Age-related changes in risk factor effects on the incidence of
thromboembolic and hemorrhagic stroke Robert D. Abbott a,b,c,e, * , J. David Curb b,c,e , Beatriz L. Rodriguez b,c,e , Kamal H. Masaki b,c,e , Jordan S. Popper b,e , G. Webster Ross b,c,d,e , Helen Petrovitch b,c,d,e a Division of Biostatistics and Epidemiology, University of Virginia School of Medicine, Charlottesville, Virginia 22908, USA b Pacic Health Research Institute, 846 S. Hotel Street, Suite 303, Honolulu, Hawaii 96813, USA c Department of Medicine, John A. Burns School of Medicine, University of Hawaii, 1356 Lusitana Street, 7th Floor, Honolulu, Hawaii 96813, USA d Department of Veterans Affairs, PO Box 50188, Honolulu, Hawaii 96850, USA e The Honolulu Heart Program and the HonoluluAsia Aging Study, Kuakini Medical Center, 347 N. Kuakini Street, Honolulu, Hawaii 96817, USA Received 27 November 2001; received in revised form 9 May 2002; accepted 18 November 2002 Abstract We examined the changes in risk factor effects on the incidence of thromboembolic and hemorrhagic stroke as they may occur with age. Findings were based on repeated risk factor measurements at four examinations over a 26-year period in 7589 men in the Honolulu Heart Program. After each examination, 6 years of follow-up were available to assess risk factor effects on the incidence of stroke over a broad range of ages (4593 years). As compared with normotensive men, the risk of thromboembolic stroke in the presence of hypertension declined from a 7-fold excess in men aged 45 to 54 years to a 1.4-fold excess in men aged 75 (P .001). Adverse effects of diabetes and atrial brillation seemed to be equally important across all ages, whereas a protective effect of physical activity increased with age. Except for men with atrial brillation, the incidence of thromboembolic stroke increased signicantly with age regardless of risk factor status, including men with normal blood pressure (P .001). Although hemorrhagic events were less common, positive relations with cigarette smoking seemed to strengthen with age, whereas those with hypertension tended to decline. Our ndings suggest that strategies for the prevention of stroke may need to account for changes in risk factor effects as they occur with age. Control of diabetes and the encouragement of active lifestyles in the elderly seem to be especially important.
2003 Elsevier Inc. All rights reserved.
Keywords: Stroke; Thromboembolic; Hemorrhagic; Risk factors; Aging 1. Introduction Rapid increases in the risk of stroke and changes in the prevalence and distribution of important risk factors are known to occur with advancing age [111]. Whether these age-related effects alter the association between a risk factor and stroke is not clear. It may be that the association between stroke and some risk factors remains constant re- gardless of age, whereas the importance of other risk factors declines or increases from middle adulthood to late life. In addition, if stroke risk factors become sufciently pervasive in the elderly population, the identication of independent effects on the risk of stroke could become difcult. The purpose of this report is to examine the way in which risk factor effects on the incidence of thromboembolic and * Corresponding author. University of Virginia Health System, Depart- ment of Health Evaluation Sciences, P.O. Box 800717, Charlottesville, VA 22908-0717. Tel.: 434-924-1687; fax: 434-924-8437. E-mail address: rda3e@virginia.edu (R.D. Abbott). 0895-4356/03/$ see front matter
2003 Elsevier Inc. All rights reserved.
doi: 10.1016/S0895-4356(02)00611-X hemorrhagic stroke can change over a broad range of ages from middle adulthood to late life in men enrolled in the Honolulu Heart Program. 2. Methods 2.1. Study background From 1965 to 1968, the Honolulu Heart Program began following 8006 men of Japanese ancestry living on the island of Oahu, Hawaii for the development of coronary heart disease and stroke [12,13]. At the time of study enroll- ment, participants received a complete physical examination when they were aged 45 to 68 years. Procedures were in accordance with institutional guidelines and approved by an institutional review committee. Informed consent was obtained from the study participants. Information on cardiovascular events that occurred after the baseline examination was obtained through a comprehen- sive system of surveillance of hospital discharges, death R.D. Abbott et al. / Journal of Clinical Epidemiology 56 (2003) 479486 480 certicates, autopsy records, and at repeat examinations given in the course of follow-up. For this report, subjects were followed for the rst occurrence of a thromboembolic or hemorrhagic stroke. 2.2. Diagnosis of stroke A diagnosis of stroke was made when a neurologic decit was accompanied by blood in the cerebrospinal uid or by evidence of an infarct or hemorrhage. Subjects were fol- lowed for the rst occurrence of a fatal or nonfatal thrombo- embolic or hemorrhagic stroke. Possible strokes (neurologic decits lasting at least 24 hours but less than 2 weeks or of unknown duration) were not included among the stroke events because of a lack of diagnostic certainty. Subjects who experienced focal neurologic episodes attributed to other conditions (e.g., blood dyscrasias, neoplastic disease, head injury, surgical accident, meningoencephalitis, fat embolism, epilepsy, or cardiac arrest) were not included among the victims of stroke. A stroke was considered thromboembolic if the focal neurologic decit occurred usually without prolonged un- consciousness, nuchal rigidity, fever, pronounced leukocyto- sis, or blood in the spinal uid. Identication of hemorrhagic stroke was based on a focal neurologic decit accompanied by loss of consciousness, headache, and blood present in the spinal uid obtained by an atraumatic lumbar puncture or on the basis of computerized tomography or surgical ndings. All diagnoses were reviewed and conrmed by a study neurologist and the Honolulu Heart Program Morbid- ity and Mortality Review Committee. Only men who were free from coronary heart disease and stroke at the time of study enrollment were considered for follow-up for a future stroke event. Prevalent coronary heart disease was dened to include angina pectoris, coro- nary insufciency, and unequivocal ndings of a myocardial infarction. The nal sample size included 7589 men. Further details on the diagnosis of coronary heart disease and stroke are provided in earlier publications [1315]. 2.3. Collection of risk factor information Risk factor information included hypertensive status, total cholesterol level, diabetes, body mass index (BMI) (kg/ m 2 ), cigarette smoking status, alcohol intake (oz/mo), and physical activity. Risk factor measurements were made at the time of study enrollment (19651968) and updated at physical examinations that occurred at 6 (19711974), 15 (19801982), and 26 (19911993) years into follow-up. Approximately 90% and 80% of the surviving members of the original Honolulu Heart Program cohort participated in the 6th and 26th year anniversary examinations, respectively. The examination that occurred 15 years into follow-up in- cluded a randomsample of men who were selected to partici- pate in the Cooperative Lipoprotein Phenotyping Study [16]. Among the risk factors considered in this report, hyper- tension was dened as systolic or diastolic blood pressures 160 and 95 mm Hg, respectively, or based on the use of antihypertensive medication. To be considered normoten- sive, systolic and diastolic blood pressures needed to be 140 and 90 mm Hg, respectively. Men who were neither normotensive nor hypertensive were classied as having borderline hypertension. Study participants were also classi- ed as having diabetes on the basis of a medical history (physician-diagnosed or based on the reported use of insulin or the receipt of oral hypoglycemic therapy). Assessment of overall metabolic output during a typical 24-hour period used the physical activity index, which is a weighted average of the number of hours spent per day in ve different activity levels (basal, sedentary, slight, moder- ate, and heavy). In addition to the Honolulu Heart Program [17,18], the same index has been usedin the Framingham[19] and Puerto Rico [20] heart studies and is known to be in- versely related to the risk of coronary heart disease and stroke. The physical activity index was derived at all exami- nations except the examination that occurred 6 years into follow-up (19711974). Removing physical activity in this report does not change the associations that were observed between the other factors and the risk of stroke. High-density lipoprotein cholesterol, serum glucose, triglycerides, and other risk factors were not examined because they were measured too infrequently during the course of follow-up to enable an assessment of their effects across a broad range of ages. Further description of the risk factors is provided elsewhere [2123]. 2.4. Statistical methods To assess the effects of a risk factor on the incidence of stroke, proportional hazards regression models were used [24]. Although such models allow for the adjustment of other factors, they also allow for the effect of a risk factor to vary with time as risk factors become updated with age. Risk factors were also compared across 10-year age ranges based on standard analysis of covariance methods [25]. All reported P values were based on two-sided tests of sig- nicance. In general, the examination cycles used in this report were undertaken at least 6 years apart. As a result, four 6-year, non-overlapping and conditionally independent [26,27] follow-up periods could be created after each risk factor update. No subjects were followed for longer than 6 years, nor was there overlap with subsequent periods of follow-up. Follow-up for the last period (which began from 19911993) was available to 1998. For this latter period, follow-up was available for nearly 6 years. Curtailing follow- up to shorter 5-year periods had only negligible effects on the reported ndings. Follow-up was further restricted to men who were free of knowncoronaryheart diseaseandstroke at the beginningof each period. With each risk factor modeled as a time-varying covariate, subjects could contribute up to four 6-year person intervals. The intervals of follow-up were pooled to enable R.D. Abbott et al. / Journal of Clinical Epidemiology 56 (2003) 479486 481 the calculationof 6-year incidence rates of stroke accordingto an updated age. This allowed for the calculation of incidence rates across a broad range of ages (4593 years) as age was updated from the time of study enrollment (19651968) to the examination that occurred 26 years later (19911993). Such pooling permitted an assessment of risk factor effects on stroke as they might change with increasing age. Here, cross-product terms between a risk factor and age as a contin- uous variable were modeled. Tests for interaction between a risk factor and age on the risk of stroke as both variables were updated with time were then based on the estimated regression coefcients and standard errors that were associ- ated with these terms. 3. Results Among the 7589 men free of stroke and coronary heart disease at the time of study enrollment, there were 18,070 6-year person intervals of follow-up. The overall average age at the beginning of the follow-up intervals was 61 10 years (range 4593 years). Across the combined person intervals, 396 strokes were observed (21.9/1000 in 18,070 intervals at risk of disease). The average time to a stroke event was 3.0 1.7 years (range 0.16 years). Among the strokes, 276 were thromboembolic (15.3/1000), 96 were hemorrhagic (5.3/1000), and the remaining 24 were of un- known origin. Table 1 further describes the 6-year incidence of each stroke subtype and all strokes combined according to age. For each event, the 6-year incidence increased consistently with age (P .001). For all strokes combined, the incidence observed in men aged 45 to 54 years (9.0/1000) was doubled at ages 55to64years (17.8/1000) andmorethantripledat ages 65 to 74 (33.4/1000). For men aged 75 to 93 years, there was more thana 5-foldexcess of stroke (48.1/1000) relative to the youngest men. Similar magnitudes of increase occurred for thromboembolic and hemorrhagic events. Table 2 provides data on average risk factor values that were observed at the beginning of the 6-year intervals of follow-up, also according to age when such data were mea- sured. Signicant associations were observed between the Table 1 Six-year incidence (rate/1000) of thromboembolic and hemorrhagic stroke according to age Six-year incidence of stroke (rate/1000) a Six-year person intervals Thromboembolic Hemorrhagic All Age, y at risk stroke stroke strokes 4554 5430 6.6 (36) b 2.4 (13) 9.0 (49) 5564 6972 11.5 (80) 5.0 (35) 17.8 (124) 6574 3358 24.1 (81) 7.4 (25) 33.4 (112) 7593 2310 34.2 (79) 10.0 (23) 48.1 (111) Overall 18,070 15.3 (276) 5.3 (96) 21.9 (396) a Incidence of each stroke event increased signicantly with advancing age (P .001). b Number of events. risk factor levels and age. The strongest associations seemed to occur for men who were hypertensive when follow-up began. Of those men who were aged 45 to 54 years, 20.6% were hypertensive, whereas in the oldest group of men, the percentage was more than doubled (53.1%, P .001). Although atrial fibrillation was less common in this cohort of Japanese-American men, the percentage of men with atrial brillation increased consistently with age (P .001). Although the frequency of diabetes seemed to increase with age as well (P .001), the greatest increase occurred after age 54. Differences in alcohol consumption were modest in men 75 years of age and increased in those who were older. Risk factors that declined with age included total cholesterol, BMI, cigarette smoking, and the physical activ- ity index. The latter changes occurred consistently with advancing age (P .001). The crude 6-year incidence of thromboembolic stroke (rate/1000) is given in Table 3 within risk factor and selected age stratum. Tests of signicance were based on proportional hazards regression models. Except for men with atrial - brillation, stroke incidence increased signicantly with age within each risk factor stratum. Further adjustment for the other risk factors (including BMI) failed to alter these ndings. Although the risk of stroke tended to rise with increasing hypertension severity within each age stratum, associations weakened signicantly with advancing age (P .001). The effects of cigarette smoking on the risk of stroke also seemed to weaken with age. In contrast, the effect of atrial brillation on the risk of stroke seemed to be strongest in men 65 years versus men who were younger. The interaction between age and atrial brillation was not statistically signicant (P .404); however, this could be the consequence of lim- ited statistical power due to the infrequency of atrial brilla- tion that was observed within all age strata (Table 2). Although a protective effect of physical activity on the risk of stroke seemed unclear or modest within each age group, when modeled as a continuous risk factor, inverse associa- tions increased signicantly with age (P .046). The protec- tive effect of physical activity on the risk of stroke became statistically signicant in men 75 years (P .032). The adverse effect of total cholesterol on the risk of stroke was strongest inmenaged55to74years where large increases in morbidity and mortality from stroke begin to occur. By age 75, the relation was absent. For diabetes, whereas associ- ations with stroke were signicant for men 55 years, the risk of stroke was fairly constant across all age strata. Men with diabetes tended to have twice the risk of stroke as compared with men without diabetes. Similar data on the effect of age, hypertension, total cho- lesterol, cigarette smoking, and alcohol intake on the risk of hemorrhagic stroke are described in Table 4. Within all risk factor strata, except for hypertensive men and men with total cholesterol levels 200 mg/dL, stroke incidence increased signicantly with age. After adjusting for the other R.D. Abbott et al. / Journal of Clinical Epidemiology 56 (2003) 479486 482 Table 2 Average cardiovascular risk factors according to age Age, y Risk factor 4554 5564 6574 7593 Borderline hypertensive, % a 17.3 18.4 21.2 19.4 Hypertensive, % a 20.6 29.9 41.4 53.1 Total cholesterol, mg/dL b 219 38 c 216 38 207 35 190 34 Diabetes, % a 8.5 13.0 14.5 12.2 Atrial brillation, % a 0.3 0.5 1.3 2.8 Body mass index, kg/m 2 24.1 3.1 23.7 3.1 23.4 3.1 23.1 3.2 Cigarette smoker, % b 47.8 38.2 23.2 7.4 Alcohol intake, oz/mo a 14.9 25.3 14.1 24.9 14.6 30.6 18.3 37.6 Physical activity index b 32.9 4.8 32.6 4.3 31.6 4.1 30.6 4.4 a Signicant increase with advancing age (P .001). b Signicant decline with advancing age (P .001). c Mean standard deviation. risk factors (including diabetes, BMI, and the physical activ- ity index), the effects of age on the risk of hemorrhagic stroke remained statistically signicant only for men with borderline hypertension (P .008) and for consumers of alcohol (P .048). As with thromboembolic stroke, the positive relation be- tween hypertension and hemorrhagic events seemed to weaken with age (P .040). In contrast to thromboembolic events, elevated total cholesterol seemed to be protective against hemorrhagic stroke up until age 75. Assessments of interactions and quadratic effects were not statistically signicant. In addition, the association between cigarette smoking and hemorrhagic stroke seemed to strengthen with advancing age. In men aged 75 to 93, there was nearly a 4-fold excess of hemorrhagic events in men who smoked cigarettes (P .006). Men who drank alcohol were also at an increased risk of hemorrhagic stroke versus nondrinkers at all ages, but in men 75 years of age, the effects became insignicant. Figs. 1and2 tendtoconrmthe reported ndings in Tables 3 and 4 after risk factor adjustment. The relative risk of thromboembolic stroke in hypertensive versus normotensive men declined with advancing age (P .001), whereas the protective effect of physical activity tended to increase (P .027) (Fig. 1). Although the positive effect of diabetes on promoting stroke seemed to be strongest in men aged 75 to 93, relative risks were similar across all ages. Declines in the association between cigarette smoking and the risk of thromboembolic stroke were modest (P .591). For hemorrhagic stroke (Fig. 2), additional risk factor adjustment removedanyapparent interactioneffect betweena risk factor and age on the risk of a hemorrhagic event. Upper condence limits for hypertension and alcohol intake are not provided in instances where they are quite high and their contribution to the patterns of association in Fig. 2 are inconsequential. For all age ranges up to 75 years, hyperten- sion continued to be signicantly associated with an increased risk of hemorrhagic stroke. At 75 years, it was not. Cigarette smoking in men 75 to 93 of age also was signicantly related to an increased risk of stroke, whereas it was weak in younger men (75). Only for men aged 65 to 74 years was alcohol drinking related to an increased risk of hemorrhagic stroke (P .047). 4. Discussion Common risk factors for cardiovascular disease undergo rapid changes in frequency and distribution with advancing age. Diabetes may be one of the most important risk factors that continue to have a consistent relation with the risk of thromboembolic stroke throughout life. Attention to the risk of an embolic event due to atrial brillation is also critical because of the large increases in stroke incidence that can be attributed to this condition [28,29]. As noted in an earlier report from the Honolulu Heart Program (and based on longer follow-up), physically active lifestyles in older middle age seem to be protective against stroke, and current ndings suggest that this effect extends to the elderly population [18]. Cigarette smoking has also long been known to promote thromboembolic and hemorrhagic events, and although it becomes a less common habit with advancing age, its effects on hemorrhagic events persists [30]. In contrast, rates of hypertension increase rapidly with age, although its effect on stroke seems to decline. This does not imply that treating hypertension to reduce the risk of stroke in the elderly population is less important than in those who are younger. Rather, it is the more powerful effect of age on promoting disease in normotensive individuals that seems to diminish the perception that treating hyperten- sion has health benets with advancing age. In conjunc- tion with the low percentage of normotensive elderly men and the increased risk of stroke that occurs naturally with advancing age, demonstration that hypertension has an effect on the risk of stroke becomes more difcult in older men than in those who are younger. In the Honolulu Heart Program, for men aged 75 to 93 years, nearly 75% were borderline hypertensive or hypertensive. Over half had hypertension. Unfortunately, the effects of other important risk factors on cardiovascular disease could also become harder to mea- sure in the elderly population because of the pervasiveness R.D. Abbott et al. / Journal of Clinical Epidemiology 56 (2003) 479486 483 Table 3 Six-year incidence (rate/1000) of thromboembolic stroke according to age and risk factor status Age, y Test for Risk factor 4554 5564 6574 7593 trend Hypertension Normotensive 2.7 4.7 19.1 28.4 0.001 a Borderline 5.3 17.2 15.5 26.8 0.002 Hypertensive 19.7 19.7 33.1 40.0 0.001 Test for trend 0.001 b 0.001 0.017 0.195 0.001 c Total cholesterol 200 mg/dL 6.0 10.3 16.4 38.3 0.001 200239 mg/dL 7.1 11.4 27.0 29.5 0.001 240 mg/dL 6.8 13.9 37.0 32.7 0.001 Test for trend 0.425 0.030 0.003 0.287 0.591 Diabetes Absent 6.2 10.2 22.0 29.1 0.001 Present 10.8 19.8 36.9 71.2 0.001 Test for risk 0.250 0.012 0.042 0.001 0.584 factor effect Atrial brillation Absent 6.5 11.4 22.5 31.5 0.001 Present 71.4 27.8 139.5 114.8 0.165 Test for risk 0.015 0.839 0.001 0.001 0.404 factor effect Cigarette smoker Nonsmoker 3.2 7.9 18.5 32.1 0.001 Smoker 10.4 17.3 40.3 31.8 0.001 Test for risk 0.002 0.001 0.001 0.915 0.470 factor effect Physical activity index 30 8.1 13.3 29.4 40.5 0.001 3033 5.1 13.6 20.6 17.1 0.004 33 7.0 14.4 24.4 28.3 0.001 Test for trend 0.363 0.782 0.211 0.032 0.046 a P value for testing changes in the risk of stroke with increasing age. b Pvalue for testing changes in the risk of stroke across risk factor strata. c P value for testing differences in risk factor effects on the risk of stroke with increasing age. of the risk factor. It may be that in such situations, experience from studies of therapies and risk factors in younger cohorts could have relevance in older individuals where the absence of a risk factor is uncommon. Findings fromthe FraminghamStudy and an earlier report from Honolulu further indicate that, although the percentage of thromboembolic strokes that are attributed to hypertension declines with age, the percentage attributed to atrial brilla- tion increases with age [28,29]. Although hypertension and atrial brillation increase in prevalence with advancing age, the strength of the association between atrial brillation and a thromboembolic event seems to be stronger than it does for hypertension. Although screening and treatment for hy- pertension and atrial brillation is important, the impact of each on the risk of stroke undergoes important changes with age, largely due to increases in prevalence that are excessive for hypertension but relatively modest for atrial brillation. In the Honolulu sample, the effect of age on the risk of stroke was least apparent when atrial brillation was present (as compared with other risk factors), which is a possible consequence of more direct mechanisms that link atrial - brillation to embolic outcomes. Such ndings are difcult to interpret because stroke susceptibility could still reect a dynamic process with changing vulnerabilities to changes in risk factor exposures that occur with age or with the risk factors themselves. Com- plex associations between hypertension and cardiovascular morbidity and mortality with advancing age have also been described elsewhere [3138]. Effects of hypertension could be operating in competition with other inuences that un- dergo rapid gains in frequency as aging occurs. Stroke sus- ceptibility in the elderly population may be a sign of accumulated vascular damage or weakening that has accrued throughout middle and later adulthood. Others have sug- gested that mid-life risk factors that promote atherosclerosis may have lasting importance with advancing age. Investiga- tors from the Atherosclerotic Risk in Communities Study suggest that hypercholesterolemia, hypertension, and the use of cigarettes in middle adulthood may have residual effects on the development of subclinical atherosclerosis in later life [39]. In the Framingham Study, carotid stenosis in men whose average age was 75 years had less of an associa- tion with late-life cholesterol levels as compared with con- centrations measured earlier [40]. Recent data from the Honolulu Heart Program further question the need to lower cholesterol to excessively low concentrations in the elderly population [41]. With regard to coronary heart disease, data from Hawaii suggest that desirable cholesterol levels in the elderly population may not be a marker of a healthy risk Table 4 Six-year incidence (rate/1000) of hemorrhagic stroke according to age and risk factor status Age, y Test for Risk factor 4554 5564 6574 7593 trend Hypertension Normotensive 1.2 0.8 2.4 6.3 0.005 a Borderline 1.1 5.5 5.6 11.2 0.006 Hypertensive 7.2 12.0 12.9 11.4 0.379 Test for trend 0.003 b 0.001 0.003 0.357 0.040 c Total cholesterol 200 mg/dL 4.2 6.6 10.3 12.8 0.002 200239 mg/dL 1.3 2.9 6.2 3.1 0.080 240 mg/dL 2.0 5.8 1.8 13.1 0.072 Test for trend 0.047 0.074 0.036 0.373 0.226 Cigarette smoker Nonsmoker 1.8 3.3 5.9 8.2 0.001 Smoker 3.1 7.9 11.7 31.8 0.001 Test for risk 0.326 0.009 0.093 0.006 0.412 factor effect Alcohol drinker Nondrinker 1.7 2.2 3.2 6.7 0.039 Drinker 2.7 6.4 9.7 12.6 0.001 Test for trend 0.005 0.001 0.014 0.975 0.030 a P value for testing changes in the risk of stroke with increasing age. b Pvalue for testing changes in the risk of stroke across risk factor strata. c P value for testing differences in risk factor effects on the risk of stroke with increasing age. R.D. Abbott et al. / Journal of Clinical Epidemiology 56 (2003) 479486 484 Fig. 1. Estimated relative risk of thromboembolic stroke with advancing age comparing hypertensive men with normotensive men, diabetic men with nondiabetic men, cigarette smokers with nonsmokers, and men in the top versus bottom tertiles of the physical activity index. Relative risk estimates have been adjusted for each of the other risk factors, including BMI and alcohol intake. factor prole, particularly if past cholesterol levels were high [42]. Risk factors that increase rapidly in frequency with ad- vancing age, such as lowankle/brachial blood pressure index (ABI) (0.9), could be one sign of an accumulation of a lifetime of generalized vascular damage. As opposed to an isolated single risk factor effect that might be more important in younger years and where vascular disease may be less apparent, a lowABI in late life could reect systemic damage from a variety of sources that include hypertension, diabetes, and past cigarette smoking [43]. In the elderly population, carry-over effects from past cigarette smoking and adverse lipid proles may be harder to measure by screening for risk factors that exist in late life, particularly in subjects who were once heavy smokers or in those who experienced large reductions in total cholesterol with advancing age. In the Honolulu Heart Program, the prevalence of an ABI 0.9 was observed in 6.3% of men aged 71 to 74 years and in more than 25% in men older than 85 years [43]. Low ABI was also signicantly related to the risk of stroke in men without hypertension or diabetes and in noncigarette smok- ers [43]. Although hypertension is thought to be the primary risk factor for stroke, data from the Honolulu Heart Program R.D. Abbott et al. / Journal of Clinical Epidemiology 56 (2003) 479486 485 Fig. 2. Estimated relative risk of hemorrhagic stroke with advancing age comparing hypertensive men with normotensive men, men with total cholesterol levels at 240 versus 200 mg/dL, smokers with nonsmokers, and alcohol drinkers with nondrinkers. Relative risk estimates have been adjusted for each of the other risk factors, including diabetes, BMI, and the physical activity index. suggest that its role as a marker for disease may become less clear as an independent risk factor with advancingage but no less important as a condition in need of prevention and treatment. Consistent risk factor effects on the risk of stroke include the effects of diabetes and atrial brillation, but these and other classic risk factors do not seem to account for the increasing incidence of thromboembolic stroke that occurs with advancing age in men with normal blood pres- sures. Other factors not available in the current study could gain in importance and may eventually equal or exceed the impact that hypertensionhas onthe riskof stroke inthe elderly population. The current ndings that active lifestyles are associated with a reduced risk of stroke in the elderly popula- tion, combined with earlier reports that describe the health benets of walking in nonsmoking retired men, suggest that encouraging active lifestyles in older individuals warrants special emphasis [44,45]. Acknowledgments This study was supported by the National Heart, Lung, and Blood Institute (contract NO1-HC-05102 and grant U01-HL-56274), the National Institute on Aging (contract R.D. Abbott et al. / Journal of Clinical Epidemiology 56 (2003) 479486 486 NO1-AG-4-2149), and by the American Heart Association of Hawaii (grant HIGS-16-97). References [1] American Heart Association. Heart and stroke facts. Dallas: American Heart Association National Center; 1992. [2] Burt VL, Cutler JA, Higgins M, et al. Trends in the prevalence, awareness, treatment and control of hypertension in the adult US population: data from the health examination surveys, 1960 to 1991. Hypertension 1995;26:609. [3] Joint National Committee on Prevention. Detection, evaluation and treatment of high blood pressure. Arch Intern Med 1997;157:241346. [4] Harris MI, Flegal KM, Cowie CC, et al. Prevalence of diabetes, impaired fasting glucose, and impaired glucose tolerance in U.S. adults: the Third National health and Nutrition Examination Survey, 19881994. Diabetes Care 1998;21:51824. [5] Cornoni-Huntley JC, Harris TB, Everett DF, et al. An overview of body weight of older persons, including the impact on mortality: The National Health and Nutrition Examination Survey I: epidemiologic follow-up study. J Clin Epidemiol 1991;44:74353. [6] Ferrara A, Barrett-Conner E, Shan J. Total, LDL, and HDL cholesterol decrease with age in older men and women: the Rancho Bernardo Study 19841994. Circulation 1997;96:3743. [7] Wilson PWF, Anderson KM, Harris T, et al. Determinants of change in total cholesterol and HDL-C with age: the Framingham Study. J Gerontol Med Sci 1994;49:2527. [8] Wilson PWF, Hoeg JM, DAgostino RB, et al. Cumulative effects of high cholesterol levels, high blood pressure, and cigarette smoking on carotid stenosis. N Engl J Med 1997;337:51622. [9] Hershcopf RJ, Elahi D, Andes R, et al. Longitudinal study in serum cholesterol in man: an epidemiologic search for an etiology. J Chron Dis 1982;35:10114. [10] Weijenberg MP, Feskens EJM, Kromhout D. Age-related changes in total and high-density-lipoprotein cholesterol in elderly Dutch men. Am J Public Health 1996;86:798803. [11] Abbott RD, Yano K, Hakim AA, et al. Changes in total and high- density lipoprotein cholesterol over 10- and 20-year periods (The Honolulu Heart Program). Am J Cardiol 1998;82:1728. [12] Kagan A, Harris BR, Winkelstein Jr W, et al. Epidemiologic studies of coronary heart disease and stroke in Japanese men living in Japan, Hawaii and California: demographic, physical, dietary, and biochemi- cal characteristics. J Chron Dis 1974;27:34564. [13] Yano K, Reed DM, McGee DL. Ten-year incidence of coronary heart disease in the Honolulu Heart Program: relationship to biologic and lifestyle characteristics. Am J Epidemiol 1984;119:65366. [14] Kagan A, Popper JS, Rhoads GG. Factors related to stroke incidence in Japanese men: the Honolulu Heart Study. Stroke 1980;11:1421. [15] Kagan A, Popper J, Reed DM, et al. Trends in stroke incidence and mortality in Hawaiian Japanese men. Stroke 1994;25:11705. [16] Castelli WP, Cooper GR, Doyle JT, et al. Distribution of triglyceride and total LDLand HDLcholesterol in several populations: the Coopera- tive Lipoprotein Phenotyping Study. J Chron Dis 1977;30:14769. [17] Donahue RP, Abbott RD, Reed DM, et al. Physical activity and coronary heart disease in middle-aged and elderly men: the Honolulu Heart Program. Am J Public Health 1988;78:6835. [18] Abbott RD, Rodriguez BL, Burchel CM, et al. Physical activity in older middle-aged men and reduced risk of stroke: the Honolulu Heart Program. Am J Epidemiol 1994;139:88193. [19] Kannel WB, Sorlie PD. Some health benets of physical activity: the Framingham Study. Arch Intern Med 1979;139:85761. [20] Garcia-Palmieri MR, Costas R, Cruz-Vidal M, et al. Increased physi- cal activity: a protective factor against heart attack in Puerto Rico. Am J Cardiol 1982;50:74955. [21] Yano K, Reed DM, Curb JD, et al. Biological and dietary correlates of plasma lipids in elderly Japanese men in Hawaii. Arteriosclerosis 1986;6:42233. [22] Roberston TL, Kato H, Rhoads GG, et al. Epidemiologic studies of coronary heart disease and stroke in Japanese men living in Japan, Hawaii, and California: incidence of myocardial infarction and death from coronary heart disease. Am J Cardiol 1977;39:23943. [23] Knutsen R, Knutsen SF, Curb JD, et al. The incidence of coronary heart disease in the Honolulu Heart Program. J Clin Epidemiol 1988;41:293302. [24] Cox DR. Regression models and life tables. J R Stat Soc 1972; 34:187202. [25] Lane PW, Nelder JA. Analysis of covariance and standardization as instances of prediction. Biometrics 1982;38:61321. [26] Wu M, Ware J. On the use of repeated measurements in regression analysis with dichotomous responses. Biometrics 1979;35:51321. [27] Abbott RD. Logistic regression in survival analysis. Am J Epide- miol 1985;121:46571. [28] Wolf PA, Abbott RD, Kannel WB. Atrial brillation as an independent risk factor for stroke: the Framingham Study. Stroke 1991;22:9837. [29] Curb JD, Abbott RD, MacLean CJ, et al. Age-related changes in stroke risk in men with hypertension and normal blood pressure. Stroke 1996;27:81924. [30] Abbott RD, Yin Y, Reed DM, et al. Risk of stroke in male cigarette smokers. N Engl J Med 1986;315:71720. [31] Taylor JO, Cornoni-Huntley J, Curb JD, et al. Blood pressure and mortality in the elderly. Am J Epidemiol 1991;134:489501. [32] Mattila K, Haavisto M, Rajala S, et al. Blood pressure and survival in the very old. BMJ 1988;296:8879. [33] Langer RD, Ganiats TG, Barrett-Connor E. Paradoxical survival of elderly men with high blood pressure. BMJ 1989;298:13568. [34] Staessen J, Bulpitt C, Clement D, et al. Relation between mortality and treated blood pressure in elderly patients with hypertension: report of the European Working Party on High Blood Pressure in the Elderly. BMJ 1989;298:15526. [35] Coope J, Warrender TS, McPherson K. The prognostic signicance of blood pressure in the elderly. J Hum Hypertens 1988;2:7988. [36] Siegel D, Kuller L, Lazarus NB, et al. Predictors of cardiovascular events and mortality in the Systolic Hypertension in the Elderly Pilot Project. Am J Epidemiol 1987;126:38599. [37] Lindholm L, Lanke J, Bengtsson B. U-shaped association between mortality and blood pressure in a thirteen year prospective study. Fam Pract 1986;3:38. [38] Abbott RD, Curb JD, Rodriguez BL, et al. Age-related changes in risk factor effects on the incidence of coronary heart disease. Ann Epidemiol 2002;12:17381. [39] Nieto FJ, Diez-Roux A, Szklo M, et al. Short- and long-term predic- tion of clinical and subclinical atherosclerosis by traditional risk factors. J Clin Epidemiol 1999;52:55967. [40] Wilson PWF, Hoeg JM, DAgostino RB, et al. Cumulative effects of high cholesterol levels, high blood pressure, and cigarette smoking on carotid stenosis. N Engl J Med 1997;337:51622. [41] Schatz IJ, Masaki K, Yano K, et al. Cholesterol and all-cause mortality in elderly people from the Honolulu Heart Program: a cohort study. Lancet 2001;358:3515. [42] Hakim AA, Curb JD, Burchel CM, et al. Screening for coronary heart disease in elderly men based on current and past cholesterol levels. J Clin Epidemiol 1999;52:125765. [43] Abbott RD, Rodriguez BL, Petrovitch H, et al. Ankle/brachial blood pressure in elderly men and the risk of stroke: the Honolulu Heart Program. J Clin Epidemiol 2001;54:538. [44] Hakim AA, Petrovitch H, Burchel CM, et al. Effects of walking on mortality among nonsmoking retired men. N Engl J Med 1998; 338:949. [45] Hakim AA, Curb JD, Petrovitch H, et al. Effects of walking on coronary heart disease in elderly men: the Honolulu Heart Program. Circulation 1999;100:913. Reproducedwith permission of thecopyright owner. Further reproductionprohibited without permission.