Schematic illustration of two patients with oral squamous cell carcinoma who each have a single metastatic cervical lymph node. Both are therefore assigned the same pathological nodal category (pN1) and, at equal lymph node yield, the same metastatic node count, the same lymph node ratio and the same log odds of positive lymph nodes. In patient A the node lies in level II, in patient B in level IV. The quantitative descriptors are identical; the topographic patterns are not. Node sizes in the drawing are schematic and not to scale, and the illustration reproduces no observation from the present cohort. Created in BioRender. Schlegel, N. (2026) wyue6qy .

Metastatic node count, lymph node ratio (LNR) and log odds of positive lymph nodes (LODDS) are proposed as refinements of pN staging. Whether they also describe how widely disease is distributed across the neck is untested.

Level-resolved pathology from 289 patients operated for oral squamous cell carcinoma between 2017 and 2022 was analysed; 80 had at least one metastatic node documented by level. Involvement was reported conditional on documented dissection. An analysis restricted to patients with at least two metastatic nodes, below which multilevel disease is arithmetically impossible, was specified before any model was fitted. A permutation null model compared the observed distribution with random placement of the same metastatic nodes.

Twenty-eight of 80 patients (35.0%) had multilevel and 11 (13.8%) bilateral disease. Conditional on dissection, levels IV and V were involved in 23.4% and 23.6% of patients, respectively, compared with 13.8% and 16.2% of the entire node-positive cohort. As these levels were dissected in only approximately one-third and fewer than half of patients, respectively, the conditional estimates reflect clinically selected subgroups. Multilevel disease was associated with T3/T4 category (OR 5.43, 95% CI 1.19–24.85) and metastatic node count (OR 2.91, 95% CI 1.58–5.39). After restricting the analysis to patients with two or more metastatic nodes (n = 38; 10 with single-level disease), node count (OR 1.34, 95% CI 0.90–2.01), LNR, and LODDS were no longer associated with multilevel disease, whereas the association with T3/T4 category persisted (OR 6.06, 95% CI 1.15–31.84). The observed proportion of multilevel disease did not differ significantly from that expected under random placement (38.2% expected; p = 0.35).

Quantitative nodal metrics provide information distinct from topographic spread. The restricted comparison is small, so this is an absence of a detectable association rather than proof of none. Level-resolved reporting, per patient and conditioned on dissection, captures a dimension they do not.