New malaria control policies and child mortality in Senegal: reaching Millennium Development Goal 4

Abstract

Background

The Demographic Surveillance System established in 1962 in Niakhar, Senegal, is the oldest in Africa. Here we analyze trends in overall child mortality, malaria and other causes of death in Niakhar from the beginning of data collection up to 2010.

Methods

Following an initial census, demographic data have been updated yearly from 1963 to 2010. From 1984, causes of death were determined by the verbal autopsy technique.

Results

During the period 1963-2010, infant and under-5 mortality rates declined from 223‰ to 18‰ and from 485‰ to 41‰, respectively. The decrease was progressive during the whole observation period except during the years 1990 to 2000 when a plateau and then an increase was observed. Malaria attributable mortality in under-5 children dropped from 13.5‰ during the 1992-1999 period to 2.2‰ in 2010. During this period, all-cause mortality in under-5 children declined by 80%.

Interpretation

Inadequate treatment for chloroquine-resistant malaria and an epidemic of meningitis in the 1990s were the two factors that interrupted a continuous decrease in child mortality. Direct and indirect effects of new malaria control policies, introduced in 2003 and completed in 2006/2008, are likely to have been key cause of the recent dramatic decrease in childhood mortality.

Introduction

Child mortality in tropical Africa has fallen dramatically since the 1960s.¹ However, it remains the highest in the world for several reasons including the fragility of the health system in many countries of the continent.^2,3 Overall under-5 mortality in West Africa decreased by only 18% between 1990 and 2007. ⁴ This has led some commentators to suggest that in such countries, achieving the Millenium Development Goals may prove very difficult, especially the fourth Millenium Development Goal (MDG4) of cutting under-5 mortality by two thirds between 1990 and 2015.^{3, 5-7} One of the leading causes of death in African children is malaria, and mortality attributable to this disease increased at least 2-fold during the 1990s associated with the spread of chloroquine resistance.^8,9 The replacement of chloroquine by more effective drugs was initiated in some African countries between 2000 and 2003, but it was not until 2006 that artemisinin-based combination therapy (ACT) was deployed on a large scale in Africa.^10,11

The oldest Demographic Surveillance System (DSS) in Africa, started in 1962, is in Niakhar, Senegal, a rural area in the Sahel.^12-14 Following an initial census, data on births, deaths and migrations have been collected prospectively at least yearly from 1963 to the present. In 1984, the study area was extended and systematic recording of causes of deaths through verbal autopsy was introduced. In this paper, we describe trends in overall child mortality since the origin of the DSS and analyze the trends in causes of deaths since 1984. After a decade of stagnation in the 1990s, a new, dramatic decline in child mortality has been observed during the past few years. This decline was temporally related to the deployment of new malaria control policies, suggesting that malaria has both a direct and indirect effect on overall mortality, and that MDG4 can be achieved primarily through malaria control and immunization in poor rural areas of Africa.

Methods

Study area and population

The Niakhar study area is located in the district of Fatick, Senegal, 120 km southeast of the capital city of Dakar. The population covered by the DSS lives in 30 villages made up of several different hamlets and comprised 43,000 inhabitants on 1st July 2010. Most of the population (96.4%) belong to the Sereer ethnic group and live traditionally on one food crop (millet) and raise one cash crop (groundnuts) and a few cattle. The climate is typical of the sub-Sahel. Rains are concentrated during a four-month period (July to October), and annual rainfall averaged 506 mm for the period 1962-2010 (457 mm for the period 1984-2010). A detailed description of the Niakhar DSS is given elsewhere. ¹⁴

Healthcare system

There are three health posts within the study area (the first opened in 1953, the last in 1983), and two outside the area, which provide first-line health care services including curative care for common illnesses, immunization, prenatal care, delivery, and management of malnutrition. These health posts are staffed by nurses and midwives. The closest health centre is located at Fatick, 20-30 km from the study area. Surgery, Caesarean sections and blood transfusions are available only in the regional hospitals of Kaolack and Diourbel, 60-70 km from the study area.

Malaria in Niakhar is endemic with a seasonal peak of transmission from August to November.¹⁵ Until November 2003, chloroquine was the first line treatment of malaria in Niakhar as elsewhere in Senegal. The first cases of chloroquine resistance were detected in 1992 and subsequently the proportion of RII and RIII strains increased rapidly, reaching 10% in 1993, 15% in 1994 and 29% in 1996.⁸ A study conducted in a random sample of the population of the 30 villages of the study area in 1995 indicated that the prevalence of malaria in asymptomatic under-five children was 36% in February, 33% in June and 79% in November, with 95% of the infections being due to Plasmodium falciparum. ¹⁶ In November 2003, the Senegalese Ministry of Health introduced the combination amodiaquine + sulfadoxine/pyrimethamine (AQ+SP) as a replacement for chloroquine for the first line treatment of malaria in all health facilities in Senegal. In May 2006, artemisinin-based combination therapy with artesunate + amodiaquine (AS+AQ) was deployed in Senegal as a replacement for AQ+SP. ¹⁷ In June 2008 long-lasting insecticide-treated nets (ITNs) were distributed to all children and universal coverage with ITNs was achieved in the area in 2009.^17,18

Although there is still no resident physician, midwife, laboratory facilities or emergency transportation, the health care available to the study population has improved considerably between 1962 and 2010 due in most part to the activities of national programmes, such as Primary Health Care (PHC), the Expanded Programme of Immunization (EPI), and the Maternal and Infant Protection (MIP) programme. Specific interventions have been short-term and generally involved only a sub-sample of children in the study population. These included clinical trials of new vaccines against measles and whooping cough, ^19,20 conducted in 1989-1996 and trials of seasonal intermittent preventive treatment of malaria in children conducted in 2002, 2004 and 2006.²¹

The demographic surveillance system

In December 1962, a demographic survey of the 65 villages forming the administrative area of Niakhar (population size ≈ 33,000) was undertaken.^12,13 Annual surveys were carried out up to 1969 when surveillance was restricted to 8 villages (population size ≈ 4,300). In 1983, surveillance was extended to 30 villages (population size: 23,391 in 1984 and 43,000 in 2010), 19 of which were part of the initial study.¹⁴

Demographic events collected included pregnancies, births, deaths, marriages and migrations. From 1963 to 1987, the population census was updated on a yearly basis. From 1987 to February 1997, each compound was visited once a week as part of the follow-up of vaccine trials; demographic information was collected during these weekly visits and its accuracy was checked during the annual census. Since March 1997, demographic data have been collected every quarter or twice a year. Infant and child mortality rates for the 1963-1999 period have been published previously.¹⁴ For the period 1963-1983, mortality rates were based on data from only the eight villages under continuous surveillance and are presented for five-year periods resulting in a smoothing of the mortality trend. For the period 1984-2010, annual probabilities of dying, based on the 30 villages of the Niakhar area, have been computed. The death rate (_nM_x) between age x and age x + n was computed as the ratio of deaths to the person-time accrued over the same period. Death rates were transformed into probabilities of dying (_nq_x) between age x and x + n using the standard equation:

$${}_{\mathrm{n}}\mathrm{q}_{\mathrm{x}}=\left(n\mathrm{x}{}_{\mathrm{n}}\mathrm{M}_{\mathrm{x}}\right)∕[1+(n-{}_{n}a_x\left)x{}_{n}M_x\right]$$

were _na_x is the mean number of person-years lived in the interval (x, x+n) by those dying in the interval.

Since 1984, all deaths which occurred in the study population have been investigated using the verbal autopsy technique. The same post-mortem questionnaire was used from 1984 to 2010 and the causes of deaths were determined from the responses to these questionnaires and from any additional medical information available. Each questionnaire was reviewed by at least two physicians and the most likely cause of death was defined by consensus between at least two reviewers. Every death for which the cause was classified as ‘undetermined’, and a sample of deaths of each determined cause, were re-examined by two or three physicians in 2009 in order to check that similar criteria for attributing causes of deaths had been used during the whole study period. When there was a discrepancy, the cause of death was reattributed by consensus between at least two reviewers. Malaria deaths for the period 1984-1995 ⁸ and the main causes of deaths for the period 1989-2000 ²² have been published previously. The main criteria used for attributing causes of deaths are given elsewhere.²²

Results

All-cause mortality

During the period 1963-1972, infant and child mortality rates were very high, and almost half of all children died by the age of 5 years (Table 1). During the 1970s and the 1980s, both infant and child mortality decreased markedly, and the probability of dying before the fifth birthday fell to 179‰ in 1990, i.e. a 62% reduction within two decades (Fig. 1). In contrast, infant and child mortality remained almost unchanged during most of the 1990s and increased dramatically in 1998 and 1999, with a probability of dying before age 5 years reaching a maximum of 295‰ in 1998. Under 5 mortality was still high in 2000 (1q0 = 83‰, 5q0 = 198‰) but it has decreased rapidly in the following years. In 2010, the probability of dying before the age of 5 years was only 41‰, i.e. an 80% reduction compared to the average value for the period 1992-1999. Fig 2 shows that the characteristic seasonal peak in under 5 deaths which occurred each year from September to November, was no longer evident in 2008-2010.

Table 1.

Trends in overall mortality, Niakhar, 1963-2010. Neonatal mortality (deaths between 0 and 28 days per 1,000 living births) and probability of dying (deaths per 1,000 children per year) before age 1 (1q0), between ages 1 and 5 (4q1), between birth and age 5 (5q0), and between ages 5 and 10 (5q5).

Year	neonatal	1q0	4q1	5q0	5q5
1963-1967	-	223	340	485	60
1968-1972	-	214	342	480	53
1973-1977	-	182	291	421	53
1978-1982	-	167	242	369	52
1984	58	129	197	301	29
1985	46	128	259	354	36
1986	50	106	154	244	24
1987	56	121	173	273	30
1988	73	125	131	239	15
1989	45	84	111	186	16
1990	46	96	92	179	15
1991	30	76	132	198	17
1992	29	85	130	204	27
1993	39	86	139	214	27
1994	39	69	102	164	32
1995	36	79	123	193	30
1996	37	81	118	190	34
1997	23	62	93	150	25
1998	22	103	214	295	66
1999	25	88	203	273	51
2000	25	83	125	198	39
2001	27	67	91	151	20
2002	24	66	84	145	18
2003	25	65	71	132	11
2004	17	36	53	88	10
2005	16	45	79	121	22
2006	16	36	62	96	13
2007	7	31	47	76	8
2008	9	22	34	55	8
2009	7	13	17	30	2
2010	9	18	23	41	4

Figure 1.

Changes in infant (1q0) and under-5 (5q0) mortality in Niakhar area.

Figure 2.

Seasonal pattern of deaths among children under 5 years of age in 1995, 2000, 2005 and each year since 2008 (note the disappearance of the seasonal peak in 2008-2010).

Cause specific mortality rates

Between 1984 and 2010, a total of 1,521 deaths were attributed to malaria in children 0-9 years of age. Figure 3 shows the evolution of malaria mortality in these children. From 1984 to 1987, the average annual malaria mortality rate was 7.1 deaths per thousand children under 5 years and 1.1 per thousand in children 5-9 years old. From 1988 to 1991, the average annual malaria mortality rate was 5.4 per thousand per year in children under 5 years and 1.2 per thousand per year in children 5-9 years old. From 1992 to 1999, a large increase in the number of malaria deaths was observed, reaching 19.7 per thousand per year in children under 5 years in 1998 and averaging 13.5 per thousand per year in children under 5 years for the whole period (12.1 per thousand per year in children under 5 years and 2.7 per thousand per year in children 5-9 years for the period 1992-1997). Malaria mortality has decreased markedly from 2000 onwards. It was 10.5, 7.6, 6.6 and 2.0 per thousand per year in under 5 year old children during the periods 2000-2003, 2004-2005, 2006-2007 and 2008-2010 respectively. In 2010, it was only 2.2 per thousand in children under 5 years of age and 0.3 per thousand in children aged 5-9 years.

Figure 3.

Trends in mortality rates attributable to malaria and their relationship to the malaria control interventions. 1. Emergence of chloroquine resistance; 2: Sulfadoxine-pyrimethamine made available for second line treatment; 3: Amodiaquine + sulfadoxine-pyrimethamine introduced for first line treatment; 4: Artesunate + amodiaquine introduced for first line treatment; 5: Widespread deployment of impregnated bednets.

Figure 4 shows the evolution of mortality attributable to meningitis. Between 1984 and 2010, a total of 292 deaths were attributed to this disease in children aged 0-9 years. Most of these deaths (96 in children under 5 years of age and 80 in children 5-9 years) occurred during an outbreak between 1998 and 2000. In 1998 and 1999, at the peak of the epidemic, mortality due to meningitis averaged 8.2 per thousand per year in children under 5 years of age and 6.6 per thousand per year in children 5-9 years.

Figure 4.

Trend in mortality rates attributable to meningitis.

Mortality rates due to diarrhoeal diseases and acute respiratory infections (ARI) are shown in Table 2. Diarrhoeal diseases in children under 5 years of age decreased considerably during the study, from 14.8 per thousand per year during the period 1984-1991 to only 2.3 per thousand per year during the period 2009-2010. The decrease was also marked for deaths from ARI, which fell from 5.4 per thousand per year during the period 1984-1991 to 0.2 per thousand per year during the period 2009-2010.

Table 2.

Mortality rates (per 1,000 person-years) from diarrhoeal diseases and acute respiratory infections (ARI), Niakhar, 1984-2010.

Age (Years)	Period	Diarrhoea	ARI
0-4	1984-1991	14.8	5.4
	1992-1999	12.4	3.8
	2000-2003	8.5	3.3
	2004-2008	3.0	0.6
	2009-2010	2.3	0.2
5-9	1984-1991	0.9	0.2
	1992-1999	0.9	0.3
	2000-2003	0.2	0.1
	2004-2008	0.4	0.0
	2009-2010	0.1	0.0

Discussion

In most rural areas of West Africa, very high childhood mortality was still common in the 1960s, with about a half of all children dying before the age of 5 years.¹ Most of these deaths were attributable to infectious diseases, especially neo-natal tetanus, diarrhoea, whooping cough, malaria, measles, pneumonia or meningitis.^1,23 The decline in child mortality that we have observed in Niakhar during the 1970s and the 1980s was also observed in other parts of Africa and was generally attributed to the implementation of immunization programmes and better access to chloroquine for treatment of malaria.^24,25 In the Niakhar area, a major immunization campaign against measles from 1978 to 1982, the introduction of EPI in the early-mid 1980s and campaigns for promoting malaria chemoprophylaxis and presumptive treatment of fever with chloroquine are likely to be the main reasons for the initial decrease of mortality observed from the mid 1970s up to the early 1990s. The rapid progress made over these two decades contrasts with the setback observed during the 1990s.

Analysis of causes of deaths by the verbal autopsy method suggest that an increase in malaria deaths and an epidemic of meningitis were successively responsible for the plateau in mortality and then the peak of deaths that occurred during the period 1992-1999. Both the sensitivity and the specificity of verbal autopsy technique vary considerably according to causes of deaths and epidemiological context.²⁶ Although severe malaria and meningitis have some symptoms in common, such as fever and impairment of consciousness, we believe that correct estimates of deaths attributable to these two diseases were obtained in Niakhar using this method. In the Sahel and sub-Sahel, rains occur only during a short period of the year, and the seasonal peak of children deaths with a picture of high fever, seizure and/or coma occurring a few weeks after the malaria vectors increase massively in numbers allows a diagnosis of malaria to be made with a much better sensitivity and specificity than in areas where rains occur year round. Malaria mortality increased almost three-fold during the 1990s, and this was temporally related to the emergence of chloroquine resistance in P. falciparum.⁸ Meningitis caused by Neisseria meningitis is well known for causing devastating epidemics in the meningitis belt of Africa.^27,28 The dramatic outbreak that occurred in Niakhar in 1998-1999 was responsible for the return to mortality rates in children under 5 years of age of almost 300‰ in the late 1990s. As usual for epidemic meningitis, cases occurred primarily during the last part of the dry season, thus facilitating the differential diagnosis from malaria. An immunization campaign launched by the regional health authorities in 1999 contributed to the control of the outbreak.

The decrease in childhood mortality that has occurred between 2000 and 2010 is striking. There has been no change in the functioning of the DSS nor in the structure of the study populations during this period. Rains were either similar or higher during this period than during the 1990s. Vaccine coverage was very high in the early 1990s - approximately 80% of children were fully vaccinated and figures for the 2000s were either similar or lower. No additional health post has been opened and only limited improvements in facilities and services have occurred. Sociologic changes have been limited with most villagers maintaining a traditional way of life. Fertility remains high; the mean number of births per woman was 8.0 in 1984-1986, 7.0 in 1999-2000 and 6.4 in 2006-2008. The only major change in health care delivery that has occurred during the 2000s has been the launching of improved malaria control policies. AQ + SP replaced chloroquine in all health facilities in Senegal in November-December 2003 and this combination proved very effective both for curing malaria attacks and preventing new infections for at least 3 weeks after treatment.²¹ Artemisin-based combination therapy with AS + AQ was deployed in May 2006 and has also been very effective.¹⁷ Its introduction was accompanied by the implementation of guidelines for the diagnosis and treatment of children presenting with a febrile illness which requires a positive rapid diagnostic test before treatment with artemisinin combination therapy. In addition, long-lasting ITN coverage increased dramatically in June 2008 in the Niakhar area, as in most other rural areas of Senegal, as they were distributed free to children and mothers and became available at low cost to others in 2008 before becoming free to everyone in 2009. ^17,18 Forty percent of the population of the study area reported using ITN in 2008 during the rainy season compared with less than 5% in preceding years. In December 2008, the prevalence of malaria among Niakhar children under 5 years of age was only 2% compared to 79% in November 1995 and 31% in December 2003. These data from Niakhar support other reports from Africa. During recent years, there have been substantial changes in malaria control throughout Africa.^11,29 Between 2006 and 2010 ACTs were deployed in all African countries, and this was often accompanied by the mass distribution of ITNs.^30-32 Recent dramatic decreases in malaria prevalence and mortality have also been reported from The Gambia, Kenya, Zanzibar and São Tomé and have been attributed in large measure to changes in malaria control policies although other biological and environmental factors may have contributed.^29,33-35

The decrease in mortality in the Niakhar study area observed during the 2000s has involved decreases in all the main causes of childhood deaths, including neonatal mortality, and not only malaria. In particular, deaths due to diarrhoeal diseases and ARI in children under the age of 5 years decreased 5- and 7-fold, respectively during the period 2004-2010 compared to the period 1992-1999 in the absence of any specific interventions directed at these conditions. Deaths due to ARI are often difficult to distinguish from those due to malaria through verbal autopsy,²⁶ but this is not the case for deaths due to diarrhoeal diseases. SP given for the presumptive treatment of malaria attacks could have had some impact on diarrhoeal diseases and ARI but it was used widely only in 2004 and 2005 and thus seems to be an unlikely cause for the dramatic changes that have been observed. Since there was no clear change in the treatment of these diseases, we suggest that the decrease in the incidence of these conditions has been due, at least in part, to the indirect consequence of improved malaria control.

There have been both historical and recent examples of a reduction in overall mortality following implementation of effective malaria control programs that was much higher than expected.^36,37 In the 1950s, almost complete elimination of malaria in Sri Lanka and Guyana was accompanied by dramatic reductions in overall mortality that were greater than could be attributed to a direct reduction in deaths from malaria. In Guyana, deaths from respiratory infections fell in parallel with those from malaria. In the 1980s, successful control of malaria with either seasonal chemoprophylaxis or insecticide-treated nets in The Gambia reduced overall mortality in children by nearly 50%, a much larger reduction that was anticipated, and reduced deaths attributed to pneumonia and diarrhoea as well as those attributed to malaria.³⁸ More recently, an effective malaria control program on the island of Bioko, which employed both ITNs and indoor residual spraying, reduced under 5 year child mortality from 152 per 1000 to 55 per 1000 over a 4 year period.³⁹ How a reduction in the incidence of malaria could have such a marked effect on overall mortality is uncertain but there is evidence that malaria impairs the immune response, increasing susceptibility to other infections, ⁴⁰ and that continuous exposure to malaria impairs weight gain during the malaria transmission season.⁴¹

To what extent are the results from Niakhar described in this paper representative of the rest of Senegal? International funding for malaria control in Senegal increased from US $ 1 million in 2004 to US $ 30 millions in 2010.¹⁷ Six millions ITNs were distributed between 2006 and 2010, almost reaching universal coverage in rural areas.^17,18 National estimates of malaria attributable mortality decreased 6 fold between 2003 and 2009.⁴² Overall under-five mortality was reduced by 30% between 2005 and 2008/2009.¹⁷ In all regions of Senegal, these dramatic changes corresponded to the deployment of combination therapy and ITNs.¹⁷ Although pyrethroid resistance in Anopheles gambiae emerged dramatically in Senegal in 2010, ⁴³our data suggest that it had up to now had only a limited impact on child mortality in Niakhar.

Experience from this study and a similar one undertaken in The Gambia⁴⁴ indicate that provision of simple health care measures and, in particular, effective control of malaria can be sufficient to allow poor rural areas of Africa to meet MDG4.