BACKGROUND The specific cell type in primary lung cancer (LC) has important consequences, both for treatment and prognosis. Our objective has been to evaluate the degree of accuracy of several preoperative techniques in the LC diagnosis. METHODS We have studied 442 diagnostic techniques realized in 360 patients who underwent thoracotomy for LC in our hospital (1988-1994). Twelve sputum cytologies, 93 transthoracic needle biopsies (TNB), 73 bronchial aspirated cytologies, 30 bronchial brushing, 205 bronchial biopsies (BB) and 29 transbronchial biopsies were included. In each case the cellular type of LC, obtained through these techniques, was compared with the result derived from the study of surgical piece (reference diagnosis). To calculate the degree of concordance between both diagnoses we used the kappa coefficient (K). RESULTS The overall concordance between the cellular type observed in the different preoperative techniques and the definitive result obtained by thoracotomy was 0.61. The worst test was TNB (K = 0.41). Sputum cytology and BB presented a good histopathological precision (K = 0.75 and 0.70 respectively). The rest of techniques showed moderate results. On the different histologies, the best result was obtained in squamous carcinoma (K = 0.68) and the worst in undifferentiated large cell carcinoma (LCC) (K = 0.39). Small-cell lung cancer only showed a moderate concordance (K = 0.58). The coexistence and coincidence of two or more tests with the same cellular type were associated with a higher proportion of accuracy (0.97 vs 0.73; p = 0.00002; odds ratio: 12.02). CONCLUSIONS The preoperative histopathological diagnoses should be interpreted with caution, especially those obtained by TNB and those where it is implied the LCC. In these two circumstances and in cases in which the knowledge of the cellular type have relevance, we think that the initial diagnosis should be reinforced with a second result. The precision of the preoperative diagnoses in small-cell lung cancer must be reevaluated.