Multidisciplinary management of colorectal brain metastases

A retrospective study




The incidence of brain metastases (BM) from colorectal cancer (CRC) is increasing, and the management of this previously rare complication at a single institution is reported.


The records of all patients with BM from 1994 to 2005 were reviewed, and 49 patients (33 men, 16 women) with 102 BM from CRC were identified. Associations between patient and tumor characteristics, treatment modality, and survival were assessed.


The median age at diagnosis of BM from CRC was 66 years. Forty patients (82%) had other systemic disease. The median survival after a diagnosis of BM from CRC was 5.1 months. Fifteen patients (31%) underwent surgery at some point, 14 patients (29%) underwent stereotactic radiosurgery (SRS), and 42 patients (86%) received whole-brain radiotherapy during their management. Seven patients (14%) underwent upfront SRS. On multivariate analysis, a longer interval from diagnosis of CRC to diagnosis of BM was associated significantly with shorter survival (p = .01). Sex, Karnofsky performance status, tumor location, recursive partitioning analysis class, and initial treatment modality did not have an impact on survival.


Because BM from CRC are a late-stage phenomenon, the majority of patients in the current study had other systemic involvement, and survival after CNS involvement was poor. The results indicated that a high prevalence of systemic disease limits the proportion of patients who are strong candidates for upfront SRS, thereby limiting the impact that this modality has on outcomes in this population as a whole. Late development (>1 year after the primary tumor diagnosis) of CNS involvement may predict for poorer survival after therapy for patients with BM from CRC. Cancer 2008. © 2008 American Cancer Society.

Each year in the United States, colorectal cancer (CRC) accounts for approximately 150,000 new diagnoses and 55,000 deaths, making it the second-leading cause of cancer-related deaths after lung cancer.1 Although brain metastases (BM) are an important cause of morbidity and mortality and may affect 10% to 30% of all adult patients with cancer,2 BM from CRC are less common and generally are a late-stage phenomenon. Although CRC accounts for 10.6% of all primary cancer diagnoses and 9.8% of all cancer deaths,1 historic series have demonstrated that only 3% to 8% of all BM are caused by gastrointestinal cancers.3–9 This disparity is because of the finding that only 0.3% to 9% of patients with CRC develop BM.7, 8, 10–14 In contrast, >25% of patients with lung cancer develop BM.15, 16

Unfortunately, patients with BM from CRC do not survive as long as patients with BM from other histologies.6, 17, 18 This probably is secondary to the burden of systemic disease in patients with metastatic CRC to the brain; concomitant extracerebral metastases have been demonstrated in 77% to 97% of patients.10, 11, 17, 19 However, as improvements in the management of CRC have altered the natural history of this disease, the incidence of metastases at previously uncommon sites, such as the brain and bone, is increasing.20 Therefore, careful consideration needs to be given to the management of this previously rare complication.

In the current study, we evaluated a group of patients with BM from CRC who underwent modern, multimodality therapies, including stereotactic radiosurgery (SRS). Although previous studies17, 19 reported large cohorts of patients with BM from CRC, those reports analyzed patients who were treated before the incorporation of SRS for BM. Furthermore, in recent studies of SRS for BM from CRC,21, 22 selected patients were deemed appropriate to receive SRS only. To our knowledge, this is the first large, unselected group of patients with BM from CRC reported since the use of SRS has become commonplace. We evaluated survival, treatments, and prognostic factors that affected survival in this group of patients and assessed the impact that the addition of SRS has had on the management and outcomes of these patients.


Between 1994 and 2005, all patients with BM from CRC were identified from the comprehensive Cleveland Clinic BM database, which is an Institutional Review Board-approved registry of more than 1200 patients with BM. Patients recorded in this database either were diagnosed and treated for primary and secondary tumors at our institution and at outside hospitals or were referred to the Cleveland Clinic for treatment of BM. The relevant patient records were reviewed retrospectively with respect to the following variables: age, sex, site of primary tumor, date of primary CRC diagnosis, and date of BM diagnosis. Computed tomography (CT) or magnetic resonance imaging (MRI) studies confirmed the presence of BM in every patient.

Prognostic Factors

The interval between diagnosis of the primary CRC tumor and diagnosis of BM was calculated. The following variables also were reviewed: the presence and location of other systemic metastases, whether the primary tumor was controlled or uncontrolled, Karnofsky performance status (KPS), and neurologic deficits/symptoms present at the time of BM diagnosis. By using the CT/MRI findings, the numbers of BM were noted along with the location (supratentorial vs infratentorial) and the maximum tumor dimension.

Patients also were assigned a Radiation Therapy Oncology Group (RTOG) recursive partitioning analysis (RPA) classification23 at the time of their BM diagnosis. This classification uses 4 criteria to stratify patients: 1) KPS, 2) primary tumor status, 3) the presence of extracranial disease, and 4) age. Patients in RPA Class I have a KPS ≥70, age <65 years, controlled primary tumor, and no extracranial metastases. Patients in RPA Class III have a KPS <70. All other patients are in RPA Class II. In addition, the management of each patient was reviewed, including the treatments and the order in which treatments were used.

Statistical Analysis

The survival from original CRC diagnosis and the survival from the time of central nervous system (CNS) involvement were calculated, as measured from the date of CT/MRI diagnosis. Survival distributions were calculated by using the Kaplan-Meier method for the entire cohort and with respect to possible prognostic factors. Associations between patient and tumor characteristics and treatment modality were assessed by using the chi-square test and the Kruskal-Wallis test. The log-rank test was used to assess the impact on survival of individual factors. The Cox proportional-hazards model with stepwise variable selection and a significance level of p = .10 for determining the entry and retention of predictors was used to assess the impact of multiple factors. All factors summarized in Table 1 were included in this analysis. All statistical tests were 2-tailed, and p values were considered statistically significant when <.05.

Table 1. Characteristics of 49 Patients With Brain Metastases from Colorectal Cancer
CharacteristicNo. (%) CharacteristicNo. (%)
  • KPS indicates Karnofsky performance status; RPA, recursive partitioning analysis.

  • *

    Of 111 tumors, only 102 have either a size or a location denoted, because 9 small tumors in patients with multiple metastases are not described.

KPS  No. of brain metastases/patient* 
503 (6) 123 (47)
607 (14) 29 (18)
709 (18) 39 (18)
8018 (37) 44 (8)
9012 (24) 52 (4)
   91 (2)
RPA class  91 (2)
I1 (2)   
II38 (77)   
III10 (20) Location of brain metastases 
   Supratentorial65 (64)
Neurologic deficits/symptoms  Infratentorial36 (35)
Ataxia15 (31) Missing1 (1)
Motor14 (29)   
Mental6 (12)   
Seizures6 (12) Greatest tumor dimension, cm 
Aphasia5 (10) No.72
Visual4 (8) Missing30
Weight Loss3 (6) Median2
Memory2 (4) Range0.5-6.0
Sensory2 (4)   
Metastatic burden  Metastatic burden 
Brain/systemic  Systemic site 
Single/absent6 (12) None9 (18)
Single/present17 (35) Lung23 (47)
Multiple/absent3 (6) Liver16 (33)
Multiple/present23 (47) Bone6 (12)
   Other17 (35)


Patient Characteristics

Forty-nine patients were identified, including 33 men (67%) and 16 women (33%). The median age at diagnosis of CNS involvement was 66 years (range, 36-85 years). In total, 102 lesions were identified, characterized, and confirmed (Table 2). Ninety-six of those lesions were present at presentation; 6 patients had a second lesion diagnosed and treated from 3 months to 12 months after the initial diagnosis of CNS involvement. Two patients had their BM diagnosed before the CRC primary tumor was identified (2 weeks and 5 months prior). For the other 47 patients, the median interval between CRC diagnosis and BM diagnosis was 23.2 months (range, 0-116 months). Primary tumors arose from the colon in 42 patients and from the rectum in 7 patients. Primary tumor histology was adenocarcinoma in all 43 patients for whom this information was obtainable; for 6 patients, this information was unavailable. American Joint Committee on Cancer staging information from the time of initial CRC diagnosis was available in 29 patients: 2 patients (7%) had stage I disease, 4 patients (14%) had stage II disease, 12 patients (41%) had stage III disease, and 11 patients (38%) had stage IV disease.

Table 2. Prevalence of Brain Metastases in Patients Undergoing Stereotactic Radiosurgery for Colorectal Cancer Brain Metastases
No. of lesionsNo. of patients (%)
17 (50)
22 (14)
34 (29)
51 (7)

At the time of BM diagnosis, 30 patients (61%) had a KPS >70 (Table 2). The most common neurologic symptoms or signs were ataxia and motor weakness. Most metastases were supratentorial (64%). The primary tumor was controlled in 41 of 49 patients (84%) at the time of CNS involvement. Forty patients (82%) had other systemic disease; the lung and liver were involved most often (47% and 33%, respectively). Twenty-two patients (45%) had only 1 BM. Only 1 patient was in RTOG RPA Class I. Ten patients (20%) were in RPA Class III, and 38 patients (77%) were in RPA Class II.


Fifteen patients (31%) underwent surgical resection other than biopsy during their management. Eleven patients underwent gross total resections, and 3 patients underwent subtotal resections; in 1 patient, the extent of resection was unknown. In 2 of the 15 patients who underwent resection, the procedure was salvage surgery after whole-brain radiotherapy (WBRT) (n = 1 patient) or SRS (n = 1 patient).

Fourteen patients (29%) underwent SRS during their treatment course, and most of those patients had 1 lesion treated (Table 3). Seven patients underwent SRS as their primary treatment for BM, and 7 patients underwent SRS as salvage therapy (Table 4). The prescription dose ranged from 1200 centigrays (cGy) to 2400 cGy, and the median dose was 2010 cGy. The dose was based on the maximum tumor dimension according to the RTOG 9005 guidelines, although 2 patients received lower doses because of the proximity of critical structures. The median tumor volume treated per lesion was 1.3 cc (range, 0.03-13.1 cc). SRS was linear accelerator (LINAC)-based before 1996; after which, the gamma knife (Elekta, Norcross, Ga) was used.

Table 3. Treatment Modalities Employed and Sequences in 49 Patients Treated for Colorectal Brain Metastasis
Sequence of treatmentsNo. of patients (%)
  1. WBRT indicates whole-brain radiotherapy; SRS, stereotactic radiosurgery.

WBRT only23 (47)
SRS only4 (8)
Surgery only2 (4)
Surgery→WBRT7 (14)
WBRT→SRS5 (10)
WBRT→surgery3 (6)
SRS→surgery1 (2)
Surgery→WBRT→SRS1 (2)
WBRT→surgery→SRS1 (2)
Table 4. Relation Between Patient Characteristics and Management Decisions for the Treatment of Colorectal Cancer Brain Metastases
Patient characteristicFirst treatment
Surgery (n=10)SRS (n=7)WBRT (n=32)
  • SRS indicates stereotactic radiosurgery; WBRT, whole-brain radiotherapy; KPS, Karnofsky performance status; CRC, colorectal cancer; CNS, central nervous system; RPA, recursive partitioning analysis.

  • *

    Kruskal-Wallis test.

  • Chi-square test.

No. of metastases   
   p = .02*
   p = .07
Interval from CRC diagnosis to CNS involvement, y   
   p = .09
RPA class   
   p = .04
Extracranial metastases   
   p = .07

Forty-two patients (86%) received WBRT during their management. In 32 of those patients, WBRT was the primary treatment; and, in 10 of those patients, WBRT was used as salvage therapy (Table 4). In 23 patients (47%), WBRT was the sole treatment. Patients generally received 3000 cGy in 10 fractions (n = 22 patients) or 3750 cGy in 15 fractions (n = 10 patients); however, the dose ranged from 3000 cGy to 6000 cGy given in 10 to 30 fractions. Five patients received boosts of 600 cGy (n = 2), 1000 cGy (n = 2), or 1250 cGy (n = 1). Overall, the median total dose delivered (including boosts) was 3250 cGy (range, 3000-6000 cGy).

Patient Characteristics Dictate Treatment Decisions

Before assessing the impact of different treatments on survival in patients who have BM from CRC, we analyzed the correlations between patient characteristics and the treatment modalities used to manage the patients' BM. All patients who had >3 metastases or who were in RPA Class III received WBRT as initial therapy, as indicated in Table 1. In addition, patients with a KPS <70, extracranial metastases, and a longer interval between primary diagnosis and CNS involvement tended to receive WBRT as their initial treatment, although these were not statistically significant findings. Age, sex, and location of metastases did not significantly affect the initial course of treatment received by patients.

Outcomes and Prognostic Factors

The median survival after the diagnosis of BM from CRC for all 49 patients was 5.1 months, with a 40% 6-month survival rate and a 13% 12-month survival rate (Table 5). Survival from the date of primary CRC diagnosis was calculated for 47 of the 49 patients, because 2 patients had their primary CRC diagnosed after their BM. This demonstrated a median survival after primary diagnosis of 28.4 months, with a 98% 6-month survival rate and an 89% 12-month survival rate. Radiographic follow-up data were available for 60 lesions from 33 patients: Ten of those lesions (17%) resolved completely, 13 lesions (22%) had partial responses, 5 lesions (8%) had minor responses, 11 lesions (18%) were stable, and 21 lesions (35%) showed progression. Unfortunately, causes of death were largely unknown. Of 10 patients with documentation, 9 patients (90%) died from systemic disease, and 1 patient died from CNS progression.

Table 5. Analyses of Prognostic Factors in Colorectal Cancer Brain Metastases*
FactorNo. of patients% 6-month survival% 12-month survivalMedian, mop
  • KPS indicates Karnofsky performance status; RPA, recursive partitioning analysis; SRS, stereotactic radiosurgery; WBRT, whole-brain radiotherapy.

  • *

    Survival distributions were calculated by using the Kaplan-Meier method for the entire cohort and with respect to possible prognostic factors.

  • Log-rank test.

  • Interval from colorectal cancer diagnosis to brain metastases diagnosis.

  • §

    P = .14 for infratentorial versus supratentorial.

Sex    .63
Age, y    .04
KPS    .31
Interval, y    .01
Tumor location    .28§
No. of metastases    .05
Extracranial metastases    .41
RPA class    .87
First treatment    .96

Table 5 summarizes survival for patient groups divided by each potential prognostic factor. When each factor was considered individually, age ≥65 years and an interval of ≥1 year between the diagnosis of CRC and the diagnosis of BM both appeared to have a negative effect on survival. In addition, the number of metastatic sites appeared to be associated with survival; patients who had >3 metastases had better survival than patients who had from 1 to 3 metastases. It is noteworthy that all of those patients received WBRT as their initial treatment. In multivariate analysis, only the interval between primary CRC diagnosis and BM diagnosis was an independent predictor of survival (Fig. 1).

Figure 1.

Survival after central nervous system involvement stratified by interval from diagnosis of colorectal cancer to diagnosis of brain metastases. Survival distributions were determined by using the Kaplan-Meier method.


Patients with BM from CRC have a poor prognosis, because they often have substantial extracranial metastatic disease. Traditionally, the therapeutic goal in many of these patients has been to palliate debilitating neurologic symptoms, because most of these patients die of systemic disease. However, new advances in metastatic CRC management—including the incorporation of monoclonal antibody therapies bevacizumab,24 cetuximab,25 and panitumumab26—are enhancing the outcomes of patients with systemic disease. Although these targeted therapies have improved management of systemic disease, they are not as effective for BM given the restrictions on delivery into a tumor caused by the blood-tumor and blood-brain barriers. In light of these advances, the management of BM from CRC may deserve reconsideration.

The current study characterizes the management and outcomes in a large group of these patients undergoing multimodality treatment and attempts to identify prognostic factors to guide therapy. The median survival for all 49 patients was 5.1 months (range, 1-18 months), which is consistent with previous reports of patients with BM from CRC in which the median survival ranged from 2.8 months to 6 months.11, 17, 21, 22, 27, 28 Analysis of numerous prognostic factors in the current study identified age <65 years, shorter interval from CRC diagnosis to BM diagnosis, and >3 metastases as positive prognostic factors when they were considered as individual variables. However, 8 patients had >3 metastases, which limited our ability to reach a firm conclusion. In multivariate analysis, only the interval between primary CRC diagnosis and BM diagnosis was an independent predictor of survival. This likely reflects the finding that most patients with BM from CRC have extracranial disease. Those patients who develop BM further out from the primary diagnosis likely have a higher burden of systemic disease that has been treated with numerous chemotherapeutic regimens; the tumor also may have become resistant to therapy with continued treatment. Indeed, patients with extracranial disease tended to receive less aggressive treatment (Table 1) and had a median survival of 4.1 months, whereas patients without extracranial metastases had a median survival of 6.1 months (P = .41) (Table 5). The number of patients without extracranial disease was small, however (n = 9), and the lack of a statistically significant difference in survival should be interpreted cautiously. A more rigorous determination of systemic versus neurologic causes of death was not possible in this study.

The identification of prognostic factors to guide the management of BM and future clinical trial designs is an area of importance. In 1997, the RTOG published an RPA that allows for the prognostic classification of patients with BM.23 Although that system was based largely on data in BM from nonsmall cell lung cancer, the findings have been validated in BM from small cell lung cancer29 and BM from renal cell cancer.30 To date, only 1 published study has examined this RPA classification as a prognostic factor in BM from CRC. Amichetti et al. reviewed 23 patients who had BM treated with WBRT between 1999 and 2004.28 Nine patients were in RPA Class II, and 14 patients were in RPA Class III; no patients were in RPA Class I. Those authors observed no significant difference in outcomes between Class II and Class III. The current study mirrored those findings: Only 1 patient was in RPA Class I, whereas 38 patients were in RPA Class II, and 10 patients were in RPA Class III. No difference in survival was demonstrated between patients in Class III versus patients in Class I/II. Given the relatively small number of patients with BM from CRC, it is difficult draw firm conclusions regarding the utility of the RPA classification system for prognosis in this setting. However, taken together, these studies suggest that few patients with BM from CRC fall into the RPA Class I category. This likely reflects the differences in metastatic patterns between CRC and other histologies. BM often are the sole site of failure in nonsmall cell lung cancer, whereas systemic disease usually precedes BM in CRC, because tumor cells generally must pass through both the liver and the lung before reaching the brain hematogenously.10, 31 Therefore, the presence of extracranial disease precludes patients from an RPA Class I classification.

Current treatment recommendations for BM from CRC do not differ from those of other histologies. Treatment options include surgery with or without WBRT, SRS (LINAC-based or gamma knife), WBRT alone, or steroid therapy. WBRT is the most common therapy used for BM, especially in patients with multiple lesions. Studies of WBRT alone in patients with CRC have demonstrated poor outcomes, with survival ranging from 2.2 months to 4 months.5, 10, 17, 27, 28 In the current study, all patients with >3 metastases and all patients in RPA Class III received WBRT as their primary treatment. Despite these adverse characteristics, patients who received upfront WBRT demonstrated a median survival (5.3 months) that was similar to that of patients who underwent surgery or SRS as their primary treatment. Nine of these WBRT patients underwent salvage surgery or SRS.

Surgical intervention for BM from CRC currently is recommended for symptomatic and/or single, surgically accessible lesions in relatively healthy patients. Indeed, numerous reports have indicated that selection for surgical resection is a positive prognostic factor,5, 10, 19, 27 although the improvements in survival likely relate to a low tumor burden and the general health of the patients who are selected for surgical resection. For example, Farnell et al.17 reported that 45 patients who had a solitary BM from CRC who underwent surgical resection had a median survival of 45 weeks compared with 17 weeks for patients who had a single lesion and received WBRT alone. However, the patients who underwent surgical resection also had a lower prevalence of extracranial disease (69% vs 89%).

In the current study, 10 patients underwent surgical resection as their primary treatment. Compared with patients who received WBRT as a primary modality, those who underwent surgery had fewer metastases (p = .02), tended to have a better RPA classification (p = .08), and had a shorter interval from initial CRC diagnosis to CNS involvement (p = .13) (Table 1). It is noteworthy, however, that these patients did not have a significantly better median survival (5.2 months) than the patients who received upfront WBRT (5.3 months). This is especially surprising in light of the finding that 8 of these 10 patients received additional WBRT (7 of 8 patients) or SRS (1 of 8 patients) after surgery, whereas 23 of the 32 patients who received WBRT got no further therapy. Although surgery clearly plays a role in BM from CRC causing a significant mass effect, these data suggest that technologic advances in radiation oncology, including the availability of SRS salvage after WBRT, may enhance outcomes in patients who have multiple tumors and/or surgically inaccessible lesions.

The role of SRS as a primary modality for treatment of BM from CRC largely remains undefined. Two recent reports have retrospectively assessed the efficacy of SRS on BM from CRC. Schoeggl et al.22 reported the outcomes of 35 patients with CRC who underwent SRS for their BM; 13 of those 35 patients also received WBRT. They demonstrated a local control rate of 94%. Although the median survival of 6 months after SRS was disappointing, the treatment-related morbidity was minimal, and those authors concluded that SRS provides patients with a higher quality for their remaining life.

Another report21 focused on SRS for BM from gastrointestinal tract tumors. In that study, 39 patients were included: 25 patients with CRC, 11 patients with esophageal cancer, 1 patient with duodenal cancer, 1 patient with jejunal cancer, and 1 patient with cholangiocarcinoma. This patient group had less systemic disease burden than typical cohorts of CRC patients with BM; only 44% of patients had extracranial metastatic disease, which likely reflects a selection bias. Eighty percent of these patients also received WBRT, and 15% underwent surgical resection. The median survival after SRS was 5 months for the group as a whole and for the 25 patients with CRC. This median survival was less impressive than that achieved by the group who underwent SRS for BM from other primary cancers (lung, 10 months; renal cell, 11 months; breast, 13 months; melanoma, 7 months).32–35 The local control rate was 84% and there were few side effects. However, the patients in that study who had active extracranial disease had a median survival of only 2 months after SRS. Therefore, the authors suggested that SRS should not be recommended routinely in patients who have simultaneous intracranial and extracranial metastatic CRC unless management of the systemic disease is feasible.

We were interested in examining whether the incorporation of SRS into the multimodality treatment of BM from CRC resulted in better outcomes in this group of patients as a whole compared with historic series that used surgery and/or WBRT. The median survival in our series after BM diagnosis was 5.1 months for all patients. This is consistent with previous reports on patients with BM from CRC who were treated before the use of SRS was commonplace. Farnell et al.17 reported on a group of 150 patients with CRC metastatic to the brain. The majority of those patients (82%) had concomitant extracerebral metastases, and an overall median survival of 4.8 months was noted. None of those patients underwent SRS. Another report5 that reviewed 100 patients demonstrated a median survival of 5 months. Again, no patients underwent SRS.

Our patient population was similar to the populations in the reports described above and had a very similar median survival. In fact, only 14% of the patients in the current study underwent SRS as their initial treatment modality; because most patients (82%) had extracranial systemic disease, only 1 patient was in RPA Class I at the time of BM diagnosis. It is noteworthy that, in the current study, patients who were in RPA Class III, who had a lower KPS, who had extracranial disease, and who had more metastases were more likely to receive WBRT than SRS as an initial treatment (Table 1). Despite these adverse characteristics (although they were not prognostic in the current study), the median survival for this group of patients was not significantly lower than the median survival for those patients who received SRS as their primary modality. It must be emphasized that these data do not assess the efficacy of SRS in BM from CRC in any prospective manner, and the number of patients was limited. In individual patients with low-volume, well controlled, systemic disease, SRS likely will have an impact on local control and survival and should be considered. In addition, SRS may be useful for patients who have recurrent BM from CRC after either primary surgical resection or WBRT. In fact, the 7 patients in the current study who underwent SRS as a salvage procedure had a median survival of 4.5 months (range, 1.3-12.4 months) after the salvage procedure, demonstrating that SRS may have value for patients who fail after initial therapy. However, taken as a whole, these data suggest that the systemic burden of CRC in patients with BM from CRC results in a minority of patients who are strong candidates for upfront SRS; therefore, this technology largely has not had an impact on the outcomes of this patient population.

To our knowledge this is the first large, unselected consecutive group of patients with BM from CRC reported since the use of SRS has become commonplace. Despite the availability of this modality, the median survival (5.1 months) of this patient cohort was similar to that reported in historic series using surgery and/or WBRT. Because BM from CRC are a late-stage phenomenon, the majority of patients (82%) had other systemic involvement. The high prevalence of systemic disease in this patient population limits the proportion of patients who are deemed strong candidates for upfront SRS. However, upfront SRS may be useful in carefully selected patients, and SRS may play a role in the treatment of recurrent lesions. In multivariate analysis, an interval of >1 year from initial diagnosis to CNS involvement was identified as a negative prognostic factor in patients undergoing treatment for BM from CRC.