Dr. Halperin evaluates progress in the treatment of adults with malignant glioma, with emphasis on older patients and those with a poor prognosis. He provides a good review of past clinical trials in this area, including, among others, trials conducted by the Brain Tumor Study Group (subsequently called the Brain Tumor Cooperative Group [BTCG]), with which we have been involved.
The evidence summarized by the author shows why, since the 1970s, the BTCG has considered "standard" therapy for adult malignant glioma to be definitive surgical resection followed by external-beam radiotherapy plus adjuvant carmustine(Drug information on carmustine) (BCNU). Other agents (eg, procarbazine(Drug information on procarbazine) [Matulane]) have shown similar results, but no chemotherapy regimen tested in the BTCG randomized trials has resulted in significantly longer survival than is seen with carmustine. With this therapy, median survival following initial surgery has been approximately 1 year for this patient population. The failure to prolong survival much beyond this point has been disappointing, and has led to explorations of other approaches, as outlined by Halperin. One of these is interstitial radiotherapy boost, which we discuss further below.
Halperin's main thesis is that more modest treatment regimens are indicated for older patients with a poor prognosis. There are four important aspects to his discussion. We will examine each of these in turn.
How Generalizable Are Randomized Trial Results?
First, Halperin argues that results from randomized trials are valid only for the "highly selected subgroup" of patients who enter them. As illustrated in Figure 1 of his review, not all patients with a presumptive diagnosis of malignant glioma will end up being eligible for a specific trial. Certainly, one has to look at the eligibility criteria of a trial when deciding to what patient population the results can be generalized. However, any randomized trial addresses a treatment decision in a defined setting, and the fact that there are patients to whom that question is irrelevant does not in any way diminish the usefulness of the results.
In Halperin's example comparing two possible options for post-radiotherapy treatment, the question is of no relevance for patients who deteriorate (or die) before or during radiotherapy. The results of the trial provide guidelines for post-radiotherapy treatment, which is the problem to which they were addressed. In the face of uncertainty, randomized trials offer evidence as to which of two (or more) alternative approaches is preferable; we don't need to answer such questions for patients who never reach that decision point.
Because an important fraction of patients with malignant glioma, espec-ially glioblastoma multiforme, do deteriorate during the time of conventional radiotherapy, we need to find better treatments to be administered early in the disease course. Early therapy has formed the basis of regimens tested by the BTCG, in which patients simultaneously begin chemotherapy and radiotherapy.
Does Therapy Benefit Only Good-Prognosis Patients?
Second, Halperin argues that only those patients with a favorable prognosis will be affected by therapy. Is evidence for this conjecture convincing? The fact that a certain characteristic influences a patient's prognosis in the absence of therapy does not necessarily mean that it will modify the effect of the therapy. For example, a treatment that increases survival time by 50% could extend expected survival from 12 to 18 months for a better-prognosis patient and from 6 to 9 months for a worse-prognosis patient. Halperin may properly argue that the benefit for the latter patient is not worth the cost (in toxicity or resources), but that is a separate issue; it does not mean that there is no benefit.
Looking at BTCG data on the efficacy of adjuvant chemotherapy , namely, the survival comparison in trial 7201 of radiotherapy plus carmustine vs radiotherapy alone, and the similar comparison in trial 7501 of radiotherapy plus carmustine vs radiotherapy plus methylprednisolone(Drug information on methylprednisolone), we found a consistent, but modest, benefit of carmustine in increasing "long-term" (eg, 18-month) survival.
If we apply regression models to these data, we find no statistically significant interactions of treatment with the major prognostic variables: age, performance status at randomization, histopathologic category, or symptom duration. This indicates that the treatment effect did not vary according to such prognostic factors. If we tabulate percentages of 18-month survivors for four categories of initial Karnofsky performance status (ie, 90 to 100, 70 to 80, 50 to 60, and equal to or less than 40), we find that the observed percentage of patients surviving equal to or more than18 months is greater for the radiotherapy plus carmustine group in all four categories (ie, for both good- and poor-prognosis patients). The same holds true for two categories of symptom duration (more than 6 months and equal to or less than 6 months) and three of the four categories of age (equal to or less than 44 years, 45 to 54 years, 55 to 64 years, and 65+ years); only in the oldest age category do the data fail to show a greater survival for the chemotherapy group.
The problem with looking at many subsets (with relatively small numbers of patients) is that the results may be due to the play of chance. In the absence of a significant interaction, we conclude that our results apply across our patient population. Because BTCG trials accepted patients with Karnofsky performance status as low as 40 or 50, the results apply to a wider range of patients than do studies excluding patients with scores below 70.
In the recursive partitioning analysis of data from the Radiation
Therapy Oncology Group , cited by Halperin, treatment-related
variables (extent of surgery and radiotherapy dose) entered the
regression tree only for glioblastoma multiforme patients over
age 50 with good performance status. Contrary to Halperin's statement,
this is not a group with a particularly favorable prognosis; two
of the three factors indicate a poorer prognosis. Furthermore,
this type of analysis can show different patterns in related subsets
that some may consider more reflective of chance than of
Is Short-Course Therapy Adequate?
Third, Halperin cites pilot data in support of his thesis that short-course therapy (usually radiotherapy alone) is as efficacious as longer-term chemoradiotherapy for poor-prognosis patients. These are uncontrolled studies, and thus, at best, are only suggestive. To support this contention, one would need properly controlled trials showing a lack of benefit of "standard" therapy among such patients.
Are Survival Gains Worth the Costs?
Fourth, while we find the above aspects of Halperin's argument unconvincing, there remains his main theme that therapies to date have not produced large gains in survival, and that for elderly patients or those with a short life expectancy, such modest gains in survival may not be worth the costs in terms of toxicity, quality of life, or resources.
Many would agree with this assessment. Indeed, it motivates the search for new treatment approaches. Preliminary results from our own BTCG trial 8701 , presented at the annual meeting of the American Society for Clinical Oncology in May 1994, showed a further modest gain in survival from using interstitial radiotherapy boost (temporary iodine-125 implants), in addition to conventional radiotherapy plus carmustine. The difference in survival curves was statistically significant, with a gain in median survival of 2.8 months. What the results will be when the data mature during the upcoming year remains to be seen.
In any case, the need for more efficacious therapies for malignant glioma still exists. Until these are found, each patient, in conjunction with his or her family members and physicians, must make the decision as to whether the gains from more aggressive therapy are worth the costs. In medicine we are constantly confronted by such decisions, and randomized trials continue to provide valid information to assist in the process.