Population-Based analysis for newly diagnosed hepatocellular carcinoma with brain metastases

Background: There is little population-based data on hepatocellular carcinoma (HCC) with brain metastases at initial diagnosis published. This study aimed to estimate incidence of brain metastases in initial metastatic HCC and its impact on prognosis. Methods: Newly diagnosed HCC cases from 2010 to 2015 in the Surveillance, Epidemiology, and End Results (SEER) database were screened for the presence of brain metastases. Data were strati�ed by age and ethnicity. Multivariable logistic and Cox regression were used to identify factors associated with brain metastases and factors associated with overall survival (OS) and cancer-specic survival (CSS), respectively. Kaplan-Meier method and log-rank test were used for survival analysis. Results: 141 cases presenting with brain metastases were identi�ed, accounting for 0.35% of all HCC cases and 2.37% of cases with metastatic HCC disease. The incidence rate was highest among cases with age 50-59 (2.74%), respectively. Ethnicity was not associated with the presence of brain metastases at the time of HCC diagnosis. However, African American patients presented signi�cantly lower disease-specic survival (median time: 1month; interquartile range (IQR):0-3.0 months). Initial lung or bone metastasis was independently associated with an increased risk of the presence of brain metastases (odds ratio (OR) 12.62, 95%CI 8.40-18.97), but not associated with worse OS and CSS among brain metastases cases. Conclusions: The study shows population-based incidence and survival of brain metastases at diagnosis of HCC. Brain metastases are most prevalent in initial metastatic HCC patients with lung or bone metastasis. The results may contribute to consider screening of the brain among HCC with initial lung or bone metastasis.


Background
Hepatocellular carcinoma is the most common form of primary liver cancer (80%), which is the sixth leading cause of cancer death worldwide and the second largest mortality contributor. 1,2Incidence rates continue to increase rapidly for hepatocellular carcinoma, by about 3% per year in women and 4% per year in men. 3 The most common hematogenic pathway of metastases for primary liver tumors is over the hepatic veins and thus to the lungs, bones and following that the brain 4,5 .Brain metastases from hepatocellular carcinoma (HCC) were thought to be very rare but data only based on estimates.Kim et al.   reported in 1998 an incidence of 0.2% of HCC with brain metastases. 6In recent years, it was expected that the incidence of brain metastases might increase over time, probably as a result of advances in early detection and treatment of hepatocellular carcinoma, which have led to a prolonged overall survival. 4,7,8o retrospective studies showed, that the incidence of brain metastases in hepatocellular carcinoma were 4.1% and 7.7%, respectively. 9,10In addition, a meta-analysis including 2538 patients with brain metastases originated from gastrointestinal cancer found that 9.18% of patients had any type of liver cancer. 11Few studies focused on population-based estimates of the incidence and prognosis of hepatocellular carcinoma with synchronous brain metastases at the time of diagnosis.[14][15][16][17] To better de ne the incidence and survival of synchronous brain metastases in newly diagnosed HCC cases the SEER database was reviewed.

Data sources and Study population
The SEER database was used for analysis.SEER database included population-based cancer registries information on cancer incidence, treatment and survival, covering nearly 28% of the United States population. 18Using the latest released data of the SEER database, 45597 cases (≥18 years old) who were diagnosed with hepatocellular carcinoma between January 1, 2010, and December 31, 2015 were identi ed.5824 cases with unavailable information on brain metastases at time of diagnosis were excluded, leaving 39773 cases for analysis of the incidence.Of those, 141 (0.35%) had synchronous brain metastases upon primary diagnosis of HCC.Among the 39773 cases, 79 cases with undisclosed survival information were excluded, leaving 39694 cases for the Cox regression analysis.One case who had an unknown survival time was excluded, leaving 140 cases for the survival analysis (Figure 1).
Variables of interest were gender, age at diagnosis, ethnicity, year of diagnosis, AFP status, primary tumor stage, primary tumor size, tumor multiplicity (presence of multiple lesions), metastatic sites to lung, bone, and surgery of the primary site.Gender was classi ed as male versus female.Ethnicity was classi ed as Caucasian, African American, Other (American Indian/AK Native, Asian/Paci c Islander) and unknown.
Data were strati ed by age at diagnosis.All cases were divided into ve categories: ages 30-49 years, ages 50-59 years, ages 60-69 years, ages 70-79 years, and ages 80+ years.Incidence rates were calculated after strati cation by age at diagnosis.In SEER database, AFP status was divided into three categories: negative, positive, and other (test not done, borderline, and unknown).Further relevant biomarkers were not available in the database, SEER does not provide data on tumor recurrence or subsequent sites of disease involvement.

Statistical analysis
Baseline characteristics for all HCC cases versus those who had synchronous brain metastases were compared with Pearson Chi-square test or Fisher exact test as appropriate.Factors of the presence of brain metastases and predictors associated with OS and CSS were identi ed.Multivariable logistic regression was used to identify whether demographics and tumor characteristics were associated with the presence of brain metastases among all HCC and initial metastasis cases.Gender, age at diagnosis, ethnicity, year of diagnosis, AFP status, primary tumor stage, primary tumor size, tumor multiplicity, and metastatic sites to lung, bone were included in the multivariable logistic regression analysis.Surgery of the primary site was not included in the multivariable logistic regression analysis as the surgery was performed after the diagnosis of brain metastases.Survival analysis was assessed with the Kaplan-Meier method, and the results were compared with the log-rank test.A Cox proportional hazards model was applied for multivariate survival analysis.The forest plot was formulated by Microsoft Excel 2013 according to the results of multivariate analysis.The data were analyzed by SPSS version 20.0 for Windows (SPSS Inc., Chicago, IL, USA).All P values were two-sided, and P-value <0.05 was considered as statistically signi cant.

Clinical and demographic characteristics
Of the 39773 HCC cases 141 (0.35%) had synchronous brain metastases at time of primary HCC diagnosis.Median age of brain metastases cases was 61 years (range, 33-92 years; standard deviation (SD),11.3years), and median age of all HCC cases was 63 years (range, 18-106 years; SD, 10.9 years).Cases with age 18-30 years had no brain metastases.Baseline clinical and demographic characteristics of all HCC and brain metastases cases were signi cantly different between groups except for gender, age at diagnosis, ethnicity, and year of diagnosis (Table 1).Most of cases were male (78.0%) and Caucasian (72.4%).Of all initial brain metastases cases most were aged 50-59.40.4% of cases were tumors >5 cm in size.48.9% of cases were positive in preoperative AFP status, and 14.9% were negative.Initial lung or bone metastases were signi cantly associated with brain metastases (44.7%; all HCC: 8.3%) and this was also the case with synchronous metastases to both lung and bone (12.8%; all HCC: 1.0%).

Incidence
The incidence of HCC with synchronous brain metastases at time of diagnosis strati ed by age is provided in Table 2.The 141 brain metastases cases accounted for 0.35% of all HCC cases and 2.37% of all initial metastatic cases.Although there were only 11 cases with primary brain metastases in ages 30-49, this accounted for an incidence rate of 0.47% in this age group.
On multivariable logistic regression analysis, initial metastasis to one site (vs no initial metastasis to lung or bone; OR 12.62, 95%CI 8.40-18.97,p<0.001), initial metastasis to two sites (vs no initial metastasis to lung or bone; OR 28.97, 95%CI 16.23-51.74,p<0.001) were independent predictors of the presence of brain metastases in all HCC cases (Figure 2).Initial metastasis to two sites (vs no initial metastasis to lung or bone; OR 1.88, 95%CI 1.08-3.30,p=0.03) was independently associated with an increased risk of the presence of brain metastases in all initially metastasized cases.No signi cant differences were observed for gender, ethnicity, and age at diagnosis, year of diagnosis, AFP status, primary tumor stage, primary tumor size, and tumor multiplicity.

Survival
The 1-year OS was 8.4% and the 1-year CSS was 16.3% in cases with synchronous brain metastases, respectively (Figure 3A and 3B).Median survival of all HCC cases was 11.0 months (IQR: 2.0-39.0months).For cases with synchronous brain metastases, the median survival was 2.0 months (IQR: 0-6.0 months; standard error (SE), 0.35 months).Cases with no synchronous brain metastases had a signi cantly better survival compared to cases with synchronous brain metastases (median survival: 12 vs 2 months, p<0.001, Figure 3C).Median survival of brain metastases and all initial metastatic cases strati ed by age at diagnosis were presented in Table 2. Overall survival of synchronous brain metastases strati ed by age at diagnosis was signi cantly worse for the age group 70-79 (p=0.028, Figure 3D).
On multivariate survival analysis (Supplementary table), ethnicity (African American vs Caucasian; HR 1.68; 95%CI, 1.06-2.89;P = 0.03) and the presence of multiple lesions (HR 3.52; 95% CI, 1.28-9.69;P=0.01) were signi cantly associated with the worse OS for HCC cases with synchronous brain metastases.For the liver cancer speci c survival, the results of multivariable cox regression indicated that ethnicity (African American vs Caucasian; HR 2.07; 95%CI, 1.16-3.70;P = 0.01) was still signi cantly associated with the inferior CSS among synchronous brain metastases cases.Gender, age at diagnosis, year of diagnosis, AFP status, primary tumor stage, primary tumor size, and metastatic sites to lung, bone, as well as surgery of the primary site were not associated with the prognosis of OS and CSS.

Discussion
Brain metastases in HCC patients are rare but the prognosis is devastating.This nding is rarely described in the literature 15 , making the need to screen more effectively pressing.In non-symptomatic patients a brain scan is usually not part of the standard work up of newly diagnosed HCC cases.In the present study, the pattern of synchronous brain metastases in newly diagnosed initial metastatic hepatocellular carcinoma was shown to be extremely aggressive when other extrahepatic metastases exist.This nding re-enforces the hypothesis that metastasized tumors tend to further metastasize to other organs.The study revealed that 0.35% of all HCC cases and 2.37% of all initially metastatic HCC cases had brain metastases.The incidence proportion of brain metastases was highest among HCC cases with age 30-49.The median survival of cases with brain metastases was 2.0 months.According to the National Cancer Institute (NCI), the median age for the time of a cancer diagnosis is 66 years and advancing age is the most important risk factor for cancer overall. 19However, concerning primary liver cancer, it cannot be stressed enough that HCC develops almost always in cirrhotic livers with Hepatitis being one of the main underlying reasons.Nowadays NASH is one of the main risk factors for HCC.Despite potent antiviral therapy there is a global increase in HCC.Therefore, a much younger group of patients is at risk to develop a metastatic disease.
Cagney et al 14 were the rst group to perform population-based epidemiologic study on brain metastases in systemic malignancy at time of diagnosis in the United States including cases from 2010 to 2013.They reported that the incidence of brain metastases was 0.36% among all hepatobiliary cancer and 1.77% among all initially metastatic cases, respectively.However, this study did not speci cally examine HCC cases.Han et al 20 retrospectively analyzed 5015 HCC patients who were diagnosed in South Korea between 2001 and 2012 and found that 0.65% of all HCC cases developed brain metastases .However, the incidence and survival were not strati ed by age at diagnosis, and factors associated with the presence of brain metastases were unknown.In the current study, a trend toward higher proportion of younger cases (<50 years) with brain metastases was found.A possible explanation for this nding is attrition bias due to lack of brain imaging in older patients with advanced HCC.When analyzing factors associated with OS and CSS it was revealed that initial metastasis to one site (either lung or bone) as well as initial metastasis to both sites (lung and bone) were independently associated with a higher risk for the presence of brain metastases.Given the lack of routine brain screening in HCC patients with extrahepatic metastases, our nding that 2.37% of all initial metastatic cases had brain metastases at diagnosis likely underestimated the prevalence.Thus, brain involvement should be potentially investigated in all stage IVB patients during staging.
In this study, only ethnicity (African American vs Caucasian) was signi cantly associated with worse OS and CSS in cases with synchronous brain metastases.Compared with African-American patients, Caucasian patients had a signi cantly longer median survival time.This is in line with Jones et al 21 and who found that African American in the US presented at more advanced stages with larger tumors, and had shortest survival after HCC diagnosis.Caucasian patients had 36% reduced mortality rate compared to African American.It is supported by data of population-based studies that found that African American had more advanced stage at diagnosis and lower rates of receiving treatment in the US 22,23 .An exact explanation of the disparity is not clear, but this may associate with underinsurance or lack of any insurance at all. 21e presence of multiple lesions was independently associated with OS but was not associated with CSS on multivariate analysis, a nding which only implies that underlying conditions (e.g.chronic liver disease) could be the reason for this disparity in survival factors.
There are limitations to our study.First, only cases with brain metastases of hepatocellular carcinoma at initial diagnosis were included since SEER does not provide data on tumor recurrence or subsequent sites of disease involvement.An analysis of HCC cases that developed brain metastases during the course of their disease is not possible. 14In addition, National Comprehensive Cancer Network (NCCN) guidelines for hepatocellular carcinoma do not include the use of routine brain screening of brain metastases, most likely asymptomatic HCC cases with brain metastases will have been ignored.Therefore, the real incidence of brain metastases in HCC cases at time of diagnosis is potentially underestimated.One might assume that the existence of pulmonary and/or bone metastases would explain the rate of brain metastases, however this has not been proven yet for HCC as it has been for metastases of colorectal cancer. 24cond, since data on treatment of cases with brain metastases unavailable in the SEER database no further analysis is possible.The role of systemic treatment is not clear since only case reports on chemotherapy have been published to date 25 .Effects of sorafenib on brain metastases in HCC patients are not independently reported so far.However, results from retrospective studies suggest that surgery, whole-brain radiation therapy (WBRT), stereotactic radiosurgery (SRS) might prolong survival 17,[26][27][28] .In fact, sorafenib as additive to the afore-mentioned treatments has shown some promising primary results in brain metastases from renal cancer 29 .Third, data on tumor characteristics, such as tumor grade and AJCC stage are incomplete in the SEER database.Our major nding of AFP association can thus be critically seen when taking into account that 30% of cases didn't have a reported AFP status in our patient cohort.Although the fact that the tumor size doesn't affect the incidence of brain metastases is present in this database, there are other examples of poor prognosis associated with HCC larger than 5cm 30 .
Another limitation is the marginal incidence variance among age groups which makes the biological relevance of this nding somewhat questionable.However, stakeholders must face the challenge of screening guidelines (brain CT or MRI) for patients with HCC.A nal limitation to this manuscript is the lack of a validation dataset.Such dataset would ideally comprise of patients with similar socioeconomic factors in the same health system, but the access to The National Cancer Database (NCDB) US cancer registry les are only accessible to members of any Commission on Cancer Accredited Facility.

Conclusions
The presented data provide updated estimates on the incidence and survival of patients with brain metastases in hepatocellular carcinoma in the United States.Although a possible causality between lung/bone and brain metastases can be disputed, these showed a correlation in our exploratory analysis.
The current study aims to contribute to the consideration of routine screening of the brain among initially metastasized HCC patients with lung or bone metastasis.

Declarations
Ethics approval and consent to participate N/A (The SEER Database did not prerequisite any ethics approval since patient data are pseudonymised) Consent for publication N/A

Figures
Figure 1 Flow

2 Figure 2A :Figure 3 Figure
Figure 2 Figure 3A: Kaplan-Meier analysis of overall survival for synchronous brain metastases cases.Figure 3B: Kaplan-Meier analysis of liver cancer-speci c survival for synchronous brain metastases cases.Figure 3C: Kaplan-Meier analysis of overall survival for synchronous brain metastases among all HCC cases.

Figure 3D :
Figure 3D: Kaplan-Meier analysis of overall survival for synchronous brain metastases by age at diagnosis.

Table 1 .
Characteristics of 39773 HCC and 141 brain metastases cases in SEER database 2010-2015 a Not done or bordeline or unknown.AFP, alpha fetoprotein; NOS, not otherwise specified.

Table 2 .
Number and incidence of brain metastases at time of HCC diagnosis stratified by age at diagnosis