From 1982 to 1991, 200 patients with squamous-cell esophageal cancers underwent operation. Sixteen prognostic parameters were prospectively collected and analyzed. The main aim of this study was to define the best preoperative criteria for selection for curative surgery and which patients would benefit most from radical resection. Squamous-cell head and neck carcinoma was associated in 43.5% of cases. Of these, 24% preceded the esophageal carcinoma while 19.5% were synchronous. Resection was curative (i.e. grossly complete) in 144 cases, palliative (leaving gross tumor behind) in 25 cases, and judged impossible in 31 cases. Twenty-one patients died during their hospitalization. Five-year survival, according to the Kaplan-Meier method, was 18 +/- 3.6% for all patients, and 27 +/- 4.8% for patients undergoing curative resection (including postoperative mortality). Palliative resections or invasion of adjacent organs, but not resection with positive lymph nodes, was thought to be synonymous with incurability. Multivariate analysis showed that the two principal statistically significant characteristics were the stage as determined by roentgenological and endoscopic evaluation (P = 0.003) (clearly assessable during initial work-up) and type of resection (complete or not) (P = 0.04). Tumor diameter, as determined on CT-scan (< or > than 30 mm), was at the limit of statistical significance (P = 0.09). A second multivariate analysis of prognostic factors was conducted in 128 patients undergoing resection with curative intent and who survived the initial postoperative period. Roentgenologic/endoscopic staging alone emerged as being statistically significant (P = 0.001).(ABSTRACT TRUNCATED AT 250 WORDS)