Currently, no further therapy in addition to surgery is recommended in completely resected NSCLC stage I patients. However, the 5-year survival rate at this stage has been reported to be approximately 60%, i.e. 40% of patients had a lower survival rate. The aim of the study was to identify those patients at increased risk by using the tumor markers CYFRA21-1 and CEA as prognostic factors. One hundred and fifty-three stage I NSCLC patients, who were treated exclusively by surgery between 1996 and 1998, entered this retrospective study. It was shown, by multivariate analysis, that elevated CYFRA 21-1 (>3.3ng/ml) and CEA (>9.8ng/ml) levels were associated with a worse outcome in 21.3% and 13.1% of the patients under study, respectively. The corresponding 3-year survival rates were found to be 60.2% for increased CYFRA 21-1 levels (p=0.029) and approximately 40% for increased CEA levels (p=0.022), compared to a rate of 78.4% and 79.0% in case of normal marker levels, respectively. The relative risk (95% confidence interval) was found to be 2.156 (1.08-4.29) for elevated CYFRA 21-1 and 2.707 (1.15-6.36) for elevated CEA. The detection rate for the identification of patients with worse outcome increased when a combination of both markers was used. Thereby, it was possible to identify 32% of patients where one or both markers were elevated. The 3-year survival rate was 55.7% in this group compared to that of 82.5% in those patients where both markers were in the normal range (p=0.0014). In order to consider the degree of marker elevation that is thought to reflect tumor burden, we introduced a tumor marker index (TMI) corresponding to the geometric mean of normalized CYFRA21-1 and CEA levels (marker value divided by diagnostic cut-off). Thereby, we were able to identify 3 groups of patients at different risk levels: the first group (22.7%) had a 3-year survival rate of 96.7%, the second group (42.6%) had one of 77.2% and the third group (34.7%) had one of only 55.7%. In conclusion, elevated CYFRA 21-1 and CEA levels were able to identify a group of curatively operated NSCLC patients who were at high risk of early death. Those patients may benefit from more aggressive treatment approaches. The group of patients with a 3-year survival rate of 96.7% probably does not need further treatment.