| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH |
|
|
|
|||||||
|
|||||||||
TRACK to:
|
|
Electronic letters published:
![[Read Comment]](/icons/misc/sectionDOWN.gif){kind=icon}
Cardiovascular Disease, Psychology, Physiology: Opportunities for Collaboration toward Population Prevention?
Why the Different Effects Across Gender and Cardiovascular Outcomes?
Dichotomizing
|
|
|||
|
Edward P. Post, Ann Arbor, MI Assistant Professor, University of Michigan Medical School and Ann Arbor VA
Send response to journal:
|
Player and colleagues present a nicely written paper from an appropriate dataset, on the associations between trait anger and long-term stress and development of hypertension and overt CHD. Their sensitivity analysis demonstrates a reasonable degree of robustness across alternative ways of constructing study variables. Moreover, their central findings are largely concordant with the extant literature about psychosocial distress and development of hypertension and CHD. There is also a developing literature about the role of physiologic factors, in particular serotonin, on vascular physiology and platelet aggregation. Serotonin also figures prominently in depression, which can correlate with long-term stress. Additionally, hypertension is a prominent risk factor for development of CHD. Clearly, the associations are myriad and causal pathways complex and at best partially understood. The authors of this work make a compelling case for the risk associations they demonstrate, using data that provides a high-level and long-term view. This view would be relevant to insights for developing a prevention approach, tailored to development of conditions with significant morbidity and mortality that constitute a public health threat. However, these risks as immediately constructed are not constructs that clinicians routinely measure. Mental health symptoms and conditions (e.g., depressive symptoms and syndromes) may be a link to a more identifiable anchor in routine practice, and research literature in these areas tends to focus on short- to medium-term time intervals. Finally, studies on physiologic phenomena such as associations between serotonin and cardiovascular reactivity and pathogenesis have historically been separate from clinical research, often conducted over very short time frames and in animal models. Thus, the challenges to proceeding from the authors' careful work are daunting: to integrate several distinct ways of conducting research, with vastly different time horizons and intermediate goals. The overarching goal is nonetheless compelling, given the importance of cardiovascular disease on population health in the United States. In a new climate of medical research that stresses multidisciplinary collaboration and translational and public health targets, the time may be ripe to begin to integrate our best resources in neurochemistry, clinical psychiatry and personality studies, and preventive cardiology toward long-term studies that target such important health outcomes. Competing interests: None declared |
|||
|
|
|||
|
Laura D Kubzansky, Boston, USA Associate Professor, Harvard School of Public Health
Send response to journal:
|
Dr. Player and colleagues have described a thoughtful study of whether anger and long-term psychological stress are associated with progression from prehypertension to either hypertension or CHD. The findings were more consistent for men, with anger associated with increased risk of developing both hypertension and coronary heart disease. For women however, anger was not associated with increased risk of developing either outcome. The gender differences in these findings are somewhat striking and deserve greater attention. The authors briefly speculated that failure to find an association may have been due to having too few cases of hypertension among women. Given the number of studies looking at anger and different cardiovascular outcomes in the ARIC sample, our overall understanding of these associations would be enhanced by thinking about how various findings fit together. In fact, the current findings are consistent with previous work in ARIC, where anger does not seem to be strongly associated with increased risk of cardiovascular outcomes, such as arterial stiffness or atherosclerosis.1, 2 Moreover, in other samples, anger has been only weakly associated with risk of hypertension among women.3 It is unlikely that power alone is enough to explain the present findings. It would be interesting to know how levels of anger correspond across men and women, or to consider whether the experience of anger plays out differently for men and for women. It was also surprising to see that long-term psychological distress, which more closely resembles depression, was associated with developing CHD but not with hypertension. One explanation might be that mechanisms by which long -term psychological stress influence CHD may be different from those leading to hypertension. However, this seems unlikely given that different types of distress have been linked with development of hypertension in other samples.4 Alternatively, could the findings be attributed to the somewhat restricted sample used for this study, including only individuals in a pre-disease state? More generally, it appears that anger has a stronger association with cardiovascular risk among men than women, while other psychosocial factors have more similar effects. This may be particularly relevant for the debate on whether we should continue to look at specific emotions versus shared components of distress in relation to cardiovascular risk.5 Understanding whether apparent differences are consistent or if they might depend on aspects of the sample or types of measures used will enhance our ability to develop strategies for intervention and prevention. References 1. Williams JE, Couper DJ, Din-Dzietham R, Nieto FJ, Folsom AR. Race- gender differences in the association of trait anger with subclinical carotid artery atherosclerosis: the Atherosclerosis Risk in Communities Study. Am J Epidemiol 2007;165(11):1296-304. 2. Williams JE, Din-Dzietham R, Szklo M. Trait anger and arterial stiffness: results from the Atherosclerosis Risk in Communities (ARIC) study. Prev Cardiol 2006;9(1):14-20. 3. Yan LL, Liu K, Matthews KA, Daviglus ML, Ferguson TF, Kiefe CI. Psychosocial factors and risk of hypertension: The Coronary Artery Risk Development in Young Adults (CARDIA) Study. JAMA 2003;290(16):2138-48. 4. Jonas BS, Franks P, Ingram DD. Are symptoms of anxiety and depression risk factors for hypertension? Longitudinal evidence from the National Health and Nutrition Examination Survey I Epidemiologic Follow-Up Study. Archives of Family Medicine 1997;6:43-9. 5. Carney RM. Psychological risk factors for cardiac events: Could there be just one? Circulation 1998;97(2):128-9. Competing interests: None declared |
|||
|
|
|||
|
Amanda L Golbeck, Missoula, MT, USA Professor, School of Public and Community Health Sciences, The University of Montana
Send response to journal:
|
The authors use a number of dichotomized continuous variables in their analysis. These include an outcome variable (hypertension based on systolic and diastolic blood pressure), as well as demographic (age into lower versus higher), risk (smoking into current or not, obesity based on BMI, hyperlipidemia based on LDL, sufficient exercise based on total number of minutes per week), and psychosocial (social networks, long-term psychological stress, and trait anger into low/moderate versus high) variables. Dichotomizations can be more or less arbitrary. Among the less arbitrary in this study are ones based on professional guidelines, but even these are based on cut points that may be subject to change over time (hypertension(1), obesity, hyperlipidemia, and sufficient exercise) or may not apply to the whole population (e.g., hypertension for those without and with diabetes(1)). Dichotomizations can be based on ‘optimal’ cut points derived from the data, but these can lead to serious bias(2). While dichotomization has been common, particularly among clinical researchers, it has other disadvantages from a statistical point of view that are well described in the literature. In particular the dichotomization of continuous predictors in regression has been declared to be a bad idea, simplicity being achieved at a cost(2). The most noticeable costs are loss of power and residual confounding. The authors of the current article are to be commended for conducting a cut point sensitivity analysis. They entered covariate risk factors into their regression models as continuous variables, and then compared the results with their initial models where they entered the covariates as dichotomous variables. They report having found similar results. However, looking more closely at Supplemental Table 1 in comparison with Table 4, focusing on the association of trait anger with progression to CHD, the 99% confidence intervals are (1.19-3.32) versus (0.99-2.93), using continuous versus dichotomized continuous covariates, respectively. Although the numbers are similar, they lead to different conclusions regarding associations within a hypothesis testing context. The AHR [which appears to be mislabeled as AOR on Supplemental Table 1] confidence interval using dichotomized continuous covariates covers 1, whereas the corresponding CI using continuous covariates does not. Thus, the authors have provided an empirical example of cost in terms of statistical significance that can be paid by dichotomizing. 1 Wang, T.J. and Vasan, R.S. (2005), Epidemiology of Uncontrolled Hypertension in the United States. Circulation 112:1651-1662. 2 Royston, P., Altman, D.G., and Sauerbrei, W. (2006), Dichotomizing Continuous Predictors in Multiple Regression: A Bad Idea. Statistics in Medicine 25(1):127-141. Competing interests: None declared |
|||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH |
TRACK: