Importance: The 8th edition TNM includes extranodal extension (ENE) as a new parameter for both clinical and pathological N-classification for non-viral related head and neck cancer, including squamous cell carcinoma of oral cavity (OSCC). Due to lack of robust data on reliability and prognostication of clinical/radiological ENE, it only refers to those with unambiguous clinical/radiological evidence of gross ENE, such as dermal involvement or soft tissue invasion with deep fixation/tethering to underlying muscle or adjacent structures. Assessing the accuracy of identifying radiological ENE (rENE) against pathological ENE (pENE) and its prognostication would aid treatment planning and potentially refine N-classification in the future.
Objective: To evaluate the accuracy of identifying rENE on pre-operative computed tomography (CT) or magnetic resonance imaging (MRI) against pENE and to determine the prognostic value of rENE on overall survival (OS).
Design, Setting, Participants: This was a retrospective study involving a cohort of patients with available preoperative imaging who had undergone cervical neck dissection (ND) for OSCC in our institution from 2010 through 2015. Two neuroradiologists blinded to the pathologic report reviewed the imaging independently. rENE was defined as ill-defined borders.
Main Outcomes and Measures: Impact of imaging-surgery interval, imaging modalities (CT vs MR), and intra-rater (>3 months apart) and inter-rater concordance of rENE were assessed. The diagnostic accuracy of rENE against pENE (identified from the ND specimen) were calculated. The frequencies of additional radiological features (invasion of adjacent structures, matted lymph nodes) on those with rENE was also calculated. OS were compared between those with and without rENE.
Results: Among the 508 consecutive OSCC patients, rENE and pENE were identified in 57 and 121 cases, respectively (52 cases with pENE did not have rENE). The diagnostic accuracy of rENE against pENE was identical (73%) for cases with the imaging-surgery interval ≤4 weeks (n=276) and 4.1-8 weeks (n=207) but lower for those >8 weeks (48%). Intra-rater concordance was excellent (0.94) while inter-rater was good (r=0.79). Excluding 25 cases with >8 weeks imaging-surgery interval, the remaining 483 cases were eligible for further analysis. The sensitivity, specificity, PPV, NPV, and accuracy of rENE against pENE was 52%, 96%, 93%, 66% and 73%, respectively. Within the rENE+ subset (52/57 had pENE), 9 (15%) showed invasion of adjacent structures (all had pENE) and 11 (20%) were identified as being matted nodes (10 had pENE). rENE assessed on CT had improved accuracy when compared to MR (80% vs 68%, p=0.0111). Median follow-up was 3.0 years. Three-year OS was significantly inferior for those cN+ with rENE (n=57) vs without rENE (n=211) and both were lower than cN0 (n=240) patients (3-year OS: 33% vs 67% vs 80%, p<0.001), which was similar to those pN+ with pENE vs without pENE vs pN0 (3-year OS: 37% vs 68% and 87%, p<0.001).
Conclusions and Relevance: This large cohort study shows a very high specificity (>90%) but low sensitivity (52%) of rENE for pENE, with good inter-rater and intra-rater reliability. Similar to pENE, the presence of rENE is associated with reduced survival in OSCC.