The influence of measurement errors in tumor markers

Abstract

Measurement error is inherent in the collection of tumor markers. In general, when measurement error is present we know that the regression parameters are bias, but for a logistic model there are other concerns. This research sought to answer what happens to Specificity, area under the curve (AUC), Sensitivity, and the classification accuracy when measurement error was present. We found that there was better discrimination for tumor markers highly correlated with the dichotomous outcome variable; and Specificity, or true negatives, decreased as measurement error increased indicating an increase in the number of false negatives in the presence of measurement error.