Maternal pre-pregnancy body mass index and risk of preterm birth: a collaboration using large routine health datasets.

BACKGROUND: Preterm birth (PTB) is a leading cause of child morbidity and mortality. Evidence suggests an increased risk with both maternal underweight and obesity, with some studies suggesting underweight might be a greater factor in spontaneous PTB (SPTB) and that the relationship might vary by parity. Previous studies have largely explored established body mass index (BMI) categories. Our aim was to compare associations of maternal pre-pregnancy BMI with any PTB, SPTB and medically indicated PTB (MPTB) among nulliparous and parous women across populations with differing characteristics, and to identify the optimal BMI with lowest risk for these outcomes. METHODS: We used three UK datasets, two USA datasets and one each from South Australia, Norway and Denmark, together including just under 29 million pregnancies resulting in a live birth or stillbirth after 24 completed weeks gestation. Fractional polynomial multivariable logistic regression was used to examine the relationship of maternal BMI with any PTB, SPTB and MPTB, among nulliparous and parous women separately. The results were combined using a random effects meta-analysis. The estimated BMI at which risk was lowest was calculated via differentiation and a 95% confidence interval (CI) obtained using bootstrapping. RESULTS: We found non-linear associations between BMI and all three outcomes, across all datasets. The adjusted risk of any PTB and MPTB was elevated at both low and high BMIs, whereas the risk of SPTB was increased at lower levels of BMI but remained low or increased only slightly with higher BMI. In the meta-analysed data, the lowest risk of any PTB was at a BMI of 22.5 kg/m2 (95% CI 21.5, 23.5) among nulliparous women and 25.9 kg/m2 (95% CI 24.1, 31.7) among multiparous women, with values of 20.4 kg/m2 (20.0, 21.1) and 22.2 kg/m2 (21.1, 24.3), respectively, for MPTB; for SPTB, the risk remained roughly largely constant above a BMI of around 25-30 kg/m2 regardless of parity. CONCLUSIONS: Consistency of findings across different populations, despite differences between them in terms of the time period covered, the BMI distribution, missing data and control for key confounders, suggests that severe under- and overweight may play a role in PTB risk.

Multiple imputation using linked proxy outcome data resulted in important bias reduction and efficiency gains: a simulation study.

BACKGROUND: When an outcome variable is missing not at random (MNAR: probability of missingness depends on outcome values), estimates of the effect of an exposure on this outcome are often biased. We investigated the extent of this bias and examined whether the bias can be reduced through incorporating proxy outcomes obtained through linkage to administrative data as auxiliary variables in multiple imputation (MI). METHODS: Using data from the Avon Longitudinal Study of Parents and Children (ALSPAC) we estimated the association between breastfeeding and IQ (continuous outcome), incorporating linked attainment data (proxies for IQ) as auxiliary variables in MI models. Simulation studies explored the impact of varying the proportion of missing data (from 20 to 80%), the correlation between the outcome and its proxy (0.1-0.9), the strength of the missing data mechanism, and having a proxy variable that was incomplete. RESULTS: Incorporating a linked proxy for the missing outcome as an auxiliary variable reduced bias and increased efficiency in all scenarios, even when 80% of the outcome was missing. Using an incomplete proxy was similarly beneficial. High correlations (> 0.5) between the outcome and its proxy substantially reduced the missing information. Consistent with this, ALSPAC analysis showed inclusion of a proxy reduced bias and improved efficiency. Gains with additional proxies were modest. CONCLUSIONS: In longitudinal studies with loss to follow-up, incorporating proxies for this study outcome obtained via linkage to external sources of data as auxiliary variables in MI models can give practically important bias reduction and efficiency gains when the study outcome is MNAR.