The prevention of low birth weight is a public health priority in many developing countries (1), where the condition is largely attributable to intrauterine growth retardation, as compared with prematurity in developed countries (2). Prematurity and intrauterine growth retardation have different risk factors and different prognoses for infant survival and long-term morbidity. There have been few population-based studies on low birth weight, especially those designed to distinguish between prematurity and intrauterine growth retardation; those that have been conducted have often had an inadequate control of confounding and a lack of statistical power, resulting in inconclusive evidence for the determinants of intrauterine growth retardation in developing countries. Studies in Latin America (3) and India (4) have shown that the following are the major risk factors for intrauterine growth retardation: maternal illiteracy; young maternal age; primiparity; poor obstetric history in previous pregnancies; smoking; short birth intervals; low maternal weight; short stature; moderate and severe anaemia; an absence of antenatal care; vaginal bleeding; and maternal hypertension. By and large these findings are in accordance with the evidence of Kramer's meta-analysis (1).
In order to plan an effective preventive programme in Pakistan, where there is a marked prevalence of low birth weight (1), it is important to generate population-based estimates of relative and attributable risk of term intrauterine growth retardation.
Materials and methods
We conducted a community-based prospective study of pregnancy outcomes in four urban squatter settlements in Karachi (Chanesar Goth, Essa Nagri, Grax, and Orangi) in order to assess the level of and risk factors for term intrauterine growth retardation. The settlements are included in a primary health care project run by the Department of Community Health Sciences, Aga Khan University. Household listings and basic demographic information are available for the entire communities. Beginning in August 1990, a total of 1000 women were recruited into the study after they had given their informed consent. A target sample for each field site was established, based on the number of married women in the reproductive age range and the pregnancy rate. A quick survey of all married women of reproductive age was conducted to determine their pregnancy status, and pregnancy tests were performed on those women whose status was doubtful. Pregnant women were recruited irrespective of gestational age. Because the target sample size for each site could not be achieved using this strategy, we followed up non-pregnant women prospectively through monthly household surveys, questioned them about their menstrual history, and, if they were pregnant, recruited them into the study. This process was stopped once the targets had been reached.
Doctors and nurses were specially trained to conduct interviews and physical examinations. For the women recruited, detailed information was collected on socioeconomic and demographic factors, diet before and during pregnancy, consanguinity, and medical and obstetric characteristics; also, all the women underwent complete physical examinations including blood pressure determinations, anthropometric measurements, and vaginal examinations. The laboratory investigations that were carried out included measurement of haemoglobin levels, urine analysis, and stool examination. Since the women delivered mainly at home, attended by traditional birth attendants, it was imperative to learn about births as soon after delivery as possible in order to obtain the various measurements and evaluations of the newborn babies. The participants were repeatedly reminded to inform their community health workers of births at the earliest opportunity. These workers were trained to carry out various measurements on newborn babies and to apply Parkin's method of estimating gestational age (5). …