Stereotactic ablative body radiotherapy (SABR) for effective palliation of metastases: factors affecting local control

Abstract

We analyzed factors associated with inferior local control following stereotactic ablative body radiotherapy (SABR) for palliation of metastases.

We reviewed records of patients receiving SABR for metastases at Duke University from 2006-2010. Biologically effective dose (BED) was calculated using the linear-quadratic model. Toxicity was assessed by CTCAE v4.0. The Kaplan-Meier method was used to estimate overall survival (OS) and local control (LC) within subgroups (primary or salvage SABR). Univariate (UVA) and multivariate (MVA) regression analysis was used.

Fifty and 33 patients received primary and salvage SABR, respectively. 105 lesions were treated (52 spine, 27 lung, 7 liver, 11 other); 67 primary SABR and 38 salvage. Median clinical follow-up was 11.1 months and 10.3 months with imaging of the treated lesion. One patient received SABR x3 and died from toxicity. 88% of symptomatic patients improved after SABR. 1-year LC and OS were 83% and 50%, respectively. Primary SABR had higher BED and was associated with improved LC on UVA (HR 3.0, p=0.01) and MVA (p=0.02); treatment site and histology were not.

SABR results in effective palliation of metastases regardless of prior treatment. In the absence of prior EBRT, SABR can be delivered with higher BED and may be associated with better outcomes.

1. INTRODUCTION

Thoughtful and effective palliation of symptomatic metastatic disease takes on greater importance with increasingly effective systemic therapies. Conventional external beam radiotherapy (EBRT) is known to be efficacious as a palliative local therapy for a variety of extra-cranial metastatic sites. In cases of treatment failure after EBRT, stereotactic ablative radiotherapy (SABR) [1] is often employed to treat targets in close proximity to dose-limiting critical normal structures.

A substantial body of literature exists to support the safety and efficacy of extra-cranial SABR. Evidence-based guidelines recommend definitive lung SABR for medically inoperable stage I non-small cell lung cancer (NSCLC) [2]. Increasing evidence is emerging to support the use of definitive SABR for prostate cancer [3, 4] and for inoperable pancreatic cancer [5]. Lung [6, 7], liver [8, 9], and spinal [10-14] metastases have been effectively treated with SABR. Five hundred spinal metastases were treated with a single fraction of 12.5 – 25 Gy resulting in 88% radiographic control, 86% improvement in symptoms, and no case of myelitis [11]. In addition, one to five metastatic sites have also been treated with upfront SABR in patients with limited metastatic disease, yielding local control rates of approximately 40-75% [15, 16]. An ongoing randomized Radiation Therapy Oncology Group (RTOG 06-31) study compares upfront SABR against EBRT for palliation of spinal metastases.

Many factors may influence local control following SABR including site of metastasis, radioresistant histology (e.g. melanoma[17], renal cell carcinoma[18]), and SABR dose. Dose-limiting critical normal structures heavily influence the SABR prescription dose. This effect is exaggerated when SABR is used following treatment failure after EBRT [19]. Moreover, SABR dose-response relationships have been suggested for treatment of stage I non-small cell lung cancer [20] and spinal metastases [13]. Thus, we analyzed factors, including but not limited to histology and prior EBRT, associated with superior local control following SABR for metastatic lesions. The effect of prior EBRT on SABR dose was also evaluated.

2. METHODS

This study was approved by the Duke University Institutional Review Board. Patients were identified in a prospectively maintained database of SABR. All records and data were reviewed retrospectively. All patients were treated palliatively with SABR either alone (i.e. primary SABR) or after local progression following EBRT (i.e. salvage SABR) from June 2006 to July 2010. A variety of metastases were treated (Table 1), although spine and lung SABR predominated. Concurrent systemic therapy was not used.

Table 1.

Patient characteristics

Characteristic	Primary SABR	Salvage SABR
Lesions Treated	67	38
Median Age (yrs)	65 (range 32-90)	62 (range 32-86)
Gender
Male	37 (55)	17 (45)
Female	30 (45)	21 (55)
Site Treated
Spine	26 (39)	26 (68)
Lung	23 (34)	4 (11)
Liver	7 (10)	0
Abdomen	7 (10)	1 (3)
Sacrum	0	3 (8)
Other*	4 (6)	4 (11)
Histology
Adenocarcinoma	24 (36)	17 (45)
Renal Cell Ca	16 (24)	6 (16)
Squamous Cell Ca	10 (15)	9 (24)
Melanoma	4 (6)	1 (3)
Other†	13 (19)	5 (13)

^*Extremity, Pelvis, Head and Neck, Sacrum, Rib, Mediastinum

^†Transitional Cell, Non-small cell, Paraganglioma, Unknown, Thymoma, Adenoid Cystic, Anaplastic, Wilms, Schwannoma, Lymphoma, Small Cell, Pheochromocytoma

2.1. SABR Technique

Simulation was tailored to the site of SABR and was at the discretion of the treating physician. For spinal lesions, patients were simulated supine utilizing a customizable body mold. Contrast-enhanced computed tomography (CT) images were obtained with patients free-breathing. Magnet resonance imaging (MRI) and positron emission tomography (PET) were fused with the planning CT as indicated in the treatment planning software, Eclipse (Varian Medical Systems, Palo Alto, CA). Lesion volumes, target expansions, and dose prescriptions have been previously described [12] and were at the discretion of the treating physician. For lung SABR, all patients were immobilized supine in a customizable body mold and underwent 4-D CT simulation. Breath-hold technique was generally utilized when superior/inferior target motion exceeded 1.5 cm, and free-breathing technique with an internal target volume (ITV) was utilized otherwise. All SABR treatments were performed using mega-voltage linear accelerators with daily kilo-voltage image guidance.

2.2. Follow-Up

Patients were seen weekly during radiotherapy and every 3-6 months thereafter. Surveillance imaging was common and scheduled at the discretion of treating physicians. Disease progression and late toxicity were assessed using both the clinical exam and radiographic follow-up (RECIST criteria, version 1.1), when available. Toxicity was graded retrospectively using the Common Toxicity Criteria, version 4.0.

2.3. Statistical Analysis

All time-to-event analyses were measured from the date of SABR delivery. Radiographic progression within or adjacent to the CTV was classified as a local recurrence. Clinical progression, i.e. pain, without evidence of treatment-related toxicity was also scored as a local recurrence. Time to local recurrence was defined from date of SABR to the date of first local recurrence. Patients without local recurrence were censored at the date of last follow-up or death, whichever occurred first. The Kaplan-Meier method was used to estimate survival probabilities.

Biologically effective dose (BED) was calculated using the following formula, where n and d represent number of fractions and dose per fraction, respectively:

An alpha/beta ratio of 10 Gy was used to estimate acute effects and 3 Gy to estimate late effects.

Univariate cox proportional hazards regression models were used to test for significant predictors of local recurrence. In addition, multivariate proportional hazards regression models were used to determine if primary SABR was a significant predictor of local control while controlling for site and histology. Two-sided p-values less than 0.05 were used to determine statistical significance.

3. RESULTS

Eighty-three patients with 105 metastatic lesions were treated with SABR between June 2006 and July 2010. Sixty-seven lesions were treated with primary SABR (no EBRT) and 38 lesions were treated with salvage SABR following treatment failure after EBRT. Patient characteristics are listed in Table 1. Median clinical follow-up was 11.1 (range 0.2 – 47.9) months and follow-up with a radiographic study to assess the treated lesion was 10.3 (range 0.5 – 44.5) months. Fifty-four patients have died and there have been 27 local recurrences.

3.1. Dose

A variety of fractionation schedules and dose prescriptions were used. For the entire cohort, the mean biologically effective dose (BED) was 155.7 Gy₃ and 76.0 Gy_10. The biologically effective dose (BED) was higher with the use of primary SABR (158.7 Gy₃ and 69.5 Gy₁₀) compared to salvage SABR (77.2 Gy₃ and 38.3 Gy₁₀) (see Table 2).

Table 2.

Mean biologically equivalent dose according to treatment group.

	Primary SABR	Salvage SABR	p
All Treated Lesions
EBRT BED Gy 3	N/A	73.0	n/a
EBRT BED Gy 10	N/A	49.3	n/a
SABR BED Gy 3	158.7	77.2	p<0.01
SABR BED Gy 10	69.5	38.3	p<0.01
Spine (n=52)
SABR BED Gy 3	88.8	61.4	p<0.01
SABR BED Gy 10	39.6	31.7	p<0.01
Lung (n=27)
SABR BED Gy 3	254.6	194.7	p<0.01
SABR BED Gy 10	108.6	83.8	p<0.01

Spinal lesions treated with conventional EBRT prior to salvage SABR received a mean total dose of 35 Gy with average fraction size of 3.3 Gy. Primary SABR to spinal lesions was given in 1 to 5 fractions (median 1) to a mean total dose of 18.5 Gy yielding mean BEDs of 88.8 Gy₃ and 39.6 Gy₁₀. Comparatively, salvage spinal SABR was given in 1 to 5 fractions (median 3) to a mean total dose 18.9 Gy; the mean BEDs were 61.4 Gy₃ and 31.7 Gy₁₀ .

Conventional EBRT for lung metastases was prescribed to a mean total dose of 46.3 Gy in 2.2 Gy fractions. Primary lung SABR was prescribed to a mean total dose of 46 Gy in 3 to 5 fractions (median 4) yielding mean BEDs of 254.6 Gy₃ and 108.6 Gy₁₀. Salvage SABR delivered to lung lesions was given in 3 to 5 fractions (median 4) to a mean total dose of 39.3 Gy; mean BEDs were 194.7 Gy₃ and 83.8 Gy₁₀.

3.2. Clinical Endpoints

Median survival for all patients was 12 months (95% confidence interval [CI] 7-24) and median time to local recurrence was 31 months (range, 18-unbounded). Actuarial local control and overall survival were 83% (95% CI 71-91%) and 50% (95% CI 39-61%) at one year. Eighty-eight percent of symptomatic lesions were improved following SABR (crude rate); 87% and 94% of patients with symptoms improved after SABR with or without prior EBRT (p=0.40). Treatment type (salvage versus primary SABR) and SABR BED > 75 Gy₁₀ were the only significant predictors of local recurrence in univariate regression models (Figure 1 and Table 3). Subset analysis for histology and anatomic treatment site as variables did not predict outcome (Table 3). A proportional hazards multivariate analysis confirmed salvage SABR as a significant predictor of local recurrence when adjusted for treatment site and histology (p=0.02). As noted above, BED was strongly correlated with treatment type. No multivariate cox proportional hazards model contained two or more significant factors.

Figure 1.

The proportion of patients without local failure according to treatment type . Tick marks represent patients who were censored.

Table 3.

Hazard ratios for the association of time-to-local failure with clinical predictors.

Variable	Hazard Ratio#	p-value
Primary vs. Salvage SBRT	2.3 (1.1 – 5.0)	0.03
SABR BED > 75 Gy 10	2.6 (1.0 – 7.7)	0.04
RCC*/Melanoma vs. Other	1.7 (0.7 – 3.8)	0.20
Spine vs. Other	1.2 (0.6 – 2.5)	0.70
Adenocarcinoma vs. Other	1.0 (0.5 – 2.1)	1.00

^*Renal cell carcinoma

^#95% CI in parenthesis

3.3. Toxicity

Four patients developed Grade 2 nausea during treatment. Two patients received liver SABR, one patient received SABR for a right lower lobe lung lesion, and one patient was treated for local recurrence in the left nephrectomy bed; all four patients were treated with primary SABR. No other acute toxicities were observed.

Two patients developed pulmonary toxicity suspected to be treatment related. One patient was scored with Grade 5 pulmonary toxicity after receiving SABR to three distinct sites in the left lung. This patient had a history of T3 N1 M0 esophageal cancer treated with neoadjuvant concurrent chemotherapy and radiation therapy (50.4 Gy) followed by surgery. Postoperatively, the patient was dyspneic, required home oxygen, and eventually returned to baseline following a prolonged oral steroid taper. A left upper lobe lung metastasis was successfully treated with four 10 Gy fractions three years later. One year later, two additional metastases in the same lobe were treated concurrently to 48 Gy in four fractions each. Left lung V20 and mean lung dose was 2% and 2.4 Gy for the first course and 19% and 11.2 Gy for the second, respectively. A restrictive ventilatory defect and left upper lobe opacification requiring home oxygen were noted three months after the last course of SABR. Six months after SABR, CT chest revealed a cavitary lesion in the left upper lobe and bronchoscopic lavage and needle biopsy showed no viable tumor. The patient died secondary to respiratory failure.

Grade 2 radiation pneumonitis was diagnosed in a patient who received 54 Gy in 3 fractions to the right upper lobe for metastatic lung squamous cell carcinoma. Right lung V20 was 19% and bilateral lung V20 12%. The patient developed dyspnea without cough and fevers 2 months after treatment. Symptoms resolved following a slow tapering course of oral prednisone. The patient died secondary to neutropenic sepsis while receiving salvage chemotherapy six months after SABR.

One patient with metastatic renal cell carcinoma was treated with EBRT to vertebral bodies T2-T5 to 31.5 Gy, followed 8 months later by SABR for persistent disease at T4 using four fractions of 6.5 Gy each. The patient developed a vertebral compression fracture and underwent kyphoplasty of T5 and T6 nine months later.

No case of radiation-induced myelitis was observed.

4. DISCUSSION

In this study, patients presenting with symptoms from metastatic disease experienced a high rate of palliation with either primary or salvage SABR. Not surprisingly, primary SABR was delivered with a higher BED. When local control was defined as freedom from either clinical or radiographic progression, primary SABR resulted in higher local control compared to salvage SABR delivered following EBRT failure for metastatic lesions. Primary SABR was a significant predictor of local control on multivariate analysis whereas treatment site and histology were not.

As a retrospective review, our report has several limitations. Studies of SABR with multiple anatomic treatment sites are inherently challenging because the dose/fractionation regimen of SABR is not standardized and is typically at the discretion of the treating physician. Patient selection criteria were not standardized. Equivalent accuracy and precision of conditions for SABR treatment delivery cannot be achieved in all patients. For instance, patients writhing in pain or with significant respiratory compromise may not be able to reliably reproduce immobilization from simulation to treatment. Mixed histologies and varied treatment sites may introduce uncontrolled biases. There were not sufficient events to allow for site-specific analyses. The radiobiological principles that underpin the linear quadratic model may not apply to extreme hypofractionation and BED comparisons used in this study [21]. In addition, our results must be evaluated cautiously given that a contemporaneous cohort of patients treated with EBRT alone was not available for analysis.

The prescribed biologically effective dose for salvage SABR was lower than for primary SABR, likely to observe dose constraints for previously irradiated normal tissues. Doses used for salvage SABR, however, were below published thresholds in studies suggesting dose-response relationships. For instance, Ryu et al. demonstrated a trend towards improved symptom control with spine SABR doses ≥ 16 Gy (versus lower) in a single fraction (BED 33.6 Gy₁₀) [13]. In the current series, primary spine SABR had BED 39.6 Gy₁₀ versus 31.7 Gy₁₀ for salvage SABR. Dose response relationships in the treatment of lung metastases are not well established and are commonly extrapolated from outcomes for early stage NSCLC treated with SABR. Onishi et al. demonstrated local recurrence rates of 8% and 43% in patient with early stage NSCLC undergoing SABR with BED >= 100 Gy₁₀ and < 100 Gy₁₀ [20], respectively. Primary SABR for lung metastases in this study had BED 108.6 Gy₁₀ versus 83.8 Gy₁₀ for salvage. Greater dose per fraction has also been shown to improve local control for metastasis-directed SABR [16].

Lower BED may be one explanation for inferior clinical and radiographic local control for salvage SABR. There are several other possible explanations as well. First and foremost, recurrent lesions may be enriched in inherently radioresistant clones, such as cancer stem cells. Lesions treated with salvage SABR following EBRT failure may possess unfavorable biology. Prior EBRT may also affect the tumor and stroma microvasculature and lead to higher levels of tumor hypoxia at the time of salvage SABR. Primary SABR may provide an advantage by delivering a more potent treatment regimen capable of overcoming acquired mechanisms of radioresistance. Nevertheless, given the possible uncontrolled biases, this finding is hypothesis-generating and requires further study.

Rates of local control in the current study were lower compared to contemporary studies reporting on metastasis-directed SABR. At 1-year, local control rates for lung metastases treated with primary SABR are reported between 80-100% [6, 7] compared to 71% in the current study. Rusthoven et al. reported local control of 96% at 2-years with most patients receiving 60 Gy in 3 fractions (BED 150 Gy₁₀) [7]. Using a regimen with a lower BED, 50 Gy in 10 fractions (BED 75 Gy₁₀), Okunieff et al. reported local control of 83% [6]. In both studies, a substantial proportion of patients (83% and 62%, respectively) had limited metastatic disease. An intermediate BED (108.6 Gy₁₀) and more extensive disease burden may in part explain higher rates of recurrence following primary SABR in this study. Moreover, both primary and salvage SABR resulted in superior local control when the BED was greater than 75 Gy_10. In addition, lung metastases from breast cancer have been reported to have relatively higher rates of progression free survival [6, 15] and only one metastatic breast cancer lesion was present in this series.

Spinal metastases treated with primary and salvage SABR had 1-year actuarial local control rates of 70% and 60%, respectively, compared to 90% and 88% as reported by Gertzen et al. [11]. In the study by Gertzen et al, all patients were treated with single-fraction SABR and the mean BED was 60 Gy₁₀ . Two reports examining SABR for spinal metastases were reported by investigators at the MD Anderson Cancer Center [10, 14]. In the first, 38, 15, and 10 patients underwent definitive, postoperative, and salvage SABR, respectively, with a mean BED of 49.7 Gy₁₀ [10]. Overall, actuarial and radiographic 1-year local control was 84% with no differences between the three groups. A more recent report describes a cohort of 14 and 47 patients undergoing definitive and postoperative SABR, respectively, demonstrated an actuarial 18-month local control of 88% [14]. The minimum BED for single fraction SABR was 41.6 Gy₁₀ in this study. Lower SABR dose and a high rate of radiographic surveillance may explain higher rates of recurrence in our study.

Response assessment in patients undergoing palliative SABR is ill defined. Patient reported analgesia most closely reflects the palliative goal of therapy. Failure to achieve durable pain control following spinal SABR was the only factor independently associated with inferior local control in a recently reported series [14]. On the other hand, radiographic assessment may identify treatment failure earlier thereby allowing intervention, such as SABR, before clinical deterioration. That said, anatomic radiographic assessment may be an unreliable measure of viable tumor, particularly in the lung. Metabolic imaging, such as ¹⁸F-FDG-PET, may improve predictive accuracy and remains under active investigation. In this study, both radiographic and clinical signs of progression were included when determining local control. This may have lowered the rate of local control compared to other contemporary studies with less rigorous surveillance.

An uncontrolled factor in this study is the use of systemic therapy. A vast majority of patients received systemic therapy before and/or after radiotherapy while none received concurrent drug therapy with primary or salvage SABR. The number of fractions delivered for EBRT, salvage SABR, and primary SABR in this series were 13.0, 3.2, and 2.9, respectively. While local effects of systemic therapy before or after SABR remain unclear, it is commonly accepted dogma that drug therapy is the only effective treatment for so-called micrometastases that remain below the threshold of clinical detection. In our study, both univariate and multivariate analyses suggest that primary SABR will be an effective and efficient treatment modality for palliation of metastases. Importantly, primary SABR has the potential to offer the patient and care team an agile local treatment option that reduces interruptions in systemic therapy [14, 22].

Evidence is rapidly emerging to support the use of SABR as a treatment modality for numerous primary and metastatic malignant lesions. It is encouraging that a high rate of symptomatic improvement was observed following palliative SABR regardless of prior treatment. This hypothesis-generating study suggests that higher biologically equivalent doses are generally utilized for upfront SABR. While EBRT followed by salvage SABR was associated with inferior local control, treatment site and histology were not.