The purpose of this study was to determine the prognostic significance of a high pretreatment serum CYFRA 21-1 level (a cytokeratin 19 fragment) adjusted for the effects of well-known co-variables in non-small-cell lung cancer (NSCLC). This meta-analysis based on individual updated data gathered comprehensive databases from published or unpublished controlled studies dealing with the prognostic effect of serum CYFRA 21-1 level at presentation in NSCLC of any stage (nine institutions, 2063 patients). Multivariate regression was carried out with the Cox model. The proportional hazard assumption for each of the selected variables retained in the final model was originally checked by log minus log plots baseline hazard ratio. The follow-up ranged from 25 to 78 months. A total of 1616 events were recorded. In the multivariate analysis performed at the 1-year end point, a high pretreatment CYFRA 21-1 level was an unfavourable prognostic determinant in all centres except one (Hazard ratio (95% confidence interval): 1.88 (1.64-2.15), P<10(-4)). Other significant variables were stage of the disease, age and performance status. Within the first 18 months, the procedure disclosed a nearly similar hazard ratio for patients having a high pretreatment serum CYFRA 21-1 level (1.62 (1.42-1.86), P<10(-4)). For patients who did not undergo surgery, the hazard ratio during the first year of follow-up was 1.78 (1.54-2.07), P<10(-4). Finally, in the surgically treated population, at the 2-year end point, a high pretreatment CYFRA 21-1 and a locally advanced stage remained unfavourable prognostic determinants. In conclusion CYFRA 21-1 might be regarded as a putative co-variable in analysing NSCLC outcome inasmuch as a high serum level is a significant determinant of poor prognosis whatever the planned treatment.