[Influence of sociodemographic variables in the enrollment of subjects in a population case-control study]

Selection in representation of the study base was quantified in a case-control study of laryngeal cancer among residents of the City of Torino, Italy. Two-hundred-ninety-nine out of 408 eligible cases (73.3%) and 608 out of 1104 (55.1%) eligible controls were interviewed. All interviewed and non interviewed subjects were compared in a logistic regression model with respect to age, sex, educational level, civil status and area of birth. No large differential selection with respect to these variables was found between interviewed and non interviewed cases. Lower proportions of interviewed subjects were found among older controls and controls with lower education. These results suggest the possibility of selection bias in this case-control study. In particular, biased measures of effects might be estimated when analyzing variables related to educational level.