Int. J. Radiation Oncology Biol. Phys., Vol. 47, No. 2, pp.

291–298, 2000
Copyright © 2000 Elsevier Science Inc.
Printed in the USA. All rights reserved
0360-3016/00/$–see front matter

PII S0360-3016(99)00507-6

CLINICAL INVESTIGATION Brain

SINGLE DOSE RADIOSURGICAL TREATMENT OF RECURRENT

PREVIOUSLY IRRADIATED PRIMARY BRAIN TUMORS AND BRAIN
METASTASES: FINAL REPORT OF RTOG PROTOCOL 90-05

EDWARD SHAW, M.D.,* CHARLES SCOTT, PH.D.,† LUIS SOUHAMI, M.D.,‡ ROBERT DINAPOLI, M.D.,§
ROBERT KLINE, PH.D.,㛳 JAY LOEFFLER, M.D.,¶ AND NANCY FARNAN, B.S.†
*Department of Radiation Oncology, Wake Forest University School of Medicine, Winston Salem, NC; †Radiation Therapy Oncology
Group, Philadelphia, PA; ‡Department of Radiation Oncology, McGill University, Montreal, Quebec, Canada; §Department of
Neurology and 㛳Division of Radiation Oncology, Mayo Clinic, Rochester, MN; and 㛳 Joint Center for Radiation Therapy, Boston, MA

Purpose: To determine the maximum tolerated dose of single fraction radiosurgery in patients with recurrent
previously irradiated primary brain tumors and brain metastases.
Methods and Materials: Adults with cerebral or cerebellar solitary non-brainstem tumors < 40 mm in maximum
diameter were eligible. Initial radiosurgical doses were 18 Gy for tumors < 20 mm, 15 Gy for those 21–30 mm,
and 12 Gy for those 31– 40 mm in maximum diameter. Dose was prescribed to the 50 –90% isodose line. Doses
were escalated in 3 Gy increments providing the incidence of irreversible grade 3 (severe) or any grade 4 (life
threatening) or grade 5 (fatal) Radiation Therapy Oncology Group (RTOG) central nervous system (CNS)
toxicity (unacceptable CNS toxicity) was < 20% within 3 months of radiosurgery. Chronic CNS toxicity was also
assessed.
Results: Between 1990 –1994, 156 analyzable patients were entered, 36% of whom had recurrent primary brain
tumors (median prior dose 60 Gy) and 64% recurrent brain metastases (median prior dose 30 Gy). The
maximum tolerated doses were 24 Gy, 18 Gy, and 15 Gy for tumors < 20 mm, 21–30 mm, and 31– 40 mm in
maximum diameter, respectively. However, for tumors < 20 mm, investigators’ reluctance to escalate to 27 Gy,
rather than excessive toxicity, determined the maximum tolerated dose. In a multivariate analysis, maximum
tumor diameter was one variable associated with a significantly increased risk of grade 3, 4, or 5 neurotoxicity.
Tumors 21– 40 mm were 7.3 to 16 times more likely to develop grade 3–5 neurotoxicity compared to tumors
< 20 mm. Other variables significantly associated with grade 3–5 neurotoxicity were tumor dose and Karnofsky
Performance Status. The actuarial incidence of radionecrosis was 5%, 8%, 9%, and 11% at 6, 12, 18, and 24
months following radiosurgery, respectively. Forty-eight percent of patients developed tumor progression within
the radiosurgical target volume. A multivariate analysis revealed two variables that were significantly associated
with an increased risk of local progression, i.e. progression in the radiosurgical target volume. Patients with
primary brain tumors (versus brain metastases) had a 2.85 greater risk of local progression. Those treated on a
linear accelerator (versus the Gamma Knife) had a 2.84 greater risk of local progression. Of note, 61 % of
Gamma Knife treated patients had recurrent primary brain tumors compared to 30% of patients treated with
a linear accelerator.
Conclusions: The maximum tolerated doses of single fraction radiosurgery were defined for this population of
patients as 24 Gy, 18 Gy, and 15 Gy for tumors < 20 mm, 21–30 mm, and 31– 40 mm in maximum diameter.
Unacceptable CNS toxicity was more likely in patients with larger tumors, whereas local tumor control was most
dependent on the type of recurrent tumor and the treatment unit. © 2000 Elsevier Science Inc.

Stereotactic radiosurgery, Recurrent glioma, Recurrent brain metastases.

Reprint requests to: Edward Shaw, M.D., Department of Ra- icine. Institutions participating in this study included: McGill
diation Oncology, Wake Forest University School of Medicine, University (31 patients), Mayo Clinic (30 patients), Joint Center
Medical Center Boulevard, Winston-Salem, NC 27157-1030. for Radiation Therapy (14 patients), Western Ontario University
Tel: (336) 716-4647; Fax: (336) 716-5972; Email:eshaw@ (14 patients), Toronto-Bayview (11 patients), Johns Hopkins Uni-
wfubmc.edu versity (10 patients), University of Rochester (10 patients), Med-
Acknowledgments—The authors gratefully acknowledge the assis- ical College of Wisconsin (8 patients), University of California-
tance of the following individuals from the RTOG Operations San Francisco (7 patients), New York University (6 patients),
Office: Walter Curran, M.D., Stephanie Davies, R.T.T., Michael University of South Florida (6 patients), Washington University (6
Gillin, Ph.D., Elizabeth Martin, R.T.T., Linda Martin, R.T.T., patients), Monetfiore (5 patients), Albert Einstein (4 patients),
Elaine Pakuris, Michelle Szydlowski, R.T.T., and Lisa Worrells, Dartmouth (2 patients), Latter Day Saints Hospital (2 patients),
P.A. The authors also acknowledge the secretarial assistance of and Sacramento (2 patients).
Mrs. Pam Cregger from Wake Forest University School of Med- Accepted for publication 30 June 1999.

291
292 I. J. Radiation Oncology ● Biology ● Physics Volume 47, Number 2, 2000

INTRODUCTION Table 2. RTOG CNS toxicity criteria

In 1989, the Radiation Therapy Oncology Group (RTOG) Grade Definition

Brain Tumor Committee made the decision to initiate a
1 Mild neurologic symptoms; no medication required
stereotactic radiosurgery clinical trials program. The first 2 Moderate neurologic symptoms; outpatient medication
radiosurgery study, RTOG Protocol 90-05, was designed to required (e.g., steroids)
treat patients with recurrent previously irradiated primary 3 Severe neurologic symptoms; outpatient or inpatient
brain tumors and brain metastases with escalating doses of medication required
single fraction radiosurgery to establish the maximum tol- 4 Life-threatening neurologic symptoms (e.g.,
uncontrolled seizures, paralysis, or coma); includes
erated dose. The choice of this study as the first radiosurgery clinically or radiographically suspected
clinical trial was based on several factors. The numbers of radionecrosis and histologically proven
patients eligible for the study were anticipated to be high, radionecrosis at the time of an operation
since local failure is common following conventionally 5 Death
fractionated radiotherapy for primary brain tumors and
brain metastases. Few other therapeutic options exist for
these patients, the maximum tolerable dose of single frac- Inc., Atlanta, GA). Rigid head immobilization with a ste-
tion radiosurgery had yet to be defined in this setting, and reotactic head frame was required. Patients then had a CT or
the resultant data was anticipated to be utilized in future MRI scan for radiosurgical treatment planning. The target
RTOG radiosurgery studies. In 1995, the initial report of volume was defined as the contrast enhancing tumor volume
RTOG Protocol 90-05 was published (1). This article up- without a margin of surrounding brain tissue. The radiosur-
dates the second report of the study (2) and provides final gical dose was delivered in a single fraction. Dose was
results. determined by the maximum diameter of the tumor volume
in any plane of the treatment planning scan. Initial doses
MATERIALS AND METHODS were as follows: Arm 1, ⱕ 20 mm, 18 Gy; Arm 2, 21–30
Eligibility criteria and pretreatment evaluation mm, 15 Gy; and Arm 3, 31– 40 mm, 12 Gy. Dose was
To be eligible, patients ⱖ 18 years old must have re- prescribed to the 50 –90% isodose line, which was to en-
ceived partial or whole brain fractionated external beam compass the entire enhancing target volume. Dose escala-
radiotherapy ⱖ 3 months prior to study entry for either a tion was based on the incidence of acute CNS toxicity
primary brain tumor or brain metastases. CT or MRI scan defined by RTOG criteria (Table 2). Unacceptable toxicity
evidence of a cerebral or cerebellar solitary recurrence ⱕ 40 was considered to be irreversible grade 3 (severe), any grade
mm in maximum diameter was required. All patients had a 4 (life threatening), or grade 5 (fatal) RTOG CNS toxicity
Karnofsky Performance Status (KPS) of ⱖ 60%, a Neuro- occurring within 3 months of protocol radiosurgery. Ten to
logic Function Score (NFS) of 0 –3 (Table 1), and a life 15 patients were entered per treatment arm. The dose was
expectancy of ⱖ 3 months. Patients with multiple lesions escalated in 3 Gy increments providing there was not an
were allowed on study as long as there was one dominant excess of unacceptable toxicity. Central nervous system
lesion, and only that lesion would be treated with radiosur- toxicity occurring after 3 months, defined as chronic toxic-
gery. Ineligibility criteria included recurrent brainstem tu- ity, was also recorded.
mors and planned additional local or systemic therapy
within three months of protocol radiosurgery. Pretreatment Follow-up
evaluation consisted of a neurologic history and examina- A neurologic history and examination, assessment of
tion, a recording of corticosteroid medication (steroid) dos- KPS and NFS, and a CT or MRI scan with contrast was
age, and a CT or MRI scan of the brain with contrast. performed 2–3 months following protocol radiosurgery, and
at 3-month intervals in the first post-treatment year, and at
Protocol treatment 6-month intervals thereafter. All clinical and radiographic
Protocol radiosurgery could be administered with a linear data were reviewed jointly by a radiation oncologist (ES)
accelerator (linac) or Gamma Knife (Elekta Instruments, and a neurologist (RD). Local tumor progression was de-
fined as the radiographic appearance of a new or increasing
enhancing lesion within the radiosurgical target volume,
Table 1. Neurologic function score regardless of KPS or NFS. Adjacent progression was de-
fined as the radiographic appearance of a new or enhancing
Score Definition lesion within five millimeters of the radiosurgical target
volume. Progression in the brain but not within or adjacent
0 No neurologic symptoms
1 Minor neurologic symptoms to the radiosurgical target volume was termed non-adjacent
2 Moderate neurologic symptoms; fully active intracranial progression. Treatment failure also included the
3 Moderate to severe neurologic symptoms; development of systemic metastatic disease. Operation was
less than fully active encouraged when a clinical or radiographic diagnosis of
4 Severe neurologic symptoms; inactive
local tumor progression versus radionecrosis was made.
 Radiosurgery for recurrent brain tumors ● E. SHAW et al. 293

Operative biopsy or resection specimens were classified as Table 3. Accrual by treatment arm and dose
showing tumor, necrosis, or both tumor and necrosis. Any
Tumor diameter Treatment arm/dose/accrual
patient having an operation yielding pathologic evidence of
necrosis (with or without tumor) was considered to have a ⱕ 20 mm Arm 1—⬎ Arm 4—⬎ Arm 7
grade 4 CNS toxicity. Treatment toxicity was also assumed 18 Gy 21 Gy 24 Gy
when the KPS or NFS worsened with a stable or decreasing n ⫽ 12 n ⫽ 18 n ⫽ 10
21–30 mm Arm 2 —⬎ Arm 5 —⬎ Arm 8 —⬎ Arm II
contrast-enhancing lesion within the radiosurgical target 15 Gy 18 Gy 21 Gy 24 Gy
volume but increasing peritumoral edema and was graded n ⫽ 15 n ⫽ 15 n ⫽ 13 n ⫽ 12
according to RTOG CNS toxicity criteria. 31–40 mm Arm 3 —⬎ Arm 6—⬎ Arm 9
Physics and Quality Control Assessment. The following 12 Gy 15 Gy 18 Gy
information was recorded on each patient: maximum dose n ⫽ 21 n ⫽ 22 n ⫽ 18
(MD) in Gy within the target volume, prescription dose
(PD) in Gy, MD/PD ratio, prescription isodose volume
(PIV) and tumor volume (TV) in mm3, and the PIV/TV tion until death. The survival distribution was estimated
ratio. The MD/PD ratio, a measure of dose heterogeneity using the product-limit estimator (3). The time-adjusted
within the target volume, and the PIV/TV ratio, a measure incidence of necrosis was estimated using cumulative inci-
of dose conformity of the treated volume relative to the dence methodology (4).
target volume, have been previously defined (1). When Univariate analysis of effects of covariates on develop-
volume data were not provided, the PIV and TV were ment of grade 3 or worse acute and chronic CNS toxicity
estimated using volume formulae for a sphere or ellipse. All was performed using the Wilcoxon rank sum test (5). Co-
physical parameters of protocol radiosurgery, including the variates that had an associated significance level of ⱕ 0.25
above, were verified by a physicist (RK). were included in the multivariate model. A polychotomous
Compliance with protocol radiosurgery was determined regression model that assumes a multinormal distribution
by assessing coverage of the target volume by the prescrip- was fit to the toxicity data (6). Odds ratios were derived
tion isodose line on serial axial images from the treatment from the coefficients obtained through the polychotomous
planning scan. Coverage was scored as acceptable (no de- regression model (7).
viation from the protocol), minor deviation (marginal cov-
erage of the tumor volume), or major deviation (partial miss
RESULTS
of the tumor volume).
Patient characteristics
Statistical methods Between August 1990 and December 1994, 168 patients
The trial was designed to escalate the dose of single from 17 institutions were entered on the protocol. Accrual
fraction radiosurgery on a given treatment arm when the was rapid enough that ⬎ 10 –15 patients were entered on
incidence of acute CNS toxicity was considered acceptable. some of the treatment arms. Twelve patients were excluded,
Ten patients were initially entered on a given treatment arm. mainly because of a tumor size ⬎ 40 mm or the presence of
That arm was then closed for toxicity assessment. If no ⱖ 2 lesions that were felt to require radiosurgical treatment.
patient developed unacceptable acute toxicity as previously This left 156 analyzable cases. The accrual by treatment arm
defined, the dose for that tumor size was then escalated. is shown in Table 3. The type of recurrent tumor was a
This sample size assured, with 97% confidence, that the true primary brain tumor in 36% of patients and brain metastases
unacceptable acute toxicity incidence was ⬍ 30%. If one or in 64%. Half of the recurrent primary tumors were astrocy-
two of the first 10 analyzable patients developed unaccept- tomas. The remainder were a mixture of other glial neo-
able acute toxicity, that arm was reopened and an additional plasms. Two thirds of the recurrent metastases were adeno-
5 analyzable patients were accrued. If there were ⱕ 3 carcinoma of breast origin, or adenocarcinoma or squamous
unacceptable acute toxicities in the entire group of 15 ana- cell carcinoma of lung origin. Essentially all patients with
lyzable patients, the true incidence of unacceptable toxicity recurrent primary tumors had single lesions, whereas 22%
was ⬍ 30% with 87% confidence. If ⱖ 3 of 10 or ⱖ 4 of 15 of those with recurrent metastases had multiple lesions,
patients developed unacceptable acute toxicity, that arm although only the dominant one received protocol radiosur-
was considered too toxic, and there was no further dose gery. The median dose of previous fractionated radiation
escalation. therapy was 60 Gy in the case of recurrent primary tumors
Analyzable patients were defined as those who met all and 30 Gy for recurrent metastases. The median interval
eligibility criteria, received their assigned dose of protocol from the completion of previous radiotherapy to protocol
radiosurgery, and had adequate follow-up data submitted to radiosurgery was 17 months. At the time of study entry, the
assess acute CNS toxicity. The change in KPS or NFS was KPS was 60 –70% in 29% of patients and 80 –100% in 71%.
taken from the time of on-study to the status closest to a The NFS was 0 in 15% of patients, 1 in 47%, 2 in 22%, and
given point in time. For statistical analyses, target volume 3 in 17%. The median tumor volume was 8.2 cc for the
was dichotomized at the median volume in cubic centime- entire group but ranged between 3.6 cc for tumors ⱕ 20 mm
ters (cc). Survival was measured from the time of registra- in maximum diameter (Arms 1, 4, and 7) to 6.6 cc for those
294 I. J. Radiation Oncology ● Biology ● Physics Volume 47, Number 2, 2000

Table 4. Technical differences between linear accelerator and Table 6. Change in Karnofsky Performance Status and
gamma knife treated patients Neurologic Function Score following protocol radiosurgery

Linear Gamma Change versus KPS NFS

accelerator knife baseline (% of patients) (% of patients)

No. patients 125 31 ⱕ 3 mos ⬎ 3–12 mos ⱕ 3 mos ⬎ 3–12 mos

Lesion type (% patients) Better 15 7 16 12
Recurrent primary 30 61 Same 46 36 52 34
Recurrent metastasis 70 39 Worse 39 57 31 53
No. isocenters
Median 1 5
Range 1–4 1–1
PIV/TV Ratio (% patients) able acute and chronic CNS toxicity is shown in Table 5. In
⬍1 9 0 the patients with ⱕ 20 mm tumors, the radiosurgical dose
1 to ⬍ 2 51 77 was escalated from 18 Gy to 21 Gy and then to 24 Gy
ⱖ2 40 23
MD/PD Ratio (% patients)
without reaching an unacceptable incidence of acute CNS
1 to ⬍ 2 93 6 toxicity. However, dose was not escalated beyond 24 Gy
ⱖ2 7 94 because of investigators’ reluctance to give a single radio-
Protocol compliance (% patients) surgical dose of 27 Gy. Therefore, the maximum tolerable
No deviation 84 dose was considered to be 24 Gy, although it is unknown
Minor deviation 11
Major deviation 5
whether unacceptable toxicity would have occurred at 27
Gy. In the patients with 21–30 mm tumors, the radiosurgical
dose was escalated from 15 Gy to 18 Gy, then to 21 Gy, and
finally to 24 Gy. Although 24 Gy proved to be unacceptable
21–30 mm (Arms 2, 5, 8, and 11) to 17.9 ccs for those
in terms of acute CNS toxicity, the maximum tolerable dose
31– 40 mm (Arms 3, 6, and 9). Eighty percent of patients
was lowered from 21 Gy to 18 Gy because of an unaccept-
were treated with a linac, while 20% underwent Gamma
able incidence of chronic CNS toxicity observed at the 21
Knife radiosurgery, all of whom were from Mayo Clinic or
Gy dose level. In the patients with 31– 40 mm tumors, the
University of California-San Francisco. Differences be-
radiosurgical dose was escalated from 12 Gy to 15 Gy and
tween linac and Gamma Knife treated patients, as well as
then to 18 Gy. The maximum tolerable dose was 15 Gy,
compliance with protocol radiosurgery, are shown in Table
because both unacceptable acute and chronic toxicity were
4. Of note, 30% of linac treated patients had recurrent
observed at 18 Gy. In total, 35 unacceptable acute and
primary brain tumors compared to 61% of patients treated
chronic toxicities were observed in the 156 patients (22%
on a Gamma Knife.
incidence), of which 15 were irreversible grade 3 (severe)
(10% of all patients, 42% of unacceptable toxicity group),
CNS Toxicity
16 were grade 4 (life-threatening) (10% of all patients, 46%
At the time of analysis, living patients had 1.0 to 5.7 years
of unacceptable toxicity group), and 4 were grade 5 (fatal)
follow-up (median, 3.0 years). The incidence of unaccept-
(3% of all patients, 11% of unacceptable toxicity group).
The grade 3 toxicities were cases of irreversible edema
Table 5. Incidence of Grade 3, 4, and 5 CNS toxicity by tumor requiring inpatient administration of steroids. They oc-
size and treatment arm curred with a median time of 4.5 months following protocol
Incidence of Grade 3, 4, and 5 CNS Toxicity radiosurgery (range, 0.5–36 months). The grade 4 toxicities
were cases of radionecrosis requiring operation in 15 pa-
% of Patients With tients, and 1 case of clinically and radiographically sus-
Toxicity pected radionecrosis (which was subsequently confirmed at
No. of
Tumor size* Arm Dose patients Acute Chronic Total autopsy). They occurred with a median time of 5.0 months
following protocol radiosurgery (range 4.0 –27 months).
ⱕ 20 mm The actuarial incidence of radionecrosis was 5%, 8%, 9%
1 18 Gy 12 0 8 8 and 11% at 6, 12, 18 and 24 months following radiosurgery,
4 21 Gy 18 0 11 11
7 24 Gy 10 0 10 10
21–30 mm
2 15 Gy 15 7 7 13 Table 7. Imaging response following protocol radiosurgery
5 18 Gy 15 0 20 20
Change versus Imaging response
8 21 Gy 13 8 31 38
baseline (% of patients)
11 24 Gy 12 33 25 58
31–40 mm ⱕ 3 mos ⬎ 3–12 mos
3 12 Gy 21 5 5 10 Complete response 4 6
6 15 Gy 22 0 14 14 Improved 38 34
9 18 Gy 18 17 33 50 Unchanged 42 28
Worse 16 32
* Maximum tumor diameter.
 Radiosurgery for recurrent brain tumors ● E. SHAW et al. 295

Table 8. Failure patterns–First sites of failure progression was 48%, and the two-year actuarial incidence
of local progression was 50%. A multivariate analysis was
Any In-Volume (Local)*–40%
Local alone–27% performed to assess potential prognostic variables contrib-
Local ⫹ adjacent to target volume–6% uting to local tumor progression, including type of recurrent
Local ⫹ non-adjacent intracranial–3% tumor, number of lesions present, maximum tumor diame-
Local ⫹ adjacent to target volume ⫹ non-adjacent ter, tumor volume, treatment unit, and tumor dose. Of these,
intracranial–3% tumor type and treatment unit were statistically significant
Local ⫹/⫺ adjacent to target volume ⫹/⫺ non-adjacent
intracranial ⫹ systemic–2% (p ⫽ 0.021 and 0.018, respectively). Recurrent primary
Any Adjacent–19% brain tumor patients had an odds ratio of 2.85 compared to
Any Non-Adjacent Intracranial–23% metastases, that is, 2.85 times higher risk of progressing
Any Systemic–19% locally. Patients treated on a linear accelerator were 2.84
No Failure–23% times more likely to have local progression compared to
* Overall incidence of local tumor progression was 48%. those treated on a Gamma Knife. Figure 1a shows the time
to local progression for linear accelerator and Gamma Knife
treated patients, broken down by those treated for recurrent
respectively. The 4 grade 5 toxicities were initially acute or primary brain tumors (Figure 1b) or brain metastases (Fig-
chronic grade 3 or 4 CNS toxicities that ultimately proved to ure 1c). The other failure patterns, including progression
be fatal. They occurred with a median time of 2.5 months adjacent to the target volume, nonadjacent intracranial pro-
following radiosurgery (range 1.0 – 65 months). gression, and systemic progression are also shown in Table
A multivariate analysis was performed to assess potential 8. Most of the patients developing non-adjacent intracranial
prognostic variables contributing to unacceptable CNS tox- progression or systemic progression had brain metastases as
icity, including recurrent tumor type, number of lesions their type of recurrent tumor.
present, maximum tumor diameter, tumor volume, prior The median survival time for all patients following pro-
radiotherapy dose, KPS, NFS, treatment unit, tumor dose, tocol treatment was 8.8 months, with one- and two-year
MD/PD ratio, and PIV/TV ratio. Maximum tumor diameter actuarial survival rates of 35% and 15%, respectively. Two
was one variable significantly associated with toxicity (p ⫽ patients (1%) were alive beyond 5 years. The survival was
0.054). Larger tumors had a greater risk of unacceptable somewhat better for patients with recurrent primary brain
CNS toxicity than smaller ones. Compared to tumors ⱕ 20 tumors versus those with recurrent brain metastases (median
mm in maximum diameter, tumors 21–30 mm had an odds survival time, 11.4 versus 7.5 months; one-year survival
ratio of 7.3, that is, a 7.3 times higher risk of developing rate, 49% versus 26% and two-year, 27% versus 8%, re-
unacceptable CNS toxicity. Tumors 31– 40 mm had an odds spectively).
ratio of 16.0. Other significant toxicity-associated variables
were tumor dose (p ⫽ 0.007) and KPS (p ⫽ 0.055). Both
DISCUSSION
increasing dose and higher KPS were associated with a
greater risk of developing unacceptable CNS toxicity, al- This study demonstrated the maximum acutely (and
though the latter finding is counter-intuitive. chronically) tolerable doses of single fraction radiosurgery
Data about steroid use was available in a subset of pa- for patients with recurrent previously irradiated primary
tients at points in time ⱕ 3 months and ⬎ 3 months brain tumors and brain metastases as 24 Gy, 18 Gy, and 15
following protocol radiosurgery. Twenty-seven percent of Gy for tumors ⱕ 20 mm, 21–30 mm, and 31– 40 mm in
patients not on steroids at the time of radiosurgery had to maximum diameter. The 24 Gy maximum dose for the ⱕ 20
subsequently start them, but the other 73% were able to mm tumors represents the maximum dose radiation that
remain off steroids following radiosurgical treatment. Sixty- oncologists were willing to administer to patients (i.e., in-
eight percent of patients on steroids at the time of protocol vestigators were reluctant to escalate to 27 Gy), rather than
radiosurgery had to continue their use, but the other 32% the dose above which there was unacceptable toxicity, since
were able to discontinue them. Changes in KPS and NFS, dose limiting CNS toxicity, acute or chronic, was not ob-
also available in a subset of patients, are shown in Table 6. served at 24 Gy in ⱕ 20 mm tumors. The study also
A minority of patients had an improvement in either KPS or demonstrated the feasibility of completing a cooperative
NFS, whereas one third to one half had stabilization or group, multi-institutional radiosurgery clinical trial using
worsening of these parameters. both linac, and Gamma Knife technology. Overall, 84% of
treatments were performed without any deviation from the
Response, failure patterns, and survival protocol, 11% with a minor deviation, and 5% with a major
Imaging response in the three months following protocol deviation.
radiosurgery was assessed and compared to response ⬎ 3 Twenty-two percent of patients developed unacceptable
months up to one year post-treatment. These data are shown CNS toxicity following protocol radiosurgery. The devel-
in Table 7. Forty percent of patients had local tumor pro- opment of unacceptable CNS toxicity was significantly
gression within the radiosurgical target volume as their first more likely in patients with larger tumors, based on maxi-
site of failure (Table 8). The total incidence of local tumor mum tumor diameter. This result is somewhat different than
296 I. J. Radiation Oncology ● Biology ● Physics Volume 47, Number 2, 2000

Fig. 1. Time to local progression as a function of treatment unit, Gamma Knife or linear accelerator, for (a) all patients,
and for patients with primary brain tumors (b) or brain metastases (c).

our initial report of this protocol, in which tumor volume metastases (newly diagnosed and recurrent), Loeffler and
(rather than diameter) and dose inhomogeneity (i.e., the Alexander reported a 4% incidence of reoperation, which
MD/PD ratio) were the two variables significantly associ- translated into an 18-month actuarial risk of 10% (11),
ated with unacceptable CNS toxicity (1). Foote et al., in a similar to that observed in our series (9% at 18 months).
series of 36 patients with acoustic neuroma treated with a However, Flickinger et al. observed a ⬍ 1% incidence of
Gamma Knife, observed a significantly higher incidence of reoperation for radionecrosis in a multiinstitutional Gamma
cranial neuropathy in tumors ⱖ 21 mm compared to smaller Knife users group study of 116 patients with radiosurgically
ones (8). Irreversible cerebral edema requiring inpatient treated solitary brain metastasis (12). While death from
steroid administration and necrosis requiring an operation radiosurgically induced radionecrosis has also been reported
were the most common unacceptable CNS toxicities in our (13), reoperation can benefit a patient’s symptoms and sur-
series. Our two-year incidence of radionecrosis was 11%. vival (9). Cerebral edema requiring steroids has been re-
Steroid dependence and reoperation have been reported in ported to occur as soon as immediately following single
other series of patients receiving single dose radiosurgery dose radiosurgery in 16% of 58 patients with recurrent
for recurrent previously irradiated primary brain tumors and glioblastoma multiforme reported by Alexander et al. (14),
brain metastases. Shrieve et al. found a 22% reoperation but as late as one year following radiosurgery in several
rate (two-year actuarial risk 48%) in 86 patients following series of patients undergoing radiosurgical treatment of
radiosurgical treatment of recurrent glioblastoma multi- brain metastases, newly diagnosed and recurrent (15, 16). In
forme, which was similar to the 23% reoperation rate in 22 our series, irreversible grade 3 (severe) CNS toxicity sec-
patients with recurrent glioblastoma reported by DeSalles et ondary to cerebral edema occurred in 15 of the 35 patients
al.(9, 10). In 196 patients receiving radiosurgery for brain with unacceptable CNS toxicity at a median time of 4.5
 Radiosurgery for recurrent brain tumors ● E. SHAW et al. 297

months following protocol radiosurgery. Twenty-two per- Table 9. RTOG radiosurgery clinical trials
cent of patients on steroids at the time of protocol radiosur-
Current/
gery were able to discontinue their use. The observation that Protocol Phase Disease Schema target accrual*
steroids could be discontinued in a significant percentage of
patients following radiosurgery for brain metastases has 93-05 III GBM RT/BCNU 108/200
also been made by Adler et al. (15) and Mehta et al. (17). ⫹/⫺ RS†
95-08 III Brain RT ⫹/⫺ RS† 172/262
On the other hand, half to 64% of patients undergoing metastases
radiosurgery for recurrent glioblastoma multiforme remain
steroid dependent following treatment (10, 18). Steroid de- * As of 3/21/00.
pendency and radionecrosis may be less likely following
†
Radiosurgery (RS) doses are 24 Gy for tumors ⱕ 20 mm in
radiosurgery for recurrent brain tumors with the use of maximum diameter, 18 Gy for tumors 21–30 mm, and 15 Gy for
tumors 31– 40 mm; the dose of fractionated external beam radia-
fractionated treatment. Shrieve et al. reported on 15 patients tion therapy (RT) in 93– 05 is 60 Gy in 30 fractions, and in 95– 08
with recurrent previously irradiated gliomas treated with is 37.5 in 15 fractions.
fractionated stereotactic radiosurgery, 20 –25 Gy in 4 –5
fractions. Though follow-up was short (1 to 21.5 months),
no patient had required prolonged steroid use or reoperation While such inhomogeneity is undesirable in standard, large
for radionecrosis (19). field external beam radiation therapy, it may have effec-
Local tumor control within the radiosurgical target vol- tively boosted the central, hypoxic, more radioresistant
ume was achieved in 52% of patients (Tables 7 and 8). This portion of the tumor, accounting for the better local control.
falls between the range of local control rates reported in Besides the MD/PD ratio, the other technical difference was
other series of single dose radiosurgical treatment of recur- the PIV/TV ratio, a measure of conformity. While Gamma
rent previously irradiated primary brain tumors and brain Knife treatments tended to be more conformal than those on
metastases, 38% for recurrent glioblastoma (14), and 82– the linac, based on a 23% compared to a 40% PIV/TV ratio,
94% for brain metastases (11, 12, 18). One of the prognostic this is unlikely to account for the difference in local control.
variables associated with a significantly increased risk of Our series assessed quality of life following protocol
local tumor progression was predictable, namely recurrent radiosurgery by serially assessing Karnofsky Performance
primary brain tumors (versus recurrent brain metastases). Status and Neurologic Function Score (Table 6). Improve-
The other variable not predicted was the treatment unit. ment in KPS or NFS occurred in 13% of patients. Stability
Patients treated on a linac had a 2.84 times higher risk of of KPS or NFS was observed in 41% of patients while a
local tumor progression compared to those who underwent similar number (45%) had worsening of KPS or NFS. The
Gamma Knife radiosurgery, despite having a more favor- nature of changes in KPS and NFS are certainly multifac-
able mix of tumors (30% recurrent primary brain tumors vs. torial, and depend upon things such as tumor control and
61% in Gamma Knife treated patients), an observation that treatment toxicity.
has not been made previously. One possible explanation lies Currently, there are two ongoing RTOG radiosurgery
in the inherent inhomogeneity that exists in the dosimetry of clinical trials, including Protocols 93-05 and 95-08 (Table
Gamma Knife radiosurgery. As seen in Table 4, 94% of 9), in which data being collected should provide further
Gamma Knife treated patients had an MD/PD ratio of two, information on the observations made in this study. Institu-
since dose was prescribed to the 50% isodose line. This tions performing linear accelerator or Gamma Knife radio-
results in a dose gradient of 100% from the center of the surgery who are not members of RTOG are encouraged to
tumor to the periphery, that is, there are internal “hot spots.” join and take part in these important clinical trials.

REFERENCES
1. Shaw E, Scott C, Souhami L, et al. Radiosurgery for the 6. BMDP Statistical Software Manual, Vol. 2, Berkeley, CA:
treatment of previously irradiated recurrent primary brain tu- University of CaliforniaPreW; 1992.
mors and brain metastases: Initial report of Radiation Therapy 7. Agresti A. Categorical data analysis. New York: John Wiley
Oncology Group Protocol 90-05. Int J Radiat Oncol Biol Phys & Sons; 1990.
1996;34:647– 654. 8. Foote RL, Coffey RJ, Swanson JW, et al. Stereotactic radio-
2. Shaw E, Scott C, Souhami L, et al. Radiosurgery for the surgery using the Gamma Knife for acoustic neuromas. Int J
treatment of previously irradiated recurrent primary brain tu- Radiat Oncol Biol Phys 1995;32:1153–1160.
mors and brain metastases: Final report of Radiation Therapy 9. Shrieve DC, Alexander E, Wen PY, et al. Comparison of
Oncology Group Protocol 90-05 (abstract). Int J Radiat Oncol stereotactic radiosurgery and brachytherapy in the treatment of
Biol Phys 1995;32(Suppl. 1):145. recurrent glioblastoma multiforme. Neurosurg 1995;36:275–
3. Kaplan EL, Meier, P. Nonparametric estimation from incom- 284.
plete observations. J Amer Stat Assoc 1958;53:457– 481. 10. DeSalles A, Brekhus S, Goetsch S, et al. Steroid dependence
4. Caplan RJ, Pajak TF, Cox JD. Analysis of the probability and and quality of life after radiosurgery for glioblastoma multi-
risk of cause-specific-failure. Int J Radiat Oncol Biol Phys forme (abstract). Acta Neurochir 1993;122:155.
1994;29:1183–1186. 11. Loeffler JS, Alexander E. Radiosurgery for the treatment of
5. Staffact User Manual. Cambridge, MA: Cytel Software; 1992. intracranial metastases. In: Alexander E, Loeffler JS, Lunsford
298 I. J. Radiation Oncology ● Biology ● Physics Volume 47, Number 2, 2000

LD, editors. Stereotactic radiosurgery. New York: McGraw in brain metastases. In: Lunsford LD, editor. Stereotactic
Hill, Inc.; 1993. p. 197–206. radiosurgery update. New York: Elsevier Scientific Publishing
12. Flickinger JC, Kondziolka D, Lunsford LD, et al. A multi- Co., Inc.; 1992. p. 429 – 434.
institutional experience with stereotactic radiosurgery for sol- 17. Mehta MP, Rozental JM, Levin AB, et al. Defining the role of
itary brain metastasis. Int J Radiat Oncol Biol Phys 1994;28: radiosurgery in the management of brain metastases. Int J
797– 802. Radiat Oncol Biol Phys 1992;24:619 – 625.
13. Engenhart R, Kimmig BN, Hover KH, et al. Long-term fol- 18. Mehta MP, Masciopinto J, Rozental J, et al. Stereotactic
low-up for brain metastases treated by percutaneous stereo- radiosurgery for glioblastoma-multiforme: Report of a pro-
tactic single high-dose irradiation. Cancer 1993;71: spective study evaluating prognostic factors and analyzing
1353–1361. long-term survival advantage. Int J Radiat Oncol Biol Phys
14. Alexander E, Coffey R, Loeffler JS. Radiosurgery for gliomas. 1994;30:541–549.
In: Alexander E, Loeffler JS, Lunsford LD, editors. Stereotac- 19. Shrieve DC, Tarbell NJ, Goumnerova L, Loeffler JS. Hypo-
tic Radiosurgery. New York: McGraw-Hill, Inc.; 1993. p. fractionated stereotactic radiotherapy for recurrent gliomas in
207–219. children and adults: A biological compromise between radio-
15. Adler JR, Cox RS, Kaplan L, et al. Stereotactic radiosurgical surgery and conventionally fractionated radiotherapy. In:
treatment of brain metastases. J Neurosurg 1992;76:444 – 449. Kondziolka D, editor. Radiosurgery 1995. Basel: Karger;
16. Kihlstrom L, Karlsson B, Lindquist C. Gamma knife surgery 1996. p. 158 –164.