Pretherapeutic identification of patients likely to benefit from neoadjuvant chemotherapy for head and neck epidermoid cancer is of interest. We retrospectively analyzed the pretherapeutic computed tomographic (CT) scans of lymph nodes of 70 patients with head and neck cancer. All 70 patients were clinically classified as having stage IV disease. The purpose of our analysis was to compare the prognostic value of CT node density with that of the following factors: age, T and N categories, Eastern Cooperative Oncology Group performance status, tumor site, histopathologic type of disease [squamous cell carcinoma (SCC) or undifferentiated carcinoma of nasopharyngeal type (UNCT)], and type of local-regional treatment. A simple two-grade nodal density grading system was devised. The density of normal adjacent muscle was chosen as the density standard. A node was classified grade 1 if less than 33% of the node consisted of hypodense zones. A node was classified grade 2 if more than 33% of the node consisted of hypodense zones. Patients with grade 1 nodes had a complete response rate of 68% (21/31) compared with 8% (3/39) for those with grade 2 nodes (P less than .0001). The only other factor associated with complete node response was UCNT (P less than .03). However, node density remained the significant prognostic factor after adjustment for histopathologic type. Follow-up ranged from 16 to 44 months, with a median of 29 months. Patients with grade 1 nodes had a median survival time of 32 months versus 13 months for those with grade 2 nodes (P less than .01). A prospective study should validate the prognostic value of CT node density and its possible use in determining optimal multimodal therapy for advanced head and neck cancers.