ADJUSTMENT FOR BIAS DUE TO ERRORS IN EXPOSURE ASSESSMENTS IN MULTICENTER COHORT STUDIES ON DIET AND CANCER - A CALIBRATION APPROACH

An advantage of multicenter cohort studies on diet and cancer is that these may include populations over a wide range of dietary exposure. W ith some simplifying assumptions, the information from such multicente r studies may be divided into 1) estimated relationships within each o f the separate cohorts, between individual-level measurements of dieta ry exposure and disease outcome, and 2) an estimated between-cohort re lationship, between the mean intake measurements and mean incidence ra tes. Errors in the dietary exposure measurements may lead to different amounts of bias in each of these estimated relationships, in particul ar when dietary questionnaire methods cannot easily be standardized. A calibration approach can be used to adjust for such differences in bi as. This will improve the relative weighting of within- and between-co hort components of evidence for a diet association.