BACKGROUND The Intergroup Rhabdomyosarcoma Study Group (IRSG) studies began in 1972 and initially used a clinicopathologic system to place patients into prognostic groups. Because of interest in the development of a pretreatment staging system for assessing the posttreatment outcomes of patients with this disease, potential staging elements were retrospectively evaluated in a subset of 505 patients who participated in IRS-II, an IRSG clinical trial. METHODS Using the IRS-II data, a TNM pretreatment staging system was developed and used to stage prospectively the patients who were entering IRS-III, a subsequent treatment protocol of the IRSG. Failure free survival and overall survival were compared by pretreatment stage in IRS-III as a means of evaluating this TNM staging. RESULTS The TNM staging system described the tumor (T) in terms of lesion size (< 5 cm or > or = 5 cm) instead of invasiveness, because these two features were not independent of each other. The clinical status of regional lymph nodes (N) was included in the staging system, as was the presence or absence of metastatic disease (M). The latter feature was extremely important, as expected. The anatomic site of the primary tumor also proved to be an important staging element. Classification of patients by tumor size, clinical status of regional lymph nodes, presence or absence of metastatic disease, and location of the primary tumor (at a favorable or unfavorable anatomic site) created four prognostically distinct staging categories that were relatively equal in size. In a prospective evaluation of this staging system with IRS-III patients, the pretreatment staging lost some prognostic impact. The survival of patients with smaller lesions at unfavorable anatomic sites without clinically involved lymph nodes (Stage II) was similar to that of patients with primary tumors at favorable anatomic sites (Stage I). CONCLUSIONS A pretreatment TNM staging system for childhood rhabdomyosarcoma, developed with data from IRS-II, was not as predictive of patient outcome when applied prospectively to patients treated in the IRS-III trial. These findings could be due to differences in the management strategy used for IRS-III or the statistical variability in the model-fitting process used to develop the staging system. This demonstrates the need for continual reevaluation of staging systems as patient evaluation and treatment innovations are developed.