In the present study, a series of a priori
-defined, theoretical factor solution of HADS were used to examine their psychometric conformity and psychophysiological relationship with CVC in elderly males residing in the community. Our results indicated that the flat three-factor model of HADS performed better in fitting the data but provided only a limited advantage over the hierarchical three-factor model and the original two-factor model. However, factors derived from the tripartite model offered an opportunity to scrutinize the counterbalancing effects of anhedonic depression and autonomic anxiety on CVC. In the hierarchical tripartite model, with lower negative affectivity, anhedonic depression and autonomic anxiety exerted counterbalancing effects on CVC along with corresponding indirect effects conferred by negative affectivity on CVC. In contrast, higher negative affectivity alone may reflect the overall deleterious effect of co-occurring anxiety and depression on CVC. Negative affectivity may be the unifying and potentially deleterious element linking individual trait negative emotions to impaired CVC [58
]. The present study further illustrated how negative affectivity moderated the relationship between negative emotions and CVC.
Theory-derived factor solutions have been previously utilized to investigate complex relationships between physiological indicators and co-occurring depression and anxiety. For example, the three hypothesized symptom groups in the tripartite model reflect highly distinctive patterns of brain activity [59
]. CVC is another commonly applied psychophysiological indicator. The CVC has been of significant value in illuminating the process of basic dimensions of psychopathology and in predicting adverse health outcomes [62
]. To the best of our knowledge, this is the first study that applies a series of a priori
-defined, theory-based factor solutions to examine the relationship between CVC and co-occurring affective symptoms. Interestingly, our results suggest that clusters of affective symptoms derived from tripartite models as well as brain activity reflect distinctive patterns of CVC.
In terms of CVC, the results here seem to offer an advantage in utilizing the tri-dimensional scoring of HADS. This helps to disentangle the complex relationship between psychopathology and physiologic markers when applying the flat tripartite model in studies that explore the associations among CVC and co-occurring depression and anxiety. For example, by extracting symptoms of autonomic anxiety from the original anxiety subscale, we demonstrated that the absence of a correlation between the original HADS anxiety subscale and CVC resulted from the heterogeneous relationships between anxiety symptoms and CVC. From the perspective of the hierarchical arrangement of the three factors of the tripartite model, additional information is acquired as primacy is conferred on the factor of negative affectivity. In the present study, negative affectivity not only indirectly affected CVC via autonomic anxiety but also moderated the direct effect of autonomic affectivity on CVC. In other words, the higher order construct of affective symptoms may possess a determinant role in the way that lower order affective constructs affect CVC and eventually lead to adverse health outcomes.
According to the conceptual organization of the tripartite model, negative affectivity explains the covariation between depression and anxiety; however, depression and anxiety have distinct features not shared with each other or with negative affectivity [26
]. With respect to cardiac autonomic control, the distinctive relationships between each tripartite factor of HADS and CVC, as noted in the present study, support this presumption. Paradoxically, the positive association between autonomic anxiety and CVC in the preset study subverts the typical understanding that elevated anxiety accompanies sympathetic activation and parasympathetic withdrawal. Several explanations help support this interesting finding.
First, the issues of construct and linguistic validity of autonomic anxiety in HADS may lead to the misallocation of symptom items and cause a biased association between autonomic anxiety and CVC. From the viewpoint of construct validity, Caci et al. (2003) argued that items assigned by Dunbar et al. (2000) were not suitable markers for the constructs of negative affectivity and autonomic anxiety. Furthermore, they were regarded as measuring the same thing [63
]. However, in both the Chinese and non-Chinese literature examining the psychometric properties of HADS, the three items selected as markers of autonomic anxiety by Dunbar et al. (2000) nearly matched some forms of anxiety constructs, such as generalized anxiety, psychic anxiety, or panic [64
]. Therefore, issues of item misallocation and language translation are insufficient to explain the positive correlation between autonomic anxiety and CVC as noted in this study. Despite arguments about inadequate psychometrical validity, the present study suggests that autonomic anxiety and negative affectivity as defined by Dunbar et al. (2000) seem to be different things in psychophysiological terms.
Another possible explanation is that Ahern et al. (2001) observed that inactivation of the right hemisphere by intracarotid sodium amobarbital administration resulted in a significant decrease in HRV. They suggested that the right hemisphere had a greater role in regulating cardiac function, perhaps by modifying parasympathetic effects [65
]. Since autonomic anxiety is associated with the activation of the right parietal region [60
], it is reasonable to infer that the positive association between autonomic anxiety and right hemisphere activation may contribute to increased CVC.
In the analysis of conditional indirect effect, we found that the positive contribution to CVC from autonomic anxiety was present only when scores of negative affectivity were low. The association tended to be reversed when negative affectivity scores were higher. Because negative affectivity in the tripartite model refers to “temperamental sensitivity to negative stimuli” and is related to neuroticism, low negative affectivity may represent less emotional arousal and a better coping style during stress. Constructive coping is related to better vagal tone [66
]. Hence, in subjects with low negative affectivity, autonomic anxiety may merely be an external marker for good psychological flexibility in response to stress, which is associated with high vagal tone [67
]. Several mild physical stresses have been reported to counteract the negative effects of aging and to increase longevity [68
]. Herrmann et al. (2000) also reported the protective effect reflected in the HADS anxiety subscale among subjects who received routine exercise testing [69
]. When subjects have high negative affectivity (a score of 4 for negative affectivity in this study), the combined effect on CVC from all factors of the tripartite model remains detrimental to health.
There are some limitations to the present study. First, although the flat tripartite model showed a better fit than did the hierarchical tripartite model and the original two-factor model, the difference was modest. Moreover, the goodness-of-fit indices of RMSEA and CFI suggested that none of these competitive models had a reasonably good fit. Empirically, the tripartite model of the European version of HADS provided a better and more acceptable fit to the data than did the two-factor model. In contrast, in the current study sample of elderly Chinese males, some elements may have affected the conformity of the tripartite model to the data. Structural differentiation of self-reported depression and anxiety is less easily performed in an older community sample than in younger populations [70
]. In a Chinese population with coronary heart disease, Wang et al. (2006) also found structural ambiguity in the Chinese version of HADS [39
]. They suggested that before any clear advantage of the tripartite models over the bi-dimensional models could be demonstrated, the amount of time consumed and the lack of a comparison to related literature rendered the tri-dimensional scoring approaches of HADS highly premature [39
]. However, from the standpoint of CVC, the advantages of the analytic strategy of the tri-dimensional approach of HADS override the shortcoming of structural ambiguity. Because the present study did not aim to offer factor structures that conformed to the data with the most optimal goodness-of-fit, we preserved the original model specifications proposed by Dunbar et al. (2000); therefore, no model revisions were performed. Potential sources of specification errors are number of factors, indicators, and error theory. Further studies with larger sample sizes and indicators for latent variables should help to examine the conformity of the tripartite model among elderly Chinese. However, the development of psychometric measurements is usually driven by philology. Perhaps the validation of paper-and-pencil tests with biological markers, as in the present study, maximizes the practical utility of psychometric tools and goes beyond philology.
Second, the study subjects came from an urban community. Their scores for depression and anxiety were too low to generalize the study finding to a general clinical population. Subjects with higher negative affectivity were so few that the statistical power to detect a conditional indirect effect was compromised. However, the total effect of negative affectivity on CVC remained statistically significant when it was 4 or more. As such, the effect was strong enough to be detected even in a small sample. Nonetheless, further research with study subjects that include clinical patients are still necessary in order to determine whether negative affectivity actually moderates the association of autonomic anxiety with CVC.
Third, gender differences are noted in the link between depression and cardiovascular disease. The association between poorer CVC and depressive symptoms has been selectively observed in elderly males only of Chinese ethnicity [19
]. However, other studies have demonstrated a female predominance in the link between depression and cardiovascular disease [5
]. With respect to gender differences, whether or not different pathophysiological processes exist in different ethnic groups is unclear.
Finally, the confounding effect of medications could not be totally controlled in the present study. Because most classes of antidepressant have robust suppressive effects on CVC [71
], subjects taking antidepressant were excluded in the present study. In addition to antidepressants, there are other medications with definite or probable effects on CVC, such as beta-blockers [76
] and antipsychotics [77
]. Unfortunately, we did not collect information on these medications, and future studies must assess these confounds.