This study demonstrates that weight change during the inter-pregnancy interval is strongly associated with the risk of experiencing a range of pregnancy complications in the second pregnancy. This was true for women experiencing particular complications for the first time and for all women irrespective of their previous first pregnancy history. The data herein for women encountering a complication for the first time in the second pregnancy can be directly compared with the previous ten-fold larger population based study of Villamor & Cnattingius . For women who gained 3 or more BMI units, this comparison reveals a striking similarity between studies in the adjusted odds ratios for major pregnancy complications including pre-eclampsia (1.85 vs 1.78), gestational hypertension (1.82 vs 1.76), and LGA (1.70 vs 1.87), and a comparable OR for emergency caesarean delivery (present study) compared with all caesareans combined (1.78 vs 1.32). In contrast to previous studies [15, 20], inter-pregnancy weight change was not associated with the risk of stillbirth but our inability to detect a relationship may partly be due to a combination of the generally low incidence rate of this complication and the considerably smaller size of our study population. In the present study, inter-pregnancy weight loss was associated with a 46% higher risk of spontaneous preterm delivery while a large BMI gain was modestly protective (35% lower risk). Both the direction and magnitude of these effects were very similar to those reported previously , albeit examining BMI changes in a different way and in a large study population of more than 200,000 women. In the latter study which used self-reported weight and height data, moving from a normal to an underweight BMI classification between pregnancies was associated with a 50% higher risk of spontaneous preterm delivery while moving from normal to overweight or obese categories was associated with a 20% lower risk of an early birth in both cases. In addition to confirming the general robustness of these previous observations, herein for the first time we demonstrate that inter-pregnancy weight loss is associated with a 65% higher risk of an SGA birth. Although not directly comparable, weight loss before pregnancy, calculated as an annual average between 20 years of age and start of pregnancy approximately 9 years later, has recently been shown to increase the risk of an SGA delivery by 76% .
We additionally examined whether the above relationships between inter-pregnancy weight change and the risk of adverse pregnancy outcomes were dependent on a women’s baseline BMI at the first pregnancy. In general the risks associated with weight gain (namely gestational hypertension, caesarean delivery and LGA birth) were higher in women who had a healthy BMI (<25) during the first pregnancy than for overweight women, emphasizing that individuals do not necessarily have to become obese per se to increase their risk of adverse outcome at the second pregnancy. The exception was the risk of pre-eclampsia which was three-fold higher in women who were overweight at baseline but not different from the weight stable reference group in women who had a healthy weight at this time. This is in sharp contrast to previous data indicating that the effect of inter-pregnancy BMI change on pre-eclampsia was independent of baseline BMI at the first pregnancy . As the approach to the data analysis was identical the reason underlying this discrepancy is unknown but may relate to subtle differences in the diagnosis and hence classification of pre-eclampsia between data bases, or to the relatively higher percentage of women in our population with a large inter-pregnancy BMI gain (3% more in both initially healthy and overweight categories). With respect to those complications associated with inter-pregnancy weight loss, the risk of a SGA birth was independent of BMI at baseline, suggesting that any downward movement between BMI categories is likely to increase the risk of this complication. In contrast the risk of spontaneous preterm delivery following inter-pregnancy weight loss was apparent only in women who had a healthy BMI at baseline. The lack of effect in women who were initially overweight (BMI >25) is in complete agreement with previous work where moving from obese to overweight or normal, and overweight to normal BMI categories did not alter the risk of spontaneous preterm delivery relative to the BMI stable normal reference group .
The public health implications of these findings are considerable. The natural assumption is that women who were obese at the start of the first pregnancy are more vulnerable to excessive weight gain during that pregnancy and in the subsequent postpartum period. While this was true for approximately 25% of our study population, a considerable number of initially overweight and obese women (18 and 20%, respectively) lost weight between pregnancies thus theoretically negating the risk of many of the maternal and perinatal complications associated with weight gain but increasing their risk of a SGA birth. In partial support a recent retrospective study involving more than 700,000 women has specifically shown that gestational weight loss protects against pre-eclampsia and emergency caesarean section but increases the risk of prematurity and SGA in all but the most severely obese women . We have no information on the underlying causes of weight change in either direction in the present study but likely candidates not controlled for in the analysis presented here include inappropriate gestational weight change, diet, physical activity, breastfeeding and socio-economic status. Further it is possible that the use of customized birth weight centiles (based on maternal height and weight at booking) to define birth weight extremes may subtly alter the reported relationships. While the consequences of large inter-pregnancy weight gains are arguably more serious for a women’s long term health our results suggest that a degree of caution is required if health professionals promote weight loss at this time, particularly as the effectiveness of dietary and lifestyle intervention strategies aimed at improving pregnancy outcome among both normal and obese women remain largely unproven [30, 31].
Together the present and previous studies linking inter-pregnancy BMI change in both directions with a range of contrasting pregnancy complications implies a causal relationship, although we cannot determine whether it is weight change or the behaviours leading to it that confer extra risk. Nevertheless the present study further suggests that some of these relationships may be mediated in part by the placenta. Placental size, morphology, blood flow and nutrient transport functions primarily determine the growth trajectory of the fetus  and placental and fetal weight are strongly correlated . Thus in light of the relationship between inter-pregnancy weight loss and SGA (1.76), and between weight gain and LGA ( and this study, OR 1.87 and 1.83, respectively), it is perhaps unsurprising that our analysis also reveals that weight loss was associated with a similarly higher risk of low placental weight (OR 1.79), and weight gain with a greater risk of having a large placenta (OR 1.76). The implication is that these changes in maternal nutritional status between successive pregnancies impact maternal nutrient reserves at the start of the second pregnancy and hence the placental growth trajectory. Alterations in placental growth may in turn be on the causative pathway to some of the pregnancy complications studies herein. In partial support, studies in a highly controlled sheep model demonstrate that placental weight is profoundly influenced by BMI at conception (irrespective of subsequent gestational intake) and is closely associated with birth weight at delivery . Moreover in term infants, placental weight has recently been shown to be an important intermediary between maternal conditions of overnutrition (namely pre-pregnancy obesity, excessive gestational weight gain and gestational diabetes mellitus) and increased birth weight .
A considerable strength of our study was that weight and height were measured at the first clinic appointment on both occasions and weight was adjusted to a standard gestational age for all maternities. This was a close approximation of pre-pregnancy BMI as it was recorded by clinically trained staff thereby negating recall bias. Moreover the data for the maternal weight corrections was originally derived from women with singleton live births between 32 and 42 weeks gestation from the same geographical area , and while arguably sub-optimal for women with stillbirth or post-term delivery is considerably better than no adjustment. Both maternities were at a single hospital and a number of known important confounders were measured and the analyses adjusted accordingly. A further strength is the uniform criteria used in the AMND to record pregnancy complications. On the other hand the study population after exclusions for missing data was relatively small and of low ethnic diversity. Moreover the low event rate for rarer complications such as stillbirth may have limited the power to detect all potential effects. Nevertheless the missing data do not appear to have introduced population bias as the proportion of women per initial BMI category at first pregnancy, the frequency of pregnancy complications in relation to first pregnancy BMI, and the relationship between inter-pregnancy BMI change and the primary incidence of specific pregnancy complications is commensurate with previous publications by other groups using other data sources [5–9, 15–22]. Moreover the data was collected over a time range where changes in obstetrical practice might have been expected (eg. criteria for caesarean section) but this potential bias should have been minimized by including year of delivery in the adjusted model. Although the inter-delivery interval was similar between baseline BMI categories, it is acknowledged that the availability of this parameter in years rather than months is not ideal and may have limited the accuracy of adjusting for this parameter in the model.