Results

Nine hundred and eighty seven singleton live born infants were recruited between July 1994 and January 1996. Birthweight in preterm infants ranged from 0.78 to 3.00 kg and 59/62 (95%) were low birthweight. Overall, the relative risk (95% confidence interval) of low birthweight for primigravidae compared with gravida 2 women was 2.23 (1.43–3.50), p<0.001, and compared with all multigravidae it was 2.83 (2.00–4.01), p<0.0001. BMI was calculated in 466 women and 29 (6.2%) were underweight (<20 kg/m²). BMI was strongly correlated with weight during pregnancy (Pearson’s r=0.91; 95% CI 0.89–0.92), but weakly correlated with height (r=0.09; 95% CI 0.01–0.17).

MALARIA

The prevalence of prematurity, IUGR, and malaria infection of both the placenta and maternal blood at delivery were significantly higher in primigravidae than multigravidae (table 1). Although numbers were small, no evidence of a decrease in birthweight or increase in preterm deliveries was observed in gravida ⩾6. Median (IQR) birthweight was 2.85 kg (2.50–3.10) when the placenta was infected, and when not, 2.98 kg (2.67–3.27; p<0.0001). Median maternal haemoglobin after delivery when the placenta was infected measured 101 g/l (91–110) and when not 107 g/l (94–118; p<0.001). Median maternal haemoglobin after delivery was similar in each gravida group, although there was a tendency for more primigravidae than multigravidae to have haemoglobin values <100 g/l (41.6 vs 35.9%; p=0.10). Placental infection was highly correlated with maternal peripheral blood infection (p<0.001) and the differences in birthweight and maternal haemoglobin were similar if maternal infection rather than placental infection was analysed. Haemoglobin during pregnancy tended to decline as gestation increased (n=392; r= -0.10 (95% CI −0.20 to 0.00). Haemoglobin values during pregnancy cannot be compared with values after delivery as different methods of measurement were used.

The prevalence of placental and maternal peripheral blood parasitaemia was similar in those who had and had not taken chloroquine during pregnancy (186/791 (23.5%) vs20/68 (29.4%), p=0.35 and 167/900 (18.6%)vs 15/76 (19.7%), p=0.92, respectively). The density of parasitaemia in each site was not related to chloroquine prophylaxis. However, median (IQR) maternal haemoglobin concentrations after delivery were significantly greater in women who gave a history of chloroquine prophylaxis (n=862, 105 g/l (93–117)) than in those who did not (n=61, 99 g/l (82–107); p=0.0002). The usage of chloroquine and iron/folate was highly correlated (p<0.001) and similar differences in haemoglobin concentrations were observed if iron/folate rather than chloroquine usage was analysed.

DETERMINANTS OF PRETERM DELIVERY AND BIRTHWEIGHT IN TERM INFANTS

In univariate analysis both preterm delivery and IUGR were significantly associated with lower BMI, lower gravidity, the mother smoking during pregnancy, and working in a garden (table 2). Paternal employment in skilled work was negatively associated with prematurity and IUGR. Other predictors of prematurity were reduced maternal haemoglobin both during and after pregnancy and no chloroquine and iron/folate. Gestation (beyond 37 weeks) had the strongest association with birthweight in term infants (table 2) (fig 1). The mother or father in paid employment, greater maternal height, and a greater number of antenatal clinic visits were associated with increased birthweight in term infants whereas malaria infection at delivery were associated with IUGR. Maternal age was not significantly associated with birthweight when gravidity was included as a confounder. Only haemoglobin measured ⩽28 weeks gestation were included in the analysis of haemoglobin during pregnancy and these measurements had been performed at a similar gestation in preterm and term infants (median 22.5 and 23.0 weeks, respectively; p=0.62). The gestation at which maternal weight had been measured was also similar for preterm and term infants (median 22.9 and 20.9 weeks respectively; p=0.22). Median (IQR) birthweight was significantly greater in boys (2.99 kg (1.31–4.66) than girls (2.90 (1.90–4.22); p=0.0076).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Birthweight according to gestation (987 infants). Median values are indicated by dense band; box represents interquartile range, with whiskers extending up to 1.5 × IQR from upper or lower limits of box; open circles represent infants with birthweights beyond this range.

Birthweight was greater in the infants of women living in Madang town (median (IQR) 3.03 kg (2.69–3.36)) than in rural villages (2.88 (2.54–3.20)) or settlements (2.92 (2.63–3.24); p<0.001), but residence was not related to gestation (p=0.86). Birthweight was greater in the infants of Highland women (3.18 kg (2.76–3.49)) than all other women (p=0.011; other coastal regions 3.03 (2.70–3.38), Madang 2.89 (2.56–3.18), and Sepik 2.90 (2.62–3.21)). A greater proportion of Highland women lived in town (15, 47%) vs190 (21%), respectively; p=0.001) and a lower proportion had worked in gardens (13, 34%) vs 500 (53%), respectively; p=0.024). Median (IQR) gestation in women of Highland ethnicity was 39 (39–40) weeks and was similar to that in all other women (39 (38–40); p=0.22). Highland women had greater haemoglobin concentrations both during pregnancy and after delivery than other women—for example, median (IQR) haemoglobin after delivery 115 g/l (94–127) vs 105 g/l (92–116), respectively; p=0.001. Smoking was more common among Highland women, but all other factors, including anthropometric measurements and parasitological indices, were similar.

Variables with an independent and significant association with prematurity identified by forward stepwise logistic regression analyses (excluding BMI) are shown in table 3. When only the 466 records with a value for BMI were analysed, only maternal haemoglobin after delivery (OR=0.96, 95% CI 0.94–0.99; p=0.005) and BMI (OR=0.84, 95% CI 0.70–1.01; p=0.042) were significantly predictive of preterm delivery. Similarly, factors independently associated with birthweight are shown in table 4. The same factors were identified when BMI was omitted from the analysis and women without a BMI value were included. In multigravidae a previous miscarriage or stillbirth was also independently associated with birthweight (β=−127 g; 95% CI−222 to−31; p=0.009).

Discussion

Assessment of the impact of malaria on the outcome of pregnancy is complicated because the greatest incidence of infection occurs during the second trimester, and placental and peripheral blood parasitaemia may have resolved by the time of delivery.8 This may account for the apparent absence of an effect of malaria infection at delivery on birthweight in a small study of hospital deliveries in the Sepik region of Papua New Guinea.16 Malaria is an environmental factor which affects primigravidae more than multigravidae and the risk of low birthweight among primigravidae relative to multigravidae may reflect more accurately the impact of malaria during pregnancy than the presence of either placental or maternal peripheral blood parasitaemia at delivery.17 Both malaria infection at delivery and low birthweight were greatly increased in primigravidae compared with multigravidae in this study and this pattern is similar to that reported in African women.8 Overall, malaria infection of the placenta was associated with a reduction in birthweight of 130 g—a difference similar to that found in Africa8 and Thailand.18 Low gravidity was an independent predictor of both preterm delivery and IUGR in multiple regression analysis, suggesting that malaria infection contributed to both outcomes.

In contrast to developed countries, neonatal mortality in preterm infants in developing countries remains high11 19 and has been reported to be much higher in preterm than growth retarded infants.11 12 Although acute malaria has been observed to cause preterm labour,20 21 the few epidemiological studies which have considered malaria as a cause of prematurity have produced conflicting results. Maternal infection was not associated with preterm delivery in Thailand19 or in a small study in the Sepik region of Papua New Guinea.22 Although gestational age was not related to either placental or maternal parasitaemia at delivery in the Ivory Coast, maternal antibody titres against P falciparum were significantly higher in mothers of preterm than term infants.10 Placental and cord blood parasitaemia were more common in preterm than term infants in Malawi,23 and parasitaemia and maternal anaemia were associated with preterm delivery in Mozambique.24

In this study a third of all low birthweight infants were preterm and reduced maternal haemoglobin after delivery was the factor most strongly predictive of preterm delivery. Malaria is a major cause of maternal anaemia in this region and the highest rate of preterm deliveries occurred among anaemic women in a previous study in Madang.25 Hookworm infection, although common, does not cause anaemia in this population because worm burdens are low.26 Homozygous α⁺-thalassaemia is common in this population and, although associated with mild anaemia, it prevents severe malarial anaemia in children.27 The impact of α⁺-thalassaemia on maternal anaemia and the outcome of pregnancy will be reported separately.

The prevention of malaria increased maternal haemoglobin concentrations in several studies.28 Chloroquine prophylaxis increased haemoglobin in this study and in the previous study of Madang women.25 The finding that chloroquine prophylaxis was a significant independent predictor of delivery at term needs to be interpreted with caution as women delivering at term would have had greater opportunity to take this drug than those delivering before term. As observed before,29 chloroquine did not reduce either placental or maternal peripheral blood infection at delivery. The choice of an alternative antimalarial drug to chloroquine depends on several factors, and randomised control trials assessing various regimens have been reviewed recently.28 Additional strategies to prevent malaria during pregnancy, such as insecticide impregnated bed nets,30 and the targeting of interventions to primigravidae, should also be considered.

Poor maternal nutritional status was reported to be associated with reduced birthweight in previous studies in the Highlands31and coastal regions of Papua New Guinea.32 33 In this study lower BMI was an independent predictor of both preterm delivery and IUGR and lower maternal height were also independently associated with IUGR. BMI serves as a measure of chronic energy deficiency and, as expected, was highly correlated with maternal weight but not height.34 Maternal malnutrition was probably underestimated in this study because BMI was calculated from weight measured during pregnancy rather than from pre-pregnancy weight which would be expected to be lower. Low maternal height is a consequence of growth failure during childhood and adolescence.

Gestation was the strongest predictor of increased birthweight in term infants and this suggests that interventions which prolong gestation may increase birthweight in term infants as well as reduce preterm delivery. Greater birthweight among Highland women delivering on the coast than indigenous coastal mothers has been noted before and considered to be due, in part, to longer gestation.31 32Gestation at delivery was not related to ethnic origin in this study and better intrauterine growth in Highland women was likely to be a consequence of their more favourable economic status. Around 70% of women living in Madang villages receive at least some antenatal care.25 Antenatal clinic attendance was negatively associated with IUGR and this association was independent of chloroquine prophylaxis and markers of family wealth. Previous studies have shown increased perinatal mortality among infants of women who do not receive any antenatal care32 and lower average birthweight in village than hospital deliveries.22 Smoking has been identified as the most important single cause of low birth weight in developed countries9 and was significantly and independently associated with both preterm delivery and IUGR in this study.

In summary, we have identified several determinants of prematurity and IUGR which could be modified by community health services in Madang Province in an attempt to reduce infant mortality. Integrated antenatal care should be promoted and focus on effective measures to prevent malaria, especially important in primigravidae and for the prevention of pregnancy anaemia. Improving nutritional status in girls and young women, and during pregnancy, and discouraging smoking, would also be expected to prevent low birthweight in this population.