|Home | About | Journals | Submit | Contact Us | Français|
Stereotactic radiosurgery for the treatment of brain metastases is commonly delivered without regard to primary cancer histology. This study sought to determine if the primary site of origin for brain metastases affected the propensity for local failure.
A total of 83 patients with 200 brain metastases were examined retrospectively for predictors of infield failure. Tumor, patient, and treatment characteristics were analyzed including primary tumor histology, radiosurgical dose and age. Cox proportional hazards models, univariate and multivariate analyses were used to identify predictors of local failure.
Freedom from local failure for the entire population was 83% and 65% at 6 and 12 months, respectively. Multivariate analysis revealed that breast cancer brain metastases have a significantly lower risk of local failure than melanoma (HR = 0.31, p< 0.001). Additionally, multivariate analysis revealed that increasing dose lowered risk for local failure (HR = 0.87, p<0.001).
Melanoma histology leads to a higher rate of local failure. Higher prescription dose results in higher incidence of local control.
Stereotactic radiosurgery (SRS) represents an effective treatment modality for brain metastases because of excellent local control outcomes and the ability to avoid cognitive toxicities caused by whole brain radiotherapy (WBRT).[2, 3] Dosing guidelines based on tumor volume established by Shaw et al, have been used for the past two decades in order to avoid radiation necrosis. Aside from tumor volume, other characteristics have not been universally considered in the SRS treatment planning process. In spite of generally excellent local control outcomes for patients with SRS, with local control rates ranging from 74%- 95% in resected and intact brain metastases, there are still patients with early and unexpected treatment failures. While predictive models have been developed to aid in predicting which patients may have higher propensity for distant brain failure, there is a paucity of data surrounding factors other than tumor volume that affect local failure.
The radiosensitivity variability of brain metastases based on histology is a concept that has existed for several decades. Renal cell carcinoma, melanoma and sarcoma are classically considered to be radioresistant histologies. One hypothesis that may explain the improvement in tumor control with SRS over conventionally fractionated radiotherapy is that there may be changes in the intracellular target of radiation with high doses per fraction. Several mechanisms have been proposed that may explain why there is a difference in local control between histologies including tumor microenvironment, oxygenation, protein expression, intercellular signaling pathways, and a difference in extent of invasion into the brain parenchyma. Regardless of mechanism, there is accumulating data to suggest that different histologies may have a differential sensitivity to radiosurgery. This serves as one of a very small number of studies that explicitly looks at primary tumor histology and its relationship to local failure rates in brain metastases after SRS treatment.[11-13]
The goal of this study was to elucidate factors beyond tumor volume that may affect local failure, including the histology of the primary tumor; therefore, a retrospective analysis was conducted examining potential predictors of local treatment failure after SRS to brain metastases.
Eighty-three consecutive patients were treated with upfront radiosurgery with definitive intent from 2003-2013, including 200 total brain metastases, at the Comprehensive Cancer Center of Wake Forest University. Patients were retrospectively reviewed via electronic medical records for the presence of local failure. Patients were selected on the basis of having received Gamma Knife Radiosurgery (GKS) for the treatment of brain metastases and for whom volumetric treatment data were available. All had a follow-up imaging schedule of every 3-4 months post-treatment using MRI with contrast. The parameters of interest that were characterized and evaluated for this study included age, sex, WBRT before SRS, treatment volume, prescribed treatment dose, and primary site histology. This study was approved by the Wake Forest University Institutional Review Board.
All were treated with either the Leksell Model B, C, or Perfexion units (Elekta AB). Prior to radiosurgery, patients underwent a high-resolution MRI of the brain. Treatment planning was performed using the Leksell GammaPlan Treatment Planning System (Elekta AB). A median dose of 18 Gy (range 8-22 Gy) to the 50% isodose line was prescribed. Prescription dose was determined based on the guidelines previously published by Shaw et al.
Patients were followed clinically and with MRI performed at 4-8 weeks after SRS. If there was no sign of treatment failure at that point, clinical and imaging follow-up was done every 3 months thereafter. Local failure was determined based on either being pathologically proven after a surgical procedure or with imaging evidence of at least a 25% increase in estimated bi-perpendicular diameter on an axial MRI slice and an increase of perfusion on perfusion-weighted imaging.
Patient and treatment characteristics were summarized using descriptive statistics including the median and interquartile range (Quartile 1: 25th- Quartile 3: 75th percentile) for continuous data, count, and frequency data for categorical/ordinal data. A univariate cox proportional hazards analysis was used to identify factors which impacted the risk for infield failure. Covariates which were significant at the p<0.2 level were considered for inclusion into the multivariate model. Interactions across covariates selected from the univariate analysis were tested using product terms. Interacting covariates were then stratified across resultant levels to identify potential differential effects on the hazard for local failure in each group. The final multivariate model was constructed using a guided backwards stepwise selection technique. All covariates in the multivariate model were tested for and met cox proportional hazards assumptions. All analyses were performed using SAS software, Version 9.3 (Cary, NC).
Population characteristics are listed in Table 1. The population consisted of 83 patients comprising 200 total disease sites: 144 sites from female patients and 56 sites from male patients. Median age was 54 years, 4 months. Intracranial metastasis histologies included; lung, breast, melanoma, renal, colorectal, larynx, ovarian, and thyroid cancer The largest primary cancer subpopulations included lung (41.5%) and breast (28.5%) carcinomas as well as melanoma (23%). Median tumor volume was 482.9 mm3(range 1.02-52770 mm3).
Median follow-up time was 27.6 months. Of the 200 disease sites included, 90 cases of local failure and 110 cases of local control were observed during the follow-up period. Median patient survival was 15.2 months and median time to local failure was 15.9 months. Freedom from local failure at different time points for the entire population is detailed in Table 2.
Freedom from local failure for each primary site is detailed in Table 2. Log rank test demonstrated a statistically significant difference in freedom from local failure when comparing breast to either melanoma or lung patients, but not when comparing lung to melanoma cases (breast vs. melanoma: p=0.003, lung vs. melanoma: p=0.55). Figure 1 provides a Kaplan-Meier analysis of primary histology vs. time to local failure for breast, lung, melanoma, and all other primary sites. Tumors originating from breast cancer exhibited better outcomes when compared to all other primary sites.
A multivariate analysis was performed to assess for variables that affected local failure (Table 3). Multivariate analysis revealed that breast cancer brain metastases have a significantly lower adjusted hazard (aHR) for local failure than melanoma (aHR = 0.31, (95% CI 0.14-0.67), p< 0.01). Primary lung cancers did not show a significant effect on local failure when compared to melanoma (aHR 0.83 (95% CI 0.44-1.54), p=0.55). Age did not have an effect on local failure (aHR 0.99 (95% CI 0.97-1.02), p=0.67). Dose prescribed to the 50% isodose line had an effect on the local failure rate (HR=0.87, (95% CI 0.81-0.94), p<0.01) for all primary sites.
The current series revealed a significant relationship between local failure and tumor histology, implicating melanoma as having the highest relative risk for local failure after SRS. These findings agree with findings from in vivo [14-16] and in vitro [17-21] studies, as well as with clinical studies of fractionated radiotherapy and radiosurgery. There are several potential mechanisms for worsened local control in melanoma patients. A pathologic study performed by Baumert et al suggested that melanoma brain metastases invade further into brain parenchyma than other histologies.
This study is not without limitations; as a retrospective study, it is subject to patient selection biases. The patient number, 83, was sufficient to analyze 200 total metastases of varying histologies. Table 1 illustrates the breakdown of histologies, the majority of which included lung, breast, and melanoma primaries. It is difficult to conclude much from the more rare histologies including colorectal, ovarian, thyroid, renal and larynx cancers. In addition, there may be unknown and unmeasured factors that are influencing local control. However, the current study adds to existing literature that demonstrates that melanoma has worsened local control after SRS than other histologies. A recent series from South Korea has suggested that patients with melanoma have increased local failure rate when compared to other radioresistant histologies. In this series, the authors found that patients with melanoma that experienced local failure generally also experienced an intratumor bleed from tumor progression. Also supportive of our findings, a recent multi-institutional analysis of brain metastases by Rades et al. reported that melanoma patients experienced worse local control than non-small cell lung cancer.
A major clinical question that arises from the results of the present series is what can be done to help improve the therapeutic ratio of SRS to further optimize the local control rate, particularly in the larger tumors or ones with melanoma histology. While the dose delivered via SRS could theoretically be increased, prescription doses are currently constrained by the volume of the tumor being treated to avoid radionecrosis. Use of concurrent systemic agents is a strategy that has been assessed in both lung cancer and renal cell cancer with encouraging results. Surgical resection of accessible lesions followed by SRS to the resection cavity results in local control that is comparable to the use of SRS in small lesions. A recent series examining clinical outcomes in melanoma brain metastases demonstrated a high rate of neurologic death even among patients with local control, suggesting that the clinical problem with melanoma is more complex than optimizing local control. More investigation is required to determine the need for escalated dose or volume in metastatic lung cancer cases and whether increasing radiosurgery dose to radioresistant histologies such as melanoma may improve local control. Our results on dose do suggest that consideration of dose escalation may improve local control rates in melanoma metastases to the brain. However, given multiple other patient outcome factors, these findings suggest more investigation is warranted as to whether radiosresistant histologies may be well served with dose escalation in radiosurgery.
Nomenclature: Stereotactic radiosurgery (SRS); whole brain radiotherapy (WBRT); Gamma Knife Radiosurgery (GKS); adjusted hazard (aHR)
Biostatistical services supported by the Comprehensive Cancer Center of Wake Forest University NCI CCSG P30CA012197 grant. The authors also wish to thank Bonny B. Morris, for her thoughtful manuscript review.