Background: It has been suggested that childhood events increase the risk of suffering from an eating disorder (ED) in adolescence or young adulthood. Our study here aimed at investigating whether there was an association between stressful childhood events and eating disorders later in life. Methods: Our population-based study was a part of the HeSSup (The Health and Social Support) postal survey in 2003, a follow-up survey with a random sample of working-aged subjects drawn from the Finnish Population Register in 1998. Participants in this study reported having been suffering from ED (N = 374), while controls (N = 18,639) reported no history of eating disorder in the questionnaire both in 1998 and 2003. The questionnaire included six questions related to childhood adversities. Results: Each childhood stressor increased the risk of developing ED cumulatively (one adversity OR 1.7 versus all six OR 8.3). Intrafamiliar conflicts (OR 2.0), being afraid of a family member (OR 1.5) or long-term illnesses in the family (OR 1.4) increased the ED-risk statistical significantly. Conclusion: Children exposed to stress are more likely to suffer from eating disorders as adults than those not exposed.
Anorexia nervosa (AN) with estimated prevalence of 0.2% - 0.9% and bulimia nervosa (BN) with 0.5% - 4.5% of sufferers are the most well-known forms of eating disorders (ED) [
EDs are markedly more common among women than men [
Psychological [
Prior research suggests that adverse childhood events increase the risk of suffering from alcoholism, depression, cardiovascular diseases, migraine and asthma [
It may be speculated that the more adverse childhood events an individual has had the greater is the risk for an eating disorder later in life. According to our knowledge, cumulative effects of childhood adversities on EDs have not been studied before. The study here aims at establishing whether there is a connection between adverse childhood events and ED.
The Health and Social Support Study (HeSSup) is a retrospective cohort study into childhood predictors of adult psychosocial health in a Finnish general population sample. The HeSSup population comprised a random sample of 64,797 individuals drawn from the Finnish Population Register in four age groups: 20 - 24, 30 - 34, 40 - 44 and 50 - 54 years. The survey was carried out by postal questionnaire in 1998. Forms were returned by 25,895 individuals (response rate 40%). A follow-up questionnaire was sent in 2003 to all those who responded to the first inquiry. A total of 19,319 individuals responded to it (response rate 75%), and this study here made use of these latter responses.
Young women and elderly men were found to be the most active survey respondents [
In the structured questionnaire, the participants were asked an eating disorders-related survey question: “Has a doctor ever told you that you have or have had an eating disorder (anorexia, bulimia)?” Alternative replies were “yes” or “no”. Those who answered “yes” in 2003 were classified as persons with a history of an eating disorder. All in all, the data included 374 persons with an eating disorder (
The control group was formed out of persons who responded “No” to the question about eating disorders (n = 18,945). This group was further narrowed down to include only those who had not reported having suffered from an eating disorder in the original 1998 survey. The control group thus comprised individuals who had reported both in 1998 and 2003 that they were not suffering or had not previously suffered from an eating disorder. A total of 18,639 persons thus formed the control group.
The subjects were asked to recall their childhood adversities in terms of the following structured questions: “Did your parents divorce?”, “Did your family have long-lasting financial difficulties?”, “Did serious conflicts arise in your family?”, “Were you often afraid of some member of your family?”, “Was someone in the family seriously or chronically ill?”, “Did someone in the family have problems with alcohol?” The alternative answers were “yes”, “no” or “I do not know”. Those having answered “Yes” were analysed as one group and those having responded either “no” or “don’t know”/“can’t say” were classified as another group.
A sum variable containing all six sections related to childhood stressors was also built from childhood adversities. With the help of the sum variable the total number of “yes” answers of the subjects was calculated.
Subjects were classified into four age groups: 25 - 29, 35 - 39, 45 - 49 and 55 - 59-year-olds. For the logistic regression analysis, they were further divided into two groups: those under 40 years and those aged 40 or older.
Marital status was classified into three groups: the single and those having a relationship while the “others” included the divorced, the separated and the widowed. Education was classified as “high” for those who had completed high school, and “low” for those who had not. Beck’s depression inventory (BDI) was used in assessing whether the respondents suffered from depression [
Body Mass Index (BMI) was calculated on the basis of the self-reported heights and weights of the respondents. BMI was then classified into three categories: BMI < 18.5 kg/m2, 18.5 - 29.9 kg/m2 and ≥ 30.0 kg/m2. BMI was considered normal when within the range 18.5 - 29.9 kg/m2 whereas those with BMI of 30 or more were assessed as obese.
Smoking status was classified into two categories: non-smokers and regular smokers. Last group included also those who had stopped smoking. Alcohol use was charted with a question “In the past 12 months, have you ever passed out due to drinking too much alcohol?” Alternative answers included “never”, “once”, “2 - 3 times”, “4 - 6 times” and “7 times or more”. The variable was further classified into two categories. The first group was formed of those, who had never passed out due to alcohol use during the past year, and all the others formed the second group, i.e. persons who had passed out due to their alcohol use at least once during the past year.
The statistical significance of differences between the target groups and the controls were tested by χ2 test. Odds ratios (OR) with 95% confidence intervals (CI) were calculated by logistic regression analysis to measure the risk of belonging to an eating disorder group. Analyses were made using the SPSS system for Windows, release 20.0.
All participants were asked to sign a consent form to gather follow up data and in order to search Finnish health registers for their data. The joint ethics board of the University of Turku and Turku University Hospital assessed that no medical ethical approval was required to conduct the HeSSup research.
The characteristics of the eating disorder and control groups are presented in
Depression was associated with a seven-fold risk of reporting an ED. The risk was also greater for women, those living without a partner, the young, and those who had passed out due to alcohol consumption during the past year. On the contrary, smoking and having a higher education were not associated with a risk of reporting an ED (
Among women intrafamiliar conflicts and a family member’s long-term illness had an association with ED. Among men only economic difficulties were associated with ED. When both genders are analysed together intrafamiliar conflicts, being afraid of a family member and a family member’s long-term illness were associated with ED (
Accumulation of childhood adversities was associated with ED (
Childhood stressors and ED have been found to be linked. Those having experienced multiple childhood adversities had more often been diagnosed with eating disorder than those who had experienced none or only few adversities. The association is stronger the more accumulated adversities there have been in an individual’s life. Prior research has already found the connection between childhood adversities and eating problems [
Those reporting an eating disorders are mostly women, which is in line with earlier research findings [
According to our findings, those reporting an eating disorder also reported depression more than controls, which is in accordance with prior research findings [
Eating disorder | Control group | ||||
---|---|---|---|---|---|
(n = 367 - 374)* | (n = 18,300 - 18,639)* | ||||
n | % | n | % | p-value | |
Gender | <0.001 | ||||
Female | 309 | 82.6 | 11,364 | 61.0 | |
Male | 65 | 17.4 | 7275 | 39.0 | |
Age group (years) | <0.001 | ||||
25 - 29 | 141 | 37.7 | 4647 | 24.9 | |
35 - 39 | 77 | 20.6 | 4284 | 23.0 | |
45 - 49 | 77 | 20.6 | 4653 | 25.0 | |
55 - 59 | 79 | 21.1 | 5055 | 27.1 | |
Marital status | <0.001 | ||||
Single | 101 | 27.1 | 2853 | 15.4 | |
In a relationship | 214 | 57.4 | 13,914 | 75.0 | |
Other | 58 | 15.5 | 1792 | 9.7 | |
Education | 0.530 | ||||
Low | 152 | 40.8 | 7858 | 42.4 | |
High | 221 | 59.2 | 10,687 | 57.6 | |
BMI (kg/m2) | 0.005 | ||||
<18.5 | 11 | 3.0 | 258 | 1.4 | |
18.5 - 29.9 | 297 | 80.7 | 16,028 | 85.3 | |
≥30.0 | 60 | 16.3 | 2502 | 13.3 | |
Smoking | 0.060 | ||||
Non-smoker | 146 | 39.0 | 8185 | 43.9 | |
Smoker | 228 | 61.0 | 10,454 | 56.1 | |
Passed out due to alcohol | <0.001 | ||||
Never | 294 | 78.6 | 15,957 | 85.6 | |
At least once | 80 | 21.4 | 2682 | 14.4 | |
BDI | <0.001 | ||||
<19 points | 283 | 77.1 | 17,590 | 96.1 | |
≥19 points | 84 | 22.9 | 710 | 3.9 |
*Every respondent did not answer all questions.
OR (95% CI) | p-value | |
---|---|---|
Depression | <0.001 | |
Beck < 19 | 1 | |
Beck ≥ 19 | 7.34 (5.70 - 9.49) | |
Gender | <0.001 | |
Male | 1 | |
Female | 3.04 (2.34 - 3.98) | |
Marital status | <0.001 | |
In a relationship | 1 | |
Single or others | 2.20 (1.79 - 2.71) | |
Passed out due to alcohol | <0.001 | |
Never | 1 | |
At last once | 1.62 (1.26 - 2.08) | |
Age group | <0.001 | |
≥40 years | 1 | |
<40 years | 1.52 (1.23 - 1.87) | |
Smoking | 0.060 | |
Non-smoker | 1 | |
Smoker | 1.22 (0.99 - 1.50) | |
Education | 0.530 | |
Low | 1 | |
High | 1.07 (0.87 - 1.32) |
Female (n = 309) | Male (n = 65) | All (n = 374) | ||||
---|---|---|---|---|---|---|
OR (95% CI) | p-value | OR (95% CI) | p-value | OR (95% CI) | p-value | |
Parental divorce | 1.12 (0.84 - 1.50) | 0.424 | 0.66 (0.30 - 1.43) | 0.292 | 1.06 (0.81 - 1.38) | 0.684 |
Economic difficulties | 0.89 (0.68 - 1.17) | 0.407 | 1.83 (1.07 - 3.15) | 0.028 | 1.02 (0.80 - 1.29) | 0.898 |
Intrafamiliar conflicts | 2.20 (1.64 - 2.95) | <0.001 | 1.20 (0.61 - 2.38) | 0.594 | 2.01 (1.54 - 2.63) | <0.001 |
Being afraid of a family member | 1.34 (0.99 - 1.81) | 0.056 | 1.72 (0.83 - 3.57) | 0.148 | 1.46 (1.11 - 1.93) | 0.007 |
Long-term illness of a family member | 1.31 (1.03 -1.67) | 0.029 | 1.66 (1.00 - 2.77) | 0.052 | 1.37 (1.01 - 1.70) | 0.005 |
Alcohol problem of a family member | 1.09 (0.83 - 1.44) | 0.547 | 1.07 (0.57 - 2.00) | 0.843 | 1.12 (0.87 - 1.44) | 0.374 |
OR (95% CI) | p-value | |
---|---|---|
No adversity | 1 | |
1 adversity | 1.68 (1.24 - 2.73) | 0.001 |
2 adversities | 2.10 (1.51 - 2.92) | <0.001 |
3 adversities | 3.31 (2.37 - 4.63) | <0.001 |
4 adversities | 3.33 (2.28 - 4.86) | <0.001 |
5 adversities | 3.48 (2.16 - 5.61) | <0.001 |
6 adversities | 8.31 (4.20 - 16.46) | <0.001 |
Respondents’ year of birth was inquired about in the survey whereas it is not known when any of the inquired events occurred. Thus it is not known how old the subjects where when faced with the reported childhood adversities and thus this study cannot distinguish those subjects having experienced childhood adversities during the sensitive period from the other participants [
Subjects with ED and controls differed from each other with regard to age. Largest category of respondents among those reporting eating disorders were the youngest, while controls were evenly divided between all age groups. EDs occur typically in young people [
HeSSup (The Health and Social Support) survey was a research study with a sample of randomly selected Finns. Respondents were thus not selected into the study based on their possible contacts with primary or specialist care services, which means that our data describe eating disorders on a population-based level. Based on non-response bias analysis, the research data are in accordance with general Finnish population and the respondents were not specifically selected to the survey [
To ensure good repeatability, the survey is described in detail. In an effort to minimize confidence error, confidence intervals and statistical significance were used. The confidence intervals of the results derived via logistic regression analysis were small, which points to the similarity of the respondents. Deviation among respondents was small. Effort was made to control the confounding factors by adjusting age, gender and degree of depression.
To confirm the reliability of responses regarding childhood adversities, the Cohen’s kappa coefficient (κ) was used to assess associations between the 1998 and 2003 questionnaires. κ varied between 0.621 and 0.903 among participants [
The question inquiring about eating disorders did not differentiate between types of ED. It is thus entirely possible that childhood stress factors influence some type of ED but not all.
Error sources, which cannot be influenced while doing the research, are always possible in surveys. Respondents may have understood a question in various different ways due to which answers may be distorted. Surveys are always based on participants’ recall-and experience-dependent answers. Due to this reason, the authenticity of responses to different questions cannot be directly compared to each other. Even though the survey questionnaire emphasizes the importance of a doctor having diagnosed the ED, it is still possible that a part of the respondents understood the concept of ED in a different manner than what the researchers intended. A person may think that they suffer from ED due to being overweight. Assessed internationally, the Finns answer freely even to fairly personal questions, and the resulting findings can be thus regarded as reliable and generalizable.
The question concerning childhood adversities was not time-limited. It is thus possible that for some respondents childhood signifies primary school, while for others childhood ends when one turns 18 years old, and for yet another group of respondents it ends with the moving out from parental home. Furthermore, for some respondents such an amount of time may have elapsed since their childhood that they may not recall all events from that era or they misremember them. Both study and control groups may include such respondents thus reducing the possibility of research bias.
Our findings on alcohol may also be regarded as a limitation in this study. Passing out due to alcohol represents quite an extreme use of alcohol. We do not have the information on the standard units of alcohol consumed per time.
Childhood adversities have been found to be associated to the onset of several diseases. It is known today that also eating disorders are among these conditions. The more accumulated adversities an individual has experienced in their childhood, the stronger the connection is.
In the future, it might be best to concentrate on those children and young adults having experienced several adversities during their childhoods. By supporting their families, the onset of disease might be prevented or at least its degree of severity reduced. EDs are a long-term illness for many, and some never recover entirely. The onset of disease takes place early in life, at which time it often has negative consequences in one’s studies, in selecting a vocation and building a family.
It would be important to charter other childhood adversities in addition to those covered in the present study. The better we manage to investigate the risk factors involved in the onset of eating disorders, the more efficiently we can concentrate our resources to those most in need of support. It would also be essential to establish whether there are critical periods during which the children or young adults in question would especially benefit from help and support.
The authors declare that they have no competing interests.
VS drafted the manuscript, conducted the literature search, analyzed all data and wrote the paper. MS and KM participated in the formulation of the aims of the study, in the interpretation of the main findings and in drafting the manuscript. All authors read and approved the final manuscript.
Kristiina Helander has made a substantial contribution to the translation process. Centre for General Practice, Pirkanmaa Hospital District has supported the work financially.
VeeraSillanpää,KariMattila,MarkkuSumanen, (2015) Childhood Adversities Associated with Eating Disorders Reported by Adults. Health,07,1327-1336. doi: 10.4236/health.2015.710147
BED: Binge Eating Disorder; BDI: Beck’s Depression Inventory; BMI: Body Mass Index; CI: Confidence Interval; ED: Eating Disorder; HeSSUp: Health and Social Support Study; OR: Odds Ratio.