Response

We would like to thank Dr Balderramo for his interest in our study, and we thank the editors for an opportunity to preview and respond to his letter. Dr Balderramo brings up several important points and questions, which we can respond to. The first question raised relates to the isolated accuracy of probe-based confocal laser endomicroscopy (pCLE) compared with the accuracy of tissue sampling. Previous studies have examined the “isolated” performance of pCLE to diagnose cholangiocarcinoma by consensus and individual review of edited images in a blinded fashion.1 and 2 In this study, we recognized that the treating physician is never unbiased when performing pCLE, having met and examined the patient and reviewed all previous studies. We therefore chose to study the clinical impression of the physician at the time of pCLE and after tissue sampling returned because these are the metrics that drive decision making. Thus, the performance of pCLE alone cannot be determined in this study design because it cannot be isolated from the clinical impression of the treating physician at the time of pCLE, as we have done in prior studies. We apologize for any confusion generated by Table 3 in our article, where the last column shows the performance of ERCP plus tissue sampling of the blinded reviewer (second investigator), who had access only to single images of ERCP, CT, and a brief clinical vignette along with the tissue sampling results. Dr Balderramo asks for the clinical ERCP impression of the second investigator, which we had chosen not to present. This blinded investigator’s impression of ERCP underperformed compared with the primary investigator, with accuracy, sensitivity, and specificity of 69%, 78%, and 53%, respectively. This is not surprising, given the limited data that were transmitted by electronic packages. We did this analysis to estimate routine clinical care without pCLE. Dr Balderramo also asked whether prior stenting of the common bile duct has an impact on the performance of pCLE. This was previously addressed in a separate study that showed slightly better performance in unstented patients.3 For the purposes of this response, we calculated the performance of ERCP plus pCLE in patients with prior stents (accuracy 80%, sensitivity 89%, specificity 71%) and in patients without prior stents (accuracy 85%, sensitivity 88%, specificity 79%) and found a nonsignificant trend toward better accuracy of pCLE in patients who had not been previously stented. Finally, we agree with Dr Balderramo that the accuracy of ERCP with pCLE should be compared with the accuracy of ERCP and tissue sampling. In our study, this could be done only by comparing data obtained from the treating physician who performed pCLE with data from the blinded second investigator, as shown in Table 3 of our article. This difference in accuracy between these 2 groups was not significant (82% for ERCP plus pCLE vs 79% for ERCP plus tissue sampling). We stand by our conclusion that the addition of pCLE to clinical impression and tissue sampling may allow for more accurate assessment of patients with indeterminate biliary strictures.