THE PREVALENCE OF CONGESTIVE

ORIGINAL CONTRIBUTION Increased Pulse Pressure and Risk of Heart Failure in the Elderly Claudia U. Chae, MD Marc A. Pfeffer, MD, PhD Robert J. Glynn, ScD Gary F. Mitchell, MD James O. Taylor, MD Charles H. Hennekens, MD, DrPH THE PREVALENCE OF CONGESTIVE heart failure (CHF), an important cause of morbidity and mortality, will increase with the aging of the US population. The incidence of CHF doubles in each decade after the age of 45 years, and CHF is the leading hospital diagnosis in those older than 65 years. 1 Risk factors such as hypertension have been well described. 2 However, data are sparse regarding other risk factors for CHF specifically in the elderly. Pulse pressure (PP), measured as the difference between systolic blood pressure (SBP) and diastolic blood pressure (DBP), rises markedly after the fifth decade due to arterial stiffening with age, which results in a progressive rise in SBP and fall in DBP as the elastic capacity of the aorta diminishes. 3-6 Arterial stiffness increases afterload 7,8 and myocardial oxygen demand, 9 impairs ventricular relaxation, 10 and causes subendocardial ischemia in the presence 11,12 or absence 13 of coronary artery stenoses. These effects may increase the susceptibility of the aging heart to ischemia and ventricular dysfunction. Arterial stiffness has been shown to be a risk factor for cardiovascular disease (CVD). In prospective cohorts, PP is associated with risk of myocardial infarction (MI) 14-16 and CVD mortality 17 in normotensive and hypertensive populations, as well as reinfarction and total mortality after MI in patients with left ventricular Context Arterial stiffness increases with age. Thus, pulse pressure, an index of arterial stiffening, may predict congestive heart failure (CHF) in the elderly. Objective To study prospectively the association between pulse pressure and risk of CHF. Design Prospective cohort study. Setting The community-based East Boston Senior Health Project, East Boston, Mass. Patients A total of 1621 men and women (mean [SD] age, 77.9 [5.0] years) free of CHF who had blood pressure measurements taken in 1988-1989 and were followed up for 3.8 years. Main Outcome Measure Incidence of CHF as ascertained by hospital discharge diagnosis (n = 208) and death certificates (n = 13). Results After controlling for age, sex, mean arterial pressure, history of coronary heart disease, diabetes mellitus, atrial fibrillation, valvular heart disease, and antihypertensive medication use, pulse pressure was an independent predictor of CHF. For each 10-mm Hg elevation in pulse pressure, there was a 14% increase in risk of CHF (95% confidence interval, 1.05-1.24; P =.003). Those in the highest tertile of pulse pressure ( 67 mm Hg) had a 55% increased risk of CHF (P =.02) compared with those in the lowest ( 54 mm Hg). Pulse pressure was more predictive than systolic blood pressure alone and was independent of diastolic blood pressure. Conclusion Pulse pressure, an easily measurable correlate of pulsatile hemodynamic load, is an independent predictor of risk of CHF in this elderly cohort. JAMA. 1999;281:634-639 www.jama.com dysfunction. 18 Measures of arterial stiffness are strongly correlated with left ventricular hypertrophy. 19,20 However, none of these studies examined the possible relationship of arterial stiffness with CHF. We addressed the hypothesis that increased arterial stiffness may contribute to CHF in the elderly, by prospectively examining whether PP is an independent predictor of CHF risk in a community-based elderly cohort. METHODS Study Population The study population consisted of participants in the East Boston Senior Health Project, East Boston, Mass, 1 of the 4 centers of the National Institute on Aging s Established Populations for Epidemiologic Studies of the Elderly. A community census, conducted in 1982-1983, identified 4497 East Boston residents aged 65 years orolder, ofwhom3809(85%) participated in baseline interviews and in-home evaluations. Follow-up in-home examinations were done in 1985-1986 and were done in 1988-1989, and telephone interviews Author Affiliations: Division of Preventive Medicine (Drs Chae, Glynn, and Hennekens) and the Cardiovascular Division (Drs Pfeffer and Mitchell), Department of Medicine, Brigham and Women s Hospital, the Cardiology Division, Department of Medicine, Massachusetts General Hospital (Dr Chae); the Department of Ambulatory Care and Prevention, Harvard Medical School (Dr Hennekens), the Departments of Biostatistics (Dr Glynn) and Epidemiology (Dr Hennekens), Harvard School of Public Health, Boston, Mass; and the East Boston Neighborhood Health Center, East Boston, Mass (Drs Taylor and Hennekens). Corresponding Author and Reprints: Claudia U. Chae, MD, Division of Preventive Medicine, Brigham and Women s Hospital, 900 Commonwealth Ave E, Boston, MA 02215-1204 (e-mail: chae.claudia@mgh.harvard.edu). 634 JAMA, February 17, 1999 Vol 281, No. 7 1999 American Medical Association. All rights reserved.

were completed in the other years. The Medicare Provider Analysis and Review files from the Health Care Financing Administration (HCFA) provided information on all hospital admissions and 5 discharge diagnoses per participant admission from 1985 to 1992. Subjects with unavailable HCFA data (n = 19) were excluded. Vital status through December 1992 was determined through local death notices, records of the East Boston Neighborhood Health Center, and yearly linkage with the National Death Index. For the entire cohort, mortality follow-up was complete, and 99.2% of death certificates were obtained and coded by a single nosologist using the International Classification of Diseases, Ninth Revision (ICD-9). In 1988-1989, 2348 subjects provided follow-up data either in person or through a proxy, 208 refused or were lost to follow-up, and 1253 had died. Of these, 1810 had blood pressure (BP) measurements. To define a cohort free of evidence of prior CHF, we excluded those with a prior HCFA diagnosis of CHF (n = 138) and those without a prior HCFA diagnosis of CHF but who were taking combinations of digoxin and loop diuretics (n = 25) or digoxin, loop diuretics, and angiotensin-converting enzyme inhibitors (n = 6). We also excluded 20 subjects with missing medication data. This resulted in a cohort of 1621 subjects without known CHF who had complete BP measurements and were followed up prospectively for 3.8 years (median). Exposure Variables Using a standard mercury sphygmomanometer, trained interviewers took 3 BP measurements at 30-second intervals after the participant had been seated for at least 5 minutes, using the protocol from the Hypertension Detection and Follow-up Program. 21 Details of quality control of BP measurements have been published previously. 22 The averages of 3 SBP and DBP measurements were used. Mean arterial pressure (MAP) was defined as {[SBP + (DBP 2)]/3}. Pulse pressure was defined as SBP DBP. We assessed demographic characteristics, medical history, functional abilities, and health habits. 23 Interviewers inspected all medication containers for prescription and nonprescription drugs taken in the prior 2 weeks. Drugs were coded using the Drug Product Information Coding System. Hypertension was defined as any of the following: self-report of high BP, previous HCFA diagnosis, use of antihypertensive medications, or an SBP greater than 160 mm Hg or a DBP greater than 90 mm Hg on examination. Diabetes mellitus (DM) was defined by selfreport, use of insulin or oral hypoglycemic agents, or previous HCFA diagnosis. Prevalent CHD (ICD-9 code 410-414, prior percutaneous transluminal coronary angioplasty or coronary artery bypass grafting), valvular heart disease, and atrial fibrillation (AF) were defined by HCFA diagnoses. Smoking status (current, past, never) and alcohol use (none in past year, none in past month, 15 g/d, 15 g/d) were categorized based on self-report. Outcome Measures In a median of 3.8 years of follow-up, 221 subjects reached the end point of first hospitalization with CHF, defined by ICD-9 code 428 in any of 5 fields for any hospitalization after the subject s 1988-1989 evaluation (n = 208) or death with CHF on the death certificate but no prior HCFA diagnosis of CHF (n = 13). Statistical Methods We used analysis of variance for continuous variables and tested for linear trends for categorical variables to examine study population characteristics stratified by tertiles of PP. Linear regression models were used to determine predictors of PP. Participants contributed follow-up time from the date of their baseline examination in 1988-1989 to the date of first hospitalization with CHF, death, or end of follow-up, whichever came first. Hazard ratios (HRs) were calculated using Cox proportional hazards models, controlling for age, to evaluate the association of BP parameters with CHF risk. Continuous variables were examined in categories chosen a priori to determine if they had a linear relationship with the outcome. The categoriesforsbp( 120[referent],120-139, 140-159, or 160 mm Hg) and DBP ( 70, 70-79 [referent], 80-89, or 90 mm Hg) were chosen to reflect clinically relevant values. Because clinically relevant values are unknown for PP, tertiles of less than 54 (referent), 54 to 67, and greater than 67 mm Hg were chosen. Tests for trend were performed by assigning an ordinal variable for each category of the BP parameter and modeling this as a continuous variable in separate Cox proportional hazards models. If a significant linear relationshipwasfound, thevariablewasused as a continuous variable. Other clinical predictors of CHF risk and potential confounders were tested individually for associations with CHF risk in separate Cox models, after controlling for age. The relationships between BP parameters and risk of CHF were examined in multivariate Cox proportional hazards model, controlling for age, sex, CHD, and significant univariate predictors of CHF risk. For each HR, 2-sided P values and 95% confidence intervals (CIs) were calculated. In analyses assessing the relationship of PP with other CVD events and CHF, we excluded participants with prior CHD at baseline (n = 167), leaving a total of 1454 subjects of whom 190 developed CHF, 118 had nonfatal or fatal MIs (defined as the primary or underlying cause of death on the death certificate, without prior known CHD), and 68 died with CVD attributed as the immediate or underlying cause of death. In multivariate proportional hazards models, we examined the relationship of PP with MI and total CVD (259 first events of MI, CHF, or CVD death). To assess whether the association between PP and CHF risk is mediated through CHD, subjects with MI occurring before or after the incident CHF hospitalization were excluded, or MI was controlled for. Effect modification by sex, DM, and CHF risk factor status (defined as having any of the following: DM, CHD, AF, valvular heart disease, or antihypertensive medication use) was also examined. RESULTS Clinical Characteristics The study population (n = 1621) was elderly (mean [SD] age, 77.9 [5.0] years) 1999 American Medical Association. All rights reserved. JAMA, February 17, 1999 Vol 281, No. 7 635

and predominantly female (66.1%). Prevalent conditions at baseline included hypertension (54.6%), DM (19.3%), CHD (10.3%), AF (2.8%), and valvular heart disease (1.8%). Nearly half (45.6%) were taking antihypertensive medications, and 12% were current smokers. Mean (SD) BP parameters included SBP of 137 (19), DBP of 75 (10), MAP of 96 (11), and PP of 62 (17) mm Hg. Age, MAP, presence of DM, hypertension, vascular heart disease, and use of antihypertensive medication significantly increased with tertiles of PP (TABLE 1). Sex, CHD, smoking status, and AF were not related to PP. In a multivariate regression model, independent predictors of PP included age, MAP, and DM (all P.001), antihypertensive medication use (P =.002), and valvular heart disease (P =.008). Using Spearman rank order correlation coefficients (r), PP was correlated with MAP (r = 0.41; P.001) and SBP (r = 0.83; P.001); the correlation of PP with DBP was small although statistically significant (r = 0.07; P =.004). Systolic blood pressure and DBP were also correlated (r = 0.44; P.001). Table 1. Clinical Characteristics Stratified by Pulse Pressure Tertile* BP and CHF Risk Age was a strong univariate predictor of CHF risk (HR, 1.44/5 y; 95% CI, 1.28-1.61; P.001). After controlling for age (FIGURE), systolic blood pressure had a graded relationship with increasing risk of CHF (P for trend,.001) with a 12% increase in risk for every 10-mm Hg rise in SBP (95% CI, 1.05-1.20). Diastolic blood pressure, however, tended to have a U-shaped relationship with CHF risk. Compared with subjects whose DBP ranged from 70 to 79 mm Hg, those with DBPs less than 70, 80 to 89, and greater than 90 mm Hg had HRs of 1.60 (95% CI, 1.17-2.20; P =.003), 1.10 (95% CI, 0.76-1.60; P =.60), and 1.69 (95% CI, 1.05-2.78; P =.04), respectively. Pulse pressure had a significant linear relationship with CHF risk (P for trend,.001), with an age-adjusted 17% increase in risk for every 10-mm Hg rise in PP (95% CI, 1.09-1.26). Subjects in the highest tertile of PP had a 75% increase in risk of CHF (95% CI, 1.25-2.45; P =.001) compared with those in the lowest. The relationship of PP to CHF risk was not materially altered by the presence (to control for the steady component of BP) 4 or absence of MAP in the model. MAP had no association with CHF in these data. After adjusting for age, use of antihypertensive medications, DM, valvular heart disease, and AF were significant predictors of CHF (TABLE 2). Sex, smoking, and alcohol use were not associated with CHF risk. Coronary heart disease was included as a covariate in the multivariate models with the ageadjusted association of borderline significance (P =.07). In a multivariate proportional hazards model controlling for age, sex, MAP, CHD, and significant univariate predictors of CHF risk (DM, AF, valvular disease, antihypertensive medication use), PP was a significant independent predictor of CHF (P for trend,.003) (Table 2). Every 10-mm Hg rise in PP was associated with a 14% increase in risk of CHF (95% CI, 1.05-1.24). Subjects in the highest tertile (PP 67 mm Hg) had a 55% greater risk of CHF (P =.02) compared with those in the lowest (PP 54 mm Hg). We addressed the relative importance of PP, SBP, and DBP in predicting CHF risk in multivariate proportional hazards models examining BP parameters alone and in combination (TABLE 3). When PP and SBP were examined in the same model, PP remained a significant independent predictor of CHF risk (HR per 10-mm Hg increase in PP, 1.16; 95% CI, 1.00-1.35; P =.049), while SBP did not Figure. Age-Adjusted Blood Pressure Parameters and Congestive Heart Failure Risk Hazard Ratio 2.5 2.0 1.5 1.0 0.5 0 Pulse Pressure, mm Hg 54 54-67 67 Characteristic (n = 534) (n = 547) (n = 540) P Value Age, mean ± SD, y 77.3 ± 4.8 77.9 ± 5.2 78.6 ± 4.9.001 Mean arterial pressure, 90.9 ± 9.0 95.3 ± 10.1 101.5 ± 11.7.001 mean ± SD, mm Hg Male sex 33.0 37.5 31.3.51 Diabetes mellitus 13.3 17.2 27.2.001 Hypertension 44.0 49.5 70.2.001 Antihypertensive medication 38.8 44.6 53.3.001 Coronary heart disease 9.2 10.8 10.9.34 Atrial fibrillation 2.4 2.2 3.7.20 Valvular heart disease 1.1 1.5 2.8.04 *Data are presented as percentages unless otherwise indicated. P values given for analysis of variance of continuous variables and test for linear trend for categorical variables. <120 120-139 140-159 160 <70 70-79 80-89 90 <54 54-67 >67 SBP, mm Hg DBP, mm Hg PP, mm Hg Univariate relationships of age-adjusted systolic blood pressure (SBP), diastolic blood pressure (DBP), and pulse pressure (PP) to congestive heart failure risk. 636 JAMA, February 17, 1999 Vol 281, No. 7 1999 American Medical Association. All rights reserved.

(HR/10 mm Hg, 0.97; 95% CI, 0.84-1.11; P =.63), suggesting that PP provides additional predictive information independent of SBP alone. In a separate multivariate model of PP and DBP, the effect of PP (HR/10 mm Hg, 1.12; 95% CI, 1.04-1.20; P =.003) on CHF risk was independent of DBP. In this model, a DBP of less than 70 mm Hg (HR, 1.54; 95% CI, 1.12-2.12; P =.007), which is an additional correlate with large artery stiffness, 3 was also independently associated with CHF risk. In these analyses, PP was independent of SBP and DBP, was more informative than SBP alone, and was additive to DBP in predicting CHF risk. PP and CHD Events In prior prospective studies in middleaged cohorts, 14-18 PP was a predictor of MI and CVD death. In this elderly population, after excluding those with prevalent CHD at baseline, PP was associated with small numbers of incident MI, but this was of borderline statistical significance after multivariate adjustment (HR per 10-mm Hg increase in PP, 1.11; 95% CI, 0.99-1.24; P =.07). For total CVD, a combined end point of CHF, MI, or death due to CVD, PP was a significant independent predictor of total CVD risk (HR/10 mm Hg, 1.12/10 mm Hg; 95% CI, 1.03-1.21; P =.006). After excluding subjects with MI occuring before or after the incident CHF presentation, PP remained a significant predictor of CHF (multivariate HR/10 mm Hg, 1.15; 95% CI, 1.02-1.29; P =.02). When those subjects were included and MI was controlled for, the HR was 1.11 (95% CI, Table 2. Proportional Hazards Models of Pulse Pressure and Congestive Heart Failure Risk* Age-Adjusted Models Multivariate Model Variable HR (95% CI) HR (95% CI) P Value Age, per 5 y 1.44 (1.28-1.61) 1.37 (1.22-1.54).001 PP, per 10 mm Hg 1.17 (1.09-1.26) 1.14 (1.05-1.24).003 MAP, per 1 mm Hg 1.01 (1.00-1.02) 1.00 (0.98-1.01).63 Diabetes mellitus 1.88 (1.40-2.50) 1.66 (1.24-2.23).001 Valvular heart disease 3.27 (1.73-6.16) 2.65 (1.34-5.26).005 Atrial fibrillation 3.02 (1.85-4.93) 2.84 (1.69-4.77).001 Antihypertensive medication 1.60 (1.22-2.08) 1.45 (1.10-1.91).009 Male sex 1.24 (0.94-1.63) 1.34 (1.01-1.78).04 Coronary heart disease 1.42 (0.97-2.08) 0.88 (0.57-1.36).56 *HR indicates hazard ratio; CI, confidence interval; PP, pulse pressure; and MAP, mean arterial pressure. Table 3. Multivariate Proportional Hazards Models of Blood Pressure Components and Risk of Congestive Heart Failure* Variable Coefficient HR (95% CI) P Value Pulse pressure, per 10 mm Hg.12846 1.14 (1.05-1.24).003 Systolic BP, per 10 mm Hg.08470 1.09 (1.02-1.16).01 Pulse pressure, per 10 mm Hg.15108 1.16 (1.00-1.35).049 Systolic BP, per 10 mm Hg.03367 0.97 (0.84-1.11).63 Diastolic BP, mm Hg 70.46810 1.60 (1.16-2.19).004 80-89.13439 1.14 (0.79-1.66).48 90.51856 1.68 (1.02-2.76).04 Pulse pressure, per 10 mm Hg.11109 1.12 (1.04-1.20).003 Diastolic BP, mm Hg 70.43464 1.54 (1.12-2.12).007 80-89.13535 1.15 (0.79-1.66).48 90.50624 1.66 (1.01-2.73).046 *All models controlled for age, sex, diabetes mellitus, atrial fibrillation, coronary heart disease, valvular heart disease, and use of antihypertensive medications. HR indicates hazard ratio; CI, confidence interval; and BP, blood pressure. Controlled for mean arterial pressure and the above variables. Models using diastolic blood pressure included 70 to 79 mm Hg as reference category. 1.01-1.22; P =.03). Thus, the association of PP with CHF risk was independent of MI and was observed in those without evidence of CHD. Effect Modification There were no significant interactions between PP and DBP, sex, DM, antihypertensive medication use, or CHF risk factor status. However, the association between PP and CHF risk was somewhat stronger in men and those without CHF risk factors (TABLE 4). Diabetes mellitus did not appear to alter the association between PP and CHF risk. Pulse pressure was not a significant predictor of CHF risk in the subgroup of patients taking antihypertensive medications. This is most likely a chance finding in this subgroup, or possibly a reflection of the effects of treatment, but we do not have adequate power to address this issue and had not specified this hypothesis a priori. COMMENT Pulse pressure, an easily measurable correlate of arterial stiffness and pulsatile hemodynamic load, 3-5 was an independent predictor of CHF risk in this prospective cohort. Pulse pressure was the single most informative BP parameter, as it best captures the physiological sequelae of arterial stiffness as reflected in disproportionately elevated SBP, decreased DBP, or both. These data support the hypothesis that the relative importance of BP components in predicting CVD risk varies with age, with PP emerging as a better marker for vascular overload in the elderly. 3,24 Prior studies of risk factors for CHF focused on systolic or diastolic hypertension, 2 and limited data exist regarding the role of specific BP components. In the Framingham Heart Study, PP was a better predictor than SBP of CHF risk in men, especially after age 65 years, but not in women. 25 In elderly patients with isolated systolic hypertension, a significant subgroup of those with age-related vascular stiffening and increased PP 3,4 treatment reduced the risk of fatal or nonfatal CHF by 49% in 4.5 years of follow-up. 26 1999 American Medical Association. All rights reserved. JAMA, February 17, 1999 Vol 281, No. 7 637

Pulse pressure reflects the complex interaction between intermittent cardiac ejection and the dynamic properties of the large arteries. 3-5 The distensible, elastic aorta transiently stores part of the stroke volume during each systole and transmits this volume forward in diastole as its walls recoil. This limits endsystolic pressure and wall tension 8 and maintains diastolic perfusion pressure. The progressive loss of aortic compliance with age, due to elastin fragmentation and increased collagen deposition, 27 results in elevation in PP, from the associated rise in SBP and fall in DBP. Several plausible mechanisms may explain an association between increased PP and CHF. Increased afterload 7,8 and premature return of reflected waves in late systole 28 impose a greater load on the ventricle, reducing ejection fraction 8 and increasing myocardial oxygen demand. 9 Diastolic relaxation is also impaired. 10 Conduit vessel stiffness may increase vulnerability to ischemia as well. The increase in systolic load and myocardial work, coupled with the fall in diastolic coronary perfusion pressure, results in subendocardial ischemia with 11,12 and without coronary artery stenoses. 13 Arterial stiffness is strongly correlated with left ventricular hypertrophy in normotensive 20 and hypertensive 19 subjects and with left ventricular mass in population-based cohorts, 14,29 a plausible consequence of its excess hemodynamic load and a known risk factor for CHF and CVD events. 30 In addition to hemodynamic consequences, the association between elevated PP and CHF risk may in part be mediated by CHD. Arterial stiffness is correlated with atherosclerosis 31,32 and may promote primary atheroma development, possibly mediated by greater cyclic wall stress. 33,34 Several prospective studies 14-18 in middle-aged populations support the hypothesis that higher PP is associated with increased CHD, but none addressed its relationship with CHF. In our data of the elderly, the association between PP and CHF was independent of CHD, suggesting that it may be mediated by other adverse effects of arterial stiffness on hemodynamics, systolic and diastolic function, and myocardial perfusion. These findings have potential preventive and therapeutic implications regarding modifiability of arterial stiffness and perhaps risk of CHF. For example, angiotensin-converting enzyme inhibitors cause greater regression of left ventricular hypertrophy than other vasodilators despite lowering MAP to the same degree. 35 This may be due to reduction in pulsatile load and premature wave reflection 35 or modulation of reninangiotensin system effects on arterial compliance. 36 Nitrates 37 and low-salt diets 38 improve arterial distensibility. Adverse effects on coronary blood supply, including decreased diastolic perfusion pressure 11,13 and a greater proportion of myocardial perfusion occurring in systole, 12,39 may have implications for treatment of elderly patients with CHD, in whom hearts perfused by wide PPs may be more sensitive to reductions in SBP and MAP. 39 As recently described, 12 increased pulsatile load during acute coronary occlusion resulted in a 3-fold decline in ejection fraction, marked fall in SBP, and doubling of ischemic mass. This may contribute to the greater severity of CHF and left ventricular dysfunction seen in the elderly after MI, independent of the severity of coronary stenosis. 12 Several issues regarding our study methods should be addressed. Brachial artery BPs by sphygmomanometry are less accurate in measuring central pressure than more invasive or technologically advanced methods. However, this type of BP measurement is easily obtained in a clinical setting, and in an elderly population, such as ours, peripheral and central PP are better correlated than in younger subjects. 3,4 We also used the average of 3 BPs taken by Hypertension Detection and Follow-up Program 21 protocol to increase the accuracy of our measurements. In other prospective studies, 14,18 even a single, casual BP measurement yielded significant positive associations between PP and CVD events. While we used HCFA diagnoses to identify end points, a validation study found that ICD-9 codes for CHF (428, 402) had sensitivities and positive predictive values, respectively, of 0.89 and 0.71 in any position, and 0.85 and 0.87 in the principal position. 40 In a study of Medicare admissions in 1991 and 1992 in 4 states, the principal diagnosis of CHF (ICD-9 code 428) was confirmed by clinical criteria in 95.7% of cases. 41 When we restricted our analysis to CHF as the principal diagnosis, the magnitude of association was not materially altered, although the CI widened due to the smaller number of end points. Hospitalization is a reasonable but imperfect indicator of incident CHF. While most patients require hospitalization at some point in their disease course, this method is biased toward selection of sicker patients and will not identify nonhospitalized cases of CHF. A potential limitation is that some subjects may have had CHF at baseline. Given the natural history of CHF, with its poor prognosis and high rate of Table 4. Pulse Pressure and Congestive Heart Failure (CHF) Risk Within Subgroups* Variable No. HR (95% CI) P Value Taking antihypertensive medication 739 1.05 (0.94-1.17).42 No antihypertensive medication 882 1.29 (1.13-1.47).001 Female 1071 1.10 (0.99-1.22).07 Male 550 1.21 (1.04-1.41).01 Diabetes mellitus 312 1.16 (1.00-1.35).06 No diabetes mellitus 1309 1.13 (1.02-1.25).02 Any CHF risk factor(s) 1031 1.11 (1.02-1.22).02 No CHF risk factors 590 1.42 (1.14-1.76).001 *HR indicates hazard ratio; CI, confidence interval. Hazard ratio reported for every 10-mm Hg increment of pulse pressure after adjustment for age, mean arterial pressure, sex, coronary heart disease, and other significant predictors of CHF risk. CHF risk factors include coronary heart disease, atrial fibrillation, valvular heart disease, diabetes, and hypertension. Adjusted for age, mean arterial pressure, and sex. 638 JAMA, February 17, 1999 Vol 281, No. 7 1999 American Medical Association. All rights reserved.

hospitalization, 1 we sought to address this by excluding subjects at baseline with a prior HCFA diagnosis in the preceding 3 years or with evidence of active drug therapy consistent with CHF treatment. Whether our findings apply to populations in middle age requires further study. The pathophysiology of arterial stiffness suggests that the association between PP and CHF risk may apply predominantly to the elderly. We do not have data about left ventricular function in these subjects but speculate that a link may exist between progressive arterial stiffness, with its adverse effects on ventricular relaxation, loading conditions, and myocardial perfusion, and the increase in diastolic heart failure observed in the elderly. In summary, in this prospective study of a community-based elderly cohort, PP was independently and linearly associated with risk of CHF. Pulse pressure may be a clinically useful tool to identify elderly subjects at risk for CHF, who may benefit from treatment to improve arterial compliance. More research is needed to further examine the relationship of arterial compliance with its cardiovascular sequelae, as well as the possible benefits of therapies specifically designed to prevent or treat the vascular pathology associated with elevated pulse pressure. Funding/Support: This study was supported in part by contract N01AG02107 from the National Institute on Aging, Bethesda, Md. Dr Chae was supported by training grant HL-07575 from the National Heart, Lung, and Blood Institute, Bethesda. Acknowledgment: The authors wish to thank Christine Albert, MD, for her thoughtful comments on the manuscript. REFERENCES 1. Congestive heart failure in the United States: a new epidemic [data fact sheet]. Bethesda, Md: National Heart, Lung, and Blood Institute Information Center; September 1996. Available at: http://www.nhlbi.nih.gov/nhlbi/cardio/other/gp/chf.htm. Accessed January 7, 1999. 2. Levy D, Larson MG, Vasan RS, Kannel WB, Ho KKL. The progression from hypertension to congestive heart failure: The Framingham Study. JAMA. 1996;275: 1557-1562. 3. Franklin SS, Gustin W IV, Wong ND, et al. Hemodynamic patterns of age-related changes in blood pressure. Circulation. 1997;96:308-315. 4. Smulyan H, Safar ME. Systolic blood pressure revisited. J Am Coll Cardiol. 1997;29:1407-1413. 5. O Rourke MF. Arterial Function in Health and Disease. Edinburgh, Scotland: Churchill Livingstone Inc; 1982. 6. Kelly R, Hayward C, Avolio A, O Rourke M. Noninvasive determination of age-related changes in the human arterial pulse. Circulation. 1989;80:1652-1659. 7. Milnor WR. Arterial impedance as ventricular afterload. Circ Res. 1975;36:565-570. 8. Urschel CW, Covell JW, Sonnenblick EH, Ross JR Jr, Braunwald E. Effects of decreased aortic compliance on performance of the left ventricle. Am J Physiol. 1968;214:298-304. 9. Kelly RP, Tunin R, Kass DA. Effect of reduced aortic compliance on cardiac efficiency and contractile function of in situ canine left ventricle. Circ Res. 1992; 71:490-502. 10. Kohno F, Kumada T, Kamabayashi M, et al. Change in aortic end-systolic pressure by alterations in loading sequence and its relation to left ventricular isovolumic relaxation. Circulation. 1996;93:2080-2087. 11. Watanabe H, Ohtsuka S, Kakihana M, Sugishita Y. Coronary circulation in dogs with an experimental decrease in aortic compliance. J Am Coll Cardiol. 1993; 21:1497-1506. 12. Kass DA, Saeki A, Tunin RS, Recchia FA. Adverse influence of systemic vascular stiffening on cardiac dysfunction and adaptation to coronary occlusion. Circulation. 1996;93:1533-1541. 13. Buckberg GD, Fixler DE, Archie JP, Hoffman JIE. Experimental subendocardial ischemia in dogs with normal coronary arteries. Circ Res. 1972;30:67-81. 14. Darne B, Girerd X, Safar M, Cambien F, Guize L. Pulsatile versus steady component of blood pressure. Hypertension. 1989;13:392-400. 15. Madhavan S, Ooi WL, Cohen H, Alderman MH. Relation of pulse pressure and blood pressure reduction to the incidence of myocardial infarction. Hypertension. 1994;23:395-401. 16. Fang J, Madhavan S, Cohen H, Alderman MH. Measures of blood pressure and myocardial infarction in treated hypertensive patients. J Hypertens. 1995; 13:413-419. 17. Benetos A, Safar M, Rudnichi A, et al. Pulse pressure: a predictor of long-term cardiovascular mortality in a French male population. Hypertension. 1997; 30:1410-1415. 18. Mitchell GF, Moye LA, Braunwald E, et al, for the SAVE Investigators. Sphygmomanometricallydetermined pulse pressure is a powerful independent predictor of recurrent events following myocardial infarction in patients with impaired left ventricular function. Circulation. 1997;96:4254-4260. 19. Girerd X, Laurent S, Pannier B, Asmar R, Safar M. Arterial distensibility and left ventricular hypertrophy in patients with sustained essential hypertension. Am Heart J. 1991;122:1210-1214. 20. Saba PS, Roman MJ, Pini R, Spitzer M, Ganau A, Devereux RB. Relation of arterial pressure waveform to left ventricular and carotid anatomy in normotensive subjects. J Am Coll Cardiol. 1993;22:1873-1880. 21. Hypertension Detection and Follow-up Program Cooperative Group. Blood pressure studies in 14 communities. JAMA. 1977;237:2385-2391. 22. Glynn RJ, Field TS, Satterfield S, et al. Modification of increasing systolic blood pressure in the elderly during the 1980s. Am J Epidemiol. 1993;138: 365-379. 23. Cornoni-Huntley J, Brock DB, Ostfeld AM, eds. Established Populations for Epidemiologic Studies of the Elderly: Resource Data Book. Bethesda, Md: National Institutes of Health; 1986. Publication 86-2443. 24. Franklin SS, Weber MA. Measuring hypertensive cardiovascular risk: the vascular overload concept. Am Heart J. 1994;128:793-803. 25. Kannel WB, Ho K, Thom T. Changing epidemiological features of cardiac failure. Br Heart J. 1994;72 (suppl):s3-s9. 26. Kostis JB, Davis BR, Cutler J, et al, for the SHEP Cooperative Research Group. Prevention of heart failure by antihypertensive drug treatment in older persons with isolated systolic hypertension. JAMA. 1997; 278:212-216. 27. Schlatmann TJM, Becker AE. Histologic changes in the normal aging aorta. Am J Cardiol. 1977;39:13-19. 28. Ting CT, Brin KP, Lin SJ, et al. Arterial hemodynamics in human hypertension. J Clin Invest. 1986; 78:1462-1471. 29. Gardin JM, Arnold A, Gottdiener JS, et al. Left ventricular mass in the elderly: the Cardiovascular Health Study. Hypertension. 1997;29:1095-1103. 30. Levy D, Garrison RJ, Savage DD, et al. Prognostic implications of echocardiographically determined left ventricular mass in the Framingham Heart Study. N Engl J Med. 1990;322:1561-1566. 31. Farrar DJ, Bond MG, Riley WA, Sawyer JK. Anatomic correlates of aortic pulse wave velocity and carotid artery elasticity during atherosclerosis progression and regression in monkeys. Circulation. 1991; 83:1754-1763. 32. Hirai T, Sasayama S, Kawasaki T, Yagi S. Stiffness of systemic arteries in patients with myocardial infarction. Circulation. 1989;80:78-86. 33. Witteman JC, Grobbee DE, Valkenburg HA, et al. J-shaped relation between change in diastolic blood pressure and progression of aortic atherosclerosis. Lancet. 1994;343:504-507. 34. Lyon RT, Runyon-Hass A, Davis HR, Glagov S, Zarins CK. Protection from atherosclerotic lesion formation by reduction of artery wall motion. J Vasc Surg. 1987;5:59-67. 35. Mitchell GF, Pfeffer MA, Finn PV, Pfeffer JM. Equipotent antihypertensive agents variously affect pulsatile hemodynamics and regression of cardiac hypertrophy in spontaneously hypertensive rats. Circulation. 1996;94:2923-2929. 36. Dzau VJ, Safar ME. Large conduit arteries in hypertension. Circulation. 1988;77:947-954. 37. Latson TW, Hunter WC, Katoh N, Sagawa K. Effect of nitroglycerin on aortic impedance, diameter and pulse-wave velocity. Circ Res. 1988;62:884-890. 38. Avolio AP, Clyde KM, Beard TC, et al. Improved arterial distensibility in normotensive subjects on a low salt diet. Arteriosclerosis. 1986;6:166-169. 39. Saeki A, Kass DA. Systolic flow augmentation in hearts ejecting into a model of stiff aging vasculature. Circ Res. 1995;76:132-141. 40. Fisher ES, Whaley FS, Krushat WM, et al. The accuracy of Medicare s hospital claims data: progress has been made, but problems remain. Am J Public Health. 1992;82:243-248. 41. Ayanian JZ, Weissman JS, Chasan-Taber S, Epstein AM. Quality of care for two common illnesses in teaching and nonteaching hospitals. Health Aff (Millwood). 1998;17:194-204. 1999 American Medical Association. All rights reserved. JAMA, February 17, 1999 Vol 281, No. 7 639