The outlook for the majority of adult patients with malignant gliomas has not improved since the Brain Tumor Study Group (BTSG) trials of the 1970s ascertained a benefit from 55 to 60 Gy of fractionated external-beam radiotherapy and carmustine(Drug information on carmustine) (BiCNU) following surgery [1-5]. The median survival time of patients who participate in formal clinical trials is approximately 1 year [2-4,6-12]. This dismal figure is almost certainly an overestimate, since powerful selection factors lead patients with better prognostic factors to enter clinical trials [8,9,13,14]. Patient age, functional status, and tumor grade are the most important prognostic factors [1,7,15]. It is only in the subgroup of patients with more favorable prognostic signs that selection of therapy has any impact on survival .
In 1960 Bouchard and Pierce reported that life expectancy in patients with glioblastoma multiforme was better in those who received radiotherapy than in those treated with surgery alone . Unfortunately, only 4% of patients in the combined-modality treatment group were alive at 10 years.
The BTSG conducted several randomized prospective postoperative clinical trials in patients with malignant gliomas [1,3,5]. In the first of these studies, the four treatment arms included surgery plus: supportive therapy alone, carmustine alone, 50 to 60 Gy of whole-brain radiotherapy alone, and carmustine plus radiotherapy. The vast majority of the patients (90%) were classified as having glioblastoma multiforme, 9% had anaplastic astrocytoma, and 1% had other anaplastic gliomas.
The approximate median survival times, in weeks, for the four treatments were: supportive care, 14; carmustine, 19; radiotherapy, 36; and carmustine plus radiotherapy, 35. The difference in survival times between radiotherapy and either carmustine alone or supportive care was statistically significant. This study thus constituted the first demonstration, in a randomized trial, that radiotherapy significantly increased median survival time for patients with glioblastoma multiforme.
In a subsequent BTSG trial (trial 7201), median survival time for treatment that included methyl-CCNU but not radiotherapy was 24 weeks. Significantly better survival times were seen with conventional radiotherapy alone (36 weeks), radiotherapy plus methyl-CCNU (42 weeks), or conventional radiotherapy plus carmustine (51 weeks).
In the third major BTSG study (trial 7501), radiotherapy plus methylprednisolone(Drug information on methylprednisolone) achieved a median survival time of 40 weeks, and radiotherapy plus methylprednisolone plus carmustine attained a survival time of 41 weeks. Significantly better survival times were reported with both radiotherapy plus procarbazine(Drug information on procarbazine) (47 weeks) and radiotherapy plus carmustine (50 weeks) [1-3].
Over a decade after BTSG trial 7201, the Central Nervous System Cancer Consortium (CNSCC) conducted a randomized trial of conventional radiotherapy plus carmustine vs conventional radiotherapy plus diaziquone (AZQ). The median survival time from randomization of all patients was 351 days, corresponding to approximately 69 weeks after diagnosis. Since randomization in this study occurred approximately 8 weeks after completion of radiotherapy, or 15 to 20 weeks after diagnosis, this median survival time is based on a subpopulation of patients who survived and maintained a Karnofsky performance status (KPS) of at least 50% for 4 to 5 months after diagnosis. This explains why the overall median survival in this study exceeds that reported in most other trials [6,11].
Adjuvant Chemotherapy--Among the most extensively examined areas in the treatment of malignant gliomas is the use of adjuvant chemotherapy. Phase II trials of the nitrosoureas documented a 10% to 40% response rate for recurrent tumors . Because of this activity, carmustine and methyl-CCNU were utilized in adjuvant chemotherapy trials. Based on the previously cited BTSG trials, many would argue that the best standard therapy of patients with malignant gliomas should include nitrosourea chemotherapy. However, only one of the three BTSG randomized trials that evaluated adjuvant chemotherapy and included a no-chemotherapy arm showed a statistically significant benefit of chemotherapy.
There are almost a limitless variety of "new" chemotherapy programs to be tested in malignant gliomas or "old" programs to be recycled with slight changes. A host of single- and multi-agent programs have been tried, including carmustine; lomustine; methyl-CCNU; bleomycin(Drug information on bleomycin) (Blenoxane); procarbazine, lomustine(Drug information on lomustine), plus vincristine (Oncovin); AZQ; and mitomycin(Drug information on mitomycin) (Mutamycin) plus mercaptopurine(Drug information on mercaptopurine) [11,12,18].
To approach the question of the role of adjuvant chemotherapy in malignant gliomas, Fine et al performed a meta-analysis of the combined results from 16 randomized trials involving more than 3,000 patients . They found a survival benefit attendant to the use of chemotherapy, and that benefit occurred earlier in patients with anaplastic astrocytoma than in those with glioblastoma multiforme.
The survival times for malignant glioma following surgery, conventional radiotherapy, and chemotherapy seem to have plateaued over the past 25 years. In the free market of ideas, various "new" therapies are held out as innovative and potentially efficacious. Some of these alleged new ideas are, in fact, previously investigated, venerable concepts that have resurfaced [19,20]. Examples include multiple daily fractions of external-beam radiotherapy ; interstitial brachytherapy1 [21-25]; localized or whole-body hyperthermia [26,27]; dose escalation with conventional fractionated radiotherapy [28,29]; radiosensitization with metronidazole(Drug information on metronidazole), bromodeoxyuridine (BUDR), iododeoxyuridine (IUDR), and neutron/boron capture [30-34]; radiolabeled monoclonal antibodies; high-dose chemotherapy with autologous bone marrow rescue; and gene therapy with retroviral vectors.
Radiosurgery--Stereotactic radiosurgery is a popular area of current clinical investigation. Radiosurgery utilizes multiple beams to tightly concentrate radiation dose to the tumor with relative sparing of normal tissue. The most commonly used radiation sources for radiosurgery are high-energy x-rays produced by a linear accelerator; cobalt-60 gamma rays, as provided by the commercially available Gamma Knife (Elekta Instruments, Decatur, Georgia); and particle beams, such as the Harvard cyclotron beam.
Either before or following fractionated external-beam radiotherapy, a radiosurgery "boost" may be given to the bulk of the tumor-provided that the tumor volume does not exceed the limits of the technology, ie, equal to or less than 3 to 4 cm. Fewer than 20% of malignant gliomas meet this criterion [9,13,35]. Multiple noncoplanar beams can be used as a boost for such larger lesions. To improve the tolerance of normal tissues for radiosurgery of larger volumes, the therapy is more highly fractionated. Thus, at some point, highly fractionated radiosurgery intellectually merges with precision conventional external-beam radiotherapy.
Study after study shows that patient age, tumor grade (glioblastoma multiforme vs anaplastic astrocytoma), and patient performance status are the pretreatment variables most predictive of outcome [10-12,15,36,37]. In many clinical trials, the benefit of a new therapy becomes less apparent in patients over 45 to 55 years of age who have poor performance status. This observation is underscored by the fact that only a select subgroup of malignant glioma patients are entered into clinical trials-individuals who are unlikely to be representative of the broader population. If the data supporting the benefit of full-dose radiotherapy plus carmustine, as evidenced by clinical trials, are derived from a highly select population, then conventional wisdom in support of such therapy may fairly be called into question.
Selection Bias in External-Beam Radiotherapy Plus Chemotherapy Clinical Trials--Researchers from the University of Western Ontario investigated selection bias in clinical trials of anaplastic glioma . They collaborated with our group at Duke in a prospective randomized clinical trial comparing AZQ to carmustine following surgery and radiotherapy for malignant gliomas [6,11]. Because of the Ontario provincial cancer care network, it was possible to assess the percentage of patients in a catchment area of approximately 1.1 million persons who ultimately entered the trial. It was also possible to evaluate how the loss of patients from the study biased the survival predictions.
Of 217 initial patients with a clinical and radiographic diagnosis of malignant glioma, 20 were too old and disabled to undergo a biopsy to establish a tissue diagnosis. Of the remaining 197 adults who had a biopsy-proven supra-tentorial malignant glioma, the investigators studied how many remained eligible for the study.
The eligibility criteria were a Karnofsky performance score 50% or greater, the absence of other medical conditions precluding chemotherapy, and signed informed consent. At diagnosis, 100% of the 197 patients met these eligibility criteria. Three weeks later, at the start of radiotherapy, 68% were still eligible. Six weeks later, at the end of radiotherapy, 47% were eligible. Eight weeks later, when offered randomization between the two drugs, 40% of the patients remained eligible. Seventy percent of these remaining patients agreed to be randomized (28% of the initial 197 patients).
The major reason patients became ineligible for the study were a decrease in KPS, followed by significant medical problems precluding participation and irregularities in the administration of external-beam radiation that violated the protocol. Study patients lived significantly longer than nonstudy patients (60 vs 25 weeks, P = .0001). One can imagine that the results might have been even more dramatic if an age cut off, ie more than 70 years, had been included as a criterion for eligibility. When one considers the fact that patients with progressive tumor during radiotherapy are excluded from randomized studies of post-radiotherapy chemotherapy, case selection is seen to be even more profound.
Selection Bias in Brachytherapy Clinical Trials--In a corollary study, the Western Ontario group collaborated with Gutin of the University of California at San Francisco and Leibel of Memorial Sloan-Kettering Cancer Center . They took records of 101 malignant glioma patients treated with conventional fractionated external-beam radiotherapy and adjuvant chemotherapy and asked two experienced surgeons and a radiotherapist to designate each patient as either eligible or ineligible for adjuvant brachytherapy. None of the patients had received such therapy. Overall, 32% of the patients were deemed eligible for brachytherapy. In a similar review conducted by the Radiation Therapy Oncology Group (RTOG) of their experience, 25% of patients were eligible for adjuvant brachytherapy . Eligible patients lived longer than ineligible ones (16.6 vs 9.3 months), were younger, and had larger initial surgical resections and better function. These findings suggest that better outcome following adjuvant brachytherapy for malignant glioma is strongly influenced by patient selection.
Recently, Curran et al published a recursive partitioning analysis of prognostic factors uncovered in three RTOG malignant glioma trials involving 1,578 patients . Patients were grouped according to a variety of prognostic factors.
Median survival time ranged from a high of 58.6 months for the 9% of patients less than 50 years of age who had anaplastic astrocytoma and normal mental status to a low of 4.6 months for the 17% of patients 50 years of age or older who had a KPS less than 70% and normal mental status and who received 54.4 Gy or less of external-beam radiation therapy. Since the RTOG trials were confined to patients 70 years of age or younger, it is likely that the study overrepresented the patient cohort with better prognostic factors.
Thus, inclusion of patients with poor prognostic factors in a study may obscure the potential benefit of aggressive therapy in patients with the more favorable prognosis. Conversely, inclusion of patients with good prognostic factors may falsely promote the usefulness of aggressive therapy.
The effects of case selection are summarized in Figure 1. This analysis of data from the published literature clearly shows that patients ultimately treated in clinical trials at academic medical centers constitute a small minority of malignant glioma patients. We are subjecting a large number of patients to forms of therapy that have the potential to benefit very few. As the incidence of brain tumors among the elderly increases, this problem will become more apparent .