Efficacy of neuroradiological imaging, neurological examination, and symptom status in follow-up assessment of patients with high-grade gliomas

J Neurosurg 93:201 207, 2000 Efficacy of neuroradiological imaging, neurological examination, and symptom status in follow-up assessment of patients with high-grade gliomas EVANTHIA GALANIS, M.D., JAN C. BUCKNER, M.D., PAUL NOVOTNY, M.S., ROSCOE F. MORTON, M.D., WILLIAM L. MCGINNIS, M.D., ROBERT DINAPOLI, M.D., PAULA SCHOMBERG, M.D., AND JUDITH R. O FALLON, PH.D. Departments of Oncology, Neurology, and Health Sciences Research, Mayo Clinic, Rochester, Minnesota; and the North Central Cancer Treatment Group, Rochester, Minnesota Object. It is standard practice for the oncological follow-up of patients with brain tumors (especially in the setting of clinical trials) to include neurological examination and neuroradiological studies such as computerized tomography (CT) or magnetic resonance (MR) imaging in addition to evaluation of the patients symptomatology and performance score. The validity of this practice and its impact on the welfare of patients with high-grade gliomas has not been adequately assessed. The purpose of this study is to provide such an assessment. Methods. The authors studied 231 similarly treated patients who were participating in three prospective North Central Cancer Treatment Group or Mayo Clinic trials who developed progressive disease during follow up. According to the protocol, the symptom status, performance score, results of neurological examination, and CT or MR status were recorded prospectively in each patient at each evaluation (every 6 8 weeks). At progression, 177 (77%) of 231 patients experienced worsening of their baseline symptoms or they developed new ones. In the remaining 54 asymptomatic patients (23%), neuroradiological imaging revealed the progression. Asymptomatic progression was more likely to be detected on MR imaging compared with CT studies (p 0.01). In no asymptomatic patient was progression detected on neurological examination alone. The median survival time after tumor recurrence was 13.3 weeks in symptomatic patients compared with 41.7 weeks in the asymptomatic group (p 0.0001). Asymptomatic patients were more aggressively treated, with surgery (p 0.0001) and second-line chemotherapy (p 0.0002). Multivariate analysis of survival time following first progression by using both classification and regression trees and Cox models showed that treatment at recurrence was the most important prognostic variable. Conclusions. Symptoms are the most frequent indicators of progression in patients with high-grade gliomas (77%). All asymptomatic progressions were detected on neuroradiological studies; MR imaging was more likely than CT scanning to reveal asymptomatic recurrences. Survival after disease progression was significantly longer in asymptomatic patients and could be related both to treatment following progression and to other favorable prognostic factors such as performance score. KEY WORDS glioma magnetic resonance imaging computerized tomography tumor recurrence clinical deterioration H ISTORICALLY, surveillance studies conducted to detect recurrent disease are a standard part of clinical trials and are frequently incorporated into the routine oncological follow-up review of patients with cancer. Recently, cumulative evidence gathered on various malignancies such as breast cancer 5,7 and melanoma 18 has indicated that early detection of asymptomatic tumor recurrences does not have an impact on patient survival, and subsequent follow-up guidelines on different tumor types have been modified to reduce the frequency of imaging and laboratory testing in follow up. 9 Abbreviations used in this paper: CART = classification and regression trees; CT = computerized tomography; ECOG = Eastern Cooperative Oncology Group; MR = magnetic resonance; NCCTG = North Central Cancer Treatment Group. It is common practice in the oncological follow up of patients with brain tumors (especially in the setting of clinical trials) to include neurological examinations and neuroradiological studies such as CT or MR imaging in addition to the evaluation of symptomatology and performance score by the patient s primary oncologist. The validity of this practice and its impact on the welfare of patients with primary brain tumors has only been partially examined. Torres, et al., 15 examined the impact of surveillance scanning in 23 children with recurrent medulloblastoma. They concluded that in this population, surveillance scanning is of little clinical value because it detected only a minority of tumor recurrences, and no patient who suffered a recurrence survived. Chuah, et al. (unpublished data) and Fulton (unpublished data) reported data on 97 consecutive patients who were treated with chemotherapy 201

E. Galanis, et al. TABLE 1 Characteristics at clinical trial entry in 268 patients with high-grade gliomas* Statistical Analysis An analysis dataset was constructed by merging selected pathological and clinical variables from the NCCTG glioma pathology databank with clinical files from the three most recently completed primary high-grade glioma trials conducted by the Mayo Clinic and/or by the NCCTG. 2,13,14 These included: 1) key dates date of onset of symptoms, date of registration and randomization, date of tumor progression, and date of death; 2) baseline patient characteristics age at study entry, sex, race, baseline performance score, and baseline Mini-Mental State Examination; 3 3) tumor characteristics grade and cell type of the primary tumor as described in the neuropath- Asymp- Excluded Evaluated tomatic Symptomatic Characteristic (37 patients) (231 patients) p Value (54 patients) (177 patients) p Value age (yrs) 0.58 0.0001 mean 55 54 47 56 median 55 56 46 58 sex (%) 0.61 0.75 male 65 59 61 58 female 35 41 39 42 baseline PS (%) 0.99 0.0019 0 35 37 54 32 1 49 45 37 47 2 13 15 7 17 3 3 3 2 4 primary cell type (%) 0.24 0.75 astrocytoma 94 94 91 94 oligoastrocytoma 6 5 7 6 NA 0 1 2 0 primary grade (%) 0.63 0.14 3 19 15 22 13 4 81 84 78 86 NA 1 0 1 extent of primary resection (%) 0.55 0.31 biopsy 30 24 17 27 subtotal resection 51 49 52 47 gross-total resection 19 27 31 26 * NA = not available; PS = performance score. Comparison of excluded patients with evaluated patients. Comparison of asymptomatic with symptomatic patients. with or without surgery for their first recurrence of supratentorial malignant glioma and subsequently followed with neuroradiological imaging at different time intervals. The fact that the difference in time to progression and survival favored the group that underwent more frequent imaging was attributed to lead-time bias (that is, artificial lengthening of survival as a result of screening) and these authors concluded that there is no evidence that less frequent imaging is detrimental. The purpose of this study was to evaluate the effectiveness of follow-up neuroradiological imaging, neurological examination, and patients symptom status in detecting progression in patients with high-grade gliomas (Grades 3 and 4). Patient Population Clinical Material and Methods We studied all 395 patients participating in the three most recently completed primary high-grade glioma trials conducted by the Mayo Clinic and by the NCCTG. 2,13,14 Central pathological review confirmed the histological diagnosis in all patients. Progressive disease had developed in 268 of these patients on follow-up review. All patients had received adjuvant radiation therapy and nitrosoureabased chemotherapy. According to the protocol, symptom status, neurological examination, and CT/MR status were recorded prospectively at each evaluation (every 6 8 weeks). Also according to the protocol, tumor progression on neuroradiological imaging was defined as a greater than 25% increase in tumor size with stable or increased doses of corticosteroid medications in patients with measurable disease, or an increase in tumor size documented by two independent observers in patients for whom adequate follow-up data were available and who were receiving stable or increased doses of corticosteroid medications. In addition, patients with neurological deterioration on consecutive follow-up examinations and who were receiving stable or increased doses of corticosteroids were also deemed to have tumor progression regardless of its appearance on neuroradiological imaging. Additional information collected from the NCCTG records at the time of this study included location and laterality of recurrent tumor, steroid dose at recurrence, neuroradiological mode used to establish recurrent disease (CT or MR study), and treatment at recurrence. 202

Evaluation of follow-up modalities in patients with gliomas ological review, and the ability to evaluate it on neuroradiological imaging (measurable or nonmeasurable disease); 4) treatment characteristics extent of primary resection, use of corticosteroid medications at study entry, treatment site (that is, NCCTG member), treatment regimen, and total dose of radiation therapy and first and last dates of its administration; 5) serial evaluation data including dates of evaluation, and values of the following variables at each evaluation: symptom status, neurological examination status, MR/CT status, objective status, ECOG 17 performance scale score, and neurological status; and 6) outcome variables time to disease progression and overall survival. Because some of the key clinical variables were defined differently among the various protocols, it was necessary to review these variables and reformat them to make them as consistent as possible across all three studies. For this analysis, patients were defined as symptomatic when they had new or worsening symptoms and as asymptomatic when there was no change in their baseline symptomatology at the time of progression. For patient subsets defined by key variables, Fisher s exact tests and Wilcoxon s tests were used to compare the distributions of various baseline characteristics. Time-toevent distributions were estimated using Kaplan Meier curves 8 and tested for equality with log rank tests. 12 Multivariate analyses in which Cox 4 and CART models 1,10 were used were performed to identify prognostic factors associated with both recurrence and survival times within the subsets. Lead- and Length-Time Bias The terms lead- and length-time bias are extensively used in this study and are defined as follows: lead-time bias is an estimate of survival time that has been artificially inflated because screening has simply lengthened the time between diagnosis and death by enabling an earlier diagnosis but without actually having an impact on cancer-specific mortality. Length-time bias occurs because the proportion of slow-growing lesions diagnosed during frequent screening sessions is greater than the proportion of those diagnosed during routine medical care. The inclusion of a greater number of slow-growing tumors makes it seem that the screening and early treatment programs are more effective. The longer the preclinical phase, the stronger the potential for this type of bias. TABLE 2 Status on neuroradiological studies and neurological examination at the time of tumor progression in 231 patients with glioma Asymptomatic Symptomatic Wilcoxon s Status (54 patients) (177 patients) p Value CT/MR (%) 0.12 regression 0 0 stable 0 7 progression 100 93 neurological (%) 0.0001 improved 5 0 stable 72 11 worse 23 89 Results Of the 268 patients, in 231 (86%) there was sufficient follow-up data at progression, and these patients were included in the analysis. The 37 excluded patients were not significantly different from the 231 patients included in the study in terms of clinical characteristics (Table 1), time to progression, or survival time. Most (94%) of the 231 patients had pure astrocytic tumors, whereas 6% also had an oligodendroglial component. Most (84%) of the tumors were Grade 4, whereas 15% were Grade 3 (Table 1). Of the 231 patients, 177 (77%) had new or worsening symptoms at the time of progression (symptomatic), whereas 54 (23%) were asymptomatic, that is, there was no change in their symptom profile during the study. In all asymptomatic patients, progression was detected on neuroradiological imaging (CT or MR study); none was detected on neurological examination alone. In 21 of the 177 symptomatic patients, tumor progression was established by clinical evaluation alone in the setting of rapid clinical deterioration that could not be attributed to other reversible causes. Seven of these 21 patients did not undergo a neurological examination at the time of progression. In the remaining 14 patients, results of the neurological examination were worse at progression. Table 1 summarizes demographic characteristics for all patients as well as for the symptomatic and asymptomatic subsets. The patients who were symptomatic at recurrence were older (p 0.0001) and had a worse performance status at study entry (p = 0.0019). Table 2 displays the CT or MR study results and neurological examination status at the time of progression for the symptomatic compared with the asymptomatic patients. In the symptomatic patients, results of neurological examination were worse at the time of progression in 89% and remained stable in 11%. In asymptomatic patients, the results of neurological examination were worse in 23%, stable in 72%, and better in 5%. All asymptomatic patients with worsening results on neurological examination also had tumor progression that was detected on neuroradiological imaging. Table 3 provides additional information regarding the characteristics of the symptomatic compared with asymptomatic patients at the time of progression. Asymptomatic progression was detected more often on MR imaging than on CT scanning (p 0.01). Of the 119 patients in the study in whom CT scanning was used for follow-up neuroradiological imaging, 23 (19.3%) were asymptomatic at progression. In contrast, of the 91 patients who underwent MR imaging as part of their follow up, 31 (34.1%) were asymptomatic at progression. In addition, patients who were symptomatic at progression were more likely to be treated with dexamethasone and to have lesions in the parietal lobe (p = 0.03 and p = 0.09, respectively). Asymptomatic patients were more likely to have either surgery (p 0.01) or chemotherapy (p 0.01) after recurrence, whereas symptomatic patients were more likely to receive no treatment at recurrence (p 0.01). The few patients who were treated with stereotactic radiosurgery at recurrence were included in the radiation therapy category. Asymptomatic patients who received treatment at recurrence lived approximately 138 days longer compared with asymptomatic patients who did not receive treatment (p = 0.03). 203

E. Galanis, et al. TABLE 3 Percentage of patients with specified characteristics at tumor progression* CART model (Fig. 3). This model, in addition to identifying major prognostic variables, permitted us to describe the relationships among them. By using the same 22 independent variables as described in Table 5, with our CART model we identified seven prognostic groups (designated by boxes in Fig. 3) that were defined in terms of only seven variables: no treatment given at progression, ECOG performance score at progression, age at trial entry, dexamethasone dose at progression, CT scan obtained at progression, MR image obtained at progression, and symptoms at progression. Nearly all of these were the same patient-condition and medical-management variables identified with the Cox modeling process. The group with the most favorable outcome was young patients (age 31 years) with good performance scores ( 3) who received treatment. The group with the most unfavorable outcome consisted of patients receiving higher dexamethasone doses ( 17.5 mg) who received no treatment at progression. The CART model provides important additional information, however: age at trial entry and ECOG performance score at progression are prognostic factors for patients who received treatment at progression, whereas dexamethasone dose at progression, CT/MR studies at progression, and symptoms at progression are prognostic factors for patients who did not receive treatment at progression. Asymp- Symp- Overall tomatic tomatic Variable (231 pts) (54 pts) (177 pts) p Value neuroimaging method (%) 0.01 CT 52 43 54 MR 39 57 34 none 9 0 12 PS at progression (% [229 pts]) 0.01 0 13 38 6 1 28 43 23 2 22 17 24 3 23 2 29 4 14 0 18 location (%) 0.09 frontal 28 32 28 temporal 27 33 25 parietal 34 24 37 occipital 4 0 5 midline 5 7 4 spine 2 4 1 side (%) 0.28 not applicable 6 9 5 rt 58 50 60 lt 36 41 35 dexamethasone (%) 0.03 no 44 57 40 yes 56 43 60 dexamethasone dose (mg [124 pts]) 0.79 mean 10.1 13.0 9.5 median 8.0 8.0 8.0 treatment at recurrence (%) surgery 15 35 9 0.01 chemotherapy 27 48 21 0.01 radiation therapy 4 9 3 0.06 none 56 22 67 0.01 * Pts = patients. Figures 1 and 2 show time to progression and time to death after progression, respectively, and Table 4 summarizes the corresponding medians and log rank probability values. The median time to progression from study entry was longer by 36 days in asymptomatic patients (p = 0.01). The median survival time after progression in the asymptomatic group was also significantly longer, at 292 compared with 93 days (a difference of 199 days). To investigate the observed survival differences after progression in the 231 patients in whom sufficient information was available at progression, multivariate analyses were performed using both Cox and CART models with the independent variables listed in Table 5. With the best Cox model we identified eight variables as significantly (p 0.05) associated with survival after progression; six were observable at progression (ECOG performance score, symptom status, CT scan, MR image, dexamethasone dose, and no treatment at progression) and two were obtained at primary diagnosis (ECOG performance score at trial entry and primary tumor histological grade). It is notable that both symptom status at progression and treatment at progression appear to have independent prognostic significance when a Cox model of multivariate analyses is used. We expanded our multivariate analysis by also using a FIG. 1. Graph showing Kaplan Meier curves estimating time to progression for the patients in the study. Discussion Our study brings up several important considerations in the follow-up care of patients with high-grade gliomas. Of the 231 patients in the study, 54 (23%) had no change in their symptoms when tumor progression was detected on neuroradiological imaging, either CT or MR studies. In 23% of these patients, results of neurological examination were also worse, whereas in 72% it was stable and in 5% it improved (Table 2). Thus, neurological examination is less sensitive than CT or MR studies in detecting tumor progression in the absence of new symptoms. In contrast, the majority of symptomatic patients (89%) also showed worsening on neurological examinations conducted at the time of progression. In the subgroup of 14 symptomatic patients in whom progression was established on clinical 204

Evaluation of follow-up modalities in patients with gliomas TABLE 4 Median time to progression and survival times for symptomatic and asymptomatic patients with glioma Median Time Median Survival Median Overall to Progres- Time After Pro- Survival Time Factor sion (days) gression (days) (days) asymptomatic 183 292 540 symptomatic 147 93 290 log rank p value 0.01 0.0001 0.0001 FIG. 2. Graph showing Kaplan Meier curves estimating survival from progression (PROG) for the study participants. evaluation only, results of neurological examination were always worse when it was performed. Both CT and MR studies were used for scheduled follow up of the patients. As may be expected, MR imaging is more likely to enable detection of asymptomatic recurrences compared with CT scanning (p 0.01). The percentage of patients who were asymptomatic at progression was significantly higher in the MR than in the CT group. Of the 119 study patients in whom CT was used for follow-up neuroradiological imaging, 23 (19.3%) were asymptomatic at progression. In contrast, of the 91 patients in whom MR imaging was part of their follow up, 31 (34.1%) were asymptomatic at progression. Given our additional data that showed improved survival for asymptomatic patients in part as a result of more aggressive management, MR imaging should be considered the modality of choice when the patient s overall condition as well as the location of the tumor would allow intervention in the event that early progression is detected. Our study also demonstrates that routine surveillance neuroradiological imaging (in the range of 2 3 months) is appropriate for patients with high-grade gliomas; the evidence from patients with breast cancer and melanoma showing that frequent follow-up testing is unnecessary 5,7,16 cannot be extrapolated to patients with high-grade gliomas. In a previous study in which the effectiveness of neuroradiological imaging was used in the follow up of children with medulloblastoma, the conclusions were based on only 23 patients with recurrent disease. 17 Even so, patients in that study who had asymptomatic disease recurrence lived longer than patients with symptomatic disease recurrence. Although in that study the patients received follow-up care at similar intervals, the authors attributed the survival differences to lead- and length-time biases and suggested that surveillance scanning was of little clinical value, an opinion that generated several objections. 6,11 The study by Chuah, et al., which was reported only in abstract form at the Sixth Canadian Neuro- Oncology Meeting (unpublished data), was hampered by lead-time biases because follow-up practices in their institution changed over time. The difference in survival between symptomatic and asymptomatic patients in our study most likely is not due to lead- or length-time bias because the follow-up interval was predetermined and it was the same for all patients in the study from the time of entry. In addition, we addressed length-time bias by including time to progression in our multivariate analysis survival models, where it proved to have no independent prognostic value. In our study we demonstrate that symptom status at progression defines separate subpopulations of patients with high-grade glioma with different survival times. Once again, we emphasize the need for very careful design of Phase II trials in recurrent high-grade glioma because the treatment under investigation may appear more or less effective depending on the characteristics of the patients tested. As we have demonstrated using two different multivariate analysis models, additional factors such as age, tumor grade, and performance score play an important role in determining survival after tumor progression and should be TABLE 5 Probability values for independent variables in Cox models of survival after disease progression in the 231 patients included in the study* Best Model Independent Variable p Value Risk Ratio age at trial entry (yrs) NS ND sex NS ND ECOG PS at trial entry 0.0352 0.813 primary tumor surgery biopsy only NS ND subtotal resection NS ND primary tumor histological grade 0.0085 1.693 primary tumor histological cell type NS ND time to progression NS ND ECOG PS at progression 0.0082 1.233 symptoms at progression 0.0257 1.541 CT scan obtained at progression 0.0003 0.386 MR image obtained at progression 0.0044 0.452 tumor location frontal lobe NS ND temporal lobe NS ND parietal lobe NS ND rt hemisphere NS ND dexamethasone administered at progression NS ND dexamethasone dose at progression 0.0013 1.022 surgery performed at progression NS ND therapy administered at progression chemotherapy NS ND radiation therapy NS ND none 0.0035 1.664 * ND = not done; NS = not significant. Values greater than 1 denote factors associated with a higher risk of death; values less than 1 denote factors associated with a lower risk of death. 205

E. Galanis, et al. FIG. 3. Depiction of the CART model showing factors associated with survival time from progression. Squares identify the seven prognostic groups. Within each circle or square, the top number gives the relative death rate compared with the total group. Values greater than 1 indicate shorter survival; values less than 1 indicate longer survival. The bottom number is a ratio showing the number of patients who died (numerator) to the total number of patients in the group (denominator). Patients ages were measured in years. Dex = dexamethasone; PS = performance score. taken into account when trials in patients with recurrent high-grade glioma are designed; however, treatment at recurrence, despite the limited modalities available, also appears to be an independent factor affecting survival. It is notable that the survival of asymptomatic patients who received treatment at recurrence is significantly longer compared with asymptomatic patients who did not receive treatment (p = 0.03), even when adjusted for other factors associated with survival. Because of our study design, we cannot address the impact of follow-up intervals different from those used in our protocol. Although the most valid way to study possible survival differences associated with different followup strategies would be through a large prospective randomized trial, it is unlikely that such a trial will ever be performed for high-grade gliomas. It would require not only a large number of patients with high-grade glioma but also standardization of treatment approaches at recurrence, a task not easy to achieve and hampered by ethical limitations. In addition, our study does not address the impact of detection of asymptomatic progression in patients with high-grade glioma after their first recurrence or in patients with low-grade gliomas. We are in the process of evaluating separately these latter populations of patients. What about the impact of frequent neuroradiological imaging on quality of life? Theoretically, frequent imaging could have both a positive effect (by reassuring the patient) and a negative effect (by triggering anxiety). Data that could address this question are not available for patients with glioma. In patients with breast cancer, available data indicate that follow-up testing did not have an impact (either positive or negative) on the patients quality of life. Conclusions Clinical symptoms are the most frequent indicator of progression in patients with high-grade glioma. Neuroradiological imaging aided detection of asymptomatic progression in 23% of the patients; MR imaging was more sensitive than CT scanning and neurological examination was less sensitive than neuroradiological imaging. The longer survival after progression observed in symptomatically stable patients seemed to be related both to treatment following progression and to other favorable prognostic factors such as initial performance score, performance score at progression, tumor grade, and age of the patient. Acknowledgments We acknowledge Debra Kvittem from NCCTG Data Monitoring Unit, Jill Burton, and other NCCTG Data Managers for help with data collection and study organization, and Gail Prechel for her help in the preparation of the manuscript. References 1. Breiman L, Friedman JH, Olshen RA, et al: Classification and Regression Trees. Belmont, CA: Wadsworth, 1984 2. Buckner JC, Schomberg PJ, McGinnis WL, et al: Phase III study of radiation therapy (RT) plus BCNU with or without recombinant interferon alpha (IFN) in high-grade glioma: an NCCTG study. Proc Am Soc Clin Oncol 16:385a, 1997 (Abstract) 3. Cockrell JR, Folstein MF: Mini-Mental State Examination (MMSE). Psychopharmacol Bull 24:689 692, 1988 4. Cox DR: Regression models and life tables. J R Stat Soc B 34: 187 202, 1972 5. Del Turco MR, Palli D, Cariddi A, et al: Intensive diagnostic follow-up after treatment of primary breast cancer. A randomized trial. JAMA 271:1593 1597, 1994 6. Friedman HS, Kun LE: More on surveillance of children with medulloblastoma. N Engl J Med 332:191, 1995 (Letter) 7. The GIVIO Investigators: Impact of follow-up testing on survival and health-related quality of life in breast cancer pa- 206

Evaluation of follow-up modalities in patients with gliomas tients. A multicenter randomized controlled trial. JAMA 271: 1587 1592, 1994 8. Kaplan EL, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457 481, 1958 9. Kattlove H, Liberati A, Keeler E, et al: Benefits and costs of screening and treatment for early breast cancer. Development of a basic benefit package. JAMA 273:142 148, 1995 10. LeBlanc M, Crowley J: Relative risk trees for censored survival data. Biometrics 48:411 425, 1992 11. Lindsey KL: Surveillance scanning of children with medulloblastoma. N Engl J Med 331:483, 1994 (Letter) 12. Mantel N: Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chemother Rep 50:163 170, 1966 13. Rajkumar SV, Buckner JC, Schomberg PJ, et al: Phase I evaluation of preirradiation chemotherapy with carmustine and cisplatin and accelerated radiation therapy in patients with highgrade gliomas. Neurosurgery 44:67 73, 1999 14. Rajkumar SV, Buckner JC, Schomberg PJ, et al: Phase I evaluation of radiation combined with recombinant interferon alpha- 2a and BCNU for patients with high-grade glioma. Int J Radiat Oncol Biol Phys 40:297 302, 1998 15. Torres CF, Rebsamen S, Silber JH, et al: Surveillance scanning of children with medulloblastoma. N Engl J Med 330: 892 895, 1994 16. Weiss M, Loprinzi CL, Creagan ET, et al: Utility of follow-up tests for detecting recurrent disease in patients with malignant melanomas. JAMA 274:1703 1705, 1995 17. Zubrod CG, Schneiderman M, Frei E, et al: Appraisal of methods for the study of chemotherapy of cancer in man: comparative therapeutic trial of nitrogen mustard and triethylene thiophosphoramide. J Chron Dis 11:7 33, 1960 Manuscript received November 16, 1999. Accepted in final form April 27, 2000. This study was conducted as a collaborative trial of the NCCTG and Mayo Clinic and was supported in part by Public Health Service Grants CA-25224, CA-37404, CA-15083, and CA-35101, and the Linse-Bock Foundation. Address reprint requests to: Evanthia Galanis, M.D., Mayo Clinic, 200 First Street Southwest, Rochester, Minnesota 55905. email: galanis.evanthia@mayo.edu. 207