Dispositional Optimism and the Risk of Cardiovascular Death: The Zutphen Elderly Study

Background:  Dispositional optimism, defined in terms of life engagement and generalized positive outcome expectancies for one's future, may be related to lower cardiovascular mortality. We aimed to determine whether dispositional optimism is a stable trait over time and whether it is independently related to lower cardiovascular mortality in elderly men.

Methods:  In a cohort study with a follow-up of 15 years, we included 545 (61.4%) of 887 men, aged 64 to 84 years, who were free of preexisting cardiovascular disease and cancer and who had complete data on cardiovascular risk factors and sociodemographic characteristics. Dispositional optimism was assessed using a 4-item questionnaire in 1985, 1990, 1995, and 2000. In Cox proportional hazards models, the first 2 years of observation were excluded.

Results:  Optimism scores significantly decreased over 15 years, but showed temporal stability (reliability coefficients, 0.72 over 5 years and 0.78 over 15 years; P < .001). Optimists in 1985 had a hazard ratio for cardiovascular mortality of 0.45 (top tertile vs lowest tertile; 95% confidence interval, 0.29-0.68), adjusted for classic cardiovascular risk factors. The risk of cardiovascular death was inversely associated with increased tertiles of dispositional optimism (P < .001 for trend). Similar results were obtained using 1990 data after additional adjustment for depression (assessed by the Zung Self-rating Depression Scale).

Conclusion:  Dispositional optimism is a relatively stable trait over 15 years and shows a graded and inverse association with the risk of cardiovascular death.

Evidence from prospective cohort studies suggests that optimism is a predictor of well-being and physical health. It has been associated with better health outcomes in patients with ischemic heart disease,^1-4 and with a lower risk for all-cause mortality^5-7 and cardiovascular morbidity^8,9 and mortality.^7,8 Previous studies, however, did not adjust for all potential confounding variables and for depression. Optimism has been conceptualized in 2 rather dissimilar ways, as explanatory style optimism⁵ and as dispositional optimism.^1,2,10,11 We focused on dispositional optimism, which is defined in terms of generalized positive expectancies for one's future, and studied its association with cardiovascular mortality in apparently healthy elderly men.

Because optimism may affect mortality indirectly (eg, through the promotion of behavior and lifestyles promoting better health), we adjusted for various classic cardiovascular risk factors and sociodemographic characteristics. Also, depression may be considered a confounder or an intermediary factor, because a low level of dispositional optimism and a high level of depressive symptoms are partly interdependent,^10,12 and depression and depressive symptoms are predictive of cardiovascular mortality in previously healthy subjects and in patients.^13-17 Finally, an optimistic disposition is nonspecific and might be induced by a wide range of life events and mood swings. Because within-person variation and random error tend to bias hazard ratios toward the reference value of 1, long-term variation could result in a considerable underestimation of the true association between optimism and mortality. We, therefore, studied the temporal stability of dispositional optimism over 5- to 15-year periods.

The cohort of the Zutphen Elderly Study consists of men born between January 29, 1900, and August 30, 1920. The Zutphen Elderly Study started in 1960 as the Dutch contribution to the Seven Countries Study on lifestyle, biological risk factors, and cardiovascular diseases in middle-aged men. In 1985, 367 of 555 men who were still alive were reexamined. In addition, a new random sample of 711 men of the same age also living in Zutphen, the Netherlands, but not belonging to the original cohort was invited to participate. The invitation resulted in a total target population of 1266 men aged 64 to 84 years, of whom 887 (response rate, 70.1%) participated in the study in 1985; 560 surviving men were reexamined (response rate, 78.0%) in 1990, 343 (response rate, 74.1%) in 1995, and 171 (response rate, 72.8%) in 2000. The study was approved by the Medical Ethics Committee of the University of Leiden, Leiden. Informed consent was obtained from all participants.

Information on the vital status of the participants until July 1, 2000, was obtained from municipal population registries. Causes of death were obtained from the Central Bureau of Statistics until July 1, 1990. From March 5, 1985, to July 1, 2000, information was also obtained from the participants' general practitioners. Information was verified with either hospital discharge data or information from the Netherlands Cancer Registry. The initial coding of the causes of death was done by 3 physicians and the final coding by an experienced clinical epidemiologist. Because it is often difficult to determine the underlying cause of death in elderly people, the primary and secondary causes of death were included in the analyses. Cardiovascular disease was defined as codes 390 to 459 according to the International Classification of Diseases, Ninth Revision (ICD-9).

The medical history was obtained using a standardized questionnaire based on the cardiovascular survey questionnaire developed by Rose and Blackburn.¹⁸ Data on disease prevalence at baseline were verified from hospital discharge data and written information from the general practitioner, and were subsequently uniformly coded. Nonresponders at follow-up examinations received a short questionnaire on disease history. We excluded men previously diagnosed as having myocardial infarction (110 men [12.4%]), angina pectoris (90 men [10.1%]), stroke (34 men [3.8%]), transient ischemic attack (18 men [2.0%]), heart failure (37 men [4.2%]), and malignant neoplasm (codes 140-172 and 174-208; 63 men [7.1%]). Diabetes mellitus was included as a covariate. This left 545 (61.4%) of 887 men in 1985 with complete information on dispositional optimism and potential confounders. (The numbers do not add up to 342, because several men had more than 1 diagnosis at baseline.) One man was lost to follow-up and was included in the analyses, but censored after 5.1 years of follow-up.

In 1985, 1990, 1995, and 2000, dispositional optimism was assessed using a questionnaire consisting of the following 4 items: “I still expect much from life,” “I do not look forward to what lies ahead for me in the years to come,” “My days seem to be passing by slowly,” and “I am still full of plans” (our translations). The Central Bureau of Statistics had previously collected data on these 4 items in life situation surveys in the Netherlands in 1976 and 1982.¹⁹ The response format was a 3-point scale of frequency: “fully in agreement” (score, 0), “partially in agreement” (score, 1), and “not in agreement” (score, 2). The additional answer category “do not know” was also coded as the midpoint (score, 1). The scores of 2 questions needed to be reversed, so that higher scores indicated greater optimism. A mean score (range, 0-2) was calculated for each time point. Internal consistency was moderate (Cronbach α, .62 in 1985, .65 in 1990, .71 in 1995, and .63 in 2000).

Surveys were conducted between March 1 and June 30 in 1985, 1990, 1995, and 2000. Information on self-rated health, living arrangement, education, family history of stroke or myocardial infarction, and physical activity was obtained with a questionnaire that participants were asked to complete at home. Self-rated health was coded along 3 levels (feeling healthy, feeling rather healthy, or feeling moderately or not healthy). With the exception of 1985, depressive symptoms were assessed using the Zung Self-rating Depression Scale,²⁰ consisting of 20 items on a 4-point scale. Living arrangement (living with a wife or partner, other living arrangement, or living alone) and education (lower vocational or general education, middle or higher vocational or general education, or college or university) were coded along 3 levels. A validated questionnaire on physical activity designed for retired men was used to calculate the total minutes spent in physical activity per week (of an intensity of >2 kcal/kg per hour).²¹ Cigarette smoking (never, former, or current smoker) and alcohol use (0, 1-29, or ≥30 g/d of alcohol) were coded along 3 levels.

Medical examinations were performed by 5 trained physicians, and included body mass index (BMI), blood pressure measurements (in duplicate), and medical history. The mean arterial pressure was calculated as follows: [systolic pressure + (diastolic pressure × 2)]/3. The use of antihypertensive medication was also assessed. Nonfasting total cholesterol and high-density lipoprotein (HDL) cholesterol levels were analyzed enzymatically in a standardized laboratory.

Physical activity, BMI, mean arterial pressure, and total and HDL cholesterol levels were continuous; all other variables were categorical. Because the physical activity score was skewed, it was logarithmically transformed and the geometric mean (10th and 90th percentile) is given. All other data are given as number (percentage) or mean ± SD. The baseline characteristics of the participants were compared between tertiles of dispositional optimism by using the χ² test and the 1-way analysis of variance, as appropriate. A Friedman test for related samples was used to examine changes in dispositional optimism over time. Reliability coefficients (ie, an intraclass correlation coefficient with a 1-way random effects model with single-measure reliability) were used to examine temporal stability.

The Kaplan-Meier method was used to present crude cardiovascular mortality. The proportional hazard assumption was satisfied using the log minus log graphical method (for which continuous variables were provisionally categorized). Hazard ratios with 95% confidence intervals of cardiovascular mortality were estimated by Cox proportional hazards models. Because causal inference is limited by potential reverse-causation bias, lag time analysis excluded the first 2 years of observation from all survival analyses. Three multivariate models were tested in 1985. Model 1 adjusted for age; 2, for age, BMI, diabetes mellitus, smoking status, mean arterial pressure, antihypertensive medication, and total and HDL cholesterol levels; and 3, for all the variables given for model 2 and physical activity, alcohol use, family history of stroke or myocardial infarction, self-rated health, living arrangement, and education. In 1990, we adjusted for all time-dependent confounders previously mentioned in addition to the Zung Self-rating Depression Scale score. We tested for linear trend across tertiles of dispositional optimism. Two-tailed P < .05 was considered statistically significant. A commercially available software program was used (SPSS 10.0; SPSS Inc, Chicago, Ill).

Table 1 presents the baseline characteristics. Higher optimism scores were associated with younger age, a lower Zung Self-rating Depression Scale score in 1990, better self-rated health, a higher physical activity score, less often living alone, and a higher level of education. Optimism was also positively associated with total cholesterol level.

The mean dispositional optimism scores decreased from 1.50 in 1985 to 1.27 in 2000. When optimism scores were compared across 4 assessments, a significant (P < .001) decrease in dispositional optimism scores over 15 years was observed (n = 115 with complete data).

Among participants who provided paired dispositional optimism scores, the reliability coefficient for dispositional optimism was 0.72 between 1985 and 1990 (n = 358 pairs), 0.69 between 1990 and 1995 (n = 223 pairs), and 0.70 between 1995 and 2000 (n = 122 pairs) (P < .001 for all). The overall reliability coefficient of 0.78 (over 4 time points) also indicated long-term consistency of considerable strength (P < .001; n = 115 with complete data). The 5-year reliability coefficients were 0.69 for physical activity, 0.88 for total cholesterol level, 0.88 for HDL cholesterol level, 0.77 for mean arterial pressure, and 0.92 for BMI between 1985 and 1990 (n = 383 pairs). For the Zung depression scores, the reliability coefficients were 0.63 between 1990 and 1995 (n = 218 pairs) and 0.77 between 1995 and 2000 (n = 116 pairs).

During the 15-year follow-up, 373 (68.4%) of 545 men died, of whom 187 (50.1%) died of cardiovascular causes. Kaplan-Meier analysis of survival according to the tertile of optimism score showed that men with a low level of dispositional optimism had higher cardiovascular mortality rates compared with men with a high level of dispositional optimism in 1985 (Figure).

Compared with men in the lowest tertile of dispositional optimism, those in the top tertile had a 55% lower multivariate-adjusted hazard ratio of cardiovascular mortality (Table 2), which was attenuated after adjustment for self-rated health, physical activity, and other covariates.

To test whether the lower risk of cardiovascular death could be explained by depressive symptoms, we additionally adjusted a multivariate model for the Zung depression score using the 1990 data. This resulted in an inverse association with dispositional optimism, with a multivariate-adjusted hazard ratio with a time-dependent variable of 0.49 (Table 3). Next, 77 (24.6%) of 313 men were excluded from the analyses because of Zung index scores of 50 or higher, indicating mild to severe depression. This resulted in multivariate- and fully adjusted hazard ratios (without adjustment for the Zung Self-rating Depression Scale score) of cardiovascular mortality of 0.40 (95% confidence interval, 0.20-0.77; P = .007 for trend) and 0.50 (95% confidence interval, 0.24-1.02; P = .06 for trend), respectively.

Our results demonstrate a strong and consistent association between dispositional optimism and an about 50% lower risk of cardiovascular mortality in elderly men during 15 years of follow-up. The association was attenuated after adjustment for cardiovascular risk factors and depressive symptoms. In a recent study,²² elderly subjects with a “positive life orientation” had a higher survival rate, also independent of depressive symptoms. We found relationships to be graded, which is consistent with previous findings in a different group of elderly men and women.⁷ Moreover, hopelessness, to some extent the reverse of optimism, has been associated with an increased risk for cardiovascular disease^23-25 and the progression of atherosclerosis.²⁶

The reliability coefficient—recorded in paired samples over 5-year intervals—of about 0.7 for dispositional optimism is reasonably good. The test-retest correlation coefficients for dispositional optimism assessed by the Life Orientation Test¹⁰ were also 0.7 in groups of undergraduates,¹² middle-aged women,²⁷ and elderly men²⁸ over up to 3-year intervals. Moreover, dispositional optimism was relatively unaffected by receiving either good or bad news after breast cancer surgery in women.²⁹ These and our results indicate that dispositional optimism changes only moderately over time, although steadily decreasing. They support the idea that dispositional optimism is a relatively constant trait aspect of a given individual's personality, and is sufficiently stable for long-term prediction of cardiovascular mortality in the present and previous studies.^7,8

On a mechanistic level, the intriguing question is how low optimism may lead to cardiovascular death. Optimism was not associated with cardiovascular risk factors such as BMI, hypertension, diabetes mellitus, and HDL cholesterol levels, although it was positively associated with total cholesterol levels. Also, several sociodemographic characteristics were significantly associated with optimism, and the inverse association was attenuated when adjusting for self-rated health and physical activity, which are potential causal intermediates. Another possible mechanistic explanation is that optimism is related to better coping behavior, goal-directed efforts (such as better self-care), vitality, and emotional flexibility.³⁰ Optimism correlated positively with persistent problem-focused coping and seeking social support,^10,11 with better treatment adherence,^3,11 and with increasing exercise in a cardiac rehabilitation program.³¹ Alternatively, the increased cardiovascular mortality rate may be mediated by biological consequences or predecessors of low optimism that were not measured, such as the effects of genetic factors, the hypothalamic-pituitary-adrenal axis, the autonomic nervous system, inflammation, endorphins, cardiac arrhythmia, or platelet activity.

Potential limitations of our study merit consideration. We tested only a single 4-item scale, which does not necessarily represent dispositional optimism as measured with more recent scales, such as the Life Orientation Test¹⁰ or the Scale of Subjective Well-being for Older Persons,⁷ which were created and validated after our study was initiated. The questionnaire was kept similar in the different surveys to avoid introducing method variation bias. Although all 3 questionnaires share a focus on the future, our 4-item scale may have a tendency toward reflecting life engagement, vitality, and feeling a purpose in life. The Zutphen Elderly Study cohort is composed of healthy, white, Dutch men, which may limit the generalizability of our findings to other ethnic groups and to women; indeed, the effect of optimism on overall mortality was recently reported to be smaller in women.⁷ It remains possible that the overall association with mortality can partly be ascribed to residual confounding. Although the prospective study design and graded associations suggested a causal relationship between optimism and cardiovascular mortality, the possibility of reverse causation deserves attention. First, bias due to preexisting disease and illness-related lowering of optimism was reduced, because subjects with prevalent cardiovascular disease and cancer were excluded from the analyses. Second, we restricted the outcome to subjects who died only after a lag period of 2 years. This study has several strengths. We had an almost complete mortality follow-up and included sufficient men to estimate cardiovascular risk reliably. The validity of our measurements is supported by the finding of the expected baseline associations of dispositional optimism with sociodemographic characteristics, the graded inverse association with cardiovascular mortality, the robust internal consistency, and the reliability of dispositional optimism.

Based on the present and previous findings,^1-9 a low subjectively perceived level of optimism should be added to the list of independent risk markers for cardiovascular mortality in elderly men. Optimism can be estimated easily and is stable over long periods. It is yet to be established whether interventions aimed at improving an older individual's level of optimism may reduce the risk of cardiovascular mortality.

Correspondence: Erik J. Giltay, PhD, MD, GGZ Delfland, Institute of Mental Health, PO Box 5016, 2600 GA Delft, the Netherlands (giltay@dds.nl).

Accepted for Publication: July 8, 2005.

Author Contributions: Dr Giltay had full access to the data and takes responsibility for the integrity and accuracy of the data analyses.

Financial Disclosure: None.

Funding/Support: The Zutphen Elderly Study was supported by the Netherlands Organisation for Health Research and Development, The Hague.

Role of the Sponsor: The funding body had no role in data extraction and analyses, in the writing of the manuscript, or in the decision to submit the manuscript for publication.