Papers

Mothers' birth weight and survival of their offspring: population based study

BMJ 1997; 314 doi: https://doi.org/10.1136/bmj.314.7091.1376 (Published 10 May 1997) Cite this as: BMJ 1997;314:1376
  1. Rolv Skjærven, professora,
  2. Allen J Wilcox, branch chiefb,
  3. Nina øyen, associate professorc,
  4. Per Magnus, professord
  1. a Section for Medical Informatics and Statistics University of BergenArmauer Hansens Building N-5021 Bergen Norway
  2. b Epidemiology Branch National Institute of Environmental Health Sciences PO Box 12233 Research Triangle Parkc NC 27709 USA
  3. c Department of Public Health and Primary Health Care University of Bergen Armauer Hansens Building N-5021 Bergen Norway
  4. d Department of Epidemiology National Institute of Public Health PO Box 4404 Torshov N-0403 Oslo Norway
  1. Correspondence to: Dr Skjærven
  • Accepted 24 January 1996

Abstract

Objective: To test the hypothesis that a baby's survival is related to the mother's birth weight.

Design: Population based dataset for two generations.

Setting: Population registry in Norway.

Subjects: All birth records for women born in Norway since 1967 were linked to births during 1981-94, thereby forming 105 104 mother-offspring units.

Main outcome measures: Perinatal mortality specific for weight for offspring in groups of maternal birth weight (with 500 g categories in both).

Results: A mother's birth weight was strongly associated with the weight of her baby. Maternal birth weight was associated with perinatal survival of her baby only for mothers with birth weights under 2000 g. These mothers were more likely to lose a baby in the perinatal period (odds ratio 2.3, 95% confidence interval 1.4 to 3.7). Among mothers with a birth weight over 2000 g there was no overall association between mother's weight and infant survival. There was, however, a strong interaction between mother's birth weight, infant birth weight, and infant survival. Mortality among small babies was much higher for those whose mothers had been large at birth. For example, babies weighing 2500-2999 g had a threefold higher mortality if their mother's birth weight had been high (≥4000 g) than if the mother had been small (2500-2999 g).

Conclusion: Mothers who weighed less than 2000 g at birth have a higher risk of losing their own babies. For mothers who weighed ≥2000 g their birth weight provides a benchmark for judging the growth of their offspring. Babies who are small relative to their mother's birth weight are at increased risk of mortality.

Key messages

  • A mother's birth weight influences perinatal survival of her own babies but only for mothers with birth weight <2000 g

  • There was a strong interaction between mothers' birth weight, infant birth weight, and infant survival

  • Babies that are small relative to their mother's birth weight are at increased risk of perinatal death

Introduction

Low birth weight is strongly associated with poor perinatal survival and may be related to poor health outcomes much later in life.1 2 A woman's birth weight is known to be directly related to the weight of her babies.3 4 5 It is unclear, however, whether a mother's birth weight has any predictive value for the survival of her babies. We evaluated the direct relation between maternal birth weight and the perinatal survival of her babies. We also investigated the influence of maternal birth weight on the baby's perinatal survival specific for birth weight.

Subjects and methods

Data were extracted from the Medical Birth Registry of Norway, a population based registry that compiles data on all births in Norway. The registry started in 1967 and comprises data on over 1.6 million births–more than one third of Norway's total population.

In Norway a unique personal number is given at birth to all citizens. Thus, it is possible to link the birth records of women born early in the registry period to the birth records of their children.

All birth records of women born in Norway in 1967 or later were linked to the birth records for the period 1981-94, thereby forming 105 104 units of mother and child (maternal age 14-28 years). Most of the babies (71%) were first births. These data were previously analysed with data up to 1989, comprising 18 342 mother-offspring units.5

Perinatal mortality was defined as all registered stillbirths plus live born babies who died within the first week of life. Stillbirths are reportable to the Medical Birth Registry of Norway from 16 weeks of gestation, and gestational age is calculated from the last menstrual period.

Linked units where the mother or the child was a twin, triplet, etc, were excluded from the analysis, leaving 101 579 mother-offspring units. Birth weight of the baby was missing in 0.10%, and birth weight of the mother was missing in 0.22%.

Logistic regression analyses were carried out by using BMDP.6 All odds ratios were calculated with the lr program in BMDP, and χ2 tests (maximum likelihood ratio tests) were used to test associations. Most analyses were done with categorical variables. As part of our data comprised non-independent births to the same mother the subset of mothers with first and second births (25 463 mother-offspring units) were analysed with generalised linear models for dependent data by using Oswald7 as available through S-PLUS.8

Results

Perinatal mortality by maternal birth weight

Table 1 shows perinatal mortality in the babies by categories of maternal birth weight. Mortality increased threefold among the babies of mothers who weighed less than 1500 g and twofold among babies of mothers who weighed 1500-1999 g at birth compared with the babies of mothers with higher birth weights. Among the maternal birth weights of ≥2000 g there were hardly any differences in mortality of the babies. Figure 1 shows the effects of maternal birth weight, with 3500-3999 g as reference category, as odds ratios (95% confidence intervals).

Table 1

Perinatal mortality by maternal birth weight. Data from Medical Birth Registry of Norway

View this table:
Fig 1
Fig 1

Perinatal mortality by eight categories of maternal birth weight expressed as odds ratios with 95% confidence intervals; reference category 3500-3999 g

Perinatal mortality specific for weight in groups of maternal birth weight

Table 2 presents data on all registered perinatal deaths in the second generation, stratified by birth weight of both mother and child (500 g categories). Although some groups had few or no deaths, a pattern of differentiated mortality by maternal birth weight emerged. Among mothers with birth weight ≥2000 g, weight specific mortality of small babies increased with increasing maternal birth weight. This was the case for every category of babies weighing <3500 g, which is close to the mean birth weight in Norway. For example, babies weighing 2500-2999 g had a threefold higher mortality if their mother's birth weight had been high (≥4000 g) than if the mother had been small (2500-2999 g).

Table 2

Perinatal mortality specific for weight by categories of maternal birth weight. Data from Medical Birth Registry of Norway

View this table:

Smaller than average babies

Using logistic regression we analysed perinatal mortality by maternal weight among babies with birth weight <3500 g (for categories, see table 2). Table 3 summarises the results presented as odds ratios (95% confidence intervals). The corresponding odds ratios for all babies are also given for comparison.

Table 3

Odds ratios (95% confidence intervals) for effects of maternal birth weight on perinatal mortality among all babies and among those with birth weight < 3500 g. Data from Medical Birth Registry of Norway

View this table:

Unlike the pattern for births as a whole, there was a strong association between mothers' birth weight and survival of babies weighing <3500 g, even after we excluded the lowest category of maternal birth weight (χ2 32.4; 4 df; P<0.0001). The association was strengthened after we adjusted for babies' birth weight (χ2 48.0; 4 df; P<0.0001). Thus, the effect of maternal birth weight among babies with birth weight <3500 g was considerably stronger than in the total.

Figure 2 shows the perinatal mortality specific for weight by categories of maternal birth weight. (Rates based on fewer than two deaths are not shown.) Although we used categorical variables in the models (table 3), the predicted rates followed a near linear trend for the maternal weight categories for ≥ 2000 g. The estimated regression coefficient corresponded to an odds ratio of 1.29 (1.20 to 1.39) for each 500 g increase–that is, an estimated odds ratio of 2.8 for perinatal mortality for equal sized babies for mothers with birth weight ≥4000 g v 2000-2499 g at birth.

Fig 2
Fig 2

Observed perinatal mortality (per 1000 births) by maternal birth weight in categories of birth weight of offspring

Only the first category of maternal birth weight (<2000 g) departed from this trend. Practically all deaths (15 of 17; table 2) for mothers with birth weight <2000 g were among babies who themselves weighed <2000 g. Of the 15 babies who died, 11 were preterm (<37 weeks' gestation).

We could not find any effect of maternal birth weight when the babies weighed above the mean (3500 g) (χ2 1.1; 4df; P = 0.89). We found a significant interaction (P = 0.04) when we compared the regression slopes of the four groups with birth weight <3500 g with this last category.

An analysis of the effect on perinatal mortality of maternal birth weight in families with two births by using generalised linear models,7 8 modelling the non-independence between the births, gave an odds ratio of 1.22 (1.08 to 1.36) for each 500 g increase. In this analysis we excluded mothers with birth weight <2000 g. Also, in an analysis of second births with adjustment for birth weight of the first birth the effect of maternal birth weight persisted (1.33 (1.13 to 1.57) for each 500 g).

Importance of relative birth weight

A mother's birth weight is a strong predictor of her child's birth weight. In our data the birth weight of the babies increased an average of 28 g for every 100 g of maternal birth weight. This linear relation seems to apply well for maternal birth weights ≥ 2000 g. As a consequence, a baby of 3000 g might be on the 30th centile for mothers who had themselves been small at birth but only on the 9th centile for mothers who were large at birth (table 4). Thus, at exactly the same birth weight, one baby might be small for gestational age while another might be appropriate for gestational age and with correspondingly lower mortality, all depending on the birth weight of the mother. This large difference in relative birth weight among babies of the same absolute birth weight presumably explains the large differences in mortality by maternal birth weight.

Table 4

Centiles for birth weight of babies by categories of maternal birth weight. Data from Medical Birth Registry of Norway

View this table:

Discussion

Adverse generational effect

Perinatal events are linked through generations; a baby's birth weight is strongly associated with the mother's birth weight.3 4 5 Previous studies have suggested the possibility of higher mortality among offspring of mothers born small, but those studies were based on small numbers without significant results.9 10 A recent study reported an association between maternal birth weight and infant survival for black but not white subjects.11 In this large registry of Norwegian births we found that a mother's birth weight directly predicts the survival of her offspring if the mother was very small at birth (<2000 g). This may provide an example of an “adverse generational effect” in which prenatal experience affects a woman's subsequent reproductive success.12 Most perinatal deaths in babies of mothers who had been small at birth occurred among babies with similarly low birth weight (<2000 g). When the baby is heavier the prognosis is better: only two of the overall 17 losses to mothers who had weighed <2000 g were among babies weighing ≥2000 g.

Deviation of birth weight from expected weight

In addition to the effect of very low maternal birth weight on survival of the baby, there is an effect in the opposite direction. Mothers who were heavier at birth were more likely to lose their baby at any given weight below the mean. Among these smaller babies there is about a threefold higher relative risk as mother's birth weight increases. This suggests that a baby's survival is related not only to its absolute birth weight but also to its weight relative to its “expected” weight.

The higher risk among small babies of heavier mothers is counterbalanced by the smaller number of low weight babies born to mothers who had been large at birth. Thus, the overall perinatal risk to babies of heavier mothers is not increased.

Mothers who were large at birth tend to have children who are large at birth. If a woman with high birth weight has a small baby, that baby is small not only in absolute terms but also in relation to its expected size. Given two babies with the same small birth weight the one that is also relatively small–that is, in relation to mother's birth weight–has the higher risk of mortality. These data show the importance of relative birth weight–that is, weight relative to the expected weight for that infant–in addition to absolute birth weight.13 14 This concept of relative birth weight adds another dimension to the interpretation of birth weight. In our study mothers' birth weight was a strong predictor of an infant birth weight, and therefore infant survival depends in part on how small the baby is in relation to its weight as predicted by the mother's birth weight.

While the concept of relative birth weight may not be familiar to clinicians, it is in fact implicit in the widely used “standards” for birth weight by gestational age. Weight by gestational age is used to identify infants who are small for gestational age for special attention and follow up during pregnancy and after birth. “Small for dates” is defined by centiles, which are units of relative weight. Moreover, the criteria for small for dates are commonly adjusted for factors that are predictive of birth weight, such as parity or sex of the infant.15 Maternal birth weight is a much stronger predictor of infant weight than either parity or sex and thus could be valuable in setting additional criteria for small for dates. What is “relatively small” for a mother who was small at birth herself is not the same for a mother who was large at birth. Accordingly, the criteria for small for dates differ among women, and this notion of relative birth weight, or heterogeneity in expected birth weight between women, can be extended to standards for fetal growth. In a previous study data on weight by gestational age that used the Norwegian medical birth registry16 we proposed an adjustment of standards of small for dates according to the weights of previous children to the same mother. This suggestion stemmed from the high correlation in birth weight between siblings born to the same mother. While this adjustment is useful, it is not applicable to first born infants. Mother's birth weight may be useful not only to identify “small” infants born to mothers who are expected to deliver larger babies but also to reduce the number of “false positive” cases of small for dates among mothers expected to deliver smaller babies.

Biological mechanisms

The biological mechanisms by which a mother's birth weight affects her infant's weight are not known. Broadly, three sets of factors are possible.

Firstly, environmental influences (for instance, diet and smoking) may persist across generations.

Secondly, genetic factors may contribute to normal variability in growth. Reports of a parallel association between father's and baby's birth weight3 17 suggest that fetal genes have a role in the correlation between birth weights of parents and children. In a study of birth weights of babies of twins, fetal genes accounted for more than half of the total variance.18

Thirdly, a set of factors may operate through maternal body size or metabolism.17 19 Sanderson et al found that adjustment for maternal height and weight in pregnancy reduced, but did not eliminate, the risk for infant death associated with low maternal birth weight.11 These data were not available in the Norwegian material. Consideration of these mechanisms is important for understanding the normal variability in fetal growth and judging the effects of relative birth weight on morbidity and mortality.

These same maternal and paternal factors that affect babies' birth weight presumably contribute to the correlation of birth weights among siblings.20 21 The correlation between birth weights of siblings (r = 0.50)22 is considerably greater than seen here between maternal and infant weights (r = 0.21). Not surprisingly, a baby's birth weight relative to its sibling's weight is also important to its risk of mortality.23 At any given birth weight, weight relative to sibling weight makes a strong additional contribution to survival. This effect is especially apparent among babies weighing <3000 g, similar to the present analysis.23 Wang et al recently showed that maternal birth weight and the birth weight of a previous sibling are at least partial independent predictors of low birth weight (<2500 g).24 This result is confirmed in our larger study. Especially, we found the odds ratios for low birth weight in second births when only the mother was of low birth weight, only the first sibling was of low birth weight, and both the mother and the first sibling were of low birth weight were 2.1, 6.1, and 12.4, respectively. The perinatal mortality among the low birth weight infants in families without previous low birth weight, however, was higher than in the three family types with low birth weight (odds ratio 1.3).

Barker has proposed that intrauterine growth is related to cardiovascular and other diseases in the adult.1 2 The excess perinatal mortality among babies of mothers born small may be an example of early life causes, where prenatal experience may affect later health.25 The aspects of maternal birth weight that lead to higher perinatal mortality of her babies may also contribute to a higher risk of adult diseases, but other explanations (as discussed above) are possible.

Population based dataset

Our data come from a large population based birth registry. Mothers and babies are linked by unique personal numbers, which do not rely on the mothers' recall. In other studies non-responders are common among categories of mothers with high risk for an adverse outcome of pregnancy.9 As our data require the mother to have been born in Norway the increased immigration to Norway during the past decade does not influence the data. The distribution of maternal age, however, may represent a bias. Among the linked births we identified mother-offspring units only for the mothers born in 1967 or later, thus excluding babies of mothers older than 28 years. Adjustment for five year categories of maternal age, within the available range, did not change the results. As 28% of mothers were included more than once in this dataset (because they had more than one birth) not all mother-offspring units were independent.Modelling this non-independence gave no substantive effect on the rates or confidence intervals. If anything, the effects of maternal birth weight among the smaller babies were stronger among second births than among first births.

Exclusion of early stillbirths between 16 and 21 weeks (or birth weight <500 g) slightly reduced the direct effect of a perinatal loss for the smaller mothers (birth weight <2000 g) from 2.31 to 2.26 (odds ratios with 3500-3999 g as reference). If anything, the relative effect of birth weight–that is, as reported in table 3–was slightly increased.

Variation in mortality

Despite our large sample, some groups of maternal and child birth weight provided only sparse data for perinatal mortality. Variation in mortality may be more complicated than the modelled results suggest. Still, mothers who weighed <2000 g at birth are at substantially increased risk of losing their own baby. Furthermore, high maternal birth weight is indicative of a large baby, and small babies born to those mothers are at higher risk than their weight alone would suggest. To the extent that a woman's birth weight can be obtained easily and accurately, it may offer clinically useful information on the risk of mortality of her babies. The biological pathways by which maternal birth weight affects the weight of her babies, however, need to be better understood.

Acknowledgments

We thank Drs Clarice Weinberg and Mathew Longnecker for helpful comments and Dr Håkon Gjessing for help with the S-PLUS analyses.

Funding: Norwegian Research Council.

Conflict of interest: None.

References

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
View Abstract