In reference to accuracy and reliability of the flow coupler in postoperative monitoring of head and neck free flaps

Abstract

We were interested to read the paper by Fujiwara and colleagues published in the April 2018 issue of The Laryngoscope. The authors aimed to assess the accuracy and reliability of the flow coupler relative to the implantable arterial Doppler probe in postoperative monitoring of head and neck free flaps. Both the venous flow coupler and arterial Doppler were employed in 120 consecutive head and neck free flap cases. When Doppler signal loss occurred, flaps were evaluated by physical exam to determine whether signal loss was a true positive necessitating operating room takeback. Sensitivity, specificity, and false positive rate (FPR) were recorded for each device. Based on their results, permanent signal loss (PSL) occurred in the flow coupler in all takebacks; PSL occurred in the arterial Doppler only in the case of arterial thrombosis. For the flow coupler, sensitivity was 100%, specificity 86.4%, and FPR 13.6%. For the arterial probe, sensitivity was 9.1%, specificity 97.1%, and FPR 2.9%. They concluded that the flow coupler can be a valuable tool in postoperative monitoring of head and neck free flaps. However, these results have nothing to do with reliability. Reliability (precision, repeatability) and validity (accuracy) are two completely different methodological issues. FPR and false negative rate as well as sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive and negative likelihood ratios are among the estimates to assess validity (accuracy) of a diagnostic test and have nothing to do with reliability. Moreover, for clinical purposes, PPV and NPV are more important than sensitivity and specificity. Finally, it is crucial to know that in clinical care, the diagnostic added value of a test should be considered by applying the area under the curve of receiver operating characteristic. The reason is that all the mentioned validity (accuracy) estimates can be excellent with clinically negligible diagnostic added value. Reliability (precision, reproducibility) as a different methodological issue should be assessed using appropriate tests. For qualitative variables, weighted j can be applied with caution. Regarding quantitative variables, intraclass correlation coefficient and Bland-Altman plot are among the well-known approaches. As a takeaway message, any conclusion regarding accuracy and reliability should be supported by the above-mentioned methodological and statistical issues. Otherwise, misinterpretation may occur.