Brain Metastases from Colorectal Cancer: A Systematic Review of the Literature and Meta-Analysis to Establish a Guideline for Daily Treatment

Simple Summary Brain metastases (BM) from colorectal cancer (CRC) are rare. There is little available information regarding incidence, risk factors, prognostic factors, treatment, and overall survival (OS). In this systematic review we performed a research of the current literature and exposed an average incidence of 2.10%. The most-reported risk factors for developing BM were KRAS mutations and lung metastases. The majority of patients with brain metastases did not show neurological symptoms. Treatment options included surgery, radiation, or chemotherapy. While patients who received surgery had prolonged survival, the best survival time was found with a multimodality treatment regimen including neurosurgery. Abstract Colorectal cancer (CRC) is the third most common malignancy worldwide. Most patients with metastatic CRC develop liver or lung metastases, while a minority suffer from brain metastases. There is little information available regarding the presentation, treatment, and overall survival of brain metastases (BM) from CRC. This systematic review and meta-analysis includes data collected from three major databases (PubMed, Cochrane, and Embase) based on the key words “brain”, “metastas*”, “tumor”, “colorectal”, “cancer”, and “malignancy”. In total, 1318 articles were identified in the search and 86 studies matched the inclusion criteria. The incidence of BM varied between 0.1% and 11.5%. Most patients developed metastases at other sites prior to developing BM. Lung metastases and KRAS mutations were described as risk factors for additional BM. Patients with BM suffered from various symptoms, but up to 96.8% of BM patients were asymptomatic at the time of BM diagnosis. Median survival time ranged from 2 to 9.6 months, and overall survival (OS) increased up to 41.1 months in patients on a multimodal therapy regimen. Several factors including age, blood levels of carcinoembryonic antigen (CEA), multiple metastases sites, number of brain lesions, and presence of the KRAS mutation were predictors of OS. For BM diagnosis, MRI was considered to be state of the art. Treatment consisted of a combination of surgery, radiation, or systemic treatment.


Introduction
Colorectal cancer (CRC) is the third most common type of malignant tumor worldwide, and in 2018, 880,792 deaths were reported due to CRC worldwide [1]. The incidence of CRC increases in an age-dependent manner, with the average age being 72-76 years at

Results
The database search identified 2018 articles. After removing duplicates, 1318 articles were left for further investigation. After screening by title and abstract for suitability, 328 manuscripts were left. Articles were read in full text to check for inclusion criteria. Eighty-six papers matched the inclusion criteria and were used to perform the metaanalysis. None of these were randomized controlled trials (RCTs), as they mostly involved retrospective analyses. Articles were grouped in different categories to perform the metaanalysis (Table 1). Table 1. Categories and numbers of articles that matched the regarding category (numerous articles matched more than one category). OS: overall survival; BM: bone metastases.

Category
Articles Found

Diagnosis 4
Risk factors for developing BM 17 Overall survival 43 Factors for poor OS 25 Treatment modalities 18

Symptoms
Six studies described symptoms in patients with BM from CRC. The initial symptoms of brain metastases were highly variable and were mostly not described in further detail. The most commonly reported symptoms were epileptic seizures, signs of increased intracranial pressure, or neurological symptoms [30][31][32][33][34][35]. Some patients did not show any symptoms at the time of diagnosis (Table 3). While Berghoff et al. found a ratio of 96.8% of asymptomatic patients, Kim D. et al. described only 5.3% of patients as being without symptoms [31,34]. Shindorf et al. performed a study that screened patients with metastatic CRC for BM, regardless of whether neurological symptoms were present. They showed that 76% of the patients with BM were asymptomatic [28].

Diagnostic Techniques
Possible imaging modalities for diagnosis of BM are CT, MRI, or PET-CT. FDG-PET-CT is commonly performed as a whole-body examination, in which BM can appear as an incidental finding. Screening for BM is usually performed with cranial MRI [36]. We identified two studies that compared whole-body PET-CT to whole-body MRI to detect metastases from CRC. In these studies, PET-CT was superior for identifying lymph node metastases, for example locoregional to the primary tumor, whereas MRI was superior for detecting lesions of <1cm, especially BM [37,38]. As even smaller lesions and a meningeal carcinomatosis can be missed by imaging, diagnostic spinal fluid examination is proposed to identify tumor cells or DNA [39,40] (see Figure 3).

Prognostic Factors
Nineteen studies depicted risk factors for developing BM; these studies are listed in Table

Lung Metastases
Eleven studies evaluated LM as a risk factor for developing brain metastases and reported a positive correlation (Table 5). In three of these studies only the abstracts were available but no full-text manuscripts. Eight studies were available in full text and are summarized in Figure 4. Altogether, 691 patients were examined. Of these patients, 400 had LM at diagnosis of BM, leaving 291 without LM at diagnosis of BM. The odds ratio was 1.81 (95%CI 1.47-2.22). Furthermore, a statistically significant difference was seen between the two groups (p <0.00001). The high heterogeneity may be caused by the small study populations ( Figure 4).

KRAS Mutation
In five studies KRAS mutation was investigated as a risk factor for developing BM (Table 6). In one of them only the abstract was available. Five studies which reported on the KRAS mutation status of BM patients are shown in Figure 5. A total of 166 patients had a KRAS mutation analysis. In total, 114 patients with BM had a KRAS mutation (68%) and 52 patients had a KRAS wild-type (32%). The odds ratio was 4.47 (95%CI 2.83-7.05). The overall effect showed a significant difference (p <0.00001). The high heterogeneity may be caused by the small study populations ( Figure 5).

Survival
In 43 studies an overall survival (OS) with a range from 2 to 9.6 month from the time of BM diagnosis was determined. The median OS was 5.3 months (95%CI 4.6-5.9). In total there were 3611 patients with BM that were included in the OS analysis. The smallest study population reported on five patients, while the largest study included 475 patients with BM. The analyzed studies are shown in Figure 6 and Table 7 [9- 12,14,15,18,20,[23][24][25][26][27]29,34,45,46,.

Survival
In 43 studies an overall survival (OS) with a range from 2 to 9.6 month from the time of BM diagnosis was determined. The median OS was 5.3 months (95%CI 4.6-5.9). In total there were 3611 patients with BM that were included in the OS analysis. The smallest study population reported on five patients, while the largest study included 475 patients with BM. The analyzed studies are shown in Figure 6 and Table 7 [9- 12,14,15,18,20,[23][24][25][26][27]29,34,45,46,. Table 7. Study characteristics and OS. Studies reported OS in months, the number of patients with CRC, and the number of patients with BM from CRC [9- 12,14,18,20,[23][24][25][26][27]29,34,45,46,. *only the abstract was available.  Table 7. Cont. Twenty-five studies investigated factors of poor OS in patients with BM ( Table 8). The most common factors were advanced age, low Karnofsky performance status (KPS), and extracranial metastases, as well as multiple BM. Four studies described a significant reduction in OS in patients with advanced age. Two studies did not report the exact OS.  [7,9,12,27,45,46,52,55,62,64,65,70]. Eleven studies reported multiple BM as a risk factor for poor OS. Four manuscripts did not provide the exact OS. Four studies compared one BM with more than one BM, those of Lu et al.

Treatment
Altogether 18 studies evaluated different treatment modalities (Table 9). They investigated the influence of the treatment on the OS.

Discussion
In this study we reviewed the current literature to analyze incidence, risk factors, treatment strategies, and overall survival in patients with BM from CRC.
Our systematic review confirmed that BM are rare in colorectal cancer patients. The incidence in the included studies ranged from 0.1% up to 11.5%. Zullkowski et al. described an incidence of 11.5% in their study population, which differed greatly from the other studies. This divergence might be due to their patient selection and the small study collective. In patients with metastatic disease, one study reported a BM incidence of 14.6% [28]. This matches our results, which showed an association of BM with extracranial metastases. The wide range of values for reported incidence might be due to the large number of asymptomatic BM patients. A lack of symptoms like nausea, vomiting, headaches, or reduced vision can lead to a late diagnosis.
Therefore, in studies with restricted cranial imaging brain metastases may be undetected, whereas in studies that perform cranial imaging more generously BM might be detected earlier.
Accordingly, studies that evaluated a screening program for BM described 96% of BM patients as being asymptomatic [28,31].
Considering these findings, we propose a systematic screening program for CRC patients ( Figure 9). Performing cranial imaging on every patient with CRC would not only lead to great number of physiological MRIs but would also be a financial burden for the health system. Therefore, we recommend cranial imaging in patients with symptoms or if risk factors are present (KRAS mutation, pulmonary metastases, rectal cancer, or positive CEA level). Our screening strategy is shown in Figure 9.
For other malignant diseases that more frequently lead to BM like breast cancer, studies proved higher survival rates and better treatment options by detecting BM early. Cagney et al. recommend screening for BM in patients with metastatic breast cancer [94] and Komorowski et al. reported that asymptomatic patients with metastatic breast cancer and HER2-overexpression profited from BM screening [95]. Morikawa et al. proved in their analysis that early detection of asymptomatic BM from breast cancer was associated with higher survival rates [96].
In patients with non-small-cell lung cancer the ESMO (EUROPEAN SOCIETY FOR MEDICAL ONCOLOGY) guidelines recommend brain imaging to screen for BM [97]. In comparison, the ESMO guidelines for metastatic CRC do not provide a recommendation regarding screening for BM. The ESMO guidelines for rectal cancer recommend cranial imaging in symptomatic patients [5,98]. Knowing that early diagnosis of BM in CRC leads to better survival rates, a screening program in patients with more than one In patients treated with best supportive care the OS was the lowest, at 0.43-2 months [26,57,64].
If surgery was not possible, a procedure for local control such as stereotactic radiosurgery or gamma-knife radiosurgery provided better OS in patients with 1-3 metastases [86][87][88][89][90]. With these procedures, local control of BM was possible in up to 95% of patients [90].
Finkelmeier et al. and Berghoff et al. reported that a combination of chemotherapy or radiation with bevacizumab prolonged survival rates and reduced neurological symptoms [91,92]. A recently published study by Amin et al. showed that immunotherapy in combination with radiation led to a longer survival of 34%, but no further information about the type of immunotherapy was provided [93].

Discussion
In this study we reviewed the current literature to analyze incidence, risk factors, treatment strategies, and overall survival in patients with BM from CRC.
Our systematic review confirmed that BM are rare in colorectal cancer patients. The incidence in the included studies ranged from 0.1% up to 11.5%. Zullkowski et al. described an incidence of 11.5% in their study population, which differed greatly from the other studies. This divergence might be due to their patient selection and the small study collective. In patients with metastatic disease, one study reported a BM incidence of 14.6% [28]. This matches our results, which showed an association of BM with extracranial metastases. The wide range of values for reported incidence might be due to the large number of asymptomatic BM patients. A lack of symptoms like nausea, vomiting, headaches, or reduced vision can lead to a late diagnosis.
Therefore, in studies with restricted cranial imaging brain metastases may be undetected, whereas in studies that perform cranial imaging more generously BM might be detected earlier.
Accordingly, studies that evaluated a screening program for BM described 96% of BM patients as being asymptomatic [28,31].
Considering these findings, we propose a systematic screening program for CRC patients (Figure 9). Performing cranial imaging on every patient with CRC would not only lead to great number of physiological MRIs but would also be a financial burden for the health system. Therefore, we recommend cranial imaging in patients with symptoms or if risk factors are present (KRAS mutation, pulmonary metastases, rectal cancer, or positive CEA level). Our screening strategy is shown in Figure 9. risk factor for developing BM should be evaluated according to our recommended screening strategy (Figure 9). A meta-analysis by Li et al. evaluated the diagnostic criteria for BM in lung cancer patients. Gadolinium-enhanced MRI had a higher sensitivity than 18FDG PET/PET-CT for the diagnosis of BM [99]. Pope et al. described the high sensitivity of cranial MRI in detection of BM independently of the primary tumors, and therefore recommended it as first choice for diagnosis as well as monitoring of therapy response [100]. In line with the results of this meta-analysis we would advise screening for BM from CRC with cranial MRI. For other malignant diseases that more frequently lead to BM like breast cancer, studies proved higher survival rates and better treatment options by detecting BM early. Cagney et al. recommend screening for BM in patients with metastatic breast cancer [94] and Komorowski et al. reported that asymptomatic patients with metastatic breast cancer and HER2-overexpression profited from BM screening [95]. Morikawa et al. proved in their analysis that early detection of asymptomatic BM from breast cancer was associated with higher survival rates [96].
In patients with non-small-cell lung cancer the ESMO (EUROPEAN SOCIETY. FOR MEDICAL ONCOLOGY) guidelines recommend brain imaging to screen for BM [97]. In comparison, the ESMO guidelines for metastatic CRC do not provide a recommendation regarding screening for BM. The ESMO guidelines for rectal cancer recommend cranial imaging in symptomatic patients [5,98]. Knowing that early diagnosis of BM in CRC leads to better survival rates, a screening program in patients with more than one risk factor for developing BM should be evaluated according to our recommended screening strategy (Figure 9).
A meta-analysis by Li et al. evaluated the diagnostic criteria for BM in lung cancer patients. Gadolinium-enhanced MRI had a higher sensitivity than 18FDG PET/PET-CT for the diagnosis of BM [99]. Pope et al. described the high sensitivity of cranial MRI in detection of BM independently of the primary tumors, and therefore recommended it as first choice for diagnosis as well as monitoring of therapy response [100]. In line with the results of this meta-analysis we would advise screening for BM from CRC with cranial MRI.
In case of BM, a number of risk factors affect the OS. In this analysis we found that a positive CEA level, a low KPS, and the presence of extracranial metastases and multiple BM predicted a poor OS. Edwards et al. evaluated the OS of elderly patients with various solid tumors. There was a great association between poor KFS and shorter OS [101]. Hwang et al. described, besides other factors, the influence of low KPS on the OS in metastatic cancer patients [102]. Furthermore, a few studies described an association between CEA level and survival after curative treatment for BM [103][104][105]. CEA may be suggestive of metastatic disease which is associated with poor OS [106][107][108].
The best survival rates were found in patients with no extracranial metastases and a multimodal therapy regimen. If neurosurgical resection is possible, it leads to better OS rates if performed with additional radiation, chemotherapy, or targeted therapy. If neurosurgical resection is not possible, the number of BM is essential for defining the best treatment option. In patients with 1-3 BM, radiosurgery or gamma-knife radiosurgery is recommended, whereas patients with more than three BM should receive whole-brain radiation [86][87][88][89][90]. Our recommended treatment algorithm is shown in Figure 10. In case of BM, a number of risk factors affect the OS. In this analysis we found that a positive CEA level, a low KPS, and the presence of extracranial metastases and multiple BM predicted a poor OS. Edwards et al. evaluated the OS of elderly patients with various solid tumors. There was a great association between poor KFS and shorter OS [101]. Hwang et al. described, besides other factors, the influence of low KPS on the OS in metastatic cancer patients [102]. Furthermore, a few studies described an association between CEA level and survival after curative treatment for BM [103][104][105]. CEA may be suggestive of metastatic disease which is associated with poor OS [106][107][108].
The best survival rates were found in patients with no extracranial metastases and a multimodal therapy regimen. If neurosurgical resection is possible, it leads to better OS rates if performed with additional radiation, chemotherapy, or targeted therapy. If neurosurgical resection is not possible, the number of BM is essential for defining the best treatment option. In patients with 1-3 BM, radiosurgery or gamma-knife radiosurgery is recommended, whereas patients with more than three BM should receive whole-brain radiation [86][87][88][89][90]. Our recommended treatment algorithm is shown in Figure 10. Figure 10. Assessment of therapy algorithm. BSC: best supportive care; SRS: stereotactic radiosurgery; GKRS: gamma knife radiosurgery; WBRT: whole-brain radiation. * Evaluate if neurosurgical resection is reasonable for the oncological therapeutic regime. The indication should be defined by an experienced neurosurgeon considering the size, number, and location of the metastases as well as symptomatology. ° The indication for SRS or GKRS should be considered individually for every patient. The DEGRO (Deutsche Gesellschafts für Radioonkologie) guidelines recommend SRS for a single BM <3 Figure 10. Assessment of therapy algorithm. BSC: best supportive care; SRS: stereotactic radiosurgery; GKRS: gamma knife radiosurgery; WBRT: whole-brain radiation. * Evaluate if neurosurgical resection is reasonable for the oncological therapeutic regime. The indication should be defined by an experienced neurosurgeon considering the size, number, and location of the metastases as well as symptomatology. • The indication for SRS or GKRS should be considered individually for every patient. The DEGRO (Deutsche Gesellschafts für Radioonkologie) guidelines recommend SRS for a single BM <3 cm or 2-4 BM <2.5 cm for patients with life expectancy >3 months [109]. Lee et al. and Yamamoto et al. described how SRS for patients with up to 15 BM dependent on their position and size was associated with survival benefit and reduced risk of neurocognitive deterioration as compared to WBRT [110,111]. This study has some limitations. Not all articles that were suitable by abstract screening were available in full text. We included them anyway in our analysis if adequate data were available in the abstract. Furthermore, all suitable studies were performed retrospectively, which could lead to a publication bias. The majority of studies included a low number of patients, as seen in the study overview in the appendix.
As BM are rare in patients with CRC, a number of studies lasted more than 10 years to reach the calculated study population. As immunotherapy has developed and changed rather quickly over the last decade, treatment modalities and recommendations might have changed during the study duration, which can also pose a risk of bias. The time interval of each included study is shown in the study overview in the appendix.

Conclusions
BM due to CRC represent a rare condition, but if patients develop BM, their overall survival is poor. The vast majority of patients (up to 96%) are asymptomatic, which can lead to late diagnosis. Therefore, we encourage the use of a screening program for patients with risk factors for developing BM. This way, BM can be detected early on and therapy options are superior. A multimodal treatment strategy provides the best OS, and can include surgery with/without radiation, chemotherapy, or targeted therapy. Nevertheless, new studies with a higher number of patients are necessary to obtain valid information about incidence, OS, and the best treatment strategies.