Patients with lung cancer frequently suffer from brain metastases at the time of presentation. This condition affects approximately 10% of non-small cell lung cancer (NSCLC) patients (Newman and Hansen, 1974;Sorensen et al, 1988). Surgery is feasible only for a small proportion of these patients. Whole brain radiotherapy has been, hitherto, the generally recommended treatment in inoperable patients. The survival of NSCLC patients with brain metastases is poor, reported to be between 3 to 6 months in patients treated with medical therapies, either radiotherapy or chemotherapy (compared to 6-10 months in other advanced NSCLC (Paesmans et al, 1995;Shepherd, 1999)). Furthermore, brain metastases at the time of presentation of lung cancer seems to be a worse prognosis (Sorensen et al, 1988) than metachronous brain metastases.New therapeutic strategies are needed to improve the outcome of these patients. The knowledge of prognostic determinants might be important in both clinical trials and routine practice (Komaki et al, 1993;Charloux et al, 1997;Paesmans et al, 1997;Merrill et al, 1999). In the former setting, prognostic co-variables must be taken into account in survival analyses; by way of illustration, in a given randomized trial, the statement that a difference in survival is related to the effects of the treatment must be supported by a proportional hazards model demonstrating that this effect does not depend on well-known prognostic determinants (Depierre et al, 1999;Furuse et al, 1999). In the second setting, a therapeutic decision might be influenced by the state of prognostic variables (Komaki et al, 1993).Here we especially take a look at the prognostic significance of 2 specific serum markers, CYFRA 21-1 and neuron-specific enolase (NSE). The prognostic value of CYFRA 21-1 (a fragment of cytokeratin subunit 19) in this disease has been suggested (Pujol et al, 1993;Wieskopf et al, 1995;Brechot et al, 1997). NSE, the γ-subunit of enolase, has been widely investigated as a marker of small cell lung cancer (SCLC;Jorgensen et al, 1989). Although only a small proportion of NSCLC presented with a high NSE level, this marker might indirectly reflect; i) a neuroendocrine component of the disease in favour of tumour heterogeneity; ii) a Summary A multi-centre retrospective study involving 4 French university institutions has been conducted in order to identify routine pretherapeutic prognostic factors of survival in patients with previously untreated non-small cell lung cancer and brain metastases at the time of presentation. A total of 231 patients were recorded regarding their clinical, radiological and biological characteristics at presentation. The accrual period was January 1991 to December 1998. Prognosis was analysed using both univariate and multivariate (Cox model) statistics. The median survival of the whole population was 28 weeks. Univariate analysis (log-rank), showed that patients affected by one of the following characteristics proved to have a shorter survival in comparison with the opposite status...