Descriptive characteristics of the study sample, categorized by high and low parental emotional care are shown in . Females who reported high parental emotional care were younger (p = .04), reported lower depressive symptomatology (p = .02) and smoking prevalence (p = .009), and higher educational levels (p = .04), compared to females with low parental emotional care. Males reporting high parental emotional care had a non-significantly higher likelihood of smoking (p = .10) compared to those reporting lower parental emotional care. Parental emotional care scores did not differ between men and women (score=8.5 for males and 8.3 for females; t-test p=0.58).
Characteristics of study participants, according to quality of parental emotional care.
When the study population was classified by quartiles of the care scores, mean 10-year CHD risk among female offspring decreased with increasing parental emotional care score (p for trend = .02). Among male offspring, parental emotional care was not conclusively associated with elevated 10-year CHD risk (p = .12) ().
Calculated 10-year coronary heart disease (CHD) risk score (% risk for coronary heart disease in the next 10 years) according to quartile of parental emotional care.
In unadjusted linear regression analyses among female offspring, average parental care score was inversely associated with CHD risk. In order to obtain results on the non log-transformed scale, regression beta coefficients were exponentiated and interpreted in terms of percent change of the non-transformed variable (29
). Thus, a one unit increase in average parental emotional care score in females resulted in a 5.7% decrease [1-exp (-0.059)] in CHD risk (p
= .002). Further adjustment for childhood SEP, which is expected to influence the quality of parental emotional care, reduced the magnitude of the regression coefficient to 0.047, i.e. a decrease in CHD risk of 4.6% (p
= .009). Additional adjustment for adulthood SEP, depressive symptomatology, and BMI did not further change the results (p
= .004). Therefore, if the average 10-year CHD risk among females in our study sample is 2.5%, then our models suggest that a one-unit increase in parental emotional care would decrease this risk by 4.6% after adjusting for covariates, resulting in a new 10-year CHD risk of 2.4%. A larger decrease of 5 points on the care scale would result in a 10-year CHD risk of 3.2%. In males, a one unit increase in reported parental emotional care score resulted in a statistically non-significant increase of 2.5% in average CHD risk (p
= .22), and this association was not markedly influenced by further adjustment for covariates ().
Multivariable-adjusted linear regression analyses for the association between parental emotional care score and log-transformed 10-year coronary heart disease risk
In regression analyses for the association between the care score and individual components of the Framingham risk algorithm, we found that among female offspring, a one unit increase in reported parental emotional care decreased the odds of smoking in unadjusted analyses (odds ratio = 0.84, p = .004), and the association remained significant after adjusting for childhood SEP, but was no longer significant in the fully-adjusted model (). Furthermore, there was a decrease in depressive symptomatology (β = -0.036, p = .01) per unit increase in care score, but this was no longer significant in the fully adjusted model (). Among males, only depressive symptomatology was significantly associated with the care score (β = -0.071, p = .001) ().
Linear and logistic regression analyses for associations between parental emotional care score and CHD risk factors.
The variance (R2) in 10-year CHD risk score explained by the independent variable and each covariate in linear regression analyses is shown in . For female offspring, BMI explains almost 25% of the variance in CHD risk while all other variables explain between 5.0 and 9.8%. Parental emotional care explained 5.8% of the variance in 10-year CHD risk score. This showed that it explained fairly comparable amounts of variance to other proposed social and psychological determinants of CHD (depression, childhood SEP and educational attainment) but substantially less than that for obesity. Among males, all of the covariates independently explained only between 0.7 and 4.8% of the variance in CHD risk; the care score explained only 1.5% of the variance.
Univariate (unadjusted) linear regression associations between the independent variables and logarithmic 10-year CHD risk, in female and male offspring
In secondary analyses that assessed maternal and paternal care components separately, instead of as a mean score for both parents, associations between emotional care and CHD risk among female offspring in the full model were similar to that for the average care score for both parents: a one unit increase in maternal care score resulted in a 3.5% decrease in CHD risk (β = -0.035, p = .01), and a one unit increase in paternal care score caused a 3.6% decrease in risk (β = -0.036, p = .02). Among male offspring, while no significant association was found between maternal care and risk (β = 0.012, p = .58), a weak positive association was observed between paternal care and risk, in the fully-adjusted model (β = 0.033, p = .05).
Sensitivity analyses demonstrated that additional adjustment for race/ethnicity did not markedly change associations of mean parental emotional care score with calculated 10-year CHD risk (females: β = -0.047, p
= .01; males: β = 0.009, p
= .67 in fully adjusted models). Further sensitivity analyses evaluated whether the association between parental emotional care and calculated 10-year CHD risk among females was primarily due to the association between parental care and smoking (as seen in ). Multivariable regression analyses were performed on associations between parental emotional care and the calculated 10-year CHD risk with the smoking component removed from the Framingham algorithm. Findings remained strong for associations in females (see Table, Supplemental Digital Content 1
, for details of the analysis). Table A