The prognostic importance of tumor size was studied in 510 patients with malignant glioma (80% with glioblastoma multiforme) in the Valid Study Group of Study 80-01 of the Brain Tumor Study Group (now the Brain Tumor Cooperative Group [BTCG]). The endpoint was length of survival from randomization, which occurred within 3 weeks of definitive surgery. Following randomization, patients were scheduled to receive radiotherapy (RT) (6,020 cGy) during a 7-week period, along with continuing courses of chemotherapy. Computed tomographic (CT) scan information was available for 124 patients preoperatively, 300 patients postoperatively (preradiation), and 218 patients 9 weeks post-RT (+/- 3 weeks). Tumor size was determined as area (length x width) on the contrast-enhanced scan and survival was compared by log rank statistics. Preoperative tumor area was unrelated to survival (P = .48), but postoperative area was significantly prognostic (P less than .0001); the smaller the residual tumor, the longer the patient lived. Patients with a 75% or greater resection, as determined by measuring the difference between the preoperative and the postoperative scans, tended to have better survival, but the difference was not significant (P = .16). The post-RT area was strongly related to survival (P less than .00001). The percent change in area between the pre- and post-RT scans was also prognostic. Tumor size was of prognostic importance independent of the other known prognostic variables: age, Karnofsky performance score, and whether the tumor was glioblastoma or anaplastic astrocytoma. We conclude that the amount of tumor remaining after surgery is an important baseline variable at the start of RT, and that the tumor size 9 weeks following RT is also prognostic. Surgical resection is most important when it leaves the least amount of residual tumor.