All patients treated in NABTC phase II trials for recurrent disease during the period February 1998 through November 2008 were included in this study (Table ). Because of differences in the types of patient data that were collected, data for patients treated before January 2005 were analyzed as one data set (older studies), and patients treated more recently were analyzed as a second data set (newer studies). Some of the older studies included temozolomide (TMZ) as one of the treatment agents. Because this is now recognized as an effective agent, data from the older studies were stratified on the basis of whether TMZ was part of the treatment regimen. None of the newer studies included TMZ, so this stratification was not necessary. Some studies included both a phase I and a phase II component; for the purposes of this analysis, all patients treated with the recommended phase II dose and who met the phase II eligibility criteria were included even if they were enrolled in the phase I portion of the study. Patients treated with the other phase I doses (lower or higher than the recommended phase II dose) were excluded.
Numbers of patients undergoing resection on North American Brain Tumor Consortium phase II trials
Standard trial entry criteria included confirmed high-grade glioma (grades III and IV) and a Karnofsky performance status score (KPS) ≥60. All protocols required central pathology review; in the few cases in which tissue samples were not available for central review, local pathology designation was accepted. Diagnosis was based on the most recent surgery for which data were available at the time of protocol registration, and only those patients with a diagnosis of GBM were included in this analysis. Patients could have been enrolled in >1 protocol and, if so, were included for each protocol in which they were enrolled. In total, 49 patients enrolled in >1 protocol, accounting for a total of 100 observations. Analyses were repeated including these patients only once, either for the first or the last protocol in which they were enrolled; because the results were substantially the same, only results from the primary analysis are presented (ie, with patients included for each protocol in which they were enrolled).
For all studies, progression was defined using the criteria of Macdonald et al.5
Because the primary end point for these studies was PFS6, either evaluable disease (unidimensionally measurable lesions or margins not clearly defined) or measurable disease (bidimensionally measurable lesions with clearly defined margins) was allowed for patients not undergoing surgery at recurrence. For patients who underwent surgery, there was no requirement for the presence of residual tumor postoperatively. Progression was determined by the local institutional investigator and was defined as a new lesion or an increase in tumor size of ≥25% for measurable disease and clear worsening for evaluable disease. Failure to return for evaluation due to death or deteriorating condition was considered to represent progression. In this case, the date that the patient was declared off-treatment due to progression was used as the progression date.
For assignment to the surgery versus nonsurgery groups for this study, patients were initially categorized on the basis of whether they were on the surgery arm of a trial that included a subset of patients treated with the experimental therapy prior to surgery. For those not on the surgery arm of a trial, the time of the most recent surgery was determined. For the older studies, the date of the progression that qualified the patient for the study was not known; it was assumed that if the surgery occurred within 30 days of registration, it was for the most recent progression, and the patient was included in the surgery group. For the newer studies, the date of the qualifying progression was known, and patients were placed in the surgery group if the date of surgery was later than the date of progression. Eighteen patients met the criteria for being in the surgery group but only underwent a biopsy; these patients were excluded from the analysis. Seventeen patients on the surgery arm were not able to receive treatment after surgery; these patients were excluded because they received no treatment with therapeutic intent.
PFS and OS were measured from time of study registration for patients who did not undergo surgery on protocol and for those who underwent surgery prior to study enrollment. For patients who did undergo surgery as part of a study, the date of first postoperative treatment was used as the baseline date. Because of the retrospective nature of this analysis, the date of the start of postoperative chemotherapy was not available for 17 patients. In those patients, we imputed the date using the latest of date of first response assessment minus 8 weeks (12 cases), surgery date (2 cases), and registration date (when surgery date was not known; 3 cases). Surgery date (or registration date) represented conservative estimates, because postoperative chemotherapy would have to have been started after these dates. The median time from earliest possible chemotherapy start date to imputed start date for those where response assessment minus 8 weeks was used was 16.5 days (range, 1–29 days). This is consistent with the expected time from surgery and is not sufficiently long to substantially affect results. Therefore, no additional sensitivity analyses were conducted.
Patients not known to have died were censored for survival as of the last date known alive. In absence of a progression date, death ≤30 days after the end of treatment was considered date of progression. If a patient was removed from treatment for a reason other than progression, that patient was censored for further evaluation of progression as of the date of starting other therapy (if that was known); if not, the date of removal from treatment was used. In cases in which follow-up for progression was not consistent once off-treatment, the off-treatment date was used.
Studies typically required repeated imaging every 8 weeks. Because the actual timing of the scans could vary, rules were developed to determine whether there was sufficient information to declare that a patient met the criteria for success for PFS6. Specifically, if the PFS duration was ≤26 weeks, then treatment was considered a failure; if the duration of PFS was >30 weeks, then treatment was considered a success. For those patients who progressed between 26 and 30 weeks, we looked for an indication of when the last stable scan was documented. Documentation of progression-free status at the 24-week scan was sufficient to declare success. If the information was not available and PFS was <28 weeks, then we assumed the treatment failed before week 26. For the 3 patients remaining, information on the last progression-free scan was not available, and PFS6 was considered to be unknown.
Both PFS and OS were estimated using the Kaplan-Meier method. To allow for some variability in timing of the scans, we summarized PFS status at 9, 18, and 26 weeks. Survival curves comparing outcome based on surgical status were created. For those analyses including patients in studies involving TMZ, the data were stratified on the basis of whether TMZ was part of the treatment. For the primary end point of PFS6, comparison between the surgery and nonsurgery groups was based on logistic regression. Time-to-event analyses were conducted using the Cox proportional hazards model. All models included adjustment for age, KPS, and TMZ, when appropriate. All P values presented in this article are 2-tailed.