| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
(Circulation. 1999;100:e88-e94.)
© 1999 American Heart Association, Inc.
Circulation Electronic Pages |
Identification of Risk Factors in Hypertensive Patients
Contribution of Randomized Controlled Trials Through an Individual Patient Database
From the Clinical Pharmacology Department (F.G., J.-P.B., F.B.), Claude Bernard University, Lyon Hospitals, Lyon, France; the National Heart, Lung, and Blood Institute (J.C., L.F., E.S.), National Institutes of Health, Bethesda, Md; the Department of Community Health Sciences (T.E.), Dalby/Lund, Sweden; Veterans Administration Medical Center (K.K.), San Francisco, Calif; the Hypertension and Cardiovascular Rehabilitation Unit (R.F.), Leuven, Belgium; Washington University School of Medicine (M.P.), St Louis, Mo; London School of Hygiene and Tropical Medicine (S.P.), London, UK; and the Division of Epidemiology (R.P.), University of Minnesota, Minneapolis. Dr Coope is in general practice, Bollington, UK.
Correspondence to Dr François Gueyffier, Service de Pharmacologie Clinique, Faculté RTH Laënnec, Rue 6, Paradin BP8071, 69376 Lyon, France. E-mail fg{at}upcl.univ-lyon1.fr
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
Background—Predicting individual risk is needed to target preventive interventions toward people with the highest probability of benefit over a given time period. We assessed which prognostic factors should be used in predicting risk for hypertensive patients and in searching for treatment modifiers.
Methods and Results—Data from 24 390 hypertensive participants who constituted the control groups from 8 controlled trials (1726 deaths over 5 years) were analyzed in multivariate survival models. Outcomes were coronary heart disease death, stroke death, and cardiovascular death. We explored systematically the heterogeneity of results between trials. Left ventricular hypertrophy was electrocardiographically confirmed to be a powerful risk factor and should be included in risk scoring. Height, glomerular filtration rate, and serum uric acid deserve further exploration. Body mass index and heart rate were not confirmed as independent cardiovascular risk factors in this population. The association between male sex and coronary heart disease death was significantly stronger in British cohorts. The lack of prognostic value of diastolic blood pressure was explained by an interaction with age, with a positive association before 65 years and a negative association thereafter. Previous antihypertensive treatment was a significant risk factor.
Conclusions—Clinical trials provide valuable information for risk prediction. Carefully exploring the heterogeneity among trials is a way to assess the generalizability of findings. This approach, if systematically performed, should increase the ability to identify risk modifiers and to predict individual therapeutic benefit.
Key Words: hypertension • trials • epidemiology • stroke • coronary disease
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Arational prevention strategy against cardiovascular diseases includes targeting interventions toward people with a high probability of benefit from intervention.1 The starting point of this approach is the hypothesis that the treatment effect model2 of antihypertensive drugs is multiplicative. This means that the relative treatment effect estimates (relative risk, hazard ratio, and odds ratio) are the most generalizable, or stable, and that these estimates are constant over the risk levels. In such a situation, the predicted absolute benefit is simply obtained by multiplying the predicted risk without treatment by the relative treatment effect estimate. Such a simple method to predict treatment benefit has been proposed to regulate the prescription of antihypertensive drugs3 in New Zealand.
The Individual Data Analysis of Antihypertensive Intervention Trials (INDANA) database includes baseline and follow-up characteristics of participants enrolled in randomized controlled trials that assessed the effect of blood pressure–lowering drugs on clinical outcomes. The main objective of the project is to identify and describe the profiles of responders and nonresponders to these drugs in terms of cardiovascular risk. These profiles will include the individual characteristics either linked to the risk (risk factors) alone or together with those modifying the treatment effect on a multiplicative scale. Under the assumption of a multiplicative treatment effect, absolute benefit is proportional to risk; thus, risk factors modify the size of the absolute benefit. This modification is called an arithmetic interaction, since the association disappears when one changes the treatment effect scale. Biological interactions are defined as the change in the size of treatment effect that cannot be explained by a simple change of scale, either multiplicative or additive. The INDANA approach can also be defined as an attempt to refute the hypothesis of the multiplicative effect, through looking for biological interactions.
This approach includes 3 principal steps: (1) identifying the significant risk factors, which would be used as adjustment covariates in the following step; (2) identifying the factors modifying the treatment effect on a multiplicative scale, ie, biological interactions; and (3) exploring the impact of these interactions on the efficiency of treatment decisions. The first step led us to assess the prognostic impact of the main cardiovascular risk factors and some other characteristics, measured in the control groups of the trials included in INDANA. The pooled analysis of data from several cohorts from different sources enables us to explore the general value of results, through the heterogeneity of findings. If a given risk factor always yields the same association with a given event, without any significant change from one cohort to another, its prognostic value is likely to be general, or generalizable. The results of the first step are presented here. Its 2 objectives are as follows: (1) to assess which prognostic factors should be retained in treatment decision making because they predict cardiovascular outcomes and (2) to explore in a systematic way the generalizability of results through their heterogeneity across trials.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Participants
The data extracted from the INDANA database included participants from the control groups of 8 trials4 5 6 7 8 9 10 11 (Table 1
). The data from 3 other
trials in the database have not been used: the Systolic
Hypertension in the Elderly Program Pilot12 (SHEP-Pilot)
study did not offer baseline data other than age, sex, and blood
pressure; the Veterans Administration–National Heart, Lung, and Blood
Institute feasibility trial13 (VA-NHLBI) had only a very
small number of events; and because of the lack of follow-up after the
first end point, the data from the Australian14 trial are
potentially biased when the terminating end points are analyzed
separately.
|
In 3 of the remaining 8 trials, the data required to analyze survival without nonfatal outcomes were not available: in the European Working Party on High Blood Pressure in the Elderly trial7 (EWPHE), nonfatal outcomes were documented only when the participants were on treatment; in the Hypertension, Detection, and Follow-up Program9 (HDFP) and in the Multiple Risk Factor Intervention Trial4 (MRFIT), only fatal events were accurately dated. Thus, the present study focuses on total and cause-specific mortality.
The main characteristics of each trial have been published previously,15 and more details can be found in separate reports from each trial.
Outcomes
Among the 7 outcomes adopted in the INDANA
protocol,15 only the results concerning fatal
cardiovascular outcomes are considered in the
present study: fatal strokes, fatal coronary events, and
cardiovascular mortality.
Potential Risk Factors
The independent covariates considered were measured at entry
into each trial. Some were continuous: age (years), systolic
and diastolic blood pressure (mm Hg), heart rate (bpm),
body mass index (kg/m2), height (m), serum total
cholesterol (mmol/L), glomerular filtration
rate (mL/min), and serum uric acid (mmol/L); others were dichotomous:
sex (male 1, female 0); smoking status (current smoker 1, other 0);
myocardial infarction history, stroke history, diabetes history, or
antihypertensive treatment history (history 1, no history 0); presence
of left ventricular hypertrophy criteria on ECG
(present 1, absent 0); and ethnic origin (African American 1, other
0). Glomerular filtration rate was determined from serum
creatinine, age, sex, and weight by the Cockroft
formula16 :
GFR=(140-age)xweightx(1.228/creatinine)x(0.85+0.15xsex),
where age is given in years, weight in kilograms, and
creatinine in millimoles per liter.
Most of the variables were collected during standardized clinical interview(s) and examination(s) (age, sex, ethnic origin, smoking status, cardiovascular history, height, and weight) and reflect the information one can observe in good clinical practice. Blood pressure measurements were always the basic inclusion criteria and thus were collected with a particular attention. Protocols of measurements varied greatly between trials in terms of their number, the use of random zero sphygmomanometer, and the time lapse between different measurements. ECG-defined left ventricular hypertrophy was the association of tall R waves (corresponding to Minnesota code 3.1) with abnormalities of repolarization (corresponding to Minnesota code 4.1 to 4.3 or 5.1 to 5.3). Laboratory data were gathered in various settings: in some trials, laboratory measurements were performed centrally for ECG and biochemistry; in others, these measurements were performed in the center for recruitment and follow-up.
Statistical Analysis
Hazard ratios associated with the presence of each covariate
were computed through a Cox proportional hazards model17
in univariate and multivariate approaches.
Analyses were stratified by trials, which allowed us to take
into account differences in risk across trials without any assumption
of proportionality of hazards along time.
Because different approaches may select different sets of covariates, 3 different stepwise approaches were used to select covariates in multivariate models: backward and forward approaches and those based on best score, with P=0.05 as threshold to retain covariates.
A first model was parameterized on the basis of 11
covariates available in all trials, namely, age, sex, smoking status,
systolic and diastolic blood pressure, stroke,
myocardial infarction or diabetes history, body mass index, height, and
serum total cholesterol. The results concerning these
covariates are given from this model. Other models were built to
explore the predictive value of 6 covariates that were not available
for all trials (see Table 1
for the availability of covariates
by trial), ie, glomerular filtration rate, serum uric acid,
ethnic origin, left ventricular hypertrophy,
history of antihypertensive treatment, and heart rate. The results of
these models are given only for these covariates.
The heterogeneity of the prognostic role of a given covariate across trials was explored by testing the significance of the increase in fit of a model involving interaction terms between covariate and trials compared with the same model without these terms. When the increase in fit was significant, a backward stepwise procedure explored which trial(s) was responsible for the interaction. The significance threshold to keep an interaction in the model was 0.05.
The improvement of fit due to the inclusion of covariates in the full model was illustrated in comparing the observed risk among quartiles of predicted risk between 2 models: one including only sex and age as covariates and one also including the other significant covariates available in all trials.
The software used for data management and statistical analysis was SAS.18
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
The number of participants included in the analysis was 23 415, among 24 390 included in the control groups of the trials considered (Table 1
). This difference is related to missing
values for baseline covariates (mainly serum cholesterol
and body mass index). In addition, the data from MRFIT were included
only for hypertensive participants, whose baseline
diastolic blood pressure was >90 mm Hg. Standardized
hazard ratios were estimated by using standard deviations computed on
the pooled data and rounded in order to illustrate the specific
importance of each risk factor in this population.
Results From the Full Multivariate Models
The hazard ratios for sex showed the only significant
discrepancy between the 3 outcomes (Figure 1): whereas being a man was the strongest
risk factor for coronary heart disease death, with a hazard
ratio 
4, there was no significant difference between men and women
for stroke death. Some covariates did not reach significance for stroke
death, possibly because of the relatively low numbers associated with
stroke death: there was almost 3 times more coronary heart
disease death than stroke death (Table 1). Stroke history was a
risk factor for stroke and coronary heart disease death,
whereas myocardial infarction history was a risk factor mainly for
coronary heart disease death. The prognostic importance of left
ventricular hypertrophy from ECG records
appeared similar or greater than that of smoking status or diabetes.
Compared with other people, African Americans had an increased risk of
cardiovascular death. Having already been treated with
antihypertensive drugs was also a significant risk factor.
|
For continuous covariates (Figure 1, right), age was by far the
most important risk factor. Systolic blood pressure was better
correlated with stroke than with coronary heart disease death.
Total cholesterol was not associated with stroke death but
with the 2 other outcomes (cardiovascular death and
coronary heart disease death). A decrease in
glomerular filtration rate and an increase in uric acid
were associated with a greater risk for the 3 outcomes; the magnitude
of this risk was similar to that of blood pressure and total
cholesterol. Height was inversely correlated with the risk
of coronary heart disease and cardiovascular
death.
Three covariates were found not to be related to cardiovascular fatal outcomes: heart rate, body mass index, and diastolic blood pressure.
The gain in predictive power due to the addition of the other
covariates to age and sex (Figure 2) is
more impressive for myocardial infarction and
cardiovascular death than for stroke death or all-cause
death.
|
Heterogeneity of Prognostic Value of Each Risk
Factor Between Trials
Nine covariates showed no significant
heterogeneity for their prognostic roles across trials:
smoking status, stroke history, antihypertensive treatment history,
ethnic origin, total cholesterol, glomerular
filtration rate, diastolic blood pressure, body mass index,
and heart rate. The other 8 were associated with some differences
between trials for at least one outcome. Concerning
cardiovascular death, these differences are indicated
on the right of Table 2. The trials in
which the hazard ratio is significantly different from the others are
listed. We give both the hazard ratio in the remaining trials (residual
hazard ratio) and the ratio that is to be multiplied by the residual
hazard ratio (added hazard ratio) to yield the specific hazard ratio in
the trials showing a significant heterogeneity. For
example, the hazard ratio of cardiovascular death for
men compared with women was 2.63 for the population as a whole (Figure 3). However, this hazard ratio was only
1.57 (residual risk ratio) when the stronger effect of sex in the
Coope8 and UK Medical Research Council
(UK-MRC)5 6 trials was isolated. The hazard ratio
computed was 2.28x1.57=3.58 for Coope, 5.04 for UK-MRC1, and 2.68 for
UK-MRC2.
|
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
In the present study, we demonstrate the first step of the modeling procedure to explore whether individual patient characteristics are associated with the intensity of the effect of blood pressure–lowering drugs in terms of clinical outcome: we identified the significant risk factors for cause-specific cardiovascular death and explored the heterogeneity of our findings across trials.
Advantages of Multivariate Analysis
The comparison of results from univariate and
multivariate analyses shows how taking into
account other risk factors may modify the estimate of their association
with risk. For example, stroke history adjusted with other risk factors
multiplied the risk of cardiovascular death by 2.82,
whereas univariate modeling suggested a stronger
association (hazard ratio 3.62) (Table 2). On the other hand, an
association may be revealed or reinforced after adjustment: an increase
in 1 mmol/L of serum total cholesterol corresponded to
a hazard ratio of 1.11 for cardiovascular death, which
is highly significant, whereas the univariate approach did
not retain cholesterol as a risk factor
(P=0.09), with the estimated hazard ratio being 1.05 (Table 2). The most striking example of the effect of allowing for
other covariates concerns height, which is positively and significantly
correlated with coronary heart disease in the
univariate approach but significantly and negatively
correlated with the same outcome in the multivariate
approach (Figure 3). This finding is explained by the fact that,
on average, men are taller than women are and at higher risk of
coronary heart disease. As a consequence, we must concentrate
the discussion on results from multivariate
analysis.
Limitations
The present study has several limitations: (1) It uses data
from cohorts that were not specifically designed to assess the
prognostic value of any of the studied characteristics, and the
collection of baseline data was not standardized between trials. For
example, blood pressure readings analyzed here were obtained
most carefully and with more frequent measures than in traditional
observational surveys, but the number of measurements used, and the
time lapse between measurements, varied across trials. The resulting
differences in measurement errors may explain a part of the
heterogeneity of the association slopes between trials.
This drawback may be prevented by prospectively planning the
meta-analysis before the inception of the included trials
(prospective meta-analyses19 20 ), which allows
standardizing procedures across trials. (2) The average follow-up is 5
years, which is not very long compared with the usual longitudinal
studies. It is planned to increase the follow-up duration for most
trials included in the INDANA database for mortality data. (3) Many
analyses have been performed; this increases the likelihood
of significant positive results by chance. (4) Statistical modeling
is simplistic and rigid compared with the complexity of biology. We
could have increased the complexity of statistical models by including
interaction terms to enhance the fit of the model to the data. However,
such an approach may increase the singularity of the model and decrease
the general value of the findings. Eventually, it may increase the gap
between models and biological reality. Modeling should be seen here as
an attempt to find out the best combination of risk factors that have
the strongest and the most general prognostic value. (5) The cohorts
analyzed in the present study represent European
and North American populations, with a reasonably large proportion of
African American participants in HDFP9 and
SHEP10 ; it may not be appropriate to
extrapolate the results to other populations. (6) The selection of
people by blood pressure criteria and the selection procedure for
clinical trials in general yield samples that are not
representative of the general population. However, the
variety of exclusion criteria between trials, with large inclusion
criteria in one of the biggest trials (HDFP), leads to a very
heterogeneous population for the present study. This
fact enables us to explore efficiently the general value of each
potential risk factor. (7) Although atrial fibrillation is a common
risk factor for stroke, we are unable to account for its contribution
because it was an exclusion criteria in most trials. This is
unfortunate for several reasons: hypertension is a risk factor for
atrial fibrillation, and both are powerful risk factors for
stroke21 (ischemic stroke for atrial fibrillation
and both ischemic and hemorrhagic stroke for hypertension). The
prevalence of both conditions increases with age, so that their
association is frequent in the elderly. Antiplatelet and
anticoagulant agents reduce the risk of stroke in atrial
fibrillation but increase the risk of
hemorrhage.22 Unfortunately, there are no data
regarding the respective benefit and a possible interaction between
these treatments yet.
Complementary Analyses and Implications for Future
Research
Our results confirm the prognostic impact of commonly identified
cardiovascular risk factors23 such as
older age, male sex, current smoking, higher systolic blood
pressure, diabetes, and history of cardiovascular
disease. Diastolic blood pressure had no independent
predictive value in multivariate models, ie, when
systolic blood pressure was taken into account. Several
explanations can be proposed: (1) Selecting people according to a
narrow range of values for a given characteristic decreases the
possibilities of assessing its prognostic value. (2) People in control
groups were to be treated by blood pressure–lowering drugs if their
blood pressure exceeded a given threshold. This procedure is likely to
decrease the apparent prognostic value of blood pressure, but such
reasoning is also applicable to systolic blood pressure. (3)
There was a significant interaction between age and
diastolic blood pressure for the occurrence of
cardiovascular death. This suggested that
diastolic blood pressure was associated with risk in
younger people only. This association weakened with age and disappeared
at 65 years, reversing thereafter. Being a man
represented a higher risk of death, especially from
coronary heart disease death. However, this association was
stronger in the British cohorts, most of all in the UK-MRC1 study. This
finding was not explained by differences in age or other common risk
factors between studies, since they were taken into account. Thus,
further study is required. Possibly, specific risk scores by
geographical area and/or by sex could be established.
Cholesterol, height, and ECG-determined left ventricular hypertrophy were associated with all outcomes except stroke death. A negative association with hemorrhagic stroke24 that masks the positive association with ischemic stroke probably explains the lack of association between cholesterol and stroke.
Potentially modifiable risk factors are generally those of most interest as potential therapeutic targets. However, when risk factors are used to predict the risk of individuals, even those factors that are not modifiable should be taken into account, such as age, sex, and height. The negative correlation between height and cardiovascular mortality has already been described,25 26 although findings have been inconsistent.27 However, the significant qualitative heterogeneity of the association of height with coronary heart disease death we have observed puts into question its use in risk prediction. Epidemiological evidence has usually indicated an increase of cardiovascular risk with body mass index.28 29 In the present study, body mass index was significantly (and negatively) associated only with all-cause death, without significant heterogeneity between trials. Testing the significance of an additional quadratic factor did not support a J- or U-shape relation to explain the lack of association with cardiovascular death.
Cardiovascular death is the most common cause of death among people with end-stage renal failure.30 The independent prognostic value of glomerular filtration rate suggests that the effect of renal failure is not mediated by only the risk factors that were taken into account and that are associated with renal failure, such as age, high blood pressure, and high uric acid level. The hazard ratios associated with the increase in 1 SD of serum uric acid or serum total cholesterol are of the same magnitude. The significant association of serum uric acid with all outcomes raises the question of the possible therapeutic benefit from drugs that decrease its serum concentration. The presence of ECG-defined left ventricular hypertrophy doubled the hazard of coronary heart disease death and cardiovascular death. The prognostic importance of this information, so simple to obtain, was already shown in epidemiological studies, as in the Framingham study.31 It should be included in any risk score that is used for deciding whether to prescribe any drug to prevent cardiovascular events.
The prognostic importance of previous antihypertensive treatment may reflect self-selection (or physician referral) of participants enrolled in clinical trials, ie, an unavoidable selection bias. That bias may also interact with the prognostic importance of blood pressure because of the remaining effect of previous drugs: the blood pressure measured would, in fact, be lower than the "real" blood pressure. However, the association between previous treatment and risk of death is stronger for coronary death than for stroke death, whereas the opposite would be expected if it were only due to a bias in blood pressure measurement. Ethnic origin had some impact on the risk of cardiovascular death but a greater impact on the risk of noncardiovascular death, which may reflect some confounding effect, such as that of socioeconomic status. The impact of ethnic origin was not apparent in the hypertensive participants of the MRFIT study.
Conclusions
We confirmed the prognostic importance of classic
cardiovascular risk factors by analyzing pooled cohorts
from control groups of clinical trials in hypertension. In addition,
the prognostic value of a smaller height, a higher serum uric acid, or
a lower glomerular filtration rate was strongly suggested
and deserves to be explored further. However, quantitative and even
qualitative disparities between cohorts for some risk factors, as seen
elsewhere,32 point to the need to assess whether a
quantified hazard ratio for a given risk factor can be appropriately
used in predicting the risk of people in various settings.
Meta-analyses based on individual patient data from either
observational studies33 or controlled clinical trials,
through multivariate modeling, constitute an
appropriate tool to explore the general value of risk factors.
|
Acknowledgments |
|---|
This study was supported by a grant from the French Ministry of Foreign Affairs and from the Association Régionale de Cardiologie du Rhône. The constitution of the INDANA database was possible thanks to grants from the Association pour la Promotion de la Recherche et de l'Evaluation en Thérapeutique, the Société Française d'Hypertension Artérielle, the Fondation pour la Recherche Médicale, and the Hospices Civils de Lyon.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
1. Alderman MH. Blood pressure management: individualized treatment based on absolute risk and the potential for benefit. Ann Intern Med. 1993;119:329–335.
2. Boissel JP, Collet JP, Lièvre M, Girard P. An effect model for the assessment of drug benefit: example of antiarrhythmic drugs in postmyocardial infarction patients. J Cardiovasc Pharmacol. 1993;22:356–363.[Medline] [Order article via Infotrieve]
3. Jackson R, Barham P, Bills T, Birch T, McLennan L, MacMahon S, Maling T. The management of raised blood pressure in New Zealand. Br Med J. 1993;307:107–110.
4.
Multiple Risk Factor Intervention Trial Research
Group. Multiple Risk Factor Intervention Trial: risk factor changes and
mortality results. JAMA. 1982;248:1465–1477.
5. Medical Research Council Working Party. MRC Trial of treatment of mild hypertension: principal results. Br Med J. 1985;291:97–104.
6. MRC Working Party. Medical Research Council trial of treatment of hypertension in older adults: principal results. Br Med J. 1992;304:405–412.
7. Amery A, Birkenhäger W, Brixko P, Bulpitt C, Clément D, Deruyttere M, De Schaepdryver A, Dollery C, Fagard R, Forette F, Forte J, Hamdy R, Henry JF, Joossens JV, Leonetti G, Lund-Johansen P, O'Malley K, Petrie J, Strasser T, Tuomilheto J, Williams B. Mortality and morbidity results from the European Working Party on High Blood Pressure in the Elderly Trial. Lancet. 1985;i:1349–1354.
8. Coope J, Warrender TS. Randomised trial of treatment of hypertension in elderly patients in primary care. Br Med J. 1986;293:1145–1151.
9.
Hypertension Detection, and Follow-up Program
Cooperative Group. Five-year findings of the Hypertension Detection and
Follow-up Program, I: reduction in mortality of persons with high blood
pressure, including mild hypertension. JAMA. 1979;242:2562–2571.
10.
SHEP Cooperative Research Group. Prevention of stroke
by antihypertensive drug treatment in older persons with isolated
systolic hypertension: final results of the Systolic
Hypertension in the Elderly Program (SHEP). JAMA. 1991;265:3255–3264.
11. Dalhöf B, Lindholm L, Hansson L, Shersten B, Ekbom T, Wester PO. Morbidity and mortality in the Swedish Trial in Old Patients with Hypertension (STOP-Hypertension). Lancet. 1991;338:1281–1285.[Medline] [Order article via Infotrieve]
12.
Perry HM Jr, McFate Smith W, McDonald RH, Black D,
Cutler JA, Furberg CD, Greenlick MR, Kuller LH, Schnaper HW,
Schoenberger JA, Vogt TM, Wolf PA, Hulley SB. Morbidity and mortality
in the Systolic Hypertension in the Elderly Program Pilot
Study. Stroke. 1989;20:4–13.
13. Perry HM, Goldman AI, Lavin MA, Schnaper HW, Fitz AE, Frohlich ED, Steele B, Richman HG. Evaluation of drug treatment in mild hypertension: VA-NHLBI feasibility trial. Ann N Y Acad Sci. 1978;304:267–288.[Medline] [Order article via Infotrieve]
14. Management Committee. The Australian therapeutic trial in mild hypertension. Lancet. 1980;i:1261–1267.
15. Gueyffier F, Boutitie F, Boissel JP, Coope J, Cutler J, Ekbom T, Fagard R, Friedman L, Perry HM, Pocock S, Prineas R, Schron E. INDANA: a meta-analysis on individual patient data in hypertension: protocol and preliminary results. Therapie. 1995;50:353–362.[Medline] [Order article via Infotrieve]
16. Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron. 1976;16:31–41.[Medline] [Order article via Infotrieve]
17. Cox DR. Regression models and life tables (with discussion). J R Stat Soc. 1972;34(series B):187–202.
18. SAS User's Guide, Version 6. 4th ed. Cary, NC: SAS Institute; 1990.
19. Cornu C, on behalf of the TAS-PP Group. Thrombolysis in acute stroke pooling project (TAS-PP): a prospective pooled individual patient database. Control Clin Trials. 1994;15:99S. Abstract.
20. WHO-ISH Blood Pressure Lowering Treatment Trialists' Collaboration. Protocol for prospective collaborative overviews of major randomized trials of blood-pressure-lowering treatments. J Hypertens. 1998;16:127–137.[Medline] [Order article via Infotrieve]
21.
Atrial fibrillation investigators. Risk factors for
stroke and efficacy of antithrombotic therapy in atrial fibrillation:
analysis of pooled data from five randomized controlled trials.
Arch Intern Med. 1994;154:1449–1457.
22.
Glasziou P, Irwig LM. An evidence based approach to
individualising treatment. Br Med J. 1995;311:1356–1359.
23. Levy D, Wilson PWF, Anderson KM. Stratifying the patient at risk from coronary disease: new insights from the Framingham Heart Study. Am Heart J. 1990;119:712–717.[Medline] [Order article via Infotrieve]
24. Neaton JD, Wentworth DN, Cutler J, Stamler J, Kuller L. Risk factors for death from different types of stroke: Multiple Risk Factor Intervention Trial Research Group. Ann Epidemiol. 1993;3:493–499.[Medline] [Order article via Infotrieve]
25.
Rich-Edwards JW, Manson JE, Stampfer MJ, Colditz GA,
Willett WC, Rosner B, Speizer FE, Hennekens CH. Height and the risk of
cardiovascular disease in women. Am J
Epidemiol. 1995;142:909–917.
26. Hebert PR, Rich-Edwards JW, Manson JE, Ridker PM, Cook NR, O'Connor GT, Buring JE, Hennekens CH. Height and incidence of cardiovascular disease in male physicians. Circulation. 1993;88(pt 1):1437–1443.
27.
Kannam JP, Levy D, Larson M, Wilson PWF. Short stature
and risk for mortality and cardiovascular disease
events: the Framingham Heart Study. Circulation. 1994;90:2241–2247.
28. Manson JE, Colditz GA, Stampfer MJ, Willett WC, Rosner B, Monson RR, Speizer FE, Hennekens CH. A prospective study of obesity and risk of coronary heart disease in women. N Engl J Med. 1990;322:882–889.[Abstract]
29.
Rimm EB, Stampfer MJ, Giovannucci E, Ascherio A,
Spiegelman D, Colditz GA, Willett WC. Body size and fat distribution as
predictors of coronary heart disease among middle-aged and
older US men. Am J Epidemiol. 1995;141:1117–1127.
30. Kasiske BL, Guijarro C, Massy ZA, Wiederkehr MR, Ma JZ. Cardiovascular disease after renal transplantation. J Am Soc Nephrol. 1996;7:158–165.[Abstract]
31. Levy D. Left ventricular hypertrophy: epidemiological insights from the Framingham Heart Study. Drugs. 1988(suppl 5):1–5.
32. Menotti A, Keys A, Blackburn H, Kromhout D, Karvonen M, Nissinen A. Comparison of multivariate predictive power of major risk factors for coronary heart diseases in different countries: results from eight nations of the Seven Countries Study, 25 year follow-up. J Cardiovasc Risk. 1995;3:69–75.
33. The Pooling Project Research Group. Relationship of blood pressure, serum cholesterol, smoking habit, relative weight and ECG abnormalities to incidence of major coronary events: final report of the Pooling Project. J Chronic Dis. 1978;31:201–306.[Medline] [Order article via Infotrieve]
This article has been cited by other articles:
 |
|
 |
|
  P. Verdecchia, G. Carini, A. Circo, E. Dovellini, E. Giovannini, M. Lombardo, P. Solinas, M. Gorini, A. P. Maggioni, and the MAVI Study Group Left ventricular mass and cardiovascular morbidity in essential hypertension: the MAVI study J. Am. Coll. Cardiol., December 1, 2001; 38(7): 1829 - 1835. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. P. Greenwood, E. M. Scott, J. B. Stoker, and D. A. S. G. Mary Hypertensive left ventricular hypertrophy: relation to peripheral sympathetic drive J. Am. Coll. Cardiol., November 15, 2001; 38(6): 1711 - 1717. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Laurent, P. Boutouyrie, R. Asmar, I. Gautier, B. Laloux, L. Guize, P. Ducimetiere, and A. Benetos Aortic Stiffness Is an Independent Predictor of All-Cause and Cardiovascular Mortality in Hypertensive Patients Hypertension, May 1, 2001; 37(5): 1236 - 1241. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||