A prospective survey of the outcome of pregnancy in a rural area of the Gambia.

Bulletin of the World Health Organization, 65 (5): 635-643 (1987) i World Health Organization 1987 A prospective survey of the outcome of pregnancy in a rural area of the Gambia* A. M. GREENWOOD,' B. M. GREENWOOD,2 A. K. BRADLEY,' K. WILLIAMS,3 F. C. SHENTON,4 S. TULLOCH,S P. BYASS,5 & F. S. J. OLDFIELD6 The outcome ofpregnancy was studied in 672 women over a 1 -year period in a rural area of the Gambia where medical resources were very limited, prior to the introduction of a primary health care programme. Maternal mortality was very high (22 per 1000), mainly caused by postpartum haemorrhage and infections. Stillbirth and neonatal death rates were also high (35 and 65 per 1000, respectively); prematurity and infections were the main causes ofdeath in neonates. First or late pregnancies, ages under 20 or over 40 years, and multiple pregnancies were all associated with a poor outcome ofpregnancy. Women in these groups should therefore be encouraged by traditional birth attendants and by the staff of rural antenatal clinics to deliver in a health centre or hospital. Despite wide recognition that in rural Africa preg- nancy is a hazardous time for both mother and child (1), there are few data quantifying the degree of risk. Most of the information about maternal mortality is related to hospital-based surveys although, in many rural areas, few women deliver in a hospital or health centre, even when there have been complications during the pregnancy. In these circumstances, most stillbirths and early neonatal deaths occur at home and information on the outcome of pregnancies can be obtained only by asking these women about their past obstetric history or, more accurately, by making direct observations if this should be possible. Few such prospective studies have been undertaken. In Machakos, Kenya, direct observations on the out- come of pregnancy in 4716 rural women over a 4-year period (1975-78) indicated a relatively favourable outcome and maternal mortality was low (2). In the Gambia, demographic records in two villages, Keneba and Manduar, showed a high * From the Medical Research Council (MRC) Laboratories, Fajara, Banjul, Gambia. Requests for reprints should be sent to Dr A. M. Greenwood at this address. 'Epidemiologist. 2 Director. 3Senior Technician. 4Research Officer. 5Computer Manager. 6 Director of Medical Services, Medical and Health Department, Banjul, Gambia. perinatal mortality which changed little during a 25-year period of observation from 1951 to 1975 (3). Since 1975, a paediatrician and a midwife have been working in Keneba and infant and childhood mortalities declined dramatically (4). Keneba and Manduar now no longer reflect the situation pre- vailing in the other rural areas of the country. In 1980 the government's Medical and Health Department embarked on an ambitious primary health care programme to establish a village health worker (VHW) and a traditional birth attendant (TBA) in every village with a population of 400 or more. As a baseline for future assessments of the effects of this programme on maternal and infant mortality a survey was carried out on the course and outcome of pregnancy in women living in a rural area with very limited medical resources. We tried to identify the risk factors for a poor outcome of pregnancy which could help TBAs and the staff of rural antenatal clinics to select pregnant women who require special care. MATERIALS AND METHODS Study area The study was undertaken in 41 villages and hamlets near the town of Farafenni in North Bank Division; the nearest village was 12 km and the furthest was 35 km from Farafenni. There are no 4824 -635- A. M. GREENWOOD ET AL. tarred roads in the area but all the study villages are accessible by laterite roads throughout the year. An occasional taxi travels between the larger villages and Farafenni but at the time of this survey the residents of more isolated villages could reach the town only by foot, bicycle, horse or donkey cart. The geo- graphical and climatic features of the area have been described elsewhere (5). At the time of this survey, from April 1982 to March 1983, the medical resources of the Farafenni area were very limited. The medical facilities of the town comprised a government dispensary, manned by a dresser-dispenser and a midwife, one private medi- cal practitioner, and a number of small pharmacies. Fortnightly maternal and child health clinics were held in two villages situated 16 km to the west and 20 km to the east of Farafenni. Expectant mothers had to travel up to 20 km to reach one or other of these clinics. Patients seen at the Farafenni dis- pensary who required further treatment had to be sent across the river by ferry to Banjul, ajourney of nearly 200 km, which could take several hours. Study population The population in the study area belonged to three ethnic groups, Mandinka, Wollof and Fula, with Mandinka predominating. Heads of households were mainly subsistence farmers, the women being respon- sible for most of the rice cultivation which is very arduous during the latter part of the rainy season. A house-to-house enumeration in November and December 1981 indicated that the total population of the study villages and hamlets was 12 313 of whom 2800 were women aged 15-45 years. Only 4% of women in the child-bearing age had had any formal education. Identification ofpregnant women Two methods were used to identify pregnant women. (a) An individual was appointed in each village and hamlet to be responsible for recording all births and deaths in the village and to identify all pregnant women. This information was given to one of a team of MRC-employed field assistants who were each responsible for data collected in a group of villages and hamlets. (b) At the start of the surveillance period, which was from April 1982 to March 1983, urine samples were collected for pregnancy testing from all women in the reproductive age group (15-45 years) excluding those who had delivered during the pre- ceding 12 months. Approximately 90%-coverage was achieved. Urine tests were repeated in September 1982 before the second clinical survey. Once a woman had been identified as pregnant, she was visited by a field worker with an initial antenatal questionnaire concerning her previous obstetric history. The intervals between births and the pro- portion of children surviving were calculated from data collected retrospectively in this way. Monthly surveillance Each woman identified as pregnant was visited once a month by a field worker with a morbidity questionnaire concerning current complaints and visits to a health facility during the preceding month. On completion of the questionnaire the oral tempera- ture was measured, using an electronic thermometer, and if this was > 38.0 OC a blood film was obtained. If a woman was found to have delivered during the period since the last visit, a final questionnaire was completed, concerning where the delivery had occurred, who had assisted, and the outcome. Clinical surveys On two occasions, once during the dry season (March 1982) and once at the end of the rainy season (October/November 1982), all the women who were known to be pregnant were examined in their village by the same physician, and their weights and heights were recorded. Completion rates of over 95% were obtained for each survey. A fingerprick blood sample was taken for parasitological, haematological and serological measurements and a urine sample was collected for chemical analyses. Treatment was given for any illness identified during the survey. All women found to have a positive serological test for syphilis were treated with penicillin. Identification of maternal and infant deaths The number of abortions, stillbirths, early neonatal and maternal deaths were recorded by the field assistants who presented the postnatal questionnaire. Deaths later in the neonatal period were detected during the course of a survey into the causes of infant and early childhood deaths in the study area. When a stillbirth or a maternal or infant death was reported to the project epidemiologist a visit to the household was made by a physician and a detailed history obtained from the family of the events that had led up to the death. This information, supplemented by any clinic records that were available, was used to try to determine the most likely cause of death. 636 PREGNANCY OUTCOME IN RURAL GAMBIA 637 Laboratory methods Urine was tested for protein and sugar.a Pregnancy tests were carried out using a simplified version of a commercial latex agglutination test;b by using only 20 Al of each reagent, dispensed with a micropipette, it was possible to carry out approximately 40-50 tests with each 1 ml of reagent. Thick blood films were stained with Giemsa; 100 high-power fields were scanned before a film was considered to be negative. The erythrocyte volume fraction was measured with a microhaematocrit centrifuge. Sera were tested for syphilis using a rapid plasma reagin (RPR) card testc as a screening procedure. Sera found positive by the RPR test were tested by the more specific treponemal haemagglutination (TPHA) test; dagglutination at a dilution of 1:80 or more was considered positive in the TPHA test. Tetanus antibodies were measured by a passive haemagglutin- ation assay. Antibodies to rubella were determined by a radial haemolysis assay;e sera which gave an area of haemolysis equal to or greater than that of the positive control were considered positive. Antibodies to cytomegalovirus (CMV) were determined by immunofluorescence using MRC 5 cells infected with the AD 169 strain of virus. An anti-whole- immunoglobulin fluorescein conjugate was used. In this assay sera which gave fluorescence at atitre of 1:2 or greater were considered positive. Antibodies to malaria and to toxoplasma were measured by ELISA tests. For the malaria assay an antigen prepared from placentae infected with Plasmodium falciparum was employed. An IgG peroxidase conjugate was used. For the toxoplasma test a commercial antigen pre- parationf was employed and both IgG and IgM alkaline phosphatase conjugates were used. Sera which gave an absorbance of 0.2 or greater in an ELISA were considered positive. Sera were screened for hepatitis B surface (HBs) antigen using a haemag- dglutination assay. RESULTS Outcome ofpregnancy During the 1-year surveillance period, 789 preg- a Labstix from Miles Laboratories (Ames Division), Slough, England. bPrognostician from Organon Teknika, Cambridge, England. cHynson, Westcott & Dunning; Becton Dickinson (UK) Ltd, Oxford, England. dBurroughs Wellcome, Wellcome Diagnostics, Dartford, Kent, England. eRubascreen from Northumbria Biologicals, Northumberland, England. f Virion from Dynatech Laboratories, Billinghurst, Sussex, England. Table 1. The outcome of pregnancy in 672 women in a rural area of the Gambia during the period April 1982 to March 1983 Outcome Number Rate per lOOOa Maternal death 15 22.3 Intra-uterine death 4 - Abortion 1 5 - Stillbirth 23 34.9 Early neonatal death (< 1 week) 26 40.9 Late neonatal death (1-4 weeks) 15 23.6 Total: neonatal death 41 64.6 Perinatal death 49 74.5 a The stillbirth and perinatal death rates are calculated per total births (658); the neonatal death rates per live births (635). Maternal deaths are defined as deaths during pregnancy, delivery, or within 6 weeks of delivery, which were related to the pregnancy. nant women were identified; 108 of them were still pregnant at the end of the survey and 9 had moved out of the study area. Thus, the outcome of pregnancy was known for 672 women. Maternal deaths, still- births and neonatal deaths were all frequent (Table 1). Only a few abortions were recorded but this figure is almost certainly an underestimate as few women reported their pregnancy until after the first trimester. Maternal deaths Fifteen women died from a cause that was probably related to pregnancy. The mean age of the mothers who died (28.5±8.0 years) was very similar to the mean age of 657 pregnant women who survived (27.3±6.4 years). Death occurred more frequently among primigravidaee,-t-o7i- 6x 9and. among women who had 5 or more previous preg- nancies (8 out of 176; 45 per 1000) than among women who had had 1-4 previous pregnancies (5 out of 420; 12 per 1000) (X2=6.5; 2 df; P<0.05). Only one woman died in hospital, two died on the way to hospital, and the remainder died at home. Eleven women had attended an antenatal clinic at least once during their pregnancy. Likely cases of maternal death are shown in Table 2; postpartum haemorrhage was the most important, being responsible for five deaths (33 %). Infections were responsible for another four deaths and sudden collapse at or shortly before delivery for a further three. Eight women gave birth to a live child, including one pair of twins. All these 9 childrenAdiedbefome-raching the age of 1 year. A. M. GREENWOOD ET AL. Table 2. Features of 15 maternal deaths in rural Gambian women Antenatal Person Time of Age clinic assisting Place of Place of death (after (years) Parity visits Probable cause of death at delivery delivery death delivery) Outcomea 33 3 2 Postpartum haemorrhage Untrained Home Home 4 hours Stillbirth TBA 22 2 0 Amenorrhoea for 3 No one - Roadside - Intrauterine death months, severe abdominal pain, collapse 40 6 3 Postpartum haemorrhage Relative Home Home 1 hour Live bom, died 3/52 36 5 2 Postpartum haemorrhage Relative Next-door Next-door 1 hour Live born, died 3/52 with retained placenta compound compound 28 6 3 Sepsis and jaundice Relative Home Home 4 days Live born, died 3/52 30 8 1 Anaemia, heart failure, Trained Hospital Home 8 weeks Twins, liveborn, died chronic renal failure midwife 1/12 and 3/12 26 6 2 Hepatic coma Trained Hospital Home 5 days Live bom, died 10/7 midwife 30 6 4 Tuberculosis Relative Home Home 2 weeks Live bom, died 6/12 41 12 1 Postpartum haemorrhage Relative Home Home 4 hours Stillbirth 21 3 1 Premature delivery, Relative Home Home 5 weeks Stillbirth convulsions 34 5 1 24 weeks' gestation, No one - Home - Intrauterine death acute toxic illness, ? S. typhi 34 3 2 36 weeks' gestation, No one - Home - Intrauterine death sudden severe abdominal pain 21 1 0 36 weeks' gestation, Untrained - Roadside - Intrauterine death sudden weakness TBA and collapse 16 0 0 Sudden death at Untrained Home Home Instant of Live born, died 6/52 delivery, ? amniotic TBA delivery fluid embolus 16 0 0 Postpartum haemorrhage Untrained Home Home 1 hour Live born, died 1/12 TBA 10n7-at 10 days; 3/52 and 6/52-at 3 and 6 weeks; 1/12, 3/12 and 6/12-at 1, 3 and 6 months. Stillbirths and neonatal deaths Twenty-three stillbirths and 41 neonatal deaths were recorded. Fig. 1 shows the stillbirths and neonatal deaths by month of death. The stillbirth rate showed little change with season but significantly more neonatal deaths occurred during the 6 months of the rainy season and immediate post-rainy season (June-November) (32/343 live births) than during the dry 6 months of the year (December-May) (9/288 live births) (X2=8.9; 1 df; P<0.01). Twenty-four of the 41 neonatal deaths occurred in males and 17 in females. Information on the circumstances of death was obtained for 38 of the 41 infants who died during the neonatal period. Only 7 (18%) died in a hospital or health centre and only 11 (29%) had received any 10- i 5- z 0- Rain Z_ - Neonatal death illbirth A M J J A S O N D J F M Motnth Fig. 1. Number of stillbirths and neonatal deaths recorded among 672 pregnant Gambian women every month from April 1982 to March 1983. 638 PREGNANCY OUTCOME IN RURAL GAMBIA form of western medicine for their final illness. Presumptive causes of death established by post- mortem questionnaire and by examination of record cards are shown in Table 3. Prematurity was the main cause of death in 12 infants, including a pair of twins and a set of triplets. Nine of these 12 infants died within the first 48 hours of life. Feeding problems accounted for the deaths of 5 further infants. In- fections were the main cause of death among infants who survived the first week of life (10/15 deaths). A total of 13 neonates probably died from an in- fection; 5 probably had pneumonia, 4 a generalized septicaemia associated with septic skin lesions, 1 meningitis, 1 neonatal syphilis, 1 acute gastroenteritis and 1 neonatal tetanus. The mother of the latter child had not been immunized. Factors influencing the outcome ofpregnancy Four of the 672 women had multiple pregnancies, comprising 3 sets of twins and 1 set of triplets. Only 4 of these 9 infants survived the neonatal period and only 2 reached the age of 1 year. Factors influencing the outcome ofpregnancy were assessed for 649 women with a singleton pregnancy whose outcome was known, excluding those who had an abortion or intra-uterine death. The results of these comparisons are summarized in Table 4. The out- come of pregnancy was influenced by both age and parity. A poor outcome (stillbirth or neonatal death) was recorded significantly more frequently in women aged less than 20 years and in those 40 years old or more than in women aged 20-39 years (x2= 10.9; 3 df; P<0.05). Neonatal mortality was especially high in those aged 40 years or more (5/24). A poor outcome was recorded more frequently in primi- gravidae and in multigravidae than in those with 1-4 previous pregnancies (X2= 11.7; 2 df; P<0.01). The outcome of pregnancy was not influenced by ethnic group. A higher proportion of women whose preg- nancy ended in a stillbirth had a history of a previous Table 3. Presumptive cause of death in 41 infants who died during the first month of life Cause of death Number Prematurity and death at < 1 week 12 Nutritional problems and failure to thrive 5 Birth trauma 1 Congenital abnormalities 1 Infection 13 No obvious cause 6 No information obtained 3 Total 41 stillbirth than did women whose pregnancy ended in a live birth (7/16 compared with 135/458) but this difference is not statistically significant (X2=2.4; 1 df; not significant). The mean duration between the onset of the index pregnancy and the previous preg- nancy was similar in women with a bad outcome of pregnancy (41.2±14.7 months) and in those with a good outcome (38.0±14.9 months). However, 29 women whose previous child had died had a sig- nificantly shorter birth interval (26.6±12.7 months) than did 357 women whose previous child was still alive (mean birth interval, 38.0±14.9 months) (t=4.0; P<0.001). The proportion of women who had an antenatal card and who were known to have visited an antenatal clinic on at least one occasion during their pregnancy did not differ significantly between women with a good outcome of pregnancy (436/590; 74%) and those with a bad outcome (41/59; 69%). Ninety percent of women who had an antenatal card had a record of administration of at least one dose of tetanus toxoid during the current or previous pregnancies; the Table 4. Risk factors for a poor outcome of pregnancy (stillbirth or neonatal death), based on a study of 649 women with a singleton pregnancy Statistically significant Data suggestive but not risk demonstrated statistically significant No risk demonstrated First pregnancy Previous stillbirth Symptoms in pregnancy Five or more previous pregnancies Short stature, high weight Oedema, raised blood pressure Age: < 20 years Prolonged labour Abnormal erythrocyte volume fraction Age: >40 years Chloroquine administration in pregnancy Malaria parasitaemia Positive serology for syphilis 639 A. M. GREENWOOD ET AL. level of tetanus immunization did not differ between the two groups. An antenatal card record of chloro- quine administration during the current pregnancy was noted significantly more frequently among women with a bad outcome of pregnancy (8/41) than among those whose child survived the neonatal period (26/436) (X2=8.4; 1 df; P<0.01). Field workers completed an average of2.9 monthly morbidity questionnaires both for women with a good outcome of pregnancy and for those with a bad outcome. Complaints were frequent in both groups (fever, abdominal pain, weakness, swelling of the feet, and dysuria in descending order of frequency) but no symptom differed significantly in prevalence between the two groups. An oral temperature of 38.0 IC or greater was recorded at the time of administration of morbidity questionnaires on only 8 of the 1813 times when temperature was recorded. Only two of these women had malaria parasitaemia. All 8 febrile women had a good outcome of pregnancy. A cross-sectional clinical survey of all pregnant women in the study area carried out on two occasions during the study period was not helpful in identifying risk factors for a poor outcome of pregnancy. An abnormal fetal position was detected in 26 out of 307 women with a palpable fetus but only one of these 26 pregnancies had a bad outcome. Oedema was noted in 11 out of 35 women (31 %) with a bad outcome and in a similar proportion of women with a good outcome (155/433; 36%). Systolic hypertension (140 mmHg or greater), diastolic hypertension (90 mmHg or greater) and proteinuria were all found infrequently and were recorded in 17, 26 and 11 out of 475 women, respectively. Only 3 women had hyper- tension and proteinuria; 1 had a stillbirth. Glycosuria was not detected. Trends towards a higher stillbirth rate in short women (1.5 m or less) and towards a higher neonatal death rate among women in the heaviest decile were observed but, overall, the outcome of pregnancy was not influenced signifi- cantly by height or weight. Blood was collected for haematological, parasito- logical and serological determinations during each clinical survey. Malaria parasitaemia was found in 3/37 women (8%) with a bad outcome of pregnancy and in 29/422 (7%) of those with a good outcome. The mean erythrocyte volume fraction of 31 women with a poor outcome of pregnancy (32.8±3.6) was very similar to that in 353 women with a good out- come (32.2±4.4). No difference between groups was found when the comparison was restricted to the .155 women seen during the last trimester. All 25 women with an erythrocyte volume fraction of 25 or less had a good outcome of pregnancy. The results of serological investigations carried out on 237 blood samples obtained from 254 pregnant women during the first clinical survey are shown in Table 5. Few differences were observed between women with a good and bad outcome of pregnancy. A higher IgM titre of toxoplasma antibodies was found in the mothers of infants who died in the neonatal period but no other significant differences were found. Although positive serological tests for syphilis were frequent, these did not not predict a bad out- come for the pregnancy. Because of this rather Table 5. Results of serological tests in 237 pregnant women in relation to the outcome of their pregnancy Outcome of pregnancy No. of children Stillbirth or early Antibody testa alive at 1 month neonatal death Late neonatal death Syphilis: RPR 57/214 (27)b 3/16 (19) 1/7 (14) RPR+TPHA 29/214 (14) 1/16 (6) 1/7 (14) Toxoplasma: IgG 133/195 (68) 7/15 (47) 6/6 (100) IgM 24/195 (12) 1/15 (7) 3/6 (50) CMV 200/209 (96) 13/15 (87) 6/6 (100) Rubella 197/208 (95) 15/16 (94) 6/6 (100) Malaria 213/214 (100) 16/16 (100) 6/6 (100) HBs antigen 30/193 (16) 4/16 (25) 2/6 (33) 'See text for details. bFigures in parentheses are percentages. 640 PREGNANCY OUTCOME IN RURAL GAMBIA surprising finding these tests were also carried out on a further 221 samples collected during the second clinical survey. Once again, no correlation with the outcome of pregnancy was found. The prevalence of HBs antigen was about twice as high in women with a poor outcome of pregnancy as in those with a good outcome, but the numbers were small and this dif- ference is not statistically significant. Mean titres of toxoplasma, malaria and CMV antibodies did not differ between women with a poor outcome of pregnancy and those with a good outcome. Information on the circumstances of delivery was obtained by questionnaire for 59 women with a bad outcome of pregnancy and for 590 women with a good outcome. The proportion of women delivered by a midwife in a health centre or hospital was low and was similar in each group (5% and 4%, re- spectively). Histories of prolonged labour, excessive bleeding and tearing were all given more frequently by women with a bad outcome of pregnancy than by women with a good outcome, but none of the differences between the two groups is statistically significant. DISCUSSION In this study we tried to identify all the pregnant women in a group of 41 villages and hamlets in North Bank Division of the Gambia during a one-year period and to determine the outcome of their pregnancies. We believe that by using a combination of registration methods and urine tests on all at-risk women very few pregnancies were missed. Very high maternal, stillbirth and perinatal death rates were recorded. We have found few other comparable sets of data, collected during the course of community surveys in Africa, with which to compare our find- ings. In a comprehensive survey of the relationship between fertility, birth intervals and fetal mortality, Armagnac & Retel-Laurentin (6) recorded an abortion rate of210 per 1000 and a stillbirth rate of50 per 1000, respectively, in a rural area of Burkina Faso. In the relatively prosperous Machakos area of Kenya, Voorhoeve et al. (2), using survey techniques similar to our own, found a stillbirth rate of 30 per 1000, a neonatal death rate of 23 per 1000 and a maternal death rate of only 0.8 per 1000, a figure 30 times lower than that recorded in our study. In the Gambia, a study carried out in the villages of Keneba and Manduar over a 25-year period from 1951 to 1975 gave stillbirth rates of 64 and 89 per 1000, neonatal mortality rates of 85 and 50 per 1000, and maternal mortality rates of 11 and 10 per 1000, respectively (3). Thus, our data suggest that there has been little or no improvement in the outcome of pregnancy in the rural areas of the Gambia during the past few years. The maternal death rate in the Fara- fenni area is about 200 times greater than that recorded in industrialized countries, the neonatal death rate 5-10 times higher, and the stillbirth rate about 3 times higher. The maternal mortality rate recorded in our study is very high, even by the standards of other developing countries, e.g., a mortality of 1 per 1000 live births in Jamaica (7). Caution is required in making generalizations from the findings obtained in a restricted population during a survey period of only one year, but our findings suggest that in rural Gambia as many of 1 out of 8 women still die in pregnancy or in childbirth. Only 2 of the 15 women who died had been seen in a hospital or health centre during their final illness so that maternal death rates based on hospital or health centre records would have grossly underestimated the size of the problem. Even before the introduction of a new primary health care programme, rural health services in the Gambia were relatively effective, as shown by the high proportion of pregnant women seen in our study who had attended an antenatal clinic and received tetanus immunization. It is likely that in some other parts of Africa where rural health programmes are less well developed, maternal mortality is at least as high as in the Gambia. To try to determine the ways by which these maternal deaths might have been prevented, we obtained as much information as possible about each woman who died. In 11 of the 15 cases death occurred either before or within 4 hours of delivery and was associated with haemorrhage or sudden collapse. Because of lack of transport and the absence of resuscitation facilities at the nearest dispensary it is unlikely that these deaths could have been prevented. Although 11 ofthe 15 women who died had been seen at least once at an antenatal clinic, none had been referred to a health centre for delivery. The stillbirth rate among Farafenni women was not especially high and was similar to that reported in the more prosperous community of Machakos, Kenya (2). A history of a previous stillbirth was obtained more frequently from women whose pregnancy ended in a stillbirth than in women with a successful out- come of pregnancy and it is our clinical impression that in the study community there are several women who experience recurrent stillbirths. A number of possible infective causes of stillbirth were con- sidered. Malaria was unlikely because no seasonal variation in the incidence of stillbirths was observed and, in the Gambia, malaria is very seasonal. Al- though a high prevalence ofpositive antibody tests for syphilis was found, this infection did not appear to be as significant a cause of stillbirths as in rural Burkina Faso (6) and urban Zambia (8). 641 A. M. GREENWOOD ET AL. Deaths among neonates fell into two main groups. Prematurity was the main cause in infants who died during the first few days after birth, and infection was the min cause of death in neonates who survived the first week of life. A history suggestive of death from neonatal tetanus was given by only one family, which accords with the high prevalence of tetanus antibodies in pregnant Farafenni women. Maternal death was an important cause of neonatal and infant mortality. All 9 children born to mothers who died failed to reach the age of 1 year. A maternal death is thus, almost inevitably, a double tragedy. We tried to identify risk factors that might help traditional birth attendants and midwives working in rural antenatal clinics with few facilities to identify at-risk pregnant women who might benefit from delivery in a health centre. Because our sample size was relatively small, only risk factors exerting a strong effect would have been detected. As expected, we found that primigravidae, women with 5 or more previous pregnancies, women over the age of 40 years, or women with multiple pregnancies were all at risk. Our attempts to detect other risk factors were unsuccessful. Monthly administration of morbidity questionnaires by field staff, together with the measurement of temperature, did not provide any helpful indicators. Examination of each pregnant woman by a physician on one occasion during preg- nancy was also unhelpful; very few women had any signs of pre-eclampsia or other recognized risk factors. Determination of the erythrocyte volume fraction and serological tests for a variety of infections known to cause congenital infections were, in the main, unhelpful although a tendency to a poorer outcome was noted in women who were HBs antigen positive. Thus, it is likely that in the Farafenni area the outcome of pregnancy is determined largely by obstetric factors and that infections such as malaria and syphilis, which might be amenable to specific interventions, do not play a major role in causing stillbirths or neonatal deaths. From data in Table 2 it is clear that in many of the maternal deaths the attendants at delivery had no possibility of alleviating the problem. Transport from outlying villages is extremely difficult to obtain in an emergency and at the time of this study, the Farafenni Health Centre had inadequate staff and facilities for dealing with obstetric disasters. The Government Hospital in Banjul, the nearest place with blood transfusion and obstetric services, is reached only after a journey of several hours, including crossing of the River Gambia by ferry. Consequently the importance of identifying the at-risk mother well before delivery should be emphasized and training given to TBAs and visiting midwives who conduct antenatal clinics in the health centres. Mothers at risk must be en- couraged to attend the health centre nearest to them for delivery, transport facilities should be improved, and a place should be made available for mothers to stay near a health centre prior to delivery. Whenever possible, major health centres should be upgraded to include blood transfusion facilities and the services of an obstetrician. ACKNOWLEDGEMENTS We thank the field and laboratory staff at the MRC Field Station, Farafenni, for their help with this project, Dr C. Bello for help in measuring the antibodies to cytomegalovirus, Mr R. W. Snow for calculating the birth intervals, Mr R. Hayes for help with the study design and for reviewing the manuscript, and Miss M. Faal for secretarial assistance. RESUME GAMBIE: UNE ENQUETE PROSPECTIVE SUR LISSUE DES GROSSESSES DANS UNE REGION RURALE L'issue des grossesses a fait l'objet d'une dtude dans une region rurale de Gambie disposant-de ressources medicales tres limitees, avant introduction d'un programme de soins de sante primaires. Cette etude visait egalement 'a identifier les facteurs de risque pouvant etre utilises par les accoucheuses traditionnelles et le personnel des dispensaires ruraux de soins prdnataux pour selectionner les femmes enceintes necessitant des soins particuliers. On a enregistrd pendant un an (1982-1983) l'issue des grossesses de 672 femmes habitant 41 villages et hameaux situes aux environs de la ville de Farafenni, sur la rive nord du fleuve Gambie, 'a 100 km de la c6te. Les femmes en- ceintes ont ete identifiees soit par les agents des villages, soit grace aux analyses d'urine effectuees a deux reprises au cours de la periode d'etude chez toutes les femmes agees de 15 a 45 ans. Par consequent, il est peu probable que de nombreuses femmes enceintes aient dchappe a l'etude. Un agent de terrain leur a rendu visite une fois par mois jusqu'a leur accouchement et, a deux reprises au cours de l'annee, toutes celles qui etaient presentes ont ete examinees dans 642 PREGNANCY OUTCOME IN RURAL GAMBIA 643 leur village par un medecin. Le taux de d6ces maternel enregistre a ete de 22 pour 1000, ce qui est tres eleve. Les primipares et les femmes ayant deja mene au moins cinq grossesses presentaient un risque plus eleve de deces. Hemorragies et infections ont etd les principales causes de deces maternel. Neuf enfants vivants sont nes de 15 meres d6cedees. Tous ces enfants sont d6cedes avant d'atteindre I'age d'un an. On a enregistre 23 mortinaissances et 41 deces neonatals (taux de morti- natalite, 35 pour 1000; taux de deces neonatal, 65 pour 1000). Les principales causes de d6ces des nourrissons morts au cours de la premiere semaine ont ete la prematurite et ses complications, les deces des nourrissons ayant survecu a cette pdriode etant essentiellement consecutifs a des infections. On a recherche les facteurs ayant eu une influence sur l'issue de la grossesse chez 649 femmes ayant mene une grossesse simple. L'age comme la parite ont eu une in- fluence sur cette issue, qui a ete le plus souvent defavorable (mortinatalite ou d6ces neonatal) chez. les femmes tres jeunes et les femmes agees. Les enquetes de morbidite n'ont ete d'aucune utilite pour identifier les femmes a risque et peu d'anomalies ont Wtd decelees au cours des enquetes transversales menees par un medecin. L'issue de la gros- sesse n'etait pas liee au volume globulaire. On a trouv6 une forte pr6valence de la seropositivite dans les 6preuves de depistage de la syphilis et de l'hepatite B. On a observe que l'issue de la grossesse avait tendance a etre defavorable chez les femmes HBsAg-positives, mais non chez les femmes seropositives pour la syphilis. nI est probable que dans la zone d'etude, l'issue des grossesses est largement determinee par des facteurs obstetricaux et que les infections telles que paludisme et syphilis, que l'on peut corriger par des interventions sp&cifiques, ne constituent pas une cause importante de mortinatalite ou de deces neonatal. REFERENCES 1. ROSENFIELD, A. & MAINE, D. Maternal mortality-a neglected tragedy. Where is the M in MCH. Lancet, 2: 83-85 (1985). 2. VOORHOEVE, A. M. ET AL. Machakos project studies. Agents affecting health of mother and child in a rural area of Kenya. XVI. The outcome of pregnancy. Tropical and geographical medicine, 31: 607-627 (1979). 3. BILLEWICZ, W. Z. & MCGREGOR, I. A. The demo- graphy of two West African (Gambian) villages, 1951-75. Journal of biosocial science, 13: 219-240 (1981). 4. LAMB, W. H. ET AL. Changes in maternal and child mortality in three isolated Gambian villages over ten years. Lancet, 2: 912-914 (1984). 5. GREENWOOD, B. M. ET AL. Ethnic differences in the prevalence of splenomegaly and malaria in the Gambia. Annals oftropical medicine andparasitology (in press). 6. ARMAGNAC, C. & RETEL-LAURENTIN, A. Relations between fertility, birth intervals, foetal mortality and maternal health in Upper Volta. Population studies, 35: 217-235 (1981). 7. WALKER, G. J. A. ET AL. Maternal mortality in Jamaica. Lancet, 2: 486-488 (1986). 8. WATTS, T. & HARRIS, R. R. A case-control study of stillbirths at a teaching hospital in Zambia, 1979-80: antenatal factors. Bulletin ofthe World Health Organ- ization, 60: 971-979 (1982).