点击显示 收起
the H.N. Neufeld Cardiac Research Institute (D.W., U.G.), Sheba Medical Center, Tel Hashomer, Israel
Department of Epidemiology and Preventive Medicine (U.G.), Tel Aviv University, Israel.
Abstract
Background and Purpose— Studies of blood pressure (BP) indices as disease predictors have offered conflicting conclusions. We compare pulse pressure (PP), systolic BP (SBP) and diastolic BP (DBP), and mean arterial pressure (MAP) as risk markers for long-term mortality with emphasis on stroke.
Methods— Male civil servants (40 to 65 years of age; n=9611) were examined in 1963 and followed up until 1986. Multivariate analysis was used to assess the association between BP indices and subsequent mortality. Stroke mortality was analyzed separately for initially normotensive (SBP140 and DBP90 mm Hg), hypertensive (SBP>140 and DBP>90 mm Hg), and men with isolated systolic hypertension (ISH; SBP>140 and DBP90 mm Hg).
Results— During follow-up, 3167 men died, including 932 of coronary heart disease and 339 of stroke. All 4 BP indices were predictive of fatal stroke among hypertensive and normotensive men (hazard ratios fluctuated between 1.59 and 2.51). In models with 2 BP indices among normotensive men, SBP but not DBP remained a predictor of stroke mortality. MAP and PP were independent predictors of stroke mortality. Among hypertensive men, SBP and DBP were independent predictors of stroke mortality (HRs, 1.68 and 1.51, respectively). MAP but not PP remained a predictor of stroke mortality. In men with ISH, the 4 BP indices predicted fatal stroke, with HRs fluctuating between 1.24 and 2.04.
Conclusions— All 4 BP indices were predictors of stroke mortality among hypertensive and normotensive men, with DBP possibly the weaker predictor among the latter. Models with 2 BP indices yielded complex associations.
Key Words: blood pressure coronary heart disease mortality prospective studies stroke
Introduction
Systolic blood pressure (BP) is widely recognized as a major contributor to the risk of vascular and other diseases, in particular stroke. Among the elderly, 65% of hypertension occurs in the form of isolated systolic hypertension (ISH).1 ISH or increased pulse pressure (PP; the difference between systolic BP and diastolic BP ) represent mainly large arterial stiffening.2 Although several studies have shown the strong association of ISH or PP with increased stroke and cardiovascular morbidity and mortality,3–9 the value of PP as an independent risk factor remains controversial. The current study aimed to compare SBP, DBP, PP, and mean arterial pressure (MAP; defined as 1/3x(SBP+2DBP)) as risk markers for mortality with an emphasis on stroke mortality in working men examined in 1963 and followed up over 23 years.
Materials and Methods
In 1963, 10 059 men 40 to 65 years of age, tenured civil servants and municipal employees, were examined in the framework of the Israeli Ischemic Heart Disease study.10,11 They constituted 84.7% of 11 876 originally sampled in a random stratified sample of 6 predefined geographical areas. Participants underwent clinical and blood biochemical evaluations in 1963, 1965, and 1968. The measurement of BP was performed in the recumbent position on the right arm. SBP was measured at the onset of Korotkoff sounds, and DBP was measured on their disappearance. The second of 2 BP recordings, undertaken 45 to 75 minutes after arrival, is used herein. Preference of 0 end-digits in BP recordings is evident (40% and 41% of SBP and DBP recordings, respectively; supplemental Figure I, available online at http://stroke.ahajournals.org).
Men with a previous heart attack or definite angina10,11 at baseline (n=448) were excluded from analysis. Overall, 9611 men were included in analysis. The underlying cause of death was determined on a case-by-case basis by a review panel until 1970 and by the use of the International Classification of Disease (ICD) codes thereafter. Deaths from presumed stroke were based on ICD codes 430 to 438. Information on mortality after 1970 was derived from the Israeli Mortality Registry, which is virtually complete for the last 35 years in terms of death reports. The likelihood of migration among Israeli men holding a tenured civil servant position during the 1960s and 1970s is extremely low, given hardships in entering countries where they would presumably seek another future, the sociopolitical environment, their family ties, and their advancing ages. We assume loss to mortality follow-up is negligible.
Regression Dilution
The estimated association between BP indices and subsequent mortality was corrected for regression dilution bias (RDB).12 We used the 1968 BP values, arranged by quintiles as assessed in 1963, to derive an estimate of the "shrinkage" of the distance between mean BP levels of the upper and lower quintiles on re-examination. The ratio of interquintile range in 1963 to that of 1968 yields an estimated regression dilution factor (RDF; supplemental Table I, available online).
Statistical Analysis
The standard population for direct age-adjustment was based on the distribution of the entire study sample arranged by 5-year age groups. A linear trend of age-adjusted mortality rates by quintiles of BP indices was tested by the Mantel extension test, using quintile medians as scores. Adjusted hazards associated with 1 SD increments of BP indices were estimated using the Cox proportional hazards model. Goodness of fit was assessed by comparing actual to predicted quintiles of risk.13 The assumption of proportional hazards, examined on the basis of Schoenfeld residuals,14 was violated for DBP and SBP in models predicting stroke mortality. For this reason, these models were stratified into: (1) both BPs elevated (SBP>140 and DBP>90 mm Hg), termed "hypertensive" throughout this article; (2) normotensive (SBP140 and DBP90 mm Hg); (3) ISH (SBP>140 and DBP90 mm Hg); and (4) isolated diastolic hypertensive (SBP140 and DBP>90 mm Hg). Group 4 was too small (n=373; 11 deaths) for meaningful analysis. Covariates11 included age, diabetes mellitus, cigarette smoking, serum cholesterol, and socioeconomic status (SES).15 We used the likelihood ratio test to screen for quadratic associations between BP indices and stroke mortality.
Akaike’s information criterion (AIC) was used to compare the fit of models16 by subtracting the AIC value of a model with BP component(s) from the same model including only covariates other than BP component(s). A larger decrease in AIC can be interpreted as a better fit. Statistical analysis was done using Stata version 8.0 (StataCorp LP).
Results
Means (SDs) of BP indices in 1963 were (mm Hg): 134.8 (20.3), 83.7 (11.0), 100.8 (13.2), and 51.1 (14.2) for SBP, DBP, MAP, and PP, respectively. During 23 years of follow-up, 3167 men died, including 932 of coronary heart disease (CHD), 339 of stroke, 674 of cancer, and 1222 of other causes. The means of all BP indices were higher among those who died during follow-up than in survivors. The average SBP, adjusted for age, among those who died during follow-up was 8.4 mm Hg higher than among survivors (mean±SDs of 140.0±22.0 versus 131.6±16.3 mm Hg, respectively), the average DBP was higher by 4.0 mm Hg (86.4±12.4 versus 82.4±9.0 mm Hg), and the average PP and MAP were higher by 4.2 mm Hg (53.6±14.8 versus 49.4±11.7 mm Hg) and 7.4 mm Hg (105.7±15.2 versus 98.3±11.3 mm Hg), respectively.
The mean age and age-adjusted mortality rates within quintiles of the BP indices are presented in Table 1. The RDFs, calculated as the ratios of interquintile differences in 1963 to those of 1968, were 1.242, 1.539, 1.328, and 1.449 for SBP, DBP, MAP, and PP, respectively.
Figure 1 shows age-adjusted all-cause mortality rates in quintiles of SBP and DBP. Mortality rates increase linearly across SBP quintiles, whereas no linearity is seen across DBP quintiles.
The strongest associations between BP indices and age-adjusted mortality were found for stroke mortality (Table 1). A 4.8-fold increase in risk for stroke mortality was found comparing the lowest and highest SBP quintiles. A 2.9-fold risk was found for the corresponding quintiles of DBP. PP was associated with a 4.6-, 2.3-, and 1.9-fold increase in age-adjusted stroke, CHD, and all-cause mortality rates, respectively (P0.015 for linear trend for BP indices and all causes of mortality).
There were 130 (2.0%), 125 (8.6%), and 73 (5.7%) stroke deaths among normotensive, hypertensive, and ISH subjects respectively. Mean values and SDs of BP components in each group are given in Table 2. Hazard ratios (HRs) for stroke mortality are shown in Table 3. We present 4 models, each incorporating a single BP index. Adjusting for age, diabetes, smoking, SES, and correcting for RDB, all 4 BP indices were predictive of fatal stroke among hypertensive and normotensive men. Among the latter, DBP could be the more weakly associated parameter (HR, 1.59; 95% CI, 0.92 to 2.75 per 1-SD increment). In alternative models, including 2 BP indices, DBP was weakened in a model controlling for SBP. MAP and PP remained predictors of stroke mortality among normotensive men. In hypertensive men, SBP and DBP were independent predictors of stroke mortality (HRs, 1.68 and 1.51; 95% CI, 1.24 to 2.29 and 1.05 to 2.18). MAP, but not PP, remained a predictor of stroke mortality. In patients with ISH, the HRs for all the BP components were >1 for stroke mortality but fell short of a 0.05 type-1 error yardstick. An additional analysis excluding 344 men who received antihypertensive medication in 1968 revealed associations consistent with a similar pattern, although this group experienced an extremely high death rate (66% died, 11.6% of stroke, during follow-up). Models with either single index of BP or 2 indices did not differ significantly in their fit as reflected by changes in AIC values between models including or excluding BP components (Table 3).
Among the entire study sample, goodness-of fit was improved in models incorporating linear as well as quadratic terms of SBP, DBP, or PP versus models with linear terms only (SBP 2=3.60, P=0.058; DBP 2=3.32, P=0.069; PP 2=9.09, P=0.003). This did not hold true for MAP (2=0; P=0.96).
All 4 BP indices were predictive of CHD and all-cause mortality (Table 4). In models with 2 BP indices, SBP and DBP remained predictors of mortality. MAP but not PP remained a predictor of mortality. Finally, Figure 2 presents HRs for the association of each BP index with stroke, CHD, and all-cause mortality stratified by 5 age groups. The HRs for all-cause mortality generally declined with increasing age. CIs for stroke mortality were relatively large.
Discussion
Strong associations between indices of BP and stroke morbidity and mortality have been reported. In the current study, all 4 BP indices were predictive of stroke mortality among hypertensive and normotensive men. Among the latter, DBP was the more weakly associated parameter. Among normotensive men, DBP was a weak predictor of stroke mortality in a model controlling for SBP. Among hypertensive men, PP was not a predictor in a model controlling for MAP. In patients with ISH, the HRs for all the BP components were >1 for stroke mortality but fell short of a 0.05 type-1 error, possibly because of the small size of this group. In contrast with previous studies,17,18 we found evidence for nonlinearity of the association between SBP, DBP, or PP, and stroke mortality. MAP presented linear associations with stroke mortality. The HRs for the association of all 4 BP indices with all-cause mortality generally declined with increasing age.
Elderly subjects17 showed 34% and 21% increased risk for stroke mortality per 1-SD change in SBP and PP, respectively. A stronger association (HR, 1.61) was reported for PP from the Corfu cohort of the Seven-Countries Study.19 Mazza et al20 found an HR of 1.50 associated with a PP of 74 mm Hg among elderly males and females. Among Chinese steel workers,21 an adjusted HR of 13.3 for stroke in the upper SBP quartile compared with the lowest quartile was higher than that found for the corresponding quartiles of DBP, MAP, and PP.
In the current study, all 4 BP indices were predictive of CHD and all-cause mortality. SBP and DBP were independent predictors of mortality. Controlling for MAP, PP was not a predictor of mortality. Strandberg et al22 reported 32-year CHD mortality HRs of 1.48, 1.47, 1.41, and 1.23 per 1-SD change in MAP, DBP, SBP, and PP, respectively. HRs for CHD of 1.38, 1.35, and 1.14 for PP, SBP, and DBP, respectively, were reported from the Framingham Study.23 Lee et al24 concluded that PP was a more accurate predictor of 29-year cardiovascular mortality than SBP or DBP among the elderly (60) but not among younger males. Casiglia et al25 found that PP predicted coronary mortality only among elderly women.
A number of studies examined the informativeness of various BP indices. Our goodness-of-fit findings show no preference for either BP index in the prediction of stroke mortality. MAP is slightly more informative in the prediction of CHD and all-cause mortality in single BP index models and equal to models with DBP and SBP or PP. Franklin et al23 and Millar et al3 showed small differences in model fit, predicting CHD by SBP versus PP. In the meta-analysis performed by the Prospective Studies Collaboration,26 PP was found less informative in the prediction of IHD and stroke mortality than SBP, DBP, or MAP. Pastor-Barriuso et al27 showed complex associations of PP with all-cause and cardiovascular mortality, depending on age, SBP, and DBP and discourage the use of PP for diagnostic or therapeutic decisions.
In the current study, subjects were last examined 5 years after baseline. We have information regarding use of antihypertensive medication in 1968 and none about BP changes, use of medication, and disease onset during the ensuing 18 years. Preference of 0 end-digits results in the misclassification of subjects with recordings at cutoff points of DBP of 90 mm Hg or SBP of 140 mm Hg and increases the inaccuracy of associations reported herein regarding prediction of stroke mortality. However, if the rounding error to the nearest 0 is random, the estimates of the reported associations could only be reduced compared with the true associations. Another limitation is the absence of women. Conversely, our mortality follow-up yielded a large number of deaths. In addition, repeated measurement of BP allowed for calculating RDFs and correcting the HRs for the latter. This correction increased the estimated associations.
Summary
The current study suggests that SBP, DBP, PP, and MAP are associated with CHD and all-cause mortality, as well as with stroke mortality, among hypertensive and normotensive men. Among the latter, DBP could be the more weakly associated parameter. Controlling for SBP, DBP was predictive of stroke mortality among hypertensive but not among normotensive men. Controlling for MAP, PP was predictive of stroke mortality among normotensive but not among hypertensive men.
References
Kannel WB. Prevalence and implications of uncontrolled systolic hypertension. Drugs Aging. 2003; 20: 277–286.
Smulyan H, Safar ME. The diastolic blood pressure in systolic hypertension. Ann Intern Med. 2000; 132: 233–237.
Millar JA, Lever AF, Burke V. Pulse pressure as a risk factor for cardiovascular events in the MRC Mild Hypertension Trial. J Hypertens. 1999; 17: 1065–1072.
Nawrot TS, Staessen JA, Thijs L, Fagard RH, Tikhonoff V, Wang JG, Franklin SS. Should pulse pressure become part of the Framingham risk score J Hum Hypertens. 2004; 18: 279–286.
Benetos A. Pulse pressure and cardiovascular risk. J Hypertens. 1999; 17 (suppl 5): S21–S24.
Gasowski J, Fagard RH, Staessen JA, Grodzicki T, Pocock S, Boutitie F, Gueyffier F, Boissel JP. Pulsatile blood pressure component as predictor of mortality in hypertension: a meta-analysis of clinical trial control groups. J Hypertens. 2002; 20: 145–151.
Mallion JM, Hamici L, Chatellier G, Lang T, Plouin PF, De Gaudemaris R. Isolated systolic hypertension: data on a cohort of young subjects from a French working population (IHPAF). J Hum Hypertens. 2003; 17: 93–100.
Staessen JA, Thijs L, O’Brien ET, Bulpitt CJ, de Leeuw PW, Fagard RH, Nachev C, Palatini P, Parati G, Tuomilehto J, Webster J, Safar ME. Ambulatory pulse pressure as predictor of outcome in older patients with systolic hypertension. Am J Hypertens. 2002; 15: 835–843.
Vaccarino V, Berger AK, Abramson J, Black HR, Setaro JF, Davey JA, Krumholz HM. Pulse pressure and risk of cardiovascular events in the systolic hypertension in the elderly program. Am J Cardiol. 2001; 88: 980–986.
Groen JJ, Medalie JH, Neufeld HN, Riss E, Bachrach CA, Mount FW, Smith H. An epidemiologic investigation of hypertension and ischemic heart disease within a defined segment of the adult male population of Israel. Isr J Med Sci. 1968; 4: 177–194.
Goldbourt U, Yaari S, Medalie JH. Factors predictive of long-term coronary heart disease mortality among 10 059 male Israeli civil servants and municipal employees. A 23-year mortality follow-up in the Israeli Ischemic Heart Disease Study. Cardiology. 1993; 82: 100–121.
Clarke R, Shipley M, Lewington S, Youngman L, Collins R, Marmot M, Peto R. Underestimation of risk associations due to regression dilution in long-term follow-up of prospective studies. Am J Epidemiol. 1999; 150: 341–353.
Hosmer DW Jr, Lemeshow S. Applied Survival Analysis. New York, NY: John Wiley & Sons; 1999: 225–230.
Schoenfeld D. Residuals for the proportional hazards regression model. Biometrika. 1982; 69: 239–241.
Eaton CB, Medalie JH, Flocke SA, Zyzanski SJ, Yaari S, Goldbourt U. Self-reported physical activity predicts long-term coronary heart disease and all-cause mortalities. Twenty-one-year follow-up of the Israeli Ischemic Heart Disease Study. Arch Fam Med. 1995; 4: 323–329.
Akaike H. A new look at statistical model identification. IEEE Trans Auto Control. 1974; 19: 716–723.
Psaty BM, Furberg CD, Kuller LH, Cushman M, Savage PJ, Levine D, O’Leary DH, Bryan RN, Anderson M, Lumley T. Association between blood pressure level and the risk of myocardial infarction, stroke, and total mortality: the cardiovascular health study. Arch Intern Med. 2001; 161: 1183–1192.
Haheim LL, Holme I, Hjermann I, Leren P. Risk of fatal stroke according to blood pressure level: an 18-year follow-up of the Oslo Study. J Hypertens. 1995; 13: 909–913.
Panagiotakos DB, Chrysohoou C, Pitsavos C, Menotti A, Dontas A, Skoumas J, Stefanadis C, Toutouzas P. Risk factors of stroke mortality: a 40-year follow-up of the Corfu cohort from the Seven-Countries Study. Neuroepidemiology. 2003; 22: 332–338.
Mazza A, Pessina AC, Pavei A, Scarpa R, Tikhonoff V, Casiglia E. Predictors of stroke mortality in elderly people from the general population. The Cardiovascular Study in the Elderly. Eur J Epidemiol. 2001; 17: 1097–1104.
Zhang XF, Attia J, D’Este C, Yu XH. Prevalence and magnitude of classical risk factors for stroke in a cohort of 5092 Chinese steelworkers over 13.5 years of follow-up. Stroke. 2004; 35: 1052–1056.
Strandberg TE, Salomaa VV, Vanhanen HT, Pitkala K, Miettinen TA. Isolated diastolic hypertension, pulse pressure, and mean arterial pressure as predictors of mortality during a follow-up of up to 32 years. J Hypertens. 2002; 20: 399–404.
Franklin SS, Khan SA, Wong ND, Larson MG, Levy D. Is pulse pressure useful in predicting risk for coronary heart disease The Framingham Heart Study. Circulation. 1999; 100: 354–360.
Lee ML, Rosner BA, Weiss ST. Relationship of blood pressure to cardiovascular death: the effects of pulse pressure in the elderly. Ann Epidemiol. 1999; 9: 101–107.
Casiglia E, Tikhonoff V, Mazza A, Piccoli A, Pessina AC. Pulse pressure and coronary mortality in elderly men and women from general population. J Hum Hypertens. 2002; 16: 611–620.
Lewington S, Clarke R, Qizilbash N, Peto R, Collins R. Age-specific relevance of usual blood pressure to vascular mortality: a meta-analysis of individual data for one million adults in 61 prospective studies. Lancet. 2002; 360: 1903–1913.
Pastor-Barriuso R, Banegas JR, Damian J, Appel LJ, Guallar E. Systolic blood pressure, diastolic blood pressure, and pulse pressure: an evaluation of their joint effect on mortality. Ann Intern Med. 2003; 139: 731–739.