STATISTICAL PROBLEMS IN THE GENETICS OF DIABETES

  • Hanis, Craig L, (PI)

Project: Research project

Description

Genetic and epidemiological studies of non-insulin dependent diabetes
mellitus (NIDDM) are complicated by several factors; these include: 1)
diagnostic criteria based on truncation of continuous distribituion (i.e.,
diagnosis based on elevated fasting blood glucose on more than one
occasion), 2) ascertainment of families based on diabetic probands, 3)
etiological heterogeneity, and 4) association with other traits such as
obesity. In response to these so called "pitfalls", the aims of this
proposal are to develop statistical methodologies which explicitly
incorporate these issues and then to apply the results to pedigree data on
NIDDM obtained from Starr County, Texas; a population which is 97%
Mexican-American and characterized by age and sex specific prevalences of
NIDDM which are three- to five-fold higher than the U.S. population at
large. Accomplishement of the aims of this proposal will required
extension of statistical theory, implementation of computing algorithms and
programs, evaluation of the theory numerically, and application to real
data. This research will permit evaluation of the sources of familial
aggregation of NIDDM, coaggregation of NIDDM and obesity, and
heterogeneity. Failure to explicitly account for the statistical issues
posed by diagnostic criteria based on cutpoints of a continuous
distribution and ascertainment of families when such is the case can lead
to inferences of genetic and/or environmental etiology which are artifacts
of inadequate correction and do not reflect underlying biological
processes. For example, ascertainment of nuclear families having one child
whose phenotypic value is above some cutpoint will lead to significant
spouse pair correlations when in fact none may exist. This can result in
inferences of common environmental effects or assortative mating which are
functions of sampling and not biology. Furthermore, these issues are
significant in that they also characterize studies of most common complex
diseases of man which aggregate in families; indeed, the Lipid Research
Clinic and several Specialized Centers of Research employ truncated
sampling. Also, diseases besides diabetes are defined by arbitrary
cutpoints including obesity and hypertension.
StatusFinished
Effective start/end date7/1/836/30/86

Funding

  • National Institutes of Health

Fingerprint

Diabetes
Diabetes Mellitus
Obesity
Dependent
Diagnostics
Pedigree
Hypertension
Evaluation
Lipids
Glucose
Truncation
Biology
Blood
Fold
Family
Methodology
Computing

Keywords

  • Medicine(all)