Background The prognostic stratification of laryngeal squamous cell carcinoma (LSCC) patients is inadequate as patients affected by tumors with similar clinicopathological parameters, and undergoing similar treatment, may differ in prognosis. The current study was undertaken to assess the immunohistochemical expression of epidermal growth factor receptor (EGFR), E-cadherin, and phosphatase and tensin homolog deleted in chromosome ten (PTEN) in 52 retrospective LSCC cases. We also investigated the prognostic implication of deregulated marker expression in relation to the clinicopathological features of LSCC, especially neck nodal metastasis.Results Decreased histologic differentiation, supraglottic tumor location, low or negative EGFR, and low or negative E-cadherin expression were significant predictors of lymph node metastasis (LNM) (P < 0.001, P = 0.011, P < 0.001, and P = 0.033, respectively), whereas patients' age, increased T-stage, cartilage invasion, muscle invasion, and differences in PTEN expression were not predictors of LNM. A receiver operator characteristic curve analysis showed that the predictability of EGFR and E-cadherin expression in discriminating LNM-positive cases was good and poor, respectively. Three models were developed using only the variables that revealed significant association with LNM in bivariate analysis and we tested their predictability to discriminate LNM-positive from LNM-negative cases. All models were significantly better than the basic model containing the constant only (P < 0.001 for the three models). A receiver operator characteristic analysis was performed, and it revealed that the predictive power of the first (tumor grade and site) and second (tumor grade, site, and EGFR) models ranged from good to excellent, whereas that of the third model (tumor grade, site, EGFR, and E-cadherin) was excellent. The number of correctly diagnosed cases (overall accuracy) increased only in the second compared with the first model, but the third model had an overall accuracy equal to that of the second model.Conclusion Nodal metastasis can be predicted by combining tumor grade and site in an equation, and if an immunohistochemical marker is to be added to increase the ability to identify patients with clinically negative nodes, our results suggest it to be EGFR. Although E-cadherin did not increase LNM predictability when added, this may be due to the limitation of the sample size. Thus, according to our results, we recommend an evaluation of these promising molecular markers on a larger sample size in future studies.