Radiotherapeutic treatment of brain metastases in patients with gynecologic malignancies: A nationwide analysis

Abstract

Objectives: Brain metastases are a rare finding in patients with primary gynecologic malignancies having a poor outcome despite treatment. We sought to use the National Cancer Database (NCDB) to characterize the incidence of brain metastases and types of treatment administered.

Methods: We queried the NCDB from 2010-16 for patients with endometrial, cervical, and ovarian primaries with brain metastases at diagnosis treated with radiation—whole brain radiation (WBRT) or stereotactic radiosurgery (SRS). We tabulated baseline characteristics and performed a multivariable logistic regression to identify predictors of SRS. Multivariable Cox regression was used to identify predictors of death. Propensity matching was done to account for indication bias.

Results: We identified 765 patients meeting above criteria, representing <1% of patients. Of patients with brain metastases, 287 received radiation to the brain. The median age was 60 (40-90). The majority of patients (80%) received WBRT. On multivariable logistic regression the only predictor of SRS was receipt of chemotherapy. Median follow up was 4.9 months. The overall survival for the entire cohort was 5.2 months. On Cox regression predictors of improved survival were receipt of chemotherapy, no extracranial disease, and SRS. After propensity matching, median survival was 8.3 months compared to 6.3 months in favor of SRS.

Conclusions: This study represents the largest series to date of patients with brain metastases from gynecologic malignancies, confirming an incidence of <1% and overall poor prognosis. Radiation remains a viable treatment option.

Background

Brain metastases are the most common tumor found within the brain and are estimated to affect up to 30% of patients throughout a diagnosis of cancer[1]. Cancers of the female genital tract including endometrial, cervical, and ovarian cancers make up about 15% of malignancies diagnosed in women each year. Despite how common these malignancies are in women, they very rarely spread to the brain; and when they do the prognosis is often very poor[2]. This rarity and the resultant outcomes are likely due to what is considered the “neuorphobic” nature of these malignancies—indicating the uncommon occurrence and correlation with widespread systemic disease.[3] Treatment options for brain metastases from any primary malignancy include surgery, whole brain radiation therapy (WBRT), stereotactic radiosurgery (SRS), or some combination thereof depending on multiple factors including size and number of lesions, location, and severity of symptoms.[4,5] Given the rarity of this phenomenon we sought to use a large national dataset (the National Cancer Database (NCDB)) to better characterize the incidence of brain metastases in this population as well as the types of treatment administered.

Methods

The methods for analyzing the NCDB have been described previously.[6,7] This project was a retrospective review of the NCDB which is entirely de identified and thus exempt from institutional review board review. The NCDB is a national database spanning approximately 1,500 cancer centers in the United States and is estimated to capture up front treatment and staging data on close to 70% of newly diagnosed cancer patients each year. As stated above, the data used in this study are derived from a de-identified NCDB file. As required by the governing body, we must state that the American College of Surgeons and the Commission on Cancer have not verified and are not responsible for the analytic or statistical methodology employed, or the conclusions drawn from these data by the investigators.

We queried the NCDB from 2010-2016 for patients with a diagnosis of endometrial, cervical, or ovarian cancer with brain metastases at time of diagnosis. We then further analyzed the subset of patients with brain metastases that received brain directed radiation (either WBRT or SRS). Data for the anatomic location, radiation technique, dose, and fractionation are all included within the database. Complete selection criteria are outlined in Figure 1.

Figure 1.

CONSORT diagram showing selection criteria.

A variety of socioeconomic factors are also recorded in the NCDB and were tabulated and used in our analysis. Race was broken down into three categories including Caucasian, African American, or other. The Charlson/Deyo comorbidity index was used to quantify any co existing comorbid conditions.[8] In 2010, the NCDB began recording presence of brain metastases, lung metastases, bone metastases, and liver metastases. We used these categories as a surrogate and recorded patients as having extracranial metastatic disease if they were listed as having bone metastases, liver metastases, or lung metastases at time of diagnosis in addition to brain metastases. Within the NDCB, socioeconomic data in the patients’ residence census tract are listed as quartiles of the percentage of persons with less than a high school education and median household income based on zip code. The facility type (community cancer center, comprehensive community cancer center, and academic/research program) was assigned according to the Commission on Cancer accreditation category. Locations were assigned based on data provided by the US Department of Agriculture Economic Research Service and stratified as metropolitan, urban, or rural. Insurance status is documented in the NCDB as it appears on the admission page ranging from none, private insurance, and governmental (Medicare/Medicaid). We also recorded whether or not systemic therapy was delivered (either hormonal therapy, chemotherapy, or a combination). Please note that only delivery and timing of systemic therapy is listed in the NCDB and not specific systemic agents or number of cycles.

Data were analyzed using Medcalc Version 18 (Ostend, Belgium). Summary statistics are presented for discrete variables. Overall survival is the only outcome recorded in the NCDB and was calculated in months from diagnosis to date of last contact or death. Multivariable logistic regression was used to determine predictors of receipt of SRS. Kaplan-Meier curves were used to calculate cumulative probability of survival.[9] Log-rank statistics were used to test whether there was a statistically significant difference in the cumulative proportions across groups. A Cox proportional hazards model was used for multivariable survival analysis in a forward elimination process.[10] Adjusted hazard ratios and 95% confidence intervals are reported, using an a level of 0.05 to indicate statistical significance.

Due to high likelihood of significant selection bias we also calculated a propensity score and performed an adjusted survival analysis to account for indication bias due to lack of randomization between WBRT and SRS.[11] As described above, multivariable logistic regression was used to calculate the conditional probability of receiving SRS, which simultaneously generates a propensity score; also indicative of likelihood of receiving SRS. The propensity model included observable variables associated with treatment selection on multivariable logistic regression. Patients were then matched using an exact match based on propensity score, resulting in 53 pairs. Kaplan Meier analysis was then completed on this set of matched pairs to compare overall survival.

Results

Using the selection criteria above we identified a total of 765 patients with brain metastases at diagnosis, representing <1% of the entire population with those diagnoses. The majority (53%) of patients had a diagnosis of endometrial cancer. The median age was 60 (40-90). Of the 765 patients, 287 (38%) received brain directed radiation with 80% receiving WBRT. The median WBRT dose was a standard 30 Gy (28-35 Gy) in 10 fractions (8-14). The median SRS dose was 22.5 Gy (18-60 Gy) in 1 fraction (1-3). Fifty-nine percent of patients had extracranial metastatic disease and 57% received systemic therapy. The median time to start of radiation was 29 days (8-61), 28 days (10-54), and 30 days (14-51) for endometrial, cervical, and ovarian primaries, respectively. Table 1 outlines all of the baseline characteristics.

Table 1.

Baseline patient characteristics

Characteristics	No. (%)
Race
White	242 (84)
African American	34 (12)
Other	11 (4)
Comorbidity Score
0	212 (74)
1	55 (19)
≥2	20 (7)
Insurance
Not Insured	20 (7)
Private Payer	131 (46)
Government	131 (46)
Unrecorded	5 (1)
Education %
≥29	88 (31)
20 to 28.9	69 (24)
14 to 19.9	66 (23)
<14	63 (22)
Treatment Facility type
Community cancer program	24 (9)
Comprehensive community cancer program	85 (31)
Academic/research program	165 (60)
Treatment facility location
Metro	220 (79)
Urban	51 (18)
Rural	9 (3)
Income, US dollars
<30,000  30,000 to 35,000  35,000 to 45,999	63 (22) 64 (22) 74 (26)
>46,000	85 (30)
Distance to treatment facility, miles
≤12 miles  >12 miles	153 (53) 134 (47)
Age distribution, years
≤60	157 (52)
>60	130 (45)
Year of Diagnosis
2010-11	92 (32)
2012-13	87 (30)
2014-15	108 (38)
Radiation
WBRT	230 (80)
SRS	57 (20)
Primary Malignancy
Endometrial	178 (62)
Ovarian	53 (18)
Cervical	56 (20)
Systemic therapy (hormonal and/or chemotherapy)
No	116 (40)
Yes	171 (60)
Extracranial Disease
No	119 (41)
Yes	168 (59)

On multivariable logistic regression the only predictor of SRS use was receipt of chemotherapy. Please see Table 2 for complete results of that analysis. The median follow up was 4.9 months for all patients. On multivariable Cox regression predictors of improved survival were lack of extracranial disease, chemotherapy delivery, and receipt of SRS (Table 3). There was no difference in survival based on primary diagnosis, and we did not directly compare them due to the smaller sample sizes. To confirm there was a benefit to any radiation, we compared survival between those patients that did and did not receive radiation (either WBRT or SRS) with a median survival of 5.2 months compared to 3.2 months, in favor of radiation (p=0.0012) (Figure 2). As described in the methods a propensity score was taken from the multivariable logistic regression for SRS and 53 pairs were matched using an exact match on the score. Kaplan Meier analysis on that set of pairs showed a trend towards improved survival with SRS, median of 8.3 months compared to 6.3 months, p=0.12 (Figure 3).

Table 2.

Multivariable Logistic Regression for Receipt of SRS

Characteristic	OR (95% CI)	P
Age
≤60	Reference
>60	1.95 (0.92-4.10)	0.078
Systemic Therapy
No	Reference
Yes	2.07 (1.01-4.26)	0.048
Comorbidity Score
0	Reference
1	0.76 (0.30-1.97)	0.57
≥2	0.25 (0.03-2.10)	0.20
Distance to facility, miles
≤12	Reference
>12	1.07 (0.50-2.27)	0.87
Extracranial Disease
No	Reference
Yes	0.59 (0.29-1.20)	0.15
Facility Type
Community Cancer Center	Reference
Comprehensive Community Cancer Center	1.81 (0.34-9.48)	0.48
Academic/Research Program	4.00 (0.81-19.81)	0.09
Grade
Well differentiated	Reference
Moderately differentiated	0.08 (0.004-1.41)	0.084
Poorly Differentiated	0.34 (0.05-2.35)	0.28
Education, % w/o high school diploma
≥29	Reference
20-28.9	2.59 (0.88-7.65)	0.08
14-19.9	2.29 (0.68-7.75)	0.18
<14	2.52 (0.66-9.55)	0.18
Income, USD
<30,000	Reference
30,001-35,000	0.46 (0.13-1.55)	0.21
35,001-45,999	0.42 (0.12-1.43)	0.16
≥46,000	0.37 (0.09-1.51)	0.16
Insurance
None	Reference
Private	10.56 (0.62-181.07)	0.10
Government	9.61 (0.56-1656.14)	0.12
Not recorded	25.00 (0.83-270.56)	0.052
Location
Metropolitan	Reference
Urban	0.89 (0.32-2.46)	0.82
Rural	3.70E-010	0.99
Primary Diagnosis
Endometrial	Reference
Cervix	0.65 (0.23-1.86)	0.43
Ovarian	0.71 (0.28-1.79)	0.47
Race
Caucasian	Reference
African American	0.82 (0.27-2.54)	0.73
Other	3.25 (0.57-18.44)	0.18
Year of Treatment
2010-11	Reference
2012-13	0.61 (0.25-1.48)	0.27
2014-15	1.25 (0.55-2.85)	0.60

Table 3.

Multivariable Cox Regression for Predictors of Survival

Characteristic	Hazard Ratio (95% CI)	P
Chemotherapy
No	Reference
Yes	0.2263 (0.17-0.30)	<0.0001
Extracranial Disease
No	Reference
Yes	1.91 (1.45-2.51)	<0.0001
Radiation
WBRT	Reference
SRS	0.63 (0.45-0.87)	0.0052
Year of Treatment
2010-11	Reference
2012-13	0.74 (0.56-0.99)	0.04
2014-15	1.08 (0.77-1.53)	0.67

Figure 2.

Survival comparing no brain directed radiation to any brain directed radiation. With radiation median survival was 5.2 months compared to 3.2 months, p=0.0012.

Figure 3.

Propensity matches survival curves comparing WBRT and SRS in patients with brain metastase from gynecologic malignancies. After matching, median overall survival was 8.3 months vs 6.3 months for SRS and WBRT, respectively. p=0.12. The corresponding one year survival rates were 25% and 30%, again in favor of SRS.

Discussion

This study represents the largest series to examine the incidence and outcomes of patients with brain metastases from a primary gynecologic malignancy. The phenomenon is very rare, as <1% of patients with a new diagnosis of endometrial, cervical, or ovarian cancer were documented as having brain metastases at that time. The rarity is further supported by the fact that all of the current literature on this topic is limited to very small case series. Not surprisingly, we did show a small but significant benefit to any radiation treatment compared to no brain directed radiation. It is interesting to note that only 38% of patients were recorded as having brain directed radiation. This could simply be a matter of how data is collected; for example, perhaps patients were treated with chemotherapy first, and then later given “salvage” WBRT, which would not by definition be recorded in the NCDB. On the other hand, the data from the NCDB may indeed reflect real world practice and perhaps radiation is used less frequently compared to patients with brain metastases from other solid malignancies. This notion is perhaps supported by times ranging up to 60+ days for administration of radiation seen in the dataset. Our analysis did show a likely benefit to SRS, although that treatment was likely a surrogate for patients with more limited intracranial disease who would be expected to have better outcomes. Similarly, patients receiving chemotherapy had improved survival (and were more likely to receive SRS) likely signaling a better performance status. Of interest, in our small series we did not show a higher predilection for whole brain radiation in patients with ovarian cancer, although that could be due to small sample size. It is also worth noting, all of these patients had stage IV disease so systemic therapy would obviously be indicated and likely not administered in patients unable to tolerate the treatment.

As mentioned above due to the rarity of brain metastases in this patient population all of the data comes from small retrospective series. One of the earliest series examined outcomes in 27 patients with brain metastases from endometrial and ovarian cancer treated with SRS using the Gamma Knife (2/3 of the patients had WBRT as well).[12] The median survival was similar to that seen here, 7 months from diagnosis. In terms of local control all intracranial lesions were controlled, implying that cause of death was likely systemic illness. The group from Harvard also published outcomes on 8 patients with 20 lesions treated with SRS using the Cyberknife^TM SRS system in both a single fraction and fractioned manner.[13] Of note, in this study no patients had cervical cancer, one had endometrial cancer, and the remainder had ovarian cancer. In addition, only one patient had received prior WBRT. Median survival in that series was a more impressive 29 months (likely due to small sample size and primary tumor distribution) with a 1 year local control rate of 91%.

A more recent study from Korea analyzed outcomes from 20 patients treated over more than a decade with a variety of radiation techniques (SRS, WBRT, WBRT and surgery).[14] In that study less than 2% of patients with brain metastases were from a primary gynecologic malignancy, with a mix of endometrial (20%), cervical (10%), and ovarian primaries (70%). Interestingly, in this series median survival was 28 months, again likely due to the preponderance of ovarian cancer patients. Also, patients that had surgical resection had a median survival of 37 months, likely due to decreased intracranial burden and oligometastatic disease. Local control was not specifically reported, but progression free survival was a median of 15 months. Lastly, a multinational study from China examined outcomes in 12 patients treated in Canada, China, or India, all with brain metastases from ovarian cancer.[15] All of these patients had developed metastatic disease after diagnosis and median survival was 17 months from time of brain metastasis. For patients treated with surgery or SRS, median survival was 25.6 months compared to 6 months (p=0.006), again likely due to other confounding factors such as performance status, disease burden, and less intracranial disease.

Given the retrospective nature of the NCDB there are limitations to our study that we must mention. Mainly this includes what can be a significant selection bias in terms of radiation treatment, specifically SRS. In addition, the NCDB does not code the size or number of brain metastases present which would influence the type of treatment delivered and eventual outcome. The database does not record other outcomes pertinent to this type of study such as local failure, distant brain failure, and any salvage therapy which are all integral to treating patients with brain metastases. It was with these limitations in mind and the nature of the dataset to frame this project as more of an examination of the incidence and trends in treatment in this patient population rather than as a direct comparison of different treatment techniques.

Conclusions

This study represents the largest series to date of patients with brain metastases from gynecologic malignancies, confirming an incidence of <1% at diagnosis with an overall poor prognosis. Radiation remains a viable treatment option, with a potential benefit for SRS in appropriately selected patients (good performance, limited disease, and ability to receive chemotherapy).