Risk thresholds for selective neck management in cN0 lower lip SCC: a decision-analytic prediction model

Management of the clinically node-negative (cN0) neck in lower lip squamous cell carcinoma (LLSCC) remains controversial because the risk of occult metastasis must be balanced against the morbidity of elective neck dissection (END). We aimed to identify predictors of early cervical nodal failure under an observation-first strategy and to translate them into a threshold-based decision-support framework.

In this multicenter retrospective cohort, 208 consecutive LLSCC patients were screened, and 74 cN0 patients treated with primary tumor resection without upfront END were included. The primary endpoint was early cervical nodal failure within 12 months, defined as failure of surveillance. A full postoperative multivariable model including DOI, tumor size, and closest histopathologic margin was fitted for individualized postoperative risk estimation. A simplified DOI-size risk stratification scheme was constructed for clinically intuitive group-level interpretation and threshold-based decision support. Discrimination, calibration with bootstrap validation and shrinkage, ROC-derived thresholds, and decision curve analysis were performed. The simplified DOI-size strata were interpreted against a prespecified 20% elective-treatment threshold, with sensitivity analyses at 15% and 25%.

Early nodal failure occurred in 15 of 74 patients (20.3%). DOI (OR 1.95 per mm, 95% CI 1.19 to 3.21, P = .008) and tumor size (OR 3.06 per cm, 95% CI 1.30 to 7.18, P = .010) independently predicted failure, whereas closest histopathologic margin showed an inverse but non-significant association (OR 0.74 per mm, 95% CI 0.52 to 1.06, P = .102). The full DOI-size-margin model achieved an apparent AUC of 0.85 with acceptable calibration after optimism correction, with a shrinkage slope of 0.86 and calibration-in-the-large of 0.03. ROC-derived thresholds were DOI ≥ 7 mm and tumor size ≥ 3 cm, while DOI ≥ 6 mm was retained as a conservative clinical boundary for simplified bedside stratification. The simplified DOI-size stratification separated 12-month nodal failure risk into 0.0% in the low-risk group (0/12), 12.2% in the intermediate-risk group (5/41), and 47.6% in the high-risk group (10/21). Only the high-risk group exceeded the prespecified 20% decision boundary. Decision curve analysis demonstrated superior net benefit of the full postoperative model compared with the reduced DOI-size model and treat-all or treat-none strategies across clinically relevant thresholds.

In cN0 LLSCC managed without upfront END, a full postoperative prediction model incorporating DOI, tumor size, and closest histopathologic margin provides individualized estimation of early surveillance failure risk. DOI and tumor size remain the dominant clinically interpretable variables for simplified bedside risk stratification, while closest histopathologic margin contributes postoperative risk refinement in the full model. When embedded within a threshold-based framework, this combined approach supports selective rather than routine elective neck management. External validation is warranted.