Traces of parenthood but not pregnancy loss in UK Biobank structural brain MRI data

Abstract

Pregnancy induces neuroanatomical changes in the human brain. Earlier studies detected traces of motherhood decades after childbirth. It is unclear whether these reflect persisting traces of pregnancy or effects of parenthood. We investigated effects of past birth and of pregnancy loss in women, and effects of fatherhood in men, using univariate and machine learning analyses on 205 regional brain volumes. A group of mothers and an age-matched sample of nulliparous women (N=4357 per group, mean age 63 years) from the UK Biobank, with no past pregnancy losses, showed significant volumetric group differences in 14 regions at Bonferroni-adjusted =.05. Likewise, we identified 18 significant group differences between age-matched samples of fathers and non-fathers of the same size (mean age 63.4), with 9 regions overlapping between sexes. Brain-wide association statistics for past live birth in mothers and those for fatherhood correlated (r = 0.55). XGBoost machine learning models trained to classify parenthood status separately in both datasets showed performance that was low, but significantly above chance (10-fold cross validation: AUC=0.56, p<1e-5 Motherhood classifier, AUC=0.54, p<1e-5, Fatherhood classifier, 10k permutations). We tested the motherhood classification model on an independent test sample comprising of four age-matched groups: 1. women who have never been pregnant, 2. women with past pregnancy loss but no live births, 3. women with live births but no pregnancy loss, and 4. women who experienced both. Class probability was significantly associated with live births, but not past loss. These findings may suggest that neuroanatomical patterns of past childbirth partly also reflect traces of parenthood and not solely persisting traces of past pregnancy, although a more detailed characterization of pregnancy loss data would be needed for full confirmation of this interpretation. Therefore, further research is needed to quantify the extent and understand the nature of these changes, particularly considering the known vulnerability for mental disorders associated with reproductive events.