In patients with malignancies, thrombocytosis has previously been related to disease stage, histological type, and survival. In the present study, the prevalence of thrombocytosis and the prognostic information provided by platelet counts were analysed in a large cohort of patients with primary lung cancer. At the time of diagnosis, pretreatment platelet counts were retrospectively recorded in 1,115 consecutive patients with histologically proven primary lung cancer. All patients were reviewed regarding histological type, tumour, node, metastasis (TNM) classification stage and survival. The prevalence of thrombocytosis in patients with lung cancer was compared with that in a series of 550 consecutive out-patients with benign lung disorders. In 269 surgically resected patients, postoperative platelet counts were recorded 1-3 months after resection of the tumour. In the follow-up period, thromboembolic episodes diagnosed either clinically or at autopsy were recorded. The overall prevalence of thrombocytosis (> 400 x 10(9) platelets.L-1) in the patients with lung cancer was 32%. The frequency of thrombocytosis was significantly higher compared with the control subjects (32 vs 6%; p < 0.0001). Platelet counts differed significantly among subgroups defined by the TNM classification, with the proportion of patients with > 400 x 10(9) platelets.L-1 greatest in the more advanced TNM stages (stage I and II 23% vs stage III and IV 37%; p < 0.0001). Patients with thrombocytosis had a significantly poorer survival than patients with normal platelet counts (p < 0.0001). In a multivariate survival analysis (Cox model), thrombocytosis continued to correlate strongly with poor survival even when adjusted for histological type, sex, age, and TNM stage (p < 0.001). In surgically resected patients, the frequency of preoperative and postoperative thrombocytosis differed significantly (23.0 vs 8.9%; p < 0.0001). Survival rate was significantly reduced in patients with preoperative thrombocytosis (p = 0.005). Thrombocytosis was not associated with an increased incidence of thromboembolism. In conclusion, thrombocytosis is an independent prognostic factor of survival in patients with primary lung cancer. We suggest that platelet counts should be included in future multivariate analyses of survival in patients with lung cancer.