The Weight of Trauma: Navigating Collider Stratification Bias in the Association Between Childhood Maltreatment and Adult Body Mass Index

In Canada, one-third of children do not enjoy a safe and healthy childhood.1Raising Canada. Top 10 Threads to Childhood in Canada and the Impact of COVID-19.; 2020. https://childrenfirstcanada.org/wp-content/uploads/2021/03/RaisingCanadaReport_Final_Sept.pdfGoogle Scholar This ranks Canada in the bottom 25% of affluent countries for child mental and physical well-being.2UNICEF Canada. Worlds Apart: Canadian Summary of UNICEF Report Card 16.; 2020. https://www.unicef.ca/sites/default/files/2020-08/UNICEF Report Card 16 Canadian Summary.pdfGoogle Scholar Childhood maltreatment, which encompasses all types of physical and emotional ill-treatment, sexual abuse, neglect, negligence, and exploitation, is a serious public health problem in Canada and globally.1Raising Canada. Top 10 Threads to Childhood in Canada and the Impact of COVID-19.; 2020. https://childrenfirstcanada.org/wp-content/uploads/2021/03/RaisingCanadaReport_Final_Sept.pdfGoogle Scholar,3World Health Organization. Child maltreatment. Published 2022. Accessed April 16, 2024. https://www.who.int/news-room/fact-sheets/detail/child-maltreatment#:∼:text=It includes all types of,of responsibility%2C trust or power.Google Scholar Experiencing childhood maltreatment can have enduring effects on health throughout the life course. These effects may include the adoption of risky health behaviours, poor mental and physical health, and chronic inflammation.4Hughes K. Bellis M.A. Hardcastle K.A. et al.The effect of multiple adverse childhood experiences on health: a systematic review and meta-analysis.Lancet Public Heal. 2017; 2: e356-e366https://doi.org/10.1016/S2468-2667(17)30118-4Abstract Full Text Full Text PDF PubMed Scopus (2665) Google Scholar,5Wiss D.A. Brewerton T.D. Adverse Childhood Experiences and Adult Obesity: A Systematic Review of Plausible Mechanisms and Meta-Analysis of Cross-Sectional Studies.Physiol Behav. 2020; 223112964https://doi.org/10.1016/j.physbeh.2020.112964Crossref PubMed Scopus (121) Google Scholar In turn, these factors can lead to increases in body weight5Wiss D.A. Brewerton T.D. Adverse Childhood Experiences and Adult Obesity: A Systematic Review of Plausible Mechanisms and Meta-Analysis of Cross-Sectional Studies.Physiol Behav. 2020; 223112964https://doi.org/10.1016/j.physbeh.2020.112964Crossref PubMed Scopus (121) Google Scholar, which may have downstream consequences, including the risk of adult obesity and other chronic diseases (e.g., type 2 diabetes, cancer, and cardiovascular disease).6Valenzuela P.L. Carrera-Bastos P. Castillo-García A. Lieberman D.E. Santos-Lozano A. Lucia A. Obesity and the risk of cardiometabolic diseases.Nat Rev Cardiol. 2023; 20: 475-494https://doi.org/10.1038/s41569-023-00847-5Crossref PubMed Scopus (56) Google Scholar,7Hruby A. Manson J.E. Qi L. et al.Determinants and Consequences of Obesity.Am J Public Health. 2016; 106: 1656-1662https://doi.org/10.2105/AJPH.2016.303326Crossref PubMed Scopus (450) Google Scholar Given the high prevalence of obesity in Canada, the identification of early-life contributing factors, such as childhood maltreatment, is needed to promote healthy development and reduce the burden of chronic disease.6Valenzuela P.L. Carrera-Bastos P. Castillo-García A. Lieberman D.E. Santos-Lozano A. Lucia A. Obesity and the risk of cardiometabolic diseases.Nat Rev Cardiol. 2023; 20: 475-494https://doi.org/10.1038/s41569-023-00847-5Crossref PubMed Scopus (56) Google Scholar,8Twells L.K. Gregory D.M. Reddigan J. Midodzi W.K. Current and predicted prevalence of obesity in Canada: a trend analysis.C Open. 2014; 2: E18-E26https://doi.org/10.9778/cmajo.20130016Crossref PubMed Google Scholar Understanding the relationship between childhood maltreatment and health outcomes is important to inform the development of social policies and programs that aim to ensure a safe and happy childhood for all children in Canada. In this issue of the Canadian Journal of Cardiology Pediatric & Congenital Heart Disease, St-Arnaud and colleagues conducted a secondary analysis of the BEL-AGE prospective cohort study to evaluate the association of childhood maltreatment with body mass index (BMI) among older adults with chronic illness.9St-Arnaud V. Chicoine A.X. Tardif J.C. Busseuil D. D’Antono B. Childhood maltreatment and body mass index in older adults with chronic illness.Can J Cardiol Pediatr Congenit Hear Dis. 2024; (Published online)Google Scholar Childhood trauma (encompassing physical, sexual, and emotional subtypes) was assessed upon enrollment, whereas BMI was measured at both baseline and 5 years after enrollment. This study included 1,232 older adults, with a mean age of 61 years. Nearly one-third (32.0%) of participants reported moderate to extremely severe scores on at least one sub-scale of the childhood maltreatment questionnaire, with 35.6% being classified as having obesity at the time of enrollment. Contrary to previous studies, childhood maltreatment was not significantly associated with BMI at baseline (β=0.004, 95% confidence interval [CI]: -0.006 to 0.014) or at follow-up (β=0.001, 95% CI: -0.003 to 0.006) after adjusting for various sociodemographic, lifestyle, and clinical factors. No effect modification of this relationship by sex or coronary artery disease status was evident. The authors leveraged data from a large Canadian cohort of older adults diagnosed with a chronic illness, thus capturing an underrepresented population within the child maltreatment research field. Using a validated questionnaire to assess childhood trauma, alongside the objective measurement of anthropometric parameters by trained clinical staff, greatly reduced the potential for differential misclassification.10Delgado-Rodriguez M. Llorca J. Bias.J Epidemiol Community Heal. 2004; 58: 635-641https://doi.org/10.1136/jech.2003.008466Crossref Scopus (812) Google Scholar To fully contextualize the findings of this study, it is essential to consider its limitations. Almost all included patients were White (98.5%), thus preventing the exploration of effect modification by race. This may also limit the generalizability of study findings to the racially diverse Canadian population. A review of child protective services reports has demonstrated ethnocultural variation in both reported and substantiated forms of maltreatment, which suggests that the association between childhood maltreatment and cardiometabolic health may vary across racialized groups.11Lavergne C. Dufour S. Trocmé N. Larrivée M.C. Visible minority, Aboriginal, and Caucasian children investigated by Canadian protective services.Child Welfare. 2008; 87: 59-76http://www.ncbi.nlm.nih.gov/pubmed/18972932PubMed Google Scholar Furthermore, the use of BMI to accurately reflect cardiometabolic health risks is an ongoing topic of debate.12Maximova K. Chiolero A. O’Loughliin J. Tremblay A. Lambert M. Paradis G. Ability of different adiposity indicators to identify children with elevated blood pressure.J Hypertens. 2011; 29: 2075-2083https://doi.org/10.1097/HJH.0b013e32834be614Crossref Scopus (28) Google Scholar,13Chiolero A. Paradis G. Maximova K. Burnier M. Bovet P. No use for waist-for-height ratio in addition to body mass index to identify children with elevated blood pressure.Blood Press. 2013; 22: 17-20https://doi.org/10.3109/08037051.2012.701376Crossref Scopus (23) Google Scholar This is largely due to its inability to comprehensively incorporate individual body composition, fat distribution, and metabolic profiles, which have been shown to differ according to age and race.14Caleyachetty R. Barber T.M. Mohammed N.I. et al.Ethnicity-specific BMI cutoffs for obesity based on type 2 diabetes risk in England: a population-based cohort study.Lancet Diabetes Endocrinol. 2021; 9: 419-426https://doi.org/10.1016/S2213-8587(21)00088-7Abstract Full Text Full Text PDF PubMed Scopus (153) Google Scholar, 15Heymsfield S.B. Peterson C.M. Thomas D.M. Heo M. Schuna J.M. Why are there race/ethnic differences in adult body mass index–adiposity relationships? A quantitative critical review.Obes Rev. 2016; 17: 262-275https://doi.org/10.1111/obr.12358Crossref PubMed Scopus (236) Google Scholar, 16Harris T.B. Weight and Body Mass Index in Old Age: Do They Still Matter?.J Am Geriatr Soc. 2017; 65: 1898-1899https://doi.org/10.1111/jgs.14952Crossref Scopus (6) Google Scholar, 17Danese A. Tan M. Childhood maltreatment and obesity: systematic review and meta-analysis.Mol Psychiatry. 2014; 19: 544-554https://doi.org/10.1038/mp.2013.54Crossref PubMed Scopus (510) Google Scholar Additional research to uncover racial disparities in the relationship between childhood maltreatment and cardiometabolic health, specifically using alternative weight-related metrics, will inform the design of tailored public health interventions. Although the authors found no statistically significant association between childhood maltreatment and adult BMI, this finding should be carefully interpreted in the context of sample selection. In particular, this result may be partially explained by collider stratification bias, which is a form of selection bias due to restricting a sample to one level of a specific variable known as a collider.18Lu H. Cole S.R. Howe C.J. Westreich D. Toward a Clearer Definition of Selection Bias When Estimating Causal Effects.Epidemiology. 2022; 33: 699-706https://doi.org/10.1097/EDE.0000000000001516Crossref Scopus (34) Google Scholar This underrecognized type of bias is commonly visualized through the use of directed acyclic graphs (DAGs). In brief, DAGs are graphical tools that are used to describe potential or known pathways between an exposure and outcome by connecting variables (known as nodes) with unidirectional arrows (known as edges). The use of DAGs to reduce bias through study design and analytical decisions is becoming increasingly popular across various scientific fields, including child maltreatment and cardiometabolic research.19Jaen J. Lovett S.M. Lajous M. Keyes K.M. Stern D. Adverse childhood experiences and adult outcomes using a causal framework perspective: Challenges and opportunities.Child Abuse Negl. 2023; 143106328https://doi.org/10.1016/j.chiabu.2023.106328Crossref Scopus (1) Google Scholar, 20Austin A.E. Desrosiers T.A. Shanahan M.E. Directed acyclic graphs: An under-utilized tool for child maltreatment research.Child Abuse Negl. 2019; 91: 78-87https://doi.org/10.1016/j.chiabu.2019.02.011Crossref PubMed Scopus (25) Google Scholar, 21Li H. Zheng C. Zhang Y. Yang H. Li J. The directed acyclic graph helped identify confounders in the association between coronary heart disease and pesticide exposure among greenhouse vegetable farmers.Medicine (Baltimore). 2023; 102e35073https://doi.org/10.1097/MD.0000000000035073Crossref Scopus (0) Google Scholar, 22Tennant P.W.G. Murray E.J. Arnold K.F. et al.Use of directed acyclic graphs (DAGs) to identify confounders in applied health research: review and recommendations.Int J Epidemiol. 2021; 50: 620-632https://doi.org/10.1093/ije/dyaa213Crossref PubMed Scopus (318) Google Scholar We refer the reader to the seminal article by Greenland et al. (1999) and other helpful pieces for an in-depth introduction to DAGs.23Greenland S. Pearl J. Robins J.M. Causal diagrams for epidemiologic research.Epidemiology. 1999; 10: 37-48Crossref PubMed Google Scholar, 24Glymour M. Using causal diagrams to understand common problems in social epidemiology. In: Oakes MJ, Kaufman JS, eds. Methods in Social Epidemiology. 2nd Editio. Jossey-Bass; 2006:393-428.Google Scholar, 25Digitale J.C. Martin J.N. Glymour M.M. Tutorial on directed acyclic graphs.J Clin Epidemiol. 2022; 142: 264-267https://doi.org/10.1016/j.jclinepi.2021.08.001Abstract Full Text Full Text PDF PubMed Scopus (68) Google Scholar, 26Shrier I. Platt R.W. Reducing bias through directed acyclic graphs.BMC Med Res Methodol. 2008; 8: 70https://doi.org/10.1186/1471-2288-8-70Crossref PubMed Scopus (999) Google Scholar Collider stratification bias refers to a mechanism in which a common effect of two unrelated variables collide at a third variable, referred to as the collider, creating a spurious statistical association between the two unrelated variables.27Banack H.R. Mayeda E.R. Naimi A.I. Fox M.P. Whitcomb B.W. Collider Stratification Bias I: Principles and Structure.Am J Epidemiol. 2024; 193: 238-240https://doi.org/10.1093/aje/kwad203Crossref Scopus (0) Google Scholar By conditioning on a collider, which is commonly achieved through adjustment, stratification, or restriction in observational studies, the causal effect of interest can be biased. Figure 1 provides the general and study-specific structure of collider stratification bias, delineating potential biasing pathways of the causal relationship of interest (X → Y). In the present study, the selection of older individuals with a chronic illness (including, but not limited to, cardiovascular disease, arthritis, diabetes, and asthma) into the study sample may have led to collider stratification bias, thus providing a potential explanation for the unexpected null finding between childhood maltreatment and adult BMI. According to DAG theory, all non-causal pathways connecting X to Y must be blocked to provide an unbiased effect estimate between X and Y (i.e., only the causal pathway X→Y should remain open and unblocked). Blocking a non-causal pathway can be achieved by conditioning on a variable along a non-causal pathway (which was performed using restriction in the present study). In our fictitious example (Figure 1A), the non-causal pathway X←L→Y can be blocked by conditioning on L, a known and measured confounder (or a common cause of both X and Y). In comparison, the non-causal pathway connecting X to Y that includes a collider C (X→C←U→Y) is not blocked when the collider is conditioned on.28Hernán M.A. Monge S. Selection bias due to conditioning on a collider.BMJ. 2023; 381p1135https://doi.org/10.1136/bmj.p1135Crossref Scopus (8) Google Scholar Instead, conditioning on a collider induces a spurious association between X and U, thus opening a non-causal pathway from X to Y (X--U→Y). In other words, a non-causal path that includes a collider is already blocked (and thus the collider should not be conditioned on), however, it can be opened by conditioning on that collider. Re-blocking this non-causal pathway would require conditioning on all common causes of the collider and Y, which may include various unknown and/or unmeasured variables (U). Therefore, after conditioning on both L and C in Figure 1A, two pathways connecting X to Y remain open: (1) the causal pathway (X→Y) and (2) the non-causal collider-induced pathway (X--U→Y). For this reason, conditioning on a collider through restriction (i.e., restricting a sample to a certain singular stratum of a collider) is expected to lead to selection bias, which can distort the direct causal effect of interest (X → Y).27Banack H.R. Mayeda E.R. Naimi A.I. Fox M.P. Whitcomb B.W. Collider Stratification Bias I: Principles and Structure.Am J Epidemiol. 2024; 193: 238-240https://doi.org/10.1093/aje/kwad203Crossref Scopus (0) Google Scholar,29Hernán M.A. Hernández-Díaz S. Robins J.M. A Structural Approach to Selection Bias.Epidemiology. 2004; 15: 615-625https://doi.org/10.1097/01.ede.0000135174.63482.43Crossref PubMed Scopus (1875) Google Scholar St-Arnaud and colleagues recognized this potential selection bias in their study, which is common in research on older adults (Figure 1B).30Banack H.R. Kaufman J.S. Wactawski‐Wende J. Troen B.R. Stovitz S.D. Investigating and Remediating Selection Bias in Geriatrics Research: The Selection Bias Toolkit.J Am Geriatr Soc. 2019; 67: 1970-1976https://doi.org/10.1111/jgs.16022Crossref Scopus (27) Google Scholar Individuals who experienced childhood maltreatment may have either died before study enrolment or reached an advanced stage of illness, rendering them unable to participate – selection bias due to death or illness. This is plausible given the well-known associations of childhood maltreatment with premature mortality and chronic diseases.31Power C. Li L. Pinto Pereira S.M. Child maltreatment (neglect and abuse) in the 1958 birth cohort: an overview of associations with developmental trajectories and long-term outcomes.Longit Life Course Stud. 2020; 11: 431-458https://doi.org/10.1332/175795920X15891281805890Crossref Scopus (14) Google Scholar,32Norman R.E. Byambaa M. De R. Butchart A. Scott J. Vos T. The Long-Term Health Consequences of Child Physical Abuse, Emotional Abuse, and Neglect: A Systematic Review and Meta-Analysis.PLoS Med. 2012; 9e1001349https://doi.org/10.1371/journal.pmed.1001349Crossref PubMed Scopus (2128) Google Scholar In the absence of conditioning on all common causes of chronic illness, survival to old age, and adult BMI (U in Figure 1B), the causal effect estimate of childhood maltreatment on BMI will be biased. While the authors controlled for various potential confounders (e.g., sociodemographic characteristics, lifestyle behaviors, clinical factors), adjusting for other unmeasured variables, such as childhood BMI and genetic factors (i.e., potential variables for U in Figure 1B), is necessary to obtain an unbiased measure of association according to this DAG. As with any study that evaluates life course epidemiologic relationships, obtaining accurate and repeated measurements of all unobserved variables poses a significant challenge. Nonetheless, restricting the study sample to individuals with a chronic illness who survived to old age may have biased (underestimated, in most cases) the true measure of association between childhood maltreatment and adult BMI due to collider stratification bias.27Banack H.R. Mayeda E.R. Naimi A.I. Fox M.P. Whitcomb B.W. Collider Stratification Bias I: Principles and Structure.Am J Epidemiol. 2024; 193: 238-240https://doi.org/10.1093/aje/kwad203Crossref Scopus (0) Google Scholar,33Weuve J. Tchetgen Tchetgen E.J. Glymour M.M. et al.Accounting for Bias Due to Selective Attrition: The Example of Smoking and Cognitive Decline.Epidemiology. 2012; 23: 119-128https://doi.org/10.1097/EDE.0b013e318230e861Crossref PubMed Scopus (350) Google Scholar In contrast to the results of this study, previous longitudinal investigations have reported on associations between experiences of childhood maltreatment and elevated risk of obesity during adulthood, even after controlling for confounders.34Power C. Pinto Pereira S.M. Li L. Childhood Maltreatment and BMI Trajectories to Mid-Adult Life: Follow-Up to in a Scopus Google I. A. et childhood experiences and life adult obesity and in the Epidemiol. 2016; PubMed Scopus Google Scholar For in a study of and older adults in the Canadian on during childhood a association with obesity measured using BMI and V. A. A longitudinal study adverse childhood experiences and obesity in using the Canadian on J Epidemiol. 2023; Scopus (1) Google Scholar This association has also been in systematic D.A. Brewerton T.D. Adverse Childhood Experiences and Adult Obesity: A Systematic Review of Plausible Mechanisms and Meta-Analysis of Cross-Sectional Studies.Physiol Behav. 2020; 223112964https://doi.org/10.1016/j.physbeh.2020.112964Crossref PubMed Scopus (121) Google A. Tan M. Childhood maltreatment and obesity: systematic review and meta-analysis.Mol Psychiatry. 2014; 19: 544-554https://doi.org/10.1038/mp.2013.54Crossref PubMed Scopus (510) Google K. S. of childhood on adult obesity: a systematic review and Rev. 2014; 15: PubMed Scopus Google Scholar The effects of and sex to vary across systematic review that the association between childhood maltreatment and obesity was in that included A. Tan M. Childhood maltreatment and obesity: systematic review and meta-analysis.Mol Psychiatry. 2014; 19: 544-554https://doi.org/10.1038/mp.2013.54Crossref PubMed Scopus (510) Google Scholar In comparison, review found no effect modification by K. S. of childhood on adult obesity: a systematic review and Rev. 2014; 15: PubMed Scopus Google Scholar study found no significant in the association between childhood maltreatment and BMI according to as by genetic H. A. Howe Investigating effect modification between childhood maltreatment and genetic risk for cardiovascular disease in the 2023; Scopus (0) Google Scholar life course epidemiologic to consider or critical of development can childhood maltreatment may lead to adverse health such as obesity in may consider trajectories of exposure to childhood maltreatment, which may to identify and understand potential for cardiometabolic health This is of as in older adults are recognized to be associated with and an elevated risk of S.B. et Weight in and of 2020; 9: PubMed Scopus Google T.A. S. D. J. R. The association of and mortality in older a systematic review and 2021; 50: PubMed Scopus Google Scholar should be in a that for potential by collider stratification the study design the development of DAGs to describe causal associations is for the of a collider, the of conditioning on this J. Lovett S.M. Lajous M. Keyes K.M. Stern D. Adverse childhood experiences and adult outcomes using a causal framework perspective: Challenges and opportunities.Child Abuse Negl. 2023; 143106328https://doi.org/10.1016/j.chiabu.2023.106328Crossref Scopus (1) Google A.E. Desrosiers T.A. Shanahan M.E. Directed acyclic graphs: An under-utilized tool for child maltreatment research.Child Abuse Negl. 2019; 91: 78-87https://doi.org/10.1016/j.chiabu.2019.02.011Crossref PubMed Scopus (25) Google Scholar such as may also be to collider stratification A. Glymour M.M. I. Using Structural to the of Adverse Childhood Social on of Heart Disease, and 2012; 23: PubMed Scopus Google Scholar These are in longitudinal with repeated of and which to (i.e., by previous T. P. in clinical research: when and to use 2017; PubMed Scopus Google Scholar research is needed to understand the relationship of childhood maltreatment with obesity and other chronic should be of such as collider stratification bias, lead to or

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.