Smoking is a well-known risk factor for esophageal cancer. However, there are few reports that directly evaluate smoking as a prognostic factor for esophageal cancer. Moreover, scarce evidence is available on whether smoking interacts with major treatment modalities of esophageal cancer. In this study we retrospectively analyzed 364 patients with esophageal squamous cell cancer who were treated between 2001 and 2005 at our institution. Background characteristics, including smoking history, were analyzed as potential prognostic factors. Of the 363 patients, 76 patients (20.9%) were non-smokers or light smokers (non-heavy), whereas 287 patients (79.1%) were heavy smokers. The 5-year survival rate for non-heavy smokers and heavy smokers was 61.8% (95% confidence interval [CI]: 49.1-72.2) vs 44.6% (95% CI: 38.2-50.9), respectively. In a multivariate Cox model (adjusted for age, gender, performance status, alcohol consumption, histology, tumor length, International Union Against Cancer [UICC] stage, and treatment), the hazard ratio for heavy smokers in comparison with non-heavy smokers was 1.73 (95% CI: 1.12-2.68; P = 0.013). When we stratified by treatment method, heavy smoking was significantly associated with poor survival only in patients treated by chemoradiotherapy (hazard ratio, 2.43; 95% CI: 1.38-4.27; P = 0.002). More importantly, a statistically significant interaction between heavy smoking history and treatment modality was observed (P = 0.041). Our results indicated that smoking history is strongly associated with poor prognosis in patients with esophageal cancer, especially those treated by chemoradiotherapy. Further investigation is warranted to explain this different prognosis.