OUP user menu

Parental education as a predictor of offspring behavioural and physiological cardiovascular disease risk factors

Elisabeth Kvaavik , Maria Glymour , Knut-Inge Klepp , Grethe S. Tell , G. David Batty
DOI: http://dx.doi.org/10.1093/eurpub/ckr106 544-550 First published online: 5 September 2011


Background: Childhood socio-economic disadvantage has been shown to be associated with an elevated rate of cardiovascular disease (CVD) events in adulthood. The objective of this study is to examine associations between mothers’ and fathers’ education and offspring CVD risk factors. Methods: The Oslo Youth Study (n = 498) was initiated in 1979. Children (age 11–15 years) attending six schools and their parents were included. Information on education was collected for parents and participants. Participants were followed through 2006 (age 40 years). Information about physical activity, diet, smoking, binge drinking, body mass index (BMI), s-cholesterol, s-triglycerides and blood pressure was collected in 1981, 1991 and 2006. Results: Fathers’ education was inversely associated with participants’ BMI at 15 and 25 years, cholesterol at 25 and 40 years, triglycerides at 25 years and systolic blood pressure at 15 and 25 years (regression coefficients −0.18 to −0.11; P < 0.05 for all). The effects were weakened after adjusting for participants’ own education. Maternal education showed no association with these risk factors. After controlling for participants’ own education, associations between parental education and behavioural risk factors in adulthood were few. Conclusion: Any impact of parental education on offspring CVD risk factors seemed to be mediated via subject’s own education. Parental education offered little predictive capacity for offspring CVD risk factors.


Several studies have shown that childhood socio-economic disadvantage is associated with an elevated rate of adult cardiovascular disease (CVD) events.1–4 These results have prompted speculation as to the underlying mechanisms. One possibility is that childhood poverty influences later CVD via established risk factors. Several studies suggest this may be the case: pre-adult deprivation has been linked to an increased prevalence of future smoking, raised blood pressure, elevated blood cholesterol, physical inactivity and obesity.5

Remarkably few prior studies have explicitly examined both mother's and father's socio-economic position (SEP) as predictors of adult CVD risk.6–8 Comparing the importance of mothers compared to father's educational status for later CVD risk factors may give insight into key mechanisms linking childhood disadvantage and adult CVD risk. For example, until recently fathers had a substantially greater influence on material circumstances in two-parent families,9 while mothers may have more influence on behavioural norms or dietary patterns. Thus the long term effects of paternal SEP may be primarily mediated by own adult SEP, while mother's education may trigger behavioural pathways less closely associated with the offspring's own attained education. Finally, there is some evidence that maternal psychological and physiological factors relate more strongly to the characteristics of offspring than do those of the father; Some prior studies report mother–offspring correlations for body mass index (BMI), blood pressure and blood lipids, to be stronger than comparable father–offspring correlations,10,11 although this is not a universal observation.12–15

In the only study of which we are aware to have examined the differential influence, if any, of mother's and father's social characteristics on offspring CVD risk factors, contrary to expectations, paternal social class was more strongly related to BMI than maternal social class.6 However, this study did not collect data on any other important CVD risk factors (e.g. elevated blood cholesterol and blood pressure, cigarette smoking), and utilized retrospective recall of parental SEP. Adult recall of parental occupational social class shows, at best, moderate agreement with data collected contemporaneously in early life, potentially attenuating the measured associations between childhood SEP and CVD outcomes.2,16

In the Oslo Youth Study,17–19 CVD risk factors were measured for children at 13 and 15 years of age and both parents reported their own education. Subsequent measurement of selected CVD risk factors took place in these offspring at ages 25, 33 and 40 years. We are therefore able to examine associations of both parent's education with CVD risk factors in offspring followed into adulthood in more detail than prior studies.


The Oslo Youth Study began in 1979 when 1016 students in combined primary and secondary schools (mean age 13 years, range 11–16 years) attending six schools in Oslo were invited to participate in a school-based health education intervention. Begun in autumn 1979 and described in detail elsewhere,17,20,21 the health intervention aimed to discourage smoking initiation and modify physical activity and dietary habits. The students were administered a questionnaire and underwent a health examination in 1979, 1981, 1991, 1999 (questionnaire only) and 2006. The two initial (1979, 1981) waves of the study were approved by the Norwegian Data Inspectorate and Oslo City Health Authorities, while the subsequent follow-ups were approved by the Norwegian Data Inspectorate and the Regional Committees for Medical Research Ethics.

Assessment of SEP

Educational attainment of each parent [elementary school (7 years), 1 year of technical college (8–9 years), high school (10 years), high/comprehensive school (12 years) and college/university (>12 years)] was ascertained from a parental questionnaire administered in 1979 and 1981 and used as a proxy of childhood SEP. Data from 1979 were used in the present analyses, with substitution from 1981 if missing. Information about participants’ own educational attainment was collected from questionnaires at ages 25, 33 and 40 years (educational level attained at respective ages was used in the analyses regarding same age). The response categories were: elementary school (9 years), high school (10–11 years), high/comprehensive school (12 years), college/university (13–17 years) and college/university (at least 17 years).

Assessment of behavioural and physiological CVD risk factors

In 1979 and 1981 data collection took place in the schools. Subsequently, questionnaires were mailed to the study participants and respondents were invited to attend a health examination at the study centre in Oslo (1991) or with their physician (2006). These examinations included measurement of height, weight, pubertal development in 1979 and 1981,22,23 cardio-respiratory fitness19 (1979 and 1981) and waist and hip circumferences (1991 and 2006). BMI was computed [weight/(height)2kg/m2].

Resting blood pressure was ascertained using a random-zero sphygmomanometer in 1979 and 1981 (two measurements), while three readings taken by two trained nurses using a Dinamap device was used in 1991. In 2006, two measurements were taken by the participant's general practitioner (type of instrument not recorded). The mean of these values for each year was used herein. Serum measurements included total cholesterol, high-density lipoprotein (HDL) cholesterol, triglycerides (1979, 1981, 1991, 2006) and glycosylated haemoglobin (HbA1c) (2006). Both fasting and non-fasting values of serum measurements were used in the analyses.

The leisure time physical activity question in the questionnaire: ‘How often do you exercise for at least half an hour to the extent that you sweat and/or are short of breath?’ had similar response categories for all years (<2 times/month; 2–3 times/month; once a week; 2–3 times/week; 4–6 times/week; daily) except for 1979 and 1981 when the response ‘4–6 times/week’ was omitted. Questionnaires in 1979, 1981 and 1991 included items about intake frequency of sugared soft drinks, chocolate, cakes, confectionary, fruit, fruit juice, vegetables and potatoes. The fruit and vegetable score for 1979 and 1981 is based on the question ‘Did you have any of these yesterday?’ followed by the food items ‘Vegetables’, ‘Fruit’, ‘Potatoes’ and ‘Juice’ with response alternatives ‘At home’ (value 1) and ‘At school’ (value 1), both gave value 2, with zero awarded if reporting having consumed none of these. The responses were summed (range: 0–8). The unhealthy eating score for 1979 and 1981 was based on the question ‘Have you had any of the following in the past week?’: ‘Pastries/cakes’, ‘Soft drink/cola’, ‘Chocolate’, ‘Other sweets/candies’ and ‘Potato crisps’ with response alternatives ‘No’ (value 0) and ‘Yes’ (value 1). An enquiry about food frequency followed: ‘How often do you eat chocolate, sweets/candies, potato crisps etc?’ with four response options ranging from ‘Never’ (value 1) to ‘Almost daily’ (value 4). There was also an item about beverages: ‘Did you have soft drinks/cola yesterday’ (range: 1–11). Similar enquiries were made in more detailed food frequency questionnaires in 1991, 1999 and 2006. Data on binge drinking was collected by questionnaires in 1991, 1999 and 2006 by the question ‘With the last six months in mind, how many times have you been drunk?’ with six response alternatives ranging from ‘None’ to ‘More than 50 times’.

Assessment of parental risk factors

Parental self-reports of body weight and height were used to compute parental BMI [weight/(height)2 kg/m2]. Parental physical activity level and smoking habits in 1979 and 1981 were assessed by self-report. The physical activity measure was identical to the one used for the offspring in 1979 and 1981 with the smoking enquiry essentially unchanged.

Statistical analyses

Pearson correlation analysis was used to explore the correlations between parental and offspring's education. Linear and logistic regression analyses were used to summarize the relation of parental education with CVD risk factors in separate models for paternal and maternal education. For both mother's and father's education, we conducted regressions with and without adjustment for participants’ own education. In final models, we show coefficients for participant's own education after adjusting for father's education.

In preliminary analyses, there was no evidence of differential associations in male and female participants so data were pooled and sex adjusted. Given the narrow age range of the participants (11–15 years in 1979), no adjustment was made for chronological age. All analyses were adjusted for sex and intervention status; data from adolescence were controlled for pubertal development. All analyses were conducted using SPSS 14.0 and PASW 18.0.

At each wave, some participants completed the questionnaire on behavioural factors, but declined the physical health assessment. Of 827 participants in 1979 (mean age 13 years), complete data on behavioural and physiological factors were available for 498 and 501, respectively, and, of these, 462 and 325 subjects in 1981 (mean age 15 years), 407 and 219 in 1991 (mean age 25 years), 329 in 1999 (behavioural factors only collected) (mean age 33 years) and 240 and 161 in 2006 (mean age 40 years). Study participants who were pregnant or reported suffering from anorexia or bulimia nervosa were excluded from analyses in 1999 (n = 13) and 2006 (n = 1). In 1981, 1991 and 2006 relatively few participated in the physical examinations. In 1991 and 2006 questionnaires were mailed to participants, while the physical examination required that participants travelled to the examination site, which probably increased the drop-out rate. In 1981, filling in the questionnaire and the health examination was completed at school. Nearly all 1981-repsondents took part in the physical examination, but a high proportion did not take part in the fitness measurements and were excluded from the analyses. Missing data raises concerns regarding selection bias. We therefore compared the baseline characteristics (1979) of participants and non-participants in the 1991 physical examination. As we have previously demonstrated,24 there were few significant differences. Participants were more likely to be female (53 vs. 44% in 1991, P = 0.086) and to have fathers with higher education (P = 0.051). There were no differences for other variables. The differences in baseline characteristics between participants and non-participants in 1999 and 2006 were similar to those observed in 1991 except from baseline systolic blood pressure which was higher (110 vs. 108 mmHg) among 1999-participants than drop-outs (P = 0.027, t-test). Of the baseline variables (1979), only sex (female), sugar score (negatively), systolic blood pressure (positively) and school predicted participation in a subsequent follow-up of the survey (P < 0.05 for all, logistic regression analyses).


The correlation coefficient between mothers’ and fathers’ education in 1981 was 0.65 (P < 0.001, n = 462). The correlations between fathers ‘education in 1981 and participants’ education in 1991, 1999 and 2006 were 0.49, 0.45 and 0.50 (P < 0.001 for all), respectively. Corresponding correlations between maternal and participants’ education were 0.43, 0.37 and 0.35 (P < 0.001 for all), respectively. Table 1 presents values of the variables included in the main analyses.

View this table:
Table 1

Characteristics of participants in the Oslo Youth Study follow-up

Year of examination19791981199119992006
Background variables (N)n = 498n = 462n = 407n = 329n = 240
Age (years)a13.0 (0.9)14.9 (0.9)25.0 (0.9)33.0 (0.9)40.0 (0.9)
Education (%)
    Elementary school (9 years)
    High school (10–11 years)20.915.812.1
    High/comprehensive school (12 years)31.025.825.0
    College/university (1–4 years)33.232.834.2
    College/university (>4 years)10.120.725.4
Fathers’ education (%)b
    Elementary school (7 years)15.3
    Elementary school, 1-year technical college (8–9 years)28.7
    High school (10 years)13.5
    High/comprehensive school (12 years)13.1
    College/university (>12 years)29.5
Mothers’ education (%)b
    Elementary school (7 years)15.7
    Elementary school, 1-year technical college (8–9 years)34.3
    High school (10 years)20.7
    High/comprehensive school (12 years)15.5
    College/university (>12 years)13.9
Behavioural variables (N)n = 498n = 462n = 407n = 329n = 240
    Fruit and vegetable intakec2.8 (1.2)2.6 (1.2)7.7 (2.7)343 (191)451 (325)
    Unhealthy eatingc6.3 (2.2)6.6 (2.3)4.5 (2.4)
    Sweet food intakec357 (489)171 (358)
    Physical activity, twice/week (%)57.862.647.737.145.4
    Smoking, daily (%)42.530.121.7
    Binge drinking, drunk ≥ 11 times last 6 months (%)22.611.911.3
Physiological variables (N)n = 501n = 325n = 219n = 329n = 161
    Tanner staged2.5 (1.2)3.8 (0.9)
    Cardiorespiratory fitness (ml/kg/min)49.5 (10.8)52.5 (10.7)
    BMI (kg/m2)18.5 (2.3)19.6 (2.4)23.3 (3.3)24.3 (3.8)24.8 (3.8)e
    Total blood cholesterol (mmol/l)4.7 (0.8)4.8 (0.8)4.8 (0.8)5.1 (0.9)
    HDL cholesterol (mmol/l)1.3 (0.3)1.4 (0.3)1.3 (0.3)1.6 (0.5)
    Triglyceride (mmol/l)0.6 (0.4)0.9 (0.4)1.3 (0.8)1.1 (0.7)
    Systolic blood pressure, (mmHg)109 (12)112 (13)129 (13)119 (13)
    Diastolic blood pressure (mmHg)58 (11)58 (11)77 (8)75 (10)
    HbA1c (%)5.3 (0.3)
  • a: Values are mean (SD) unless otherwise stated.

  • b: Father's and mother's individual education collapsed for 1979 and 1981.

  • c: Scores computed differently in 1991 compared with 1979 and 1981, in 1999 and 2006 food intake was measured in grams per day.

  • d: Scale 1–5.

  • e: Self-reported BMI among 240 participants.

Figure 1 shows the associations between parental education and participants’ behavioural risk factors, including: unhealthy eating habits, fruit and vegetable intake, leisure time physical activity and binge drinking. Mothers and fathers with higher education had offspring with a lower unhealthy eating score and higher fruit and vegetable intake at age 15 years, while father's education predicted healthier eating at age 40. Mother's and father's education each predicted physical activity at age 33, but not at other ages. The associations did not change materially after adjusting for physical activity among mothers and fathers. There were no statistically significant associations with unhealthy eating or fruit and vegetable intake at other ages, and after adjustment for own education, the association with father's education and age-40 eating patterns was non-significant.

Figure 1

Standardised regression coefficients (B) for the relation of parental educational level to unhealthy eating habits (left a) and fruit and vegetable intake (right a) at ages 15, 25, 33 and 40 years, leisure time physical activity at ages 15, 25, 33 and 40 years (b) and binge drinking at ages 25, 33 and 40 years (c): the Oslo Youth Study follow-up. *P < 0.05, **P < 0.01, ***P < 0.001

Although neither father's nor mother's education predicted binge drinking in models adjusted only for sex and intervention status, after adjustment for the subject's own educational achievement, mother's education was positively associated with binge drinking in the offspring at the age of 40 years.

Smoking was not associated with parental education after adjustment for sex, intervention status and parental smoking (results not shown), but after additional adjustment for own education, higher education for either parent was associated with higher odds of daily cigarette smoking at the age of 25 years; odds ratio (OR) and 95% confidence interval (95% CI) for daily smoking among participants was 1.18 (1.00–1.40) among those whose father had a high educational level. Corresponding value among those whose mother had a high educational level was OR = 1.27 and 95% CI = 1.06–1.53.

In table 2, we present the associations between the offspring's own education and behavioural CVD risk factors. Own education predicted sweet food intake at ages 33 and 40 years and physical activity at ages 25 and 33 years, although the former became non-significant after adjusting for father's education. Own education predicted cigarette smoking at all ages, higher education was associated with lower odds of cigarette smoking.

View this table:
Table 2

Standardised regression coefficients (95% CIs) for the relation of adult education with behavioural CVD risk factors: the Oslo Youth Study follow-up

Age (years)aNAdult education
Sex and intervention adjusted+ father's education adjusted
Unhealthy eating25407−0.08 (−0.17, 0.02)−0.07 (−0.18, 0.04)
Sweet food intake33280−0.12 (−0.24, −0.01)−0.11 (−0.24, 0.01)
40240−0.18 (−0.30,−0.05)−0.11 (−0.26, 0.03)
Fruit and vegetable intake254070.07 (−0.03, 0.17)0.04 (−0.08, 0.15)
333290.004 (−0.11, 0.11)−0.02 (−0.14, 0.10)
402400.09 (−0.04, 0.22)0.09 (−0.06, 0.23)
Physical activity254070.13 (0.04, 0.23)0.13 (0.02, 0.24)
333290.18 (0.08, 0.29)0.16 (0.04, 0.28)
402400.11 (−0.02, 0.24)0.14 (−0.01, 0.29)
Binge drinking254070.02 (−0.07, 0.11)0.01 (−0.10, 0.11)
33328−0.03 (−0.13, 0.08)−0.06 (−0.18, 0.06)
40240−0.07 (−0.19, 0.06)−0.14 (−0.28, 0.01)
Daily cigarette smoking254070.56 (0.45, 0.69)0.49 (0.39, 0.63)
333290.59 (0.47, 0.75)0.55 (0.42, 0.71)
402400.57 (0.43, 0.76)0.52 (0.37, 0.72)
  • a: Mean age at risk factor measurement.

Table 3 depicts the relation of both parents’ educational achievement and the subjects own with several physiological CVD risk factors. There was an inverse association between paternal education and offspring BMI at ages 15 and 25 years, such that children whose fathers had higher education were leaner in adolescence and adulthood (table 3). These associations were only marginally changed after adjustments for paternal BMI (results not shown). Commensurate with this observation, higher paternal education was related to more favourable levels of total cholesterol in the offspring at age 25 and 40 years, triglycerides at 25 years, and systolic blood pressure at 15 and 25 years. There was no evidence that either waist-hip ratio or HbA1c in adulthood or cardio-respiratory fitness during adolescence was associated with father's education (results not shown). Maternal education did not predict any of the physiological risk factors except BMI at age 15 years. After adjustment for participant's own education (for observations at age ≥25 years), neither father's nor mother's education significantly predicted any of the physiological risk factors.

View this table:
Table 3

Standardised regression coefficientsa (95&percnt; CI) for the relation of parental and own education to physiological CVD risk factors: the Oslo Youth Study follow-up

Age (years)aNFather's educationMother's educationAdult education
Sex and intervention adjustedb&plus;Own education adjustedSex and intervention adjustedb&plus;Own education adjustedSex and intervention adjusted&plus; father's education adjusted
BMI15325&minus;0.13 (&minus;0.23, &minus;0.03)&ndash;&minus;0.10 (&minus;0.20, &minus;0.001)&ndash;&ndash;&ndash;
25219&minus;0.13 (&minus;0.27, &minus;0.003)&minus;0.06 (&minus;0.22, 0.09)&minus;0.07 (&minus;0.20, 0.06)0.003 (&minus;0.14, 0.15)&minus;0.17 (&minus;0.30, &minus;0.04)&minus;0.14 (&minus;0.29, 0.01)
40240c&minus;0.06 (&minus;0.18, 0.07)0.01 (&minus;0.14, 0.15)&minus;0.03 (&minus;0.15, 0.09)0.02 (&minus;0.12, 0.15)&minus;0.12 (&minus;0.25, &minus;0.001)&minus;0.13 (&minus;0.27, 0.02)
Total cholesterol15325&minus;0.002 (&minus;0.11, 0.10)&ndash;&minus;0.02 (&minus;0.12, 0.09)&ndash;&ndash;&ndash;
25219&minus;0.18 (&minus;0.32, &minus;0.05)&minus;0.09 (&minus;0.24, 0.06)&minus;0.11 (&minus;0.25, 0.02)&minus;0.03 (&minus;0.18, 0.12)&minus;0.23 (&minus;0.36, &minus;0.10)&minus;0.18 (&minus;0.34, &minus;0.03)
40161&minus;0.16 (&minus;0.31, &minus;0.004)&minus;0.12 (&minus;0.30, 0.06)&minus;0.01 (&minus;0.16, 0.15)0.05 (&minus;0.12, 0.21)&minus;0.13 (&minus;0.28, 0.02)&minus;0.08 (&minus;0.25, 0.10)
HDL cholesterol15325&minus;0.10 (&minus;0.21, 0.01)&minus;&minus;0.11 (&minus;0.22, 0.002)&ndash;&ndash;&ndash;
25219&minus;0.01 (&minus;0.12, 0.11)&minus;0.10 (&minus;0.23, 0.04)0.06 (&minus;0.06, 0.17)0.01 (&minus;0.12, 0.13)0.13 (0.02, 0.24)0.18 (0.05, 0.31)
40161&minus;0.04 (&minus;0.20, 0.11)&minus;0.09 (&minus;0.27, 0.09)0.11 (&minus;0.05, 0.27)0.11 (&minus;0.06, 0.28)0.04 (&minus;0.11, 0.20)0.09 (&minus;0.09, 0.26)
Triglycerides153250.06 (&minus;0.05, 0.17)&minus;0.05 (&minus;0.06, 0.16)&ndash;&ndash;&ndash;
25219&minus;0.13 (&minus;0.26, &minus;0.01)&minus;0.04 (&minus;0.18, 0.10)&minus;0.05 (&minus;0.18, 0.07)0.04 (&minus;0.09, 0.17)&minus;0.21 (&minus;0.33, &minus;0.08)&minus;0.19 (&minus;0.33, &minus;0.04)
40161&minus;0.14 (&minus;0.29, 0.01)&minus;0.11 (&minus;0.29, 0.06)&minus;0.12 (&minus;0.27, 0.03)&minus;0.10 (&minus;0.26, 0.06)&minus;0.10 (&minus;0.24, 0.05)&minus;0.04 (&minus;0.21, 0.12)
Systolic blood pressure15325&minus;0.14 (&minus;0.24, &minus;0.03)&ndash;&minus;0.10 (&minus;0.21, 0.002)&ndash;&ndash;&ndash;
25219&minus;0.12 (&minus;0.23, &minus;0.01)&minus;0.12 (&minus;0.25, 0.02)&minus;0.07 (&minus;0.18, 0.05)&minus;0.05 (&minus;0.17, 0.08)&minus;0.07 (&minus;0.18, 0.05)&minus;0.01 (&minus;0.14, 0.12)
40161&minus;0.08 (&minus;0.23, 0.07)&minus;0.07 (&minus;0.24, 0.11)&minus;0.09 (&minus;0.24, 0.06)&minus;0.08 (&minus;0.24, 0.09)&minus;0.06 (&minus;0.21, 0.09)&minus;0.03 (&minus;0.20, 0.14)
Diastolic blood pressure15325&minus;0.03 (&minus;0.13, 0.08)&minus;&minus;0.03 (&minus;0.14, 0.08)&ndash;&ndash;&ndash;
25219&minus;0.04 (&minus;0.17, 0.09)&minus;0.09 (&minus;0.24, 0.06)&minus;0.05 (&minus;0.18, 0.08)&minus;0.08 (&minus;0.22, 0.06)0.05 (&minus;0.08, 0.18)0.09 (&minus;0.06, 0.25)
40161&minus;0.03 (&minus;0.18, 0.12)&minus;0.02 (&minus;0.20, 0.15)0.03 (&minus;0.12, 0.18)0.05 (&minus;0.12, 0.21)&minus;0.02 (&minus;0.17, 0.13)&minus;0.01 (&minus;0.18, 0.16)
  • a: Mean age (years) at risk factor measurement.

  • b: Analyses of data for age 15 years are also adjusted for maturation (Tanner scale).

  • c: Self-reported weight and height.

Participants' own education in adulthood predicted total and HDL cholesterol and triglycerides at 25 years of age as well as BMI at 25 and 40 years of age. In all cases, a higher level of education was associated with a more favourable CVD risk profile.


In a well-characterized sample of Norwegian children followed from 1979 through 2006, we found associations of paternal education with several adult CVD behavioural and physiological CVD risk factors and most of these associations were attenuated after adjusting for participants’ own education. Maternal educational level showed no consistent associations with most risk factors.

Prior studies

Few studies have examined the effect of father's and mother's education on offspring's CVD risk factors. Ball and Mishra6 found that childhood socio-economic status was inversely associated with adult BMI, with a stronger association with father's than for mother's education. Our findings for BMI are consistent with these results. Although we expected childhood diet to be more closely linked to maternal education because of mothers’ typical role in food preparation, father's education was a stronger predictor than mother's for both indicators of diet. That a significantly lower proportion of Norwegian women than men had higher education in the 70s and 80s might partly explain the weak association between maternal education and offspring CVD risk factors 20–30 years later. In contrast to other studies, we found no evidence for an effect of either parental or own education on fruit and vegetable intake in adulthood.25,26

Contrary to some studies,27–30 much of the association of early socio-economic status with later risk factors was lost when the participants’ own education was taken into account. These attenuations therefore could be due to mediation of parental SEP effects by the offspring's own education; to confounding of the associations between offspring education and CVD risk factors; or to insufficient statistical power to detect modest direct effects. Achieved education in adulthood might be a measure of early life circumstances,31 and therefore, adjusting for own adult education might represent an over-adjustment. Analyses in which a consequence of the dependent variable is included as an independent variable are commonly biased towards the null, so analyses should be interpreted with great caution.

In accordance with results from other studies, we showed that smoking in adulthood was more strongly associated with own education than parental education.32,33 Surprisingly, smoking at age 25 years was positively associated with higher parental education after adjustment for own education. Findings from other studies are equivocal, but mostly they have found that upwardly mobile subjects have more favourable lifestyles than the class they left, but less favourable than the class they join.33,34

The only other behavioural risk factor in adult life that was predicted by parental education after adjusting for own education was binge drinking at age 40 years, for which the risk increased with increased parental education. This is discordant with results from some studies,35 but an earlier study from Norway has found that alcohol consumption was inversely related to educational attainment.36

Study strengths and limitations

The strengths of this study include repeated assessments of physiological and behavioural CVD risk factors across the life course until middle age, and prospectively self-reported data on education. It is not, however, without shortcomings. First, there was attrition at each attempt to follow-up the study members. However, analyses of the characteristics of people lost to follow-up vs. those retained revealed few differences. This loss to follow-up resulted in a small sample size at age 40 years, and reduced statistical power. However, the direction of the associations was as hypothesized, with negative associations between educational level and CVD risk factors. As discussed above, our approach to testing for direct effects of parental education by adjusting for the potential mediator (own education) has been shown to be inappropriate if there are unmeasured confounders of own education and the CVD risk factors, or if own education modifies the effect of parental education on the CVD risk factors.37–39

Parental and own education was the only measure for childhood and adult socio-economic status applied in the current study. Even if it is regarded an excellent measure of social position, other measures of childhood social status, such as parental occupation, income, or measures of poverty may represent different meanings of SEP than educational level.40 Most of the parents would have completed their education by the time of the assessment, and therefore parents’ age should not be a concern. Using parental income as a measure of SEP has a life course trajectory and might therefore be a less reliable measure.

In conclusion, paternal education predicted aspects of diet, BMI and systolic blood pressure in adolescence; associations were generally weaker and less consistent for maternal education. After controlling for subject's own educational level, most associations with parental education were no longer statistically significant.


The Oslo Youth Study was initially supported by The Norwegian Cancer Society. Subsequent support includes grants from the Research Council of Norway, the EXTRA funds from the Norwegian Foundation for Health and Rehabilitation and from the Norwegian Health Association. E.K. was funded by a grant from the Norwegian Research Council (2006-2009) for current data collection and analyses; G.D.B. a UK Wellcome Trust Fellowship; M.G. is funded by the American Heart Association (10SDG2640243).

Conflicts of interest: None declared.

Key points

  • Childhood socio-economic disadvantage measured by paternal education is associated with an elevated rate of CVD risk factors.

  • Whether maternal SEP as an index of childhood SEP is associated with adult CVD risk factors is unknown.

  • Paternal, but not maternal, education predicted behavioural and physiological risk factors in their offspring. However, most associations were no longer significant after adjustments for subject's own education in adulthood.

  • Parental education offered little predictive value above subjects’ own education.

  • These findings suggest that own education is more important than parental education in shaping adult CVD risk factors, emphasizing the importance of own education.


G.S.T. and K.-I.K. designed and conducted the baseline and early follow-up surveys for the Oslo Youth Study. E.K. and G.D.B. generated the idea for the present manuscript which was developed by the co-authors. G.D.B. and E.K. wrote the first draft around analyses conducted by E.K. M.G. contributed substantially to data analyses and interpretation. All authors have contributed to critical revision of the manuscript, and have read and approved of the final manuscript.


View Abstract