Childhood maltreatment and health in the UK Biobank: triangulation of outcome-wide and polygenic risk score analyses

Ana Lucia Espinosa Dice; Rebecca B Lawn; Andrew Ratanatharathorn; Andrea L Roberts; Christy A Denckla; Ariel H Kim; Pedro A de la Rosa; Yiwen Zhu; Tyler J VanderWeele; Karestan C Koenen

doi:10.1186/s12916-024-03360-9

Childhood maltreatment and health in the UK Biobank: triangulation of outcome-wide and polygenic risk score analyses

BMC Med. 2024 Mar 25;22(1):135. doi: 10.1186/s12916-024-03360-9.

Authors

Ana Lucia Espinosa Dice^{1

2}, Rebecca B Lawn^#^{3

4}, Andrew Ratanatharathorn^#^{3

4

5}, Andrea L Roberts⁶, Christy A Denckla⁷, Ariel H Kim^{3

8}, Pedro A de la Rosa^{3

9}, Yiwen Zhu³, Tyler J VanderWeele^{3

10}, Karestan C Koenen^{3

4

7

11}

Affiliations

¹ Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, MA, USA. alespinosadice@hsph.harvard.edu.
² Stanley Center for Psychiatric Research, Broad Institute of MIT and Harvard, Cambridge, MA, USA. alespinosadice@hsph.harvard.edu.
³ Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, MA, USA.
⁴ Stanley Center for Psychiatric Research, Broad Institute of MIT and Harvard, Cambridge, MA, USA.
⁵ Department of Epidemiology, Columbia University Mailman School of Public Health, New York City, NY, USA.
⁶ Department of Environmental Health, Harvard T.H. Chan School of Public Health, Boston, MA, USA.
⁷ Department of Social and Behavioral Sciences, Harvard T.H. Chan School of Public Health, Boston, MA, USA.
⁸ Department of Molecular and Cellular Biology, Harvard University, Cambridge, MA, USA.
⁹ Institute for Culture and Society, University of Navarra, Pamplona, Spain.
¹⁰ Department of Biostatistics, Harvard T.H. Chan School of Public Health, Boston, MA, USA.
¹¹ Psychiatric Neurodevelopmental Genetics Unit, Department of Psychiatry, Massachusetts General Hospital, Boston, MA, USA.

^# Contributed equally.

Abstract

Background: Childhood maltreatment is common globally and impacts morbidity, mortality, and well-being. Our understanding of its impact is constrained by key substantive and methodological limitations of extant research, including understudied physical health outcomes and bias due to unmeasured confounding. We address these limitations through a large-scale outcome-wide triangulation study.

Methods: We performed two outcome-wide analyses (OWAs) in the UK Biobank. First, we examined the relationship between self-reported maltreatment exposure (number of maltreatment types, via Childhood Trauma Screener) and 414 outcomes in a sub-sample of 157,316 individuals using generalized linear models ("observational OWA"). Outcomes covered a broad range of health themes including health behaviors, cardiovascular disease, digestive health, socioeconomic status, and pain. Second, we examined the relationship between a polygenic risk score for maltreatment and 298 outcomes in a non-overlapping sample of 243,006 individuals ("genetic OWA"). We triangulated results across OWAs based on differing sources of bias.

Results: Overall, 23.8% of the analytic sample for the observational OWA reported at least one maltreatment type. Of 298 outcomes examined in both OWAs, 25% were significant in both OWAs and concordant in the direction of association. Most of these were considered robust in the observational OWA according to sensitivity analyses and included outcomes such as marital separation (OR from observational OWA, OR_o = 1.25 (95% CI: 1.21, 1.29); OR from genetic OWA, OR_g = 1.06 (1.03, 1.08)), major diet changes due to illness (OR_o = 1.27 (1.24, 1.29); OR_g = 1.01 (1.00, 1.03)), certain intestinal diseases (OR_o = 1.14 (1.10, 1.18); OR_g = 1.03 (1.01, 1.06)), hearing difficulty with background noise (OR_o = 1.11 (1.11, 1.12); OR_g = 1.01 (1.00, 1.01)), knee arthrosis (OR_o = 1.13 (1.09, 1.18); OR_g = 1.03 (1.01, 1.05)), frequent sleeplessness (OR_o = 1.21 (1.20, 1.23); OR_g = 1.02 (1.01, 1.03)), and low household income (OR_o = 1.28 (1.26, 1.31); OR_g = 1.02 (1.01, 1.03)). Approximately 62% of results were significant in the observational OWA but not the genetic OWA, including numerous cardiovascular outcomes. Only 6 outcomes were significant in the genetic OWA and null in the observational OWA; these included diastolic blood pressure and glaucoma. No outcomes were statistically significant in opposite directions in the two analyses, and 11% were not significant in either OWA.

Conclusions: Our findings underscore the far-reaching negative effects of childhood maltreatment in later life and the utility of an outcome-wide triangulation design with sensitivity analyses for improving causal inference.

Keywords: Childhood maltreatment; Outcome-wide analysis; Polygenic risk score; Triangulation; UK Biobank.

MeSH terms

Biological Specimen Banks
Child
Child Abuse*
Genetic Risk Score*
Humans
Self Report
UK Biobank

Abstract

MeSH terms

Grants and funding