Immunotherapy in association with stereotactic radiotherapy for non-small cell lung cancer brain metastases: results from a multicentric retrospective study on behalf of AIRO

Abstract

Background

To define efficacy and toxicity of Immunotherapy (IT) with stereotactic radiotherapy (SRT) including radiosurgery (RS) or hypofractionated SRT (HFSRT) for brain metastases (BM) from non-small cell lung cancer (NSCLC) in a multicentric retrospective study from AIRO (Italian Association of Radiotherapy and Clinical Oncology).

Methods

NSCLC patients with BM receiving SRT + IT and treated in 19 Italian centers were analyzed and compared with a control group of patients treated with exclusive SRT.

Results

One hundred patients treated with SRT + IT and 50 patients treated with SRT-alone were included. Patients receiving SRT + IT had a longer intracranial Local Progression-Free Survival (iLPFS) (propensity score-adjusted P = .007). Among patients who, at the diagnosis of BM, received IT and had also extracranial progression (n = 24), IT administration after SRT was shown to be related to a better overall survival (OS) (P = .037). A multivariate analysis, non-adenocarcinoma histology, KPS = 70 and use of HFSRT were associated with a significantly worse survival (P = .019, P = .017 and P = .007 respectively). Time interval between SRT and IT ≤7 days (n = 90) was shown to be related to a longer OS if compared to SRT-IT interval >7 days (n = 10) (propensity score-adjusted P = .008). The combined treatment was well tolerated. No significant difference in terms of radionecrosis between SRT + IT patients and SRT-alone patients was observed. The time interval between SRT and IT had no impact on the toxicity rate.

Conclusions

Combined SRT + IT was a safe approach, associated with a better iLPFS if compared to exclusive SRT.

Key Points.

1. The addition of IT to SRT for NSCLC BM may improve local control.

2. SRT for BM in patients treated with IT is well tolerated with a low radionecrosis rate.

3. Our data suggest that the IT schedule should not be modified when patients receive SRT.

Importance of the Study.

Although SRT is the standard therapy for oligometastatic patients to the brain and IT has a fundamental role in the treatment of metastatic NSCLC, a few data about the use of SRT and IT for the treatment of BM from NSCLC are currently available. To our knowledge, this is the largest existing series of patients with BM from NSCLC who were treated with SRT + IT. Despite the limitations due to its retrospective nature, our data contribute to support the feasibility of the combined approach of SRT + IT, showing its safety and efficacy. Until prospective data is available, these data may provide insight and practical guidance for clinicians caring for these patients.

Brain metastases from NSCLC account for as much as 20% of all BM. Approximately 10% of patients with NSCLC initially present with BM and about 40% develop brain lesions throughout their disease.¹ Immunotherapy has been shown to have strong activity in patients with both advanced stage and locally advanced stage NSCLC.^2–6 Although SRT is the standard therapy for patients with 1–4 BM^7,8 and its use may be considered also for patients with more than 4 lesions,^9,10 even specifically for NSCLC,¹¹ little is known about the use of SRT and IT for the treatment of BM from NSCLC.

A multicentric retrospective analysis of NSCLC patients receiving SRT and IT was conducted on behalf of Brain Tumor Group and Thoracic Oncology Group of AIRO (Italian Association of Radiation and Clinical Oncology): our study aimed to collect a relatively large number of cases and evaluate the results in terms of efficacy and safety of the combined treatment.

Materials and Methods

Data about patients treated in 19 Italian radiotherapies (RT) centers were collected. Criteria for inclusion were the following: diagnosis of BM from NSCLC, with a maximum diameter up to 4 cm, treated with concurrent SRT and IT. Only patients receiving administration of IT within 4 weeks of SRT delivery were included. Availability of a proper follow-up (clinical examination, including toxicity evaluation and MRI imaging performed, at least, 3 months after brain SRT) was an essential inclusion criterion. Patients who had received prior RT or received postoperative SRT after surgical metastasectomy were excluded. Data regarding patient demographics, tumor pathology, RT schedule, biologically equivalent dose for an alpha/beta ratio of 12 Gy (BED12),¹² radiological responses, and toxicity assessment were collected. Similarly, data about patients treated with SRT for BM that was diagnosed during systemic therapy different from IT or during follow-up were collected as a control group. All the patients of the control group were treated at the coordinating center (Azienda Ospedaliera Universitaria Careggi, Florence). The study was approved by the Ethical Committees of all the participating centers.

BM was assessed using a contrast-enhanced T1-weighted 1 to 3 mm-slice thickness MRI imaging. SRT consisted of RS if delivered in a single fraction or HFSRT if delivered in 3 to 5 fractions. Gamma Knife RS (Elekta Inc., Stockholm, Sweden), CyberKnife SRT (Accuray Inc, Sunnyvale, CA), or linear accelerator (LINAC)-based SRT were used according to the available equipment of the participating centers. The Gross Tumor Volume (GTV) was defined as the contrast-enhanced lesion at the T1-weighted MRI sequence. A margin of 1 to 3 mm was added for patients treated with LINAC SRT, a margin of 0 to 3 mm was used for CyberKnife SRT, whereas PTV was equal to GTV for patients treated with Gamma Knife RS.

Patients receiving combined modality treatment (SRT + IT) were stratified into two groups according to the timing of the combined treatment: the interval between SRT and IT ≤7 days versus >7 days. Patients who began IT at the diagnosis of BM were stratified according to the treatment sequence: IT administration before receiving SRT versus IT delivery after receiving SRT.

Medical charts and follow-up MRI imaging of the patients were collected: data about the neurological status and possible side effects of the treatment were registered. Participating centers were asked to use i-RANO criteria¹³ to define progressive disease in combined treatment patients, while for the control group we used the RANO criteria.¹⁴

Radiation-induced adverse events were reported according to the Common Terminology Criteria for Adverse Events (CTCAE v 5.0).¹⁵ In particular, we focused on the development of brain radionecrosis (RN) (Table 1), intracranial hemorrhage, and neurological symptoms. When radionecrosis was suspected, advanced imaging modalities such as perfusion MR imaging, MR spectroscopy, diffusion tensor imaging, and FDG-positron emission tomography were used to differentiate between tumor relapse and radiation necrosis.

Table 1.

CNS Radionecrosis Classification According to CTCAE Version 5.0

CNS Radionecrosis	Grade 1	Grade 2	Grade 3	Grade 4	Grade 5
Definition: a disorder characterized by a necrotic process occurring in the brain and/or spinal cord.	Asymptomatic; clinical or diagnostic observations only; intervention not indicated	Moderate symptoms; corticosteroids indicated	Severe symptoms; medical intervention indicated	Life-threatening consequences; urgent intervention indicated	Death

Statistical Analysis

The study population and the control group were compared using the Fisher’s exact test and the Mann–Whitney test for categorical and continuous variables, respectively.

Kaplan Meier method was used to estimate intracranial local progression-free survival (iLPFS), intracranial distant progression-free survival (iDPFS), and overall survival (OS). Progression-free survival was calculated from the date of SRT to the last follow-up or the date of progression; recurrent disease in the site of SRT was considered an intracranial local relapse, whereas the appearance of a new brain lesion was defined as intracranial distant relapse. OS was computed starting from the date of SRT until the last follow-up or until the death.

Differences in survival between groups were evaluated using the log-rank test. Univariate (UVA) and multivariate (MVA) models were fitted using the Cox proportional hazard regression analysis. A propensity score was calculated and added as an additional covariate to Cox regression models in the attempt to compensate for different patients having different probabilities of being assigned the treatment under investigation, which may often occur in observational studies. Given the limited sample size, we preferred a priori the use of PS as a covariate in the model over PS matching because the latter method may entail a nonnegligible loss of statistical power if there are many unmatched patients. Of note, the approach adopted here has been previously shown to be as reliable and efficient as other PS-based methods, including PS matching.¹⁶ A two-tailed P < .05 was considered statistically significant.

Differences in terms of toxicity rate were calculated with Fisher’s exact test.

All the statistical tests were performed using Stata version 14 (Stata Corp LP, College Station, TX).

Results

Patients Characteristics

One hundred patients with a total of 163 BM treated between March 2015 and November 2018, were included in our analysis (Table 2). The median age was 65 (range 33–78). Notably, 81% of patients had a diagnosis of adenocarcinoma: none of them presented a mutation of EGFR, only 2 were ALK re-arranged, whereas 44% of the cases presented an expression of PDL-1 ≥1%.

Table 2.

Characteristics of the Patients in the Combined Treatment Group and the Control Group

	SRT + IT (n = 100)		SRT (n = 50)
	N	%	N	%	P-Value
Gender					1.000
Male	63	63.0%	31	62.0%
Female	37	37.0%	19	38.0%
Age (median, interquartile range)	65	(59–70)	66	(60–72)	.751
Age category					.568
<70 years	73	73.0%	34	69.4%
≥70 years	27	27.0%	16	30.6%
Smoking					.789
Nonsmoker	25	25.0%	12	24.0%
Smoker	65	75.0%	31	62.0%
Unknown	10	10.0%	7	14.0%
Histology					.095
Adenocarcinoma	81	81.0%	46	92.0%
non adenocarcinoma	19	19.0%	4	8.0%
EGFR					.255
None	73	73.0%	24	48.0%
Any	0	0.0%	1	2.0%
Unknown	27	27.0%	25	50.0%
ALK					.510
None	66	66.0%	17	34.0%
Any	2	2.0%	1	2.0%
Unknown	32	32.0%	32	64.0%
PDL1					<.001
None	6	6.0%	19	38.0%
Any	44	44.0%	3	6.0%
Unknown	50	50.0%	28	56.0%
KPS performance status grade					<.001
100	40	40.0%	3	6.0%
90	35	35.0%	32	64.0%
80	22	22.0%	14	28.0%
70	3	3.0%	1	2.0%
Extracranial disease control at SRT					.005
No	59	59.0%	17	34.0%
Yes	41	41.0%	33	66.0%
Primary disease control at SRT					.388
No	48	48.0%	20	40.0%
Yes	52	52.0%	30	60.0%
Number of brain metastases
Single	50	50%	20	40%
>1	50	50%	30	60%	.298
Lung-mol-GPA (median, interquartile range)	2	(1.5–2.5)	2	(2–3)	.202
Fractionation					<.001
RS	60	60%	46	92%
HFSRT	40	40%	4	8%
Dose (Gy), total and per fraction (combined)					<.001
>18 Gy, single fraction	48	48.0%	39	78%
<=18 Gy, single fraction	12	12.0%	6	12%
Any Gy, more fractions	40	40.0%	5	10%
BED12
≥40 Gy	92	92.0%	47	94.0%	.752
<40 Gy	8	8.0%	3	6.0%

Abbreviations: RS: radiosurgery; HFSRT: hypofractionated stereotactic radiotherapy; KPS: Karnofsky performance score; SRT: stereotactic radiotherapy; Lung-mol-GPA score: lung molecular diagnosis specific-graded prognostic assessment; BED12: biologically effective dose with alpha/beta ratio =12. Fisher exact test was used for categorical variables; Mann–Whitney test was used for comparing median values of continuous variables. Bold values denote significance at the P < .05 level. Unknown were excluded.

Sixty percent of the cases had a score ≤2 according to molecular diagnosis specific-graded prognostic assessment (lung-mol-GPA).¹⁷ The median number of lesions treated per patient was 2. Sixty patients were treated with RS; among them, 48 patients received a single dose >18 Gy. Forty patients received HFSRT; among them, 33 cases were treated with 3 fractions with a median total dose of 28,9 Gy (range:18–33,75 Gy). Ninety-two patients were treated with a BED12 ≥ 40 Gy.¹²

Nivolumab, pembrolizumab, and atezolizumab were the most used agents (n = 55, n = 41 and n = 4, respectively).

All the 100 patients were further stratified into two groups according to the timing of the combined treatment, irrespective of the sequence: the interval between SRT and IT ≤ 7 days (n = 90) and the interval between SRT and IT > 7 days (n = 10). Forty-one (41%) patients were already receiving IT before the diagnosis of BM: for these patients, SRT was given in addition to the IT that was not modified. The remaining 59 patients began IT at the diagnosis of BM, either before receiving SRT (n = 31) or after receiving SRT (n = 28).

A control group of 50 patients with a total of 139 BM, treated with exclusive SRT between January 2013 and November 2018 at the coordinator center was analyzed. In the control group, the median number of BM was 4, the lung-mol-GPA score was ≤2 in 54% of the cases; most patients received RS with doses >18 Gy (81.3%); cisplatinum and pemetrexed was the most used chemotherapy regimen (52% of the cases).

Compared to patients treated with the combined treatment, patients treated with SRT alone more commonly received single fraction treatment (60% for SRT + IT group vs 92% for SRT alone group; P < .001); moreover, RS was more frequently used with a dose greater than 18 Gy (48% for SRT + IT group vs 81.3% for SRT alone group; P < .001).

The two groups were well balanced for the other most important prognostic factors, excepting for the stratification according to KPS score (P = .001), rate of controlled extracranial disease (41% for SRT + IT group vs 66% for SRT alone group; P = .005) and PDL1 status (P < .001).

Intracranial Progression-Free Survival and Overall Survival

The median follow-up was 23 months and 19 months for the SRT + IT group and the control group, respectively. The total number of MRI scans that were performed was 252 and 197 for the combined treatment group and the control group, respectively. For patients treated with SRT + IT, iLPFS was 89.5% and 83.9%; iDPFS was 69.7% and 55.2% and OS was 79.4% and 64.5% at 6 and 12 months after SRT, respectively. The addition of IT to SRT significantly improved the iLPFS (propensity score-adjusted HR = 0.32, CI 0.14–0.73, P = .007), while no significant differences were observed in terms of iDPFS or OS between the 2 groups (Figure 1).

Fig. 1.

Intracranial local progression-free survival (iLPFS, a), intracranial distant progression-free survival (iDPFS, b), and overall survival (OS, c) of patients treated with combined treatment SRT + IT (dotted line) or treated with SRT alone (solid line).

Univariate and Multivariate Analysis

The following parameters were included in UVA for patients treated with SRT + IT: gender, age, smoking, histology, number of BM, single or multiple fractions, KPS score, extracranial disease control, primary disease control, Lung-mol-GPA, PDL1 status, SRT dose; BED12 and interval time between SRT and IT (Table 3).

Table 3.

Univariate Analysis

			Local Progression-Free Survival (iLPFS)				Distant Progression-Free Survival (iDPFS)				Overall survival (OS)
	N	%	HR	Lower	Upper	P-Value	HR	Lower	Upper	P-Value	HR	Lower	Upper	P-Value
Gender
Male	63	63.0%	1.00				1.00				1.00
Female	37	37.0%	0.45	0.12	1.67	.232	1.05	0.54	2.06	.887	0.50	0.24	1.04	.065
Age category
<70 years	73	73.0%	1.00				1.00				1.00
≥70 years	27	27.0%	0.95	0.26	3.54	.945	0.55	0.23	1.33	.183	1.48	0.73	3.04	.279
Smoking
Nonsmoker	25	27.8%	1.00				1.00				1. 00
Smoker	65	72.2%	1.77	0.37	8.36	.472	0.79	0.37	1.67	.540	1.43	0.65	3.16	.373
Histology
Adenocarcinoma	81	81.0%	1.00				1.00				1.00
Non-adenocarcinoma	19	19.0%	2.03	0.55	7.57	.290	0.79	0.28	2.26	.661	3.13	1.55	6.34	.001
Number of BM
Single	57	57.0%	1.00				1.00				1.00
Multiple	43	43.0%	1.31	0.42	4.08	.642	1.37	0.70	2.67	.360	0.92	0.47	1.82	.814
RS or HFSRT
RS	60	60.0%	1.00				1.00				1.00
HFSRT	40	40.0%	0.45	4.43	1.41	.559	0.85	0.40	1.78	.660	2.36	1.18	4.71	.015
KPS score
100	40	40.0%	1.00				1.00				1.00
90	35	35.0%	1.29	0.34	4.85	.704	1.24	0.62	2.51	.540	1.13	0.48	2.63	.780
80	22	22.0%	2.96	0.72	12.16	.132	0.61	0.17	2.13	.438	2.81	1.18	6.67	.019
70	3	3.0%	-	-	-	-	1.62	0.21	12.54	.645	6.23	1.30	29.88	.022
Extracranial disease control at SRT
No	59	59.0%	1.00				1.00				1.00
Yes	41	41.0%	1.09	0.37	3.27	.872	0.71	0.46	1.41	.331	0.67	0.33	1.36	.264
Primary disease control at SRT
No	48	48.0%	1.00				1.00				1.00
Yes	52	52.0%	0.86	0..29	2.56	.786	0.78	0.41	1.50	.454	0.65	0.33	1.30	.222
Lung-mol-GPA score
0–2	60	60.0%	1.00				1.00				1.00
2.5-4	40	40.0%	0.56	0.17	1.82	.333	1.50	0.79	2.88	.216	0.42	0.19	0.92	.030
PDL1
Any	44	44.0%	Cannot be calculated (No events in PDL1 neg group)				1.00				1.00
None	6	6.0%					1.13	0.25	5.21	.872	0.38	0.05	2.85	.345
Dose and fractionation
>18 Gy, single fraction	48	48.0%	1.00				1.00				1.00
≤18 Gy, single fraction	12	12.0%	0.77	0.09	6.39	.811	1.01	0.35	2.97	.978	2.00	0.71	5.65	.189
Any Gy, more fractions	40	40.0%	1.36	0.42	4.40	.609	0.85	0.40	1.81	.672	2.73	1.30	5.74	.008
BED12
≥ 40	92	92.0%	1.00				1.00				1.00
<40	8	8.0%	1.10	0.22	5.43	.910	0.80	0.27	2.33	.683	0.94	0.33	2.71	.910
Timing
≤7 days	85	85.0%	1.00				1.00				1.00
>7 days	10	10.0%	1.85	0.40	8.61	.432	1.00	0.30	3.30	.999	1.70	0.70	4.12	.238

Abbreviations: iLPFS: intracranial local progression-free survival; iDPFS: intracranial distant progression-free survival; OS: overall survival; BM: brain metastases; RS: radiosurgery; HFSRT: hypofractionated stereotactic radiotherapy; KPS: Karnofsky performance score; SRT: stereotactic radiotherapy; Lung-mol-GPA score: lung molecular diagnosis specific-graded prognostic assessment; BED12: biological effective dose with alpha/beta ratio = 12. Bold values denote statistical significance at the P <.05 level.

Non-adenocarcinoma histology (squamous cell and large cell carcinoma) was found to be predictive of OS (P = .001). Patients treated with HFSRT showed a worse OS when compared to patients treated with RS (P = .015), and even more, if compared to patients treated with RS for a total dose ≥18 Gy (P = .008). Furthermore, low KPS was associated with worse survival (KPS equal to70 P = .022, KPS equal to 80 P = .019 if compared to KPS 100). Lung-mol-GPA score was also related to OS (Lung-mol-GPA ≤ 2 vs GPA = 2.5–4 P = .030).

Among patients who received IT at the diagnosis of BM (n = 59), patients receiving IT administration after SRT delivery had no significant difference in terms of OS when compared to IT administration before SRT (P = .129). A benefit for patients receiving SRT before IT was statistically significant (P = .037) among the patients with both intracranial and extracranial progression (n = 24).

At MVA, non-adenocarcinoma histology, KPS equal to 70, and use of HFSRT (if compared to patients treated with RS for doses >18 Gy) were confirmed as unfavorable prognostic factors for OS (P = .019, P = .017 and P = .007, respectively). Among patients who received IT at the diagnosis of BM, IT administration after SRT delivery was predictive for a better OS when compared to IT before SRT but the difference was not significant (propensity score-adjusted HR 0.74, CI 0.25–2.15; P = .574). Furthermore, the time interval between SRT and IT >7 days was shown to be related to a shorter overall survival if compared with patients who received SRT ≤7 days before or after IT (propensity score-adjusted HR 6.97, CI 1.64–29.52; P = .008).

Toxicity

The combined treatment IT + SRT was well tolerated: 15, 4, and 1 patient developed grade 1 (G1), grade 2 (. G2), and grade 3 (G3) RN, respectively. No case >G3 was observed (Table 4).

Table 4.

Radionecrosis Rate

	SRT + IT (n = 100)	SRT Alone (n = 50)	Fisher Exact Test
Any grade RN	20	11	P = .83
No RN	80	39
≥G2 RN	5	2	P = 1.0
No RN or G1 RN	95	48
	RS + IT (n = 61)	HFSRT + IT (n = 39)	Fisher Exact Test
Any grade RN	15	5	P =.20
No RN	46	34
≥G2 RN	3	2	P =1.0
No RN or G1 RN	58	37
	Time Interval SRT-IT ≤7 days (n = 90)	Time Interval SRT-IT >7 days (n = 10)	Fisher Exact Test
Any grade RN	19	1	P = .68
No RN	71	9
≥G2 RN	5	0	P = 1.0
No RN or G1 RN	85	10

Abbreviations: RN: radionecrosis; G: grade.

In the control group 11 out of 50 patients experienced G1 or G2 radionecrosis (G1 n = 9; G2 n = 2). No other ≥G2 neurological toxicity was observed both in the combined treatment group and in the control group.

Considering any grade of toxicity, there was no significant difference in terms of RN rate between patients who received the combined treatment and patients who received SRT alone. No difference between patients who received RS + IT and patients who received HFSRT + IT was observed. In addition, the time interval between IT and SRT (≤7 days vs >7 days) did not have an impact on the toxicity rate.

When focusing only on ≥G2 RN, there was no significant difference between all the above-mentioned subgroups as well. Three cases presented severe IT-induced toxicity (n = 1 G3 hypothyroidism; n = 1 G3 increasing of bilirubin and n = 1 bowel perforation). The first two patients suspended IT until the adverse event (AE) recover to G1, while the third patient interrupted IT. No other severe AE due to IT was observed in our population. At the moment of the final analysis, 18 patients were still on IT treatment, whereas 81 discontinued for progressive disease.

Discussion

Approximately 50% of NSCLC patients develop BM during their disease, with a life expectancy of about 7 months starting from the diagnosis of brain lesions^{1–11,13,17,18}

Although SRT (RS or HFSRT) is the standard therapy for brain oligometastatic^7–9 and exclusive IT has shown to be active against intracranial disease,^18–20 the specific role of IT in association with SRT for the treatment of patients with BM is not well defined.

Notably, the biological rationale for using the combination of RT and IT is very strong, with evidence of a mutual synergistic effect. RT may promote anti-tumor immunity through different mechanisms, resulting in a significant increase of the antitumor effect of IT.^21–25 Firstly, radiation treatment causes the release of tumor antigens and proinflammatory signals. Secondly, the effects of ionizing radiations can also promote the cross-presentation of tumor-derived antigens by dendritic cells to T cells, leading to anticancer responses. Furthermore, the use of higher doses per fraction, as happens in SRT, may damage the blood-brain barrier and, consequently, facilitate the diffusion of systemic agents into CNS and the infiltration of the immune cells. Intriguingly, there is also some evidence that the use of hypofractionation may have a favorable impact on pro-immunogenic response.^24,26–28 On the other hand, IT may enhance the effect of RT: recent studies demonstrated that IT may modulate the tumor microenvironment normalizing the tumor vessels, thus tumor hypoxia may be reduced and radiosensitivity may increase²⁹; moreover, the use of IT may harness the immune stimulatory of radiation while mitigating its immune-suppressive effects.²⁴

Nevertheless, even though IT has revolutionized the therapeutic landscape of NSCLC,^2–6 published clinical data about the combination of SRT and IT in patients with BM from NSCLC are still scarce. Therefore, the Brain Tumor Group and the Thoracic Oncology Group of the Italian Association of Radiotherapy and Clinical Oncology proposed this retrospective study. The primary objective of this study was to understand whether the addition of IT to SRT may have an impact on the outcome and toxicity of patients with BM from NSCLC.

To our knowledge, this is the largest published series focusing on patients with BM from NSCLC who were treated with SRT + IT (Table 5).

Table 5.

Existing Series About SRT + IT for BM Including Patients With NSCLC

Author	Study Type	Primary Cancer	Total Patients N	SRT + IT N of patients	Control Group (SRT alone) Patients N	SRT: Total Dose Gy/Number of Fractions	IT Agent	iLPFS	iDPFS	OS	Toxicity SRT + IT
Kowalski et al.30	Retrospective	NSCLC Other	97 82	Any primary: 36 NSCLC: 22	143	15–24/1 Dose not specified for HFSRT	Ipi Pembro Nivo Atezo Durva	Any primary	Any primary	Any primary	1-y RN ≥G2 SRT + IT: 3.4% SRT: 7.1% P = NS
								1-year-iLPFS SRT: 90.1% SRT + IT: 100% P = .02	1y-iDPFS SRT: 44% SRT + IT: 59% P = NS	1y-OS SRT: 65% SRT + IT: 64% P = NS
Lanier et al. 33	Retrospective	Melanoma NSCLC	45 226	Any primary: 101 NSCLC: 73	Any primary: 170	16–20/1	Ipi Pembro Nivo Atezo	Any primary:	Any primary:	Any primary:	NR
								1-year-iLPFS SRT: 4% SRT + IT: 9% P = NS	1-year-iDPFS SRT: 34% SRT + IT: 54% P < .01	Median OS SRT: 6.1 m SRT + IT: 15.9 m P < .01
Chen et al. 35	Retrospective	Melanoma NSCLC RCC	70 157 33	Any primary: 79 NSCLC: 37	Any primary: 181	15–24/1 18–24/3 25/5	Anti CTLA-4 Anti-PD1	Any primary:	Any primary:	Any primary:	Acute CNS toxicity - SRT alone: G3: 3% G4: 1% - SRT-IT interval >2 weeks: G3: 0 G4: 0 - SRT-IT interval ≤ 2 weeks: G3: 3% G4: 0 p: NS Overall RN: 3% with no significant difference among groups
								1-year-iLPFS - SRT alone: 82% - SRT-IT interval >2 weeks: 79% - SRT-IT interval ≤ 2 weeks: 88% P = NS	Median iDPFS - SRTalone: 10.2 m - SRT-IT interval >2 weeks: 5.1 m - SRT-IT interval ≤ 2 weeks: 11.5 m P = NS	Median OS - SRT alone: 12.9 m - SRT-IT interval >2 weeks: 14.5 m - SRT-IT interval ≤ 2 weeks: 24.7 m P = .002
Weingarten et al. 39	Retrospective	Melanoma NSCLC RCC Breast	25 23 8 1	Any primary 57 NSCLC: NR	No	Median 20/1	Nivo Ipi Pembro Durva Treme	NR	NR	Any primary: 32 m	RN G2: 5% RN G3: 2%
Colaco et al. 40	Retrospective	Melanoma NSCLC RCC Breast Colorectal Other	56 71 16 27 7 3	Any primary: 42 patients receiving any IT before SRT or within 6 months after SRT NSCLC: NR	Any primary: 138 patients receiving no IT before SRT or within 6 months after SRT	15–24/1	Anti CTLA-4 Anti-PD1	NR	NR	Any primary	Patients receiving any IT before SRT or within 6 months after SRT had a significantly increased risk of radiological or biopsy-proven RN on univariate analysis, not confirmed at multivariate analysis
										Patients receiving any IT before SRT or within 6 months after SRT: 9.3 m Patients receiving before SRT or within 6 months after SRT: 10.0 m
Martin et al. 41	Retrospective	Melanoma NSCLC RCC	145 294 41	Any primary: 115 NSCLC: 38	Any primary: 365	18–20/1 25–30/5	Ipi Nivo Pembro	NR	NR	NR	RN ≥ G2 SRT + IT: 20% SRT: 7% P < .01
Hubbelling et al. 42	Retrospective	NSCLC	163	35	113	10–22/1	Nivo Pembro Atezo Others	NR	NR	NR	All AE - SRT >4 weeks before IT: 10% - SRT-IT interval ≤ 4 weeks: 7% - SRT >4 weeks after IT: 0 p:NS
Ahmed et al.34	Retrospective	NSCLC	17	17	No	18–24/1 25/5	AntiPD1 AntiPDL-1	NR	6 m-iDPFS 48% 6 m-iDPFS: - SRT + IT (interval SRT-IT ≤ 4 weeks) or SRT prior IT: 57% -SRT after IT: 0% P =.05	Median OS: 5.6m OS: SRT + IT (interval SRT-IT ≤ 4 weeks) or SRT prior IT vs SRT after IT P =.006	RN ≥ G2: 0%
Schapira et al.37	Retrospective	NSCLC	37	37	No	20–25/2–5 15–18/1	Nivo Pembro Atezo	1y-iLPFS - SRT + IT (interval SRT-IT ≤ 4 weeks) or > 4 weeks after IT: 100% - SRS > 4 weeks before IT: 72.3% P = .016	1y-iDPFS - SRT + IT (interval SRT-IT ≤ 4 weeks): 61.5% - SRS > 4 weeks before IT: 34.2% - SRT > 4 weeks after IT: 0% P = .04	Median: 17.6 1y-OS - SRT + IT (interval SRT-IT ≤ 4 weeks): 87.3% - SRT > 4 weeks before IT: 70% - SRT > 4 weeks after IT: 0% P = .008	Radiological or biopsy-proven RN: - SRT + IT (interval SRT-IT ≤ 4 weeks): 3% - SRT > 4 weeks before IT: 5%
Singh et al.31	Retrospective	NSCLC	85	39	46	12–24/1	Ipi Pembro Nivo Atezo	Median iLPFS SRT + IT: 3.6 m SRT: 6.4 m P = NS	Median iDPFS SRT + IT: 4.67 m SRT: 6.17m P = NS	Median OS: 11.6 SRT + IT: 10 SRT: 11.6 P = NS - SRT + IT (interval SRT-IT ≤ 4 weeks): 10m - SRT + IT (interval SRT-IT > 4 weeks): 12.1 m P = NR	Any grade RN SRT + IT: 10.2% SRT: 10% P = NS
Shepard et al.32	Retrospective	NSCLC	51	17	34	NR	NR	1y-iLPFS SRT + IT: 84.9% SRT: 76.3% P = NS	SRT + IT vs SRT P = NS	SRT + IT vs SRT P = NS	Any grade RN or intralesional hemorrhage SRT + IT: 5.9% SRT: 2.9% P = NS
Present series	Retrospective	NSCLC	150	100	50	15–24/1 18–33.75/3 20–32/4 30–40/5	Pembro Nivo Atezo	1y-iLPFS SRT + IT: 83.9% SRT: 57.5% P = .002	1y-iDPFS SRT + IT: 55.2% SRT: 59.5% P = NS	1 y OS SRT + IT: 64.5 % SRT: 67.5% P = NS	RN ≥ G2: SRT + IT: 5% SRT: 4% P = NS

Abbreviations: N: Number; SRT: Stereotactic Radiotherapy; EBRT: External Beam Radiotherapy; CHT: chemotherapy; IT: Immunotherapy; BM: Brain Metastases; NR: Not Reported; WBRT: Whole-Brain Radiotherapy; PBI: Partial Brain Irradiation; RT: radiotherapy; NSCLC: Nonsmall cell lung cancer; RCC: Renal Cell Carcinoma; AE: Adverse Event; RN: Radionecrosis; Nivo: Nivolumab; Pembro: Pembrolizumab; Atezo: Atezolizumab; Ipi: Ipilimumab; Durva: Durvalumab; Treme: Tremelimumab; CNS: Central Nervous System; iLPFS: intracranial Local Progression-Free Survival; iDPFS: intracranial Distat Progression-Free Survival; OS: Overall Survival; 1-y: 1 year; m: months; NS: not significant.

Our findings seem to confirm the synergistic effect of the combination of SRT + IT, showing that the addition of IT resulted in a significantly longer iLPFS (1y-iLPFS: 57.5 % vs 83.9 % in patients treated with SRT alone vs SRT + IT, respectively, P = .007). Notably, the control group in our study had a relatively low iLPFS if compared with prospective data from the Alliance trial from Brown et al.⁸: this difference may be explained with the retrospective nature of our data and considering that N0574 trial included BM from all primaries and excluded lesions bigger than 3 cm in diameter and located in critical sites, whereas in our series, focused on NSCLC metastases, both lesions up to 4 cm and lesions in critical sites were included. Kowalski et al.³⁰ found a similar significant advantage in terms of iLPFS (1y-iLPFS: 90.1% vs 100% in patients treated with SRT alone vs SRT + IT, respectively) but their results did not focus on BM from NSCLC only, including BM from other primaries. We found only two series including NSCLC patients exclusively, that compared the results of the combined treatment to the outcome of a control group treated with SRT alone^31,32: none of them found a significant difference in terms of iLPFS.

We found no significant difference between patients treated with SRT + IT and patients treated with exclusive SRT either in terms of iDPFS or OS. Similarly, Kowalski et al.³⁰ and Singh et al.³¹ did not report on a significant benefit in terms of freedom from new brain metastases or survival. Conversely, Lanier et al.³³ found a better iDPFS (1y-iDPFS: 34% vs 54% in patients treated with SRT alone vs SRT + IT, respectively, P < .01) and a longer OS (median OS: 6.1 months vs 15.9 months in patients treated with SRT alone vs SRT + IT, respectively, P < .01) but, notably, in their large series both melanoma and NSCLC patients were included. Noteworthy, in our series, only 41% of the patients treated with SRT + IT had extracranial disease control versus 66% for patients treated with SRT only. The fact that survival was similar between those two groups although a larger number of patients in the combined treatment group had uncontrolled disease outside of the brain, may confirm the value of adding immunotherapy.

The prognostic value of several clinical factors was investigated at UVA and MVA. Given that Lung-mol GPA¹⁷ permits giving a score also to patients who have an unknown gene status, we included it in our analysis, despite a large number of missed molecular data. Since Lung-mol GPA has important differences with the original classification¹ in terms of cut-off values and score assignment, we confirmed the choice to use this prognostic system to make our results comparable to other recent or future series.

At UVA, several factors were shown to be related to OS: non-adenocarcinoma histology, use of HFSRT, low KPS score, and Lung-mol GPA were unfavorable factors for OS.

At MVA, non-adenocarcinoma histology, KPS equal to 70, and use of HFSRT were confirmed as unfavorable prognostic factors for OS. Other authors found that the performance status of the patients was an important prognostic factor for patients treated with SRT + IT,^31,33–35 whereas Singh et al.³¹ found that Lung-mol GPA ≥ 1.5 was found to be predictive of better survival. The finding that the use of HFSRT was related to a worse outcome may be explained by the fact that HFSRT is commonly used for treating large lesions or critical sites. Unfortunately, we could not confirm this hypothesis because of the lack of data regarding the volume and location of the treated BM.

Moreover, we investigated whether the timing of the combination of SRT and IT affected the outcome of the treatment. First, we performed a subgroup analysis of the patients who started IT at the time of the diagnosis of BM (n = 59) to define the optimal sequencing of the combined treatment. We found that the delivery of SRT before IT administration resulted in improved overall survival among patients with both intracranial and extracranial progression (n = 24). In a small series of 17 NSCLC patients, Ahmed et al.³⁴ found an improved iDPFS for patients who received SRT before or during IT delivery, if compared to patients who received IT before SRT (6-months iDPFS 57% vs 0%, respectively). The superiority of giving SRT before IT over the opposite sequence, already shown for melanoma BM,³⁶ may be explained with the enhanced immune response because of radiation effects. In other words, our data suggest that IT-naive NSCLC patients with newly diagnosed BM that are candidates to IT and SRT, should receive SRT before the IT administration, especially if the patients have an extracranial disease as well. Secondly, we stratified all the patients of the combined treatment group (n = 100) according to the interval time between SRT and IT and, although the two groups were numerically unbalanced (≤7 days n = 90 vs >7 days =10), we found that patients with the interval between SRT and IT ≤7 days had a longer survival if compared to patients with the interval between SRT and IT >7 days (propensity score-adjusted P = .007). Although the definition of concurrent administration is extremely heterogeneous in the existing literature, other authors found a similar result in favor of concomitant delivery. Chen et al.,³⁵ in a large series including NSCLC, melanoma, and kidney carcinoma patients, showed significant differences for the whole group in terms of intracranial control and OS for patients treated with SRT and IT has given concurrently (ie, ≤2 weeks). Shapira et al.³⁷ found a significant improvement of iLPFS, iDPFS, and OS for patients affected with BM from NSCLC when SRT and IT are delivered within an interval of time shorter than a month. Interestingly, the difference in terms of OS between these patients and patients receiving SRT >4 weeks before or after IT administration remained significant after controlling for age, KPS, Lung-mol GPA, and systemic adverse effects ≥G2. Noteworthy, a large meta-analysis of individual patient data focusing on this issue was recently published.³⁸ The authors analyzed data coming from 534 patients with 1570 BM from melanoma, NSCLC, and renal cancer, and concluded that concurrent therapy, defined as IT and SRS being administered within one month of one another, is associated with improved outcome [1-year OS: 64.6% and 51.6% (P <.001); 1-year iLPFS: 89.2% and 67.8% (P = .09); 1-year iDPFS: 38.1% and 12.3% for concurrent and non-concurrent therapy, respectively (P = .049)]. In conclusion, our data support the delivery of SRT and IT within a short time frame given that concurrent therapy was shown to be more effective without having an impact on toxicity. This finding may suggest that altering the schedule of IT administration should not be recommended in patients receiving SRT.

Our results showed that the combined treatment is well tolerated without any severe (>G3) neurological toxicity. These results are consistent with other series that reported a ≥G3 toxicity rate ≤3% for patients treated with SRT + IT.^{30,34,35,39,40} In our series differences in terms of toxicity were not found even if any grade toxicity is considered. Conversely, in a series including NSCLC, melanoma, and renal cell carcinoma patients, Martin et al.⁴¹ found that symptomatic (ie, G2) RN occurred in 23 of 115 (20%) and 25 of 365 (6.8%) patients who did vs did not receive IT, respectively.

The comparison to the control group demonstrated that the addition of IT to SRT has no impact on toxicity rate. Other series did not find any significant difference between patients treated with the combined treatment and patients receiving SRT alone.^30–32 Noteworthy, the severe toxicity rate of the SRT + IT arm (G3 ≤3%) did not differ from the treatment-related adverse event rate reported for exclusively SRT by the largest existing series in the literature.^8,9 Furthermore, we did not find any difference in terms of adverse effects between patients treated with RS + IT or HFSRT + IT. To our knowledge, no other series reported on the lack of difference in terms of toxicity between RS and HFSRT when the radiation treatment is associated with IT. Finally, we investigated whether the timing of the combined treatment affected the safety of the treatment and we found that a shorter time interval (≤7 days) between SRT and IT did not impact the frequency or severity of neurological adverse effects. Other authors drew the same conclusion but with a longer time interval between SRT and IT, ranging from 2 weeks³⁵ to 4 weeks.^36,42 This is another important point for the clinical practice: given that concurrent treatment may be the most effective option, the fact that the concomitant administration is also safe, suggests that interruption of the IT should not be recommended in patients receiving SRT.

The most relevant limitation of this study is its retrospective nature with a high risk of selection bias and consequences for the reliability of toxicity data. Nevertheless, it should be noted that we tried to exclude strong confounding factors in our retrospective series. A lot of existing studies reported results of patients affected with BM from different primaries^{30,33,35,39–41} whereas we focused on BM from NSCLC, considering that the natural history of BM may be dramatically different according to the different primary disease.¹ We also excluded patients who had previously received any radiation treatment to the brain, while some authors included also patients who were previously treated with WBRT^{30,31,34,39,40,42}; similarly, we excluded patients who received postoperative SRT after surgical resection of a BM whereas this subgroup of patients was seldom included.^30,35,42 Furthermore, we collected data about local brain control and freedom from new BM while on the contrary, not all the literature series reported data about local intracranial control^{34,39,41–43} and distant intracranial control.^{35,39,41–43} In addition, only patients that received SRT within 28 days interval from IT delivery were included, whereas some studies enrolled patients who had received IT within longer time intervals (ranging from 3^30,32 or 6⁴⁰ months to, even, 13,3 months³⁹). In addition, we scored the toxicity according to CTCAE v 5.0 whereas other series did not report on the grade of toxicity either for RN or for other types of side effects.^36,41,42 Lastly, the presence of a control group treated with SRT alone is another strong point of our study, considering that a control group is lacking in most studies.^{31,34,36,39,42,43}

Other drawbacks of our study were: firstly, the lack of volumetric data for a better correlation between dose, volume, and response; secondly, the use of different advanced imaging modalities to diagnose RN; third, the large proportion of patients with unknown EGFR, ALK, and PDL1 status.

Nevertheless, we believe that, until prospective data is available, these data may provide insight and practical guidance for clinicians caring for these patients.

Conclusions

To our knowledge, this is the largest published series of patients with BM from NSCLC treated with SRT + IT. Despite the limitations due to its retrospective nature, our data contribute to support the feasibility of the combined approach of SRT + IT, showing efficacy and good tolerability. In case that IT is proposed at the diagnosis of BM, our data suggest an advantage of administering SRT before giving IT, at least for patients with extracranial progression as well.

Given that prospective studies are needed to validate this approach, the Brain Tumor Group and the Thoracic Oncology Group of AIRO are going to conduct an observational study focusing on this topic (Strait-luc trial, registration number NCT04787185), to confirm the efficacy and the safety of the combined treatment for BM from NSCLC.

Funding

All authors have no funding or financial support to declare.

Conflict of interest statement. All authors have no conflict of interest to declare.

Authorship statement. Processing of experimental design: S.S. Experimental design implementation: S.S. and E.O. Analysis and interpretation of the data: S.S., E.O., and S.C. Writing of manuscript at draft: S.S. and E.O. Revision of manuscript and approval of its final version: S.S., E.O., V.P., M.F.O., R.D.F., S.C., P.A., P.M., D.F., C.M., G.B., F.P., A.B., P.T., E.G., P.C., A.M., S.P., M.T., M.K., N.G.-L., I.D., G.P., P.M., E.M., L.F., P.N., U.R., V.S., L.L.