Secular Trends in Long-term Sustained Hypertension, Long-term Treatment, and Cardiovascular Mortality
The Framingham Heart Study 1950 to 1990
Background Cardiovascular morbidity and mortality result from the chronic processes involved in hypertension. However, long-term sustained (LTS) hypertension has received little attention.
Methods and Results Trends in the prevalence of LTS hypertension and its treatment were assessed in 1950, 1960, and 1970 among three cohorts of men and women in the Framingham Heart Study (Mantel-Haenszel test). Cardiovascular disease (CVD) incidence and mortality were compared between patients with LTS hypertension with and without long-term treatment by use of the χ2 test. Cox proportional hazards regression analysis was used to estimate 10-year risk of death as a function of risk factor levels and treatment. Prevalence of LTS hypertension rose from 138 to 208 per 1000 between the 1950 and 1970 male cohorts (P<.01), while prevalence fell from 253 to 198 per 1000 between the female cohorts (P<.02). Long-term treatment increased 51% between the male cohorts and 45% between the female cohorts (both P<.001). While CVD incidence was similar (26% versus 25%), all-cause mortality was significantly lower among men with long-term treatment (31% versus 43%; P<.05), and CVD mortality was less than half (13% versus 28%; P<.01). Among treated women, all-cause mortality was 21% (versus 34%; P<.01), and CVD mortality was 9% (versus 19%; P<.01). Ten-year risk of CVD death for patients with LTS hypertension with long-term treatment compared with those without was 0.40 (95% CI, 0.27 to 0.60).
Conclusions This investigation of LTS hypertension, its treatment, and its sequelae in a free-living general population confirms the reduction in CVD mortality demonstrated in more short-term clinical trials of hypertension therapy in select patient groups.
As one of the most prevalent vascular diseases in the world, hypertension poses a major public health problem. It affects more than 50 million people,1 contributing to more than 0.5 million strokes and 1.25 million coronary attacks per year in the United States alone.2 3 In 1986, more than 10 million people were treated with antihypertensive medications, at a cost of $2.5 billion.4 Epidemiological studies have shown that hypertension is a major contributor to stroke incidence and mortality,5 6 7 and clinical trials have confirmed the benefits of treatment to prevent stroke, congestive heart failure, and LVH.8 9 10 Recent clinical trials on hypertension control in older persons have also demonstrated reductions in coronary heart disease morbidity and mortality.11 12 13
A number of surveys and studies have reported trends over time in the prevalence and control of hypertension.14 15 16 17 In these reports, the hypertensive status of a subject is assessed on a single occasion. Similarly, reports from clinical trials identify subjects as hypertensive at the start of the study and compare morbidity and mortality between treatment groups during the subsequent follow-up period.9 10 11 12 13 There have been no reports of secular trends in the treatment and control of hypertension sustained for long durations, even though it is the chronic nature of hypertension that gives rise to many of its cardiovascular sequelae. The objectives of the present study were (1) to describe trends in the prevalence of LTS (≥6 of 10 years) hypertension and its treatment at time points bridging the introduction and widespread use of antihypertension medications and (2) to assess the impact of long-term treatment on 10-year CVD incidence and mortality.
The Framingham Heart Study is a prospective epidemiological study of CVD based on a two-thirds sample of the adult population in Framingham, Mass. Of the original 5209 individuals aged 28 through 62 years when assembled in 1948, 55% were female.18 Information on newly developed CVD and on risk factors is obtained through standard examinations held biennially consisting of an interview, physical examination, and laboratory tests. Information on all CVD events, hospitalizations, and deaths occurring between examinations is obtained through daily hospital and death surveillance.
Study Design and Definitions
This study uses a modified cross-sectional cohort design that measures differences over time among successive cohorts of men and women of similar ages. This report focuses on the prevalence of hypertension in 1950, 1960, and 1970 among three independent and mutually exclusive cohorts of men and women from the Framingham Heart Study. There were 589 men and 719 women 50 to 59 years old as of January 1, 1950 (the 1950 cohorts), 596 men and 799 women 50 to 59 years old as of January 1, 1960 (the 1960 cohorts), and 655 men and 830 women 50 to 59 years old as of January 1, 1970 (the 1970 cohorts). The study population totaled 4188: 1840 men and 2348 women.
Prevalence of hypertension in each cohort is presented in terms of point prevalence (hypertensive at baseline examination in 1950, 1960, or 1970) and period prevalence (hypertensive status during 1950 to 1960, 1960 to 1970, and 1970 to 1980). Hypertension is defined as SBP ≥160 mm Hg and/or DBP ≥95 mm Hg or the patient taking antihypertension medication.14 15 Hypertension is considered to be controlled if SBP is <160 mm Hg and DBP is <95 mm Hg while the patient is taking antihypertension medication and is uncontrolled if SBP is ≥160 mm Hg and/or DBP is ≥95 mm Hg regardless of treatment.14 15
The analysis of LTS hypertension is based on subjects in the 1950, 1960, and 1970 cohorts who attended the baseline examination for their cohort and at least three of the five subsequent biennial examinations, including the examination at the end of 10 years of follow-up. Of the 1950 cohorts, 434 men (74%) and 581 women (81%) met these criteria; of the 1960 cohorts, 442 men (74%) and 625 women (78%); and of the 1970 cohorts, 505 men (77%) and 708 women (85%).
A subject has LTS hypertension if he or she is hypertensive at the baseline examination plus a minimum of three of the five subsequent biennial examinations, including the examination 10 years after baseline (ie, hypertensive ≥6 of 10 years). Anyone who missed two consecutive examinations or died within 10 years of baseline is excluded. Those LTS hypertensive patients receiving antihypertension medication at four or more biennial examinations (≥6 years) are considered to have long-term treatment. LTS hypertension is considered to be controlled if SBP is <160 mm Hg and DBP is <95 mm Hg for ≥6 of 10 years while the patient is taking antihypertension medication and is uncontrolled if SBP is ≥160 mm Hg and/or DBP is ≥95 mm Hg for ≥6 of 10 years regardless of treatment.
Other Risk Factor Data
Methods of risk factor measurement and laboratory analysis are described elsewhere.19 SBP and DBP, cholesterol, glucose intolerance, and LVH at baseline of each cohort and 10 years after baseline are reported for all LTS hypertensive patients. In the 1970s, plasma cholesterol measurements were substituted for serum cholesterol in the Framingham Study. Since plasma cholesterol values have been shown to be systematically lower than those from serum,20 plasma values for the 1970 cohort were inflated by 3%.21
End-Point Definition and Ascertainment
Cardiovascular and vital status information was available for all subjects included in this report. Throughout the Framingham Study, cardiovascular events have been dated and assigned a diagnosis based on medical record review and standard criteria for each outcome18 22 consistently applied at regular meetings of a panel of physicians. CVD includes coronary heart disease (myocardial infarction, angina and coronary insufficiency, sudden and nonsudden coronary death), stroke, and other CVD (transient ischemic attacks, congestive heart failure, intermittent claudication, and other cardiovascular death).
Each cohort was followed for 20 years from baseline (1950 cohort, 1950 to 1970; 1960 cohort, 1960 to 1980; and 1970 cohort, 1970 to 1990). CVD incidence among LTS hypertensive patients was ascertained during the first 10 years of follow-up of each cohort, and mortality (all-cause and CVD) was assessed during the second 10 years of follow-up.
Trends in rates of hypertension, treatment, and cardiovascular events were analyzed by the Mantel-Haenszel test for trend.23 Student’s t test23 was used to compare baseline risk factor levels between LTS hypertensive patients in the three cohorts with and without long-term treatment and to compare changes in risk factors after 10 years between those with and without long-term treatment. The χ2 test23 was used to compare 10-year incidence of CVD and 10-year mortality (all-cause and cardiovascular) between those with and without long-term treatment. Cox proportional hazards regression analysis24 was used to determine the risk of death in 10 years for those LTS hypertensive patients with and without long-term treatment while controlling for other risk factors. A significance level of P≤.05 was used to test hypotheses, and all significance tests were two-tailed.
Trends in Baseline (Point) Prevalence of Hypertension
The prevalence of hypertension among 50- to 59-year-old men at baseline was similar between the 1950 and 1970 cohorts (Table 1⇓; P=.66), while it declined from 366 per 1000 among women in the 1950 cohort to 270 per 1000 in the 1970 cohort (P<.001). Forty-five percent of female hypertensive patients in the 1960 cohort received treatment compared with 24% of male hypertensive patients (P<.001). In the 1970 cohorts, 56% of women received treatment compared with 45% of men (P<.05). More women than men had their hypertension controlled in the 1960 cohorts (14% compared with 6%; P<.01) and 1970 cohorts (31% compared with 22%; P<.05).
The prevalence of uncontrolled hypertension was significantly lower in the 1970 male and female cohorts compared with the 1950 cohorts (P<.001 for both male and female cohorts). More women than men with uncontrolled hypertension had elevated SBP in the cohorts: 25% compared with 15% in the 1950 cohorts (P<.05) and 32% compared with 20% in the 1970 cohorts (P<.01). In contrast, more men than women with uncontrolled hypertension in each cohort had elevated DBP: 30% compared with 12% in the 1950 cohorts (P<.001), 35% compared with 23% in the 1960 cohorts (P<.01), and 24% compared with 15% in the 1970 cohorts (P<.05). Treatment rates for uncontrolled hypertension, whether due to elevated SBP or elevated DBP, were ≈20% among the male cohorts and 30% among the female cohorts.
Trends in Period Prevalence of LTS Hypertension
Trends in prevalence of LTS hypertension over 10-year periods are presented in Table 2⇓. Period prevalence was 208 per 1000 in the 1970 male cohort compared with 138 in the 1950 cohort (P<.01), while it was 198 per 1000 in the 1970 female cohort compared with 253 in the 1950 cohort (P<.02). Even though long-term treatment and control increased between the 1960 and 1970 male cohorts (P<.001 for each), more women than men with LTS hypertension in each cohort received treatment: 63% of women compared with 27% of men in the 1960 cohorts (P<.001) and 89% compared with 78% in the 1970 cohorts (P<.05). More women than men with LTS hypertension also had their hypertension controlled: 33% of women compared with 14% of men in the 1960 cohorts (P<.01) and 78% compared with 66% in the 1970 cohorts (P<.05).
The rate of uncontrolled LTS hypertension was 23% in the 1970 male cohort compared with 85% in the 1950 cohort (P<.001) and 15% in the 1970 female cohort compared with 92% in the 1950 cohort (P<.001). The percent who were uncontrolled and untreated was 4% in the 1970 male cohort compared with 35% in the 1950 cohort (P=.03) and 1% in the 1970 female cohort compared with 31% in the 1950 cohort (P<.001).
Risk Factors Among LTS Hypertensive Patients With and Without Long-term Treatment
Age and blood pressure levels at initiation of antihypertensive therapy were compared for those LTS hypertensive patients with long-term treatment identified in each cohort (Table 3⇓). Treatment was initiated at younger ages (P<.001) and at lower SBP and DBP (both P<.001) in each successive cohort.
Risk factor levels at baseline of each cohort were compared for LTS hypertensive patients with and without long-term treatment (Table 4A⇓). Among men and women in the 1950 cohorts, baseline SBP was higher for those with treatment (195 mm Hg compared with 177 mm Hg, P<.01), as was DBP (113 mm Hg compared with 101 mm Hg, P<.001). However, this situation was reversed in the 1970 cohort, when those with long-term treatment had a baseline SBP of 156 mm Hg compared with 163 mm Hg among those without treatment (P<.05 for difference). This is consistent with the trend to initiate treatment earlier noted in Table 3⇑. More LTS hypertensive patients with LVH received long-term treatment. Baseline prevalence of LVH (by ECG) among those with long-term treatment was approximately twice the prevalence of those without treatment. This difference was not statistically significant because of sample size.
Changes in baseline risk factor levels over 10 years of follow-up of each cohort were compared for LTS hypertensive patients with and without long-term treatment (Table 4B⇓). Mean changes in SBP were similar, but DBP dropped an average of 10.2 mm Hg over 10 years among those with long-term treatment compared with 7.6 mm Hg among those without (P=.01). Cholesterol fell 4.6 mg/dL and glucose intolerance rose 12% among those with long-term treatment compared with a 2-mg/dL rise in cholesterol and a 7% rise in glucose intolerance among those without (P<.03 and P<.06, respectively). The prevalence of LVH rose only 0.3% over 10 years among those with long-term treatment compared with a 4.4% rise among those without. This difference was not statistically significant.
Cardiovascular Morbidity and Mortality Among LTS Hypertensive Patients
CVD incidence during the first 10 years of follow-up of the three cohorts was 29% among men with LTS hypertension, 24% among women, and 26% among men and women combined (Table 5⇓). There was no significant difference in CVD incidence among LTS hypertensive men with and without long-term treatment (30% versus 27%), among hypertensive women (24% versus 24%), or among hypertensive men and women combined (26% versus 25%).
Table 6⇓ compares trends in all-cause and CVD mortality between LTS hypertensive patients with and without long-term treatment in each cohort. In the combined cohorts, all-cause mortality was 40% for men with long-term treatment compared with 49% for men without (P<.05), while CVD mortality was 14% among men with long-term treatment compared with 30% among men without (P<.01). Twenty-seven percent of LTS hypertensive women with long-term treatment died during 10 years of follow-up compared with 36% without long-term treatment (P<.01). Of those with long-term treatment, 9% died of CVD compared with 20% without (P<.01). When the male and female cohorts were combined, the differences in mortality rates between those LTS hypertensive patients with and without long-term treatment were significant at the P<.001 level: All-cause mortality was 31% for those with long-term treatment compared with 41% for those without, and CVD mortality was 11% for those LTS hypertensive patients with long-term treatment compared with 24% for those without.
Cox proportional hazards regression analyses were performed using risk factor levels 10 years after baseline to predict mortality during the second 10 years of follow-up. After age, sex, smoking, cholesterol level, and glucose intolerance status were controlled for, the risk of death among those LTS hypertensive patients with long-term treatment compared with those without was 0.69 (95% CI, 0.53 to 0.89), and the risk of death from CVD was 0.40 (95% CI, 0.27 to 0.60).
This is a unique study of trends in the treatment and control of LTS hypertension between 1950 and 1990. Four decades of the Framingham Heart Study were analyzed, during which time antihypertension medications were introduced and widely adopted. Although other studies have reported changes in the point prevalence of hypertension over time17 25 and trends in the treatment of short-term sustained hypertension,26 this is the first report of trends in the prevalence of hypertension sustained for ≥6 of 10 years, its treatment, and sequelae. The study of LTS hypertension is important because the prevalence rates it provides are not affected by random fluctuations in blood pressure measurements. More importantly, since many of the sequelae of hypertension reflect chronic processes that mediate the effects of blood pressure,7 9 it is essential that hypertension and treatment sustained for long periods of time be the focus of evaluation research. Trends in the treatment and control of sustained hypertension are, thus, appropriate outcome measures for hypertension detection programs.
In the present report, the prevalence of LTS hypertension increased between the 1950 and 1970 male cohorts, while prevalence decreased between the female cohorts. Since hypertension in this study is defined by SBP and DBP levels as well as the use of antihypertension medication, higher prevalence can reflect increased use of medications due to initiation of treatment at younger ages or at lower blood pressures—both trends noted in this report. On the other hand, the lower prevalence of LTS hypertension in the 1970 female cohort compared with the 1950 cohort cannot be explained by changes in the use of medications. Elsewhere,27 however, we have reported a significant reduction in the prevalence of obesity (defined as weight ≥120% of the US mean for sex, age, and height) among women in the Framingham Heart Study. Prevalence dropped from 62% of women in 1950 to 44% in 1970 (P<.001). The average body mass index [ie, weight (kg)/height (m)2] declined from 27.1 to 25.5 (P<.001). If we estimate that a 10-pound difference in weight results in a 4.5 mm Hg average change in SBP,28 secular trends in obesity may offer an explanation for the lower prevalence of LTS hypertension in women in 1970 reported here.
Major gains were made in the long-term treatment and control of LTS hypertension between 1960 and 1980. Among men, long-term treatment increased 51%. In the 1960 cohort, one half of men with long-term treatment had their hypertension controlled; by the end of follow-up of the 1970 cohort, 84% of those treated were controlled. Long-term treatment of LTS hypertension increased 45% between the female cohorts, and control of LTS hypertension rose from 50% of those treated in the 1960 cohort to 88% in the 1970 cohort. Uncontrolled LTS hypertension declined by 77% between the female cohorts, and only 1% of all LTS hypertensive patients in the 1970 cohort remained untreated during follow-up.
While three clinical trials of antihypertension drugs in older persons found significantly lower relative risk of CVD events (0.60 to 0.83) among those on active treatment compared with those on placebo,11 12 13 we did not observe a significant difference in 10-year incidence of CVD between those LTS hypertensive patients with and without long-term treatment in the three cohorts. This is not surprising, given the limited sample size and the high prevalence of LVH, a major risk factor for CVD,29 among those with long-term treatment in the present study. Although we did not assess physicians’ reasons for initiating therapy, logistic regression analysis indicated that presence of LVH was a significant predictor of hypertension therapy among LTS hypertensive patients in these cohorts.
The beneficial effects of long-term treatment of LTS hypertension on mortality end points were evaluated during the second 10 years of follow-up of each cohort. When other risk factors were controlled, risk of death was 31% lower and risk of CVD death was 60% lower among those with long-term treatment compared with those without. This finding is consistent with decreased total and vascular mortality reported in clinical trials of the pharmacological treatment of hypertension.9 11 12 13 That mortality should have been affected 10 to 20 years after baseline speaks to the chronic processes involved in hypertension and its cardiovascular sequelae. A recent report after 10.5 years of follow-up of the MRFIT study30 also indicated a delayed effect of risk factor changes on mortality end points.
Study Strengths and Limitations
This study evaluates the impact over time of the introduction of intensive antihypertension therapy for LTS hypertension in a free-living general population. It is an observational study and not a clinical trial: comparison groups were not matched, treatment was not randomly assigned, and therapies were not administered according to strict protocols. In the general population, many factors contribute to the physician’s decision to administer therapy, and compliance is less certain than in clinical trials. For these reasons, results from population-based studies may provide more realistic evaluations of therapeutic interventions.
The methods used in this report differ from those of prior evaluations of hypertension therapy from the Framingham Heart Study. Using a person-examination approach that pooled examinations over a 30-year period, Cupples and D’Agostino19 reported that for any given blood pressure, those receiving treatment had a higher risk of death in 2 years than those not receiving treatment. In contrast, we identified subjects with LTS hypertension in 1950, 1960, and 1970. We followed each of these subjects for 20 years from their baseline examination to assess CVD incidence and mortality. Relative risk of death in 10 years was estimated for those with and without long-term treatment while risk factors other than blood pressure were controlled for, since comparing risks between treated and untreated hypertensive patients with the same blood pressure levels ignores the benefit of treatment. It assumes that a blood pressure attained on medication is the same as an untreated blood pressure.31
Bias may have entered our analysis of incidence through the “healthy survivor effect,” since to be defined as a LTS hypertensive patient, the subject had to be alive and had to attend the examination at the end of his or her cohort’s first 10-year follow-up period. If more of the untreated patients died of CVD before the end of this first period and were therefore excluded from the analysis of incidence, our finding of no difference in incidence between the two groups underestimates the protective effect of long-term treatment.
The decline in cardiovascular mortality among men and women in this study was concomitant with increased treatment of LTS hypertension. Because of small sample size in each cohort, we combined the cohorts for the statistical analysis of mortality. Whether in doing so we overestimated the effect of treatment or whether, as has been suggested,28 32 there was a causal relation between hypertension treatment and the decline in cardiovascular mortality cannot be definitively answered through these analyses.
Selected Abbreviations and Acronyms
|DBP||=||diastolic blood pressure|
|LVH||=||left ventricular hypertrophy|
|SBP||=||systolic blood pressure|
This study was supported by NIH/NHLBI grant R01-HC-40423 and contract N01-HC-38038.
Reprint requests to Pamela A. Sytkowski, PhD, MSc, The Framingham Heart Study, Statistics and Consulting Unit, Boston University, Department of Mathematics, 111 Cummington St, Boston, MA 02215.
- Received August 14, 1995.
- Revision received October 26, 1995.
- Accepted October 31, 1995.
- Copyright © 1996 by American Heart Association
Health, United States, 1990. Hyattsville, Md: Government Printing Office; 1991. DHHS publication 91-1232.
Morbidity and Mortality Chartbook on Cardiovascular, Lung and Blood Disease −1990. Bethesda, Md: Government Printing Office; 1990. National Heart, Lung, and Blood Institute.
Kannel WB. Epidemiology of essential hypertension: the Framingham experience. Proc R Coll Physicians (Edinburgh). 1991;21:273-287.
MacMahon S, Peto R, Cutler J, Collins R, Sorlie P, Neaton J, Abbott R, Godwin J, Dyer A, Stamler J. Blood pressure, stroke, and coronary heart disease, I: prolonged differences in blood pressure: prospective observational studies corrected for the regression dilution bias. Lancet. 1990;335:765-774.
Collins R, Peto R, MacMahon S, Hebert P, Fiebach N, Eberlein KA, Godwin J, Qizilbash N, Taylor J, Hennekens CH. Blood pressure, stroke, and coronary heart disease, II: short-term reductions in blood pressure: overview of randomized drug trials in the epidemiologic context. Lancet. 1990;335:827-838.
Schulman SP, Weiss JL, Becker LC, Gottlieb SO, Woodruff KM, Weisfeldt ML, Gerstenblith G. The effects of antihypertensive therapy on left ventricular mass in elderly patients. N Engl J Med. 1990;332:1350-1356.
MRC Working Party. Medical Research Council trial of treatment of hypertension in older adults: principal results. BMJ. 1992;304:405-412.
Roberts J. Blood pressure of persons 18-74 years, United States, 1971-72: data from the National Health Survey (NHANES I). Vital Health Stat. 1975; No. 150. Washington, DC: Government Printing Office; 1975. US Dept of Health, Education, and Welfare publication 75-1632.
Rowland M, Roberts J. Blood pressure levels and hypertension in persons aged 6-74 years: United States, 1976-80 (NHANES II). Vital Health Stat. 1982; No. 84. Washington, DC: Government Printing Office; 1982. US Dept of Health and Human Services publication 82-1250.
Drizd T, Dannenberg A, Engel A. Blood pressure levels in persons 18-74 years of age in 1976-80 and trends in blood pressure from 1960-1980. Vital Health Stat 1986; No. 234. Washington, DC: Government Printing Office; 1986. US Dept of Health and Human Services publication 86-1684.
Dawber TR. The Framingham Study: The Epidemiology of Atherosclerotic Disease. Cambridge, Mass: Harvard University Press; 1980.
Cupples LA, D’Agostino RB. Some risk factors related to the annual incidence of cardiovascular disease and death using pooled repeated biennial measurements: Framingham Study, 30-year follow-up. In: Kannel WB, Wolf PA, Garrison RJ, eds. The Framingham Study: An Epidemiological Investigation of Cardiovascular Disease, Section 34. Washington, DC: Government Printing Office; 1987. NIH publication 87-2703.
Laboratory Methods Committee of the Lipid Research Clinics Program of the National Heart, Lung, and Blood Institute. Cholesterol and triglyceride concentration in serum/plasma pairs. Clin Chem. 1977;23:60-63.
Feinleib M, Garrison RJ, Stallones L, Kannel WB, Castelli WP, McNamara PM. A comparison of blood pressure, total cholesterol and cigarette smoking in parents in 1950 and their children in 1970. Am J Epidemiol. 1979;110:291-303.
D’Agostino RB, Kannel WB. Epidemiological background and design: the Framingham Study. In: Gail MH, Johnson NL, eds. Proceedings of the American Statistical Association Sesquicentennial 1988-89. Alexandria, Va: American Statistical Association; 1989:707-718.
Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic Research: Principles and Quantitative Methods. London, UK: Lifetime Learning Publications; 1982.
Cox DR. Regression models and life-tables. J R Stat Soc B. 1972;34:187-220.
Hennekens CH, Satterfield S, Hebert P. Treatment of elevated blood pressure to prevent coronary heart disease. In: Higgins MW, Luepker RV, eds. Trends in Coronary Heart Disease Mortality. New York, NY: Oxford University Press; 1988:103-108.
Shea S, Cook EF, Kannel WB, Goldman L. Treatment of hypertension and its effect on cardiovascular risk factors: data from the Framingham Heart Study. Circulation. 1985;71:22-30.
Sytkowski PA, D’Agostino RB, Belanger A, Kannel WB. Gender and time trends in cardiovascular disease incidence and mortality: the Framingham Heart Study 1950-1989. Am J Epidemiol. In press.
D’Agostino RB, Wolf PA, Belanger AJ, Kannel WB. Stroke risk profile adjustment for anti-hypertensive medications: the Framingham Study. Stroke. 1994;25:40-43.
Goldman L, Cook EF. The decline in ischemic heart disease mortality rates: an analysis of the comparative effects of medical interventions and changes in lifestyle. Ann Intern Med. 1984;101:825-836.