Hemorrhage of brain metastasis is a poor prognostic factor in hepatocellular carcinoma patients

It is unclear whether hemorrhage of brain metastasis is a poor prognostic factor in patients with hepatocellular carcinoma. We conducted a retrospective cohort study to compare overall survival between hemorrhage and no-hemorrhage groups of hepatocellular carcinoma patients with brain metastasis. Hepatocellular carcinoma patients with brain metastasis treated between June 2000 and June 2016 at the Cancer Hospital of Guangxi Medical University were retrospectively reviewed. Clinical characteristics and overall survival were compared between patients with (n = 11) and without (n = 25) hemorrhage of brain metastasis. Univariate and multivariate survival analyses showed hemorrhage to be a poor prognostic factor (hazard ratio = 5.812, 95% confidence interval: 1.399-24.142, p = 0.015). Patients with hemorrhage had a shorter median survival than those without hemorrhage (4 weeks vs 8 weeks, p = 0.001). These results suggest hemorrhage of brain metastasis is a poor prognostic factor in patients with hepatocellular carcinoma patients.


INTRODUCTION
Hepatocellular carcinoma (HCC) is one of the most commonly occurring cancers in Southeast Asia [1]. In China, it is also one of the top three causes of cancer death in areas where hepatitis B infections are prevalent. Moreover, the incidence of HCC is rising in Western countries [2]. The lungs, bone, and adrenal glands are common metastasis sites in HCC, whereas brain metastasis (BrM) is rare [2], with an incidence of 0.2% to 2.2% [3][4][5][6][7][8][9]. Because the prognosis of HCC patients with BrM is extremely poor [3,[5][6][7]10], prognostic factors and treatment modalities are not well defined.
BrM from HCC is fast growing, highly vascularized, and commonly associated with hemorrhage [11], though several studies suggest hemorrhage is not a prognostic factor in HCC and does not affect survival duration [5,6,10,12]. On the other hand, Han et al [7] reported that hemorrhage of BrM was associated with poor overall survival in HCC.
Recent therapeutic advances for HCC have contributed to improved survival rates [2]. As a result, the incidence of BrM is expected to increase as HCC patients survive longer [4]. We therefore conducted a retrospective cohort study to assess the prognosis of HCC patients with hemorrhage from BrM. We anticipate the results of this study may help clinicians make better treatment decisions for HCC patients.

RESULTS
A total of 39 HCC patients were diagnosed with BrM. Three of those were excluded because of incomplete data, so this study ultimately included 36 patents. All of these had patients died by the final follow-up. Table 1 summarizes patients' characteristics. Figure 1 shows the comparison of overall survival between the hemorrhage and no-hemorrhage groups. In a univariate analysis, the variables correlated with median survival after diagnosis of BrM were RPA, HCC treatment modality, hemorrhage, and BrM treatment modality ( Table 2). The results showed that there was a significant difference in median survival between the hemorrhage and no-hemorrhage groups (4 weeks vs 8 weeks, p = Clinical Research Paper www.impactjournals.com/oncotarget

DISCUSSION
Our study suggests that HCC patients with BrM hemorrhage have a poorer prognosis than those without hemorrhage. This finding suggests clinicians should pay greater attention to BrM hemorrhage when making treatment decisions.
Previous studies reported that BrM from HCC is frequently associated with hemorrhage [5-7, 10, 12], with incidences of 41.94% to 74.74%. In the present study, the hemorrhage rate among HCC patients with BrM was 30.56%. At our hospital, brain imaging is not performed only in cases with neurologic symptoms/signs, but also part of the routine evaluation of HCC patients. Consequently, 12 patients in this study were diagnosed with BrM at the time of their HCC diagnosis, which may account for the lower rate of BrM hemorrhage in our study.
Whether BrM hemorrhage significantly affects survival in HCC patients is controversial. Univariate and multivariate analyses carried out in several studies have suggested that BrM hemorrhage is not a prognostic factor associated with difference in survival [5,6,10]. For example, Hsieh et al [12] reported that the occurrence of BrM hemorrhage did not influence overall survival of HCC patients as compared to patients who did not experience BrM hemorrhage. By contrast, Han et al [7] reported that BrM hemorrhage was predictive of poorer prognosis, as patients without hemorrhage survived longer than who experienced BrM hemorrhage (13.7 weeks vs 8.1 weeks, p = 0.044 in univariate analysis). Both our univariate and multivariate analyses also indicate BrM hemorrhage is a poor prognostic factor and that HCC patients with BrM hemorrhage have a significantly shorter median survival than those without hemorrhage. In our study, 81.82% patients with BrM hemorrhage received palliative care. This may explain the poorer survival compared to earlier studies [6,7], as patients who received palliative care had a poorer prognosis than those who received therapeutic treatment. This would confound the result in the context of a treatment effect versus patient selection effect.
In this study, palliative care was associated with poorer survival than BrM treatment, including resection, whole brain radiotherapy, stereotaxic radiosurgery, or chemotherapy (4 weeks vs 11 weeks, p = 0.001). However, the best treatment modalities for BrM from HCC are not clear due to its rarity. The treatment may be similar to the general guidelines for metastatic brain tumors. For a single large lesion (<3 cm), surgical resection or stereotaxic radiosurgery should be considered with/without whole brain radiotherapy. Surgery was also a good treatment option for hemorrhagic BrM, though increased intracranial pressure and severe neurologic deficits may have existed [13]. In our study, two patients with BrM hemorrhage received resection, and they showed considerably

MATERIALS AND METHODS
This study was approved by the Ethics Committee of the Cancer Hospital of Guangxi Medical University. HCC patients treated between June 2000 and June 2016 at the Cancer Hospital of Guangxi Medical University were retrospectively reviewed. HCC was diagnosed based on pathology or radiological criteria [2]. BrM was diagnosed based on computerized tomography (CT) and/or magnetic resonance imaging (MRI), with or without pathology.
Clinical data at the time BrM was diagnosed, including age, sex, time interval from HCC diagnosis to BrM, Karnofsky performance status (KPS), Child-Pugh classification, recursive partitioning analysis (RPA) class, level of alpha fetoprotein (AFP), and extracranial metastasis, were collected. Also evaluated were data on BrM, including presenting symptoms/signs, location, number, hemorrhage, treatment modality, and survival time. BrM hemorrhage was diagnosed based on the  HCC treated: HCC treated with resection, transcatheter arterial chemoembolization, radiofrequency ablation, radiotherapy, or chemotherapy. BM treated: BM treated with resection, whole brain radiotherapy, stereotaxic radiosurgery, or chemotherapy.
pathology at surgery and/or CT/MRI. Patients were divided into hemorrhage and no-hemorrhage groups. The endpoint of this study was overall survival. The follow-up period was terminated by death or the beginning of this study (March 2017). Categorical variables were compared using the Chisquare test or Fisher's exact t-test. Continuous variables were expressed as the mean ± standard deviation and compared using Student's t-test. Prognostic factors were analyzed using log-rank test for univariate analysis; Cox regression analysis was used for multivariate analysis. Overall survival was calculated from the diagnosis of BrM of death or last day of follow using the Kaplan-Meier method and compared using the log-rank test.
Statistical analysis was performed using SPSS for Windows version 16.0 (SPSS Inc., Chicago, IL). All tests were two-sided, and values of P <0.05 were considered statistically significant.