About the Phenotypic Data Analyzed
We thank the Framingham Investigators for their careful collection of Framingham Heart Study (FHS) participant medical data and making the limited access data set available to us for our analysis.
Questions about the FHS study or the limited access data set should be directed to the Framingham Investigators.
Data on quantitative traits collected over 6 separate exams of the Framingham Heart Study (FHS) Offspring Cohort were analyzed as part of our study. Quantitative traits examined include body mass index, fasting blood glucose, plasma cholesterol, QT-interval and hypertension among others.
This phenotypic data, termed the "limited access data set", is available to all IRB-approved scientists. More information about requesting the limited access phenotypic dataset as well as the genotypes generated by the 100K Genome Scan can be found at http://www.nhlbi.nih.gov/about/framingham/policies/index.htm
About the Framingham Heart Study
The FHS study began in 1948 with the enrollment of two-thirds of the adult population of the town of Framingham, MA, a small town 15 miles west of Boston. The dedicated participants in the study return for an all day examination every few years during which measurement of heart function, lung function, blood chemistries and a variety of life style habits are collected.
Framingham Heart Study Design
High quality phenotypic data is essential for genetic association studies. For example, blood pressure measurements show considerable variation from exam to exam; this can lead to imprecise estimates of the genetic effect of a particular polymorphism. An advantage of the Framingham dataset is that it has been collected at a single center with frequent review by clinicians.
The data has also been collected over six Offspring Exam cycles. When such "longitudinal" data is used, estimates of the heritability of many traits increases and so does the power to identify the modest effects that single genes are likely to exist for complex diseases. Further, during each exam, covariate data was also collected, enabling adjustment for differences in participant exposures and exploration of gene-environment interactions.
Framingham Heart Study BMI Measurements by Exam
Phenotypic Data Cleaning
For phenotypes with well-established covariates and a normal distribution, linear regression was performed, first to remove influential outliers and then to calculate adjusted residuals. A standard box-plot was used to remove those residuals that are 1.5 times the interquartile range of the distribution. For phenotypes that are directly calculated, e.g. BMI, or insulin resistance, phenotypic values that are 1.5 times the interquartile range of the distribution were also removed. Here removal of outliers increases our ability to detect SNPs with modest effects in the association analysis, which is regression-based.