Premium
A comparison of three approaches to estimate exposure‐specific incidence rates from population‐based case‐control data
Author(s) -
Benichou Jacques,
Wacholder Sholom
Publication year - 1994
Publication title -
statistics in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.996
H-Index - 183
eISSN - 1097-0258
pISSN - 0277-6715
DOI - 10.1002/sim.4780130526
Subject(s) - statistics , incidence (geometry) , population , confidence interval , relative risk , interval (graph theory) , computer science , econometrics , mathematics , demography , geometry , combinatorics , sociology
In population‐based case‐control studies, an attempt is made to identify all incident cases diagnosed in a specified population during a fixed time interval. Assuming that this goal is met allows one to obtain measures of risk other than relative risks. In this paper, we describe three approaches to estimate exposure‐specific incidence rates. Approach 1 relies on estimating crude incidence rates of the disease in strata defined, for instance, by age and geographic area, and combining them with relative risk estimates from the case‐control data. In approaches 2 and 3, baseline incidence rates and relative risks are estimated jointly. Approach 2 is based on a pseudo‐likelihood, while, in approach 3, the problem is regarded as a missing data problem and a full likelihood is maximized. We applied these three approaches to a study of bladder cancer. Our three sets of estimates of exposure‐specific incidence rates were in close agreement, while there appeared to be greater precision with approaches 2 and 3.