Session Item

Response assessment after definitive radio-chemotherapy (RCT) for HNSCC is based on a 12-week FDG PET-CT. The 5-point Hopkins score (HS) is the only qualitative scale available for standardized reporting, albeit limited by suboptimal positive predictive value (PPV). The aim of our study was to explore the feasibility and assess the diagnostic accuracy of an experimental 6-point scale (“Xxxxx score” from the name of the city where the scale was designed)

Patients with locally advanced HNSCC undergoing RCT between 1/10 and 12/15 could be enrolled. Baseline and post-treatment PET-CT scans had to be available for each subject. The post-treatment scan had to be executed not before 10 weeks and no later than 6 months after RT completion.An indipendent qualitative evaluation of post-treatment scans was undertaken by 3 experienced nuclear medicine physicians who were blinded to patients’ information, clinical data and other imaging examinations. Response to treatment was evaluated according to HS and Xxxxx score (table 1).The primary endpoint of the study was to evaluate the PPV of Xxxxx score in assessing locoregional control (LRC). We also correlated semi-quantitative metabolic factors as included in PERCIST and EORTC criteria with disease outcome. Based on a PPV of HS of 71.1% (Marcus et al, J NucI Med 2014), applying the experimental 6-point score was hypothesized to yield a 20% increase. With a significance level α=0.05 and a power of 0.90 when the PPV is 85.3%, the required sample size with this design was 81 patients

350 patients were included from 11 centers. In accordance with EANM 1.0 guidelines,the requested acquisition time of PET-CT scans should be 60 +10 minutes after injection, therefore our final cohort consisted of 119 subjects (oropharynx: 57.1%; HPV negative: 73.1%). At a median follow-up of 42 months (range 5 – 98), the median LRC was 35 months (95% CI: 32 – 43), with a 74.5% 3-year rate. The experimental Xxxxx score had the highest diagnostic accuracy (76.5%) with a PPV for primary tumor (Tref), nodal disease (Nref) and composite TNref of 42.9%, 100% and 50%, respectively (table 2). Good inter-observer agreement was measured with the Krippendorff’s alpha coefficientat 3 time points (data not shown). A Xxxxx score of 5-6 was associated with poor OS at multivariate analysis (HR 6.0; 95% CI: 1.88-19.18; p=0.002). In addition, nodal PD according to PERCIST criteria was associated with worse LRC (OR for LR failure, 5.65; 95% CI: 1.26-25.46, p=0.024) and OS (OR for death, 4.81; 1.07-21.53, p=0.04)

In the frame of a standardized methodology, the feasibility of Xxxxx score was preliminarly validated, yielding a high nodal PPV. Prospective investigations are warranted to further evaluate its reproducibility and diagnostic accuracy. The strictly blinded approach we followed may justify why our qualitative interpretations underperformed mainly in terms of PPV for Tref, where integrating more imaging modalities and clinical feedback is useful