Menu
Menu
Menu
Menu
BACKGROUND: Although childhood abuse has been consistently associated with cardiovascular disease in later adulthood, its associations with cardiometabolic health in younger adults are poorly understood. We assessed associations between childhood physical, sexual, and psychological abuse and cardiometabolic outcomes at 18 and 25 years. METHODS AND RESULTS: We used data on 3223 participants of the ALSPAC (Avon Longitudinal Study of Parents and Children). Exposure to childhood abuse was self-reported retrospectively at 22 years. We used linear regression to assess the associations between childhood abuse and cardiometabolic outcomes at 18 and 25 years. At 18 years, physical (β 1.35 kg/m2; 95% CI, 0.66–2.05), sexual (β 0.57 kg/m2; 95% CI 0.04–1.11), and psychological (β 0.47 kg/m2; 95% CI 0.01–0.92) abuse were associated with higher body mass index. Physical abuse was also associated with lower high-density lipoprotein cholesterol (β −0.07 mmol/L; 95% CI, −0.13 to −0.01) and higher C-reactive protein (31%; 95% CI, 1%–69%), and sexual abuse was associated with higher heart rate (β 1.92 bpm; 95% CI 0.26–3.58). At age 25, all 3 types of abuse were additionally associated with higher insulin, and sexual abuse was associated with lower cholesterol (−0.14 mmol/L; 95% CI, −0.26 to −0.01). The age at which abuse occurred (<11or 11–17 years) had little influence on the associations, and when sex differences were evident, associations were stronger in men. CONCLUSIONS: Childhood abuse is associated with negative cardiometabolic outcomes even by young adulthood. Further follow-up will determine whether associations strengthen across the life course and whether sex differences persist, which is essential for targeting effective screening programs and early interventions in those who suffered abuse in childhood.
Because of the sensitive nature of the data collected for this study, requests to access the data set from qualified researchers may be sent to the ALSPAC (Avon Longitudinal Study of Parents and Children) Executive Committee at https://proposals.epi.bristol.ac.uk/. The ALSPAC is a prospective population‐based pregnancy cohort (see www.alspac.bris.ac.uk) that recruited pregnant women living in the Avon area of the United Kingdom who were due to give birth between April 1991 and December 1992. 14 In total, 14 541 pregnancies were enrolled and children, mothers, and their partners have been followed up repeatedly ever since. Please note that the study website contains details of all the data that are available through a fully searchable data dictionary and variable search tool. For full details of the data from the ALSPAC study, see http://www.bristol.ac.uk/alspac/researchers/our‐data. The study participant flow is given in Figure 1. Participants were included if they had data on at least one type of abuse and one cardiometabolic outcome. Participants pregnant at the 18‐ and 25‐year clinic assessments were excluded as pregnancy could alter BMI and cardiometabolic health outcomes, resulting in the inclusion of 3223 participants in the study. Ethical approval was obtained from the ALSPAC Law and Ethics Committee and the Local Research Ethics Committee. Consent for biological samples has been collected in accordance with the Human Tissue Act (2004). Exposure to childhood abuse (before 18 years) was retrospectively self‐reported in a questionnaire at 22 years. The questionnaire used to collect information on abuse was based on the Child Abuse Questionnaire 15 and the Sexual Experiences Survey 16 and includes the 3 main types of abuse: physical, sexual, and psychological abuse; see http://www.bristol.ac.uk/media‐library/sites/alspac/documents/questionnaires/YPB‐life‐at‐22‐plus.pdf, section H. Participants were asked about experiences occurring in childhood (before 11 years), and during adolescence (11–17 years). Details about the abuse categorization are available in Data S1. We assessed abuse in each time period (<11/11–17 years) and also combined both time periods to indicate abuse <18 years and generated a summary score varying from 0 (no experience of abuse <18 years old) to 3 (experience of all abuse types <18 years old). The study data were collected and managed using REDCap electronic data capture tools hosted at University of Bristol. 17 Height and weight were measured in research clinics at both 18 and 25 years using standard procedures. Participants fasted overnight or for a minimum of 6 hours. Total cholesterol, plasma triglycerides, and high‐density lipoprotein (HDL) cholesterol were measured using enzymatic reagents for lipid determination from the standard Lipid Research Clinics Protocol. Low‐density lipoprotein cholesterol concentrations were calculated using the Friedewald equation. 18 Serum insulin was measured with ELISA (Mercodia, Uppsala, Sweden), which does not cross‐react with proinsulin. An automated particle‐enhanced immunoturbidimetric assay (Roche UK, Welwyn Garden City, United Kingdom) was used to measure plasma glucose and CRP (C‐reactive protein). We considered household occupational social class, maternal and paternal education, ethnicity, age, and sex as potential confounders. Household occupational social class was assessed at recruitment to the study and defined based on the highest of mothers’ and their partners’ self‐reported occupation using the 1991 British Office of Population and Census Statistics classification. Maternal and paternal education were also assessed at recruitment and correspond to the highest educational attainment achieved. Race/ethnicity was classified as White/non‐White, as most participants were of White race (96%). Data were analyzed using Stata 16.1 (Stata Corp., College Station, TX, 2016). Positively skewed outcome variables were log‐transformed for analyses and back transformed for presentation of results. We investigated each type of abuse separately and assessed associations of childhood abuse with cardiometabolic health considering abuse exposure occurring (1) before 11 years, (2) between 11 and 17 years, and (3) at any age before 18 years. We used linear regression to estimate associations of childhood abuse with measures of cardiometabolic health at 18 and 25 years, unadjusted and adjusted for the potential confounders defined previously. We used the outcomes in their original units, as well as standardized measures to allow comparability across the different outcomes. We compared the associations for abuse <11 years and between 11 and 17 years by comparing the point estimates and examining whether 95% CI overlapped and by using seemingly unrelated estimation to assess the difference between the coefficients. We explored possible sex differences in the associations between childhood abuse and cardiometabolic outcomes in a model including an interaction term between each type of childhood abuse and sex. We also used linear regression to examine the association between a summary score of abuse types that occurred <18 years (ranging from 0 to 3) and the cardiometabolic outcomes. We assessed whether there was a dose‐response relationship (ie, increase in the outcomes by increase in the score of abuse) by using the score as continuous and a Wald test for linear trend. Considering that different types of childhood abuse commonly co‐occur, we performed sensitivity analysis with the types of abuse mutually adjusted to estimate their independent associations. Mental health can influence the report of childhood abuse, such that individuals with higher psychological distress are more likely to report adverse childhood experiences. 19 Therefore, we also performed sensitivity analysis adjusting for depression at the time of childhood abuse reporting. Depression was measured using the Short Mood and Feelings Questionnaire, 20 a 13‐item questionnaire with score ranging between 0 and 26, in which a greater score represents higher depression. To explore the frequency of childhood abuse occurrence, we recategorized the occurrence of each type of abuse into the following frequency categories: never, rarely/sometimes, and often/very often. As each abuse type was assessed by multiple questions, we applied the response indicating the highest frequency per type. More details are presented in Data S1. There were missing data on outcomes and covariates. The highest proportion of missing data was observed for insulin at age 18 (44.7%), followed by total cholesterol, HDL, low‐density lipoprotein, triglycerides, glucose, and CRP (43.7%) at 18 years (Table S1). To increase precision and reduce selection bias, we conducted multivariate multiple imputation using chained equations to impute missing information. Twenty cycles of regression switching were used and estimates of results were averaged across the imputed data sets according to Rubin's rules. 21 More details on the imputation models are available in Data S1. We also performed analysis in those with complete data on child abuse, covariates, and outcomes (complete cases) as a sensitivity analysis.
