Verification bias occurs when the test used as a gold standard (or to verify the diagnosis) is not the same for all groups of patients in a trial. Different patients get different gold standards.
For example, in a trial looking at high sensitivity troponins, patients with a positive troponin might go on to have an angiogram, whereas patients with negative troponins could be watched clinically. The angiogram will find more pathology, and because the test is done only when the troponin is positive, the results are biased towards a higher specificity. On the other hand, you are less likely the find pathology in the group of patients with no testing performed. Performing less testing when there is a negative troponin will find less pathology, biasing the results towards an improved sensitivity.
This post is part of a series of posts on bias in medical research. You can find the whole bias catalogue here.