Unknown primary carcinoma: natural history and prognostic factors in 657 consecutive patients.

PURPOSE To evaluate the natural history, validate previous observations, and identify prognostic factors in patients with unknown primary carcinoma (UPC). PATIENTS AND METHODS Nine hundred twenty-seven consecutive patients referred to the M.D. Anderson Cancer Center with a preliminary diagnosis of UPC were prospectively identified. A standardized evaluation narrowed the study population to 657 patients with UPC. All data were recorded and computerized for storage, retrieval, and analysis. The primary end point for the study was survival, which was calculated from the first day of patient registration. Survival curves were estimated using the Kaplan-Meier method and compared using the Cox-Mantel log-rank test. To identify important prognostic factors, univariate and multivariate analyses were conducted. RESULTS The demographics of the UPC patient population mirrored those of the general population of patients referred to our cancer center except for an excess of men among the UPC patients. Most patients had histologic or cytologic evidence of adenocarcinoma and had more than one organ site metastatically involved. Univariate and multivariate analyses identified numerous important prognostic factors with a significant influence on survival, including sex, number of organ sites involved, specific organ sites involved, and pathologic subtypes. CONCLUSION This study validated previously identified important prognostic factors for survival in UPC. Additional variables that had an impact on survival were identified and the complex interaction of the factors was explored. As patient numbers increase, this database will be able to provide further analyses of patient subsets and potentially relate specific clinical features to the evolving molecular and biochemical understanding of these malignancies.