I wish to comment on the recent paper by Fakhry et al., “Fine needle aspiration cytology and frozen section in the diagnosis of malignant parotid tumours”. This study illustrates a common problem in diagnostic accuracy studies for fine needle aspiration (FNA). In this study, cases were only included if they received histological verification. This is a common type of study design but, when applied to FNA, results in verification bias. Verification bias occurs when positive and negative findings are verified at different rates. This generally occurs in FNA because positive findings are referred to surgery and receive histological verification, but negative findings are managed clinically. Thus, when studies only include cases that receive histological verification, positive cases are verified at a higher rate than negative cases. This type of bias generally results in overestimates of sensitivity and underestimates of specificity. Statistical techniques are available to estimate the extent of bias; however, to make these estimates, authors must report the total number of FNA cases and the FNA findings. Because these were not reported, it is not possible to estimate the bias in sensitivity and specificity in the study of Fakhry et al. Predictive values are not subject to verification bias. Also, the estimates of sensitivity and specificity for frozen section obtained in this study are not subject to verification bias because all cases that received the index test were verified by histopathology. This case is unusual because it provides an interesting example where two tests (FNA and FS) were performed on the same set of cases that were all verified by the same method, but only one of the tests (FNA) was subject to verification bias.
Recent studies have shown that verification bias is common in diagnostic accuracy studies for FNA and particularly in ENT journals. Authors and reviewers need to be aware of this source of bias.