Malaria prevention during pregnancy in unstable transmission areas: the highlands of Madagascar


M. Cot INSERM U.149, Service de gynécologie-obstétrique, Hôpital Tenon, 4 rue de la Chine, 75020 Paris, France. Fax: +33 1 5601 7188; E-mail:


Malaria transmission in Madagascar is highly variable from one region to the next, and the consequences of the disease on pregnant women and their foetuses are not fully documented. In midwestern Madagascar, the high-transmission lowlands in the west of the country meet the central plateaux, where malaria is unstable because of the high altitude and annual indoor spraying of DDT since 1993. We studied five of the region's main maternity clinics. We began by interviewing sample groups of women of childbearing age living within the vicinity of each clinic. This enabled us to determine the extent to which they had accessed and made use of available maternal health services during pregnancy and delivery, and, hence, to estimate the feasibility of boosting the prophylaxis. We then spent a whole year (from June 1996 to May 1997) observing deliveries at the five clinics in order to gauge the prevalence of placental infection and its consequences on birthweight in various transmission situations. Although only between 2 and 15% of the women said that they had taken prophylaxis during their previous pregnancy, the vast majority had benefited from preventive care: 97% had attended an antenatal visit on at least one occasion and 84% had had the assistance of medical or paramedical staff during delivery, even when their homes were situated relatively far away from the clinic (76%). In total, we observed 1637 deliveries with a mean placental malaria prevalence rate of 8.1%. Individual prevalence rates, however, were found to differ significantly between the maternity clinics situated in the east (minimum 2.1%) and west (maximum 26.2%) of the region. There were also marked variations in line with the seasonal fluctuations in entomological transmission. On the whole, a greater percentage of low birthweights (LBWs) was recorded at the lowland clinics than at the highland ones (17.1% vs. 9.7%), possibly because of the higher malaria infection rate in low altitude areas. On the other hand, the relative risk of LBW linked to placental infection was far greater in the highlands [4.9 (3.3–7.3)] than in the lowlands [1.9 (1.2–3.0)]. Although the rate of placental malaria among women inhabiting the country's central plateaux may be low, it means that transmission – and, hence, the risk of LBW because of placental infection – still persists in spite of the indoor DDT spraying programme. For maximum efficacy, we recommend a combination of vector control (extended to lower altitude areas outside the current OPID zone) and preventive care – i.e. individual chemoprophylaxis – for all highland women during pregnancy.


The many studies investigating the consequences of malaria during pregnancy in stable malaria areas – particularly in sub-Saharan Africa – have helped to produce a relatively clear picture of the concomitant public health problems, reduced birthweight and, to a lesser extent, maternal anaemia, and to single out a subgroup likely to benefit from preventive care: primigravidae (Morley et al. 1964; Brabin 1991; Greenwood et al. 1994; Cot et al. 1995). Less is known about the situation in unstable malaria areas because too few subjects are afflicted to be able to conduct systematic surveys there. Most case studies have focused on sample groups of hospitalized subjects, so their findings are liable to distort estimates of incidence rates and the seriousness of clinical cases observed among pregnant women (Menon 1972; Herd & Jordan 1981; Sholapurkar et al. 1988).

Women in stable malaria areas have acquired a protective immunity, which, though partially diminished during pregnancy, remains strong enough to prevent the appearance of acute clinical symptoms in cases of infestation. Nobody, however, can tell what the situation is with regard to maternal morbidity/mortality when malaria is unstable – although it would be fairly safe to say that because local people lack such well-established immunity, clinical attacks there will be more serious and liable to jeopardize the outcome of pregnancy. Just as little is known about the effects of placental malaria infection on the state of a newborn child in low-transmission areas, not least in terms of birthweight, which has a strong bearing on neonate and infant morbidity (Kramer 1987).

This lack of scientific knowledge goes hand in hand with questions regarding the opportunity for public health policy measures to protect all women from malaria. In stable malaria areas, the vast majority of placental infections and low-birthweight neonates are found among primigravidae, which clearly puts them at the front of the queue when it comes to deciding who needs to benefit from such measures. In unstable malaria areas, the fear is that because they enjoy less protection, multigravidae are vulnerable too. Meanwhile, the impression given by the very low transmission that serious complications are scarce in those areas raises doubts as to whether the cost-effectiveness ratio makes it worth bringing preventive care into general use among all pregnant women.

Midwestern Madagascar is a transitional region featuring both types of epidemiological situations: highly prevalent, seasonal malaria transmission in the hot and humid lowlands in its western parts; unstable malaria on the central plateaux to the east, because of the high altitude and annual indoor spraying of DDT carried out for the past 4 years under the Opération de Pulvérisations Intra-Domiciliaires (OPID) programme (Jambou et al. 2001). Madagascar's health ministry currently adheres to a policy of recommending that every pregnant woman, regardless of where she lives or her gravidity, take a weekly dose of chloroquine (5 mg/kg).

We conducted a series of studies focusing on the population of pregnant women that inhabit this region in order to determine the consequences of malaria during pregnancy, and suggest a range of prevention measures suited to each type of situation. To begin with, we carried out a survey that enabled us to become acquainted with the attitudes and practices of women of childbearing age, along with the conditions in which previous pregnancies and deliveries had taken place. We then monitored five maternity clinics for a full year (1996–1997) so as to assess the effects of placental malaria infection in the different transmission situations.

Materials and methods

Interviewing women of childbearing age

Five maternity clinics that we regarded as representative of our study zone, both geographically and in terms of sample group enrolment, were selected for a preliminary survey revolving around the registers of deliveries. The clinics in question were situated along an axis running from the central highlands (1200–2000 m) in the east to lower lying areas (around 700 m) in the west (Figure 1): Betafo (altitude 1400 m), a maternity clinic that is part of a hospital and generally takes in women living in areas targeted by the OPID indoor DDT spraying programme, which at first sight effectively reduces transmission (Brutus et al. unpublished data); Soavina and Ambohimanambola (around 1300 m), which are located near Betafo, also within the OPID zone yet apparently not achieving the same degree of effectiveness in view of the persistence of significant prevalence of malaria contracted locally (Brutus et al. unpublished data); and Mandoto (900 m) and Anjoma Ramartina (700 m), which are situated further west in a lower lying region outside the OPID zone, where malaria is highly endemic and more stable.

Figure 1.

Study field (with maternity clinics highlighted in red).

Three sample groups of women of childbearing age (15–45 years) were selected per clinic. Each group was composed of around 30 subjects (depending on the size of the urban area in question) living (1) less than 1 km from the clinic, (2) in a village 1–5 km and (3) between 5 and 10 km away. These samples were representative of the entire eligible population living in the corresponding village or neighbourhood (in the case of a clinic in a large urban area). The women were asked about the conditions of their previous deliveries (place, assistance), how frequently they had visited their antenatal clinics, their motives, and a variety of socio-economic variables.

Monitoring of maternity clinics

Starting at the end of May 1996, midwives at the five maternity clinics were asked to fill in a fact sheet for each delivery taking place over the following 12 months, the main points to note being the mother's obstetric and medical history, whether antimalaria chemoprophylaxis had been taken during gestation, and the characteristics of the newborn child. Furthermore, a placental smear was taken to see if the placenta showed any signs of malaria infection (i.e. presence of parasites). Slides were extemporaneously fixed in methanol by the midwives, then collected, stained in Giemsa and examined under the microscope in the malariology laboratory at IRD (Institut de Recherche pour le Développement, formerly ORSTOM) in Antananarivo.


Interviewing women of childbearing age

In all, 422 women responded to the questionnaires asking about their practices during pregnancy. Sixty-four percent had given birth to their previous child at a maternity clinic (Tables 1 and 2). This proportion remained more or less constant, regardless of how far they lived from a health centre. It did, however, vary according to the clinic – the women's practices seemingly being influenced by their financial situation (the midwives from some maternity clinics charged for deliveries at home) and the health centre's reputation. Eighty-four percent of deliveries were assisted by medical or suitably trained paramedical staff at home or at a clinic. Few disparities were found to exist between maternity clinics with respect to medically assisted deliveries, with Betafo, which is located in a large urban area, being the only clinic posting a rate of less than 80%. This rate diminished with distance between home and clinic (93% when the home was close to the clinic and 69% when they were separated by a distance of more than 5 km).

Table 1.  Women's practices during previous pregnancy: antenatal visits and deliveries at the five maternity clinics Thumbnail image of
Table 2.  Women's practices during previous pregnancy: antenatal visits and deliveries according to distance between home and clinic Thumbnail image of

Ninety-seven percent of the women had had at least one antenatal visit (ANV) during pregnancy, and 80% had had three or more. These proportions were quite markedly influenced by the distance factor: the further away the clinic was, the lesser the degree of follow-up.

We did a logistic regression analysis to determine what factors might influence the chances of having a medically assisted delivery (Table 3). Four variables were found to be highly influential: primarily, closeness to the maternity clinic and number of ANVs; secondarily, the woman's age and factors specific to the clinic's location. A larger proportion of women benefited from specialized assistance when they were living within 1 km of the clinic (93 vs. 78%; odds ratio=3.6) and had three or more ANVs (83 vs. 66%; odds ratio=2.6). Those giving birth with the help of medical or paramedical staff tended on average to be younger (29.5 vs. 31.7 years; P=0.01).

Table 3.  Logistic regression: probability of a medically assisted delivery, according to distance from clinic, number of antenatal visits, age of mother and nearest clinic Thumbnail image of

Monitoring of maternity clinics

We monitored 1637 deliveries in the five maternity clinics. Of the 1609 placental smears taken, 8.1% showed signs of placental malaria infection. This overall percentage varied considerably according to geographical location (Table 4): the clinics in the western (non-OPID zone) parts of our study area (Mandoto and Anjoma Ramartina), where altitudes were lower (700–900 m), posted the highest infection rates (26.2 and 24.8%, respectively), while those located in the OPID-zone highlands (Ambohimanambola, Betafo and Soavina) posted the lowest rates (5.1, 2.6 and 2.1%, respectively).

Table 4.  Birthweight and rate of placental infection per clinic Thumbnail image of

Overall, the average birthweight was 2898 g, with 11.3% of neonates weighing under 2500 g. Differences between the proportion of low birthweights (LBWs) in highland and lowland maternity clinics were less marked than for the prevalence of placental infection (9.7% for the three highland clinics vs. 17.1% for the two lowland ones; P < 10–4).

Because of the considerable differences observed in the prevalence of malaria infection inside and outside the OPID zone, we divided the maternity clinics into two groups for the next stages of our analysis: the first regrouping the three clinics located at an altitude of over 1000 m (Betafo, Soavina and Ambohimanambola), and the other containing the two located at lower altitudes (Mandoto and Anjoma Ramartina).

In the lowland clinics placental infestation rates were subject to fairly strong seasonal fluctuations (Figure 2). They peaked between March and May (theoretically the prime transmission period because of the high temperatures and profuse rainfall) as well as in October–November, and plunged in July – the driest and coldest month of the year (hence the least favourable for the multiplication of Anopheles).

Figure 2.

Monthly placental infestation rates at maternity clinics within (highlands) and outside (lowlands) the OPID zone (June 1996 to May 1997).

Variations were less pronounced in the highlands. On the whole, prevalence rates remained low because of both the altitude not being conducive to Anopheles survival, and local towns and villages being sprayed with DDT. That said, they were noticeably lower in July, as well as in November and March–April.

Gravidity was found to have a major influence on placental infection. Primigravidae appeared to be the group most prone to develop such an infection, and the difference between primigravidae and multigravidae was more pronounced in the lowlands (44.8 vs. 19.3%; P < 10–4) than in the highlands (5 vs. 2.3%; P=0.01).

Gravidity also appeared to be by far the most influential factor affecting birthweight, which regularly increased with numbers of pregnancies, while proportions of LBW neonates diminished (Table 5). This was seen to be the case in lowland and highland areas alike, although women in the lowlands (unless they were grand multigravidae) were usually found to give birth to lower weight babies than women in the highlands.

Table 5.  Average birthweight and proportion of low birthweight (LBW) per parity in both areas (lowlands and highlands) Thumbnail image of

Decreasing birthweight was very closely associated with placental infection (Table 6, P < 0.001 in both highlands and lowlands). Despite the fact that malaria was far less prevalent in the highland OPID zone, weight variations attributable to placental infection were far more significant in this region than in the lowlands (398 vs. 250 g), and the same can be said of the relative risk of LBW (4.89 vs. 1.93; Table 6).

Table 6.  Average birthweight and proportion of low birthweight (LBW) according to placental infection in both areas (lowlands and highlands) Thumbnail image of

During each delivery, the mother was asked if she felt feverish. Despite the subjectivity of this question, the association between fever and placental infection was most significant for the highland maternity clinics (12.2% of placental infections among febrile women vs. 2.8% among non-febrile women, P < 0.01), which was not the case with deliveries taking place in the lowlands (25.8 and 26.3% of placental infections, respectively, P=0.96). The relative risk of placental infection linked to fever was 2.2 for women on the High Plateaux [95% confidence interval (1.23; 4.02)].


Midwestern Madagascar is the region where the epidemic zone of the high plateaux meets the stable malaria area of the lowlands. The monitoring of maternity clinics showed that placental infestation rates in the lowlands were generally high (some 25%), with seasonal variations that more or less tallied with the March–May transmission peaks noted by entomologists, and the near-total disappearance of Anopheles in July during the cold season (Brutus et al. 2001; Cot et al. 2001).

In contrast with this stable malaria situation, placental infection rates in the highland clinics remained consistently low, complying with the very low degree of entomological transmission observed in neighbouring areas (Brutus et al. unpublished data). The paradoxical observation that malaria transmission was minimal during the months of March and April – theoretically the peak period – could be attributed to the fact that these areas were in the OPID zone and, hence, had been subject to DDT spraying in December 1996 to January 1997 (the reduction in the Anopheles population usually taking 1–2 months to take effect on the parasite rate in humans). Meanwhile, the difference in prevalence rates between primigravidae and multigravidae was less pronounced in the highlands than in the lowlands. This was probably because of the fact that since malaria is unstable in high-altitude areas, women in general cannot develop an effective antimalarial immunity, as it is the case in their counterparts in the high-transmission areas to the west (with the reduced immunity during first pregnancies being less evident in the former than in the latter). It should be noted, however, that no matter how low infestation rates may be, they prove that transmission still persists and affects pregnant women attending even the most centrally located maternity clinics.

Birthweights are generally low (an overall average of 2900 g) and represent a public health problem as more than 11% of neonates (as much as 16% in the month of February) weigh less than 2500 g. Placental infection, whose prevalence decreases conversely to altitude, is associated with a particularly marked decrease in birthweight in the areas subject to DDT spraying and, moreover, a fivefold increase in the proportion of LBW neonates (rising from 9 to 43%). This phenomenon is far less evident in the lowlands, where there no such vector control campaign takes place. Similarly, the clear association between placental infection and fever during delivery (with twice the risk of placental infection among febrile subjects) is only observed in the high plateaux region, possibly because women living at high altitudes with DDT vector control have less protection and, hence, develop acute clinical symptoms. Primigravidae, wherever they live (albeit to a less conspicuous degree in the highlands), represent an especially vulnerable population.

So it is important to protect that population, regardless of transmission intensity (lowlands or foothills of plateaux). There are two possible means to do this: chemoprophylaxis and vector control. Chemoprophylaxis is the method officially recommended by the department of health. When questioned on the subject, however, very few of the women interviewed said that they had continued taking prophylaxis throughout pregnancy (only 2–15%, depending on the clinic, as residual chloroquine in the urine confirmed). Yet they are well aware of the potential problems during gestation or delivery, as 97% claimed to have paid at least one visit to an antenatal clinic in the course of their previous pregnancy, and 84% had been assisted by medical staff or paramedics trained in obstetrics during delivery, even if they lived relatively far away (5–10 km) from the maternity clinic (76% of women not living in the immediate vicinity of a health centre had medically assisted deliveries).

Our view is that contrary to what has been observed in other geographical areas (Kaseje et al. 1987; Brabin et al. 1990), the reason for the poor level of adherence to prophylactic treatment has more to do with a lack of available medicines than any lack of motivation on the part of the target population or health workers, which seems excellent in this region. We have, therefore, introduced a scheme whereby midwives and health unit managers distribute chloroquine and information in the dialect of all the pregnant women attending antenatal clinics. We are counting on this scheme – which came into effect at the start of 1999, and is currently under evaluation – to help bring about a significant improvement in neonate birthweight, not least because resistance to chloroquine in the region has been shown to be very low (R. Jambou 1998, personal communication).

Efforts are already being made to wipe out the vectors in every area over 1000 m in altitude via the health ministry's annual indoor DDT spraying programme (OPID). This measure is not specifically designed to address the needs of pregnant women, but nonetheless gives them the same degree of protection as the rest of the population.

A number of recent trials have sought to assess how the use of insecticide-treated bednets affects pregnancies in stable malaria areas. Two studies carried out in Thailand and The Gambia (Dolan et al. 1993; d'Alessandro et al. 1996) have shown a more or less pronounced decrease in maternal anaemia and, in The Gambia, a reduction in the number of premature births. Two other studies in Ghana and Kenya (Browne 1997; Shulman et al. 1998) revealed no effect whatsoever, either on the mother or on the newborn child. In any event, vector control has had little or no effect as a means of preventing LBW.

These contrasting results demand some comment. First, all but one of the studies were carried out in high transmission areas where, according to some (Mbogo et al. 1995; Snow & Marsh 1995; Trape & Rogier 1996), vector control would need to be massively effective for a fall in transmission rates to have a significant impact on public health. Second, they all opted for impregnated bednets as the means of control, whereas bednets only really work if used at the time of transmission. The anopheles population is relatively small in Midwestern Madagascar, especially in the foothills of the plateaux (Brutus et al. 2001; Cot et al. 2001), and vector control measures consist of indoor DDT spraying, the effects of which individuals are unable to escape unless they sleep outside the home.

Hence, we believe the most effective way to protect pregnant women in Midwestern Madagascar against the consequences of malaria to be chemoprophylaxis together with informing the women and raising their awareness; making the medicine available in regularly restocked health centres, plus vector control through insecticide spraying, if possible extended to peripheral lower-altitude areas.


We thank Doctors V. Hanitrasoamampionona, M. Ramiaramanana, M. Ratsimbazafy and F. Mauny, Mr G. Le Goff and Miss V. Rajaonarivelo for their help, as well as the staff of the IRD laboratory in Antananarivo, the midwives of the Betafo health district, and the pregnant women who contributed to the surveys. The Midwestern Madagascar pregnancy monitoring programme was financed by IRD (formerly ORSTOM).