Can elective neck dissection for oral cancers be planned based on depth of invasion?

Abstract

e18042 Background: Elective neck dissection (END) is the standard of care for oral squamous cell carcinoma (OSCC). Literature suggests END may be avoided in primaries with depth of invasion (DOI) less than 4 mm. The present study evaluates the accuracy of DOI in predicting nodal metastases in OSCC. Methods: This retrospective study, approved by the Institutional Ethics Committee, includes all the patients with OSCC who underwent surgical excision. All tumors were classified and DOI was measured microscopically on resected specimens according to the AJCC TNM 8th ed. Statistical analysis was performed using IBM SPSS for Windows version 26. Pathological N-classification (pN) was correlated with DOI (in mm) using Spearman’s rank-order (pN1-3) and point-biserial (pN0 vs pN+) correlation statistics. Receiver operating characteristic (ROC) curve with area under the curve (AUC) were analyzed to determine the accuracy of DOI in predicting nodal metastases (pN+). Primaries were classified into three groups based on the range of DOI (≤5 mm; >5 mm and ≤10 mm; >10 mm) according to the AJCC TNM 8th ed. Chi-square statistic was used to measure the association between DOI groups and pN status (pN0; pN+). ANOVA with post hoc Bonferroni statistics were used to compare between three DOI groups. Results: Three hundred and thirty-seven (n=337) patients were screened for inclusion. After adjusting for missing data, 254 patients were included in the final analysis. Median age at presentation, DOI and nodal yield were 46 (24-82) years, 11 (1-40) mm, and 35 (1-121), respectively. Depth of invasion failed to demonstrate a strong positive correlation with pN classification (ρ=0.31; p<0.001). DOI did not show a strong correlation with pN status (pN0 vs pN+) either (rpb=0.27; p<0.001). ROC curve analysis suggested a poor accuracy of DOI in predicting nodal metastases (AUC=0.67; 95% CI 0.6-0.73). There was a statistically significant association between DOI groups and pN status (Chi-square=26.67; p<0.001). ANOVA with post hoc Bonferroni showed statistical significance only for DOI >10 mm group (p<0.001 for DOI ≤5 mm; p=0.02 for DOI >5 mm and ≤10 mm). However, the sensitivity and specificity achieved by the ROC analysis for DOI cut-off value of 10mm in the present cohort were merely 69.3% and 57.9%, respectively. Conclusions: DOI alone is a poor predictor of nodal metastases in OSCC. Though all correlations in the present study achieved statistical significance, none demonstrated a strong association or an accuracy credible enough of clinical importance. END should remain the standard-of-care for all OSCC, including early primaries with DOI <4 mm.