Effect of gender on the association between cumulative cardiovascular risk factors and depression: results from the US National Health and Nutrition Examination Survey

Study population

In the 1960s, the US National Health and Nutrition Examination Survey (NHANES) Project began to evaluate the nutritional and health status of adults and children.9 Interviews (questionnaires on demographic, socioeconomic, nutritional and health-related factors) and laboratory testing (medical, dental and physiological measurements) comprised most of the NHANES data. The study population employed in this cross-sectional study was derived from continuous NHANES 2005–2016 evaluations (n=58 660). In this secondary analysis, we excluded individuals with missing data regarding their depression and demographic status and other related variables such as living status, self-rated health and sleep duration; the result was a study population of 20 816 participants (figure 1). Due to insufficient information on low-density lipoprotein (LDL) cholesterol and triglyceride (TG) levels in much of the collected data, we further reduced the number of participants to 9906 for the LDL and TG analysis (figure 1).

Request permission

Depression status

The definition of depression for the study was based on the responses to nine questions in the Patient Health Questionnaire-9 (PHQ-9); a participant with a final score over 15 was considered depressive, as reported in a previous study.10 Moreover, participants in the sample who reported taking antidepressants were also defined as depressive.

CVD risk factors and cumulative risk exposure

The current study included 18 reported CVD risk factors divided into three categories: lifestyle behaviours, biomarkers and chronic disease status. The lifestyle behaviour risk factors included sedentary status, smoking, alcohol use, living status, self-rated health, sleep quality and sleep duration. The biomarker category included levels for high-density lipoprotein (HDL), cholesterol, LDL, TG, uric acid, inflammatory markers (white blood cell (WBC) count and the neutrophil–lymphocyte ratio (NLR)) and adiposity indicators (body mass index (BMI) and waist circumference (WC)). The clinical thresholds for coding biomarkers from continuous to categorical variables were according to the Adult Treatment Panel III guidelines,11 National Institutes of Health guidelines12 and other studies13–15 (online supplemental table 1). The chronic disease status category included CVD, diabetes, hypertension and cancer. For depression prediction by the cumulative risk model, 18 CVD risk factors were dichotomised (ie, 1=high risk, 0=no or low risk) and summed to derive a cumulative risk index; higher scores represented higher levels of cumulative risk exposure of multiple risk factors.

Confounding variables

The potential confounders used for adjustment while investigating the association between CVD risk factors and depression were similar to those used in our previous publication16: gender, age, race, education and poverty–income ratio. The latter, the official US Census measure of poverty, evaluates the socioeconomic status of individuals and households within the poverty threshold.

Statistical analysis

Statistical differences in categorical variables between participants were examined by the Rao-Scott χ² test. The association between CVD risk factors and depression status was investigated by logistic regression. Because the effects of CVD risk factors on depression differ in specific populations, the possible interaction effect between the CVD risk factors and age, gender or existing CVD status was also examined by adding a multiplicative interaction term to the regression model. For the continuous variables in the CVD risk factors, the dose–response relationship between the variable and depression status was investigated by the three-knot restricted cubic splines.17 To examine whether the relationship between the cumulative risk index and the PHQ-9 mean score was non-linear, a quadratic term (squared cumulative risk index) was added in a one-way analysis of variance (ANOVA) that examined the statistical differences of the PHQ-9 mean scores among the groups with a specific number of risk factors. SAS V.9.4 (SAS Institute) was applied for the analyses. Two-tailed tests were used to evaluate statistical significance.

Main findings

The incidence and costs of depression continue to increase globally, highlighting the need to understand the involved risk factors better. CVD and depression are highly comorbid, suggesting that CVD risk factors may independently predict depression. Indeed, the vascular depression hypothesis indicates that late-life depression is caused by abnormalities in cerebral vessels that are partly attributed to CVD risk factors.18 Despite the inter-relationship between these two diseases and the potential role of CVD risk factors in the pathogenesis of depression, no study has systematically and comprehensively investigated the individual and cumulative effect of CVD risk factors on depression. Thus, the current work investigated the possible association between each/cumulative index of 18 CVD risk factors and depression in a nationally representative population.

We found that various CVD risk factors were independently associated with depression after a multivariable adjustment. For example, smoking, living alone, poor self-rated health, poor sleep quality, short sleep time, high WBC level, higher BMI, having sedentary behaviour, hypertension and cancer were associated with increased risk of depression after confounder adjustment. The association between several lifestyle behaviours and depression has been identified. For example, adherence to aerobic exercise consistent with public health recommendations is an effective method for treating mild to moderate depression19; likewise, our results indicated that sedentary behaviour was significantly associated with depression. Smoking has been reported as a risk factor most closely related to depression, and we consistently observed that smoking was positively associated with depression. Although depression is considered a potential predictor for smoking, a systematic review has also reported a bidirectional relationship between tobacco use and depression.20 Therefore, further studies are needed to confirm the direction of this relationship. For living status, our results showed that participants living alone had an elevated risk of depression, presumably because living with others can enhance social connections and thus help prevent or alleviate the symptoms of depression.21 Similarly, participants with poor self-rated health had a significantly higher risk of depression, agreeing with a previous study reporting self-perceived health as an important risk factor for predicting depressive symptoms among American Indians and Alaskan Natives.21 For sleep-related variables, a previous meta-analysis indicated that sleep disorder is a statistically significant risk factor for depression among older adults in the community.22 Our data also revealed a significant association between depression and poor sleep quality or short sleep duration in a nationally representative population. Lifestyle behaviours aside, several studies have investigated the association between biomarkers and depression. For instance, a systematic review of observational studies has shown mixed evidence about the association between LDL cholesterol and depression.23 In addition, higher TG levels were found in people with depression.24 In our study population, we found TG, LDL, HDL and uric acid were all associated with depression in crude logistic regression, but no association could be identified after confounder adjustments. In addition, inflammation, a common risk factor for both CVD and depression, has been observed in individuals with depressive symptoms as indicated by elevated levels of circulatory proinflammatory cytokines.25 Moreover, longitudinal studies have shown that the inflammatory markers of higher C reactive protein or interleukin 6 concentrations in children or adults are related to the elevated risk of depression during subsequent follow-up.5 Inflammation leads to changes in mood state and autonomic cardiovascular regulation, and the underlying mechanisms may involve increased serotonin turnover, oxidative stress and hypothalamic–pituitary–adrenal axis activation.26 In our analysis, logistic regression revealed that clinically higher levels of WBC, but not NLR, were associated with depression. The association between adiposity and depression is still controversial. Results from a meta-analysis suggest that depression is positively correlated with obesity.27 However, other studies only weakly support the association between obesity and depression. For instance, one Mendelian randomisation study provided no evidence of causality in obesity as a risk factor for depression.28 The results of the adiposity indicators in our overall study population supported a significantly positive association between depression and higher BMI but not WC. Moreover, the association between physical health status and depression is inconsistent in the literature. We investigated the possible association between a history of specific chronic diseases and depression; the results showed that, as expected, CVD was associated with depression. Moreover, an increased risk of depression was observed in patients with cancer, possibly because a cancer diagnosis can induce feelings of anxiety and depression.29

Extensive evidence confirms that cumulative risk predicts outcomes more accurately than any single risk factor alone.30 Therefore, the present study adopted the method of combining multiple risk factors as a cumulative risk index, constructed a cumulative risk index and tested the hypothesis that the incidence of depression increased with the cumulative risk exposure levels. For all the risk factors associated with CVD and those in the three domains, significantly higher depression risk was observed in participants with greater cumulative risk exposure.

Gender difference in disease onset is one of psychiatric research’s most widely supported observations.31 According to a recent meta-analysis of nationally representative samples, the risk of depression in women is about twice that of men,32 which is consistent with our result that females had an increased prevalence of depression compared with males. Moreover, we found that the association between CVD risk factors and depression varied with gender, based on the interaction effect of CVD risk factors and gender. Our analyses revealed that higher BMI was associated with depression in females but not males, which agreed with the findings of a stronger association between depression and obesity in women than men.33 Interestingly, the association between drinking and depression was positive in women, but negative in men, leading to its null association in the overall study population. A common limitation of studies investigating risk factors of depression is that these variables usually are studied in isolation. However, risk factors do not exist independently but usually aggregate in one person. Based on the significant interaction effect between gender and the number of CVD risk factors on depression (p_interaction<0.05), we performed cumulative risk assessment in sex-stratified subpopulations, and the results indicated that females had a similar depression risk compared with males at a low level of cumulative risk exposure (OR_adjusted=1.32; 95% CI: 0.87 to 1.99). However, a significantly higher depression risk was found for females at a high level of cumulative risk index (OR_adjusted=2.86; 95% CI: 1.79 to 4.59). Consistent results of the cumulative risk analysis were obtained if only including 10 of the CVD risk factors (sedentary status, smoking, living alone, poor self-rated health, poor sleep quality, sleep duration, WBC, BMI, hypertension and cancer) associated with depression (online supplemental table 6). It is noted that the prevalence of CVD risk factors may vary between males and females (online supplemental figure 4), and they may be prone to have different risk factors at all levels of cumulative risk exposure. However, the number of risk factors has been known to be more indicative of depression risk than the type or nature of risk factors in cumulative risk analysis.34 Differential coping styles between males and females may serve as an explanation of high cumulative risk-mediated gender disparities in depression prevalence.35

Individuals are exposed to different risk factors over their lifespan. We found significant interaction effects between age and living status, BMI or diabetes status, suggesting the associations of these factors with depression were dependent on age. Interestingly, BMI had an interaction effect with both gender and age, which may explain the inconsistency mentioned above about the association between BMI and depression in different study populations in the literature. Our analysis of gender and age-stratified population showed that BMI was generally associated with depression in women of any age range but only in men who were 40–59 years old. It has been known that people with cardiac metabolic diseases are more likely to develop depression compared with healthy individuals.36 Our data indicated that smoking and poor sleep quality had a stronger association with depression in the population without CVD. In contrast, a positive association between WC and depression could only be identified in participants with CVD. In cumulative risk analyses, there was no interaction effect between the number of CVD risk factors and age/CVD status on depression, suggesting the association between cumulative risk exposure and depression was not dependent on age or CVD status.

Limitations

The current study has several limitations. First, causality or the direction of the association between CVD risk factors and depression could not be determined because of the study’s cross-sectional design. As a result, the reverse causality of the observed associations could not be excluded. For example, depression may hinder individuals from living with others, leading to the association between living status and depression. Second, information about the chronic disease status of participants was self-reported. Third, while we have included several important confounding variables in our study, residual confounding factors may remain. Fourth, this study lacked a formal depression diagnosis.

Implications

This is the first study to investigate the individual and cumulative effects of as many as 18 CVD risk factors on depression in a nationally representative population. It illustrated gender differences in the association of cumulative risk exposure with depressive symptoms.