An Analysis of Using Pedigrees in Family Based Studies and an Exploration of Cancer Risk and Cancer Resistance UsingTwin Studies
MetadataShow full item record
AbstractIn the first section of this thesis, we explore the use of family pedigrees in association analysis. Family pedigrees were successfully used in linkage analysis to discover many genes for Mendelian traits, but less successful for identifying genes for complex diseases. The family-based association test (FBAT) can be used to test for association in pedigrees by two methods, conditioning on the parents to get the null distribution of the offsprings’ genotypes separately for each nuclear family and then combining over families, or condi- tioning on the founders in the pedigree to get the offsprings’ genotypes. In this study, we use simulations to compare the power of conditioning on the founders or parents when using the FBAT statistic to test for association in the family pedigree.We consider two scenarios where the disease outcome is represented as a simple Mendelian trait and the disease outcome is modeled as more complex as multiple factors influence the disease.
Two new results were found in our simulation study. Under the first assumption of a Mendelian disease outcome, conditioning on the founders is slightly more powerful than conditioning on the parents for detecting association. For complex diseases, the power of all of the ascertainment methods were reduced considerably, but using multiplex pedi- grees were still more powerful than trios and sib pairs when the recurrence risk ratio was between family members was relatively low. In the second part of this thesis we explore the use of twin studies in assessing cancer risk and cancer resistance. Twin studies provide unique information about familial risk of cancer. Typically, twin studies are used to quantify heritability. However, clinical application of the insight gained from twin studies needs to rely on estimates of absolute risk, as these are directly relevant for both individual and population-level decision making. We provide estimates of risk using the Nordic Twin Studies of Cancer (NorTwinCan), estimate risk ratios which can be applied to estimate the risks in a population with a different baseline risk, and compare methods of calculating the risks. Our results suggest that both models provide slightly different absolute risk estimates compared to the empirical estimates, which is not surprising since these models rely on differing modeling assumptions and condition on different covariates. While estimates of heritability of can- cer are high, for an unaffected individual the implications of having an affected co-twin remain relatively contained, and additional family history should continue to play a role in counseling and decision making.
In the last section, we explore cancer resistance using twin studies. It has been hypothesized that some individuals have a decreased risk of cancer, or a cancer resistance, resulting from genetic predisposition. Using simulation studies and twin study data, we explored the question of whether these studies can also be used to investigate a genetic predisposition to avoiding cancer. We first conduct simulations to assess the impact such a genetic predisposition would have on the proportions of cancer concordant MZ and DZ twins, we postulated a simple model wherein a fraction of the population carries an inherited and extreme resistance to cancer, and developed a likelihood-based approach to estimate this prevalence. We then applied our approach to the Nordic Twin Studies of Cancer (NorTwinCan), a cohort of over 200,000 individual twins from Sweden, Norway, Denmark and Finland. We estimate the prevalence of the ”cancer resistance” genotypes as 1.7% (95% C.I.: 1.2, 2.1%) in this population. These results are obtained using the fol- lowing assumptions; 1) all cancers are considered together, 2) the resistance genotype is fully penetrant, and 3) distributions for age of onset of cancer, censoring, and death are fixed. We do, however, provide a general framework under which these assumptions can be relaxed. Our results suggest that predisposition to avoiding cancer may be heritable, and that twin data can provide information on this hypothesis, and that the largest twin studies allow for quantitative exploration of genetic parameters.
Citable link to this pagehttp://nrs.harvard.edu/urn-3:HUL.InstRepos:37944991
- FAS Theses and Dissertations