How effective is antenatal care in preventing maternal mortality and serious morbidity? An overview of the evidence


Dr. Guillermo Carroli World Health Organization, RHR-X41, CH-1211 Geneva 27, Switzerland. Tel.: +41 22 791 3328; Fax: +41 22 791 4171; E-mail:


This is an overview of evidence of the effectiveness of antenatal care in relation to maternal mortality and serious morbidity, focused in particular on developing countries. It concentrates on the major causes of maternal mortality, and traces their antecedent morbidities and risk factors in pregnancy. It also includes interventions aimed at preventing, detecting or treating any stage along this pathway during pregnancy. This is an updated and expanded version of a review first published by the World Health Organization (WHO) in 1992. The scientific evidence from randomised controlled trials and other types of intervention or observational study on the effectiveness of these interventions is reviewed critically. The sources and quality of available data, and possible biases in their collection or interpretation are considered. As in other areas of maternal health, good-quality evidence is scarce and, just as in many aspects of health care generally, there are interventions in current practice that have not been subjected to rigorous evaluation. A table of antenatal interventions of proven effectiveness in conditions that can lead to maternal mortality or serious morbidity is presented. Interventions for which there is some promising evidence, short of proof, of effectiveness are explored, and the outstanding questions formulated. These are presented in a series of tables with suggestions about the types of study needed to answer them.


Background and aims

Maternal mortality is the health indicator that shows the greatest differential between developing and industrialised countries. The lifetime risk of death as a result of pregnancy or childbirth is estimated at one in 23 for women in some countries in Africa, compared with about one in 7000 for women in Northern Europe. Less reliable information exists on levels of serious morbidity related to pregnancy and childbirth, and on differentials between developed and developing countries. However, it is clear that those conditions that can lead to maternal mortality occur much more commonly in developing countries, and/or give rise to far higher case fatality rates.

Whether antenatal care can prevent or contribute to the prevention of maternal mortality and serious morbidity is a difficult question to answer definitively. Although systematic antenatal care was first introduced in the early 1900s in Europe and North America, and is now almost universal in developed countries, questions related to its effectiveness have only begun to be tackled comparatively recently. In 1972, Cochrane wrote: ‘By some curious chance, antenatal care has escaped the critical assessment to which most screening procedures have been subjected’. He further recommended that ‘the emotive atmosphere should be removed and the subject treated like any other medical activity and investigated by randomised controlled trials’.1 Current knowledge, doubts and recommendations for antenatal care have been explored in Effective Care in Pregnancy and Childbirth,2 in Caring for Our Future: the Content of Prenatal Care3 and in the articles by Villar and Bergsjø.4,5 These publications, however, largely deal with populations with low maternal mortality, and examine the effectiveness and potential of care in pregnancy to improve perinatal and infant mortality, and the general wellbeing of the family rather than maternal mortality and serious morbidity.

The role that pregnancy care, as distinct from delivery care, has played in this dramatic decline in maternal mortality in the developed world is not clear.6[7][8]–9 Reviews on improving the status of women, providing family planning programmes, provision of safe abortion services, strengthening antenatal care, improving emergency obstetric services, training traditional birth attendants (TBA) and community mobilisation have provided optimistic10,11 and pessimistic9,12,13 views of the potential of antenatal care to reduce maternal mortality. There is, however, a notable lack of comprehensive and critical reviews of the effectiveness of antenatal care programmes and/or of individual interventions during pregnancy to avert maternal death or severe morbidity. An article challenging aspects of routine care that do, or could improve maternal survival14 provoked responses from many parts of the world.15[16][17][18][19]–20 Solutions were suggested but no firm evidence put forward for the most pressing local problems, which included anaemia, hypertensive diseases of pregnancy (HDP) and unwanted pregnancy.

There is a widespread desire to improve maternity care services and make optimum use of women’s contact with the health services. If considerable resources are to be devoted to providing antenatal care, then it is important to identify which interventions are effective and how best to deliver them. The primary aims of the document are as follows:

• to identify those interventions in pregnancy that have been proved to be effective in terms of reducing major maternal morbidity and mortality;

• to identify promising but unproven interventions to be assessed;

• to define research priorities.

It should be noted that some interventions may not be effective in reducing maternal mortality or morbidity, but are effective in either improving general or perinatal health and, for these reasons, should also be included in antenatal care programmes. The present article updates a review first published by the World Health Organization (WHO) in 1992.21

Scope, sources and definitions

This overview draws together available information on how antenatal care could be used to reduce serious morbidity and maternal mortality, especially in areas where high levels of morbidity and maternal mortality currently arise. In deciding which interventions to include, the overview begins with the main causes of maternal mortality.

The causes of maternal deaths are often multifactorial and involve complex interactions of several medical, obstetric, health service and social factors. Attribution of deaths to a given underlying cause may be an artefact of classification when infection caused by unclean delivery leads to secondary postpartum haemorrhage, or pre-eclampsia to placental abruption, and women do not receive the necessary emergency care.22,23 Nevertheless, haemorrhage, infection (including HIV), obstructed labour, HDP and unsafe abortion are estimated together to account for at least two-thirds of all maternal deaths in developing countries.24 Although the available data on maternal mortality from developing countries are incomplete and not completely reliable, there is general agreement that they are the most important issues to tackle for preventing maternal mortality. From this basis, the overview identifies the antecedents in pregnancy and all the diagnostic, prophylactic and therapeutic interventions that might be employed at various stages to detect, prevent or treat the conditions outlined above.

For descriptive purposes, the WHO defines antenatal care as a dichotomous variable, having one or more visits with a trained person during the preg-nancy, or none. However, it may be taken to mean only the care that is routinely provided for all pregnant women at the primary care level, or every aspect of care from screening to intensive life support provided to any woman while pregnant and up to delivery. Because what is provided at the primary or secondary level varies widely even within developing countries and because effective care for one condition may involve a coordinated series of interventions at different levels, this overview does not use primary or referral level to define antenatal care. Primary care level and first referral level are used here to refer to services that are or should be available to all pregnant women, and to services that are or should be provided to women referred because of complications of pregnancy.


Evidence for the 1992 review21 was sought through searches of the published literature, and from unpublished reports provided by the WHO and several research groups working in the field of maternal health. Extensive searches of Medline and Popline on CD-ROM covering from 1982 to 1992 were carried out using as key words all the conditions and interventions considered, as well as more inclusive terms, such as ‘maternal health care’, ‘prenatal care’ and ‘maternal mortality’, to identify original research studies and reviews. The reference collections on maternal health held by the Maternal and Child Epidemiology Unit at the London School of Hygiene and Tropical Medicine (LSHTM) and by the Family Health Division of the WHO, which include many unpublished or unindexed reports, were systematically searched. Several international groups provided extensive reports of health-care and research projects on maternal health in developing countries that they had carried out, supported or evaluated (see Acknowledgements). Bibliographies and reference lists of papers from all these sources were used to identify important earlier writings. The Oxford Data Base of Perinatal Trials (ODPT) was also searched to identify randomised controlled trials (RCT) and systematic reviews of several interventions. These reviews include pooled estimates of the effects of interventions from meta-analyses of methodologically sound trials.25,26

For the present updated review, new search strategies were developed. Sources searched included the following:

• PubMed, which provides access to the PubMed database of bibliographic information, drawn primarily from MEDLINE and PREMEDLINE. In addition, for participating journals that are indexed selectively for MEDLINE, PubMed includes all articles from that journal, and not just those that are included in MEDLINE. The file contains approximately 9 million records that date back to 1966. Coverage is worldwide, but most records are from English-language sources or have English abstracts. The searches were carried out using as key words all the conditions and interventions considered, as well as more inclusive terms, such as ‘maternal health care’, ‘prenatal care’ and ‘maternal mortality’, to identify original research studies and reviews covering the period from 1966 to December 1999.

• The Cochrane Database of Systematic Reviews (CDSR) 2000, disk issue 2.

• The Cochrane Pregnancy and Childbirth Database (CCPC) 1995, disk issue 4.

• The Cochrane Controlled Trials Register (CCTR) 2000, disk issue 2.

Electronic searches of Medline and CCTR were also performed to identify trials that might have been published after the most recent update of the relevant systematic review, and for interventions where no systematic reviews were available.

In many cases, conclusions from studies in industrialised countries can be generalised to treatment of the same conditions in developing countries. For example, a drug found to be effective in controlling blood pressure in severe pre-eclampsia from trials in Europe and North America is likely to have the same effect on pre-eclampsia in Africa. However, other issues may have to be addressed before treatments are transferred, such as drug safety and the level of supervision required. In addition, it may not be possible to generalise the results because of differences in health patterns.

In this overview, the effectiveness of each intervention is considered in terms of what it is intended to achieve. Thus, for a screening test, this may be defined as its ability to discriminate between those with and without the condition in question; in contrast, a treatment’s effectiveness may be judged by its ability to prevent, cure or prevent progression of a condition or reduce case fatality. Where possible, the evidence that any type of intervention reduces mortality from a given cause is assessed. The main emphasis is on biological effectiveness or efficacy. Once the biological effectiveness is demonstrated, the effectiveness of an intervention in normal practice or as part of a programme can be explored, including operational questions of how best to deliver the treatment in the various situations encountered in developing countries.

The reliability of evidence on effectiveness could be ranked according to The Levels of Evidence and Grades of Recommendations.27 The ‘gold standard’ in assessing the effectiveness of any preventive or therapeutic intervention is the systematic review (SR) of RCTs, followed by a single RCT. When properly conducted, random, concealed allocation to a treatment group eliminates the possibility that differences in outcome between treatment groups are caused by systematic differences in underlying risk between the groups. There are treatments with such dramatic and enor-mous benefit that their efficacy is obvious without the need for a formal trial; for example, the introduction of penicillin to treat puerperal fever in the 1940s. However, most if not all interventions in obstetrics have much smaller effects and an RCT is the only way to obtain a true, unbiased estimate of their efficacy.25 However, RCTs have not been carried out on all interventions of interest. Evidence from other types of epidemiological studies on the effectiveness of interventions has been assessed for this review and the evidence extracted is ranked accordingly.

Although numerous assessments of various aspects of antenatal care in developing countries were identified, many of these did not provide any useful information on the effectiveness of the care being provided. There are several reasons for this. Some studies only set out to measure the process (availability, uptake, number of visits, etc.) of the services and not health outcomes. Others measured various aspects of ‘quality’, including the percentage of attendees who received particular tests or treatment, or who were seen before a given point in their pregnancies. While this information is obviously useful for management of services, the value of those services in terms of improved health cannot be assumed until the effects of the individual interventions or the package on the outcome for mother and infant have been demonstrated. The scale of many studies is too small, so they lack the statistical power to show an effect on important outcomes. Adverse maternal outcomes are uncommon events, and results from many centres may need to be pooled to obtain a clear answer. This can only be done when the studies are similar in several aspects and is best achieved by a planned, multicentre study. In many published reports, insufficient information is given to know whether data can be pooled. Unclear objectives and poor study design often mean that only a portion of the information needed to judge effectiveness is collected. For example, several studies of the risk approach report the percentage of women seen for antenatal care who are judged to be at high risk, or the percentage of those delivering in hospital who are at high risk, but they do not give data on outcome for high- and low-risk groups, either in terms of process (such as percentage delivering where recommended) or health (such as percentage with obstructed labour, incidence of morbidity or mortality). Few studies provide adequate details of the catchment population or possible biases in self- or health service selection for attendance at the antenatal clinic or delivery in hospital. Studies that are sufficiently large, carefully designed and executed, with clear objectives, are needed to establish the efficacy of antenatal care. Rigorous peer review of proposals and publications is needed to ensure the quality of future research.

Antenatal care programmes

Rationale and effectiveness

The rationale for antenatal care is that it is essential to screen a predominantly healthy population to detect early signs of, or risk factors for disease, followed by timely intervention. As such, it should fulfil the criteria applied to assess the usefulness of any screening programme. These are as developed from Wilson and Jungner:28

• the outcome in question is an important public health problem in the population;

• there is a detectable preclinical phase or high-risk state preceding this outcome;

• screening tests have useful likelihood ratios (positive and negative) to permit identification of those at risk and those who are not;

• there is an effective treatment available for this early disease stage;

• early intervention is more effective than waiting and treating overt disease.

Maternal mortality is undoubtedly an important public health problem in many developing countries. It is possible to identify the precursors, early signs or risk factors for at least some of the major pathogenic causes of maternal death, such as rising blood pressure, that may proceed to eclampsia. However, evidence that antenatal interventions fulfil the remaining criteria used to assess a screening programme needs careful scrutiny.

In theory, antenatal care might reduce maternal morbidity and mortality directly through the detection and treatment of pregnancy-related or intercurrent illness, or indirectly through the detection of women at increased risk of complications of delivery and ensuring that they deliver in a suitably equipped facility. However, the realisable potential of antenatal interventions to address these problems is unclear for several reasons. Most formal investigations of the effectiveness of antenatal care programmes, whether in developed or developing countries, have concentrated on the effect of care on infant outcomes, perinatal mortality, preterm delivery and low birthweight.

Analyses of historical data in England have shown that the fall in maternal mortality rates since the 1930s can be attributed, at least in part, to obstetric care, and particularly to improved delivery care reducing mortality from infection and haemorrhage.29 Similar conclusions have been reached using data from the USA.12 It is more difficult to show a relationship with the introduction of antenatal clinics or particular interventions in pregnancy, even for the more chronic problems of anaemia and hypertension.6,8

Associations between the availability and use of antenatal services have been shown in various types of epidemiological study. Although the availability, content and quality of antenatal care vary enormously among developing countries, they are generally much lower in countries with high maternal mortality.30 Hospital case series31[32]–33 and confidential enquiries into causes of maternal mortality frequently identify lack of antenatal care as a risk factor.34[35]–36 Case-control studies of maternal death in developing countries also show an association with lack of antenatal care.37,38

Despite the consensus from studies of different designs in favour of antenatal care, reservations about the extent of its true effectiveness remain unknown for several reasons. In places where antenatal care is lacking, delivery services are also likely to be poor and information systems unreliable. Confidential enquiries are intended primarily to identify deficiencies in standards, provision and utilisation of health services that are known to be factors, such as blood transfusion.39 However, these enquiries generally lack control groups, as do the case series, so they cannot be used to establish the effectiveness of a given intervention or service.

In developed countries, comparison of outcomes among women who did and did not receive antenatal care, or who first attended late vs. early in pregnancy have been shown to be confounded by socio-economic factors, education, unwanted pregnancy, maternal age and other factors that influence the outcome of pregnancy.40,41 In developing countries, further confounding factors are likely to be knowledge of, distance from, access to and utilisation of other health services, including those for delivery. No studies have been identified that control adequately for these factors.

Lastly, reports of programmes with no impact are less likely to be published. It appears pointless in the face of all these limitations to attempt to quantify the overall protective effect of an undefined and variable package of tests and treatments known as ‘antenatal care’.

Despite all these caveats, however, there may be real benefit to be had from at least some of the elements of antenatal care, and the absolute scope for benefit may be greater in developing countries where morbidity and mortality are higher. In some developing countries, far more women are seen by health workers during pregnancy than are delivered by a trained attendant.42 This is therefore a chance to reach a larger section of the pregnant population, and it should be put to the most effective use possible. In addition, in highly fertile societies, antenatal visits may afford a chance to reach a large proportion of the whole female population and an opportunity to address other health-related issues. These include awareness of services for, and information about family planning, immunisation, child health, nutrition and sexually transmitted diseases.

The risk-scoring system

Many antenatal care programmes throughout the world are based to some extent on a formal risk-scoring system, in that they attempt to screen the entire pregnant population, and provide surveillance and treatment to individual women or groups of women according to their level of need. In the context of the low risks experienced by most women in developed countries, the rationale is to limit medical interference and give them informed personal choice, while attempting to ensure universal access to high-quality care.

Although the same arguments apply to some extent in developing countries, there are important differences. The risks of childbearing are much higher for women in developing countries, to the extent that so-called ‘low-risk’ groups experience higher mortality than so-called ‘high-risk’ groups in Northern Europe.43[44]–45 In many developing countries, formal delivery facilities or higher-level delivery facilities are not sufficient to cover most of the population. The risk approach is an attempt to allocate existing resources according to a formal risk-scoring system that reflects need rather than according to access or demand.46

The effectiveness of a risk-scoring system, in its narrowest sense, is measured by its ability to identify women who are at high risk and those who are at low risk.

The ability to discriminate between women at high and low risk in all formal risk-scoring systems is poor. Only 10–30% of the women allocated to the high-risk groups actually experienced the adverse outcome for which the formal risk-scoring system predicted them to be at risk.47[48][49][50]–51 Individual factors are poor predictors of risk, probably because they are only markers for groups of women at increased risk rather than direct causes of poor outcome. Attempts have been made to improve discrimination using complex scoring systems that combine large numbers of variables.52,53 The causation of maternal deaths is multifactorial, complex and not fully understood. The relationship with identifiable risk factors such as age, parity or height is indirect, and shows continuous variation across their distribution. Reports on antenatal programmes which refer to the risk approach lack sufficient information on maternal risk factors and outcomes to calculate the positive and negative likelihood ratios of the factors used to identify women who need to deliver in a referral level facility.

However, for this strategy to be effective as a public health strategy, several other conditions must be met:

(a) the whole population must be included in primary screening;

(b) conditions screened for must include the important causes of maternal mortality and morbidity;

(c) when increased risk is detected, appropriate referral or other action must be taken;

(d) adequate services must exist at the referral level;

(e) women at risk must be able to reach the referral level facility and be motivated to do so;

(f) all care providers must be motivated to implement the system;

(g) this strategy must show proof of reducing maternal morbidity and/or mortality.

(a) The whole population must be included in primary screening. In developing countries, reaching the whole population can be difficult, not only because of physical distance but also because those most at risk are often the least likely to use health services. Screening programmes all over the world face a problem sometimes referred to as the ‘inverse care law’. It is often those at highest risk of adverse outcome who are least likely to have access to health services, particularly preventive services.54 Risk factors such as very young age, lack of education and poverty are all associated with low use of health services. This was demonstrated in a study in India, which found that women with higher risk factors were less receptive to antenatal supervision.55 In developing countries, distance from and ignorance of the existence or purpose of health services may compound this. In addition, in some areas, cultural practices may prevent high-risk women, such as very young primigravidae, from revealing their pregnancies or seeking care.56,57

(b) Conditions screened for must include the important causes of maternal mortality and morbidity. Most work on risk-scoring systems for use in antenatal care, both in developed and developing countries, has concentrated on predicting infant outcomes, with maternal health risks regarded as secondary concerns.47,48,50[51]–52,58 In fact, many of the risk factors for poor infant outcomes are also related to maternal risk, such as extremes of maternal age and parity, poor past obstetric history and medical illness. This may be an advantage from the point of view of identifying all the women who need extra care in pregnancy, but risks to mother and infant may interact in complex ways. Therefore, if interventions are undertaken to alleviate the consequences for the infant, without being aware of the maternal risk, or the other way round, then problems for both may be increased.

(c) When increased risk is detected appropriate referral or other action must be taken. Simple identification of women with increased risk does not always lead to the required action, sometimes because of the failure of health workers to recognise the importance of the risk, but also because facilities are inadequate or inaccessible, or women are unconvinced of the need to use them or of their effectiveness. The performance of programmes based on the risk approach has come under question, particularly where resources are most scarce, and there is some concern that this approach may divert attention away from improving services for all women.59 Not only must appropriate action be taken when increased risk is detected, but these interventions must also be timely to be effective. In general, formal risk-scoring systems perform better at later stages of pregnancy, when little can be done to improve the diagnosed condition.

(d) There may be real scope for benefit from the risk approach when it is effectively carried out and backed up by adequate referral services. Identifying risk will not help women if the means to prevent or treat problems are not available or if they are inaccessible. In Senegal’s Kaolak Hospital, transfusion was not available for 80% of the women referred in labour who needed it, and anaesthesia was similarly not available for 64% of those in need.60 In the long run, this sort of failure of services can be very prejudicial to the implementation of the referral system. Indeed, women may refuse transfer to a distant referral centre if their experience is that those transferred have poor outcomes. This is a problem in any system based on high risk, because women referred are at higher risk of dying than are those not referred. However, it may be compounded where services are inadequate.

(e) Women at risk must be able to reach the referral level facility and be motivated to do so. One study in three clinics in Maputo, Mozambique (one each urban, peri-urban and rural) found that, although many women were successfully assessed, referred and cared for, less than half of those in the high-risk group were actually seen at the referral centre. Furthermore, only 11% of women referred from the rural clinic for delivery in hospital actually delivered there, compared with 66% of those referred from the other two clinics.61 Distance appeared to be a stronger determinant of place of delivery than risk screening. In a 4-year community-based study of pregnancy and delivery in a rural area of Kenya,62 84% of women attended an antenatal clinic at least once but only 27% delivered in hospital. In fact, only just over one-third of women who had intended to deliver in hospital actually did so, and distance from the hospital town and the rainy season were shown to have strong negative effects on the percentage of women delivering in hospital.

However, in this study, delivering in hospital was related to primiparity, height less than 150 cm, history of a serious complication in a previous delivery, or to a previous hospital delivery, which suggests that some degree of risk selection did occur. Of the 67 women with a previous caesarean section, 73% delivered in hospital. The only high-risk factors associated with a reduced likelihood of delivering in hospital were grand multiparity and advanced age. The authors attributed this to older women’s personal experience of successful childbirth at home. The study is too small to assess any effect on maternal or perinatal mortality. Questioned while pregnant, women themselves gave safety and the availability of doctor, midwife and modern medicines as their reasons for choosing hospital delivery. Those who intended to deliver at home did so because of a lack of transport or money, or because they had previously experienced uncomplicated deliveries.

There are two additional studies showing poor referral compliance and their causes. In a descriptive survey from eastern Zaire, the referral success rate was only 33%, despite strong community participation, close links among health centres and hospital, and the absence of financial barriers. This lack of accomplishment was attributed mainly to the hospital inaccessibility and women’s own perception of their risk status.63 The remaining study showed a very weak referral system with low compliance, mainly as a result of hospital fees, irregular transport, poor roads, lack of drugs and negative staff attitudes.64

One possible strategy for overcoming poor access to delivery services because of distance or transport problems is the maternity waiting home or village near a hospital or clinic, where women can live during the last weeks of pregnancy while they await delivery. By their nature, these serve isolated rural communities. A narrative review, describing examples of maternal waiting homes in various countries around the world, and discussing issues about implementation and providing an analytical framework for planning, management and evaluation of this service intervention has been published.65 This intervention seems to be effective in reducing maternal and perinatal mortality. However, it is clear that a more rigorous evaluation of its effectiveness is needed. Other articles have addressed this issue,66[67]–68 showing that maternity waiting homes are an effective policy option to improve both women’s access to obstetric services and maternal and perinatal outcomes; Wilson et al.69 showed that, before its implementation, users should be asked to identify potential local problems. However, none of the authors has assessed reliably the effectiveness of maternal waiting homes.

(f) All care providers involved in the formal risk-scoring tasks must be trained and motivated to implement the system. In settings as different as Java,70 Burkina Faso71 and the UK,72,73 there is evidence of frequent failure to elicit important information on risk factors – particularly those based on past obstetric and medical history – and of failure to take appropriate action when evidence of increased risk is elicited. In Java, at least one in four women who have experienced previous stillbirths or child deaths were not noted by TBAs. In the UK, only 32% of the women with a previous history of delivering infants light for gestational age and 15% of all cases with previous postpartum haemorrhage were noted by medical staff.

The performance of the wide variety of checklists and antenatal record cards designed to improve the recognition of important risk factors has rarely been evaluated. In contrast, antenatal records kept by the woman herself rather than by the health worker or institution can improve the transfer of information between levels and ensure that recorded information is available when it is needed. Home-based maternity records have been developed for use in many communities, in various languages and pictorial forms.74[75][76][77]–78 Evidence shows that women lose their records less often than do institutions both in developed79 and developing countries.75 An additional advantage of these records is that they have space for information on consecutive pregnancies and intervening periods, so that women and their health may be considered in full rather than as ‘index pregnancies’.

Few rigorous evaluations of antenatal care programmes based on the risk approach have been published. One study of TBAs working within the health services in Fortaleza, Brazil, showed that they referred 97% of 117 women in their care who had an antenatal complication (pre-eclampsia, eclampsia, haemorrhage or premature rupture of membranes). This accounted for half the total referrals, the other half being for problems arising in labour. No information is given on how many were referred for investigation or treatment before labour. There were no maternal deaths in this sample and no further information on maternal outcomes.80 In this setting, TBAs seemed to be able to identify women at high risk of perinatal loss, and to act appropriately. However, they were working in shifts within the formal health services, in special delivery premises in their communities, visited by a nurse once a week, with an ambulance always available and not more than 90 minutes’ journey from the referral hospital. It would be difficult to achieve such a high standard in more isolated areas, with workers less closely integrated into the health services.

In theory, an objective formal risk-scoring system would be more efficient than a subjective clinical impression of risk. The validity and usefulness of the formal risk-scoring system remain to be determined. For a reliable assessment of the effectiveness of formal risk scoring, a large and methodologically sound RCT comparing it with routine care would be needed.

Number of antenatal visits and type of care provider

Antenatal care as it is currently delivered was originated from European models developed in the early 20th century. Most of the activities remain unchanged or have evolved since that time. As new technologies, tests and interventions became available, they were often added to the programme without proper scientific evaluation of their effectiveness. More visits were added, reaching a total of 16 in some countries. The number, timing and content of antenatal visits appear to be more a matter of ritual than evidence-based health care.

In recent years, attention has been directed to more rigorous evaluation through RCTs of the number, timing and content of antenatal visits, and the type of care provider. A systematic review of such trials shows that there are two trials from one developing country, examining reductions in the number of visits.81 The largest reduction in the number of visits, from six visits in the control group to four visits in the intervention group, was achieved in the Munjanja trial performed in Zimbabwe.82 The results of the systematic review show that there were no differences in the maternal biological outcomes, such as caesarean section, antepartum and postpartum haemorrhage, or maternal mortality. However, there was a difference favouring the intervention arm in lowering the incidence of pre-eclampsia. This lower incidence of pre-eclampsia could be attributed to a lower detection rate, rather than to an intervention effect itself. Perinatal outcomes, such as preterm delivery, low birthweight and small for gestational age, were similar in the intervention and control groups. Thus, the review concludes that, in developed countries with well-established obstetrics services, a reduction in the number of goal-oriented visits could be implemented without risking the health of the mother and her infant. The recently finished WHO Antenatal Care Trial was the largest RCT to date, with 24 000 women, evaluating a new antenatal care programme that consisted of interventions scientifically demonstrated to be effective in improving maternal and perinatal outcomes. The results of this trial will soon be available.83

The review also evaluated the effectiveness of routine antenatal care provided by midwives/general practitioners compared with care led by obstetricians/gynaecologists. Overall, outcomes were similar in both groups but, in midwife/general practitioner care, the rates of pregnancy-induced hypertension and pre-eclampsia were consistently lower than in obstetrician-led shared care. However, a lack of early recognition of fetal malpresentation was observed among the midwife/general practitioner group.81 This could be the result of missed diagnoses during antenatal visits. No detrimental effects were observed in perinatal outcomes.

Interventions in pregnancy related to major causes of maternal morbidity and mortality


Obstetric haemorrhage is one of the leading causes of pregnancy- and childbirth-related death throughout the world. Historically and geographically, as the overall level of maternal mortality increases, so does the proportion of these deaths caused by haemorrhage. Haemorrhage is usually an acute event, which can occur at any stage in pregnancy, delivery or the puerperium. The prevention of mortality from haemorrhage mainly depends on prompt treatment of its cause to prevent further bleeding, and replacement of blood loss to maintain the circulation. These, in turn, depend on access to medical, surgical, intravenous fluid replacement and blood transfusion services, as outlined in Essential Elements of Obstetric Care at First Referral Level.84

Postpartum haemorrhage is one of the major causes of maternal death, accounting for around one-third of all such deaths.85 It can be prevented by active management of the third stage of labour, including prophylactic use of oxytocics86,87 and umbilical vein injection of oxytocin for the management of retained placenta.88 Operational research on ways to improve third-stage management by all types of birth attendant in home and institutional deliveries is needed.

The role of antenatal care in preventing or ameliorating the effects of haemorrhage is limited. There are several means by which interventions in the antenatal period might theoretically influence the risk of death from haemorrhage.

• The first involves the detection of those women at increased risk of serious bleeding in labour and ensuring that they deliver in a facility that is adequately equipped.

• The second concerns the detection of adverse symptoms and signs that develop during pregnancy, and referral for prompt investigation and treatment.

• The third involves a reduction in the prevalence of anaemia, so that those women have a greater haematological reserve to withstand blood loss.

Risk assessment and referral

Identifying women at increased risk of haemorrhage will clearly only be of benefit if adequate delivery services, including active management of the third stage, umbilical vein injection for retained placenta, blood transfusion and operative facilities, are available. No direct assessment of the effectiveness of risk screening in preventing death from haemorrhage in developing countries has been identified for this review. Several risk factors for postpartum haemorrhage (PPH) are known: history of haemorrhage in a previous delivery, multiple pregnancy and grand multiparity. A history of previous PPH was found to be associated with a relative risk of 1.6 of recurrence, but only 6.3% of those with such a history suffered PPH in the index delivery.72

This study took place in Aberdeen, where active management of the third stage is the rule, and the positive predictive value might be higher where this is not the case. In Papua New Guinea, 49% of women with a previous third-stage complication recorded ante-natally went on to experience third-stage complications again.89 However, this study could be subject to considerable recall bias. Strikingly, clinicians in Scotland apparently only noted this risk factor in 15% of those in whom midwives recorded it. Hall et al.90 studied 36 312 women in Aberdeen and have pointed out that an increased risk of PPH was seen among primiparae women and with the induction of labour. Also, there is a clear tendency for PPH and retained placenta recurrence in later vaginal deliveries. Thus, women with previous history of third-stage complications should be referred to services where facilities such as trained obstetricians, anaesthesia and blood transfusions are available.

A case-control study of risk factors for PPH showed that prolonged third stage, episiotomy, arrest of labour, operative delivery and soft tissue trauma have a significant association with PPH.91 None of these could be predicted from antenatal care strategies. However, other risk factors for PPH, such as pre-eclampsia, previous PPH, multiple gestation and nulliparity, could be predicted.

Disseminated intravascular coagulation secondary to prolonged undelivered intrauterine death and clotting disorders related to pre-eclampsia or eclampsia may result in catastrophic bleeding at delivery. Women with HDP or who experience intrauterine fetal death should, in any case, be referred for hospital care.

Diagnosis and treatment

The effectiveness of interventions for haemorrhage depends on immediate recognition of the seriousness of the signs and symptoms; availability of first aid; and knowledge of and access to definitive treatment. The potential roles of antenatal care are twofold:

• enquiry about symptoms and signs of conditions such as placenta praevia and placental abruption, as well as the conditions that precede or predispose to them, at antenatal visits followed by appropriate action;

• educating pregnant women and the community to recognise the seriousness of vaginal bleeding and abdominal pain in late pregnancy, and to act immediately should these symptoms arise.

Vaginal bleeding at any time, particularly in late pregnancy, is an alarming symptom and one expected to lead to immediate efforts to seek care. However, it is not clear whether vaginal bleeding is seen as dangerous in all cultures, or as a reason to seek medical help. Traditional interpretations of its significance and traditional remedies may delay women from presenting to health workers. Anthropological research is needed to clarify this. However, it is easy to attribute to cultural barriers what are often failures to provide adequate services or failure to make them accessible. Many of the antenatal records developed for use in developing countries (‘action-oriented antenatal card’) include spaces to enter information about bleeding in pregnancy, and instructions to refer the woman immediately to the nearest health centre or hospital. These cards are often in pictorial form, or combine written and pictorial elements. No studies of the effectiveness of the cards in reducing morbidity or mortality have been identified.


The haemoglobin concentration falls slightly in mid-pregnancy as a result of physiological haemodilution, so that a greater proportion of pregnant than nonpregnant women will have a haemoglobin level below any given cutoff point. No untoward consequences of this relative reduction of haemoglobin on mother or fetus can be discerned in well-nourished women. Indeed, failure of haemodilution and high haemoglobin levels throughout pregnancy are associated with poor fetal outcome, probably because they reflect a lack of physiological adaptation to pregnancy.92

There are several interventions, such as iron, folate and iron with folate supplementation, that do prevent or reduce the fall in haemoglobin concentration and reduce the proportion of women whose haemoglobin level is below 10 g/dL in late pregnancy.93[94]–95 Routine iron supplementation improves the haematological indices in women receiving it and the results of the largest trial suggest that such supplementation may reduce the need for blood transfusion.96 This protective effect could be extremely beneficial in countries where HIV seropositivity is high and screening of blood donors is not rigorously carried out. Routine iron supplementation should be implemented in populations where iron deficiency is common. In places where megaloblastic anaemia in pregnancy is common, routine supplementation with folate or with folate plus iron is effective in improving anaemia.

In areas where maternal mortality is high, the prevalence of anaemia is also much higher than elsewhere. It is estimated that over one-half of women of childbearing age in Africa have haemoglobin levels below 11 g/dL.97 In addition to its morbidity, anaemia may contribute indirectly to mortality associated with haemorrhage and directly through heart failure. Most cases of anaemia are thought to be caused by iron deficiency, and there is probably an even larger pool of women who, although not frankly anaemic, have insufficient iron stores for the needs of pregnancy, delivery and lactation.98 In some areas, haemoglobinopathy may be an important contributor to morbidity and mortality from anaemia. Paradoxically, as health services for children improve, more girls with homozygous haemoglobinopathy are likely to survive to adulthood and need special care during pregnancy.99 Patterns of local prevalence will determine the need for screening for homozygous or heterozygous disease.

Malaria may also contribute to the development of profound anaemia through haemolysis and, in many areas, is its major cause. Susceptibility to malaria is increased in pregnancy, particularly in primigravidae. Therefore, prophylaxis, investigation and treatment of malaria are important elements of care for all pregnant women in endemic and epidemic areas. Garner and Gulmezoglu100 have reviewed the literature systematically and addressed some interesting issues in malarial infection during pregnancy. In intervention programmes for all pregnant women, routine antimalarial drug interventions have been shown to reduce the incidence of fever but little overall effect has been seen in other parameters measured. Effects in primigravidae appear to be more marked than in multigravidae. In intervention programmes for primigravidae, measures to prevent or routinely treat malarial infection have been shown to reduce antenatal parasitaemia, and prevent the prevalence of severe anaemia at 34 weeks of gestation. There is no direct evidence that measures to prevent or routinely treat malaria infection presumptively during pregnancy result in lower mortality in the mother or infant compared with prompt treatment of symptomatic illness or anaemia in the antenatal period. There is insufficient evidence specific to pregnant women to establish whether the use of insecticide-impregnated or unimpregnated bednets during pregnancy is of practical benefit.

Clearly, pregnant women who live in malaria-endemic areas need access to services that can provide treatment for illness episodes and anaemia. Therefore, it is important that policy-makers establish these services before considering investment in secondary preventive programmes.

Presumptive treatment programmes, such as the sulphadoxine–pyrimethamine regimen from the Shulman study,101 are feasible and practical. The decision on making these programmes routine policy depends on whether or not preventing severe anaemia at 34 weeks will be maintained to term, and translated into health benefits for the mother and fetus. The available evidence shows that, for perinatal death in primigravidae, the results are not statistically significant, but are a reminder that the potential impact of this intervention on mortality could be high. Therefore, we need a large trial, using a placebo-controlled design, of the effects of drug interventions on pregnancy outcome and mortality in the neonate. A factorial design could also be used to examine the impact of routine iron supplementation in malarial areas. Local patterns of prevalence should determine policy on screening for malaria and other parasitic diseases which can be the underlying cause of anaemia. Reliable data on parasitic disease, although often better than data on maternal health, may not be available. Where there is doubt, small-scale prevalence studies are then needed as a guide to practice.


Even if routine iron and folate supplementation is available, screening for moderate or severe anaemia is still needed, since women suffering from more than just mild anaemia need additional investigation and treatment. The successful management of anaemia during pregnancy should be based on a reliable method for diagnosis and surveillance of the response to the adequate treatment. Unfortunately, at present, for many women who live in rural areas, such diagnosis is only carried out by inspection of conjunctivae and mucous membranes, if at all.

There are two major types of diagnostic technique: screening tests, which indicate whether the haemoglobin is above or below a given threshold; and tests to assess the exact haemoglobin level. The likelihood positive ratio of clinical diagnosis of anaemia using subjective interpretation of the inspection of skin, nail beds, conjunctivae, lips, tongue and palms is poor, so these tests are not useful.102[103][104][105][106][107][108][109][110][111][112]–113 Qualitative techniques with cutoff values are invasive and involve blood samples being taken. The most useful test method seems to be copper sulphate densitometry, which has quite a good predictive value for anaemia.114 Other methods, such as capillary microhaematocrit measurement, although promising, require further research.115,116

Prevention and treatment

Routine prophylactic administration of iron to all pregnant women may prevent the development of anaemia in many women with frank or borderline iron deficiency, or may correct mild anaemia in many others. The potential for benefit depends on the actual prevalence of iron deficiency and that of iron-deficiency anaemia in the population. The levels of anaemia or mean haemoglobin concentration in the population, at which universal supplementation would be beneficial are not clear, partly because the level at which detriment to the mother and/or fetus takes place is also uncertain.

Haemoglobin levels can rise 0.4–0.7 g/dL per week on a dose of 120 mg ferrous salt with 5 mg folate, so that moderate iron-deficiency anaemia may theoretically be corrected by oral therapy in women attending for antenatal care in the mid-trimester. However, orally administered iron causes nausea and constipation, and the side-effects appear to be related to the amount absorbed.98 Not all studies demonstrate such marked improvement with oral therapy. However, it is not clear whether the differences in outcome are the result of variations in absorption, perhaps related to other dietary factors, nutritional status or disease; of noncompliance with treatment because of side-effects or cultural factors; or of failure of health services to deliver treatment or motivate women.117 Intravenous or intramuscular iron is also proven to be effective in resolving iron deficiency and leading to correction of anaemia.98 This avoids the problems of compliance and gastrointestinal side-effects, but requires facilities and skills for intramuscular injection or intravenous infusion and for emergency treatment of allergic reactions.

The realisable benefit will depend on dosage, coverage and compliance. The first essential step is to ensure that adequate supplies of oral iron are actually provided to pregnant women at appropriate times. Lack of supplies, poor access because of distance or limited opening times, artificial constraints on the quantities given per visit and other operational difficulties may make routine supplementation programmes ineffective. Even if supplements are provided, women must be motivated to take them.

Hypertensive disorders of pregnancy

Difficulties are encountered in comparing rates and studies of HDP (including pregnancy-induced hypertension, pre-eclampsia and eclampsia), because of inconsistencies in definition and in detection.118[119]–120 However, throughout the world, they constitute one of the leading causes of maternal and perinatal death. In areas with high maternal mortality, the proportion of all maternal deaths that are the result of hypertensive disorders is relatively low compared with the large number of deaths from haemorrhage and infection. Nevertheless, the absolute risk of death from HDP is high in these areas, because of higher incidence and higher case fatality. Moreover, mortality from HDP is not as responsive to improvements in basic delivery care as is mortality from the other main causes. Duley121 has pointed out that the maternal mortality ratios for HDP are remarkably similar in Africa, parts of Asia and Latin America, despite large differences in the all-cause maternal mortality ratios.

There are theoretical points at which interventions during pregnancy might improve maternal outcome from HDP. These include primary prevention, detection of increased risk, and early detection of any stage of an HDP with secondary prevention of progression by treatment at the primary level or referral for expert care. It must be recognised, however, that not all cases of HDP follow an orderly progression from mild to severe disease, and that women may be found to be suffering from any stage of disease, including eclampsia, without having apparently passed through the preceding stages. Saudan et al.122 have shown that about 15–25% of women initially diagnosed with pregnancy-induced hypertension developed pre-eclampsia, and this probability is higher with earlier presentation or if the woman has had a prior miscarriage. Among women with pregnancy-induced hypertension detected after 36 weeks of gestation, the risk of developing pre-eclampsia was about 10%. The natural history of HDP is not fully understood and the relative importance of the degree of hypertension, proteinuria, oedema or biochemical abnormalities as indicators of severity of disease or prognosis is unclear.123 However, gestational hypertension identifies women with a higher rate of adverse pregnancy outcomes than is found among the general population.124 Women with pre-eclampsia are at higher risk of severe maternal disease, preterm birth and small-for-gestational-age babies.

Risk factors

The geographic variation in incidence suggests the possibility of prevention, if causes amenable to change can be identified.121,123,125,126 Primiparous women are twice as likely to develop pre-eclampsia as are multiparae,127 and this risk is particularly high at extremes of age. Women who have suffered proteinuric pre-eclampsia in their first pregnancy are at increased risk in their second pregnancy. This risk is also higher when severe pre-eclampsia occurs in the first pregnancy.128 The level of risk is also higher in those with a positive family history, obesity, multiple gestation or excessive weight gain in pregnancy. However, none of these factors alone or in combination can be used confidently to predict women who will develop HDP.127 Despite frequent suggestions to the contrary, the incidence of HDP shows no consistent pattern with socio-economic class or any dietary factor, including salt intake.127,129,130 However, an inverse relationship between calcium intake and HDP has been described by Belizan and Villar.131 This observation was supported by other epidemiological and clinical studies132[133]–134 and led to the hypothesis that an increase in calcium intake during pregnancy might reduce the incidence of hypertensive disorders among women with low calcium intake.

Early detection and diagnosis

The wide variation in case fatality suggests that differences in care can improve outcome. In African countries, the case fatality from eclampsia is estimated at between 7% and 25%, whereas it is only 1.4% in Sweden.121,126 There is epidemiological evidence that improved detection and care for women with HDP has improved maternal outcomes, but there is little clear evidence of how or what specific treatments are effective. Redman123 and Douglas and Redman135 point to the fact that most eclamptic fits in the UK still occur in hospital as evidence that early detection and treatment are not effective at preventing disease progression. However, this could also be taken to show, at least for those that do not occur postpartum, that those women at risk of eclampsia are at least detected and admitted to hospital, where supportive care is available and delivery can be expedited, so reducing mortality. Analysis of historical data for Sweden suggests that the early fall in mortality from eclampsia seen there was primarily the result of improved case survival, i.e. the result of better treatment of advanced disease, whereas the continued fall over the past two decades is accounted for by a falling incidence of eclampsia.136

There is a strong sense, backed up by data from confidential enquiries around the world, that women who die of HDP have usually received substandard pregnancy care, including failure to diagnose the condition until comparatively late or failure to act on the diagnosis promptly.36,137 In Jamaica, where HDP was the commonest cause of maternal death, accounting for 31% of the total, failure to recognise the severity of the condition, to refer to a hospital, to start drug treatment or to expedite delivery were all noted as avoidable factors in a confidential enquiry.35,36 However, alternative theories are that the term ‘pre-eclampsia’ is misleading because eclampsia can precede pre-eclampsia. Supporting this, Douglas and Redman135 have shown that proteinuria was the only premonitory sign in 10% of the cases of eclampsia and that one-third of the women had only mild hypertension. Furthermore, strengthening this theory, convulsions occurred despite antenatal care and within one week of the woman’s last antenatal visit. Finally, more than one-third of the women developed seizures without either hypertension or proteinuria being documented, although they may have been present.

Roughly, half the deaths associated with HDP in the UK are related to pre-eclampsia and half to eclampsia, although the incidence of pre-eclampsia is estimated to be 100 times that of eclampsia.123 In developing countries, the great majority of fatalities are due to eclampsia,35,121,126,138,139 and the percentage thus incurred is higher where antenatal coverage is lower. This is often taken as evidence of a higher probability of progression from pre-eclampsia to eclampsia in the absence of antenatal care.121 However, the observed differences in proportional mortality from pre-eclampsia and eclampsia may be subject to selection and/or detection bias. For instance, selection bias can affect hospital-based studies, because women who fit are more likely to be admitted to hospital than are those who do not. Even population-based studies that trace all maternal deaths in the community may be subject to detection bias in ascribing cause of death. Fits are a striking and memorable occurrence, easily reported by a lay informant, and generally taken as pathognomonic of eclampsia in a pregnant or recently delivered woman who subsequently dies. A retrospective diagnosis of pre-eclampsia is more difficult to establish post mortem in the absence of antenatal recordings of hypertension, proteinuria or oedema. Carefully conducted case-control studies of women dying from (compared with women surviving) severe HDP could give estimates of the effectiveness of early detection and clues as to which elements of care are worth investigating in large-scale trials. It is difficult to separate the effects that a lack of access to early detection of HDP and to referral level services for treatment and delivery have on mortality or morbidity from HDP.

Early detection of hypertension and proteinuria is possible with relatively simple instruments. Measurement of arterial blood pressure is the most sensitive screening test for diagnosing HDP, but not all women with arterial hypertension in pregnancy have or develop proteinuric pre-eclampsia, and some women present with sudden onset of eclampsia but little preceding hypertension. In a retrospective record study in a routine antenatal clinic in Scotland, antenatal measurement of blood pressure had a likelihood positive ratio of 14.2 and likelihood negative ratio of 0.30 for pre-eclampsia during pregnancy, labour or puerperium.140 A review141 and additional studies142,143 have shown the poor ability of the tests developed to predict HDP. Furthermore, the performance of a test, which must be repeated at intervals, will depend on the interval chosen and the speed with which the condition appears.

By definition, pregnancy-induced hypertension is hypertension that occurs after 20 weeks of gestation in a previously normotensive woman, implying the need for at least one early or prepregnancy baseline measurement to differentiate this condition clearly from chronic hypertension in a pregnant woman. Blood pressure tends to fall in mid-pregnancy and rise again in the last trimester.144 In addition, it is not entirely clear whether an absolute level of diastolic pressure higher than 90 mmHg, a rise in diastolic pressure of more than 20 mmHg or a combination of the two is the best diagnostic and prognostic sign.127 However, it seems that the definition of gestational hypertension (systolic blood pressure of at least 140 mmHg with a rise of at least 30 mmHg, or a diastolic blood pressure of at least 90 mmHg with a rise of at least 15 mmHg on two or more occasions after 20 weeks of gestation) and pre-eclampsia (gestational hypertension plus proteinuria of at least 2+ or 1 g/L on dipstick or a 24-h urinary protein excretion more than 0.3 g) identify women at higher risk of pregnancy complications.124

In interpreting the effect of many screening tests, length bias can present a problem: diseases that progress more slowly and have a better prognosis are more likely to be picked up by screening than is a more acute disease with severe prognosis, and this may give a falsely optimistic view of the effect of screening on outcome (for example, screening for breast cancer). The natural history of pregnancy-induced hypertension is not well enough understood to assess this properly. On the one hand, it rarely appears before 28 weeks but, when it does, the prognosis is worse, so that more serious cases are picked up sooner by earlier measurement. For this reason, Wallenburg127 recommends monthly blood pressure measurement throughout the second trimes-ter. On the other hand, severe pre-eclampsia or eclampsia can develop very suddenly, and the relative contribution of this to the severity of disease or poor prognosis in those diagnosed late is not clear. The minimum number and timing of measurements to detect most cases reliably and early is not yet known.

In many developing countries, blood pressure meas-urement is not available to all women at the primary level. Requirements in terms of equipment and skills have been set out.145,146 Widening the availability of blood pressure measurement at the primary care level could be tackled using two approaches: the development of appropriate technology and operational research. Training health-care workers to take blood pressure measurements in a reliable, unbiased fashion can be difficult anywhere in the world.144 Routine recording of blood pressure is inaccurate and systematic error of about 10–15 mmHg would be expected, as a result of failure in standardisation of the method.147 Terminal digit preference is a well-documented source of systematic error,148[149]–150 so potentially reducing the diagnostic ability of blood pressure measurement.151

Another source of error in blood pressure measurement is the instrument used to perform the measurement. The manual mercury sphygmomanometer is the instrument of choice to date, because it is accurate, reliable, cheap and readily available in most clinics, and standards for calibration are well established. Aneroid devices present problems with calibration, so inaccuracy is an issue to bear in mind when they are used.152 There is good evidence to support the use of the fifth Korotkoff sound rather than the fourth. It is easier to teach health-care workers to recognise the fifth sound, and it is more reliable.153[154][155]–156 The use of automated instruments, although controversial, is increasing. Validation studies of automated blood pressure monitors in comparison with the gold standard mercury sphygmomanometer have given promising results.157[158]–159 However, other studies have demonstrated that blood pressure measurements with automated instruments have been consistently lower than the manual sphygmomanometer measurements.160,161

Although the classification of HDP is controversial and is an obstacle for developing a uniform message across the literature, there is little doubt that proteinuria is widely acknowledged as a reliable prognostic factor for maternal and perinatal outcomes.162 Thus, the accuracy of measuring proteinuria is a crucial point in the management of HDP. A very small percentage of women who develop pre-eclampsia may exhibit proteinuria before the rise in blood pressure, but it is not entirely clear whether or not universal urinalysis at every antenatal visit in the absence of hypertension is worthwhile. Accurate chemical tests and reagent strips are available relatively cheaply for the detection and quantification of albuminuria. However, reliable tests are not universally available or used at primary level. In recent years, there has been an increased tendency towards using dipstick urinanalysis instead of biochemical measurement of total protein excretion in a 24-h period. This could be seen as a backward step, in view of pitfalls of using dipstick urinalysis in comparison with the 24-h urine collection method.162[163][164][165]–166 If dipsticks are used for screening, then proteinuria in HDP should always be confirmed in a 24-h quantitative protein urine collection. Research into appropriate technologies and operational research on how best to provide an accessible screening system are needed. In some areas, primary health-care workers will have to refer women with positive dipstick results for investigation at the first referral level.

The difficulties of ensuring universal blood pressure estimation prompted a study of the accuracy of prediction and diagnosis of pre-eclampsia from the presence of oedema or proteinuria, or a combination of both.167 Unfortunately, even using either oedema or proteinuria, the detection rate of women who developed antenatal diastolic hypertension or of those who developed eclampsia was low. Detection of pre-eclampsia was better but also low. In view of these results, the detection of oedema and proteinuria should not be advocated as an alternative primary screening test to blood pressure measurement, although it may have a place as an interim measure in some areas until universal blood pressure screening can be provided.

Dependent oedema is common in normal pregnancy, but generalised oedema has been advocated as a sign of pre-eclampsia. Examination of all pregnant women to detect oedema alone at the primary health-care level is often recommended, particularly in settings where blood pressure measurement is not available. Examination for oedema of hands, face and sacrum does not require any equipment, and only modest skill. However, it should be taken into account that this strategy is limited by the fact that oedema of the face or hand or both was reported in 64% of normotensive women in the last trimester of gestation.168 Furthermore, significant oedema of the face and hands occurred in 30% of normotensive women, while up to 40% of women with eclampsia exhibited no oedema before the onset of convulsions.169 Thus, the potential effectiveness of this strategy has not been demonstrated, and it seems very limited, in view of the results found in the WHO collaborative study of proteinuria and oedema discussed above.167,170


Three interventions to prevent pre-eclampsia and its sequelae – calcium supplementation, fish oil and antiplatelet agents – have been subject to numerous trials. The results of systematic reviews of the evidence to date on these interventions are now outlined.

Calcium supplementation during pregnancy is associated with reduced hypertension, reduced pre-eclampsia (especially for those groups at high risk), low baseline dietary calcium intake, reduced low birthweight, reduced preterm delivery for those women at high risk of hypertension and reduced childhood hypertension in the only study that measured this outcome.171 The largest single trial to date did not show any beneficial effect of calcium supplementation on outcomes.172 However, this trial included women with low risk of hypertension and adequate dietary calcium intake, with all women receiving low-dose calcium supplementation during pregnancy. Current evidence supports calcium supplementation for women at high risk of gestational hypertension in communities with low dietary calcium intake. Further RCTs focusing on women at high risk of gestational hypertension and/or low dietary calcium intake are warranted. The long-term follow-up of the children born to the mothers participating in the studies deserves more attention to confirm or refute the findings of one of the trials.173

In a systematic review by Duley,174 it was concluded that fish-oil supplementation during pregnancy had shown a promising reduction in the risk of pre-eclampsia. However, in view of the methodo-logical limitations of the studies included, no recommendations were made about its use until the results of methodologically sound trials are available. Details of two additional trials have been published,175,176 neither of which demonstrates any differences in the proportion of women with hypertension in the fish-oil group in relation to the women in the control group. More recently, four prophylactic and two therapeutic RCTs of fish oil failed to show any improvement on the incidence of pregnancy-induced hypertension.177 Based on current evidence, fish oil is not recommended for routine use during pregnancy.

Data from 16 trials of low-dose aspirin and/or other antiplatelet agents to prevent pre-eclampsia and intrauterine growth retardation were reviewed by Collins.178 At that time, the conclusions reached were that, in view of the findings, the routine use of aspirin is not supported and results from large trials should be awaited in order to have more reliable evidence. These studies are now available and the results have reliably shown that routine use of low-dose aspirin is marginally effective in reducing the incidence of pre-eclampsia.179[180][181]–182 It is also worth pointing out that the use of low-dose aspirin has not demonstrated any bene-ficial effects in high-risk populations or in perinatal outcomes such as intrauterine growth restriction.180

The latest systematic review of 42 trials of antiplatelet agents (mainly low-dose aspirin) to prevent pre-eclampsia concluded that there was a small-to-moderate benefit.183 However, the authors pointed out that more information was still needed on which women should be treated, what dose should be used and when treatment should start.

There are other nutritional supplements that have been advocated as likely to be beneficial for pre-eclampsia. Prophylactic magnesium supplementation trials did not exhibit any benefit in preventing pre-eclampsia.184,185 RCTs of antioxidants have shown encouraging results in reducing pre-eclampsia but additional studies are needed before they can be introduced in clinical practice.186,187


Women with pre-eclampsia and eclampsia appear to experience better outcomes when they have access to and use professional care. In the past 50 years, during which mortality rates from HDP have fallen dra-matically in developed countries, treatments have been introduced and others dropped, untested by trials, and with unknown effects on the case fatality. The treatments advocated have included heavy sedation with barbiturates, enforced rest in a darkened room and prophylactic forceps delivery under sedation or general anaesthesia. Over the same period, in developed countries, general supportive measures have improved, including care and detection of complications such as clotting disorders. In addition, more effective drugs for inducing labour to expedite delivery have been developed, and the safety of anaesthesia and caesarean section has increased. The contribution of these advances to the reduced mortality from HDP is uncertain.

The treatment of established HDP – potentially life-threatening conditions – is not part of routine antenatal care by primary care workers. Women must be referred to the first level at which they can receive specialist professional care. Debate on the treatment of HDP still centres on the role of rest, antihypertensives and anticonvulsants.

Rest. Rest, either in hospital or at home, has a long history in the treatment of mild pregnancy-induced hypertension and pre-eclampsia, and has been considered by different authors. Goldenberg et al.188 summarised the existing data in 1994 and showed that bed rest was used in about 20% of all pregnancies to prevent or treat a wide variety of conditions, including chronic hypertension and pre-eclampsia, at substantial costs but with little evidence of its effectiveness. They recommended curtailment of its use unless RCTs demonstrate an improvement in a specific outcome. Systematic reviews have consistently shown that there is no evidence that bed rest improves health outcomes of pregnant women with proteinuric or nonproteinuric HDP.188[189][190]–191

Two trials of strict bed rest for proteinuric hypertension in hospital compared with hospitalisation with no restriction on movement actually appear to show an increased risk of fulminating pre-eclampsia in the strict bed-rest groups, but the numbers are small and the effect could be caused by chance.190 Questions remain about how effective rest at home or in hospital is in treating mild, pregnancy-induced hypertension. Large RCTs are needed to answer these questions reliably. Evidence from developed countries may not be directly applicable to developing countries where women are often involved in heavy physical work for much of the day and where some kind of rest at home may not be possible. However, the information to judge the effectiveness of rest for treating pre-eclampsia in developing countries does not yet exist. Hospitalisation might have an indirect beneficial effect, because women at risk of progression to more serious stages of HDP would then be within reach of medical care. The proportion of women who might benefit in this way cannot be quantified at present and neither can its inverse, i.e. the proportion who would be admitted to hospital unnecessarily.

Antihypertensive drugs. There is agreement in the literature that pharmacological treatment of pregnant women with systolic blood pressure higher than 169 mmHg and/or a diastolic blood pressure greater than 109 mmHg is beneficial. For lower systolic and/or diastolic blood pressure, there is a lack of consensus. Most of the RCTs that evaluate the effectiveness of pharmacological treatment with antihypertensives for mild-to-moderate hypertension of pregnancy suffer methodological problems, such as insufficient sample size or exclusion of a substantial proportion of women from the final analysis. It is very difficult to pool the results, because of variations in the populations included, definitions of hypertension and outcomes, and treatments (drugs, doses, length of treatment, weeks of gestation at entry to the trial and the use of two drugs).

Magee et al.192 reviewed seven trials with a total of 623 women, comparing antihypertensive treatment with no treatment for chronic hypertension in pregnancy (blood pressure > 140/90 mmHg that predates the current pregnancy or develops before the 20th week of gestation). Antihypertensives were effective in decreasing the incidence of severe hypertension, as well as the need for additional antihypertensives and admission before delivery. No clinically or statistically significant differences were seen in other maternal or perinatal outcomes. However, the individual trials are insufficient in sample size and, considered all together, the confidence intervals are wide and consistent with both a beneficial and a harmful effect. Pharmacological treatment may benefit the mother but the impact on perinatal outcomes is not clear. There was some concern about the use of beta-blockers and their effects on the dramatically increased incidence of small-for-gestational-age babies in one of the trials.193

There were 15 trials (including 1926 women) that compared antihypertensive drugs with no treatment for mild-to-moderate hypertension later in pregnancy.192 The drugs most commonly used were methyl-dopa and beta-blockers. Antihypertensives decreased the incidence of severe hypertension and of proteinuria, as well as decreasing the need for additional antihypertensives and admission before delivery. Except for caesarean section, where no difference was seen, no reliable conclusions on other maternal outcomes can be extracted. The balance between beneficial and harmful effects of antihypertensives on perinatal outcomes is not clear. Again, the increase in the number of babies that were small-for-gestational-age, although not reaching statistical significance, is of clinical concern. These findings support the conclusion of an earlier systematic review (including 23 trials) that evaluated pharmacological treatment for mild-to-moderate hypertension during pregnancy.194

More recently, von Dadelszen et al.195 investigated the relation between fetoplacental growth and the use of antihypertensives in the treatment of mild-to-moderate pregnancy hypertension. They studied this association by meta-regression analysis of randomised clinical trials. For the principal analysis, they examined the change in mean arterial pressure from enrolment to delivery, and its effect on indicators of fetoplacental growth. The main finding was that greater differences in mean arterial pressure were associated with a higher incidence of small-for-gestational-age newborns and a lower mean birthweight. Interestingly, this association could not be explained by type of hypertension, by antihypertensive agent or by mean duration of therapy. Any treatment that reduces maternal blood pressure may reduce fetal growth. Taking into account the small and not absolutely proven benefits of this therapy for mild-to-moderate hypertension during pregnancy, large and methodologically sound RCTs that address these questions are needed.

There are two main strategies for the treatment of very high blood pressure during pregnancy: expedited delivery vs. expectant management and pharmacological treatment with antihypertensives. The risks of maternal mortality, perinatal mortality and serious perinatal morbidity are increased when women present with severe hypertension before 34 weeks of gestation. The only definitive cure for the mother is prompt delivery of the baby. However, the evidence for overall benefit is not completely clear. Two trials with a total of 133 women were conducted, comparing aggressive with expectant management.196,197 The results of both trials combined have shown that pregnancy was prolonged for about 2 weeks on average.192 Although the sample size is small and the confidence intervals are wide, serious maternal morbidity was similar in both groups, and higher incidences of small-for-gestational-age newborns and serious neonatal morbidity were seen in the aggressive management group.

There is insufficient evidence to implement either strategy; larger randomised controlled trials are needed before a formal strategy could be established. Of course, expectant management can only be contemplated in a setting where emergency care, including caesarean section, is immediately available should the woman’s condition deteriorate. Prompt referral to such a centre is essential for all women with severe hypertension in pregnancy.

Severe hypertension, i.e. above 170/110 mmHg, has major implications in terms of serious maternal and perinatal morbidity and mortality. Therefore, the main objective of reducing rapidly the high levels of blood pressure is to decrease these risks. Duley et al.198 have conducted a systematic review to evaluate the effects of antihypertensive drugs used for rapid treatment of severe hypertension during pregnancy. All but one of the trials, which accounted for over half the women (627 of about 1200 women), and five of eight comparisons included hydralazine.199 All the drugs included in all the comparisons have been shown to reduce blood pressure, but there is no evidence of any advantage of one drug over the others. Until better evidence is available, the choice of antihypertensive should depend on the experience and familiarity of an individual clinician with a particular drug, and on what is known about maternal and fetal side-effects. However, ketanserin should not be used, because of its increased risk of low blood pressure; similarly, diazoxide should not be used, because it is less effective.

Large, well-designed and properly conducted trials are needed to obtain reliable information about the effects of different antihypertensive drugs used in the treatment of severe hypertension during pregnancy. In particular, these trials should address important health outcomes for mothers and babies, and not merely the effects on lowering the blood pressure. Plasma volume is reduced among women with pre-eclampsia, and some researchers have suggested that plasma volume expanders could improve maternal and perinatal outcomes. Three trials, with a total of 61 pregnant women, were included in a systematic review.200 All the trials compared a colloid solution with no plasma volume expansion. The evidence is insufficient to recommend the use of plasma volume expanders for women with pre-eclampsia.

Anticonvulsants. The available evidence shows different findings when anticonvulsants are to be administered therapeutically, following a fit, or prophylactically, in women suffering from pre-eclampsia. A systematic review, in which one large trial dominated the results,201 showed that magnesium sulphate is substantially more effective than diazepam, phenytoin or lytic cocktail for the treatment of eclampsia, particularly in reducing the recurrence of convulsions.202,203 There is also a trend toward reducing maternal mortality, but the confidence intervals are compatible not only with a beneficial effect but also with a deleterious effect.

The appropriateness of prophylactic anticonvulsants in treating pre-eclampsia is not yet clear. A systematic review that included 10 trials in which the effectiveness of anticonvulsants for preventing convulsions in pre-eclamptic women was assessed shows that anticonvulsant drugs are associated with reduced risk of eclampsia.204 Magnesium sulphate seems to be slightly better than phenytoin at reducing the risk of eclampsia, but at an increased risk of caesarean section. Studies in which magnesium sulphate was compared with diazepam were too small for a reliable conclusion to be reached.204 There is insufficient evidence to say reliably whether or not the advantages of anticonvulsants preventing fits outweigh the disadvantages of their widespread use and the consequent side-effects. A large RCT that addresses these questions is under way.205

Obstructed labour

By definition, obstructed labour is a problem specific to labour, and definitive treatment, including caesarean section, must be available to all women to improve the outcome for mother and infant. It is in situations where access to professional delivery care (including operative delivery) is poor or delayed that obstructed labour may frequently result in maternal mortality or morbidity, such as vesicovaginal fistula, as well as perinatal mortality.206 Interventions in pregnancy along this pathway can be aimed at primary prevention, or at detection of women at increased risk and their referral for delivery in a well-equipped centre.


Obstructed labour is usually the result of cephalopelvic disproportion (CPD) or malpresentation. Strategies for primary prevention of CPD include improved nutrition, beginning at the mother’s conception and continuing throughout infancy, childhood and adolescence to ensure that women will reach their full growth potential; and interventions to delay first births until women are fully grown, or at least 17 years of age. These interventions include contraception, education and improved economic opportunities for women. These are outside the scope of antenatal care, but should be part of any maternal and child health strategy.

There is some encouraging evidence that antenatal care for very young primigravidae may increase their own growth during pregnancy. In Zaria, Northern Nigeria, a combination of nutritional supplementation (iron and/or folate) and treatment of malaria increased linear growth in teenage primigravidae.207 However, more than half the subjects in this small study were excluded for noncompliance or were lost to follow-up, leaving only 69 girls in the final analysis. Some doubt has also been cast on the measurements of height in this study, because some girls increased their stature by very large amounts, and height is notoriously difficult to measure reliably.208 Pelvic growth is completed later than growth in height, and the true gain during pregnancy is not clear. Further careful RCTs are needed to see whether or not nutritional supplementation and antimalarial chemoprophylaxis in teenagers does decrease their risk of fetopelvic disproportion and obstructed labour.

Rush209 raised the possibility that, because it can increase birthweight, nutritional supplementation may increase the risk of disproportion and the need for caesarean section in chronically malnourished and stunted mothers. This hypothesis was supported by another article,210 which suggested that prenatal interventions known to increase fetal growth should not be promoted unless skilled assistance in childbirth can be assured. Although these suggestions to reduce efforts to improve fetal growth among nutritionally deprived women sound appealing, they are not supported by solid evidence. Another study211 offers some evidence to the contrary, demonstrating that maternal height has a stronger relationship with incidence of intrapartum caesarean delivery than any measure of birth size. Head size and, therefore, delivery risk were not significantly increased.

Risk assessment and referral

Many attempts have been made both in industrialised and developing countries to predict which women will develop obstructed labour, based on the detection of fetopelvic disproportion and malpresentation.

The best test of pelvic capacity is labour, and fetopelvic disproportion can be virtually ruled out in a woman who has previously delivered a good-sized infant, unless her current fetus is very large. In Zaire, multigravidae who had required intervention in a previous delivery, or who had experienced a stillbirth or neonatal death, were found to be 10 times more at risk of obstructed labour than were multigravidae without these risk factors.212 Unusually, this population showed a higher risk of ‘life-threatening fetopelvic dystocia’ (LTFPD), including transverse lie and ruptured uterus, among women having their second birth than is found in primigravidae or grand multiparae. The authors attributed this to the high proportion of low-birthweight babies in primigravidae, but it might also reflect some selection among women attending for care based on the level of difficulty of their first birth.212 A study in Papua New Guinea213 found that 23% of 119 women who had previously been through long labours experienced complications in the index delivery, compared with 12% of women without such a risk factor. However, none of these required operative delivery, the problems being third-stage complications and sepsis. Lennox89 has also found that a history of previous long labour may be poorly recorded. Only 10% of 51 women in Papua New Guinea with this risk factor had it correctly identified and recorded on action-oriented antenatal cards, compared with 100% recording of previous caesarean section.

Obstetric history does not exist for primigravidae, so they might all be regarded as at high risk. However, the majority of primigravidae will not experience prolonged labour or obstruction, and a more specific test, with higher predictive value, is needed where specialist delivery services are scarce. The simplest additional screening tests used are maternal height, foot size and age, as proxies for internal pelvic diameters. Hofmeyr214 reviewed studies from around the world and concluded that, although short stature, small foot size and very young age are undoubtedly correlated with risk of CPD and caesarean section rates, they are poor discriminatory tools. Short maternal height is associated with higher risks of difficult labour caused by CPD.215 Because maternal height is easily measured, it has been widely used to predict risk of obstructed labour. In Aberdeen, Scotland, 17% of women less than 152 cm tall were delivered by caesarean section, and 11% by rotational forceps, compared with 8% and 5%, respectively, of taller women, giving a positive predictive value of 28%.72 It is not possible to obtain estimates of sensitivity or specificity from this published report.72

The performance of maternal height as a screening test depends on the chosen cutoff point. The shorter the at-risk height is set, the better is the sensitivity, but at the cost of poorer specificity. Although 150 cm is often quoted as a cutoff point to define high risk, the most appropriate ‘at-risk height’ is actually variable between populations. Moller and Lindmark216 have shown that it also varies within populations over time, with cohort changes in average height. One of the weaknesses of maternal height in predicting CPD is that it is a risk factor both for the mother having a small pelvis, and risk of having a low-birthweight baby.

The possibility that a combination of maternal anthropometric measurements that reflect both long-term (e.g. height) and current (weight, weight for height, weight gain, arm circumference) nutritional status might discriminate better between women at risk of CPD, and low birthweight, was investigated in a large collaborative study.217 In this study, the relationship between assisted delivery (as a proxy indicator of obstructed labour) and various indicators of maternal nutritional status were analysed. Maternal height was the only one from a wide range of measurements (arm circumference, attained weight at different stages of pregnancy, prepregnancy body mass index, body mass index at different weeks of gestation and weight gained at different stages of pregnancy) that showed a clinically small but statistically significant predictive risk.

In a prospective follow-up study that included 16 850 pregnant women, Merchant et al.211 showed that very short maternal height (148 cm or less) with large fetal size (3220 g or more) are associated with a high incidence of intrapartum caesarean delivery (53%) among primiparous women. They also demonstrated that maternal height is more predictive of a risk of intrapartum caesarean delivery than is newborn head circumference, and that head circumference was more predictive than is birthweight, in agreement with previous studies of difficult delivery.

Clinical pelvic assessment by manual examination is an element of routine antenatal care in industrialised countries. If it were a reliable test, then it might be used as a secondary-level screen in conjunction with referral of women of short stature or very young age, to distinguish before labour those women at high risk of CPD. This strategy was adopted in a study of the risk approach in Sirur, India, but insufficient outcome data are provided to assess its effectiveness.218 There is some evidence that experienced examiners can identify women with severely contracted pelvises, but insufficient data are available to assess the reliability of clinical examination in identifying women at high risk of obstructed labour or its effectiveness as part of an antenatal care programme.219

The significance of nonengagement of the fetal head near term, particularly in African women, is not clear.214 In Kasongo, among 837 primigravidae who attended an antenatal clinic during the ninth month of pregnancy and had engagement recorded by an auxiliary midwife, 176 were recorded as ‘not engaged’.212 Some 17% of these women went on to exhibit LTFPD in labour compared with 1.5% of those in whom the head was engaged (P < 0.001). Non-engagement did not predict an increased risk of forceps or ventouse delivery, and data on length of labour are not given. However, the numbers were small and more data are needed.212

Clinical tests have been advocated, such as tests of whether or not the head will engage on half sitting, but their usefulness cannot be assessed from the available evidence. Even the reliability of X-ray and ultrasound pelvimetry to predict fetopelvic disproportion remains in doubt, leaving little scope for antenatal screening to improve outcome by arranging delivery appropriate to risk, unless reliable screening tests that combine acceptable levels of sensitivity and specificity, can be established.214,220

The other major immediate causes of obstructed labour are malposition or malpresentation. These are more frequent in grand multiparae and in multiple pregnancies. Referring to obstetric history, including the number of previous deliveries and difficulties experienced, is the first step in assessing risk, and can be assessed at any time during pregnancy. In the Kasongo study,212 multiparous women with a history of complicated delivery had a relative risk of 42 for LTFPD in labour, which included transverse lie and ruptured uterus. This history identified six of the 48 multiparae who had either LTFPD or abnormally prolonged labour (APL; defined as needing forceps or ventouse delivery, not by length of labour) delivering in the study hospital, yielding a sensitivity of 12.5% and a positive predictive value of 15%.

Abdominal examination and symphysis–fundus height measurement by a skilled examiner in late pregnancy can reveal large fetuses, multiple pregnancy, polyhydramnios and abnormal lie, presentation or position of the fetus. One objective of measuring the symphysis–fundus height is to detect larger fetuses and twin pregnancies to prevent obstructed labour. Different detection rates have been reported, varying from authors who strongly recommend its use221[222]–223 to those who have shown a poor predictive value for large fetuses or mode of delivery.224,225 The only RCT that assessed the effectiveness of measuring symphysis–fundus height has shown lower detection rates for low birthweight for gestational age in the measurement group (28%) than in the control group (48%).226 However, the recommendation from a systematic review227 is that it would seem unwise to abandon the use of symphysis–fundus height measurement until larger trials show that this tool is unhelpful. The closer to labour the abdominal examination is performed, the more accurately it predicts the actual presentation and position of the fetus in labour.

External cephalic version (ECV) of breech presentations at term (but not before) has been shown to reduce the rate of caesarean section and of breech delivery.214,228,229 There is good evidence to support the use of ECV at term for breech presentation. However, the number of cases studied is too small to give an accurate estimate of the risk of this procedure. ECV is recommended in those women where the value of an improved chance of a cephalic birth outweighs the potential risks of the procedure. Risks include the current and future risks of caesarean section, which are increased in settings where resources are scarce. To date, there is insufficient evidence to judge the effectiveness of ECV in transverse and oblique lies, which are much less common; it is not possible to extrapolate from the effect on breech, because the causes may be different and may be associated with a greater degree of instability after version.

More data from epidemiological studies are needed to assess the accuracy of clinical examination by primary health-care workers and specialists to detect abnormal lie, and to make an estimate of the likelihood that position would change between examination near term and labour. Large-scale studies are needed to determine the level of skill required to detect abnormal presentation near term and its predictive accuracy for malpresentation in labour. Transverse and oblique lie, with their very serious risks to mother and infant, should be easier to detect than breech, even with moderate skill or training, but not enough reliable data were found with which to examine this. Given the current state of knowledge, women presenting in this way near term must be advised to deliver in a fully equipped referral-level facility. Large-scale multicentre trials would be needed to assess the value of ECV at term to correct transverse and oblique lies, so preventing obstructed labour.

To minimise the risks of obstructed labour during antenatal care, the following interventions are recommended.

• Measure maternal height for primiparous women at first examination, especially in settings where hospital birth is not routine. Cutoff levels for hospital birth should be decided locally.

• Measure height from symphysis to fundus at every visit for all women. Values more than two standard deviations above normal require referral for hospital delivery or evaluation for twin pregnancies.

• Pregnant women with breech presentation at 37 weeks of gestation should be referred for external cephalic version.

• Women with a history of previous caesarean section and/or prolonged labour with or without instru-mental delivery should be referred for hospital delivery.5


Puerperal sepsis

The possible ways in which care during pregnancy might reduce the risk of death or long-term sequelae from puerperal infection are widely acknowledged. These include the following:

• risk detection and referral;

• health education for clean delivery in order to recognise and seek care for signs of infection or ruptured membranes;

• supply of clean delivery kits to women to prevent unclean delivery;

• detection and treatment of genital tract infection that is present prior to labour;

• recognition of and treatment after prelabour rupture of membranes.

Although reliable data are most difficult to obtain in areas where the risk is highest, the risk of infection and death appears to be related to a lack of trained assistance at delivery and poor obstetric facilities. Historical data suggest that most puerperal sepsis is related to infection introduced at delivery, and that the risk is increased in proportion to the frequency of vaginal examination and intervention during labour. The early improvement in maternal mortality from puerperal sepsis in developed countries arose primarily from improved hygiene practices at delivery. Later, the introduction of antibiotics to treat infection virtually eliminated deaths from this cause in industrialised nations.29

Following this model, the role of antenatal care in reducing mortality and morbidity from infection would be limited. Health education to promote clean delivery in the home, distribution of ‘clean delivery’ kits directly to pregnant women and the promotion of delivery by trained attendants might prevent some infection. Education might also lead to better recognition of the importance of symptoms and signs of infection after delivery, and to women seeking care earlier. However, the crucial factors that give rise to unclean delivery are probably more related to poverty and lack of facilities for any alternative than to ignorance. Therefore, the effect of health education alone is likely to be small.

In addition, the likelihood that a woman will be referred for institutional delivery, based on an assessment of the likely risk of infection at delivery, is probably slight. Many risk factors are likely to be features of the community in which a woman lives and, therefore, of the delivery care and facilities available to her, rather than features that could distinguish between women at high and low risk in any given community.

One of the most successful interventions in pregnancy is the introduction of maternal vaccination to prevent maternal and neonatal tetanus resulting from infection at delivery. Unfortunately, only 33% of the mothers in developing countries have been shown to receive two or more doses of tetanus toxoid.42 Furthermore, there is some evidence that a large proportion of women who attend antenatal care in developing countries do not receive the recommended tetanus immunisation.230 One of the priorities in antenatal care is to provide tetanus immunisation to mothers who have not been vaccinated previously. Women without written evidence of previous immunisation should receive a complete series in the present pregnancy. The first dose should be at the first visit, irrespective of gestational age, and the second dose 6–10 weeks later.

The role of pre-existing, lower reproductive tract infection in puerperal infections is not clear. Although it may not account for much acute puerperal fever and mortality, it may be associated with much more low-grade infection and chronic morbidity, including pelvic inflammatory disease and secondary infertility. Epidemiological studies are needed to clarify the prevalence of various infections and their contribution to morbidity and mortality in various developing countries.

Prolonged rupture of the membranes before delivery carries an increased risk of ascending infection. This is true in prolonged labour with ruptured membranes and with rupture of the membranes prior to full-term or preterm labour. The number of vaginal examinations performed increases the risk. Current evidence does support induction of labour in the case of uncomplicated prelabour-ruptured membranes at term or near term. Trials of induction for prelabour rupture of membranes at or near term, either using oxytocin or prostaglandin, have shown consistent beneficial results.231,232 Compared with expectant management, induction of labour by oxytocin is associated with a decreased risk of maternal and neonatal infection, and no increase in the rate of caesarean section.231 Furthermore, induction of labour by prostaglandin for prelabour rupture of membranes at or near term is associated with a lower risk of maternal infection and admission to intensive neonatal care; no differences are seen in the rate of caesarean section.232 In the comparison of prostaglandin with oxytocin for prelabour rupture of membranes at or near term for induction of labour, women with prostaglandins appear to have an increased risk of choriamnionitis and nausea/vomiting in comparison with those treated with oxytocin.233

In preterm rupture of membranes in women not in labour, prophylactic antibiotics appear to reduce the risk of subsequent maternal and neonatal infectious morbidity, and to prolong pregnancy. However, there is no proof that prophylactic antibiotics improve neonatal morbidity or mortality, so there is insufficient evidence to recommend routine antibiotics for preterm rupture of membranes.234 A large RCT that addresses relevant questions about the effects of antibiotics for preterm rupture of membranes on maternal and neonatal morbidity and mortality is under way.235 The administration of maternal corticosteroids in pre-term labour is a safe and effective intervention to reduce neonatal morbidity and mortality.236 Women with prelabour rupture of membranes should be referred quickly to a level at which skilled assessment and management are available. Depending on local circumstances, it may be appropriate to begin treatment with antibiotics and even corticosteroids before transfer.

Routine antenatal care has a limited place in prevention and management. Pregnant women can be taught about the importance of seeking care promptly after membrane rupture, but the means for them to do so must be provided if this is to be effective. No assessment of the effectiveness of this sort of health education has been found during this review. Ascending infection from the lower genital tract may predispose to membrane rupture and chorio-amnionitis.237 Therefore, investigation and treatment of vaginal discharge in those attending antenatal clinics could prevent some cases, as could screening for and treating sexually transmitted diseases (STDs). Current evidence does not support the strategy of screening and treating all pregnant women for bacterial vaginosis to prevent preterm birth and preterm prelabour rupture of membranes.238 A new trial that confirms the results of the review has been published.239 To determine whether a policy of screening for and treating asymptomatic bacterial vaginosis could prevent preterm delivery, 1953 women were randomly assigned to receive 2 g of metronidazole or placebo. This treatment, although effective in reducing bacterial vaginosis, did not reduce the incidence of preterm delivery or other adverse perinatal outcomes.

Sexually transmitted diseases

Sexually transmitted infections during pregnancy can give rise to serious maternal morbidity and to perinatal mortality and morbidity, although early infection in women is often asymptomatic. Screening of all pregnant women has long been advocated, both because of specific risks of consequences of infection in pregnancy to the woman and fetus, and as opportunistic screening of a sexually active population in an attempt to control the spread of disease. The evidence that opportunistic screening is effective in controlling the level of disease in the population is limited, but the scope for benefit to the health of women and children may be very considerable if the prevalence is high.

Information on the prevalence of syphilis, gonorrhoea and other STDs in the local population is needed before deciding whether or not universal screening should be a priority, but this is an area in which antenatal care clearly can have an important benefit for mother and child.

Syphilis. Analyses in the UK and USA have demonstrated continuing high cost:benefit ratios for a policy of routine screening for syphilis, despite low prevalence.237 Many developing-country studies have found seroprevalence rates of syphilis among pregnant women of 5–15%, which is one or two orders of magnitude higher than is found in industrialised countries today, but similar to levels reported in the early part of this century. Thus, there is considerable scope for intervention.57,240[241]–242

Schulz et al.243 estimate that systematic screening and treatment of syphilis in pregnancy would be at least as cost effective in terms of child health as is the Expanded Programme of Immunisation. A study in Zambia found a seroprevalence of 8% among those attending antenatal clinics, and showed that 58% of these pregnancies result in abortion, stillbirth, prematurity, low birthweight or congenital syphilis compared with 10% among seronegative women.57 In this population, syphilis was the single most common cause of adverse outcome of pregnancy. The study highlighted three major failings by the routine antenatal service:

• failure to screen 80% of those attending the clinic, despite the fact that serology in this clinic was supposed to be mandatory;

• failure to treat three-quarters of those found to be positive, or almost any of their partners;

• failure to detect infection acquired during pregnancy, after initial screening. Even during the intervention study, screening and treatment were suboptimal, but the percentage of adverse fetal outcomes was halved.

Several reliable tests for syphilis exist, some of which are more sensitive (Venereal Disease Research Laboratory [VDRL] and Rapid Plasma Reagin Test [RPR]) and some of which are more specific than others (Fluorescent Treponema Antibody Absorption Test [FTA-ABS], Microhaemaglutination Assay for Antibodies to Treponema Pallidum [MHA-TP]), and at varying cost. Rapid, cheap serological tests are available that give almost instant results. Using these means, women can be screened and treated in the same clinic visit, so improving the uptake of treatment. Routine screening is of no use if adequate treatment, follow-up and treatment of partners are not available. All these components should be integrated into routine antenatal care.

Operational barriers to effective screening and treatment of all pregnant women need to be identified and overcome. Screening high-risk groups, coupled with examination of women with symptoms, or whose partners report symptoms, is a ‘second best’ alternative, but only in areas of proven low prevalence. Investigation and treatment, or even presumptive treatment of women with genital ulcers and their partners, may have an impact on the transmission of HIV–although studies of this are still under way. Women at high risk of acquiring infection should be screened both early and late in pregnancy, which, in some areas, will mean repeated screening of all women. Penicillin is still a highly effective therapy for syphilis treatment, although this intervention has not been evaluated rigorously.

Gonorrhoea. Wang and Smaill237 have reviewed the effectiveness of screening for gonorrhoea in pregnant women, mainly in developed countries, and found that it is justified by the potential benefit, even where prevalence rates are not high. The risks that gonococcal infection poses to a woman’s health may be increased by pregnancy. There is some evidence that blood-borne spread, which gives rise to disseminated infection and involvement of joints and other systems, occurs more commonly in pregnant than in nonpregnant women. Ascending infection may lead to septic abortion in early pregnancy, or chorio-amnionitis and prelabour rupture of membranes. In the puerperium, gonorrhoea can cause endometritis and pelvic infection, leading to severe puerperal infection. Maternal gonorrhoea may occasionally lead to severe disseminated neonatal infection but by far the commonest complication in neonates is ophthalmia neonatorum.

Universal application of silver nitrate or tetracycline eye drops at delivery is a cheap and effective strategy for preventing blindness caused by ophthalmia neonatorum in infants.243 Other agents, such as erythromycin and povidone iodine, have also been used. An RCT has shown the beneficial effects of the povidone iodine solution over silver nitrate and erythromycin.244 However, a large RCT is needed to assess reliably the effectiveness of the these drugs. This is a low-cost treatment that could have a significant impact on child health without the need for screening, but it does nothing to improve the health of women with gonococcal infections who need effective systemic antibiotics to prevent immediate and long-term complications of infection.

Definitive diagnosis of gonorrhoea depends on the culture of a fairly fastidious organism, so necessitates access to a microbiological laboratory; however, a presumptive diagnosis can be made using microscopy. Nevertheless, Wang and Smaill237 advocate screening all pregnant women early in pregnancy by culture, with repeated testing for high-risk groups of women. Once again, the benefit of universal screening depends on local prevalence.

Investigation of symptomatic women and women whose partners have symptoms or confirmed gonococcal infection is also very important. As with syphilis, effective safe treatment for gonorrhoea exists, although knowledge of local patterns of antibiotic resistance or culture for sensitivity is needed. Presumptive treatment of symptomatic women and their partners may be necessary where laboratory diagnosis is not available, particularly if prevalence is known to be high. Local prevalence surveys may be needed to establish this, and these could be used to identify locally appropriate risk factors. However, when it is available, gonorrhoea treatment during pregnancy should be based on gonococcal culture and susceptibility testing.

The results of a systematic review245 have shown that penicillin, amoxicillin plus probenecid, ceftriaxone, cefixime or spectinomicin are suitable options for gonorrhoea treatment. Ascending infection at delivery or afterward may contribute to secondary infertility. Epidemiological research, including case-control stud-ies of pelvic inflammatory disease and secondary infertility, could be carried out to investigate this risk.

Chlamydia. Even in developed countries, the prevalence and consequences of chlamydial infection are not clear. Infants of infected mothers are at risk of inclusion conjunctivitis and pneumonia. Infection may be associated with abortion, preterm delivery and low birthweight, but a causal link has not been proved. Chlamydial infection may give rise to cervical discharge or dysuria, but it is often asymptomatic. Postpartum endometritis has also been associated with chlamydial infection.

Diagnosis relies on sophisticated laboratory tech-niques of cell culture or antigen immunofluorescence, which are rarely available to all women in developing countries. However, some simpler tests, such as the polymerase chain reaction (PCR), have also been shown to be good diagnostic tests in pregnant women.246,247 Routine screening for chlamydia during pregnancy has not been proved to be cost effective. Antibiotic therapy for chlamydial infection has been shown to be highly effective in reducing by 90% the positive cultures in comparison with placebo treatment. Brocklehurst and Rooney248 have demonstrated that amoxicillin seems a good alternative therapy for genital chlamydial infections during pregnancy in comparison with erythromycin. However, clinda-mycin and azithromycin could be given when the above-mentioned antibiotics are contraindicated or not tolerated.

Trichomonas. Vaginitis caused by trichomonas vaginalis is one of the commonest sexually transmitted diseases. This disease affects about 120 million women every year.249 Symptomatic infection is characterised by yellow discharge, dyspareunia, abnormal odour, vulvar pruritus and dysuria. Diagnosis is readily demonstrated by identification of the parasite in a wetmount smear. More expensive and sensitive diagnostic methods are available, such as culture, immunofluorescence and enzyme immunoassay. Evidence of the effects of trichomonas infection on pregnancy outcomes is controversial.250 Gulmezoglu251 concluded that metronidazole treatment, given in a one-dose schedule, provides an effective parasitological cure, but the effect of this intervention in pregnancy-related outcomes is not known.

It is not yet clear how important chlamydia and trichomonas infection will prove to be in terms of chronic morbidity, and this deserves some attention. However, the detection and treatment of syphilis and gonorrhoea are both of greater priority and more easily achieved.

HIV. In some countries of the developing world, infection with the HIV has become one of the commonest complications of pregnancy. This important subject cannot be covered in detail in a report of this length. A complete review, dealing with all the relevant aspects of HIV in pregnancy, has been published elsewhere by the WHO.252 However, some relevant issues will be addressed here. At the end of 1998, more than 33 million people were living with HIV. Almost half of them were women in their reproductive years and more than 1 million children had been infected from their mothers.253 It has been estimated that 1.5 million HIV-positive women become pregnant every year, which means that almost 600 000 children will be infected annually by mother-to-child transmission.

Pregnancy seems to have little effect on the progress of infection in asymptomatic HIV-positive women or in those with early infection, although there may be more rapid progression in women with a late stage of HIV infection.254[255]–256 Furthermore, HIV infection is not regarded as a definite indication for termination of pregnancy. Although studies have shown that HIV has little effect on maternal and perinatal outcomes in the developed world,255,257[258]–259 other studies in developing countries have demonstrated a higher risk of certain conditions related to pregnancy outcomes, such as spontaneous abortion, ectopic pregnancy, stillbirth, preterm labour, preterm rupture of membranes, urogenital tract infections and syphilis. In addition, opportunistic infections, such as tuberculosis, herpes zoster, bacterial pneumonia and Kaposi’s sarcoma, are more common in infected than in uninfected women.252

Estimates of the rate of vertical (mother-to-child) transmission vary greatly, ranging from 15% to 25% in Europe and the USA, to 25% to 40% in Africa and Asia.260,261 The lower incidence reported in developed countries could be explained by the introduction of antiretroviral therapy. Vertical transmission can happen in different stages of pregnancy, delivery or, postnatally, by breastfeeding; therefore, depending on the stage, different strategies could be implemented to prevent vertical transmission. A number of strategies for preventing vertical transmission are known and are being researched. These include the following:

• termination of pregnancy;

• behavioural interventions;

• antiretroviral drugs;

• vitamin A and other micronutrients;

• immunotherapy;

• avoidance of invasive tests (chorion villous sampling, amniocentesis, cordocentesis, external cephalic version);

• vaginal cleansing during delivery;

• caesarean section delivery;

• avoidance or early cessation of breastfeeding.252

From these, to date, there are only some interventions that have been rigorously evaluated by RCTs. Brocklehurst262 carried out an extensive systematic review, leading to the following findings.

• In four trials comparing Zidovudine with placebo and involving 1585 participants, Zidovudine showed a statistically and clinically significant reduction, halving the risk of mother-to-child transmission independently of long- or short-course therapy.

• Nevirapine was compared with Zidovudine in one trial that involved 626 women. Nevirapine showed a statistically and clinically significant benefit in reducing the incidence of mother-to-child transmission. It is worth pointing out that most of the women were breast feeding their infants.

• Delivery by elective caesarean section was evaluated in comparison with vaginal delivery in one trial that included 436 women. Caesarean section significantly lowered by about 80% the risk of mother-to-child transmission. The reduction in those women who also received antiretroviral therapy was not statistically significant. It has been suggested that, with adequate antiretroviral therapy, the advantage of caesarean section may be reduced. The risks of caesarean section for women who are already immunologically compromised have not yet been addressed.

• Hyperimmune immunoglobulin plus Zidovudine was evaluated in comparison with nonspecific immunoglobulin plus Zidovudine in one trial with 501 women. The results of this trial do not indicate any beneficial effect of this therapy. Zidovudine, Nevirapine, delivery by caesarean section and alternative infant feeding seem to be very effective interventions to decrease mother-to-child HIV transmission.

Screening of women at high risk, or in areas of high risk, has been advocated. However, the main benefits of knowing HIV status at that time were counselling about the advisability of further pregnancies and avoidance of transmitting the infection, and early detection and treatment of opportunistic infections.237 Since then, with increasing knowledge of the mechanisms of vertical transmission and the introduction of preventive interventions known to be effective, the situation has changed from the probable public health benefits of testing to the current real benefits for the individual woman and her child.263,264 In view of this, it is now absolutely necessary to provide all pregnant women with facilities for counselling and voluntary HIV testing. However, these facilities should be offered only when women who have tested positive for HIV have access to antenatal care and where there is provision of measures to prevent or cure the conditions related to this infection.

Urinary tract infection

Urinary tract infection (UTI) is a relatively common medical complication of pregnancy, and asymptomatic bacteriuria is the most prevalent of these conditions.265,266 The prevalence of this condition in pregnancy is similar to that seen in nonpregnant women, ranging from 4% to 7%,267[268][269][270]–271 although lower or higher rates have been reported in certain populations.272,273 It has been shown that, when untreated, 20–40% of pregnant women with asymptomatic bacteriuria will develop acute pyelonephritis.266,274

Screening in pregnancy for asymptomatic bacteriuria and treatment of positive cases with antibiotics is widely recommended,3,237,275 following evidence from RCTs that treatment reduces the risk of pyelonephritis, cystitis, preterm delivery and low birthweight. Whether the reductions in preterm delivery and low birthweight result from a reduction in pyelonephritis or via eradication of micro-organisms, particularly beta haemolytic streptococcus, from the cervix is not clear.275 In these trials, screening was generally carried out in the first trimester, as recommended by the report Caring for Our Future: The Context of Prenatal Care,3 although it is not clear whether the timing matters, or whether anything is to be gained by repeated screening. However, in a large study,270 the highest rates of bacteriuria were detected between the ninth and 17th weeks of gestation. The 16th gestational week seems to be the optimal time for simple screening for bacteriuria, calculated as the number of bacteriuria-free gestational weeks gained by treatment. Most bacteriuria, as for most symptomatic UTIs, is caused by Escherichia coli.

Short courses of antibiotics appear to be as effective as longer courses276 and there seems to be little to choose between ampicillin and other antibiotics in terms of effectiveness.275 However, it is important to assess local sensitivity, because ampicillin-resistant organisms are prevalent in some settings. The benefit of this strategy will depend on the prevalence of asymptomatic and symptomatic UTIs in the population. Reliable estimates from developing countries are difficult to find. A study from India, in which those attending antenatal clinics were screened for asymptomatic bacteriuria, using the Griess test for nitrites in the urine, gave a prevalence in this population of 2.7%, i.e. at the low end of the range 2–20% reported in studies from industrialised countries.277

The traditional reference test for UTIs is the quantitative culture. However, this method is expensive, time consuming and cumbersome, needing the infrastructure of a microbiology laboratory with qualified and trained staff. To solve these problems, certain screening tests have been developed. The most widely assessed screening tests are the nitrite and leukocyte esterase reagent strips tests. Although these tests combined seem to be promising for the detection of bacteriuria in asymptomatic pregnant women, there is insufficient evidence to reassure clinicians and patients that a negative result is truly negative.278 Further methodologically sound research is needed on appropriate technology for bacteriuria screening where prevalence studies indicate that screening would be worthwhile.

Interventions related to other maternal health outcomes

Pregnancy and childbearing have a profound and long-lasting effect on women’s health and influence the risk of mortality from most major causes. Pregnancy involves enormous physiological changes involving all bodily systems over a relatively short period. Therefore, it is not surprising that pregnancy influences the risk and course of many diseases. According to the 10th revision of the International Classification of Diseases, indirect maternal mortality is defined as death during pregnancy or within 42 days of its termination, resulting from previous existing disease or disease that developed during pregnancy, and which was not the result of direct obstetric causes, but was aggravated by the physio-logical effects of pregnancy.279 The definition of ‘indir-ect’ maternal morbidity is more problematic.22,280 At its narrowest, it must include those conditions that can lead to maternal mortality, but this excludes many chronic, potentially disabling morbidities. It has been argued that the definition of maternal morbidity should include all conditions that directly or indirectly arise from or are aggravated by pregnancy, childbearing and its management.280,281

Which illnesses most frequently affect pregnant women in a given setting will depend on local patterns of prevalence in the general population, as well as on the influence that pregnancy has on the course of the illness. The causes of adult mortality and morbidity in developing countries represent a neglected area of study.282,283 Many of the data that are available are from vertical programmes for the control of particular ‘tropical’ communicable diseases. This is the area in which direct extrapolation from experience in developed countries is least appropriate, because the levels and nature of morbidity encountered are very different. There exists very little information on which to judge the effectiveness of antenatal care provision in reducing maternal mortality and morbidity from conditions not directly caused by pregnancy but exacerbated by it. Nevertheless, the higher prevalence of treatable disease experienced by women in developing countries suggests that there is much greater potential for the detection and treatment of illness during pregnancy in these populations. This is particularly true of infectious and parasitic diseases. Susceptibility to many of these, including tuberculosis, malaria and hepatitis, may be increased in pregnancy and they may be important causes of maternal and perinatal mortality and morbidity.

A confidential enquiry into maternal deaths in Addis Ababa, Ethiopia, identified hepatitis as the commonest cause of indirect maternal mortality.34 In this setting, viral hepatitis appears to be associated with especially high case fatality in pregnant women. There is a high incidence of hepatitis B infection and prevalence of the carrier state in many parts of the developing world.284 Hepatitis A is usually a self-limited infection without a chronic carrier state, and perinatal transmission does not occur. Hepatitis B may cause chronic infection and infants delivered to infected mothers are at risk of neonatal hepatitis. Passive and active immunisation are highly effective in preventing perinatal transmission. Hepatitis D occurs as a coinfection or superinfection with hepatitis B. Patients infected with both viruses are at high risk of chronic liver disease. Perinatal transmission of hepatitis D can be prevented by the immunoprophylaxis used for hepatitis B. Non-A, non-B hepatitis occurs in two distinct forms: parenterally transmitted hepatitis C and enterically transmitted hepatitis E. Perinatal transmission of hepatitis C can occur, particularly in women who are concurrently infected with HIV. Neonatal immunoprophylaxis is not yet available.285 Although a chronic carrier state does not exist and perinatal transmission cannot occur, hepatitis E may be associated with high maternal mortality rates in developing countries.286 However, more information is needed on the incidence, natural history and management of the different types of viral hepatitis in pregnancy.

Tuberculosis is a leading cause of death in many developing countries,287,288 and the risk of active tuberculosis is increased in pregnancy. Rheumatic fever and other sequelae of infection are still commonplace in some developing countries. Pregnancy affects profoundly the severity of morbidity and the survival rates from these conditions. Policies for screening and treating pregnant women will depend on local prevalence and population screening and control programmes. Treatment will depend on local patterns of drug sensitivity/resistance, as well as on evidence of efficacy and safety in pregnancy, including, where available, teratogenesis. Coordination with specialist services for treatment and control may be necessary. Antenatal care programmes may allow opportunistic screening and recruitment of women into other health-promotion and health-care programmes. The effectiveness of incorporating these services into routine antenatal care is difficult to assess, and few data are available.

Conclusions and recommendations for further research

What is striking in examining the evidence for or against the effectiveness of care during pregnancy in reducing maternal mortality or serious morbidity is how little is known. This is not just a problem of how to deliver services of proven efficacy in developing country settings. The report on antenatal care in the USA3 and, more recently, reports by Villar and Bergsjø5 and Bergsjø and Villar4 emphasise the same fundamental lack of certainty about the efficacy of many established practices. They also highlight the need for good research to provide the information required for future policy-making, which is apparent from this review. These questions are more pertinent to developing countries, because staff and resources are scarce, the levels of mortality and morbidity remain much greater and the lack of reliable information even more marked.

Underlying this uncertainty is ignorance of, or lack of information about the following:

• the aetiology, pathogenesis and natural history of even some of the more common complications of pregnancy, such as pregnancy-induced hypertension;

• the levels and patterns of causation of maternal mortality in many developing countries;

• the levels and patterns of maternal morbidity;

• the biological efficacy of many treatments in current use.

In situations of high maternal mortality, all opportunities for contact between health services and pregnant women or women of childbearing age should be used to maximum effect. If these opportunities are to be seized, and if more women are to be persuaded to attend for antenatal care, then it is essential to make use of staff, resources and women’s time as effectively as possible. In the face of such large gaps in knowledge, particularly gaps in basic epidemiological data about maternal mortality and morbidity in developing countries, it is difficult to decide on the priorities for research.

Nevertheless, there are some interventions in pregnancy that have been shown to be effective in detecting, treating or preventing conditions in pregnant women that might otherwise give rise to serious morbidity or mortality. These are outlined in Table 1. They relate mainly to chronic conditions – anaemia, HDPs and infections in pregnancy – rather than directly to acute conditions such as haemorrhage or obstructed labour, which emerge close to the time of delivery. However, even in these more chronic conditions, the picture is not entirely straightforward and further research is needed.

Table 1.  Antenatal interventions known to be effective Thumbnail image of

The prevention and treatment of anaemia should be given priority. Effective remedies exist but the questions to be answered in deploying them are complex, as outlined in Table 2. Routine supplementation with iron and folate is warranted where the levels of prevalence of anaemia and iron deficiency are high. However, epidemiological studies of haemoglobin levels and pregnancy outcome, and controlled trials of supplementation are needed to determine at what level of haemoglobin, iron and folate supplementation should be administered to all pregnant women. Operational and anthropological research are called for to overcome difficulties in providing adequate supplements for sufficiently long in an acceptable form. In addition, suitable techniques and strategies to allow universal screening for anaemia need to be devised or identified and tested. Strategies to ensure the effective treatment of moderate and severe anaemia must also be devised and tested in large, thorough trials.

Table 2.  Research questions regarding potentially effective interventions for anaemia Thumbnail image of

More fundamental research is required to obtain the information necessary to plan services to reduce mortality and morbidity from HDP. Although it appears that professional care in hospital reduces mortality from eclampsia, and that magnesium sulphate reduces the recurrence of convulsions and there is a trend for reduced maternal mortality, the best treatment for this or milder stages of disease are not clear, and the effectiveness of early detection is also uncertain. Table 3 outlines the questions that need to be answered and indicates the types of study required.

Table 3.  Research questions regarding potentially effective interactions for hypertensive diseases of pregnancy Thumbnail image of

Adequate management of HDP appears to require screening of all women and prompt, appropriate referral for treatment and further investigation or surveillance, coupled with 24-h access to care for emergency cases. However, there remain questions about the best method and timing of screening, and about how to provide it effectively to all women in developing countries. Whether expedited delivery or treatment of mild disease (pregnancy-induced hypertension or mild pre-eclampsia) with rest, any group of antihypertensives or plasma expanders prevent progression to more serious disease is not clear and can only be clarified by large-scale, carefully conducted randomised trials. Similarly, the effectiveness of different types of antihypertensive drug in controlling blood pressure, preventing complications and improving survival for mother and fetus in cases of severe pre-eclampsia and eclampsia can only be determined by large-scale, meticulous, randomised trials. How to provide this care at the most peripheral health-care level possible can then be explored. Trials of preventive strategies, such as calcium supplementation, should also be carried out in populations of high-risk women or those with low calcium intake. However, epidemiological studies of aetiology are also urgently needed. In addition, there is a need for large trials to address whether the claimed benefits of prophylactic anti-convulsants outweigh the unconfirmed detriments.

Women with severe pre-eclampsia require specialist care. Guidelines for first aid and referral are needed, adapted to local circumstances and with strategies to maximise cooperation and communication between primary and first and subsequent referral levels. The role of anticonvulsants and antihypertensive drugs, and the level of training needed to supervise their use need to be clarified. The effectiveness of health education in improving recognition of the importance and gravity of symptoms such as headache, abdominal pain and visual disturbance in late pregnancy, particularly in conjunction with signs of pre-eclampsia, needs to be evaluated.

With regard to infection, the potential for improving maternal and child health through the detection and treatment of sexually transmitted or genitourinary tract infections is high. Reliable tests and effective treatments exist for infections such as syphilis and gonorrhoea, as shown in Table 1. Table 4 sets out the remaining questions that must be answered if these infections are to be dealt with effectively in antenatal care.

Table 4.  Research questions regarding potentially effective interventions for infection Thumbnail image of

A high priority must be operational research to identify the best ways of making available to all pregnant women the screening, treatment, follow-up and health education for prevention of these infections. The performance of these services must be subject to evaluation, with the collection of outcome and process measures as a priority. Better epidemiological data on levels and trends of infection are needed, as are data on the relationship between infection and outcome meas-ures for mother and infant. In the case of chlamydia and trichomonas infections, further epidemiological research on prevalence and consequences is needed, as well as the development of screening and diagnostic tests appropriate to developing-country settings. The role of screening for HIV, and the appropriate treatment with antiretroviral drugs is now clearer in terms of patient management, although epidemiological data are urgently needed on levels and trends. Although bacteriuria screening and treatment has proved beneficial, more information is needed on the prevalence and consequences of asymptomatic and symptomatic UTI in pregnant women in developing countries. Appropriate screening and diagnostic tests, and operational problems will have to be addressed in extending this screening to the entire pregnant population in areas where the prevalence warrants this.

The major questions regarding the potential of antenatal interventions to prevent or ameliorate the effects of haemorrhage and obstructed labour are set out in Tables 5 and 6. There is clearly much that can be done to improve maternal health through improved standards of, and access to delivery care. Levels of maternal mortality from haemorrhage, obstructed labour and puerperal infection are responsive to improvements in the quality, coverage, utilisation and emergency access to delivery care. The role that antenatal care might play in reducing mortality or serious morbidity from these causes is not clear and has not been tested. One major pathway through which such care might do so is by screening and referral for supervised delivery, but this could only be effective where delivery services can cope with the additional workload. Obstructed labour seems to be better predicted than other life-threatening complications using maternal height, which is certainly a cheap and simple test. However, this test suffers from poor sensitivity and specificity. There remain questions about how and when the risk factors should be measured and how these measurements could be useful in predicting obstructed labour. The effec-tiveness of the formal risk-approach strategy in re-ducing mortality or serious morbidity has not been systematically examined in trials, and there is a need to confirm or refute their supposed benefits (Table 7).

Table 7.  Research questions regarding to potentially effective interventions for formal risk approach Thumbnail image of

Basic and epidemiological research on antenatal care in developing countries is not an academic luxury; improved information on patterns of maternal health and the efficacy of investigation and treatment are essential to rational planning of effective health services to reduce maternal mortality and morbidity from their current alarming level. Many of the fundamental questions raised in this review apply to countries throughout the developed and developing world. However, developing countries are the best settings for much of the recommended research for several reasons.

The fact that incidence and mortality rates are higher in developing countries means that the questions are more urgent, but also that sample sizes could be smaller and benefits more easily seen. The information obtained would be directly relevant to the populations that could benefit most from the interventions. Research studies could provide opportunities to strengthen local expertise in research and clinical practice, and to find solutions appropriate to local needs and circumstances.

Although this review has looked at interventions during pregnancy, the effectiveness of these interventions should not be considered in isolation. Effective prevention of maternal mortality and morbidity will require improvements in all aspects of women’s health and health care throughout their lifespan.


During the preparation of this manuscript, Guillermo Carroli was supported with a grant from the UNDP/UNFPA/WHO/World Bank Special Programme of Research, Development and Research Training in Human Reproduction of the Department of Reproductive Health and Research of the World Health Organization, Geneva, Switzerland.

In particular, we would like to thank Dr Marion Hall and Professors G. Justus Hofmeyr and James P. Neilson for their very useful comments on an earlier version of this document.