Stroke A Journal of Cerebral Circulation

Stroke A Journal of Cerebral Circulation JULY-AUGUST VOL. 7 1976 NO. 4 Components of Blood Pressure and Risk of Atherothrombotic Brain Infarction: The Framingham Study WILLIAM B. KANNEL, M.D., THOMAS R. DAWBER, M.D., PAUL SORLIE, M.S., AND PHILIP A. WOLF, M.D. SUMMARY From a study of the evolution of atherothrombotic brain infarction (ABI) in the Framingham cohort of 5,209 men and women over 18 years of follow-up, it has been ascertained that hypertension is the most common and most powerful precursor. Atherothrombotic brain infarction developed in hypertensive patients seven times more often than in normotensive patients, and the risk was proportional to the blood throughout its range. Various components of blood, including systolic and diastolic blood, pulse, lability of, mean arterial and tension-time index, were analyzed in relation to ABI incidence. While all measures were associated with ABI incidence, the simple casual systolic emerged as good a predictor of ABI incidence as any other component of the. The other measures added very little to risk. FROM AN EXAMINATION of various personal attributes and living habits believed to contribute to the occurrence of atherothrombotic brain infarction (ABI) at Framingham and elsewhere, blood has emerged as the most powerful precursor. 1 " 8 Hypertensive patients have been found to have ABIs at from five to more than 30 times the rate of normotensive persons depending upon age and sex (fig. 1). This warrants a detailed analysis of the various components of blood in relation to ABI incidence. To this end a multivariate analysis of the relation of ABI incidence in the Framingham cohort to various components of blood has been carried out. The parameters of blood examined in relation to ABI incidence in this report include: systolic and diastolic, pulse (systolic minus diastolic), mean arterial (one-third systolic + two-thirds diastolic), lability of systolic (difference between two systolic s) and tension-time index (systolic X heart. The objective is to compare each of these components of blood as predictors of ABI incidence. the examining physician and one by the nurse. Only the physician's blood determinations are used in this report. Heart rate was determined from the resting ECG. Only 2% of the subjects participating were completely lost to follow-up and 85% received every possible examination. The rest were examined at less frequent intervals. Admissions to the only hospital in town were monitored daily and most of the cases suspected of having a stroke were seen on the wards by a neurologist assigned to the study. All cases were reviewed by a panel of investigators to determine if the cases met criteria for a stroke. Diagnostic criteria for a stroke employed here have been reported in detail elsewhere. 4 Minimal criteria for an ABI consisted of the abrupt onset of a focal neurological deficit (e.g., hemiparesis, aphasia, homonymous hemianopia, hemisensory defects) in the absence of a bloody spinal fluid or a source for emboli from the heart. Statistical Analysis Analysis is restricted to the segment of the population aged 45 to 74, since too few ABIs occurred outside this age Methods range for meaningful analysis. Risk of ABI was determined This report is based on 5,209 men and women aged 30 to in relation to blood at the time of each biennial examination in relation to person-years' exposure at that age 62 at entry who participated in the Framingham Study. They received cardiovascular examinations biennially where and. 4 measurements of blood, daily tobacco use, weights, The net and joint effect of each parameter of blood ECG findings and a variety of biochemical tests, including was determined using the multiple logistic model cholesterol and glucose, were obtained. Blood was and employing the Duncan-Walker maximum likelihood determined in the left arm with the subject seated using a method of analysis." The logistic model assumes that the mercury sphygmomanometer. These s were recorded at each examination over a one-hour period, two by characteristics, X,, can be characterized by the equation: probability of an event, P, as a function of measured P = 1 -r 1 + exp (-a -2B,X,). Framingham Heart Study, 123 Lincoln Street, Framingham, Massachusetts 01701, and National Heart and Lung Institute, National The parameters of the equation, a and B,, are estimated Institutes of Health, 7910 Woodmont Avenue, Bethesda, Maryland 20014. from the weighted least squares procedure of Duncan and 327

328 STROKE VOL. 7, No. 4, JULY-AUGUST 1976 Normotensive ^fl Borderline Women predominance (fig. 1), except for premature strokes under age 55. At any age in either sex the incidence was substantially greater among hypertensives than normotensives (fig. 1). Furthermore, risk was proportional to blood, from the lowest to the highest values recorded (fig. 2). No critical value of blood demarcates the ABI candidate from the general population. From our examination of ABI risk gradients in relation to blood on a logarithmic scale, a uniform increment in risk is perceived at either the low or high end of the range. For each 10- mm increase in the risk of an ABI increases about 65-64 66-74 45-64 56-64 06-74 FIGURE 1. A verage annual incidence rate for atherothrombotic brain infarction by hypertensive status: Framingham study, 18-year follow-up. Walker. The regression coefficient, B,, affords a measure of the association between increased risk of the event and levels of the characteristic. Since the magnitude of the coefficients is affected by the variance of the characteristic, standardized coefficients have been presented (coefficient X standard deviation of characteristic) to put all coefficients on the same scale. Results In the age group 45 to 74, 31 men and 46 women had ABIs which met criteria over the 18 years of follow-up. The incidence of ABI rose with age without a distinct male 100 80 60 40 20 10 74-109 1120-129 140-149 1160-169 1180-189 110-119 130-139 150-159 170-179 190-300 Systolic Blood Pressure FIGURE 2. Annual incidence rale of ABI per 10,000, men and women 45 to 74, crude rates on log scale by systolic blood ; Framingham heart study, 18-year follow-up. Influence of Components of Blood Pressure Various components of the blood may contribute to stroke incidence including systolic and diastolic, pulse, mean arterial, the lability of the TABLE la Standardized Coefficients* for Regression of ABI Years, 45 to 74- Framingham Study: 18-Year Follow-Up Univariate standardized regression coefficients Systolic blood Diastolic blood Pulse Mean arterial (1/3 0.82 It 0.646t 0.622t 0.775t SBP + 2/3 DBP) Lability of systolic 0.267 (SBP'-SBP 2 ) Tension-time index 0.671t 4,708/8 0.996t 1.054t 0.772t l.uot 0.250 0.904 0.567t 0.247 0.605t 0.442 0.311 0.508 1T - > 2.00 0.829t 0.712t 0.674t 0.816t 0.279t 0.706t TABLE lb Standardized Coefficients* for Regression of ABI Years, Women 45 to 74- Framingham Study: 18-Year Follow-Up Univariate standardized regression coefficients Systolic blood 0.784f 0.680f 0.534f 0.663f Diastolic blood 0.869f 0.650f 0.673f 0.709f Pulse 0.641f 0.570f 0.276 0.500f Mean arterial (1/3 0.856f 0.699f 0.668f 0.728f SBP + 2/3 DBP) Lability of systolic 0.536f 0.110 0.346f 0.349f (SBP'-SBP 2 ) Tension-time index 0.646f 0.578f 0.358f 0.512f 5,858/6 4,559/18 2,092/22 tt - > 2.00.

FRAMINGHAM STUDY/Kannel et al. 329 and the tension-time index. A multivariate analysis was undertaken to examine the net contribution of each of these components of to ABI incidence using the method of Duncan-Walker. Used as a single variable, any one of these components of the can be used effectively to predict ABI incidence based on the coefficients averaged over all ages (table 1 a and b). However, within certain age-sex groups, particularly the older age group, some of these coefficients are not significantly different from zero. The fact that all components of contribute to risk of ABI is very likely attributable to their high correlation with each other. The magnitude of the standardized coefficients can give an indication of the relative contribution of each component. In univariate analysis for all ages combined, the coefficients for systolic are largest in men with mean arterial running a close second in each sex (table la). In women, mean arterial is foremost (table lb). Lability of contributes the least. Bivariate logistic regression analysis comparing the con- TABLE 2a Standardized Coefficients* for Regression of ABI Years, 46 to 74- Framingham Study: 18-Year Follow-Up Bivariate standardized regression coefficients Bivariate independent variables Systolic blood 0.700 1.582f 0.160 0.857f Pulse 0.135-0.623 0.463-0.037 Systolic BP 1 + 0.801f 1.214f 0.482 0.856f systolic BP 2 Lability systolic BP 0.080-0.135 0.199 0.045 (SBpi-SBP 2 ) Systolic blood 0.716 0.769f 0.368 0.648f Tension-time index 0.123 0.345 0.247 0.253 4,708/8 tt - > 2.00. TABLE 2b Standardized Coefficients* for Regression of ABI Years, Women 45 to 74. Framingham Study: 18-Year Follow-Up Bivariate standardized regression coefficients Bivariate independent variables Systolic blood 1.218f 0.971f 1.325f 1.171f Pulse -0.463-0.327-0.902f -0.597f Systolic BP 1 + 0.636f 0.726 0.490f 0.622f systolic BP 2 Lability systolic BP 0.293-0.068 0.279 0.192 (SBPi-SBP 2 ) Systolic blood 1.117f 0.598f 0.603f 0.679f Tension-time -0.405 0.107-0.091-0.047 index (SBP X heart 5,858/6 4,559/18 2,092/22 tt =- > 2.00. tribution of other specified components of blood when systolic is taken into account reveals no additional effect of lability, pulse or tension-time index. The coefficients for systolic blood are positive and significantly different from zero, while the coefficients for the other components of blood are small or in one case negative (table 2a and b). The high correlation of these variables makes it difficult to interpret the bivariate regression coefficients. A more appropriate statistic is the likelihood ratio statistic which gives a measure of overall regression. In the univariate case this statistic is distributed chi-square with one degree of freedom. These statistics reveal that overall systolic is most closely related to ABI risk with mean arterial running a close second in men. In women, particularly, mean arterial and systolic appear to be equivalent predictors of ABI incidence (table 3a and b). The contribution to regression of the other components of blood on ABI incidence above the contribution of systolic can be measured by subtracting the likelihood ratio of the univariate (systolic as the independent variable) from the likelihood ratio of the bivariate (systolic + other specified components of ). This difference is a chi-square of one degree of freedom. In all the bivariate analyses the likelihood ratio statistic is only slightly larger than the statistic for univariate systolic blood alone, indicating no additional contribution to risk (table 4). While the chi-square for lability may differ at all ages from those of systolic alone, the increment is not large and, because of the large number of comparisons, cannot be assigned much importance. TABLE 3a Risk of ABI According to Various Components of Blood Pressure. Likelihood Ratio Statistic for Logistic Function* Univariale. 45 to 74- Framingham Study: 18-Year Follow-Up Systolic blood Diastolic blood Pulse Mean arterial Lability of systolic Tension-time index Chi-equare 1 degree of freedom 9.96 4.64 7.47 7.84 0.84 7.30 4,708/8 22.56 19.65 13.32 24.26 1.04 17.08 4.80 0.75 5.53 2.72 1.73 4.35 Duncan-Walker evaluation of logistic function parameters. 1 TABLE 3b Risk of ABI According to Various Components of Blood Pressure. Likelihood Ratio Statistic for Logistic Function* Univariale. Women 45 to 74- Framingham Study: 18-Year Follow-Up Systolic blood Diastolic blood Pulse Mean arterial Lability of systolic Tension-time index Chi-square 10.05 9.57 6.64 10.66 4.89 5.38 5,858/6 1 degree of freedom 13.11 7.62 10.87 12.08 8.64 1.94 13.46 11.86 0.24 4.41 9.12 3.51 4,559/18 2,092/22

330 STROKE VOL. 7, No. 4, JULY-AUGUST 1976 TABLE 4 Risk of ABI According to Various Components of Blood Pressure: Framingham Study: 18-Year Follow- Up. Likelihood Ratio Statistic for Logistic Function Bivariate Analysis. and Women 45 to 74 Systolic blood and pulse 10.06 SBP 1 and SBP 2 and lability SBP 10.41 Systolic blood and tension-time index 10.06 Systolic blood alone 9.96 4,708/8 24.46 28.19 23.68 22.56 Chi-square 2 degrees of freedom 5.63 5.00 5.24 10.68 10.34 Chi-square 1 degree of freedom 4.80 10.63 10.05 5,858/6 Women. 13.68 14.66 13.24 13.11 4,559/18 12.55 10.73 7.71 7.62 2,092/22 From this analysis it seems clear that the simple casual systolic is as good a predictor of ABI incidence as any other component of. Because it is more accurately determined than diastolic and yields a wider range of values, systolic is the preferred parameter for evaluating risk of ABI. Other derivations such as pulse, mean arterial and tensiontime index add little to risk or its assessment. Systolic Pressure Since systolic emerges as the best and most direct predictor of ABI incidence, a detailed look at its relation to ABI is warranted. Casual Systolic Blood Pressure The foregoing analysis is based upon casual blood s obtained during a one-hour cardiovascular examination. These casual s are strongly predictive of ABI incidence. It might be argued that only those whose casual s are fixed at an elevated level are at any increased risk and that those whose s are labile are not. To examine this, those where three s did not vary more than 20 mm Hg during the one-hour examination were excluded from the analysis and the risk assessed according to the highest value recorded. In both men and women this maximum SBP in persons with very labile s during an examination is a significant contributor to risk of ABI with standardized coefficients for the systolic of 0.8740 and 0.5495 respectively. These are significantly different from zero (p < 0.01). It is thus apparent that even such labile casual blood values are distinctly related to the risk of ABI, the risk proportional to the casual blood level. Discussion Blood is without doubt the most powerful contributor to ABI incidence. This is widely accepted but most consider the diastolic component to be the culprit. The filtration hypothesis of atheroma formation would suggest that the rate of cerebral atherogenesis might be a product of the mean arterial. If hemodynamic injury to, or flow through, sclerotic vessels is involved as the mechanism then pulse or lability of the systolic might be more important. Impaired cardiac function has been shown to contribute to stroke in persons with a compromised cerebral circulation as evidenced by CHD, congestive failure and ECG by LVH. 3 The work load of the left ventricle is reflected better by the modified tension-time index than by any other parameter of blood. 6 The data presented suggest that, considering the net and joint effects of all these parameters of blood, the risk of development of an ABI is most closely linked to the systolic or mean arterial. However, because of the high correlation of these components of blood and their lack of independence one from the other, it is difficult to dissociate their effects statistically, or for that matter experimentally. The data do indicate, however, that the casual systolic obtained in the office is highly predictive of subsequent stroke incidence. No other component of this casual blood can improve the estimate of risk so obtained. This is especially important since the key to the prevention of strokes is the early detection and control of elevated blood. Some of the generally held notions about the details of the relation of hypertension to stroke appear to be misconceptions. In the clinical evaluation of hypertension the degree and nature of the blood elevation are considered relevant in assessing its pragmatic importance. Casual systolic elevations are considered misleading and their treatment misguided. The data herein reported suggest that this is definitely not the case and that, at the very least, such persons should be closely followed rather than dismissed. When accompanied by other contributors, even a moderate, casual systolic elevation can be associated with a sizable risk of ABI. 7 Some degree of hypertension is an almost ubiquitous finding in the background of stroke victims. After 18 years in the Framingham cohort only 10 of 105 brain infarctions occurred in those with s under 140/90 mm Hg, where 42 would be expected. The fear that reduction of systemic arterial blood will somehow interfere with cerebral blood flow in persons with cerebral atherosclerosis seems unfounded. Normally, autoregulation maintains a constant cerebral blood flow until a mean arterial of 120 mm Hg is exceeded, at which time excessive autoregulation may take place and cerebral vasospasm occurs. Reduction of blood in such persons improves the cerebral circulation to the ischemic brain by reducing cerebral vasospasm. Trials of antihypertensive therapy in hypertensive stroke patients have shown an improvement rather than a reduction in cerebral blood flow and a concomitant clinical improvement. 8 The VA studies 9 have shown that, in the balance,

FRAMINGHAM STUDY/Kannel et al. 331 TABLE 5 Percent of ABI Cases Evolving From Segment of Population at Risk With Hypertension* by Age and Sex. and Women 45 to 74- Framingham Cohort 18-Year Follow-Up Age % i of pop. with (year) HBP Total (all ages) 17.1 21.9 23.1 19.7 *Blood > 160/95. % of ABI cases 41.7 65.2 42.9 53.1 % of pop. with HBP 15.0 27.7 36.6 22.8 Wormin % of ABI cases 44.4 60.9 62.5 58.9 hypertensive persons under treatment have fewer, and not more, strokes. While reduction in stroke recurrence following antihypertensive therapy has not been found in all studies, 10 none have shown an increased risk. Blood determinations are an efficient means for identifying ABI candidates even when used without any other information. Using some arbitrary designation of "hypertension" as a criterion (i.e., > 160/95 mm Hg) will select about a fifth of the population from which 50% to 60% of the ABIs will emerge (table 5). However, in screening programs using this kind of categorical criterion for "hypertension," the phenomenon of regression toward the mean is of some concern. Later measurements by the physician on the referred will fall closer to the mean of the population screened than the earlier ones. In addition, the circumstances under which the is measured has changed, and the physician often settles on the lowest he can obtain as the true one. For these reasons as many as 35% of those characterized as definitely "hypertensive" will be borderline or normal on reassessment. Although these people are at increased risk of progression to sustained hypertension and of having an ABI they will often be rejected by the physician to whom the case is referred. Thus, from a practical point of view it may be necessary for screening programs to use the lowest of several recorded s or a secondary screening as the basis for referrals. References 1. Kannel WB, Wolf PA, Verter J, et al: Epidemiologic assessment of the role of blood in stroke: The Framingham study. JAMA 214:301-310, 1970 2. Kannel WB, Gordon T, Wolf PA: Hemoglobin and the risk of cerebral infarction: The Framingham study. Stroke 3:409-420, 1972 3. Wolf PA, Kannel WB, McNamara PM, et al: The role of impaired cardiac function in atherothrombotic brain infarction: The Framingham study. Am J Public Health 63:52-58, 1973 4. Shurtleff D: Some characteristics related to the incidence of cardiovascular disease and death: Framingham study, 18-year follow-up. In The Framingham Study, An Epidemiological Investigation of Cardiovascular Disease, Section 30. Washington, D.C., U.S. Government Printing Office, 1974 5. Walker SH, Duncan DB: Estimation of the probability of an event as a function of several independent variables. Biometrika 54:167-172, 1967 6. Sarnoff SJ, Braunwald E, Welch GH Jr, et al: Hemodynamic determinants of oxygen consumption of the heart with special reference to the tension-time index. Am J Physiol 192:148-156, 1968 7. McGee D: The probability of developing certain cardiovascular diseases in eight years at specified values of some characteristics. In The Framingham Study, An Epidemiological Investigation of Cardiovascular Disease, Section 28. Washington, D.C., U.S. Government Printing Office, 1973 '8. Meyer JS, Sawada T, Kitamura A, et al: Cerebral blood flow after control of hypertension in stroke. Neurology 18:772-781, 1968 9. Freis ED: Veterans Administration Cooperative Study on Antihypertensive Agents: Effects of treatment on morbidity and hypertension. JAMA 202:1028-1034, 1970 _10. Hypertensive-Stroke Cooperative Study Group: Effect of antihypertensive treatment on stroke recurrence. JAMA 229:409-418, 1974