The current study was devised to evaluate the therapeutic potential of extended surgery for improving survival in undifferentiated thyroid carcinoma (UTC). An institutional retrospective survival analysis (July 1994 to December 1998) of 30 patients who underwent surgery for UTC with locally curative intent was done. Median and 1-year survival was 4 months and 37%, respectively. Primary patients were older (70 vs. 59 years; p = 0.026) and deceased earlier (median survival 4 vs. 20 months; p = 0.027, log-rank test) than their reoperative counterparts, suggesting a referral bias toward younger patients. Survival analysis was restricted to primary pT4 UTC, leaving 18 patients. On univariate analysis, pN and M category, degree of resection (R2 versus R0/1 and radiotherapy (0-30 Gy versus >30 Gy) were identified as parameters suitable for further testing. On multivariate analysis, pN1 was a significant prognosticator of decreased survival (RR = 5.9; p = 0.043), followed by R2 (RR = 4.1, p = 0.088) and M1 (RR = 3.6; p = 0.089). Because of low patient numbers after stratification for radiotherapy, only pN and degree of resection were analyzed on subsequent multivariate analysis. In the incomplete radiotherapy stratum, neither of the two parameters affected survival, whereas R2 and pN1 limited survival in the complete radiotherapy stratum. In primary pT4 UTC, a subset of pN0 patients with R0/1 resections and radiotherapy greater than 30 Gy seemed to benefit from extended surgery. Because pN1 and R2 patients with radiotherapy of 30 Gy or less comprised most UTC patients, only 1-year, but not median, survival improved compared to literature controls.