Premium
Relative risk models for assessing the joint effects of multiple factors
Author(s) -
Lubin Jay H.,
Gaffey William
Publication year - 1988
Publication title -
american journal of industrial medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.7
H-Index - 104
eISSN - 1097-0274
pISSN - 0271-3586
DOI - 10.1002/ajim.4700130110
Subject(s) - covariate , medicine , relative risk , poisson regression , statistics , proportional hazards model , regression analysis , cohort , econometrics , regression , demography , epidemiology , confidence interval , environmental health , surgery , mathematics , population , sociology
A goal of analyses of occupational cohort data is the specification of how covariate information relates to age‐specific disease risks. In describing this relationship, certain assumptions or models must be defined. For example, the usual standardized mortality ratio assumes a constant multiplicative increase in the age and calendar period disease rates of an exposed cohort over rates in a unexposed referent group. For analyzing several exposures, some of which may be continuous, such as cumulative dose, dose rate, duration of employment, and smoking patterns, or for analyzing complex associations between disease rate and covariates, flexible regression procedures are required. Using a crossclassification of the data and a Poisson probability model, relative risk regression methods are outlined. Breslow and Storer [1985], Guerrero and Johnson [1982], and Thomas [1981] propose general models for the relative risk as alternatives to, but which include, the usual exponential form. We review these models, discuss some limitations (in particular when there is more than one covariate) and present alternatives. Methods and models are illustrated by examining the joint effects of radon exposure and tobacco use on lung cancer mortality among a group of uranium miners.