Characteristics of the study population were presented as the mean ± SD or median (IQR) for continuous variables and the number (frequency) for categorical variables according to sex. Student’s t tests or Wilcoxon rank sum tests were used to compare the differences between groups for continuous variables. The
χ2 test was used to compare the differences between groups for categorical variables. Concordance for cardiovascular risk factors was defined as a case in which both spouses had the same response for a category of variables.
The univariate Spearman correlation coefficient (
r) and the Phi coefficient were used to assess the correlations of continuous and categorical variables within couples, respectively. Spousal correlations for metabolic indicators (BMI, SBP, DBP, FBG, TC, HDL-C, LDL-C, TG, and UA) might emerge because of the associations of these indicators with age. Therefore, we fitted a regression of metabolic indicators against age and derived the residual for these indicators adjusted for individuals’ age. We then used the residuals to calculate the spousal correlations in the metabolic indicators; namely, the correlation coefficients were corrected for the age of both partners (Model 1). Since previous studies reported that BMI is a surrogate for assortative mating [22 (
link)], we further adjusted the BMI of both spouses in Model 2 using the same method.
The
χ2 test was performed to explore the crude association of cardiovascular risk factors between couples. The odds ratios (ORs) and their corresponding 95% confidence intervals (CIs) were estimated by age-adjusted and multivariable-adjusted logistic regression models to indicate the spousal association for each given cardiovascular risk factor. The multivariable model of lifestyle factors adjusted the age, education, annual income, and geographic regions of the individuals. The multivariable model of cardiometabolic disease was further adjusted for individuals’ smoking, drinking, leisure-time physical activity, and overweight/obesity status. Family history of hypertension and diabetes was additionally adjusted in the models of hypertension and diabetes, respectively. Spousal similarity was defined as
r > 0 and its corresponding
p value <0.05 for continuous variables and ORs >1 and their corresponding 95% CIs excluding 1 for categorical variables.
To explore potential changes in spousal similarities with age, roughly representing marriage duration, subgroup analyses were performed according to the individuals’ age (20–50 years, ≥50 years). Multiplicative interaction was calculated by cross-product interaction terms in multivariable logistic regression models. A sensitivity analysis was performed, excluding couples with an age difference ≥5 years, to evaluate the robustness of the results.
All analyses were performed separately for husbands and wives. The significance level was set as a 2-sided
p value < 0.05. SAS 9.4 (SAS Institute Inc., Cary, NC, United States.) was used to conduct all analyses.
Lin B., Pan L., He H., Hu Y., Tu J., Zhang L., Cui Z., Ren X., Wang X., Nai J, & Shan G. (2023). Spousal Similarities in Cardiovascular Risk Factors in Northern China: A Community-Based Cross-Sectional Study. International Journal of Public Health, 68, 1605620.