Tumor characteristics believed to predict for deep muscle invasion after transurethral resection of superficial bladder cancer alone were analyzed in 51 patients with stage T1 (stage A) transitional cell carcinoma of the bladder. All patients were treated at Stanford University Medical Center and none had intravesical chemotherapy at any time during the median followup of 78 months. No patient had carcinoma in situ. A total of 14 patients (27%) had deep muscle invasion. None of the patients with grade 1, 5 (22%) with grade 2 and 9 (50%) with grade 3 to 4 tumors had deep muscle invasion. Comparison of the risk of muscle invasion using pathological tumor grade at diagnosis, highest grade at any cystoscopic biopsy before the diagnosis of muscle invasion or highest grade at cystoscopic biopsy immediately antecedent to the cystoscopy at which muscle invasion was diagnosed all showed similar probability of muscle invasion. Mean interval to development of muscle invasion was 6 and 12 months in the grades 2 and 3 to 4 groups, respectively. At 36 months the cumulative probability of invasion-free survival was 62% for grade 2, compared to 50% for grades 3 to 4 cancer patients (p less than 0.005, Gehan). Univariate regression analysis demonstrated grade to be the only significant factor in predicting for invasive disease (p = 0.005), with tumors in the bladder neck to be of borderline significance (p = 0.159). On multivariate logistic regression analysis, grade remained the single tumor variable predicting for invasive bladder cancer (p = 0.004). These data suggest that of routinely available data at diagnosis and during subsequent followup cystoscopic examinations, tumor grade is the most important biological predictor of progression to muscle invasive cancer.