Fragile X Premutation Pilot Study of Infertile Women
American Journal of Epidemiology
Conventional case-control analyses using single nucleotide polymorphisms (SNPs) are useful in looking for single gene associations, however studying gene interactions becomes inefficient as the number of SNPs increase. Application of more complex statistical methods allows for the analysis of many SNPs and covariates simultaneously. To study sickle cell associated vasoocclusive events (VOEs), 1,489 patients were genotyped for 353 SNPs in 160 genes. We identified 542 patients with at least one
... E and the remaining patients served as controls. CART and SGB was run on a random sample of 80% of the patients to identify genes and covariates whose interactions characterize patients with VOEs and the accuracy of that model assessed using the remaining patients as a test set. Along with age, sex and fetal hemoglobin, both CART and SGB identified KL and genes of the TGF ß/BMP pathway as being associated with the VOEs. CART was able to correctly classify 65% of patients while SGB, was able to classify 71%. When used in concert however (SGB followed by CART) the reduced model was now able to correctly classify 81% of patients and showed a marked improvement in sensitivity and specificity. While, typically, SGB provides a more accurate method of classification, it does not provide a simple graphical model as does CART. We found that using SGB in conjunction with CART retains the best of both methods; it provides a model that predicts VOEs with a good degree of accuracy and a visual representation of the interactions among genes and covariates. Genes that code for constituents of the dopaminergic neurotransmitter system have been suggested as candidates for involvement in addictive behaviors, including smoking. We hypothesized that polymorphisms associated with reduced dopaminergic neurotransmission would be more common in continuing smokers than among women who quit smoking. The study included 593 women aged 26-65 years who participated in a 12-month randomized controlled smoking cessation trial conducted in Seattle, Washington in 1993-1994. Subjects were recontacted approximately 3 years after the trial to obtain updated smoking history and biological specimens. We assessed polymorphisms involved in dopamine synthesis (TH), receptor activation (DRD2, DRD3, DRD4), reuptake (SLC6A3), and metabolism (COMT). We computed the relative risk (RR) of smoking cessation at the end of the trial ("short-term" quitting, defined as not smoking within the 7 previous days), and at a subsequent interview several years later ("longterm" quitting, defined as not smoking within the previous 6 months). We observed no association of any of the polymorphisms studied with the likelihood of quitting smoking, for either short-term or long-term smoking cessation. Although, in a few instances, the magnitude of the RRs could be considered consistent with weak associations with quitting, either the direction of effect was opposite of that hypothesized or results of the shortterm and long-term cessation endpoints were inconsistent, reducing the plausibility of a causal association. The results of this study fail to support the findings of prior studies that reported associations of variation in genes involved in dopaminergic neurotransmission with smoking status. Alcoholism and alcohol drinking habits are partly genetically determined, however, exact genes are largely unknown. Because alcohol is degraded mainly by liver alcohol dehydrogenase (ADH), we hypothesized that well known functional genetic variation in ADH2 and ADH3 genes predict alcoholism and alcohol drinking habits. This is biologically plausible because the ADH2Á2 versus the ADH2Á1 allele imply a 38 fold maximal alcohol degradation rate and because the ADH3Á1 versus the ADH3Á2 allele imply a 2.5 fold maximal alcohol degradation rate. By genotyping 9080 Caucasians we found that individuals with ADH2 slow versus fast alcohol degradation had a 34% higher alcohol intake, and had odds ratios for daily drinking on 2.5 (95% confidence limits: 1.5; 4.