Abstract
Response and remission are determined by the proportion of people who fall below a threshold score on depression rating scales. This calculation implies a possibility of false positives (FP) and false negatives (FN) depending on sensitivity and specificity of the threshold used, but also on response and remission rates. A simulation illustrates the methodological consequences of this phenomenon in a comparative trial where response and remission rates differ between groups: the probability of being misclassified differs between groups, and measures of association (relative risk and odds ratio) are biased. Alternatives are proposed to cope with this misclassification bias.