Improved time to disease progression in the brain in patients with melanoma brain metastases treated with concurrent delivery of radiosurgery and ipilimumab

ABSTRACT

Background: To identify the optimal sequencing and timing of immunotherapy (IT) and stereotactic radiosurgery (SRS) for melanoma brain metastases (MBMs).

Methods: The elapsed days between IT and SRS were correlated with local control (LC), regional brain control (RBC), time to CNS progression (TTP_CNS), overall survival (OS), and radiation necrosis (RN). Logistic regression and Cox proportional models were used for statistical analysis.

Results: Twenty-five patients with 58 MBMs underwent SRS and IT. Median follow-up was 22.7 mo (3.1–77.9 mo). A median of 2 SRS treatments of 21 Gy (range 16–24 Gy) and 4 cycles of Ipilimumab were delivered. SRS was delivered Before, After or Concurrently with IT in 9, 5, and 11 patients, respectively; 8/25 received SRS ≤30 d of IT and 17/25 were >30 d of IT. Median OS was 35.8 mo, 1- and 2-y OS was 83% and 64%, respectively, and LC was 94.8%. By timing, RBC and TTP_CNS were significantly improved when SRS was delivered ≤30 d of IT (75% vs 23.5%, p = 0.03 and median not reached vs 5.7 mo, p = 0.02, respectively). By groups, Concurrent delivery improved TTP_CNS (p = 0.04). The rate of RN was 20.7% (12/58 lesions) and RN was associated with improved OS (HR 0.21, p = 0.01).

Conclusions: High OS was found for MBM treated with SRS and IT compared to historical reports. A significant association for improved RBC and TTP_CNS was found when SRS was delivered concurrently and within 30 d of IT. Occurrence of RN was higher than SRS alone series but significantly associated with improved OS.

KEYWORDS:

• Brain metastases
• immunotherapy
• melanoma
• radiosurgery
• timing

Introduction

The development of melanoma brain metastases (MBMs) is a devastating complication of metastatic melanoma. MBMs are clinically evident in up to 40% of patients and increase up to 73% on autopsy studies.^1,2 Historically, median survival for these patients has been 6.7 mo (range 3–13 mo), varying with the number of cranial metastases and performance status.³ Regional brain control (RBC) has been reported as approximately 35% at 1 y with a reported 3-mo time to progression in the brain.⁴ Consequently, up to 54% of these patients succumb to their intracranial disease.¹

Local treatment of MBM includes surgery, whole brain radiotherapy (WBRT) and stereotactic radiosurgery (SRS). SRS alone leads to high rate (70–80%) of local control (LC) for intracranial metastases⁵ depending on metastases size and number. However, even in selected patients managed with surgery (± postoperative SRS) or SRS alone, median overall survival (OS) still only ranges between 8 and 10 mo.⁶

Recently, targeted therapy utilizing BRAF and MEK inhibitors or immunotherapy (IT) in the form of checkpoint inhibitors have become the current standard of care for systemic treatment of metastatic melanoma leading to improved survival. Immune checkpoint inhibitors (CTLA-4 inhibitor Ipilimumab; PD-1 inhibitors Pembrolizumab and Nivolumab) assist the patient's immune system in overcoming intrinsic immune-evasive tumor mechanisms, resulting in activation and proliferation of tumor-specific T cells. However, the prognosis of patients with MBM is so poor that most Phase III studies have excluded patients with active CNS disease due to their dismal prognosis and concerns that systemic treatments would not penetrate the blood brain barrier (BBB). Only one phase II trial has been published providing proof of principle of Ipilimumab activity in patients with MBM.⁷ Indeed, activated T cells have now been proven to penetrate the BBB,⁸ suggesting immunostimulatory treatments can be a potentially effective treatment against MBM.

Furthermore, standard fractionated radiotherapy (RT) does not appear to trigger antigenic signaling, but the combination of high dose per fraction RT (such as SRS or stereotactic body radiotherapy, SBRT) and IT results in increased antigen targeting, decreased numbers of inhibitory myeloid-derived suppressor cells, and increased levels circulating activated T cells.^9-12 Moreover, the phenomenon of distant tumor shrinkage in unirradiated sites after radiation, known as the abscopal effect, is modulated by the immune system, potentially via cytotoxic T lymphocytes (CTLs).^13,14 Although abscopal responses have been reported in metastatic melanoma for extracranial sites, it is unknown if privileged sites (e.g., within the BBB) can undergo “local” abscopal effects (i.e., long-term RBC improvement with optimized timing of SRS and IT).^9,15-19

Several retrospective studies have already reported that combining IT with high dose/fraction radiation for MBM improves outcomes particularly when SRS is delivered before or during IT thereby providing an antigenic “primer” for ctivated T cells.^4,20-25 Others suggest a range of time during which concurrent treatment may be helpful, such as ipilimumab infusion within 14 d or within 4 mo of SRS.²² However, the exact sequencing and timing of the two modalities to achieve maximum benefit are still unclear. Therefore, in this investigation, we sought to identify if an optimal window of time for IT/SRS co-administration exists that leads to improved brain control. Thus we retrospectively evaluated patients with MBM treated with SRS and we analyzed these patients by groups (Before, After, Concurrent) and by a timing cutoff (≤ 30 d or > 30 d). To our knowledge, is the first study to analyze the timing effect based on specific groups and timing cut-points.

Methods

Patient selection and treatment

Between 2011 and 2015, 25 patients with MBM were treated with SRS and received anti-CTLA-4 agent (Ipilimumab, Ipi). These patients, harboring 58 MBM, were treated on a Novalis® linear accelerator platform (BrainLab; Munich, Germany). Lesions were treated with a SRS median dose of 21 Gy (16–24 Gy) to the planning target volume (PTV) following guidelines based on the Radiation Therapy Oncology Group trial 90-05 (Table 1). All 25 patients received Ipi at a median dose of 3 mg/kg (range 3–10 mg/kg) every 3 weeks for a median of four cycles (range 1–13). At the time of systemic disease progression, six patients received Pembrolizumab (Pembro) at 2–10 mg/kg every 2 or 3 weeks (median 13 cycles, range 3–37); no patients received Pembro alone, and as such, data analysis was not done in relation to Pembro and SRS timing (Table 1).

Table 1. Patient demographics and treatment characteristics.

Patient characteristics
Total patients	N = 25
Total lesions	N = 58
Sex	Male	N = 20
	Female	N = 5
Age	68.5 y	Range 33–85
Anti-CTLA-4 therapy	Ipilimumab	N = 25
	# Cycles	N = 4 (range 1–13)
	Dose	250 mg/kg (range 250–289) q3 weeks
Extent of extracranial disease at time of SRS
Extensive	N = 10	≥ 3 sites
Minimal	N = 10	< 3 sites
NED	N = 5	No extracranial disease
# SRS/patient	N = 2	Range 1–6
Dose, median (range) Gy	21 Gy	Range 16–24 Gy
Treatment type	Linac based	58/58 lesions
Grouping of SRS	Before IT	N = 9
	After IT	N = 5
	Concurrent	N = 11
Timing of SRS	≤ 30 d of IT	N = 8
	> 30 d of IT	N = 17
Other therapy	Number of patients	Dose per fraction (fx)
Surgery before SRS	N = 2	16 Gy/fx
High dose per fraction radiotherapy		Dose per fraction (fx)
SRS alone	N = 10	21 Gy/fx
SRS and SBRT	N = 2	5–6 Gy/fx (sternum, shoulder)
Other systemic therapy prior to diagnosis of brain metastases
BRAF inhibitor	N = 3
Interleukin/interferon therapy	N = 4
Other systemic therapy after systemic failure
Pembrolizumab	N = 6	13 cycles (range 3–37), 2–10 mg/kg q2–3 weeks

Abbreviations: IT, immunotherapy; NED, no evidence of disease; SRS, stereotactic radiosurgery; SBRT, stereotactic body radiotherapy.

Patients were divided into groups based on the timing of SRS in relation to IT infusion (SRS Before IT vs After vs Concurrent). A secondary analysis was done based on the elapsed time between the two treatments with a cutoff of SRS occurring within ≤30 d of IT > 30 d of IT (Fig. 1).

Figure 1. Schema for grouping and timing. (A) Patients that had SRS ±30 d of Ipilimumab infusion (solid box) vs >30 d (dashed box) from Ipilimumab infusion were assessed. Additional analysis was conducted on patients who had (B) SRS Before (C) After or (D) Concurrently (interdigitated) with Ipilimumab.

At the time of failure (brain or systemic), patients received additional treatment. A small cohort (n = 6) received Pembro after systemic failure or additional radiation as detailed in Table 2: 7/25 patients received WBRT (5/7 as salvage therapy at a median of 4 mo after first SRS, and 2/7 received planned, up-front WBRT at diagnosis of MBM). Of note, 11 patients had received additional systemic therapy before SRS, including IL-2 (n = 1), Vemurafenib (n = 3), Temodar (n = 5), and IFN-alpha (n = 4).

Table 2. Characteristics of patients that received whole brain radiotherapy (WBRT).

Patient	WBRT dose	Type of WBRT	Time to salvage WBRT
1	30 Gy/10fx	Salvage	47 d
2	37.5 Gy/15fx	Salvage	189 d
3	37.5 Gy/15fx	Concurrent with SRS	n/a
4	37.5 Gy/15fx	Before SRS	n/a
5	30 Gy/10fx	Salvage	189 d
6	30 Gy/10fx	Salvage	119 d
7	30 Gy/10fx	Salvage	54 d
			Median 119.6 d (∼4 mo)

Intracranial response assessment

T1-weighted, gadolinium-enhanced magnetic resonance images (MRI) with 1-mm slice thickness of the whole brain were obtained prior to SRS treatment and at each follow-up approximately every 6–12 weeks. The TTP_CNS was defined as time to first intracranial progression of any kind. Radiation necrosis (RN) was defined based on neuroradiology interpretation of sequential scans with perfusion imaging and multidisciplinary review; time to RN was backdated from the date of consensus agreement of RN to date of first concerning finding. RECIST 1.1 criteria were used to score CNS and systemic imaging outcomes by CR (complete response), PR (partial response), SD (stable disease) or PD (progressive disease).²⁶

Statistical analysis

The impact of SRS groups (Before, After or Concurrent to IT) as well as timing (≤ 30 days, > 30 days) on four endpoints (RBC, TTP_CNS, OS and RN) was assessed. For LC, RBC and RN, a Logistic Regression model was used to test the association of each of these parameters against time to infusion. The TTP_CNS, RBC and OS were estimated using the Kaplan–Meier method. The association of OS with CNS progression and RN and the association of TTP_CNS and RBC with timing of SRS and IT were assessed using a Cox proportional hazard model. The hazard ratio (HR) and 95% confidence interval (CI) were presented. The p-value < 0.05 was considered significant. All analyses were conducted using SAS V9.1.2 software (Cary, NC).

Results

Local control of SRS treated lesions

With a median follow up of 22.7 mo (range 3–78 mo), an LC of 94.8% (55/58) of the treated lesions was observed, with a CR rate of 38% in 22 lesions, occurring at a median of 5.4 mo (range 1.7–20.5 mo). A PR was seen in 18 (31%), SD in 15 (26%), while only 3 lesions (5%) had local failure at a median time of 9.1 mo after SRS (range 2–43 mo). LC was not, however, associated with improved OS (p = 0.54) based on subgroup analysis. Of note, 97% of lesions were in-tact and 3% were post-operative cavities.

Effect of sequencing and timing of SRS with IT on time to treatment progression and regional brain control

When evaluated by groups, 9 patients received SRS Before IT (median 119 d), 5 had SRS After IT (median 52 d), and 11 patients had SRS Concurrently with IT (median 8 d). When considering all 25 patients, the median shortest elapsed time between SRS administration and IT infusion was done 14 d prior to IT.

The median TTP_CNS for all patients was 16.7 mo (Fig. 2). However, the concurrent delivery of SRS and ipilimumab was significantly associated with improved TTP_CNS compared to the other two groups (p = 0.04) When comparing the Concurrent vs Before groups, a significant difference was noted with a median TTP_CNS of 30.2 mo vs 4.5 mo, respectively (p = 0.02). Furthermore, TTP_CNS for delivery of SRS within ≤30 d of Ipilimumab was significantly better than >30 d (p = 0.02, HR 5.09).

Figure 2. Time to progression in the CNS. (A) All patients, median time to progression = 16.7 mo. (B) By groups: the Concurrent group is statistically better than the Before and After groups, p = 0.04), and (C) by timing of SRS delivery: ≤ 30 d is statistically better than > 30 d, p = 0.02.

RBC was similar between the groups (p = 0.08) with a `1- and 2-y control rates of 52.2% and 34.8%, respectively. When analyzed by timing of administration, however, RBC was significantly improved when SRS occurred ≤ 30 d of IT (75% vs 23.5%, p = 0.03). The 1-y RBC for patients with SRS ≤ 30 d was 83.3% vs 37.4% if there were >30 d between SRS and IT; the 2-y RBC rates were, respectively, 83.3% vs 15%. At last follow up, 15/25 patients had failed intracranially after the initial SRS. Salvage treatment consisted of either repeat SRS to the new lesions (n = 8), WBRT (n = 5), new systemic therapy (n = 5), and one patient opted for hospice.

Overall survival and disease progression

At time of last follow up, 48% (12/25) patients were alive with a median OS of 35.8 mo from date of first SRS to last known follow up. The 1- and 2-y OS rates were 83% and 64%, respectively (Fig. 3). There was no effect of sequencing or timing of SRS/IT on overall survival. At last follow up, 44% of patients had stable systemic disease, 4% had a partial response, and 52% had progressive systemic disease. Six patients received Pembro after systemic progression on Ipilimumab. Despite failing extracranially, this cohort of six patients had an improved TTP_CNS compared to those who never received Pembro (HR 0.21, p = 0.03). This group incidentally had a shorter median time between treatments (7.7 d before IT vs 19 d before IT in the non-Pembro group).

Figure 3. Overall survival analysis for (A) all patients, (B) by group and (C) by timing.

Overall survival and radiation necrosis

Radiographic RN was seen in 12/58 treated lesions (20.7%) in nine patients. The median dose for lesions with RN was 21 Gy (range 16 −24 Gy) and the median time to onset of RN was 14.7 mo (range 8.1–71.7 mo). Symptomatic RN occurred in only 5% of treated lesions, requiring treatment with bevacizumab (Avastin). The presence of RN, however, was associated with an improved survival with a HR of death of 0.21 (CI 0.05–0.77, p = 0.01). Patients without RN had a median OS of 22.6 mo, while the median OS of those with RN was not reached.

Discussion

This investigation has found that concurrent delivery of SRS within 1 mo of Ipi significantly improved the time to intracranial progression and thereby significantly improved regional brain control in patients with MBM. It also reports the longest median OS to date of 35.8 mo in these poor prognosis patients. Regarding timing, our data suggests an optimal window for the co-administration of SRS and IT, with SRS interdigitated between IT cycles by ≤30 d. We believe this result may be secondary to the synergy between SRS and IT, and mechanistically it could be explained by a “locoregional” abscopal effect in the brain leading to less intracranial recurrence and thus less neurologic death.

The mechanism of this synergy appears to be related to the high dose per fraction RT (SRS). This dosing is thought to be immunogenic and thus “primes” the immune system by improving antigen supply as a direct consequence of melanoma cell death.⁴ Specifically, high-dose RT optimizes local tumor microenvironment by upregulating MHC class I molecules for antigen presentation to T cells and increases antigen supply via tumor ablation (in situ vaccination). This interaction unmasks previously occult cancer antigens to T cells, generates tumor specific CTLs and increases tumor infiltrating lymphocytes. We postulate that T cells, activated by IT drugs, then penetrate the BBB and neutralize microscopic deposits of tumors throughout the brain, thus increasing the TTP_CNS, leading to improved RBC.^9-11,27-32

We, along with other authors, believe that this synergy is potentiated by a shorter interval between SRS and IT.^22,35 Mechanistic support for this theory comes from a recent study done in mice which demonstrated that when in vivo melanoma tumor cells were treated with Ipi and RT (10-20Gy/1 fraction) together, the mannose-6-phosphate receptor (MPR) was upregulated and tumor cell growth was significantly reduced compared to IT or RT alone (p = 0.0078).³³ When tumors expressed more MPR, then the effects of Ipi-induced CTLs came to fruition through the binding of their products (granzymes) to their receptors (MPR), thereby killing more tumor cells. This rapid and transient effect of MPR upregulation begins 24 h after SRS, peaks 3 d later, and normalizes to pre-treatment values by day 7, whereas the half-life of ipilimumab is 14.9 d.³⁴ This suggests that there may be an optimal window of time during which to combine the therapies with SRS hypothetically occurring within −6 days or +15 days after Ipi infusion. In our cohort, the median time between SRS and IT infusion was 14 d and this short interval could have driven our high LC and RBC. Other authors have also previously suggested the theory of an optimal window of co-administration as well. Several retrospective studies have shown that SRS occurring “within 4 mo,” “4 weeks” or “14 d” either before or after IT can reduce the volume of metastases and their results have trended toward improved overall survival; however, statistical significance for OS has been difficult to achieve with small, retrospective cohorts, similarly to our study.^22,35

Other potential explanations for the improved survival in this small cohort of patients compared to historic controls are: First, the LC and CR rates achieved were very high, 94.8% and 38%, respectively, which may have led to fewer CNS deaths. Second, additional RT using 5–8 Gy per fraction for extracranial cancer treatment (SBRT) may have played a role in ongoing antigenic priming, as eight of our patients received this modality during their entire treatment course for treatment of systemic disease. Finally, the majority of our patients (60%) had a low burden of extracranial disease at the time of diagnosis of brain metastases perhaps contributing to the lengthening of their OS.

The outcomes reported here are similar to other recent series, although our survival and intracranial control were somewhat higher (Table 3). For example, regarding timing, a recent study showed 1-y OS of 42.9% when Ipi was delivered within 14 d of SRS compared to our results of 83% when IT was given within 30 d of SRS.²² Regarding grouping, our results compare favorably with those of Kiess et al.⁴ who showed a 1-y OS of 65% and a median OS of 12.4 mo. Our data may have been an improvement on Kiess's results because we had a shorter time between treatments and lower extracranial disease burden. Therefore, the data presented in this paper suggests that closer timing of the two modalities potentially allowed for more potent synergy between treatments, leading to better brain control and less death from intracranial progression.

Table 3. Review of recent trials assessing SRS and immunotherapy for the treatment of melanoma brain metastases.

Author, year	# of patients	Best timing of SRS in relation to IT	Median overall survival (mo)	1- and 2-y OS	Regional brain control^*
Knisley et al., 2012	27	Before	21.3	2-y 47%	Not reported
Shoukat et al., 2013	11	Not reported	28.3	Not reported	Not reported
Kiess et al., 2015	46	Concurrent, Before	12.4	1-y 65%	1-y^* 31%
				Concurrent	Concurrent
				56% Before	36% Before
				40% After	8% After
Schoenfeld et al., 2015	16	Before	14.4	1-y 54%	Not reported
				2-y 25%
Patel et al., 2015	20	Within 14 d	9.1	1-y 42.9%	1-y 29.1%
				2-y 42.9%
Ahmed et al., 2016	96	Not reported	9–12^**	1-y 50–66%^**	1-y 21–38%^**
Current study, 2016	25	Concurrent, SRS within ≤ 30 d of IT	35.8 mo	1-y OS 83%	1-y 52%
				2-y OS 64%	2-y 34.8%

* If brain control rates were originally reported as brain recurrence rats, the recurrence rate was converted to a control rate by subtracting it from 100%. Asterisks are noted where this occurred.

** Range reflects values for both anti-CTLA-4 and anti-PD1 agents used.

One concern of combined modality treatment is that it may lead to late toxicity in the form of RN. However, the rate of RN in this study was 20.7%, which falls within reported ranges of 6–24% with SRS alone.^37,38 In fact, it was lower than the RN rates reported by comparable recent series of GammaKnife Radiosurgery showing rates of 30–37.5%.^22,39 Furthermore, only three of our nine patients with RN were symptomatic (5% of lesions), and these patients were successfully treated with Avastin. Interestingly, RN was associated with improved OS in our study, as the patients with RN have not yet reached median OS while those with RN had a median OS of 22.6 mo (HR 0.21, p = 0.01). Our observation on the effect of RN on survival is in line with the results of the Colaco et al. study,³⁹ which also reported higher median OS in the RN group of 23.7 vs 9.9 mo. These are preliminary data that requires more dosimetric and radiographic analysis and will be reported separately.

Limitations of this study include its retrospective, non-randomized nature and small sample size containing biases inherent in retrospective analyses. However, we feel that our study still adds to the body of literature supporting ongoing clinical research investigating radiation and IT timing that maximizes the therapeutic ratio. Finally, systemic treatments for metastatic melanoma are evolving to include patients with brain metastases. Emerging phase II data on 18 MBM patients suggests that Pembro alone can produce a 22% response rate and is safe to use after Ipi failure.⁴⁰ In fact, PD-1 inhibition in combination with SRS may yield even better responses compared to anti-CTLA-4 inhibition in these poor prognosis patients, and drug choice should be taken into consideration when combining RT and IT as well.³⁶

Conclusions

The data presented here suggests that when SRS is delivered concurrently and within 30 d of Ipilimumab infusion, it leads to improved time to CNS progression and regional brain control in patients with MBMs leading to median survival of about 3 y in this group of poor prognosis patients. The incidence of RN appears similar to SRS alone historic controls, and seems to be associated with a low rate of symptomatic RN and surprisingly long overall survival.

Significant conclusions

When SRS is given concurrently with IT, and more specifically within a 30-d window on either side of IT, time to progression in the brain and therefore regional brain control are significantly improved and potentially contribute to the lengthened overall survival seen in this cohort of patients who historically have poor survival. RN may be higher than with SRS alone in this cohort, but it appears to be associated with improved overall survival.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Author contributions

Conceptualization: TS, SS, HC and BS; methodology: TS and BS; formal analysis: HC; investigation: TS; data curation: TS; writing original draft: TS; writing review and editing: TS, BS, SS and HC; visualization: TS, BS and HC; and supervision: BS.