The role of current circumstances in the associations between

health behaviour

D a i l y s m o k i n g

The strong negative associations between parental education and the respondent’s daily smoking in both gen ders attenuated when all childhood adversities were adjusted for (Figure 6, Model II; Substudy II, Tables 4 and 5). Further, the in clusion of parental smoking in the model further slightly reduced the associations in women (Substudy II, Table 4). After adjusting for the respondent’s own education, the associa tions lost their statistical signifi cance in both genders (Figure 6; Model IV, Substudy II, Tables 4 and 5). Daily smoking was more common in women who had lived in a single-parent family as a child than in those who had lived with both single-parents (however, the difference was statistically signifi cant only in the age group 1824 years, see Table 7).

The association lost its statistical signifi cance after all childhood circumstances were adjusted for (Figure 6, Model II).

In age-adjusted models, parental divorce showed statistically signifi cant associations with daily smoking in both genders, as did parental unemployment in men and long-term fi nancial problems, maternal alcohol and mental health problems and serious confl icts within the family in women (Figure 6, Model I). Almost all associations lost their statistical signifi cance after controlling for all childhood conditions. However, in some of them, adjusting for the respondent’s own education seemed to reduce the associations further (Figure 6, Model II). The association between daily smoking and parental divorce (OR = 2.66) remained statistically signifi cant in women even when all indicators relating to current circumstances were controlled for (Figure 6, Model IV).

Daily smoking varied strongly according to parental smok ing in both genders. The risk of daily smoking was increased if both the respondents’ parents smoked, even if all childhood and current circumstances were adjusted for (Figure 6, Model IV). The effect of a smoking father on men’s smoking also remained statistically signifi cant despite all adjustments. In women, the signifi cant association between maternal smoking and the respondent’s smoking lost its statistical signifi cance when the effect of the respondent’s own education was adjusted for (Figure 6, Model III).

Figure 6. Daily smoking by childhood circumstances in men and women aged 1829 years (OR). Adjusted for age, childhood circumstances^1,2 (parental education, childhood family structure, childhood adversities and parental smoking), the respondent´s own education and other current circumstances.

* p < 0.05

1 Only statistically signifi cant childhood circumstances (p < 0.05) for either gender are presented (for others, see Substudy II, Tables 2, 4 and 5).

2 Parental primary level education compared to the highest educational category; single-parent family compared to two-parent family; both parents smoking, paternal smoking and maternal smoking compared to those whose neither parent smoked.

Adversities compared to those who did not report the specifi c problem in childhood.

H e a v y d r i n k i n g

Altough parental education was not statistically signifi cantly associated with heavy drinking, there were some apparent associations that did not quite reach statistical signifi cance. For example, women with low parental education seemed to have a lower risk and men with middle parental education a higher risk of being a heavy drinker, even in the model that adjusted for all childhood and current circumstances (Substudy III, Tables 3 and 4).

In men, the association between heavy drinking and having lived in a single-parent family in childhood (OR = 3.23) remained statistically signifi cant when parental education was adjusted for, but disappeared after other childhood adversities were included in the model (Figure 7, Model II; Substudy III, Table 2), mainly because of the effect of parental divorce. The association further decreased slightly after adjusting for the respondent’s own education (Figure 7, Model III).

In women, long-term fi nancial problems (OR = 2.43) and serious confl icts within the childhood family (OR = 2.31) were associated with heavy drinking. The association between long-term fi nancial problems and heavy drinking remained statistically signifi cant despite all adjustments. In addition, the association between serious confl icts within the family and heavy drinking remained suggestive after controlling for the other explanatory variables (Figure 6, Model IV; Substudy III; Table 3).

In men, parental divorce, parental mental health problem, parental serious illness or disability and hav ing been bullied at school were associated with heavy drinking in the age-adjusted models (Figure 6, Model I; Substudy III, Table 4). However, the associations between heavy drinking and parental divorce as well as parental serious illness or disability disappeared after controlling for the respondent’s educational level (Figure 7, Model III). The associations with paren tal mental health problems and having been bullied at school lost their statistical signifi cance when other childhood circumstances were controlled for (Figure 7, Model II). However, the association with having been bullied at school re-emerged when all factors were adjusted for (Figure 7, Model IV).

Statistically signifi cant associations were found between both parents’ alcohol problems and heavy drinking in both gen ders (Figure 7, Model I; Substudy III, Tables 3 and 4).

In women, this association lost its signifi cance when other childhood circumstances were controlled for (Figure 6, Model II). In men, the strong association remained statistically signifi cant even af ter adjusting for other childhood circumstances, own education and main activity, but after controlling for current family structure and degree of urbanisation of current residence, the association did not quite reach statistical signifi cance (Figure 8, Model IV). However, men whose both parents had an alcohol problem dif fered from the reference category at the 0.1 signifi cance level in the last model as well (Figure 8, Model IV).

Figure 7. Heavy drinking by childhood circumstances in men and women aged 1829 years (OR). Adjusted for age, childhood circumstances^1,2 (parental education, childhood family structure, childhood adversities and parental alcohol problems), the respondent’s own education and other current circumstances (main activity, current family structure and place of residence).

* p < 0.05

1 Only childhood circumstances statistically signifi cant for either gender are presented (for others, see Substudy III, Table 2).

2 Single-parent family compared to two-parent family; parental alcohol problems compared to those whose neither parents had alcohol problems. Adversities compared to those who did not report the specifi c problem in childhood.

O v e r w e i g h t a n d o b e s i t y

The clear negative association found in women between parental education and overweight (25 < BMI < 30) remained relatively unchanged after all childhood circumstances were included in the model, but lost its statistical signifi cance when adjusting for the respondent’s own educational level (Substudy IV, Table 4a). In both genders, the strong association between parental education and obesity (BMI > 30) remained strong after all adjustments, although not quite reaching statisti cal signifi cance

in women in the model adjusted for all current circumstances (Figure 8, Model V;

Substudy IV, Tables 4a and 4b). The signifi cant association between primary parental education and obesity in women lost its statistical signifi cance when the respondent’s own education was adjusted for (Figure 8, Model III).

The associations of growing up in a single parent family with both overweight and obesity were margin ally statistically signifi cant in the age-adjusted model in women (Figure 8, Model I; Substudy IV, Table 4a), but attenuated when parental education was adjusted for (Substudy IV, Table 4a). In men, childhood family structure was not associated with either overweight or obesity (Figure 8; Substudy IV, Table 4b).

Figure 8. Obesity (BMI � 30) by childhood circumstances in men and women aged 18�29 years (RRR) compared to normal weight respondents (18.5 � BMI < 25). Adjusted for age, childhood circumstances^1,2,3, the respondent’s own education, other current circumstances⁴ and health behaviour⁵.

* p < 0.05

1Only childhood circumstances signifi cant for either gender are presented (for others, see Substudy IV, Tables 3, 4a and 4b).

2Parental education, childhood family structure and 11 childhood adversities.

3Parental primary level education compared to the highest educational category; single-parent family compared to two-parent family; parental alcohol problems compared to those whose neither parents had alcohol problems. Adversities compared to those who did not report the in childhood.

4Main activity, current family structure, place of residence.

5Smoking, drinking, physical activity, use of vegetables.

RRR RRR

Women who had lived in a rural municipality in childhood were more likely to be obese than those from semi-urban or urban municipalities (Figure 8, Model I;

Substudy IV, Table 4a). When current residence and main economic activity were included in the model, the association was no longer statistically signifi cant (Figure 8, Model IV), but remained suggestive despite the rest of the adjustments (Figure 8, Model V). For men, childhood residence was not statistically signifi cantly associated with overweight or obesity (Figure 8).In women, regular parental unemployment was statistically signifi cantly associated with overweight (RRR = 2.09) (Substudy IV, Table 4a), whereas parental alcohol and mental health problems and being bullied at school were associated with obe sity in the age-adjusted model (Figure 8, Model I;

Substudy IV, Table 4a). In addition, serious confl icts within the family were associated with obesity at the p < 0.1 level. Controlling for other childhood factors decreased the RRRs (Figure 8, Model II), and only that of being bullied at school remained a statistically signifi cant predictor of obesity, despite all adjustments (Figure 8, Model V). In men, none of the childhood adversities was statistically signifi cantly associated with overweight or obesity.

6.4 Explanatory effects of childhood circumstances,