Only 5-10% of patients with lung cancer (LC) can be expected to be cured by radical treatments. In the remaining subjects the potential survival benefit of treatment must be weighed, taking into consideration the possible deterioration of quality of life (QL). Indeed, studies dealing with different aspects of QL are being increasingly reported in LC. In a few of them, the interesting observation was made that the patient-rated QL correlated well with the subsequent clinical outcome. In the present study we analyse 11 items of the Therapy Impact Questionnaire (a new instrument of QL), assessing both disease and therapy impact on physical condition, functional status, concomitant emotional and cognitive factors and social interactions. Questionnaires were completed by 128 consecutive LC patients, who had been seen, in the years 1990 through 1993, either for a newly diagnosed cancer (40 patients), or after a successful operation (15 patients), or during active and/or symptomatic treatment (73 patients). At the time of the QL assessment, a minimal set of demographic and clinical variables was recorded. Univariate tests of survival showed that stage of disease, difficulty at work or doing the housework, weight loss, performance status, difficulty relaxing, having been felt unsure, and tumor cell type were all associated, in decreasing order of significance, with prognosis. QL variables correlated well with each other, but poorly with clinical and demographic variables (an expected exception was the good correlation existing between working capacity/physical autonomy and the corresponding observer evaluation of performance status). This lack of correlations explains how QL variables maintained their significance in multivariate survival analyses. In the best multivariate model, the self-estimated difficulty at work or doing the housework followed the stage of disease, but preceded weight loss as a significant, independent, prognostic determinant. Further studies evaluating several other additional prognostic indicators are needed to better clarify the relative prognostic importance of quality of life.