BACKGROUND: Two randomised trials concerning thoracic oesophageal cancer concluded that for squamous cell carcinoma, chemoradiation alone leads to the same overall survival (OS) as chemoradiation followed by surgery. One of these trials, FFCD 9102, randomised only fit, compliant and operable responders to induction chemoradiation between continuation of chemoradiation and surgery. In the present analysis, the outcome in the patients not eligible for randomisation was calculated to determine if attempt of surgery should be recommended. METHODS: Eligible patients had operable T3-N0/N1-M0 thoracic oesophageal cancer. After initial chemoradiation, patients with no clinical response, or with contraindication to follow any attributed treatment, were not randomised. OS was studied first in the whole population of not randomised patients, and then specifically in clinical non-responders. The impact of surgery on OS was studied in these two populations. FINDINGS: Of the 451 registered patients in the trial, 192 were not randomised. Among them, 111 were clinical non-responders. Median OS was significantly shorter for non-randomised patients (11.5 months) than for randomised patients (18.9 months; p=0.0024). However, for the 112 non-randomised patients who underwent surgery, median OS was not different from that in randomised patients: 17.3 versus 18.9 months (p=0.58). Concerning clinical non-responders, median OS was longer for those who underwent surgery compared to non-operated patients: 17.0 versus 5.5 months (hazard ratio (HR)=0.39 [0.25-0.61]; p<0.0001), and again was not different from that in responding, randomised patients (p=0.40). INTERPRETATION: In patients with locally advanced thoracic oesophageal cancer, overall survival did not differ between responders to induction chemoradiation and patients having surgery after clinical failure of chemoradiation. Surgery should therefore be considered in those patients who are still operable.