Radical radiotherapy is widely used to treat inoperable non-small-cell lung cancer (NSCLC) although only a small number of patients benefit in the long run from this intensive treatment. There is a small proportion of long-term survivors who might derive advantage from even more aggressive radiotherapy combined with chemotherapy. In order to support optimal treatment selection we have carried out univariate and multivariate analyses of possible prognostic variables in the retrospective data of 502 NSCLC patients treated at one institute with external radiotherapy, both with curative and palliative intent. To obtain more accurate tools for a rational treatment decision, we identified, by using Cox's proportional-hazards model, the five most powerful determinants of overall survival and combined them to a prognostic index. On the basis of only the number of these risk factors (advanced stage, general or metastatic symptoms, poor performance status, anemia and tumor size of at least 7 cm), the patient falls into one of the six possible prognostic groups and these groups turned out to be identifiable as separate prognostic clusters. Thirty-one per cent of the patients have three or more risk factors and a median survival of 5-7 months compared with 18 months for patients without any non-favorable factor. Furthermore, the prognostic factors were so strong that multivariate analyses did not reveal the treatment selection to have any significant influence on survival. As each of the five variables have the advantage of being routinely available, our index is simple enough to be used in daily clinical practice. The clinical value of the prognostic index should be verified by using independent data.