Qualitative and quantitative methods in post-chemoradiation PET for head and neck cancer

Objective The aim of this study was to determine whether quantitative methods could aid in the evaluation of post-treatment head and neck scans, particularly taking human papillomavirus status into account. Methods Clinical readings of positron emission tomography/computed tomography scans as well as standardized uptake value (SUV)(max) (and other metrics) of nodes visible on PET conducted on a total of 172 patients with head and neck squamous cell cancer were examined. Locoregional recurrence at 2 years was assessed. In total 88 of these patients had close enough follow-up to determine whether individual nodes were positive or negative, and 233 nodes on these patients were compared to surgical pathology notes or follow-up (if no path was available). Results General negative predictive value (NPV) of complete response was 93% and an equivocal response was 89%; focusing on nodal recurrence, NPV was found to be 97% and positive predictive value (PPV) 46% if equivocal reads were treated as negative and NPV 98% and PPV 16% if equivocal reads were treated as positive. Using SUVmax of the hottest node with a cutoff of 3.4 gave NPV 97% and PPV 26%; a direct re-read (using 2 observers) gave NPV 98% and PPV 32% if equivocal reads were treated as negative, and NPV 99% and PPV 18% if equivocal reads were treated as positive. Using other first-order radiomics data such as SD and skewness did not improve this. Conclusions Quantitative data such as SUVmax does not show additional value over qualitative evaluation of response to chemoradiation in head and neck tumors.