Leukoencephalopathy after whole-brain radiation therapy plus radiosurgery versus radiosurgery alone for metastatic lung cancer




As systemic therapies improve and patients live longer, concerns mount about the toxicity of whole-brain radiation therapy (WBRT) for treatment of brain metastases. Development of delayed white matter abnormalities indicative of leukoencephalopathy have been correlated with cognitive dysfunction. This study assesses the risk of imaging-defined leukoencephalopathy in patients whose management included WBRT in addition to stereotactic radiosurgery (SRS). This risk is compared to patients who only underwent SRS.


We retrospectively compared 37 patients with non–small cell lung cancer who underwent WBRT plus SRS to 31 patients who underwent only SRS. All patients survived at least 1 year after treatment. We graded the development of delayed white matter changes on magnetic resonance imaging using a scale to evaluate T2/FLAIR (fluid attenuated image recovery) images: grade 1 = little or no white matter hyperintensity; grade 2 = limited periventricular hyperintensity; and grade 3 = diffuse white matter hyperintensity.


Patients treated with WBRT and SRS had a significantly greater incidence of delayed white matter leukoencephalopathy compared to patients who underwent SRS alone (P < .001). On final imaging, 36 of 37 patients (97.3%) treated by WBRT developed leukoencephalopathy (25% with grade 2; 70.8% with grade 3). Only 1 patient treated with SRS alone developed leukoencephalopathy.


Risk of leukoencephalopathy in patients treated with SRS alone for brain metastases was significantly lower than that for patients treated with WBRT plus SRS. A prospective study is necessary to correlate these findings with neurocognition and quality of life. These data supplement existing reports regarding the differential effects of WBRT and SRS on normal brain structure and function. Cancer 2013. © 2012 American Cancer Society.

Whole-brain radiation therapy (WBRT) has been advocated as the primary treatment for metastatic brain cancer. The benefit of WBRT has not changed markedly despite alterations in dose, fractionation schedules, or the addition of radiosensitizers.1-5 Stereotactic radiosurgery (SRS) has emerged as a minimally invasive initial treatment for brain metastases and has been applied to a variety of histologies and anatomic locations.6-13 In a prospective randomized, controlled trial for patients with 1 to 4 metastases, no difference in median survival was found between SRS alone and SRS plus WBRT.14 These results challenge the reflexive use of up-front WBRT in patients with newly diagnosed metastatic brain cancer.

Late neurotoxic effects of WBRT have been difficult to study, because patient cohorts have short life expectancies. Earlier cancer diagnosis and aggressive systemic therapies have improved survival for many cancers but have also raised concerns about long-term toxicities of treatment. WBRT injures small cerebral vasculature and neuropil,15-17 effects linked to imaging-defined white matter changes. Oligodendrocytic death, the resultant demyelination, and a failure to replenish these cells due to killing of the subependymal stem cell population are likely key features in white matter injury.18, 19 Diseases of white matter are correlated with neurocognitive dysfunction,20-22 and WBRT has deleterious effects on cognition, memory, and mood.23

The present retrospective study in patients with lung cancer was designed to evaluate the risk of developing delayed leukoencephalopathy at least 1 year after WBRT plus SRS or SRS alone for brain metastases.


Patient Population

With Institutional Review Board approval, we retrospectively reviewed data from 295 consecutive patients who underwent Gamma Knife SRS for non–small cell lung cancer brain metastases between 2005 and 2009. Sixty-eight patients were included because they had evaluable imaging obtained at around 1 or more years after WBRT or SRS. Thirty-seven patients treated with WBRT and SRS were compared with 31 patients who only underwent SRS. Data regarding age, sex, chemotherapy, WBRT, number of SRS procedures, the number of metastases treated by Gamma Knife, and survival were able to be collected.


WBRT was administered at hospitals closest to patients' homes. Details of treatment were obtained from outside facilities' records prior to SRS. Total dose used most often (65% of patients) was 30 Gy in 10 fractions of 3 Gy. Median total dose was 30 Gy (range = 20 to 37.5 Gy). No patient had repeat WBRT. The WBRT completion date was the time from which imaging follow-up was calculated. A total of 32 of the 37 patients in the WBRT plus SRS cohort underwent SRS at some time after WBRT, and 4 of the remaining 5 patients in this group had up-front SRS followed by a WBRT boost within a month. In patients who received only SRS, imaging follow-up was calculated from the date of the first Gamma Knife procedure. All patients had at least 1 Gamma Knife radiosurgical procedure. For patients who received WBRT, 18 patients had a single SRS procedure, and 19 had 2 or more. In patients who had only SRS, 16 had 1 procedure, and 14 had 2 or more. Seventeen of the patients in the WBRT plus SRS group had “boost” or “early” consolidation SRS for their brain metastases. Sixteen of the patients in this cohort had SRS as a salvage therapy for new or recurrent tumors. Our radiosurgical technique has been described.12

Imaging Evaluation

Patients were followed with magnetic resonance imaging (MRI) evaluated at a median of 1 and 2 years after treatment. Previous qualitative scales have used numerous grades or involve many anatomic locations.22, 24, 25 These are time-consuming and impractical for everyday clinical use. Quantitative analyses can improve sensitivity and precision, but require uniformity between imaging studies and proprietary software.26 We sought a simple and easy-to-use scale for evaluating MRI in clinical practice that reflected the centrifugal progression of radiation-induced white matter change from the periventricular white matter.25 A grading scale was devised to assess imaging changes using T2 or FLAIR (fluid attenuated image recovery) sequences: grade 1 = little or no white matter hyperintensity; grade 2 = limited periventricular hyperintensity; and grade 3 = diffuse white matter hyperintensity. Local white matter changes from specific tumors were not incorporated. Figure 1 shows representative MRIs. The MRIs were evaluated by authors (E.M. and P.P.) blinded to patient treatments.

Figure 1.

Representative T2-weighted magnetic resonance imaging is shown for (A) grade 1, (B) grade 2, and (C) grade 3 white matter change.

Statistical Analyses

Descriptive statistics such as mean with standard deviation and median with range were used for continuous data, and categorical data were reported by frequencies and proportions. Variables pertaining to the 2 groups were compared with appropriate statistical tests to identify significant differences (SAS version 9.3; SAS Institute, Cary, NC).

Student t test was used for normally distributed continuous data, and the Wilcoxon rank-sum test was used for nonparametric continuous data not meeting the normality assumption. The Pearson chi-square test was used for categorical data, and the Fisher exact test was used for categorical data when the cells had an expected count of less than 5.


Patient Characteristics

Patient characteristics were similar between the 2 groups (Table 1). We found no significant difference in the rate of chemotherapy treatment (P = .09), a factor implicated in white matter change and neurocognitive dysfunction.27 Patients treated with WBRT and SRS had significantly greater total brain metastases treated by Gamma Knife compared to the SRS-only cohort (median of 6 vs 2; P < .01). The number of tumors treated at the initial SRS procedures was significantly greater in patients who had WBRT plus SRS (4 vs 1, respectively; P < .001). However, the 2 cohorts did not undergo a significantly different number of SRS procedures and in those patients who underwent additional salvage SRS procedures, the numbers of treated tumors was not significantly different. The SRS marginal tumor dosing for the 2 cohorts was lower for the WBRT plus SRS cohort versus the cohort with SRS alone (18 Gy vs 20 Gy, respectively, P < .001). This, in part, reflects our practice of routinely lowering the SRS prescription dose for patients who have undergone previous WBRT. Patients who received only SRS had a slightly better median survival (28.2 months vs 25.1 months; P = .05).

Table 1. Baseline Characteristics of the Study Cohorts
CharacteristicWBRT+SRSSRS OnlyP
  • Mets indicates metastases; MRI, magnetic resonance imaging; SRS, stereotactic radiosurgery; WBRT, whole-brain radiation therapy.

  • a

    Based on student t test.

  • b

    Based on Pearson chi-square test.

  • c

    Based on Fisher exact test.

  • d

    Based on Wilcoxon rank-sum test.

Total, n = 68n = 37n = 31 
Age, Mean (standard deviation)61.4 (9.1)63.4 (12.5).46a
Men19 (51.4%)15 (48.4%).81b
Chemotherapy37 (100%)28 (90.3%).09c
No. of SRS treatments Median (range)2 (1-5)1 (1-5).83d
No. of total Mets treated by SRS Median (range)6 (1-34)2 (1-19)<.01d
No. of Mets treated at initial SRS Median (range)4 (1-13)1 (1-7)<.001d
No. of Mets treated by salvage SRS Median (range)5 (1-26)3 (1-15).50d
Marginal SRS dose, Gy Median (range)18 (12-20)20 (10-20)<.001d
Proximity of baseline MRI to treatment, mo Median (range)–0.1 (–1.4 to 8.8)–0.3 (–1.8 to 3.7).29d
Time to first graded imaging, mo Median (range)12.8 (9.7-24.7)12.3 (8.1-39.4).44d
Time to second graded imaging, mo Median (range)23.6 (14.7-38.2)24.4 (15.1-65.4).16d
Survival, mo Median (range)25.1 (12.1-59.3)28.2 (13.9-70.1).05d

Imaging Results

The groups had similar initial white matter grades (Table 2). Leukoencephalopathy was noted in 8.1% for WBRT plus SRS patients versus 12.9% for SRS only (P = .69). At first evaluated imaging (median time after treatment of 12.8 and 12.3 months, respectively, for WBRT plus SRS and SRS only), a significantly larger proportion of WBRT patients had grades 2 or 3 leukoencephalopathy compared with patients treated with SRS only (91.9% vs 12.9%, respectively; P < .0001). Near 2 years after treatment (median time to second evaluated imaging was 23.6 and 24.4 months, respectively, for WBRT plus SRS and SRS only), 70.8% of WBRT patients showed grade 3 changes (Fig. 2 and Fig. 3A-C). No patients treated with only SRS developed grade 3 changes (P < .0001), with only 1 patient progressing at all and only to grade 2 (Fig. 2 and Fig. 3D-F).

Table 2. White Matter Change Following Treatment for Brain Metastases
  • SRS indicates stereotactic radiosurgery; WBRT, whole-brain radiation therapy.

  • a

    Based on Fisher exact test.

  • b

    Based on Pearson chi-square test.

White matter score at baseline   
 134 (91.9%)27 (87.1%) 
 23 (8.1%)4(12.9%) 
White matter score at first imaging   
 13 (8.1%)27 (87.1%) 
 217 (46%)4 (12.9%) 
White matter score at second imaging   
 11 (4.2%)23 (85.2%) 
 26 (25%)4 (14.8%) 
 317 (70.8%)0<.0001b
Figure 2.

Graph depicts changes in white matter grades between patients treated by whole-brain radiation therapy (WBRT) and Gamma Knife (GK) radiosurgery versus those treated by GK only, as a percentage of the total number of surviving patients (*P < .0001; **P < .0001).

Figure 3.

T2-weighted magnetic resonance images are shown from a patient at (A) baseline and at (B) 15.8 months and (C) 26.5 months after whole-brain radiation therapy, compared with those from a patient at (D) baseline and at (E) 11.4 months and (F) 22.3 months after first stereotactic radiosurgery procedure.

Tumor burden has been implicated in neurocognitive dysfunction in metastatic brain cancer.28 Because the 2 treatment groups differed in number of tumors treated by SRS, we performed an ad hoc analysis stratifying patients by this factor in order to determine if it contributed significantly to the development of white matter change. Stratifying patients on this basis (<10 metastases; ≥10 metastases) did not change the significant association between WBRT and white matter change, suggesting that number of metastases treated by SRS is not associated with leukoencephalopathy in this patient population (Table 3).

Table 3. White Matter Change Stratified by Number of Tumors Treated by SRS
CharacteristicWBRT+SRSSRS OnlyP
  • Mets indicates metastases; SRS, stereotactic radiosurgery; WBRT, whole-brain radiation therapy.

  • a

    Based on Fisher exact test.

  • b

    Based on Pearson chi-square test.

Patients with ≥10 Mets   
 White matter score at baseline   
  111 (100%)4 (100%) 
 White matter score at first imaging   
  11 (9.1%)4 (100%) 
  27 (63.6%)0 
  33 (27.3%)0<.01a
 White matter score at second imaging   
  104 (100%) 
  23 (42.9%)0 
  34 (57.1%)0<.01a
Patients with <10 Mets   
 White matter score at baseline   
  123 (88.5%)23 (85.2%) 
  23 (11.5%)4 (14.8%) 
 White matter score at first imaging   
  12 (7.7%)23 (85.2%) 
  210 (38.5%)4 (14.8%) 
  314 (53.9%)0<.0001a
 White matter score at second imaging   
  11 (5.9%)19 (82.6%) 
  23 (17.7%)4 (17.4%) 
  313 (76.5%)0<.0001b


Delayed radiation toxicity stems from breakdown of the blood–brain barrier with vasogenic edema,29 white matter endothelial hyperplasia,30-32 and focal demyelination.29 Imaging studies have correlated histopathology with white matter abnormalities. Computed tomography identified cerebral atrophy and white matter hypodensities.33, 34 MRI revealed progressive areas of white matter T2/FLAIR hyperintensity.35 Tsuruda et al described symmetrical periventricular and subcortical areas of white matter high signal intensity following cranial radiation in 95 patients.25 They noted relative sparing in younger patients and in the brainstem and internal capsule. The incidence of white matter change has varied. Fujii et al detected it in 83% of long-term survivors treated with WBRT for brain metastases, but not before 6 months.24 Szerlip et al found white matter changes in all 30 patients examined volumetrically after WBRT.26 Aoyama et al observed leukoencephalopathy in only 10.8% of brain metastases patients randomized to therapy including WBRT, potentially owing to this cohort's limited median survival (7.5 months).14 Similar imaging changes in white matter have been linked to neurocognitive dysfunction in several disorders.22, 36

A growing emphasis in cancer therapy is on quality of life, not merely disease control. Radiation causes meaningful neurocognitive toxicity in children. Irradiating increasing volumes of supratentorial brain in children correlates with decreased intelligence.37 This neurocognitive toxicity is of sufficient concern that strategies have evolved to delay or withhold radiation entirely.38 Delayed toxicity of WBRT is less well understood in adults, likely due to poor survival rates in conditions requiring WBRT. DeAngelis et al identified 12 cases of radiation-induced dementia detected 15 to 36 months after WBRT treatment for brain metastases.39 In a secondary analysis of a randomized trial, Aoyama et al found that although neurocognitive dysfunction correlated most with metastatic tumor recurrence, WBRT had important neurocognitive toxicity.40 Kondziolka et al surveyed brain metastasis patients following WBRT and SRS or SRS alone.23 Patients whose therapy included WBRT reported significantly greater difficulties with short- (72% vs 16%) and long-term memory (33% vs 13%), concentration (61% vs 25%), and depression (54% vs 19%). Chang et al, in the first randomized, controlled trial with a neurocognitive primary endpoint, observed that patients who received WBRT plus SRS possessed a significantly greater risk of developing learning and memory dysfunction by 4 months after treatment, compared with those who were treated with SRS alone, despite higher tumor recurrence with SRS alone.41

To our knowledge, the present study reports for the first time that WBRT plus SRS harbors a significantly greater risk of leukoencephalopathy when compared to a SRS-only approach. Nakaya et al calculated the total brain radiation dose in 105 patients after single-session SRS for 10 or more metastases and estimated it to be equivalent to 8.25 Gy of fractionated radiation, a dose they concluded was unlikely to cause late neurotoxicity.42 These findings correspond with the observation that a SRS-only approach for metastases avoids the neurocognitive toxicity of WBRT.41 Our study has several limitations. It is retrospective and lacks neurocognitive and quality-of-life correlations. However, as a result of these findings, an assessment of leukoencephalopathy is being incorporated into a multicenter, prospective trial to be conducted by the North American Gamma Knife Consortium comparing WBRT plus SRS versus SRS alone for the treatment of brain metastases. Primary outcomes of this study will include neurocognitive and quality-of-life assessments. Another limitation, selection bias, may limit the generalizability of our findings. Long-term survivors of metastatic non–small cell lung cancer to the brain remain the minority. It will be important to study patients with different cancer histologies (ie, breast cancer) where long-term survivors make up a larger proportion of patients. Qualitative imaging assessments lack sensitivity when compared to volumetric analyses, resulting in potential underestimation of white matter change in the SRS-only group. Despite these limitations, the overwhelming difference between treatment groups suggests the difference is true. This work adds to mounting evidence demonstrating the differential effects of radiation therapies on normal brain structure and function. Improved understanding of these effects will allow clinicians to maximize the effectiveness of their treatments while minimizing toxicities.


No specific funding was disclosed.


Drs. Lunsford and Kondziolka are consultants with AB Elekta. Dr. Lunsford is a stockholder in AB Elekta.