The TNM classification (tumor-node-metastasis) was adopted by the American Joint Committee on Cancer and the International Union against Cancer a decade ago to avoid heterogeneity of prognostic classification schemes used for differentiated thyroid cancers. To date, however, clinical data based on this classification are lacking. We retrospectively evaluate the prognosis of 700 patients (208 men and 492 women) with papillary (89%) and follicular (11%) thyroid cancers according to the pathological TNM (pTNM) staging system, treated over a 25-yr period (1970-1995). Patients who received primary treatment at our center constituted 87.4% of the cases; the majority underwent total thyroidectomy, followed by 131I ablative therapy in high risk groups, as standard treatment. Clinical and follow-up data were obtained from the medical records and our cancer registry. Disease-free and cancer-specific survival data were analyzed by Kaplan-Meier product limit estimates and Cox proportional hazard models. Patient distribution by the pTNM system were: stage I, 516 patients; stage II, 57 patients; stage III, 104 patients; and stage IV, 23 patients. Over a mean +/- SE follow-up of 11.3 +/- 0.3 yr, the overall cancer recurrence and mortality rates were 20.5% and 8.4%, respectively. However, the respective cancer recurrence and mortality rates were distinctly different in the various pTNM stages: 15.4% and 1.7% in stage I, 22% and 15.8% in stage II, 46.4% and 30% in stage III, and 66.7% and 60.9% in stage IV tumors. Using actuarial survival plots, a clear separation in both disease-free survival and cancer-specific survival was noted among all the stages (P < 0.0001). Risk factors analyses showed a significant association between all the prognostic variables used in TNM staging (age, tumor size, extent of primary tumor, and presence of nodal or distant metastases) and the observed end points of recurrence or death from thyroid cancer. After correcting for TNM stages, the risk of cancer recurrence was halved in female compared to male patients, whereas this was 1.7-fold higher in multifocal than unifocal tumors. Conversely, cancer mortality was 3.4-fold higher in follicular than papillary thyroid cancer. In the analysis of effect of primary treatment among 492 patients with tumor more advanced than the T1N0M0 category, patients who underwent less extensive surgery (lobectomy or subtotal thyroidectomy) had a 2.5-fold risk of cancer recurrence (P < 0.0001) and a 2.2-fold risk of death (P < 0.01) compared to those who underwent total or near-total thyroidectomy. Patients not treated with 131I ablation had a 2.1-fold greater risk of cancer recurrence (P < 0.0001) than those given 131I ablation, although no difference was noted in deaths from thyroid cancer. Based on our data, the pTNM classification is useful in distinguishing patients with different prognostic outcomes. However, the small patient numbers in pTNM stages other than stages I precludes us from evaluating its usefulness as a guide for therapy. Until prospective data could be accrued from controlled treatment trials, we support the standard practice of total thyroidectomy followed by 131I ablative therapy (if focal iodide uptake was noted) in patients with papillary thyroid cancer more advanced than the T1N0M0 category or of multicentric nature and in the majority of patients with follicular thyroid cancer.