Prostate Adenocarcinoma with Brain Metastasis: A Surveillance, Epidemiology, and End Results Database Analysis 2010–2015

Background

Prostate cancer with brain metastasis is rare and usually occurs late in the course of the disease. Prostate cancer metastasis disseminates predominantly to the bones (87%), followed by distant lymph nodes (10.6%); however, it can eventually spread to other organs such as the liver (10.2%), lungs (9.1%), and brain [1]. The most common site for brain metastasis is in the leptomeninges, followed by the cerebrum and cerebellum [2,3]. In 1940, Dr. Oscar Vivan Batson first discovered the importance of the venous plexus (which was then named the Batson’s venous plexus) in the human body [4], which is a network of veins connecting the paravertebral veins to the pelvic veins and the thoracic veins to the intraspinal (basivertebral) veins [5]. These plexuses are important in understanding how prostate cancer spreads. Metastasis to the brain can be explained by the hematogenous spread, Batson’s venous plexus network, or direct extension from the sphenoid bone or adjacent sinuses [6].

Patients with brain metastasis are asymptomatic in the early stage but, later, they can present with headache, confusion, memory loss, seizures, hemiparesis, or aphasia [7]. The current National Comprehensive Cancer Network guidelines recommend magnetic resonance imaging (MRI) of the brain only for symptomatic patients. Hatzoglou et al conducted an 11-year analysis of MRI characteristics in prostate brain metastasis. A total of 21 patients were included in the study; one-third of those patients presented with a hemorrhagic pattern similar to that of other cancers, such as renal cell cancer and melanoma. The rest of the patients presented with varied patterns (from purely solid to mixed cystic and solid or ring-like patterns) [8].

Small cell histology originating from the prostate behaves like small cell cancers from any other primary site and has a high tendency to spread to the brain, compared with adenocarcinoma histology [9–11]. Prostate adenocarcinoma brain metastases have not been well described in the literature. Thus, we used Surveillance, Epidemiology, and End Results (SEER) data to determine and compare the incidence, correlation, and survival outcomes of cohorts of patients with prostate adenocarcinoma with brain metastasis and without brain metastasis.

Material and Methods

DATA SOURCE:

Data were extracted from the SEER database for all stage IV prostate cancer adenocarcinoma diagnosed between 2010 and 2015. We searched for prostate cancer patients using the International Classification of Disease for Oncology (3rd edition) histological subtype code “adenocarcinoma, NOS” 8140/3. We excluded cases that were not primary or first prostate cancer or were without active follow-up information (only autopsy or death certificate records), and histological subtype codes “small cell carcinoma, NOS” 8041/3, 8043/3, “large cell carcinoma, NOS” 8012/3, and “large cell neuroendocrine carcinoma” 8013/3. Data, when abstracted, were deidentified, and therefore, this study was exempt from institutional review board approval.

DEFINITION OF VARIABLES:

The demographic variables of interest for these patients included age at diagnosis, sex, race, histology grade, sites of distant metastases (liver, lung, bone, and brain), and 5-year survival outcomes. Race was categorized as White, Black, and other (other and unknown). Histology grades were based on the 2014 SEER Gleason conversion table. For this analysis, we combined grades 1 and 2 and grades 3 and 4. Grades 1 and 2 are groups of patients with Gleason scores from 2 to 7 and histology ranging from well differentiated to moderately differentiated. Grades 3 and 4 are groups of patients with Gleason scores from 8 to 10 and poorly differentiated histology.

Distant metastases to the bone, liver, lung, and brain were examined. Brain metastasis only was identified as “code 1” and represented “mets at diagnosis”. Brain involvement could have been single or multiple sites and could have included clinical or pathological information. No brain metastasis was identified as “code 0”, which has no clinical or pathological or imaging evidence for brain metastasis.

We further inquired for liver, lung, and bone metastases in the patients with brain metastasis at diagnosis and patients without brain metastasis at diagnosis. Surgery at metastasis site, and chemotherapy were categorized into “yes” and “no/unknown” statuses. The SEER data provided only “surgery at metastasis” but there was no information provided on the location of the surgery at metastasis site; for example, brain surgery vs visceral surgery. However, we selected patients who presented with brain metastasis at the time of diagnosis. We assumed that “surgery at metastasis” site would refer to brain surgery, since visceral surgery would rarely be performed in patients having both brain and visceral metastases.

Survival was defined as the time from diagnosis until death due to all causes. We tracked survival outcomes for 5 years. Patients with missing or unknown information and TNM/AJCC staging were excluded.

STATISTICAL ANALYSIS:

SEER *Stat software (version 8.3.5) was used to extract data. Patients were divided into 2 cohorts: brain metastasis and no brain metastasis. Descriptive analysis was conducted to evaluate the distribution of demographic variables, presence of visceral metastasis, grading, and treatment. Differences of categorical variables between the 2 groups were assessed using the chi-squared test. Overall survival distribution was estimated using the Kaplan-Meier method. The log-rank test was used for comparison of survival distributions across different subgroups. Univariate and multivariable analyses were conducted using the Cox regression model. All tests were 2-sided, and a P value of <0.05 was defined as statistically significant.

Results

BASELINE DEMOGRAPHICS:

We identified a total of 14 753 patients with stage IV prostate adenocarcinoma between 2010 and 2015 using SEER data. We divided the population into 2 main groups: patients with brain metastasis (group 1) and patients without brain metastasis (group 2). Of these, 187 patients were in group 1 and 14 566 were in group 2. The incidence of brain metastasis among the 2 groups was 1.26%. The clinicopathological characteristics of all patients are listed in Table 1. In group 1, the median age was 67 years (range, 42–92 years), with race of 75% White and 19% Black. The majority of patients (89%) had grade 3–4 and 11% had grade 1–2 metastases. Most patients in group 1 had metastases to bone (87%), followed by lung (29%) and liver (13%). Therapy received in group 1 was divided into chemotherapy and brain surgery. Only 19% of these patients underwent brain metastasis resection and 13% received chemotherapy. In group 2, the median age was 71 years (range, 34–99 years), with race of 75% White and 18% Black. The majority of patients (91%) had grade 1–2; only 9% of patients had grade 3–4 metastases. Most of these patients had metastases to bone (90%), followed by lung (7%) and liver (4%).

SURVIVAL ANALYSES:

The overall survival of group 1 was 50% after 1 year, 30.75% after 2 years, 21.4% after 3 years, and 15% after 5 years, while the overall survival of group 2 at the same time periods was 75%, 54.6%, 40.7%, and 24.7%, respectively. Median survival was 12 months in group 1 and 28 months in group 2 (P<0.001) (Figure 1). In group 1, a significant survival benefit was seen in patients who underwent resection at metastasis sites (hazard ratio 1.62, P=0.04) (Figure 2) and who received chemotherapy (hazard ratio 1.8, P=0.04) (Figure 3).

Logistic regression was used to study the association between brain metastasis (yes/no) and other risk factors (Table 2). Brain metastasis was statistically significantly associated with liver metastasis (P<0.001), lung metastasis (P<0.001) and age group (≤75 vs >75) (P=0.048), but not with race (P=0.77), bone metastasis (P=0.17), or chemotherapy (P=0.20) by univariate analysis (Table 2). Liver metastasis and lung metastasis were also significant by multivariable analyses (Table 2). There was a strong association between brain metastasis and visceral metastasis; these patients had a 2-fold to 4-fold higher risk of having brain metastasis (P<0.001) (Table 3). The odds ratio for liver metastasis was 2.85 (95% confidence interval [CI] 1.89–4.2) and for lung metastasis was 4.6 (95% CI 3.3–6.4) (Table 3). There was no association between age, race, grade, or bone metastasis and risk of having brain metastasis (Table 3).

Discussion

LIMITATIONS:

The current study has several limitations. First, the risk of selection bias is high as it is a retrospective study. Second, the SEER database does not provide details of treatment, for example, name of chemotherapy, name of hormonal therapy, exact location of surgery at metastasis site, brain resection information, and pathology information from brain resection; also, the molecular information was limited. We did not include radiation information in our study because these data were difficult to interpret with regard to which site was treated (prostate, bone metastasis, or brain metastasis site).

Conclusions

Prostate cancer with brain metastasis has a poor prognosis. Our study shows that visceral metastases are associated with a higher risk of brain metastasis and, thus, early MRI of the brain should be considered to diagnose asymptomatic brain metastasis, as early intervention would improve overall survival outcomes.