Safety of thoracic radiotherapy in patients with prior immune-related adverse events from immune checkpoint inhibitors

Abstract

Background:

Immune checkpoint inhibitors (ICIs) and thoracic radiotherapy are increasingly used to treat advanced cancers. Despite data indicating exaggerated radiation toxicities in patients with autoimmune disease, the safety of thoracic radiotherapy in patients with prior ICI-associated immune-related adverse events (irAEs) is undefined.

Patients and methods:

Patients treated from 2014 to 2020 with ICIs were queried for receipt of corticosteroids and radiotherapy. Patients who received thoracic radiation after symptomatic irAEs were assessed for ≥grade 2 radiation pneumonitis (RP). Characteristics predictive of RP were assessed using logistic regression and response relationships were modeled.

Results:

Among 496 assessed patients, 41 with irAE history subsequently treated with thoracic radiotherapy were analyzed. Most irAEs were grade 2 (n = 21) and 3 (n = 19). Median time from irAE onset to radiotherapy was 8.1 months. Most patients received stereotactic body radiation therapy (n = 20) or hypofractionated radiotherapy (n = 18). In total, 25 patients (61%) developed ≥grade 2 RP at a median of 4 months from radiotherapy and 11 months from onset of irAEs. Three months from RP onset, 16 of 24 (67%) assessable patients had persistent symptoms. Among patients with prior ICI pneumonitis (n = 6), five patients (83%) developed ≥grade 2 RP (grade 2, n = 3; grade ≥3, n = 2). The mean lung radiation dose (MLD) predicted for RP (odds ratio: 1.60, P = 0.00002). The relationship between MLD and RP was strong (area under the receiver-operating characteristic curve: 0.85) and showed an exaggerated dose-response. Among patients with an MLD >5 Gy (n = 26), 21 patients (81%) developed ≥grade 2 RP.

Conclusion:

This is the first study assessing the toxicity of radiotherapy among patients with prior irAEs from ICIs. Patients with prior irAEs were found to be at very high risk for clinically significant and persistent RP from thoracic radiotherapy. Careful consideration should be given to the possibility of an increased risk of RP, and close monitoring is recommended in these patients.

INTRODUCTION

Immune checkpoint inhibitors (ICIs) have created a paradigm shift for the treatment of advanced malignancies. Many of these patients will also be exposed to thoracic radiotherapy, particularly with the increasing utilization of stereotactic body radiation therapy (SBRT) in patients with advanced disease.¹ In these patients, radiation pneumonitis (RP) is a serious toxicity that results in substantial morbidity and is a result of a dysregulated inflammatory response.² While multiple studies have established the safety of thoracic radiotherapy with concurrent or sequential ICIs,^3–5 the safety of radiotherapy among patients with prior ICI-induced immune-related adverse events (irAEs) remains undescribed.

irAEs due to ICIs may develop in any organ system and are thought to be a consequence of autoimmunity.⁶ Prior studies have documented exaggerated radiation toxicities in patients with pre-existing autoimmune disease, attributed to a heightened pro-inflammatory state.^7,8 However, it is not currently known if a history of irAEs presents a similar risk. Therefore, we sought to assess and characterize the risk of ≥grade 2 RP among patients with prior symptomatic irAEs subsequently treated with thoracic radiotherapy.

METHODS

Patients, treatment and toxicity assessments

Consecutive patients treated with atezolizumab, nivolumab, ipilimumab, pembrolizumab, durvalumab and tremelimumab between 2014 and 2020, who subsequently received prednisone or methylprednisone followed by radiotherapy were initially queried from an institutional database. Of this cohort, patients were selected for further analysis if they received corticosteroids for the treatment of irAEs and then underwent thoracic radiotherapy targeting nodes, lung parenchyma or pleura. Patients with ≤9-week post-radiation follow-up were excluded. Patients with ≥grade 2 RP were defined as those with new or progressive respiratory symptoms within 12 months of radiotherapy that warranted treatment. Chest computed tomography (CT) imaging was reviewed for radiographic changes involving the radiated field, defined as within the 50% radiation prescription-dose volume. Toxicity grading was per common terminology criteria for adverse events (CTCAE) v. 5.0. This study was completed under an institutional review board approved protocol.

Statistical analysis

Patient and treatment characteristics at the time of radiotherapy were assessed to predict for ≥grade 2 RP using logistic regression with bootstrap resampling using 1000 iterations. The dose distributions were converted into the equivalent dose in 2 Gy fractions (EQD2) assuming α/β = 3 Gy using Computational Environment for Radiotherapy Research⁹ and the fractionation-corrected mean lung radiation dose (MLD) for each patient was extracted for analysis. Variables that presented with a P value <0.005 (Bonferroni-corrected) were considered predictors for which the response relationship models were plotted. Model discrimination was assessed by the area under the receiver-operating characteristic curve (AUC). Kaplan–Meier estimation was used to determine the 6-month RP-free survival. For comparison, the widely-used, and validated QUANTEC RP model,¹⁰ which includes the MLD, was also assessed to predict ≥grade 2 RP in this patient population.

RESULTS

Patients and treatments

Among 496 reviewed patients, 41 patients treated with thoracic radiation after developing irAEs were analyzed. Table 1 shows patient characteristics. Most patients had prior irAEs of grade 2 (n = 21) and grade 3 (n = 19). The most common irAEs included colitis (n = 18), pneumonitis (n = 6) and hepatitis (n = 6). All irAEs were treated with systemic corticosteroids with the median initial prednisone dose of 50 mg daily [interquartile range (IQR): 40–80]; five patients (12%) received additional immunosuppressive drugs (Table 1).

Table 1.

Patient characteristics at time of radiotherapy

	No. (%) of all patients (N = 41)
Median (range) age, years	65 (34–83)
Sex
Female	21 (51)
Male	20 (49)
Cancer type
NSCLC	24 (58)
Melanoma	4 (10)
RCC	3 (7)
SCLC	2 (5)
Sarcoma	2 (5)
Ovarian	2 (5)
Othera	4 (10)
Karnofsky Performance Status
60–70	4 (10)
80	14 (34)
90–100	23 (56)
Smoking history
Never	10 (24)
Former	27 (66)
Current	4 (10)
History of COPD/asthma	6 (15)
Prior ICI therapy
Pembrolizumab	12 (29)
Ipilimumab + nivolumab	11 (27)
Nivolumab	10 (24)
Atezolizumab	5 (12)
Durvalumab	1 (2)
Tremelimumab	1 (2)
Durvalumab + tremelimumab	1 (2)
Prior irAE requiring systemic steroidsb
Colitis	18 (44)
Pneumonitis	6 (15)
Hepatitis	6 (15)
Arthritis	4 (10)
Skin reaction	4 (10)
Myositis	3 (8)
Pancreatitis	1 (2)
Fever	1 (2)
Nephritis	1 (2)
Hypophysitis	1 (2)
Uveitis	1 (2)
Adrenal insufficiency	1 (2)
Prior irAE grade
2	21 (51)
3	19 (46)
4	1 (2)

COPD, chronic obstructive pulmonary disease; ICI, immune checkpoint inhibitor; irAE, immune-related adverse event; NSCLC, non-small-cell lung cancer; RCC, renal cell carcinoma; SCLC, small cell lung cancer.

^aOther cancer diagnoses included lymphoma (n = 1), urothelial (n = 1), endometrial (n = 1) and salivary gland (n = 1).

^bEleven patients had ≥1 irAE requiring systemic corticosteroids. Patients were nominally treated with systemic steroids; five patients received additional immunosuppressive therapies (n = 4 infliximab, n = 1 mycophenolate mofetil).

Patients received thoracic radiation at a median of 8.1 months (IQR: 4–14) from the onset of irAEs at which time 78% (n = 32) of patients had complete resolution of irAE symptoms. Most patients were treated with SBRT (n = 20) or hypofractionated radiotherapy (n = 18) and treatments mostly targeted lesions in the lung parenchyma (n = 28) (Table 2). Most patients received radiotherapy for the management of oligo-progressive or oligometastatic disease (n = 17 and n = 10, respectively) as well as for palliative local therapy (n = 10) (Table 2).

Table 2.

Thoracic radiotherapy characteristics

	No. (%) of all patients (N = 41)
Radiation prescription dosea
Median (interquartile range)	58 (48–84)
Radiation treatment type
SBRT	20 (49)
Hypofractionated	18 (44)
Conventional	3 (7)
Thoracic disease radiation target
Lung parenchyma	18 (44)
Parenchyma and nodes	10 (24)
Thoracic lymph nodes	9 (22)
Pleura	4 (10)
Concurrent systemic therapyb	11 (27)
Concurrent systemic corticosteroidsc	13 (32)
Mean lung dosed
Median (interquartile range)	6 (4–11)

SBRT, stereotactic body radiation therapy.

^aRadiation dose is shown as the biologic effective dose (BED) (α/β 10 = Gy).

^bSystemic anticancer therapy given during or within 14 ±days of radiotherapy, anti-programmed cell death protein 1 (PD-1) therapy (n = 5); cytotoxic chemotherapy (n = 4); tyrosine kinase inhibitor (n = 2).

^cOral steroids taken during radiotherapy course.

^dMean lung dose is shown as the equivalent dose in 2 Gy fractions (α/β = 3 Gy).

Incidence and characterization of RP

Out of 41 patients, 25 (61%) developed ≥grade 2 RP. The rate of ≥grade 3 RP was 15% (n = 6). The 6-months RP-free survival estimate was 49% (95% CI: 33% to 65%) (Figure 1A). The onset of RP symptoms was at a median of 4 months (IQR: 3–6 months) from the start of radiation, 11 months (IQR: 9–20) from the onset of irAEs and 19 months (IQR: 12–27) from starting ICI therapy. Patients with RP presented with symptoms of new or progressive non-productive cough (n = 22), dyspnea (n = 21), wheezing (n = 5) and/or chest tightness (n = 3) and most, 96% (n = 24), were treated with systemic steroids. Chest CT imaging showed new patchy or dense consolidations (n = 15), and ground-glass changes (n = 11) most pronounced in the radiation field. Figure 2 shows representative diagnostic CT images. At 3 and 6 months from the onset of RP symptoms, respectively, 16 of 24 (67%) and 9 of 15 (60%) assessable patients had persistent RP symptoms including dyspnea and/or cough.

Figure 1.

(A) Kaplan–Meier estimation (and 95% confidence bounds) of ≥grade 2 radiation pneumonitis-free survival. (B) Dose-response curve (magenta) of the mean lung dose (MLD) and ≥grade 2 radiation pneumonitis (RP). The observed RP rate is denoted by the yellow square. Also, the QUANTEC model¹⁰ (light pink) is located far below the observed RP data (green dots) indicating the inability of this widely used model to capture the observed RP distribution in this patient population. Note: error bars denote the exact binomial 95% confidence interval in each observed data quintile.

Figure 2. Radiation planning imaging showing the 100% prescription-dose volume (red) and 50% dose volume (blue) in patients treated with (A) conventional definitive, (B) hypofractionated palliative, (C) stereotactic body radiation therapy (SBRT) for oligo-progression, and (D) hypofractionated definitive radiotherapy (top row).

Pre-radiation diagnostic chest computed tomography (CT) imaging at the approximate axial slice (middle row). Diagnostic chest CT imaging at the time of pulmonary symptoms consistent with ≥grade 2 radiation pneumonitis (RP), imaging shows characteristic changes within the radiated lungs (bottom row).

Among patients with prior ICI pneumonitis (n = 6), all had resolution of prior respiratory symptoms at the time of radiotherapy, and five (83%) developed RP (n = 3, grade 2; n = 1, grade 3; n = 1, grade 4). The onset of RP among these patients was a median of 4 months (IQR: 3–6) from the start of radiation and 18 months (IQR: 10–46) from the time of ICI pneumonitis.

Predictors of RP

The fractionation-corrected mean radiation dose delivered to the lungs (MLD) significantly predicted for ≥grade 2 RP (odds ratio: 1.49, range: 1.11–8.11E165, P = 0.00002). No other explored variable including smoking history, concurrent systemic therapy, grade of prior irAEs, presence of irAE symptoms at the time of radiotherapy or the time between irAE or last ICI treatment and radiotherapy predicted for RP (Table 3). Additionally, among the 24 patients with RP, the majority had radiotherapy as the only new therapy introduced before the onset of pneumonitis symptoms. Only four patients had a change in their systemic therapy between the start of radiotherapy and the onset of RP symptoms.

Table 3.

Univariate analysis results

	≥Grade 2 radiation pneumonitis OR (range)	P value
Age at RTa	1.04 (0.94–1.28)	0.14
Sex
Female (ref.)	2.16 (0.28–5.54E44)	0.24
Male
Smoking history
Never (ref.)	3.27 (0.00–4.48E40)	0.12
Former
Current
Concurrent systemic therapy
Yes	1.17 (0.09–5.11E44)	0.51
No
Concurrent systemic corticosteroids
Yes	1.30 (0.08–2.06E44)	0.55
No
Resolution of irAE symptoms
Yes	1.30 (5E-113–2.88E44)	0.55
No
Grade of irAE
Grade 2	1.29 (0.04–7.56)	0.56
Grade 3–4
Time between irAE and RTb	1.00 (1.00–1.01)	0.48
Time between last ICI treatment and RTc	1.00 (1.00–1.01)	0.50
Mean lung dosed	1.49 (1.11–8.11E165)	0.00002

ICI, immune checkpoint inhibitor; irAE, immune-related adverse event; OR, odds ratio; ref, reference; RT, radiotherapy.

^aAge was assessed as a continuous variable.

^bTime between onset of ICI-associated irAE and RT was assessed as a continuous variable.

^cTime between last ICI treatment and RT was assessed as a continuous variable.

^dMean lung dose was assessed as a continuous variable as the equivalent dose in 2 Gy fractions (EQD2) assuming α/β = 3 Gy, which take into consideration dose and fraction size (see Methods).

The relationship between MLD and RP was strong (AUC = 0.85) and the predicted RP rates agreed with observed RP events (Figure 1B.). According to this model, an MLD of only 5 Gy would predict for a 50% risk of ≥grade 2 RP. For comparison, the QUANTEC RP model suggested the RP risk to be between 2% and 12%, considerably underestimating the observed RP event rate in the quintiles ranging from 25% to 100%. Among the 26 patients with an MLD of >5 Gy, 81% (n = 21) developed ≥grade 2 RP.

DISCUSSION

To our knowledge, this is the first study assessing the toxicity of radiotherapy among patients with prior irAEs due to ICIs. We found patients with prior irAEs to be at exquisitely high-risk for developing >grade 2 RP, with 51% developing clinically significant and persistent RP at 6 months. A strong and exaggerated dose-response between MLD and RP was also identified, suggesting an increased sensitivity to RP that is potentially modifiable.

Pneumonitis is one of the most serious toxicities of thoracic radiation. The risk of ≥grade 2 RP is estimated to be between 10% and 25% with fractionated radiotherapy concurrent with platinum-based chemotherapy and/or with sequential ICI therapy.^11,12 However, highly conformal radiotherapy, such as SBRT, is associated with a 10%−15% risk of RP, even with concurrent ICI therapy.^13,14 Therefore, our finding that 61% of patients developed ≥grade 2 RP, despite highly conformal radiotherapy, suggests that patients with prior irAEs are at a unique and increased risk for toxicity from radiotherapy.

The safety of thoracic radiation in patients with prior ICI-induced irAEs has not been previously described. However, multiple studies have addressed the risk of re-challenge with ICIs in patients who developed irAEs. With ICI re-challenge, both the recurrence of the same irAE as well as the development of a different irAE have been described.^15,16 Additionally, there are reports of other at-risk populations after ICI therapy. For example, exposure to the tyrosine kinase inhibitor osimertinib following ICI therapy has been associated with a higher rate of pneumonitis.¹⁷ Further research is necessary, but the development of irAEs may represent immune system dysregulation predisposing these patients to RP, similar to the amplified radiation toxicities seen in patients with autoimmune disease.^7,8,18

We found the MLD to predict the development of RP. The exaggerated dose-response between RP and MLD (even across the fairly-low MLDs) among patients with prior irAE supports the use of strategies to reduce the MLD far below current guidelines. Deep inspiratory breath hold, margin reduction and/or proton therapy should be considered as potential technical adjustments to mitigate the risk of RP and the associated morbidity and risk of even mortality in these patients.^19,20 Additionally, in appropriately selected patients, surgical therapy should also be considered. Future strategies exploring the preventative use of biologic agents to improve the safety profile of thoracic radiotherapy in high-risk patients are warranted.^21,22 Overall, careful consideration should be given to the possibility of an increased risk of RP when weighing the benefits of radiotherapy and close monitoring for RP is recommended.

Limitations of this study include those inherent to retrospective analyses and, therefore, confirmatory studies are necessary. This study is limited by its sample size, but this is a unique patient population poised to grow with the increased use of thoracic radiation in patients with advanced pretreated disease. Additionally, determining the diagnosis of RP was a challenge even before the widespread use of ICIs and is now, undoubtedly, more complex.²³ Prior studies have found variable imaging characteristics in patients with RP and the patients within this series have had multiple exposures.^24,25 However, we present a thorough analysis on a cohort with mostly classic symptoms and imaging evidence for RP.

In conclusion, our data suggest that patients with prior ICI irAEs are a unique subgroup at very high risk for developing clinically significant and persistent RP from radiotherapy. Careful consideration of the increased risk of RP and efforts to minimize the MLD are warranted to mitigate the risk of substantial morbidity.