Reply to the Letter to the Editor: “Multiparametric MRI of the bladder: inter-observer agreement and accuracy with the Vesical Imaging-Reporting and Data System (VI-RADS) at a single reference center”
by G. Barchetti, F. Del Giudice, G. Anello, V. Panebianco (email@example.com)Multiparametric MRI of the bladder: inter-observer agreement and accuracy with the Vesical Imaging-Reporting and Data System (VI-RADS) at a single reference center
We thank the authors Sabour and Shokri for their interest in our recently published paper . We acknowledge that the issues they raised are of potential concern, therefore we hope we can clarify some of the points they addressed.
Firstly, we are well aware of the limitations of the kappa, as it depends on the number of categories. Even if Sabour and Shokri state that the prevalence in our study is not known, we must emphasize that we reported the prevalence of muscle-invasive bladder cancer (stage T2 or higher), which is the focus of our research, in the overall population (22/103 lesions) and for each VI-RADS category for each reader. We proposed to evaluate the inter-reader agreement by means of kappa statistics, and this was considered appropriate after a thorough statistical revision. Besides, it is common and widely accepted to use k value to assess inter-reader agreement when a 5-point imaging scale is introduced into clinical practice. One of such well-known scales is the Prostate Imaging-Reporting and Data System (PI-RADS), and its reliability has been thoroughly assessed by means of k value [2, 3].
Secondly, it has been pointed out that we neglected patients’ clinical status and physicians’ opinion. In this regard, we must highlight that we gave a complete description of the clinical characteristics of our population and, most importantly, that the physician’s opinion is of utmost importance in such a delicate context. The diagnosis of muscle-invasive bladder cancer, compared to a non-muscle invasive bladder cancer, entails a completely different surgical approach, with potential consequences for the patients that have to be conscientiously taken into account.
Finally, as regards bootstrapping, we recall that a considerable number of patients in our cohort had multiple lesions and that scores of multiple lesions from the same patients are likely to be correlated. The conventional standard error is therefore not reliable, as it requires independent observations. For this reasons we used the bootstrap re-sampling procedure (number of bootstrap samples, 1000) to calculate the standard error of the k estimates, as previously described .