Hypothesis testing in the polychotomous logistic model with an application to detecting gastrointestinal cancer

Abstract

We discuss the use of the trichotomous logistic model to discriminate between patients with gastrointestinal (GI) cancer, patients with benign GI disease and 'normal' subjects, using symptoms and the concentrations of some serum proteins that are potentially indicative of malignancy as covariates. A parsimonious model can be obtained by invoking an indistinguishability hypothesis which is appropriate when a covariate is considered to have no predictive value between categories. It is shown that the polychotomous model can be re-parameterised under the null hypothesis to give a 'reduced form', which can be fitted by maximum likelihood. The validity of the use of the same methods for retrospective sampling is discussed. The approach is illustrated by the development of a logistic model to identify symptomatic and asymptomatic subjects with a high risk of GI cancer.

Genetic variation in XRCC1, sun exposure, and risk of skin cancer

“…The results did not change substantially in the matched analysis for each of the three types of skin cancer with their own matched controls. We used a likelihood ratio test (LRT) to evaluate heterogeneity in the effects of the XRCC1 genotypes on different types of skin cancer in polytomous logistic regression models (Marshall and Chisholm, 1985). To summarise multiple variables, we constructed a multivariate confounder score to create a constitutional susceptibility index for skin cancer (Miettinen, 1976).…”

A prospective study of the relationship between body mass index and cataract extraction among US women and men

“…We computed the Lagrange multiplier test statistic as a measure of the differences between type-specific coefficients. 39,40 We conducted separate analyses for each cohort and pooled the two study groups to achieve maximum statistical power. Tests for heterogeneity between the two study groups were conducted, and meta-analytic methods using a random effects model were employed to pool the relative risks from the two cohorts.…”

Sex Hormone Levels, Breast Cancer Risk, and Cancer Receptor Status in Postmenopausal Women: the ORDET Cohort

“…To test for differences in hormone/SHBG levels by receptor status, we used polychotomous logistic regression (19) Results Table 1 shows the characteristics of the 165 cases and 672 controls. Compared with controls, cases had a higher frequency of family history of breast cancer, lower frequency of breast-feeding, and lower past use of oral contraceptives.…”