Pathways of association between maternal haemoglobin and stillbirth: path-analysis of maternity data from two hospitals in England.

OBJECTIVE: To investigate the mechanisms that link maternal haemoglobin concentration with stillbirth.

DESIGN: A retrospective cohort analysis using anonymised maternity data from two hospitals in England.

SETTING: The Royal Wolverhampton NHS Trust and Guy's and St Thomas' NHS Foundation Trust.

STUDY POPULATION: 12 636 women with singleton pregnancies ≥24 weeks of gestation giving birth in the two hospitals during 2013-2015.

METHOD: A conceptual framework of hypothesised pathways through birth weight-for-gestational age and maternal infection including potential confounders and other risk factors was developed and examined using path-analysis. Path-analysis was performed by fitting a set of regression equations using weighted least squares adjusted for mean and variance. Goodness-of-fit indices were estimated.

MAIN OUTCOME MEASURES: Coefficient of association (β) for relationship between each parameter, and direct, indirect and total effects via the postulated pathways.

RESULTS: The path-model showed a significant adjusted indirect negative effect of maternal haemoglobin on stillbirth mediated via birth weight-for-gestational age (standardised estimate (SE)=-0.01; 95% CI=-0.01 to -0.001; P=0.028). The effect through maternal infection was not significant at P<0.05 (SE=0.001; 95% CI=-0.004 to 0.01; P=0.610). There was a residual direct negative effect of maternal haemoglobin on stillbirth (SE=-0.12; 95% CI -0.23 to -0.02; P=0.020) after accounting for the two pathways. Total indirect SE=-0.004; 95% CI -0.01 to 0.003; P=0.267; total direct and indirect SE=-0.13; 95% CI -0.23 to -0.02; P=0.016. The goodness-of-fit indices showed a good fit between the model and the data.

CONCLUSION: While some of the influence on risk of stillbirth acts through low birth weight-for-gestational age, the majority does not. Several new mechanisms have been suggested for how haemoglobin may be exerting its influence on the risk of stillbirth possibly involving genetic, epigenetic and/or alternative obstetric and nutritional pathologies, but much more research is needed.