A series of 71 patients with muscle invasive bladder cancer received a median of 3 cycles (range 1-6) of methotrexate, vinblastine, Adriamycin and cisplatin (M-VAC). Efficacy assessed by transurethral resection alone showed that 48% of patients were TO, 13% Tis and 54% had normalisation of initially positive urinary cytology after treatment. However, when considering transurethral resection of the bladder (TURB), cytology and non-invasive procedures (CT scan and/or ultrasound), only 21% had a clinical complete remission (cCR); 48 patients (68%) had pathological evaluation and 13 (27%) were PO after treatment. Non-responding patients had a poor prognosis: 14/30 (47%) developed metastatic disease and 13 died. In assessing the primary lesions, clinical understaging was significant. Of 15 patients who were TO cystoscopically prior to surgery, 6 (40%) had residual disease in the pathological specimen, including 4 with muscle infiltration; 23 patients (32%) remained clinically staged, only 8 of whom remain disease-free. With a median follow-up of 24 months (range 2-42+), 41 patients are alive and disease-free, including 20 with a functional bladder. The large staging error raises questions concerning studies using clinical rather than pathological endpoints as the sole criteria of efficacy.