We have shown previously that hepatocyte growth factor (HGF), which is produced by lung fibroblasts, is a potent mitogen and motogen for both normal and neoplastic bronchial epithelium, and that expression of the HGF receptor, the c-met proto-oncogene protein, is uniformly found in the human bronchial epithelium and in non-small cell lung carcinomas (NSCLCs; P. Singh-Kaw et al., Am. J. Physiol., 268: L1012-L1020, 1995). Yamashita et al. have reported an association of HGF with poor survival in invasive ductal carcinoma of the breast (Cancer Res., 54: 1630-1633, 1994). There are few prognostic markers for lung cancer, and the high recurrence rate for stage I lung cancer suggests the frequent presence of undetectable tumor burden in such patients. Criteria are needed to evaluate these patients for risk of recurrence. We have now evaluated whether HGF present in resectable lung tumors has prognostic significance. In this study, 56 primary NSCLCs, mainly adenocarcinomas, were examined for presence of HGF by quantitative Western blot. These tumors consisted of tissue from 34 stage I patients, 9 stage II patients, and 13 stage IIIa patients who underwent curative resection for primary NSCLC. Extracts of whole tumor tissue were analyzed after separation of proteins by electrophoresis and transfer of proteins to nitrocellulose membranes. Immunoreactive (ir)-HGF was visualized by reaction with a polyclonal anti-HGF antiserum and quantitated by densitometry. Lung tumor content of ir-HGF varied widely among individuals. Median ir-HGF content in tumor extracts was 15.3 ng/40 microg of tumor protein; mean ir-HGF was 27.2 ng/40 microg of tumor protein. The median and mean ir-HGF were both significantly higher in tumor tissue from patients who suffered a recurrence during the follow-up period compared with those with no evidence or residual disease; this was true of all patients (P = 0.0001) and stage I patients analyzed separately (P = 0.002). Analysis of survival curves indicated that ir-HGF levels higher than the median were associated with poor overall survival (P < 0.03). Univariate analysis showed three factors related to poor overall survival in this set of patients: ir-HGF, tumor (T) status (a measure of primary tumor size and extent), and age. Nodal (N) status and stage were only marginally related to overall survival, most likely because the majority of the patients in the study were stage I. N status, stage, and T status were related to disease-free survival, however. Multivariate Cox analysis showed that ir-HGF, T status, and age independently had a negative impact on overall survival. ir-HGF was a strong independent negative prognostic indicator (P = 0.0001) with a relative risk of 1.022 per unit of ir-HGF (ng/40 microg of protein). This demonstrates that, in this group of patients, the relative risk of ir-HGF content increased continuously as ir-HGF increased, and exceeded 10 at units of ir-HGF of 100 or more. In comparison, in this group of patients, the relative risk of a T status greater than 1 was 4.75 and that of age greater than 65 was 3.95. The combined negative effect of a T status greater than 1 and elevated ir-HGF on survival was also highly pronounced (P < 0.005). In addition, elevated ir-HGF had a negative impact on survival when patients were stratified by stage or N status. Stage I patients with high ir-HGF values had a worse outcome than stage II or stage IIIa patients with low ir-HGF values. Elevated ir-HGF was strongly associated with poor outcome for resectable NSCLC patients as a group, and also identified stage I patients with poor outcome, indicating that it could be a useful indicator of risk of relapse and death in patients who have early lung cancer. The impact of elevated ir-HGF was especially prominent in patients whose T status was greater than 1, suggesting that patients with both risk factors who are stag