Framingham Heart Study Data Research Articles

A genome-wide scan to identify loci for smoking rate in the Framingham Heart Study population.

BackgroundAlthough many years of genetic epidemiological studies have demonstrated that genetics plays a significant role in determining smoking behavior, little information is available on genomic loci or genes affecting nicotine dependence. Several susceptibility chromosomal regions for nicotine dependence have been reported, but few have received independent confirmation. To identify susceptibility loci for nicotine dependence, 313 extended pedigrees selected from the Framingham Heart Study population were analyzed by both the GENEHUNTER and S.A.G.E. programs.ResultsAfter performing linkage analyses on the 313 extended Framingham Heart Study families, the EM Haseman-Elston method implemented in GENEHUNTER provided evidence for significant linkage of smoking rate to chromosome 11 and suggestive linkage to chromosomes 9, 14, and 17. Multipoint sib-pair regression analysis using the SIBPAL program of S.A.G.E. on 1389 sib pairs that were split from the 313 extended families identified suggestive linkage of smoking rate to chromosomes 4, 7, and 17. Of these identified positive regions for nicotine dependence, loci on chromosomes 7, 11, and 17 were identified by both GENEHUNTER and S.A.G.E. programs.ConclusionOur genome-wide scan results on the Framingham Heart Study data provide evidence for significant linkage of smoking rate to chromosome 11 and suggestive linkage to chromosomes 4, 7, 9, 14, and 17. These findings suggest that some of these regions may harbor susceptibility loci for nicotine dependence, and warrant further investigation in this and other populations.

Group 6: Pleiotropy and multivariate analysis.

Analysis techniques using data on several traits simultaneously allow researchers to dissect the genetic architecture affecting correlated traits, and can increase the power to detect pleiotropic genes, i.e., genes that influence two or more traits. Several of the papers in this group from Genetic Analysis Workshop 13 presented promising univariate summaries of multiple traits that detected linkage signals that standard single-trait univariate methods did not. Other papers found linkage signals using multivariate techniques that univariate techniques missed, although this was not uniformly the case. Some papers also considered the correlation among measurements of a single trait taken at different ages to assess whether the genetic architecture of the trait changed over age. Applications of the Framingham Heart Study data identified major loci jointly influencing body mass index and high-density lipoprotein cholesterol, high-density lipoprotein cholesterol and triglycerides, total cholesterol and triglycerides, and various combinations of four traits involved in metabolic syndrome.

Effects of covariates: a summary of Group 5 contributions.

This report summarizes the contributions of Genetic Analysis Workshop 13 (GAW13) related to the use of covariates in genetic analysis. Seven papers are summarized, five of which analyzed the Framingham Heart Study Data, and two the simulated data. Five papers examined the role of covariates in linkage analysis, using a variety of statistical approaches including affected sibling pair analysis, conditional logistic regression, and variance components methods. One paper examined the impact of covariates on family-based association analysis. In each of these papers, the detection of genetic effects could be influenced by the incorporation of covariates. The final paper examined the role of transmission ratio distortion in the analysis of complex traits and the role of covariates in the variability in transmission ratio distortion. While each paper takes a different approach to the genetic analysis of complex traits, a common thread running through each is that the inclusion of covariates can have a substantial impact on the results of the analysis. Care must be taken to understand how the covariates are being used in each analysis, what assumptions are being made, and how these assumptions might affect the results and their interpretation. Finally, the results of Group 5 studies show that inclusion of covariates can increase the power to detect genes for complex traits, and has the potential to advance an understanding of the role of genes in these complex traits.

Nonparametric longitudinal allele-sharing model.

Basically no methods are available for the analysis of quantitative traits in longitudinal genetic epidemiological studies. We introduce a nonparametric factorial design for longitudinal data on independent sib pairs, modelling the phenotypic quadratic differences as the dependent variable. Factors are the number of alleles shared identically by descent (IBD) and the age categories at which the dependent variable is measured, allowing for dependence due to age. To identify a linked marker a rank statistic tests the influence of IBD group on phenotypic quadratic differences. No assumptions are made on normality or variances of the dependent variable. We apply our method to 71 sib pairs from the Framingham Heart Study data provided at the Genetic Analysis Workshop 13. For all 15 available markers on chromosome 17 we analyzed the influence on systolic blood pressure. In addition, different selection strategies to sample from the whole data are discussed.

Genome scan linkage results for longitudinal systolic blood pressure phenotypes in subjects from the Framingham Heart Study.

The relationship between elevated blood pressure and cardiovascular and cerebrovascular disease risk is well accepted. Both systolic and diastolic hypertension are associated with this risk increase, but systolic blood pressure appears to be a more important determinant of cardiovascular risk than diastolic blood pressure. Subjects for this study are derived from the Framingham Heart Study data set. Each subject had five records of clinical data of which systolic blood pressure, age, height, gender, weight, and hypertension treatment were selected to characterize the phenotype in this analysis.We modeled systolic blood pressure as a function of age using a mixed modeling methodology that enabled us to characterize the phenotype for each individual as the individual's deviation from the population average rate of change in systolic blood pressure for each year of age while controlling for gender, body mass index, and hypertension treatment. Significant (p = 0.00002) evidence for linkage was found between this normalized phenotype and a region on chromosome 1. Similar linkage results were obtained when we estimated the phenotype while excluding values obtained during hypertension treatment. The use of linear mixed models to define phenotypes is a methodology that allows for the adjustment of the main factor by covariates. Future work should be done in the area of combining this phenotype estimation directly with the linkage analysis so that the error in estimating the phenotype can be properly incorporated into the genetic analysis, which, at present, assumes that the phenotype is measured (or estimated) without error.

Lack of reproducibility of linkage results in serially measured blood pressure data.

BackgroundUsing the longitudinal Framingham Heart Study data on blood pressure, we analyzed the reproducibility of linkage measures from serial cross-sectional surveys of a defined population by performing genome-wide model-free linkage analyses to systolic blood pressure (SBP) and history of hypertension (HTN) measured at five separate time points.ResultsThe heritability of SBP was relatively stable over time, ranging from 11.6 to 23.5% (coefficient of variation = 25.7%). However, the variability in linkage results was much greater. The average correlation in LOD scores at any pair of time points was 0.46 for HTN (NPL All LOD) and 0.17 for SBP (Variance Components LOD). No evidence of reproducible linkage results was found, with a mean κ of 0.02 for linkage to HTN and -0.03 for SBP linkage. At loci with potential evidence for linkage (LOD > 1.0 at one or more time points), the correlation was even lower. The coefficient of variation at loci with potential evidence of linkage was 126% for HTN and 135% for SBP. None of 15 chromosomal regions for HTN and only one of 28 regions for SBP with potential evidence for linkage had a LOD > 1.0 at more than two of the five time points.ConclusionThese data suggest that, although heritability estimates at different time points are relatively robust, the reproducibility of linkage results in serial cross-sectional samples of a geographically defined population at successive time points is poor. This may explain in part the difficulty encountered in replicating linkage studies of complex phenotypes.

Cardiovascular risk assessment based on US cohort studies: findings from a National Heart, Lung, and Blood institute workshop.

This report was derived from a workshop on cardiovascular risk assessment sponsored by the National Heart, Lung, and Blood Institute, which addressed whether risk equations developed in the Framingham Heart Study (FHS) for predicting new-onset coronary heart disease (CHD) apply to diverse population groups. Preparation for the workshop included a reanalysis and comparison of prospective studies in several different populations in which risk factors were related to cardiovascular outcomes. Some studies included fatal and nonfatal CHD end points, whereas others contained only CHD mortality. Extensive collaboration provided as much uniformity as possible with respect to both risk factors and CHD end points. The FHS has led in defining the quantitative impact of risk factors.1 Many potential risk factors were measured and related to cardiovascular outcomes. Several risk factors proved to be strong, largely independent predictors of cardiovascular disease (CVD). These factors—advancing age, cigarette smoking, blood pressure (particularly systolic), cholesterol in total serum and HDL, and diabetes—served as the basis for the development of risk prediction equations.1 If FHS risk estimates are to be widely used, they must apply widely in the US population. To document their transportability, they must be compared with prospective studies in other populations. Although the FHS is the longest running prospective study, there are other major studies. The cardiovascular end points of these other studies have varied. Some include cardiovascular morbidity and mortality; others have only cardiovascular mortality. Among the end points, CHD is the most extensively reported; for this reason, CHD was the primary focus of the workshop. ### Multivariate Relative Risk Comparisons In preparation for the workshop, multivariate regression coefficients for each risk factor were compared in different populations with those of the FHS. Adjusted relative risk estimates make it possible to determine whether each independent risk factor confers a similar or different relative risk among different …

Modeling Multivariate Discrete Failure Time Data

A bivariate discrete survival distribution that allows flexible modeling of the marginal distributions and yields a constant odds ratio at any grid point is proposed. The distribution can be extended to a multivariate distribution and is readily generalized to accommodate covariates in the marginal distributions and pairwise odds ratios. In addition, a pseudo-likelihood estimation procedure for estimating the regression coefficients in the marginal models and the association parameters in the pairwise odds ratios is presented. We evaluate the performance of the proposed estimation procedure through simulations. For bivariate data, pseudo-likelihood estimation of the association parameter has high efficiency. Loss of efficiency in the marginal regression coefficient estimates is small when the association is not strong. For both the marginal regression coefficients and the association parameter, coverage probabilities are close to the 95% nominal level. For multivariate data, the simulation results show that the parameter estimates are consistent. Coverage probability for the regression coefficient in the marginal model is close to the 95% nominal level but is slightly less than the nominal level for the association parameter. We illustrate the proposed methods using a subset of the Framingham Heart Study data where a significant positive association was found between the failure times of siblings.

The Influence of Cigarette Smoking on the Association between Body Weight and Mortality. The Framingham Heart Study Revisited

PURPOSE: To calculate for two measures of obesity, the Metropolitan Relative Weight (MRW) and body mass index (BMI), the value at which minimum mortality occurs. This was done to retest the hypothesis, in the Framingham Heart Study data, that the association between obesity and mortality can be obscured by an interaction between the measure of obesity and smoking. In the original analysis of the Framingham data it was suggested that there was a U- or J-shaped relationship between MRW and death in smokers but a linear relationship in nonsmokers. The design and setting were those of the NHLBI Framingham Heart Study. METHODS: The 5209 members of the Framingham Heart Study underwent a baseline examination in 1948–1952 (Exam 1) and they were reexamined at approximately two-year intervals over a 30-year period. The study included both men n = 2336 and women n = 2873 in the age range of 28 to 62 years. After excluding persons with missing baseline data, the analytic sample size was 5163. Additional analyses were conducted by deleting persons with cardiovascular disease (CVD) at baseline n = 135 , the sample used by the original paper by Garrison and colleagues, and persons who died within the first four years of follow-up n = 62 . The main outcome measures consisted of thirty-year survival through Exam 16, approximately in 1980, as influenced by MRW or BMI, age, and smoking status at baseline (Exam 1). RESULTS: We were able to show that the sample sizes of male nonsmokers were too small to test the hypothesis within age groups < 40 and 40–49 years. In men ages 50–62 there was a significant age-adjusted quadratic relationship between BMI or MRW, and risk of death. The estimated BMI at the minimum risk of death for smokers (24.5) and nonsmokers (23.8) were not statistically different. Identical results were found for MRW (minimum: smokers = 112.5, nonsmokers = 111.4). In men and women ages 28–62 there appeared to be a u- or j-shaped relationship between the 30-year crude mortality rate and MRW. After excluding persons with missing data, CVD at baseline, and persons who died within the first four years of follow-up, the age adjusted estimated BMI value at the minimum risk of death was nearly identical for men and women and for smokers and nonsmokers (Men: smokers = 22.8, nonsmokers = 22.8; Women: smokers = 22.9, nonsmokers = 23.3). Additionally, the estimates of the minimum were always below the mean. Identical results were found without deleting persons with CVD at baseline and deaths in the first four years of follow-up. Identical results were found for MRW. CONCLUSIONS: Reanalysis of the Framingham Heart Study data does not support the hypothesis that there is an interaction between smoking and measures of obesity. Moreover, the estimated BMI or MRW at the minimum risk of death was similar for men and women smokers and nonsmokers alike even after deleting prevalent cases of CVD and deaths within the first four years of follow-up.

A predictive model of the health benefits and cost effectiveness of celiprolol and atenolol in primary prevention of cardiovascular disease in hypertensive patients.

This study compares the antihypertensive and lipid modifying effects of treatment of mild to moderate hypertension with celiprolol or atenolol. It also models the 5-year cardiovascular risk reduction and the cost effectiveness of monotherapy from a partial societal perspective. The effects of celiprolol and atenolol on systolic blood pressure (SBP), total serum cholesterol (TC) and high density lipoprotein cholesterol (HDL-C) were obtained from a pooled analysis of published studies. Although celiprolol and atenolol had similar effects on SBP, celiprolol reduced the ratio of TC to HDL-C by 10.2% [95% confidence intervals (95% CI) -16.4%, -4.0%) but atenolol increased the ratio by 7.7% (95% CI of 3.4%, 12.0%). The 5-year absolute risks of an initial coronary or cerebrovascular event or cardiovascular death were computed for cohorts of patients treated with either agent or remaining untreated, using an accelerated failure time (AFT) model, based on Framingham Heart Study data. Inputs to the model were age, gender, smoking status, SBP, TC and HDL-C. The change in absolute risk was estimated using the changes in SBP and TC: HDL-C obtained from the pooled analysis. Average life-months gained by therapy were computed as differences between the Kaplan-Meier survival curves estimated from the model plus differences in 5-year cardiovascular death rates multiplied by average life expectancy obtained from life tables. Direct medical costs included drug treatment, and the costs of acute care for initial coronary and cerebrovascular events deferred by therapy over the 5-year treatment period. The model shows that in the lowest-risk base case (60-year-old men who are nondiabetic and nonsmokers with SBP of 160 mm Hg and a 5-year absolute cardiovascular risk of 12%), celiprolol (271 mg/day) is 2-fold more effective than atenolol (77.4 mg/day) in reducing coronary event risk, and equally effective in reducing cerebrovascular event risk. The number of individuals that would have to be treated for 5 years to avoid 1 coronary event is about 30 for celiprolol versus 70 for atenolol. Therapy with celiprolol yields more life-months and at current prices, the cost per life-year gained by therapy is significantly lower. Both drugs are cost effective by international standards in the treatment of patients with 5-year absolute cardiovascular risk greater than 10%, and are more cost effective in those patients at higher levels of absolute cardiovascular risk. The direct medical costs of treatment for 5 years with celiprolol are the same or slightly less than treatment with atenolol at the dosages used in the clinical trials, despite a 19% higher tablet price. Both drugs are more cost effective in patients at higher levels of absolute cardiovascular risk. These findings are sensitive to the drug dosages, tablet prices and the discount rate. Based on epidemiological and clinical data, replacing atenolol with celiprolol in patients with mild to moderate hypertension, but without overt cardiovascular disease, is predicted to have similar effects on stroke risk, but to be substantially more effective in reducing the risk of coronary events at no additional direct medical cost over a 5-year treatment period.

Instrumental Variable Estimation in Generalized Linear Measurement Error Models

Abstract Instrumental variable estimation in generalized linear measurement error models are studied. For models with canonical link functions, unbiased estimating equations are derived. The maximum likelihood estimator for the normal theory, structural linear instrumental variable model is shown to be a solution to the estimating equations derived herein. Logistic regression is studied in detail. An example is given and a simulation study described for the logistic model based on the Framingham Heart Study data.

Cost-Effectiveness of Pravastatin in Secondary Prevention of Coronary Artery Disease

This study analyzed the cost-effectiveness of pravastatin in secondary prevention of coronary artery disease (CAD). The projected risk model in 445 male patients with established CAD and moderately elevated serum low-density lipoprotein cholesterol used results data from 2 placebo-controlled plaque regression trials: Pravastatin Limitation of Atherosclerosis in the Coronary Arteries and Pravastatin, Lipids, and Atherosclerosis in the Carotids. Framingham Heart Study data were used to project the risk of mortality 10 years after myocardial infarction (MI) for incremental male patients in the placebo group who had MI. A Markov process was used to estimate life-years saved, and decision analysis was used to estimate cost. Depending on the patient-risk profile, the midrange estimated cost per life-year saved with pravastatin in secondary prevention of CAD varied from $7,124 to $12,665, which is favorable compared with other widely accepted medical interventions. (Am J Cardiol 1996;78:409–414)

Maximum Likelihood Estimation for Interval-Censored Data Using a Weibull- Based Accelerated Failure Time Model

The accelerated failure time regression model is most commonly used with right-censored survival data. This report studies the use of a Weibull-based accelerated failure time regression model when left- and interval-censored data are also observed. Two alternative methods of analysis are considered. First, the maximum likelihood estimates (MLEs) for the observed censoring pattern are computed. These are compared with estimates where midpoints are substituted for left- and interval-censored data (midpoint estimator, or MDE). Simulation studies indicate that for relatively large samples there are many instances when the MLE is superior to the MDE. For samples where the hazard rate is flat or nearly so, or where the percentage of interval-censored data is small, the MDE is adequate. An example using Framingham Heart Study data is discussed.

Presentation The negative association in women between cigarette smoking and uncomplicated angina pectoris in the Framingham heart study data

As a result of a more complete analysis of available Framingham Heart Study data on the association of cigarette smoking to incidence of uncomplicated angina pectoris than has previously been published in Framingham papers, it was determined that there is a consistently negative risk factor association for women at every follow-up period from 12 to 30 years, with Framingham women cigarette smokers experiencing 30–40% lower rates of uncomplicated angina pectoris than non-smokers, the rates also declining with increasing amounts of cigarettes smoked. Our analysis confirms the absence of a predictive risk factor association in the case of men previously indicated by Framingham investigators. The published declarations by Framingham investigators of an absence of association between cigarette smoking and angina pectoris in Framingham women, is a failure to publicly recognize the distinctive negative association present in the Framingham female data. The Framingham data on the relationship of smoking to angina incidence is clearly at variance with the Surgeon General's sketchy finding of an inconsistent positive association for men and an uncertain relationship for women. It is suggested that special attention should be directed to these results of the Framingham data because of the preeminence of Framingham material world wide, because angina pectoris is the most common manifestation of coronary heart disease, because it will improve the “conventional wisdom” on this subject, and because the negative relationship found for Framingham women does not lend support to the belief held by some that smoking enhances the degree of cornonary atherosclerosis.

Relation of pooled logistic regression to time dependent cox regression analysis: The framingham heart study

A standard analysis of the Framingham Heart Study data is a generalized person-years approach in which risk factors or covariates are measured every two years with a follow-up between these measurement times to observe the occurrence of events such as cardiovascular disease. Observations over multiple intervals are pooled into a single sample and a logistic regression is employed to relate the risk factors to the occurrence of the event. We show that this pooled logistic regression is close to the time dependent covariate Cox regression analysis. Numerical examples covering a variety of sample sizes and proportions of events display the closeness of this relationship in situations typical of the Framingham Study. A proof of the relationship and the necessary conditions are given in the Appendix.

Left ventricular hypertrophy

Left ventricular hypertrophy caused by cardiovascular disease is a marker as well as a contributing factor for cardiovascular morbidity and mortality. Ventricular tachycardia, complex ventricular arrhythmias, and silent ischemia are more prevalent in patients with left ventricular hypertrophy than in those without. Echocardiography is more sensitive and more specific than electrocardiography in diagnosing the entity. Treatment of the underlying disorder may cause regression of left ventricular hypertrophy which, Framingham Heart Study data suggests, may reduce the incidence of cardiovascular events.

Epidemiology of diabetes mellitus in the elderly: The Framingham study

Factors potentially associated with adult-onset diabetes mellitus in the elderly, were reviewed, using the Framingham Heart Study data and other population data. Incidence data for diabetes mellitus in the elderly are sparse, but they indicate that the prevalence of diabetes increases greatly with age. Prevalence rates commonly exceeded 10 percent in those over the age of 60. Men and women in the Framingham Study who were overweight by more than 40 percent had twice the prevalence of diabetes mellitus compared with those of normal weight. Hypertension and coexistent vascular disease were particularly common in elderly diabetic patients, with rates markedly greater than those found among younger adult-onset diabetic patients. In those 50 years of age or older, the later development of diabetes mellitus was associated with increased levels of very-low-density Ilpoprotein cholesterol, decreased levels of high-density lipoprotein cholesterol, obesity, elevated casual glucose levels, use of diuretics, and preexisting vascular disease.

Heart Disease: The MRFIT Test

Pc;)cO T is now widely known that inconclusive results were obtained from the largest trial ever carried out to examnine the effect on mortality of multiple intervention on the major risk factors for cardiovascular disease-the Multiple Risk Factor Intervention Trial (MRFIT) (i). k *4.? The potential implications of MRFIT's generally "negative" results for public health practice are such that it is fundamental to determiine if these results constitute an indictment against primary prevention of cardiovascular disease or if, alternatively, they only reflect the trial's important-and by and large unexpected-methodologic problems. The Multiple Risk Factor Intervention Trial was the outcome of lengthy deliberations and a pilot study supported by the then National Heart and Lung Institute, which suggested that cost and sample size considerations made it impossible to test experimentally the hypothesis that serum cholesterol and fat in the diet cause heart attacks. Instead, the decision was made to carry out a large trial to answer the question: do "high risk" individuals undergoing simultaneous intervention on cigarette smoking, high cholesterol, and hypertension have a better survival than comparable individuals in whom no such intensive intervention is carried out? Hence, about 12,800 mostly white, "high risk" men aged thirty-five to fifty-seven years were randomized to either a "special intervention" group (SI) in whom an intensive effort was made to lower the three major risk factors, or to a "usual care" group (UC), referred to their doctors for treatment, but who were otherwise not meant to undergo an intensive intervention effort. "High risk" was determined on the basis of the upper is%-later changed to io%-of a risk score distribution obtained from the Framingham Heart Study data. The net effect of these eligibility criteria was inclusion of individuals who in general had at least two of the three major risk factors.

Coffee Drinking and Myocardial Infarction

To the Editor.— Much of the letter by Jick et al concerns our attempt to explain the variation among studies. As our Table 4 clearly indicated, the listing of proportions of heavy coffee-drinkers in various studies was made without precise equilibration of the number of cups of coffee drunk per day. Many persons who answered yes to the question Have you usually, in the past year, drunk more than cups of coffee per day? may have imprecisely interpreted the query as six or morecups per day? Therefore, we thought that the comparison was appropriate. However, we now agree that a strict interpretation categorizes our positive responders as drinking seven or more cups of coffee per day. Such categorization is closer to Paul's1200 or more cups per month group that included only 7% of controls. The new Framingham Heart Study data, presented by Jick et al, suggest