Predictors of linear accelerator versus Gamma Knife stereotactic radiosurgery use for brain metastases in the United States

Abstract

Introduction

Stereotactic radiosurgery (SRS) for brain metastases is predominantly delivered via single-fraction Gamma Knife SRS (GKRS) or linear accelerator (LINAC) in up to five fractions. Predictors of SRS modality have been sparsely examined on a nationwide level.

Methods

An observational cohort study was performed on patients receiving SRS for brain metastases from non-small cell lung cancer from 2010 to 2016 at Commission on Cancer-accredited hospitals throughout the United States (US). A multivariable logistic regression model characterized SRS receipt, adjusting for patient age, dose, geographic location of treatment, facility type, and distance from treatment facility.

Results

A total of 2,684 patients received GKRS, while 1,643 patients received LINAC SRS. After adjusting for significant covariates, treatment at non-academic facilities was associated with increased LINAC SRS receipt, most prominently in the Midwestern (OR=6.23;p<0.001), Northeastern (OR=4.42;p<0.001), and Southern US (OR=1.96;p<0.001). Compared to patients receiving 12-17 Gy, patients receiving doses of 18-19 Gy (OR=1.42;p=0.025), 20-21 Gy (OR=1.82;p<0.001), and 22-24 Gy (OR=3.11;p<0.001) were more likely to receive LINAC SRS; similarly, patients located within 20 miles of a radiation treatment facility were more likely to receive LINAC SRS (OR=1.27;p=0.007).

Conclusions

Despite Gamma Knife being more prominently used over LINAC for SRS, patients treated at a non-academic facility outside of the Western US or requiring increased radiation dose were substantially more likely to receive LINAC over Gamma Knife. Additionally, patients residing in close proximity to a treatment center were 27% more likely to receive LINAC, likely indicative of the increased geographic accessibility of LINAC compared with GKRS.

Introduction

The predominant modality of stereotactic radiosurgery (SRS) delivery for brain metastases is via Gamma Knife SRS (GKRS) or linear accelerator (LINAC), with GKRS typically administered in a single fraction and LINAC administered in one to five fractions (1). Clinical and demographic characteristic associated with SRS modality use have been sparsely examined on a nationwide level; such information increases in importance as SRS treatment modality becomes more diverse (2). We sought to address this void via a national analysis.

Methods

An observational cohort study was performed on patients receiving SRS for brain metastases from non-small cell lung cancer from 2010 to 2016 at Commission on Cancer-accredited hospitals in the United States (US). This study was performed using the National Cancer Database (NCDB), which provides data on radiation dosage, technique, and target, and contains de-identified data on 70% of newly diagnosed cancers in the US. The NCDB is a hospital-based cancer registry jointly by the American Cancer Society and the American College of Surgeons (3). We identified 4,327 qualifying recipients of SRS therapy who received between 12 and 24 Gy radiation. SRS was defined as single-fraction GKRS or LINAC SRS of 1-5 fractions.

Univariate and multivariable analyses were performed using R version 3.5.3. The multivariable logistic regression model adjusted for dose, distance from treatment facility, and the interaction between geographic location of treatment, and facility type. Significance was defined as a two-sided P value < 0.05.

Results

A total of 2,684 patients received Gamma Knife SRS, while 1,643 patients received LINAC SRS. Univariate analysis revealed that increased age, higher income, more education, larger urban centers, increased total radiation dose, non-academic facility, and closer proximity to the treatment facility were associated with increased receipt of LINAC SRS modality. Histology and geographic regions are also significantly associated with SRS modality use while gender, race, insurance status, or Charlson-Deyo comorbidity score were not (Table 1). LINAC SRS utilization is highest in non-academic facilities of the Midwest (73%) and Northeast US (66%). Utilization is lowest in the West with 31% use in non-academic compared with 30% in academic facilities (Table 2).

Table 1.

Logistic univariate analysis of NSCLC brain metastases for factors associated with LINAC versus Gamma Knife SRS (n = 4,327)

Term	Levels	Odds	LL	UL	p-value	overall p-value
AGE	< 55 years (ref)
	≥ 55 years	1.19	1.01	1.41	0.036
SEX	Male (ref)
	Female	1.00	0.91	1.17	0.610
RACE	White (ref)					0.700
	NH_Black	0.93	0.77	1.12	0.470
	Hispanic	0.91	0.60	1.35	0.640
INCOME†	< $46,000 (ref)
	≥ $46,000	1.14	1.00	1.29	0.050
EDUCATION†	≥ 29% (ref)
	< 29%	1.30	1.05	1.50	0.010
URBAN/RURAL†	Other (ref)
	Metropolitan, > 250,000	1.25	1.08	1.46	c
HISTOLOGY	Adenocarcinoma (ref)					0.010
	Squamous cell carcinoma	1.14	0.94	1.38	0.176
	Large cell carcinoma	1.51	1.01	2.25	0.042
	Other	0.84	0.69	1.01	0.059
Charlson-Deyo Score	< 2 (ref)
	≥ 2	1.20	0.99	1.45	0.070
Dose levels (Gy)	12-17 Gy (ref)					< 0.001
	18-19 Gy	1.69	1.33	2.14	< 0.001
	20-21 Gy	1.66	1.33	2.09	< 0.001
	22-24 Gy	3.40	2.72	4.27	< 0.001
Facility type (2) ¥	Academic (ref)
	Non-Academic	3.09	2.63	3.64	< 0.001
Facility region	Northeast (ref)					< 0.001
	South	0.97	0.82	1.13	0.662
	Midwest	1.24	1.06	1.45	0.008
	West	0.52	0.40	0.66	< 0.001
Insurance†	Private (ref)					0.230
	Medicare	1.08	0.94	1.23	0.270
	Medicaid	0.93	0.73	1.16	0.510
	Uninsured	0.80	0.54	1.16	0.250
Distance from treatment facility† ¥	≥ 20 miles (ref)
	< 20 miles	1.57	1.3	1.85	< 0.001
Year of Diagnosis (2-periods)	2010-2012 (ref)
	2013-2016	0.96	0.84	1.10	0.570

^†’missing/unknown’ omitted from analysis. Significant associations with LINAC use appear in bold italic

^¥uses dataset verified for facility (n = 2,824)

Table 2.

SRS Modality Utilization by facility type and region (n = 2824*)

		Gamma Knife	LINAC	p-value
Region:		n (%)	n (%)
All US
	Academic	1,355 (71%)	553 (29%)	< 0.001
	Non-Academic	405 (44%)	511 (56%)
Midwest
	Academic	335 (71%)	139 (29%)	< 0.001
	Non-Academic	66 (27%)	176 (73%)
Northeast
	Academic	585 (70%)	245 (30%)	< 0.001
	Non-Academic	81 (34%)	156 (66%)
South
	Academic	382 (72%)	146 (28%)	< 0.001
	Non-Academic	144 (53%)	128 (47%)
West
	Academic	53 (70%)	23. (30%)	0.920
	Non-Academic	114 (69%)	51 (31%)

^*limited to those facilities known Academic/Non-Academic affiliation.

Multivariable analysis (Table 3) revealed that patients treated at non-academic facilities were more likely to receive LINAC SRS; most prominently in the Midwestern [odds ratio (OR) = 6.23; 95% confidence interval (CI) = 4.40-8.93; p < 0.001], Northeastern (OR = 4.42; 95% CI = 3.16-6.22; p < 0.001), and Southern United States (OR = 1.96; 95% CI = 1.42-2.70; p < 0.001) (Figure 1). In the Western United States there was no increased preference for LINAC SRS treatment by facility type (OR =1.00, 95% CI = 0.52, 1.92). Dose levels of 18-19 Gy (OR = 1.42; 95% CI = 1.05-1.94; p = 0.025), 20-21 Gy (OR = 1.82; 95% CI = 1.37-2.43; p < 0.001), and 22-24 Gy (OR = 3.11; 95% CI = 2.33-4.19; p < 0.001) were associated with increased receipt of LINAC SRS as compared with 12-17 Gy. Finally, patients located within 20 miles of a radiation treatment facility were significantly more likely to receive LINAC SRS (OR = 1.27; 95% CI = 1.07-1.51; p = 0.007).

Table 3.

Multivariate logistic model for factors associated with treating NSCLC brain metastases with LINAC versus Gamma Knife SRS from 2010-2016.

Term	levels	Odds	LL	UL	p-value (level)
Facility Type Stratified by Region†					< 0.001*
	Academic (ref)
Northeast	Non-Academic	4.42	3.16	6.22	< 0.001
Midwest	Non-Academic	6.23	4.4	8.93	< 0.001
South	Non-Academic	1.96	1.42	2.7	< 0.001
West	Non-Academic	1.00	0.52	1.92	0.990
Dose levels (Gy)					< 0.001*
	12-17 Gy (ref)
	18-19 Gy	1.42	1.05	1.94	0.025
	20-21 Gy	1.82	1.37	2.43	< 0.001
	22-24 Gy	3.11	2.33	4.19	< 0.001
Distance from treatment facility ¥	≥ 20 miles (ref)
	< 20 miles	1.27	1.07	1.51	0.007

^†Stratified due to the significant interaction effect between Facility Type and Facility Region

^¥“missing’ omitted from analysis

^*reporting variable overall p-value

Figure 1.

Analysis of LINAC versus Gamma Knife SRS utilization by geographic region of the United States.

Discussion

The findings in this study strongly indicate a clustering of GKRS in academic facilities and of LINAC SRS in non-academic facilities. Interestingly, while LINAC SRS is prominent in non-academic facilities throughout most of the US, it is not as prominent in the Western United states, where it comprises approximately 30% of SRS cases; this is in stark contrast to the near 80% LINAC SRS in non-academic centers in the Northeast, Midwest, and South in 2016 (Figure 1). Patients located in the Midwest were 523% more likely, Northeastern patients were 342% more likely, and Southern patients were 96% more likely to receive LINAC when treated at a non-academic facility.

Another interesting finding is that increasing delivery dose is independently associated with LINAC over Gamma Knife receipt, indicating that smaller tumors – particularly those less than two centimeters – were more likely being treated with LINAC in accordance with RTOG 90-05 dose recommendations, which involve 24 Gy for lesions 0-2 cm, 18 Gy for lesions 2.1-3 cm, and 15 Gy for lesions 3.1-4 cm (4). While the NCDB does not provide tumor size data for brain metastases treated, the dose of radiation administered can be used as an indirect measure of tumor size given the well-known parameters established by RTOG 90-05. Finally, patients residing in close proximity to a treatment center were 27% more likely to receive LINAC, likely indicative of the increased geographic accessibility of LINAC compared with Gamma Knife (2).

Limitations of this study relate to its retrospective nature and reliance on the NCDB, the limitations of which have been documented in detail, including its absence of Karnofsky Performance Status data (5-6).

In conclusion, despite Gamma Knife being more prominently used over LINAC for SRS, patients treated at a non-academic facility outside of the Western United States, resided within 20 miles of a radiation treatment facility, or required higher doses of radiation were substantially more likely to receive LINAC over Gamma Knife.