets. The extra reduction in mortality provided by the use of Maloprim taken as a prophylactic was probably slight and difficult to detect in this study because of the very strong effect of sleeping in a treated bed-net (see footnote 6). Work not discussed in detail here shows that both the mosquito population and malaria morbidity were both reduced by the intervention. The effects of the treatment of the bed-nets with Permethrin reported here are clearly due to the treatment and not just to the use of bed-nets since a large proportion of children had been sleeping in bed- nets before the intervention began. There are two additional findings of importance to those interested in malaria-eradication prograi,mes. One is the clear additional effect on other causes of death of the treated bed net intervention. This 'frailty protection' effect is substantial but largely unexplained. More basic research is this area is called for. 17 A second finding is that not all children have to be sleeping in bed-nets for the benefits of the treatment to be felt. If the community as a whole participates in the intervention, some small rates of non-compliance need not invalidate the effectiveness of the intervention. It must be :emembered that the dipping of the nets was carried out entirely by village women under the supervision of the Village Health Worker and the Traditional Birth Attendant with the support of the women's association. No major difficulties were encountered during this process although several women did wash their nets in the middle of the intervention, thereby diluting the effects of the insecticide. The experience in the Farafenni area leads us to believe that the washing and dipping process, including making up the solution with the correct concentration, can be undertaken successfully with the minimum of supervision in other contexts. The costs of the solution are a few US cents per net dipped. There are no signs that the intervention has produced any toxicity or other adverse reactions; the chemicals are difficult to detect two weeks after the dipping has taken place. 9 There are some unanswered questions about the implications of the loss of immunity whilst protected from mild infection when using a bed-net, and the effects of the bed-nets on adult mortality and morbidity have not so far been studied in detail. Mosquitos will soon develop some resistance to Permethrin but by altering the type of insecticide used every few years, it should be possible to keep ahead of these changes. Other behavioural adaptations of mosquitos denied easy access to human blood are difficult to predict. There are some indications that adults in the intervention villages received additional bites. If the 9 RW Snow, M Jawara, CF Curtis Bull Ent Res 1987 77:279-86; RW Snow, SW Lindsay, RJ Hayes, BM Greenwood Trans R Boo Trop Med 1988. 18 mosquitos were to move to animal hosts, the cycle of malaria transmission would be broken. For estimating the mortality effect of the use of treated bed-nets elsewhere, it may be worth remembering one or two distinctive characteristics of malaria in The Gambia. One is the highly seasonal nature of transmission; we have no evidence of the results will be the same in populations in which malaria is present all year round. Secondly, the AnoRheles gambiensis does not bite duriing the day whereas other species have different biting habits. Thirdly, the high level of immunization achieved in The Gambia may make the effects of malaria control on mortality clearer than elsewhere since the number of competing risks from other infectious diseases has been limited. Finally the form of living arrangements of the Mandinka, large families living in crowded compounds with wives, co-wives and children sharing a common sleeping room, may be responsible for some of the especially strong effects of the bed-nets on malaria and overall mortality. In rural Gambia, a 1991 survey revealed that 62% of beds in PHC villages and 51% of beds in non-PHC villages had bednets around them. 10 000 10 M Aikins and S Bennett, National bednet survey, MRC, The Gambia. 19 TABLE A: Main cause of death by age of death for the pre- intervention period: Primary Health Care (PHC) and non- Primary Health Care (non-PHC) villages combined. Cause Age at death (months) All of death 0 1-5 6-11|12-23 24-35 36-47 _48-59 |60+ ARI 5 8 8 10 1 3 1 1 37 Malaria 0 2 10 14 18 10 5 11 70 Mening. 1 2 2 2 3 1 0 1 12 Diar/mal 0 1 2 3 4 0 0 0 10 Gasterol 0 4 10 9 0 1 0 1 25 Septaec. 0 2 2 3 0 1 0 0 8 Tetanus 1 0 0 0 0 0 0 0 1 Miscell. 2 1 0 1 2 0 1 1 8 No agree 7 8 8 11 2 3 0 2 46 Unknown 19 5 1 0 1 0 1 1 30 Total 36 37 38 54 37 191 9 17 247J 20 TABLE B: Main cause of death by age of death in the post- intervention period: Primary Health Care (PHC) and non- Primary Health Care (non-PHC) villages combined. Cause of Age at death (months) All death 0 1-S 6 J11 12-23 24-35 36-47 48-59 60+ ARI 4 11 0 2 2 2 0 0 21 Malaria 0 2 2 5 4 5 2 7 27 Mening. 1 2 3 0 0 0 0 0 6 Diar/mal 0 0 0 1 1 0 0 1 3 Gastero' 0 1 1 1 1 0 0 0 4 Septaec. 0 2 1 0 0 0 0 0 3 Tetanus 0 0 0 0 0 0 0 0 0 Miscell. 3 0 0 0 0 0 0 0 3 No agree 7 8 8 11 2 3 0 2 41 Unknown 19 5 1 0 1 0[ 1 1 28 [~~ ~ l I _ _ |Total 34 311 16 201 11 10 3 11 136 21 TABLE C: Main cause of death by age of death for the pre- intervention period: Primary Health Care (PHC) villages only. Cause Age at death (months) All of death o j 1-5 6-11 12-23 24-35 [36-47 148-59 60+ ARI 3 4 7 7 1 2 1 1 26 Malaria 0 2 9 11 12 6 5 7 52 Mening. 0 1 1 0 3 0 0 1 6 Diar/mal 0 0 2 3 3 0 0 0 8 Gastero' 0 3 6 7 0 0 0 0 16 Septaec. 0 0 2 2 0 0 0 0 4 Tetanus 1 0 0 0 0 0 0 0 1 Miscell. 2 0 0 1 2 0 0 0 5 No agree 4 3 2 6 5 2 0 1 23 Unknown 9 4 0 1 1 0 1 1 17 Total |I 19 17 29 38 27[ 10I 7 111 158 22 TABLE 13: Main cause of death by age of death for the pre- intervention period: Non-Primary Health Care (NPHC) villages only. Cause Age at death (months) All of death O | 1-5 [6-11 12-23 24-35 36-47 48- _ - -. ARI 2 4 1 3 0 1 0 0 11 Malaria 0 0 1 3 6 4 0 4 18 Mening. 1 1 1 2 0 1 0 0 6 Diar/mal 0 1 0 0 1 0 0 0 2 Gasterot 0 1 4 2 0 1 0 1 9 Se-ptaec. 0 2 0 1 0 1 0 0 4 Tetanus 0 0 0 0 0 0 0 0 0 Miscell. 0 1 0 0 0 ( 1 1 3 No agree 2 4 2 8 2 2 0 1 21 Unknown 11 0 1 ° 0 0 1 1 14 Total | 18[ 15| 5 16 61 5| 3| 7|1 75 23 TABLE E: Main cause of death by age of death for the post- intervention period: Primary Health Care (PHC) villages only. Cause [ Age at death (months) All of death 0 1-5 6-11 12-23 24-35 36-47 148-59 60+ ARI 2 5 0 1 1 0 0 0 9 Malaria 0 1 2 0 1 4 0 2 10 Mening. 1 1 1 0 0 0 0 0 3 Diar/mal 0 0 0 0 1 0 0 1 2 Gastero' 0 0 1 0 1 0 0 0 2 Septaec. 0 0 1 0 0 0 0 0 1 Tetanus 0 0 0 0 01 0 0 0 0 Miscell. 0 0 0 0 Qj 0 0 0 0 No agree 5 4 6 3 0 1| 0 1 20 Unknown 8 5 0 0 1 0 0 0 14 Total 171 201 9 16 101 9J 2] 6T 61 24 TABLE F: Main cause of death by age of death for the post- intervention period: Non-Primary Health Care (NPHC) villages only. Cause Age at death (months) |All of death_ of death 0 1-5 6-11 12-23 24-35 36-47 48-59 60+ ARI 2 4 1 3 0 1 0 0 11 Malaria 0 0 1 3 6 4 0 4 18 Mening. 1 1 1 2 0 1 0 0 6 Diar/mal 0 1 0 0 1 0 0 0 2 Gasterol 0 1 4 2 0 1 0 1 9 Septaec. 0 2 0 1 0 1 0 0 4 Tetanus 0 0 0 0 0 0 0 0 0 Miscell. 0 1 0 0 0 0 1 1 3 No agree 5 7 1 5 3 1 1 0 23 Unknown 9 3 1 0 0 0 0 0 13 Total 18 15 5 16 6 5 3 7 November 1991 25 Under 5 mortality 5qO 0.5- 0.4 0.2 0..1 - 0 1977 1982 1987 Date * non-PHC villages "-. PHC villages Figure 1 Main causes of death for all under 5's before and after the use of bed-nets Proportion of total 0.6 0.5.-. 0.4 . 0.3.-.-... . . . 0.2- ARI Malaria Mening Diarrh Gastro Septae Tetanus Misc DK Main cause of death Key % before % after Figure 2 Numbers of deaths by main cause in non-PHC villages before and after the use of bed-nets Number of deaths 20_ 15 ..i. ..... 00 1 lot ARI Malaria Mening Diarrh Gastro Septae Misc DisagreeUnknown Main cause of death Key Non-PHC before Non-PHC after Figure 3 Numbers of deaths by main cause in PHC villages before and after using bed-nets Number of deaths 60 50 . . 30.. 20 . ARI Malaria Moning Dlarrh G3astro Septae Mlsc DisagreeUnknown 0~~ Main cause of death Key PHC before _ PHC after Figure 4 Ratios of numbers of deaths by cause in PHC and non-PHC villages before and after the use of bed-nets Ratio PHC/non-PHC 10 80 6 - 2 I ARI Malarila Mening Dlarrh Gastro Septae Misc DleagreeUnknown Main cause of death Key R PHC/NPHC before MR PHC/NPHC after A Standard ratio Figure 5 Policy Research Working Paper Series Contact Title Author Date for paper WPS871 The Distribution of the Benefits from Dominique van de Walle March 1992 A. Bhalla Social Services in Indonesia, 1978-87 37699 WPS872 Romania's Evolving Legal Framework Cheryl W. Gray March 1992 CECSE for Private Sector Development Rebecca J. Hanson 37188 Peter G. lanachkov WPS873 Measure and Interpretation of Francois M. Ettori March 1992 F. Ettori Effective Protection in the Presence 32340 of High Capital Costs: Evidence from India WPS874 The Trade Restrictiveness Index: An James E. And6,son March 1992 G. llogon Application to Mexican Agriculture Geoffrey Bannister 33732 WPS875 Rural Finance in Developing Jacob Yaron March 1992 C. Spooner Countries 30464 WPS876 Old Debts and New Beginnings: Ross Levine March 1992 W. Pitayatonakarn A Policy Choice in Transitional David Scott 37664 Socialist Economies WPS877 Assessing Gains in Efficient Gary H. Jefferson March 1992 CECSE Production Among China's Industrial Wenyi Xu 37188 Enterprises WPS878 Adjustment and Private Investment Kazi M. Matin March 1992 D. Ballantyne in Kenya Bernard Wasow 37947 WPS879 Comprehensive Water Resources Peter Rogers March 1992 D. Ranger Management: A Concept Paper 31296 WPS880 Exchange Rate Policy, the Real Miguel A. Kiguel April 1992 R. Luz Exchange Rate, and !nflation: Lessons 34303 from Latin America WPS881 Dual and Multiple Exchange Rate Nita Ghei April 1992 R. Luz Systems in Developing Countries: Miguel A. Kiguel 34303 Some Empirical Evidence WPS882 Issues in Reforming Financial Alfredo Thorne April 1992 L. Ly Systems in Eastern Europe: The Case 37352 of Bulgaria Policy Research Workilng Paper Series Contact Title Author Date for paper WPS883 Malaria: The Impact of Treated Bed- Pedro L. Alonso April 1992 0. Nadora Nets on Childhood Mortality in the Allan G. Hill 31091 Gambia Patricia H. David Greg Fegan Joanna R. M. Armstrong Andreas Francisco K. Cham Brian M. Greenwood WPS884 Intercommodity Price Transmittal: Harold Alderman April 1992 C. Spooner Analysis of Food Markets in Ghana 30464