Traditional Multivariate Logistic Regression Research Articles

Abstract 2176: Obesity and risk of colorectal cancer: A Mendelian randomization study

Abstract Background: Obesity has been consistently linked to higher risk of colorectal cancer (CRC) in men, whereas the association is generally lower in women. Although these findings come from well conducted epidemiologic studies, it is difficult to exclude all other explanations for associations in observational studies, such as measurement error and other forms of bias, reverse causality and confounding. Mendelian randomization studies, using instrumental variables (IVs), can overcome some of the inherent limitations of observational studies and provide an unbiased and unconfounded estimate of the causal association between an exposure and outcome. Objective: We used genetic variants that are associated with body mass index (BMI) or waist-hip ratio (WHR) to examine the causal association between obesity and CRC. Methods: We used epidemiologic and genetic data from 10 226 CRC cases and 10 286 controls of European ancestry from the Genetics and Epidemiology of Colorectal Cancer Consortium (GECCO) or the Colon Cancer Family Registry (CCFR). The primary Mendelian randomization analysis was performed using a genetic risk score, derived from 77 established “BMI-increasing” genetic variants, as an instrumental variable for higher BMI. We additionally derived a risk score for WHR using 47 variants for WHR. We compared the IV odds ratio (OR) with the OR obtained using a traditional multivariable logistic regression model adjusted for age, sex, smoking status, family history, aspirin use, and HRT use (women). Results: For men and women combined, in traditional multivariable analysis, each 5 kg/m2 increase in BMI was associated with a 24% (95% confidence interval [CI]: 1.17-1.32) increase in risk of CRC. In the IV analysis, the OR was higher (IV-OR per 5 kg/m2=1.54, 95% CI: 1.15-2.06); however, the two estimates were not statistically significantly different (Pdifference=0.07). For men, there was a statistically significant association between BMI and CRC in traditional multivariable analysis (OR per 5kg/m2=1.33, 95% CI: 1.21-1.46), but no association in the IV analysis (IV-OR per 5 kg/m2=1.13, 95% CI: 0.70-1.83). The estimates however were not significantly different (Pdifference=0.56). In contrast, for women, the IV estimate (IV-OR per 5 kg/m2=1.94, 95% CI: 1.33-2.82) was significantly higher than the estimate from the traditional multivariable analysis (OR per 5 kg/m2=1.18, 95% CI: 1.11-1.25; Pdifference=0.01). In secondary analysis, we found a positive association between WHR and CRC in men (IV-OR per 0.1 unit increase=1.84, 95% CI: 1.02-3.31) but no statistically significant association in women (IV-OR per 0.1 unit increase=1.22, 95% CI: 0.75-1.98). Conclusions: Obesity is independently associated with risk of CRC and may confer greater risk of CRC in women than previously reported. The mechanisms remain largely unknown; however these results suggest that visceral abdominal fat may be particularly important for promoting CRC in men. Citation Format: Aaron P. Thrift, Sonja I. Berndt, Andrew T. Chan, Jenny Chang-Claude, Martha L. Slattery, Michelle Cotterchio, Graham Casey, John D. Potter, Polly A. Newcomb, Emily White, Hermann Brenner, Ulrike Peters, Peter T. Campbell. Obesity and risk of colorectal cancer: A Mendelian randomization study. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 2176. doi:10.1158/1538-7445.AM2014-2176

The Impact of SLHS Program on Perinatal Indicators

The federal Healthy Start program began 20 years ago, yet outcome evaluations lack sufficient rigor to draw conclusions on program impact. We evaluated the impact of the Healthy Start program on birth outcomes, prenatal care, and public services utilization. Birth record data for the St. Louis Healthy Start Program (SLHS) and non-SLHS controls (matched using a propensity score technique) were assessed for differences. Propensity score matching techniques matched SLHS to non-SLHS clients on potentially confounding variables for births from years 2006 to 2008. Traditional multivariable logistic regression on the full, unmatched sample was also conducted for comparison. Matching eliminated any prior statistical differences between groups on covariates. 168 controls and 84 SLHS participants remained in the final matched analysis group. Both analysis techniques were similar on all outcomes, revealing significant group differences for low birth weight (matched OR = 0.28, p = 0.023) and prematurity (matched OR = 0.25, p = 0.012) but not for prenatal care (matched OR = 0.76, p = 0.414), or public services utilization (matched OR = 3.31, p = 0.121). Early results for this Healthy Start project are positive in key areas directly impacting infant mortality. However, continued analysis of this program for sustained impact in these areas and ultimately, a reduction in infant mortality is needed. Additionally, more rigorous experimental and quasi-experimental evaluation designs are needed to assess the impact of other Healthy Start programs around the country.

Identification by random forest method of HLA class I amino acid substitutions associated with lower survival at day 100 in unrelated donor hematopoietic cell transplantation

The identification of important amino acid substitutions associated with low survival in hematopoietic cell transplantation (HCT) is hampered by the large number of observed substitutions compared to the small number of patients available for analysis. Random forest analysis is designed to address these limitations. We studied 2,107 HCT recipients with good or intermediate risk hematologic malignancies to identify HLA class I amino acid substitutions associated with reduced survival at day 100 post-transplant. Random forest analysis and traditional univariate and multivariate analyses were used. Random forest analysis identified amino acid substitutions in 33 positions that were associated with reduced 100 day survival, including HLA-A 9, 43, 62, 63, 76, 77, 95, 97, 114, 116, 152, 156, 166, and 167; HLA-B 97, 109, 116, and 156; and HLA-C 6, 9, 11, 14, 21, 66, 77, 80, 95, 97, 99, 116, 156, 163, and 173. Thirteen had been previously reported by other investigators using classical biostatistical approaches. Using the same dataset, traditional multivariate logistic regression identified only 5 amino acid substitutions associated with lower day 100 survival. Random forest analysis is a novel statistical methodology for analysis of HLA-mismatching and outcome studies, capable of identifying important amino acid substitutions missed by other methods.

Analysis of genome-wide association data by large-scale Bayesian logistic regression

Single-locus analysis is often used to analyze genome-wide association (GWA) data, but such analysis is subject to severe multiple comparisons adjustment. Multivariate logistic regression is proposed to fit a multi-locus model for case-control data. However, when the sample size is much smaller than the number of single-nucleotide polymorphisms (SNPs) or when correlation among SNPs is high, traditional multivariate logistic regression breaks down. To accommodate the scale of data from a GWA while controlling for collinearity and overfitting in a high dimensional predictor space, we propose a variable selection procedure using Bayesian logistic regression. We explored a connection between Bayesian regression with certain priors and L1 and L2 penalized logistic regression. After analyzing large number of SNPs simultaneously in a Bayesian regression, we selected important SNPs for further consideration. With much fewer SNPs of interest, problems of multiple comparisons and collinearity are less severe. We conducted simulation studies to examine probability of correctly selecting disease contributing SNPs and applied developed methods to analyze Genetic Analysis Workshop 16 North American Rheumatoid Arthritis Consortium data.