Abstract
Background: Cervical lymph node metastasis is one of the most important prognosticators in oral squamous cell carcinoma (OSCC). The presence of extranodal extension (ENE) in metastatic nodes further worsens the prognosis. The primary objective of this study was to develop a simple model to predict ENE in OSCC using demographic and clinicopathological parameters that are easily available in a preoperative clinical setting.
Methods: All patients who underwent surgery for the treatment of primary OSCC of the tongue or buccal mucosa with neck dissection (ND) between 1998 and 2020 from two oral pathology diagnostic units were included in the study. Chi-square tests were utilised to analyse the presence/absence of associations between each input parameter and ENE. The strengths of these associations were assessed using univariable and multivariable logistic regressions and indicated as odds ratios.
Results: Among the 1098 OSCC patients included in the study, 185 (16.85%) patients exhibited ENE. In the multivariable logistic regression model the odds for ENE were greater in females and tongue OSCC. Similarly, increasing T category and DOI as well as a dyscohesive WPOI were associated with a greater odds ratio for ENE. The AUCs from the 10-fold cross-validations to assess the predictive ability of the newly developed model ranged from 0.655 to 0.842 with a 95% CI of 0.663-0.744, indicating that the diagnostic accuracy of our model was acceptable.
Conclusion: Our model confirms that a higher risk of ENE is associated with female gender, tumours of the tongue, increasing tumour size as well as DOI and a dyscohesive WPOI.