BACKGROUNDThe objectives of this prospective trial were to evaluate the diagnostic accuracy and predictive value of magnetic resonance imaging (MRI) and to use MRI evaluation under general anesthesia (EGA) 6 to 8 weeks after chemoradiation to determine local control.METHODSEighty‐two consecutive patients with advanced‐stage squamous cell carcinoma of the oral cavity, oropharynx, hypopharynx, or supraglottic larynx were treated with selective targeted chemoradiation. All patients who completed treatment and survived had a minimum follow‐up of 3 years. MRI and EGA were performed from 6 to 8 weeks after treatment. Posttreatment MRI findings were compared with pretreatment MRI findings and were graded for risk of local recurrence/residual disease on a 4‐point scale. The diagnosis of treatment failure was based on tissue biopsies, which were obtained during EGA or later during follow‐up. The predictive value of MRI was analyzed by using a Cox proportional hazards model.RESULTSOnly 1 patient with MRI Grade 0 or 1 findings (discrete mass < 10 mm; n = 62 patients) had residual disease 6 to 8 weeks after treatment that was detected during EGA. In 5 patients with MRI findings of Grade 2a and 2b (mass > 10 mm; n = 20 patients), residual disease was detected. After 2 years, 23 patients had a local failure (28%). Twelve local failures were found among 62 patients who had MRI findings of Grade 0 and 1. Posttreatment MRI emerged as an independent predictive factor (hazard ratio, 3.0; P = .014) for local control.CONCLUSIONSPosttreatment MRI studies provided predictive information on local control in addition to pretreatment predictors. In patients with focal masses < 10 mm, the combination of response evaluation under general anesthesia and posttreatment MRI from 6 to 8 weeks after chemoradiation hardly provided more information on the local control than posttreatment MRI alone. Cancer 2006. © 2006 American Cancer Society.