Self‐reported child maltreatment and cardiometabolic risk in 30‐year‐old adults

Abstract

Background

Childhood maltreatment (CM) is associated with both dietary fat intake and obesity in later life. There is less information on associations with metabolic risk factors and specific types of CM such as physical, sexual and emotional abuse, as well as neglect.

Aims

To assess the association between five types of self‐reported CM and a range of obesity and metabolic indicators in a subsample of a birth cohort.

Methods

This was a study of 1689 adults born in a major metropolitan maternity hospital in Australia and followed up 30 years later. Body mass index, bioimpedance and fasting lipid levels/insulin resistance were measured. Details on self‐reported CM were collected using the Child Trauma Questionnaire. We adjusted for birth weight, parental income and relationship at participants' birth, as well as maternal age and alcohol or tobacco use. We also adjusted for participants' smoking, depression, educational level, marital and employment status at follow up.

Results

One‐fifth reported maltreatment (n = 362), most commonly emotional neglect (n = 175), followed by emotional abuse (n = 128), physical neglect (n = 123), sexual (n = 121) and physical abuse (n = 116). On adjusted analyses, there were significant associations for CM, particularly neglect or emotional abuse, and one or more of the following outcomes: obesity, the total cholesterol/high‐density lipoprotein cholesterol (TC/HDL‐C) ratio and HDL levels. Results for other outcomes were more equivocal.

Conclusions

Of child maltreatment types, emotional abuse and neglect show the strongest associations with obesity and several cardiometabolic risk factors, therefore highlighting the public health importance of early intervention to reduce childhood adversity.

Introduction

Childhood maltreatment (CM) is associated with a range of physical health conditions in later life. ¹ , ² These include height deficit, high dietary fat intake, obesity, poor sleep quality, asthma, chronic lung conditions, ischaemic heart disease and cancer. ³ , ⁴ , ⁵

Most information concerns obesity. For instance, CM has been associated with being overweight ² , ⁶ , ⁷ , ⁸ and abnormal body mass index (BMI) ⁹ in large cross‐sectional, ⁷ longitudinal ⁶ , ⁸ and meta‐analytic studies. ² , ⁹ , ¹⁰ One possible mechanism is through high dietary fat, which has been shown to be associated with retrospective self‐reported and prospective agency‐notified CM. ¹¹ , ¹² One consequence of dyslipidaemia is cardiovascular disease (CVD) and a recent meta‐analysis reported that CM was associated with a twofold increase in this outcome. ⁵ One limitation of many studies is that there is little adjustment for lifetime confounders. For example, socioeconomic disadvantage in early life shows a significant association with both markers of cardiovascular risk and CVD in adulthood. ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷

There is less information on specific types of CM such as physical, sexual and emotional abuse, as well as neglect. One systematic review of the long‐term consequences of sexual abuse reported an association with obesity but did not include other metabolic risk factors. ¹⁸ In terms of non‐sexual CM, another review reported that both physical and emotional abuse were associated with obesity, ⁹ while neglect showed a significant association with self‐reported obesity, but not increased waist circumference or BMI. ⁹

Therefore, we assessed the association between five types of self‐reported CM and a range of obesity and metabolic indicators in a birth cohort. These included insulin, glucose, and the ratio of total cholesterol (TC) to high‐density lipoprotein cholesterol (HDL‐C) levels as measured in a fasting blood sample. Use of a birth cohort gave the opportunity to adjust for poverty at birth as well as smoking, depression and markers of social disadvantage in adulthood.

Methods

This longitudinal birth cohort study of obesity and metabolic indicators in 30 years used data from the ‘Mater‐University of Queensland Study of Pregnancy’ (MUSP) in Queensland, Australia. ¹⁹ The University of Queensland Behavioural and Social Sciences Ethical Review Committee and Mater Misericordiae Ltd. Human Research Ethics Committee approved the study.

Data sources

As described elsewhere, baseline data were collected at the first antenatal visit from 7223 consecutive women who gave birth to live singleton babies previously at the main obstetrics hospital in Brisbane, Australia, from 1981 to 1983. ¹⁹ We report on a selected subsample of 1689 adults with data on BMI and cardiometabolic markers at 30‐year follow up.

Self‐reported CM measures

We administered the Child Trauma Questionnaire (CTQ) at the 30‐year follow up. This instrument has been used worldwide and shows consistent reliability, sensitivity, factor structure and discriminant validity. ²⁰ , ²¹ , ²² , ²³ This is reflected in findings from a meta‐analysis of CM measures that reported only two met COSMIN (COnsenus‐based Standards for the selection of health Measurement INstruments) criteria: the full and short‐form (SF) versions of the CTQ. ²⁴

The CTQ has five subscales with five items in each. ²⁵ These subscales cover emotional, physical and sexual abuse, as well as emotional and physical neglect. In common with other epidemiological studies of CM prevalence, the continuous scores from the subscales were divided into ‘none/low’ and ‘moderate/severe’. ²⁶

Obesity and cardiometabolic markers

We measured a range of cardiometabolic markers. BMI (weight (kg)/height²) was the primary outcome with height measured using a portable stadiometer and weight using the Tanita Body Composition Analyser BC‐418. Obesity was defined using standard criteria provided by Cole et al. ²⁷

Insulin, glucose and the ratio of TC to HDL‐C (TC/HDL‐C) levels were measured in a fasting blood sample. Respondents were asked to eat by 7 pm the previous evening and fast for at least 9 h before blood sampling. Samples were collected by Mater Pathology Service, Brisbane. Respondents who lived outside the Brisbane area had their samples obtained by a participating laboratory. The homeostatic model assessment of insulin resistance (HOMA‐IR) was estimated using the formula fasting insulin (m u/L) × fasting glucose (mmL/L)/22.5 and then divided into quartiles. Cholesterol oxidase/peroxidase, lipase GK/GPO/peroxidase and phosphating state/Mg²⁺ methods were used to assess serum cholesterol, triglycerides (Tg) and high‐density lipoproteins (HDL) using an Ortho Clinical Diagnostics Vitros Analyser. The TC/HDL‐C ratio was calculated. Last, truncal fat mass was measured using a bioimpedance machine.

Apart from BMI and HOMA‐IR, all markers of obesity and cardiometabolic risk were divided into quartiles of increasing (or in the case of HDL decreasing) risk. We did not use continuous scores given P‐values of less than 0.05 on both the Kolmogorov–Smirnov and Shapiro–Wilk tests indicated that the data did not have a normal distribution.

Covariates

We adjusted for participants' sociodemographic variables at baseline as follows: low birth weight (<1499 g), gender, parental racial origin, parental relationship and maternal age, as well as binge drinking during pregnancy (5+ glasses on any occasion) and smoking in the 6 months from participants' birth. This was used in preference to smoking during pregnancy to minimise indirect effects through low birth weight given evidence that parental smoking early in a child's life has a similar effect on subsequent obesity. ²⁸ We also adjusted for low family income at birth, which was defined as the lowest 20–25% of family incomes based on Australian data and consistent with the National Poverty Line. ²⁹ To these, we added participants' employment, educational and marital status at the 30‐year follow up, as well as self‐reported tobacco use and depression at follow up. We measured the latter using the depression subscale of the Personal Disturbance Scale (DSSI/sAD). This was derived from the full Delusions‐Symptoms‐States Inventory (DSSI). The depression scale consists of seven self‐reported items with a score of 3 and above indicating probable depression. ³⁰ As in previous work on this cohort, participants' responses to how many cigarettes they had smoked in the previous week were dichotomised into 0–19 and 20 or more. ³¹

Statistical analysis

Bivariate analyses were first performed on the association between maltreatment and cardiometabolic outcomes. Maltreatment in general was considered, as well as neglect and emotional, physical or sexual maltreatment. We also studied the effect of the number of reported CM types. We divided the variable into two reported types as preliminary analyses of the primary outcome (BMI) indicated that this was the threshold for increased metabolic risk (Supporting Information Table S1).

We used multinomial logistic regression to measure the strength of association between child maltreatment and BMI levels, lipid or adiposity quartiles while adjusting for covariates. Logistic regression was used to analyse HOMA‐IR. Last, we undertook sensitivity analyses of the effect of including the BMI results in the regression models of the other cardiometabolic outcomes.

Results

Of the 1689 young adults at follow up, 966 (57.2%) were female and 723 (42.8%) were male. Loss to follow up was greater among participants who were male, Indigenous, whose mothers were teenagers at the time of their birth and whose parents were not living together at baseline, as well as for low family income on study entry (Table S2). Attrition was also greater in the presence of low birth weight and maternal binge drinking during pregnancy or smoking in the initial six post‐natal months (Table S2).

Just over one‐fifth of the sample at follow up reported maltreatment (n = 362), most commonly emotional neglect (n = 175), followed by emotional abuse (n = 128), physical neglect (n = 123), sexual (n = 121) and physical abuse (n = 116). Of these, 80 (4.7%) reported more than two types. As with loss to follow up, self‐reported maltreatment was associated with measures of social disadvantage. This included younger maternal age and low family income at baseline, along with single status, being unemployed or receiving government benefits, and depression at age 30 years (Table S3). CM was also associated with low birth weight and maternal binge drinking or smoking (Table S3).

Table 1 compares the results for BMI levels by a range of sociodemographic and child maltreatment variables. Overall CM, as well as four subtypes, was associated with increasing BMI levels. The only exception was physical neglect. The presence of more than two CM reported subtypes was also associated with greater BMI levels (Table 1). Other factors that also showed a significant association were the following: male gender, parents not living together at baseline and depression at follow up (Table 1). Table S4 shows the adjusted results for overall CM using multinomial logistic regression analysis. Male gender, indigenous status and younger maternal age were associated with being overweight while CM was associated with obesity (Tables 2, S4). In terms of the subtypes, emotional neglect or abuse showed significant associations with obesity on adjusted analyses (Table 2). The same applied to participants with more than two types of CM reports (Table 2). There was no association with the other CM subcategories (Table 2).

Table 1.

Factors associated with abnormal body mass index (BMI)

		BMI
		Low or normal (n = 691)		Overweight (n = 567)		Obese (n = 431)		Significance
		n	Column %	n	Column %	n	Column %	Χ 2 †	P‐value
Male gender	Yes	241	34.9%	305	53.8%	177	41.1%	46.23	<0.0001
Indigenous status	Indigenous	14	2.1%	22	3.9%	18	4.3%	5.44	0.066
Baseline age of mother	<20	53	7.7%	61	10.8%	50	11.6%	5.75	0.056
Parental relationship	Not in a relationship	68	9.8%	46	8.1%	56	13.0%	6.51	0.039
Family income at birth	≤$10 399	176	26.5%	143	26.4%	132	32.0%	4.73	0.094
Birthweight	Low	21	3.0%	16	2.8%	18	4.2%	1.59	0.451
Maternal smoking	At 6 months	232	33.6%	191	33.7%	173	40.1%	5.97	0.049
Maternal binge drinking	In pregnancy	130	18.8%	103	18.2%	76	17.6%	0.25	0.880
Single at age 30	yes	193	27.9%	151	26.6%	107	24.8%	1.31	0.520
Tertiary education	Yes	161	23.3%	140	24.7%	106	24.6%	0.41	0.815
Work status at age 30	Unemployed/pension	36	5.2%	26	4.6%	28	6.5%	1.81	0.405
Smoking at age 30	20+	18	2.9%	16	3.2%	20	5.4%	4.43	0.109
Depression at age 30	Yes	52	7.5%	34	6.0%	54	12.5%	14.64	0.001
Emotional neglect	Yes	58	8.4%	59	10.4%	58	15.7%	8.65	0.013
Physical neglect	Yes	44	6.3%	38	6.7%	41	9.5%	5.08	0.079
Physical abuse	Yes	40	5.7%	34	6.0%	42	9.7%	8.51	0.014
Sexual abuse	Yes	46	6.7%	31	5.5%	44	10.2%	9.65	0.008
Emotional abuse	Yes	41	5.9%	39	6.9%	48	11.1%	12.21	0.002
Any case of Abuse/neglect	Yes	133	19.2%	114	20.1%	115	26.7%	9.61	0.007
No. reports	>2	24	3.5%	21	3.7%	35	8.1%	14.72	0.001

^†DF = 2.

Bold values are considered significant.

Table 2.

Adjusted associations with body mass index (BMI)

	BMI
	Overweight			Obese
	OR (95% CI)	Adjusted OR (95% CI)	P‐value	OR (95% CI)	Adjusted OR (95% CI)	P‐value adjusted OR
Neglect
Emotional	1.30 (0.88–1.90)	1.31 (0.87–1.97)	0.192	1.79 (1.21–2.63)	1.54 (1.01–2.34)	0.046
Physical	1.08 (0.68–1.69)	1.06 (0.66–1.71)	0.823	1.62 (1.04–2.53)	1.43 (0.88–2.32)	0.147
Emotional abuse	1.18 (0.76–1.88)	1.44 (0.89–2.35)	0.140	2.09 (1.34–3.23)	1.78 (1.11–2.87)	0.017
Physical abuse	1.06 (0.66–1.70)	1.14 (0.69–1.88)	0.615	1.84 (1.17–2.90)	1.61 (0.99–2.64)	0.057
Sexual abuse	0.98 (0.51–1.32)	0.91 (0.55–1.51)	0.720	1.67 (1.08–2.57)	1.48 (0.92–2.39)	0.103
Any self‐reported maltreatment	1.06 (0.80–1.40)	1.28 (0.62–2.62)	0.505	1.53 (1.15–2.03)	1.45 (1.06–2.02)	0.018
>2 reported maltreatments	1.07 (0.59–1.94)	1.23 (0.66–2.630	0.516	2.46 (1.44–4.19)	2.13 (1.20–3.81)	0.010

Variables used in adjustment:

• At baseline: gender; parental race; parental relationship; maternal age; low birth weight; maternal smoking at 6‐month follow up; binge drinking in pregnancy.

• At the 30‐year follow up: young adult's employment status; young adult's education level; young adult's marital status; cigarette smoking; Delusions‐Symptoms‐States Inventory (DSSI)‐defined depression. CI, confidence interval; OR, odds ratio.

Bold values are considered significant.

Table 3 shows the bivariate results for TC/HDL‐C ratios as an example of one of the biochemical markers. In this case, only physical abuse and both types of neglect showed a significant association with increased levels, as did male gender, work and smoking status. On adjusted analyses, both physical abuse and neglect were associated with TC/HDL‐C ratios in the second quartile, as were more than two CM reports, while physical or emotional neglect was associated with ratios in the highest quartile (Table 4). When the BMI results were included in the model, only the findings for physical neglect remained significant (odds ratio (OR) = 3.10; 95% confidence interval (CI) = 1.51–6.36; P = 0.002).

Table 3.

Factors associated with biochemical markers

		Cholesterol/high‐density lipoprotein ratio
		1st quartile (n = 409)		2nd quartile (n = 450)		3rd quartile (n = 397)		4th quartile (n = 428)		Significance
		n	Column %	n	Column %	n	Column %	n	Column %	Χ2 †	P‐value
Male	Yes	70	17.1%	138	30.7%	160	40.3%	266	62.1%	193.4	<0.0001
Indigenous status	Indigenous	9	2.3%	10	2.3%	16	4.1%	17	4.1%	4.44	0.218
Baseline age of mother	<20	42	10.3%	48	10.7%	37	9.3%	43	10.0%	0.44	0.932
Parental relationship	Not in a relationship	42	10.3%	42	9.3%	38	9.6%	42	9.8%	0.23	0.973
Family income at birth	$10 399 or less	114	29.8%	113	26.2%	122	32.1%	120	28.8%	3.59	0.309
Birthweight	Low	17	4.2%	11	2.4%	12	3.0%	17	4.0%	2.57	0.462
Maternal smoking	At 6 months follow up	130	31.8%	172	38.2%	137	34.5%	147	34.3%	4.01	0.260
Binge drinking	In pregnancy	64	15.6%	82	18.2%	70	17.6%	82	19.2%	1.89	0.595
Tertiary education	Yes	108	26.4%	107	23.8%	101	25.4%	107	25.0%	0.82	0.846
Single at age 30	yes	113	27.6%	121	26.9%	103	25.9%	126	29.4%	1.38	0.711
Work status at age 30	Unemployed/pension	12	2.9%	27	6.0%	27	6.8%	40	9.3%	14.83	0.002
Smoking at age 30	20+	4	1.0%	12	2.9%	16	4.3%	16	4.1%	8.72	0.033
Depression at age 30	Yes	30	9.7%	27	7.3%	27	8.4%	35	10.7%	2.93	0.403
Emotional neglect	Yes	30	7.7% 7.6%	54	12.0%	44	11.1 %	58	13.5%	10.06	0.018
Physical neglect	Yes	14	3.4%	37	8.2%	26	6.6%	44	11.3% 10.2%	17.12	0.001
Physical abuse	Yes	17	4.2%	38	8.4%	32	8.1%	34	7.9%	7.94	0.047
Sexual abuse	Yes	37	9.0%	33	7.3%	29	7.3%	26	6.1%	2.24	0.524
Emotional abuse	Yes	27	6.6%	38	8.4%	34	8.6%	39	9.1%	2.39	0.496
Any case of Abuse/neglect	Yes	80	19.6%	103	22.9%	95	23.9%	100	23.4%	2.72	0.437
No. reports	> 2	11	2.7%	30	6.7%	18	4.5%	26	6.1%	8.37	0.039

^† DF = 3.

Bold values are considered significant.

Table 4.

Unadjusted and adjusted analyses of the association between cholesterol/high‐density lipoprotein ratios and child maltreatment

	Cholesterol/high‐density lipoprotein ratio
	2nd quartile			3rd quartile			4th quartile
	OR (95% CI)	Adjusted OR (95% CI)	P‐value adjusted OR	OR (95% CI)	Adjusted OR (95% CI)	P‐value adjusted OR	OR (95% CI)	Adjusted OR (95% CI)	P‐value adjusted OR
Neglect
Emotional	1.76 (1.10–2.82)	1.94 (1.17–3.21)	0.010	1.59 (0.98–2.59)	1.45 (0.85–2.47)	0.174	2.08 (1.31–3.31)	1.87 (1.10–3.17)	0.020
Physical	2.58 (1.37–4.85)	2.91 (1.46–5.71)	0.002	1.99 (1.03–3.88)	1.72 (0.82–3.60)	0.152	3.39 (1.82–6.29)	3.30 (1.65–6.63)	0.001
Emotional abuse	1.33 (0.79–2.22)	1.38 (0.80–2.39)	0.242	1.34 (0.79–2.26)	1.13 (0.64–2.01)	0.675	1.48 (0.89–2.47)	1.45 (0.82–2.59)	0.205
Physical abuse	2.16 (1.19–4.28)	2.15 (1.14–4.04)	0.017	2.03 (1.11–3.73)	1.89 (0.99–3.63)	0.054	2.07 (1.13–3.78)	1.62 (0.83–3.18)	0.159
Sexual abuse	0.81 (0.50–1.32)	0.87 (0.51–1.47)	0.604	0.81 (0.48–1.34)	0.89 (0.52–1.56)	0.700	0.68 (0.40–1.14)	0.96 (0.53–1.73)	0.886
Any self‐reported maltreatment	1.22 (0.88–1.70)	1.25 (0.87–1.80)	0.216	1.29 (0.92–1.81)	1.18 (0.82–1.73)	0.368	1.25 (0.90–1.75)	1.18 (0.81–1.74)	0.392
>2 reported maltreatments	2.50 (1.15–5.42)	2.97 (1.32–6.69)	0.008	1.51 (0.64–3.56)	1.39 (0.56–3.48)	0.472	2.07 (0.92–4.62)	2.28 (0.94–5.50)	0.067

Variables used in adjustment:

• At baseline: gender; parental race; parental relationship; maternal age; low birth weight; maternal smoking at 6‐month follow up; binge drinking in pregnancy.

• At the 30‐year follow up: young adult's employment status; young adult's education level; young adult's marital status; cigarette smoking; Delusions‐Symptoms‐States Inventory (DSSI)‐defined depression. CI, confidence interval; OR, odds ratio.

Bold values are considered significant.

With regards to HDL‐C levels, emotional abuse and both types of neglect were associated with levels in the lowest quartile on adjusted analyses (Table 5). The same applied to more than two types of CM reports (Table 5). Similarly, only emotional abuse was associated with high HOMA levels on unadjusted (OR = 1.55; 95% CI = 1.09–2.21; P = 0.014) and adjusted analyses (OR = 1.48; 95% CI = 1.02–2.16; P = 0.042). The same applied to trunk fat mass where only emotional abuse was associated with readings in the highest quartile on unadjusted (OR = 2.34; 95% CI = 1.38–3.98; P = 0.002) and adjusted analyses (OR = 1.91; 95% CI = 1.06–3.45; P = 0.032).

Table 5.

Unadjusted and adjusted analyses of the association between high‐density lipoprotein levels and child maltreatment

	High‐density lipoprotein
	2nd quartile			3rd quartile			4th quartile
	OR (95% CI)	Adjusted OR (95% CI)	P‐value adjusted OR	OR (95% CI)	Adjusted OR (95% CI)	P‐value adjusted OR	OR (95% CI)	Adjusted OR (95% CI)	P‐value adjusted OR
Neglect
Emotional	0.78 (0.53–1.20)	0.88 (0.57–1.37)	0.580	0.69 (0.46–1.02)	0.79 (0.51–1.22)	0.287	0.39 (0.24–0.63)	0.43 (0.25–0.74)	0.002
Physical	0.98 (0.61–1.58)	0.98 (0.59–1.64)	0.945	0.60 (0.36–0.98)	0.70 (0.41–1.20)	0.195	0.40 (0.22–0.73)	0.46 (0.24–0.88)	0.019
Emotional abuse	1.07 (0.68–1.70)	1.33 (0.79–2.21)	0.279	0.76 (0.48–1.21)	0.76 (0.45–1.29)	0.308	0.50 (0.29–0.86)	0.50 (0.27–0.93)	0.027
Physical abuse	0.98 (0.60–1.58)	1.34 (0.79–2.27)	0.277	0.65 (0.39–1.08)	0.85 (0.49–1.48)	0.530	0.51 (0.29–0.90)	0.60 (0.31–1.14)	0.598
Sexual abuse	1.18 (0.67–2.09)	1.14 (0.61–2.12)	0.682	1.55 (0.92–2.59)	1.33 (0.75–2.36)	0.322	1.49 (0.87–2.56)	1.04 (0.57–1.90)	0.900
Any self‐reported	0.99 (0.72–1.37)	1.08 (0.75–1.54)	0.675	0.93 (0.68–1.26)	1.01 (0.72–1.44)	0.919	0.75 (0.54–1.05)	0.76 (0.52–1.12)	0.160
>2 reports	0.83 (0.47–1.47)	0.92 (0.50–1.72)	0.803	0.67 (0.35–1.23)	0.66 (0.36–1.24)	0.336	0.40 (0.19–0.81)	0.37 (0.17–0.78)	0.010

Variables used in adjustment:

• At baseline: gender; parental race; parental relationship; maternal age; low birth weight; maternal smoking at 6‐month follow up; binge drinking in pregnancy.

• At the 30‐year follow up: young adult's employment status; young adult's education level; young adult's marital status; cigarette smoking; Delusions‐Symptoms‐States Inventory (DSSI)‐defined depression. CI, confidence interval; OR, odds ratio.

Bold values are considered significant.

On including the BMI results, both forms of neglect remained significantly associated with HDL scores in the lowest quartile, the odds ratio for the emotional category being 0.48 (95% CI = 0.27–0.83; P = 0.009), as were more than two self‐reported CM events (OR = 0.45 (95% CI = 0.21–0.99; P = 0.047). However, none of the results for the other outcomes were significant.

Discussion

Obesity and dyslipidaemia, along with hypertension, and smoking, are major risk factors for cardiovascular disease, type 2 diabetes mellitus (T2DM) and cancer. ³² , ³³ Being overweight or obese also increases the risk of all‐cause mortality. ³⁴ , ³⁵

There is literature on the association between child maltreatment and self‐reported physical health status. ³ , ⁷ , ³⁶ There is also information on the association of CM with markers of obesity such as dietary intake and BMI. ⁹ , ¹² This paper adds to the literature by examining in detail the association between five types of child maltreatment and several cardiometabolic markers. Depending on the metabolic outcome, neglect, physical and emotional abuse all showed a significant association. As in other studies, increasing numbers of CM types were also associated with greater CVD risk. ³ , ³⁷ Our findings are also consistent with an earlier paper from the same cohort where childhood poverty was associated with many of the same cardiometabolic markers, especially in females. ¹⁷ This suggests that a wide range of adversity circumstances in childhood may be associated with poor cardiometabolic outcomes.

One strength of this study is adjustment for a variety of baseline demographic factors and characteristics at follow up including smoking and concurrent depression. These have been previously shown to be possible confounders in any relationship. For instance, sociodemographic factors including age, lower education, ⁷ lower qualification, unemployment ⁸ and lower income ⁷ have been associated with obesity. ⁷ , ⁸ , ³⁸ Similarly, smoking and depression are associated with both childhood abuse ³ , ³¹ , ³⁹ and an increased risk of heart disease, ⁴⁰ , ⁴¹ as well as subsequent cardiac‐related morbidity and mortality. ⁴² We also undertook sensitivity analyses of the effect of including the BMI results in the regression models of the other cardiometabolic outcomes. The results for physical or emotional neglect and TC/HDL‐C ratios or HDL levels remained significant suggesting that the association was at least partly independent of obesity.

There are several possible explanations for the association between CM and cardiometabolic markers. One is that adverse events in childhood have been associated with an unhealthy diet in adolescence ³⁸ as a result of the palatability of high energy‐density foods such as fast food, chips or salty snacks ⁴³ and their possible stress‐relieving properties. ⁴⁴ This may be compounded by ease of access to these foods for those in disadvantaged circumstances. ⁴⁵ The children may develop a craving for fatty or sugary foods ³⁸ with subsequent ‘food addiction’ ⁴⁶ , ⁴⁷ as a result of ‘emotional eating’ ⁴⁸ and ‘self‐medication’ ³⁸ to cope with depression or loneliness. ⁴⁸ In addition some of these food items may directly alter brain chemistry ⁴⁹ and neuro‐adaptive mechanisms. ⁵⁰ , ⁵¹ Child maltreatment may also have an effect on other lifestyle factors that may be relevant to metabolic risk such as physical activity. ³⁷ , ⁵²

The associations we found were largely restricted to emotional abuse and neglect rather than sexual or physical abuse. It is possible that this might represent under‐reporting by respondents due to the traumatic nature of previous events or stigma. For instance, in an earlier study from the MUSP cohort at the 21‐year follow up almost 40% of participants who had been the subject of substantiated child sexual abuse failed to recall any abuse when asked in adulthood, possibly either due to actual loss of memory or to suppression of memories as a psychological defence. ⁵³ Conversely, the nature and severity of physical or sexual abuse may increase the likelihood of action before it leads to long‐term consequences. ⁵⁴

In addition, there may be more direct mechanisms whereby persistent adversity in childhood affects the normal development of the hypothalamus–pituitary–adrenal axis leading to increased cortisol levels in adulthood and greater risk of cardiometabolic disease. ⁵⁵ Similar experiences in childhood may also lead to chronic inflammation and susceptibility to systemic disease. ⁵⁶ , ⁵⁷

There are several limitations of this study, the most important being the substantial loss to follow up over 30 years. This, in turn, was associated with markers of social deprivation, which may limit the study's generalisability. ⁵⁸ However, while loss to follow up may affect sample means, there is less of an effect on estimates of association. ⁵⁹ , ⁶⁰ For instance, analyses of the effect of differential follow up in the MUSP cohort suggest that the estimates of association are similar in the groups retained in the study and those lost to follow up. ⁶¹ These findings have now been replicated. ⁶² In addition, multiple imputation of MUSP data changes neither the estimates nor their precision, ⁶¹ , ⁶³ and this is consistent with findings of the effects of differential attrition in other longitudinal studies. ⁶⁴ Finally, we adjusted for variables that were more common among missing participants.

Another limitation is the use of self‐reported retrospective CM data and the possibility of recall and rumination bias. In particular, it has been suggested that recall may be affected by current psychological health. ⁶⁵ For example, people in poor psychological health may differentially recall early negative experiences or interpret them negatively, particularly those with depression. ⁶⁵ In turn, depression is associated with an increased risk of heart disease ⁴⁰ , ⁴¹ and related mortality. ⁴² However, our results remained significant even after adjusting for concurrent depression. In addition, there is evidence for the accuracy of retrospective self‐reported CM in terms of consistency over time, or similar risk factors and outcomes to prospective data. ⁶⁶ , ⁶⁷ , ⁶⁸ In addition, although we adjusted for a wide range of covariates, there may be other important variables such as maternal weight that were not included in the model.

Last, the meaning of the significant association between some types of CM and TC/HDLC ratios in the second quartile is unclear and may reflect multiple testing. Results for study outcomes of borderline statistical significance should therefore be viewed with caution. In addition, HOMA‐IR is more related to carbohydrate rather than fat intake.

This paper adds to our previous findings of high dietary fat intake in the same cohort as measured at age 21 years and adds further weight to the need for the primary prevention of obesity through appropriate dietary advice in those with a history of CM. ¹² A greater understanding of modifiable risk factors of obesity and associated long‐term consequences may improve outcomes, ¹⁰ especially given the limited availability of alternative treatments for obesity and other metabolic conditions. ⁶⁹ This might include a greater awareness of the possible role of CM, particularly emotional abuse and neglect, in people presenting with obesity and related cardiometabolic risk factors, as well as an understanding that CM survivors will be at greater risk.

Prevention should encompass both CM in the current generation, as well as the possibility of intergenerational transmission given the increased likelihood of CM by parents who experienced maltreatment in their own childhood. ⁷⁰ , ⁷¹ This means addressing factors related to the child, parent(s), family and community, as well as wider societal attitudes towards children and abuse. ⁷⁰ For instance, maladaptive behaviour may be modelled during childhood that individuals then apply to their own subsequent parenting practices. CM survivors may also have impaired emotional regulation leading to irritability and aggression on later life. Added to this are broader disadvantages such as poverty or inadequate social support. Interventions should be timely multidimensional and encompass all forms of CM instead of solely sexual or physical abuse. ⁷² Examples include family health, home visiting and parenting programmes, particularly for those families at greatest risk.

Conclusions

These findings suggest that obesity and cardio‐metabolic risk factors are further adverse consequences, among many, of CM, so highlighting the public health importance of the issue. In particular, these results emphasise the importance of emotional maltreatment and neglect in terms of long‐term adverse consequences.

Supporting information

Table S1. Relationship between number of child maltreatment types and metabolic risk factors.

Table S2. Factors significantly associated with loss to follow up.

Table S3. Factors associated with child maltreatment.

Table S4. Multivariate associations with Body Mass Index (BMI).