PII S0360-3016(02)02784-0

CLINICAL INVESTIGATION Brain

ANALYSIS OF REPEAT STEREOTACTIC RADIOSURGERY FORPROGRESSIVE PRIMARY AND METASTATIC CNS TUMORS

AJAY BHATNAGAR, M.D.,* DWIGHT E. HERON, M.D.,* DOUGLAS KONDZIOLKA, M.D.,*†

L. DADE LUNSFORD, M.D.,*†‡AND JOHN C. FLICKINGER, M.D.*†

Departments of *Radiation Oncology,†Neurological Surgery, and‡Radiology, Center for Image-Guided Neurosurgery,University of Pittsburgh School of Medicine, Pittsburgh, PA

Purpose: To identify and evaluate the pretreatment and patient factors that would predict for complications afterrepeat radiosurgery.Methods and Materials: The data from 26 patients who underwent reirradiation with Gamma Knife surgeryafter a previous procedure in the same or subjacent location were available for evaluation. The range of follow-upwas 1–45 months (mean 10). The mean minimal and maximal initial dose and volume for all 26 patients was 16.2Gy (range 12–22), 31.0 Gy (range 22.2–40.0), and 12.4 cm3 (range 1.20–70.84), respectively. The mean marginaland maximal repeated radiosurgery dose and volume for all 26 patients was 14.9 Gy (range 12–22.5), 29.7 Gy(range 18.0–45.0) and 12.8 cm3 (range 1.10–39.20), respectively.Results: Tumor control was significantly better statistically (p � 0.0129) for benign tumors (6 of 6, 100%actuarial rate at 4 years) compared with malignant tumors (7 of 20, 35% actuarial rate at 3 years, 3 of 4metastatic tumors and 2 of 10 primary malignant gliomas). The retreatment volume for radiosurgery correlatedsignificantly with the probability of neurologic decline (any cause) (p � 0.0181).Conclusion: Repeat radiosurgery can be performed for recurrent tumors with minimal central nervous systemtoxicity, especially for benign tumors, with reasonable tumor control. © 2002 Elsevier Science Inc.

Reirradiation, Radiosurgery, CNS tumors, Gamma Knife.

INTRODUCTION

Nearly 20,000 people in the United States develop a primarycentral nervous system (CNS) tumor each year. Most ofthese tumors are high-grade gliomas (1). In addition, thebrain is one of the most common sites of systemic spreadfrom solid tumors, such as lung cancer, breast cancer, mel-anoma, and renal cell carcinoma (2). Nearly 50% of allpatients diagnosed with cancer will develop brain metasta-ses (3, 4). The common modalities of treatment for primaryand metastatic CNS tumors are surgery, external beamradiotherapy (EBRT), and chemotherapy. Stereotactic ra-diosurgery has been proved to be an effective alternativeand, in some cases, adjunctive treatment for certain CNStumors, primary and metastatic, because it delivers a high-dose single fraction to a focused target volume, sparingnormal brain tissue (5–7). Radiosurgery is ideally suited forpatients with CNS tumors that are few in number (usually�3) and small in size, generally�3–5 cm in diameter (7).

The vast majority of patients with primary or metastaticCNS tumors die as a result of local tumor progression,despite radical intervention with surgery, radiotherapy,

and/or chemotherapy (8). Retreatment with EBRT has beenstudied, but most reports noted limited success because ofthe significant acute, subacute, and late toxicities of reirra-diation (9–13).

Few reports of radiosurgery for reirradiation after failureof whole brain or large field radiotherapy have been done(14–17). The results from the Radiation Therapy OncologyGroup (RTOG) 90-05 protocol, which analyzed radiosur-gery for the treatment of previously irradiated primary andmetastatic CNS tumors, demonstrated a maximum toleratedsingle radiosurgical dose of 24 Gy, 18 Gy, 15 Gy for tumors�20, 21–30, and 31–40 mm in maximal diameter (15).However, this trial used radiosurgery after EBRT as theinitial radiation treatment, which is not repeat radiosurgery.Recently, repeat radiosurgery for cerebral arteriovenousmalformation has been reported (18). Little information onrepeating radiosurgery for primary and metastatic CNS tu-mors is available.

The purpose of this paper was to identify and evaluate thepretreatment and patient factors that would predict for tu-mor control and complications after Gamma Knife reirra-

Reprint requests to: John C. Flickinger, M.D., Radiation Oncol-ogy B-300, 200 Lothrop St., Pittsburgh, PA 15213. Tel: (412)647-3600; Fax: (412) 647-6029; E-mail: flickingerjc@msx.upmc.eduAcknowledgment—We greatly thank Carol Rawlins for her hard

work in helping draft the manuscript; without her help, this paperwould not have been possible.

Received Sep 19, 2001, and in revised form Feb 4, 2002.Accepted for publication Feb 11, 2002.

Int. J. Radiation Oncology Biol. Phys., Vol. 53, No. 3, pp. 527–532, 2002Copyright © 2002 Elsevier Science Inc.Printed in the USA. All rights reserved

0360-3016/02/$–see front matter

527

diation. Our hypothesis was that radiosurgical reirradiationwould be effective with acceptable toxicity.

METHODS AND MATERIALS

PatientsIn a retrospective review of all patients treated at the

University of Pittsburgh School of Medicine, we identified26 patients who were reirradiated with Gamma Knife sur-gery after a prior procedure in the same or subjacent loca-tion of 3443 patients who underwent Gamma Knife radio-surgery between July 1991 and November 2000. Thedefinition of the subjacent location was based on a closeproximity to the previously irradiated site, an area that, if itwere to be treated, the Gamma Knife portal would overlapwith the previously treated field. The pretreatment data andpatient characteristics included the diagnosis, tumor loca-tion, interval between treatments, dose characteristics ofboth treatments, treatment volumes for each session, base-line and subsequent neurologic symptoms, radiographic ev-idence of a change in tumor size, documented evidence of achange in the use of steroids after retreatment, and treat-ment-related toxicities such as radionecrosis. Toxicity wasscored using the RTOG CNS toxicity scale.

Statistical analysisUsing the Statistical Package for Social Sciences soft-

ware, Kaplan–Meier survival analysis was performed forpatients divided into cohorts of malignant or benign disease,with overall survival and tumor control as the primary endpoints of the analysis. Patients were classified into themalignant or benign cohorts on the basis of their originaltumor histologic findings. In this study population, benignmeningioma, acoustic neuroma, and pituitary adenomaswere classified as benign tumors, and the remaining tumorhistologic features were designated malignant tumors.

We performed stepwise (forward conditional) multivari-ate Cox regression analyses of outcome (tumor growth,neurologic decline) to identify and model any correlationwith treatment parameters, including volume, marginal doseof both treatments, interval between treatments, and type oftumor (malignant or benign). Tumor growth was based onthe most recent MRI follow-up report showing an increasein size compared with the prior MRI. Neurologic declinewas based on the neurologic symptoms, which were classi-fied into 3 subsets (seizures, focal deficits, and headaches)before reirradiation compared with after the repeat radio-surgery. An increase in any of the neurologic symptoms ora new symptom after reirradiation was considered neuro-logic decline.

RESULTS

Patient characteristicsThe study population consisted of 26 patients who un-

derwent repeat radiosurgery between July 1991 and Novem-ber 2000. Table 1 shows the patient characteristics divided

into cohorts according to the tumor histologic findings. Therange of follow-up was 1–45 months (mean 10). The meanage for all patients in the study was 50 years. The percent-age of men and women was 54% (14 of 26) and 46% (12 of26), respectively. At last follow-up, 20 (77%) of the 26patients in the study were living. The most common tumorhistologic features were meningioma (benign and malig-nant, 23%), glioblastoma multiforme (19%), and brain me-tastases (15%); the remaining histologic features are sum-marized in Table 1. This last group included metastasesfrom any primary solid tumors, including lung, breast, andunknown primary site. The mean interval between the initialand subsequent treatment was 22.3 months for the entirepatient population. The mean marginal dose to the isocenterfor the first and second treatments for all patients was 16.2and 14.9 Gy, respectively. The mean maximal dose to theisocenter for the first and second treatments for all patientswas 31.0 and 29.7 Gy, respectively. The mean treatmentvolume for the initial and retreatment Gamma Knife radio-surgeries for the entire patient population in the study was12.4 and 12.8 cm3, respectively. Only 3 patients in thisstudy exhibited radiographic evidence of radionecrosis afterreirradiation.

Steroid useTable 2 shows the use of steroids after reirradiation for all

evaluated patients classified by malignant and benign tu-mor. The percentage of patients requiring new or an in-creased dose of steroids after reirradiation was similar forboth benign and malignant histologic findings. Patients withbenign tumors had a slightly higher rate of no use or adecrease in steroid use compared with patients with malig-nant tumors. Overall, most of the study population de-creased or ceased requirements for steroid use after reirra-diation (Table 2).

Kaplan–Meier survival analysis for overall survivalKaplan–Meier survival analysis was performed for pa-

tients according to malignant or benign disease, with overallsurvival as the end point of analysis. Figure 1 illustrates thesurvival outcome in the two groups, with the benign cohorthaving a 100% survival rate and the malignant cohort hav-ing a �65% survival rate during 10 months of follow-up. Alog–rank test of this analysis proved to be statisticallyinsignificant (p � 0.1343).

Kaplan–Meier survival analysis for tumor controlUsing tumor control as the end point, Kaplan–Meier

survival analysis was executed for the entire patient popu-lation divided according to the histologic findings (Fig. 2).This graph demonstrates the degree of tumor control for thebenign cohort; the malignant cohort had a gradual decline.The log–rank test proved to be statistically significant (p �0.0129).

528 I. J. Radiation Oncology ● Biology ● Physics Volume 53, Number 3, 2002

Table 1. Patient characteristics

GBM/AA/PNETAcoustic neuroma/pituitary adenoma Brain metastases All others* Total

n 9 8 4 5 26Age (y)

Range 14–53 20–79 56–72 45–67 20–79Mean 37 54 66 59 50

Gender (n)Female 3 5 2 2 12Male 6 3 2 3 14

Status at last follow-up (n)Alive 5 8 2 5 20Dead 4 0 2 0 6

Treatment interval (mo)Range 4–45 10–108 8–13.0 3–34 3–108Mean 16.1 52.6 10.2 13.7 22.3

Marginal dose (Gy)GK1

Range 13–20 12–20 14–20 14–20 12.0–20Mean 16.7 15.3 17.5 15.7 16.2

GK2Range 9–20 12–23 12.0–18 13–20 12–22.5Mean 15.4 14.4 14.3 16.4 14.9

GK1�2Range 22–40 24–40 28–36 27–37 22–40Mean 32.1 29.6 31.8 32.1 31.1

Maximal dose (Gy)GK1

Range 22.2–40 24–33 28–40 28–40 22.2–40Mean 30.8 29.0 35 31.3 31.0

GK2Range 18–40 20–45 24–36 22–40 18–45Mean 30.9 28.8 28.5 30.2 29.7

GK1�2Range 44–72 48–73 56–72 51–74 44–73.3Mean 61.7 57.7 63.5 61.5 59.7

Treatment volume (cm3)GK1

Range 2.8–70.8 4.3–35.3 9.9–21.4 4.8–23.8 12–70.8Mean 14.6 11.1 13.4 9.8 12.4

GK2Range 1.4–27.1 1.1–13.8 2.9–32.7 2.4–17.2 1.1–39.2Mean 18.9 7.0 19.2 6.1 12.8

GK1�2Range 7.6–107.3 4.3–38.8 14.3–43.6 8.7–27.9 4.3–107.3Mean 33.5 18.1 32.6 16.0 25.2

* Hemangiopericytoma, ependyoma, nasopharyngeal squamous cell carcinoma, and choroid plexus papilloma.Abbreviations: GBM � glioblastoma multiforme; AA � anaplastic astrocytoma; PNET � primitive neuroectodermal tumor; GK1 �

Gamma Knife treatment 1; GK2 � Gamma Knife treatment 2; GK1�2 � sum of GK1 and GK2.

Table 2. Steroid use after reirradiation among patients with malignant or benign CNS tumors

Increase in use (n) No use or decrease (n)No change in use frombefore reirradiation (n) Total (n)

Malignant 6 (30) 9 (45) 5 (25) 20Benign 2 (33) 4 (67) 0 (0) 6Total 8 (31) 13 (50) 5 (19) 26

Abbreviation: CNS � central nervous system.Numbers in parentheses are percentages.

529Repeat radiosurgery for brain tumors ● A. BHATNAGAR et al.

Cox regression analysis for tumor growthTable 3 shows the results of the multivariate Cox regres-

sion analysis of the study parameters (benign tumor histo-logic features, minimal and maximal doses of both treat-ments, interval between treatments, and treatment volumeof reirradiation) on tumor growth after reirradiation. In theinitial univariate analysis for Cox regression, benign tumorfeatures were statistically significant for an inverse correla-tion with tumor growth after reirradiation (p � 0.0147). Inthe subsequent multivariate analysis, none of these factorswas statistically significant, including benign tumor (p �0.1520). However, the treatment volume of reirradiationshowed a trend toward significance (p � 0.0883) in themultivariate analysis.

Cox regression analysis for neurologic declineTable 4 shows the results of the multivariate Cox regres-

sion analysis of the study parameters (benign tumor histo-logic features, minimal and maximal doses of both treat-ments, interval between treatments, and treatment volumeof reirradiation) on neurologic function after reirradiation.

DISCUSSION

The reirradiation of CNS tumors has been a topic ofresearch for at least the past two decades; yet, only limiteddata are available regarding the effectiveness and safety ofreirradiation using EBRT (9–13). Bauman et al. (10) re-ported only modest palliative and survival benefits but couldfind no consensus on safety of reirradiation with EBRT. Stillothers have tried to define a population more apt to benefitfrom retreatment. Cooper et al. (9) stated that reirradiationshould only be considered for patients with an excellentKarnofsky performance score and with an interval to re-treatment of at least 4 months from the initial RT (9).Recently, Veninga et al. (13) reported clinical improvementin 24% of their patients receiving reirradiation for primaryCNS tumor, with most patients preserving their quality oflife after retreatment. Veninga et al. (13) also identified 4prognostic factors predictive of outcome with reirradiation:World Health Organization score before retreatment, inter-val between treatments, tumor histologic features, and theresponse to the initial treatment.

Animal studies of reirradiation with EBRT have indicatedsubstantial repair taking place up to 4–6 months after theinitial treatment (19, 20). Several studies have demonstratedthat the retreatment tolerance of the rat spinal cord, the most

Fig. 1. Kaplan–Meier analysis of overall survival.

Fig. 2. Kaplan–Meier analysis of tumor control function.

Table 3. Cox regression analysis for tumor growth

Parameters p

Benign tumor 0.1520Treatment volume for repeat GK 0.0883Marginal dose of initial GK 0.5537Marginal dose for repeat GK 0.4215Isocenter dose for initial GK 0.3630Isocenter dose for repeat GK 0.3132Interval between treatments (wk) 0.2514

Abbreviation: GK � Gamma Knife.

Table 4. Cox regression analysis for neurologic decline

Parameters p

Volume for repeat GK 0.0181Marginal dose for initial GK 0.0666Marginal dose for repeat GK 0.9178Isocenter dose for initial GK 0.8260Isocenter dose for repeat GK 0.8338Benign tumor 0.5265Interval between treatments (wk) 0.1630

Abbreviation: GK � Gamma Knife.

530 I. J. Radiation Oncology ● Biology ● Physics Volume 53, Number 3, 2002

common animal model for studying CNS reirradiation, de-pends on the initial dose and the interval from initial irra-diation (19, 20). The higher the initial dose, the longersignificant repair seems to continue. Also, Wong et al. (21)showed that regardless of any specific interval for reirradia-tion after the initial treatment, the decrease in latent times totoxicity was progressive with increasing size of the initialinjury (21). In our human study, neither the interval nordose was significant for tumor control or neurologic decline.

A recent report evaluating cranial reirradiation was theRTOG 90-05 protocol, which analyzed radiosurgery forretreatment of patients who had previously undergoneEBRT for primary and metastatic tumors (15). The resultsdemonstrated the feasibility of radiosurgery. The maximaltolerated single radiosurgical dose ranged from 15 to 24 Gyand was dependent on the treatment volume. Recently,some reports of repeat radiosurgery have been published.Maesawa et al. (18) reported their results for repeat radio-surgery for incompletely obliterated cerebral arteriovenousmalformations. Repeat radiosurgery achieved obliteration in21 (70%) of 30 of the patients, with acceptable risk, usingsimilar or slightly higher radiation doses to achieve thesame in-field obliteration rates as for nonirradiated arterio-venous malformations (18).

This analysis of repeat Gamma Knife radiosurgery forreirradiation of progressive primary and metastatic tumorsof the CNS suggests that repeat radiosurgery is feasible witha minimal change in neurologic status. However, we iden-tified one factor (treatment volume) likely to predict forcomplications of this procedure. There were few CNS tox-icity reports in this study. Repeat radiosurgery has beenassociated with radiation necrosis, a major complication ofGamma Knife radiosurgery (22). Only 3 patients were re-ported to have radionecrosis on MRI reports, which canshow increased enhancement post Gamma Knife radiosur-gery due to irritation of the blood–brain barrier. These wereall clinically occult.

A useful surrogate for quality of life in this cohort ofpatients is the need for steroids after treatment among thesepatients. Steroids are commonly used to relieve severesymptoms and edema from tumor progression such as focaldeficits, headaches, and seizures. An increase in the use ofsteroids would indirectly reflect a patient’s changing neu-rologic status as increasing in severity, and a decrease in theuse of steroids could potentially imply that the patient’sneurologic symptoms are improving. Data on steroid use forthe total patient population revealed that most patients hadimprovement in their neurologic symptoms with a decreaseor cessation of steroids (52%), and 23% of the patients hadno further decline in neurologic status with no change in

their use of steroids. Admittedly, a limitation of this methodis that there is no record of the exact point when the changein steroid use occurred, and therefore the dynamic nature ofthe symptoms and symptom control with steroids cannot befully assessed.

Cox regression analysis for neurologic decline was an-other way we studied the neurologic status of the patientpopulation. In this analysis, we attempted to identify certainvariables that could have an impact on neurologic function.The parameters tested were histologic features, marginaland isocenter doses of both treatments, interval betweentreatments, and treatment volume of reirradiation. The onlyvariable that was statistically significant in this analysis wasthe treatment volume for retreatment (p � 0.0181). Thisresult would be expected, because a larger treatment volumeindicates a larger tumor size, which is more likely to causeneurologic symptoms than is a smaller tumor size, despiteradical efforts.

Kaplan–Meier survival analysis showed that the benigngroup had a 100% overall survival rate and complete tumorcontrol and the malignant group had lower overall survivaland tumor control. The log–rank test showed that the tumorcontrol for the benign group was significantly better statis-tically than that of the malignant group.

It is interesting that Cox regression analysis for tumorgrowth using several treatment parameters that may have animpact on local tumor growth showed that benign tumorswere significantly better controlled statistically comparedwith malignant tumors. However, in the subsequent multi-variate analysis, the differences were no longer statisticallysignificant. It is possible that this occurred as a result of thesmall patient sample size or because of the excessive num-ber of parameters tested in the Cox regression analysis. Noother parameter was statistically significant, although theretreatment volume exhibited a trend toward significance.The treatment volume is an indirect marker of tumor size,and this would be expected to correlate with tumor growth,because a larger treatment volume indicates a more unfa-vorable prognosis than a smaller treatment volume.

CONCLUSION

The results of this first report of repeat Gamma Kniferadiosurgery for benign and malignant tumors demonstratethe efficacy of Gamma Knife retreatment with minimalCNS toxicity as determined by minimal neurologic changescompared with baseline for the 26 patients in this study. Tobetter characterize the scope of treatment and patient selec-tion factors, additional investigation of the use of repeatradiosurgery for progressive benign and malignant CNStumors is warranted.

REFERENCES

1. Laws ER Jr, Thapar K. Brain tumors. CA Cancer J Clin1993;43:263–271.

2. Johnson JD, Young B. Demographics of brain metastasis.Neurosurg Clin North Am 1996;7:337–344.

3. Pickren JW, Lopez G, Tzukada Y, et al. Brain metastases: Anautopsy study. Cancer Treat Symp 1983;2:295–313.

4. Posner JB, Chernik NL. Intracranial metastases from systemiccancer. Adv Neurol 1978;19:579–592.

531Repeat radiosurgery for brain tumors ● A. BHATNAGAR et al.

5. Mehta MP, Rozenthal JM, Levin AB, et al. Defining the roleof radiosurgery in the management of brain metastases. Int JRadiat Oncol Biol Phys 1992;24:619–625.

6. Flickinger JC, Kondziolka D, Lunsford LD, et al. A multi-institutional experience with stereotactic radiosurgery for sol-itary brain metastasis. Int J Radiat Oncol Biol Phys 1994;28:797–802.

7. Alexander E II, Coffey R, Loeffler JS. Radiosurgery for gli-omas. In: Alexander E II, Loeffler JS, Lunsford LD, editors.Stereotactic radiosurgery. New York: McGraw-Hill; 1993. p.207–219.

8. Borgelt B, Gelber R, Kramer S, et al. The palliation of brainmetastases: Final results of the first two studies by the Radi-ation Therapy Oncology Group. Int J Radiat Oncol Biol Phys1980;6:1–9.

9. Cooper JS, Steinfeld AD, Lerch IA. Cerebral metastases:Value of reirradiation in selected patients. Radiology 1990;174(3 Pt 1):883–885.

10. Bauman GS, Sneed PK, Wara WM, et al. Reirradiation ofprimary CNS tumors. Int J Radiat Oncol Biol Phys 1996;36:433–441.

11. Flickinger JC, Deutsch M, Lunsford LD. Repeat megavoltageirradiation of pituitary and suprasellar tumors. Int J RadiatOncol Biol Phys 1989;17:171–175.

12. Hazuka MB, Kinzie JJ. Brain metastases: Results and effectsof re-irradiation. Int J Radiat Oncol Biol Phys 1988;15:433–437.

13. Veninga T, Langendijk HA, Slotman BJ, et al. Reirradiationof primary brain tumours: Survival, clinical response andprognostic factors. Radiother Oncol 2001;59:127–137.

14. Shaw E, Scott C, Souhami L, et al. Radiosurgery for thetreatment of previously irradiated recurrent primary brain tu-

mors and brain metastasis: Initial report of Radiation TherapyOncology Group protocol (90-05). Int J Radiat Oncol BiolPhys 1996;34:647–654.

15. Shaw E, Scott C, Souhami L, et al. Single dose radiosurgicaltreatment of recurrent previously irradiated primary brain tu-mors and brain metastases: Final report of RTOG protocol90-05. Int J Radiat Oncol Biol Phys 2000;47:291–298.

16. Loeffler JS, Kooy HM, Wen PY, et al. The treatment ofrecurrent brain metastases with stereotactic radiosurgery.J Clin Oncol 1990;8:576–582.

17. Simpson JR, Mendenhall WM, Schupak KD, et al., for theAmerican College of Radiology. Follow-up, and retreatmentof brain metastases: ACR appropriateness criteria. Radiology2000;215(Suppl.):1129–1135.

18. Maesawa S, Flickinger JC, Kondziolka D, et al. Repeatedradiosurgery for incompletely obliterated arteriovenous mal-formations. J Neurosurg 2000;92:961–970.

19. Wong CS, Hao Y. Long-term recovery kinetics of radiationdamage in rat spinal cord. Int J Radiat Oncol Biol Phys1997;37:171–179.

20. van der Kogel AJ. Central nervous system radiation injury insmall animal models. In: Gutin PH, Leibel SA, Sheline GE,editors. Radiation injury to the nervous system. New York:Raven Press; 1991. p. 91–111.

21. Wong CS, Poon JK, Hill RP. Re-irradiation tolerance in the ratspinal cord: Influence of level of initial damage. RadiotherOncol 1993;26:132–138.

22. Chin LS, MA L, DiBiase S. Radiation necrosis followinggamma knife surgery: A case controlled comparison of treat-ment parameters and long term follow up. J Neurosurg 2001;94:899–904.

532 I. J. Radiation Oncology ● Biology ● Physics Volume 53, Number 3, 2002