Sample size calculations for evaluating a diagnostic test when the gold standard is missing at random.

摘要

We appreciate the letter from Dr Skaik and the opportunity to comment. A dichotomous diagnostic test can be used both for detection of disease or equivalently to indicate absence of disease. We do not argue otherwise in our paper. Hence, we agree with the comment that application of a test can have as its purpose selection of a non-diseased group. Here, a very high or perfect (100 per cent) specificity is more important than high sensitivity. Of course, even for a good test, misclassification is expected because rarely a diagnostic test has both sensitivity and specificity close to 100 per cent.