Purpose: For patients with clinically positive cervical lymph nodes treated with primary radiotherapy, the use of planned neck dissection to improve regional control is controversial. This study analyzes patients with cancer of the oropharynx treated by radiotherapy alone or in combination with surgery, either prior to or immediately following irradiation. We examine whether combined treatment results in better regional control, factors that predict failure, and complication rates by treatment group.Methods and Materials: Between 1970 and 1994, 260 patients treated in our department received radiotherapy with curative intent for clinically node-positive base of tongue or tonsil squamous cell carcinomas. Eight patients are excluded for simultaneous cancers of the lung, breast, or thyroid that may obscure analysis of cervical lymph node status. Treatment groups are neck surgery, either neck dissection or node excision, followed by radiation (SR-RND or SR-Exc), radiation alone (R), or radiation followed by surgery for residual adenopathy (RS). By group, 17 are SR-RND, 24 SR-Exc, 158 R, and 53 RS. Nodal stage includes 82 (32%) N1, 147 (59%) N2, and 23 (9%) N3, and tumor stage consists of 13% T1, 39% T2, and 48% T3. Kaplan-Meier estimates are used to calculate regional control and overall survival, and Cox proportional hazard for the multivariate analysis. We score complications using the RTOG morbidity scoring scheme that ranges from 0 (none) to 4 (life-threatening).Results: With a median follow-up of 30 months, regional control (RC) correlates well with N-stage, as shown in Table 1. By univariate analysis treatment group correlates with RC and overall survival (OS) but not after controlling for differences among the treatment groups. Multivariate analysis shows T-stage, clinical N-stage, overall radiation treatment time, photon energy, and continued tobacco use to be significantly associated with recurrence in the cervical lymph nodes. Treatment group, age, sex, grade, fractionation, dose, nodal regression after treatment, electron boost, decade of treatment, primary site, and chemotherapy are not significant in predicting recurrence. The absence of a clinical complete response (CR) is a poor predictor of pathologic CR as demonstrated by residual disease in only 57% of RS pathologic specimens. The subgroup with a pathologic CR has a significantly better regional control (85%) than those without (50%). Patients from other treatment groups having second neck surgeries show a pathologic CR rate ranging from 0 to 60%. The incidence of complications, reported in Table 2, varies by treatment with SR-RND having the highest rate of Grade 2-4 complications.Conclusions: In a large retrospective analysis of node-positive cancer of the oropharynx, nodal treatment seems not to be a good predictor of ultimate regional control. Modifiable factors that can improve regional control include shorter treatment time and cessation of smoking. Original TNM staging is more important in predicting recurrence than complete clinical regression following treatment. In our experience patients treated with radical radiation therapy resulted in a 25% immediate post-radiotherapy neck dissection in which 57% harbored residual disease. Pathologic CR is prognostic for RC. In a setting where no outcome differences are observed by nodal treatment, differences in complications may influence choice.