Corneal Substructure Dosimetry Predicts Corneal Toxicity in Patients With Uveal Melanoma Treated With Proton Beam Therapy

International Journal of

Radiation Oncology biology

physics

www.redjournal.org

Clinical Investigation

Corneal Substructure Dosimetry Predicts Corneal Toxicity in Patients With Uveal Melanoma Treated With Proton Beam Therapy Howard J. Lee, Jr, BA,* Andrew Stacey, MD,y Todd R. Klesert, MD, PhD,z Craig Wells, MD,z Alison H. Skalet, MD, PhD,x,k Charles Bloch, PhD,{ Angela Fung, RT(T), CMD,# Stephen R. Bowen, PhD,{,** Tony P. Wong, PhD,# Dean Shibata, MD,** Lia M. Halasz, MD,{ and Ramesh Rengan, MD, PhD{ *Duke University School of Medicine, Durham, North Carolina; yDepartment of Ophthalmology, University of Washington School of Medicine, Seattle, Washington; zVitreoretinal Associates of Washington, Seattle, Washington; xCasey Eye Institute, Oregon Health and Science University, Portland, Oregon; kDepartment of Radiation Medicine, Oregon Health and Science University, Portland, Oregon; {Department of Radiation Oncology, University of Washington School of Medicine, Seattle, Washington; #Seattle Cancer Care Alliance Proton Therapy Center, Seattle, Washington; and **Department of Radiology, University of Washington School of Medicine, Seattle, Washington Received Sep 17, 2018. Accepted for publication Feb 4, 2019.

Summary In patients with uveal melanoma treated with proton beam therapy, we examine the relationships of tumorrelated factors and corneal dose-volume parameters with incidence of grade 2þ intervention-requiring corneal toxicity. Patients with anterior uveal melanomas experience a high rate of grade 2þ corneal toxicity because of increased corneal dose. Anterior location and

Purpose: This study examines the relationship between dose to corneal substructures and incidence of corneal toxicity within 6 months of proton beam therapy (PBT) for uveal melanoma. We aim to develop clinically meaningful dose constraints that can be used to mitigate corneal toxicity. Methods and Materials: Ninety-two patients were treated with PBT between 2015 and 2017 and evaluated for grade 2þ (GR2þ) intervention-requiring corneal toxicity in our prospectively maintained database. Most patients were treated with 50 Gy (relative biological effectiveness [RBE]) in 5 fractions, and all had complete six-month follow-up. Analyses included Mann-Whitney, c2, Fisher exact, and receiver operating curve tests to identify risk factors for GR2þ toxicity. Bivariate logistic regression was used to identify independent dose-volume histogram (DVH) predictors of toxicity after adjustment for the most important clinical risk factor. Results: The 6-month PBT GR2þ corneal toxicity rate was 10.9%, with half of patients experiencing grade 2 toxicity and half experiencing grade 3 toxicity, with no grade 4 events. Patients with anterior chamber tumors had a higher risk (58.3%)

Reprint requests to: Howard J. Lee Jr, 1959 NE Pacific St, Seattle, WA 98195. Tel: (509) 863-5027; E-mail: [email protected] Conflict of interest: L.M.H. reports grants from AbbVie, and A.S. reports personal fees from Castle Biosciences Inc, outside of the submitted work. A.H.S. is funded by Castle Biosciences Inc., grant P30 EY010572 Int J Radiation Oncol Biol Phys, Vol. -, No. -, pp. 1e9, 2019 0360-3016/$ - see front matter Ó 2019 Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.ijrobp.2019.02.005

from the National Institutes of Health (Bethesda, MD), and by unrestricted departmental funding from Research to Prevent Blindness (New York, NY). Supplementary material for this article can be found at https://doi.org/ 10.1016/j.ijrobp.2019.02.005.

2

International Journal of Radiation Oncology  Biology  Physics

Lee et al.

corneal dose-volume histogram parameters independently predict toxicity risk. We propose dosimetric constraints to facilitate treatment planning and toxicity mitigation.

for toxicity than those with posterior tumors (0%) or posterior tumors extending past the equator (25%, P < .0001). On univariate analysis, larger size according to Collaborative Ocular Melanoma Studies was associated with increased toxicity rate (P < .004). DVH analysis revealed that cutoffs of 58% for V25, 32% for V45, 51.8 Gy (RBE) for maximum dose, and 32 Gy (RBE) for mean dose to the cornea separated patients into groups experiencing and not experiencing toxicity with 90% sensitivity and 96% specificity. Bivariate logistic regression indicated that corneal V25, V45, and mean dose independently predicted for toxicity after adjusting for tumor location. Conclusions: Patients receiving PBT for anterior uveal melanomas experience a high rate of GR2þ corneal toxicity because of increased corneal dose. Anterior location and corneal DVH parameters independently predict toxicity risk. We propose dosimetric constraints to facilitate treatment planning and toxicity mitigation. Ó 2019 Elsevier Inc. All rights reserved.

Introduction Uveal melanomas are rare intraocular tumors that arise from the iris, ciliary body, and the choroid and may involve more than 1 of these regions. Choroidal melanomas are more common than ciliary body and iris melanomas, representing 86% of uveal melanomas.1 Uveal melanomas were historically treated with enucleation. In the 1960s, brachytherapy began to be used; by the 1980s, it emerged as a globe-sparing alternative for small and medium tumors under the Collaborative Ocular Melanoma Studies (COMS).2 Beginning in 1975, investigators at Harvard began using external beam proton radiation therapy (PBT) as an alternative definitive globe-sparing treatment modality.3 These shifts provided patients with multiple advantages. Patients maintained reasonably functional vision and a cosmetic advantage. Furthermore, as these tumors did not often undergo biopsy, the small yet possible risk of enucleation in the setting of misdiagnosis was eliminated. Globe-sparing management of melanomas with PBT is possible because of the dosimetric properties of charged particle therapy that allow selective sparing of organs at risk (OARs).4 It is especially desirable in younger patients who have a good prognosis and are at risk for late effects after completion of radiation therapy. PBT has been shown to have excellent control rates, with results indicating local control in over 95% of cases at 15 years, with a 4.1% enucleation rate.5-7 Melanoma-related mortality seems to differ based on patient age and tumor size. Patients with smaller tumors (11 mm) have 20-year mortality under 10%, whereas patients with larger tumors (>11 mm) have 20-year mortality of 33%.8 Although no randomized trials have directly compared PBT with brachytherapy, brachytherapy has been compared with helium ion beam therapy in a randomized trial, which showed improved local control, eye preservation, and disease-free survival for patients treated with charged particle therapy.9 Brachytherapy is also known to provide excellent local control rates (over 90%).10-12 Both PBT and brachytherapy

are offered to patients today with the intent of avoiding enucleation. A significant number of patients, even with large tumors, achieve long-term survival with PBT.13 Ambulatory visual acuity is preserved in a small proportion of patients, and visual decline is most strongly associated with tumor proximity to the optic disc or fovea.14 However, there may be a subset of patients, especially with anteriorly positioned tumors away from critical visual structures, in whom preservation of acuity at higher rates may be achievable. Corneal and ocular surface toxicities, including keratitis, persistent corneal epithelial defects, corneal necrosis, and persistent dry eye syndrome, are known complications of radiation therapy for uveal melanoma, especially in anteriorly positioned tumors. Persistent epithelial defects of the cornea can also be a vision-limiting posttreatment change. In the previous era of model-based treatment planning, corneal dose was difficult to accurately quantify. Treatment planning today has shifted to contour-based planning from a model-based approach, allowing for dose-volume histogram (DVH) analysis of the cornea and ultimately development of optimization approaches to minimize the risk of toxicity. However, dose-volume parametric predictors of corneal toxicities are not yet well described. The primary objective of this study is to examine comparative corneal substructure doses in patients who experienced corneal toxicity within 6 months after treatment with those who did not to develop dose constraints.

Methods and Materials Patients The medical records of 105 consecutive patients with uveal melanoma treated with PBT at a single institution between November 2015 and September 2017 were reviewed under an institutional review boardeapproved retrospective study

Volume -  Number -  2019

of prospectively captured corneal toxicity events. Patients were excluded from analysis if they received treatment to doses and fractionation other than 50 Gy (relative biological effectiveness [RBE]) in 5 fractions or 56 Gy (RBE) in 4 fractions (n Z 2), did not have magnetic resonance imaging (MRI) available (n Z 1), or had follow-up of <6 months (n Z 10). A total of 92 patients were included in this analysis, with baseline characteristics shown in Table 1. Patients underwent a B-scan ultrasound with their referring ocular oncologist at the time of diagnosis. All patients underwent staging computed tomography (CT) of the abdomen to assess for metastatic disease, with attention to the liver. Patients were included in the study if they had a clinical diagnosis of choroidal melanoma limited to the globe (no evidence of extrascleral extension) and no evidence of metastatic disease at time of diagnosis.

Treatment planning Treatment planning was conducted in a manner similar to that previously described by Hartsell et al.15 Patients underwent high-definition MRI scans of the orbit with and without contrast, using thin (1.25-mm) slices in the axial planes and volumetric acquisition. Three tantalum clips were sutured to the sclera of the eye to confirm the positioning of the eye during treatment. The locations of the clips were determined based on a 3-dimensional model of the eye, MRI, and funduscopic images and were placed at least 3 mm away from the tumor to avoid shadow within the target volume for treatment. One week after clip placement, the patients underwent CT simulation in the supine position. For patient treatment, the seated position was preferred unless the patients were too tall (>1.9 m) or too heavy (>129 kg) to be treated in the seated position. Treatment was delivered using fixed beam lines with the patient in a rotating chair. A thermoplastic immobilization mask was constructed, and patients were instructed to fix their gaze to our in-house built optical immobilization device attached to the couch of the CT bore. A thinly sliced (1.25-mm) axial CT scan was obtained with the eyes in the appropriate position; this scan covered the entire cranial contents (to aid in initial patient setup). Images from the treatment planning CT scan(s) and T1and T2-weighted MRI scans were imported into MIM version 6.0 (MIM Software Inc, Cleveland, OH). MRI and CT images were fused, and gross tumor volume was contoured by a radiation oncologist. A planning target volume (PTV) margin of 2 mm was added for setup uncertainty. A single-beam or multibeam treatment plan using Elekta XiO version 5.0 treatment planning system (Elekta, Stockholm, Sweden) was formulated to optimize PTV coverage while meeting OAR constraints, when feasible. Beam selection was based on tumor location, tumor size, and adjacent OARs. OARs, including the cornea, lacrimal gland, lens, retina, and optic nerve, were manually contoured.

Corneal toxicity after proton therapy for uveal melanoma Table 1

3

Patient characteristics and dosimetry (N Z 92) Variable

Age at time of proton beam therapy, y Sex Male Female Eye Right Left COMS size Small Medium Large Tumor position Anterior chamber Iris/ciliary body Other Anterior to equator* Posterior Planning target volume size, cm3 US-based tumor dimensions, mm Transverse Longitudinal Height/thickness Comorbidities History of diabetes History of hypertension History of hyperlipidemia History of glaucoma Dose and fractionation 50 Gy in 5 fractions 56 Gy in 4 fractions

No. (%) or median (range) 66 (13-93) 49 (53.3) 43 (46.7) 52 (56.5) 40 (43.5) 33 (35.9) 53 (57.6) 6 (6.5) 12 (13.0) 6 6 12 (13.0) 68 (74.0) 1.20 (0.40-4.36) 9.44 (1.20-24.50) 9.77 (1.10-24.50) 3.11 (0.50-11.20) 15 39 22 2

(16.3) (42.4) (23.9) (2.2)

87 (94.6) 5 (5.4)

Abbreviation: COMS Z Collaborative Ocular Melanoma Studies. * Not within the anterior chamber.

Corneal contour generation The corneal limbus has been defined surgically as the point of transition between the cornea and the sclera.16 Given that the average corneal thickness is approximately 600 mm, a 1.2-mm thickness contour was used to draw the cornea, defined as a 12-mm horizontal diameter contour along the anterior surface of the eye defined on a T2-weighted driven equilibrium (DRIVE) MRI sequence. The corneal contour extended to and included the corneal limbus, which was defined as the inflection point between the corneal and scleral curvatures. The limbus alone was defined as a 1.2mm thickness contour at the inflection point of the corneal and scleral curvatures.

Treatment delivery Image guidance for proton treatment was performed through daily orthogonal KV x-ray imaging based on bony anatomy and fiducial clips. An in-room camera system is

4

Lee et al.

also used to monitor the patient’s gaze throughout the entire treatment. Treatment was interrupted if any eye movement deviated from the correct gaze. Setup images and visual tracking and monitoring were repeated for each field. Eyelid retractors were used in some cases where the tumor was located anteriorly. Treatment was delivered using a standard 10-cm snout with brass apertures, using a fixed beam at 90 or an inclined beam line at 30 .

Outcome measurement and follow-up Toxicity grading was based on an expanded version of the National Cancer Institute Common Terminology Criteria for Adverse Events, version 4.0, with assessments of toxicity performed once during treatment and at 6 months at our proton therapy center or, in the case of patients who traveled from far out of state, with an ophthalmologist at our center or a local ophthalmologist, who would forward their toxicity assessments and clinic notes to our center at 1, 3, and 6 months. Grade 2 toxicity was defined as toxicity requiring medical intervention, such as steroid eye drops or artificial tears. Grade 3 toxicity was defined as toxicity requiring either contact bandage lens or amniotic membrane transplant. Grade 4 toxicity was defined as toxicity requiring enucleation. To rigorously ensure that interventions were performed for corneal radiation therapyerelated toxicities, patients receiving eye drops, artificial tears, and/or contact bandage lenses had their clinic notes cross-referenced to verify corneal indications in their ophthalmologists’ examinations and descriptions of toxicity. Systemic disease screening was initiated within 6 months by a medical oncologist. Intraocular surveillance was performed by the referring ocular oncologist with B-scan ultrasound at 4- to 6-month intervals. Pre-existing clinical conditions that might contribute to toxicity, such as glaucoma, diabetes, hypertension, and hyperlipidemia, were recorded.

Statistical analysis of corneal toxicity risk All patients had complete follow-up data at the acute toxicity window of 6 months, permitting analysis of toxicity incidence rather than time-to-incidence. Receiver operating characteristic analysis, along with c2 (multi-level categorical), Fisher exact (bilevel categorical), and Mann-Whitney (continuous) hypothesis testing with Benjamini-Hochberg false discovery rate correction were used to identify predictors of corneal toxicity incidence. Cutoff values in continuous DVH parameters that maximized the Youden Index, defined as sensitivity þ specificity e 1, for predicting toxicity incidence were calculated and are proposed as planning constraints. A bivariate logistic regression model for toxicity prediction was constructed to evaluate the

International Journal of Radiation Oncology  Biology  Physics Table 2 Six-month rates of corneal toxicity for patients receiving proton beam therapy, by patient and tumor characteristics Variable Tumor location Posterior Anterior to equator Anterior chamber COMS size category Small Medium Large HLD Yes No HTN Yes No T2DM Yes No

n

6-mo GR2þ corneal toxicity rate (%)

68 12 12

0.0 25.0 58.3

33 53 6

3.0 11.3 50.0

22 70

22.7 7.1

39 53

12.8 9.4

15 77

13.3 10.4

P value <.0001*

.004*

.07y .74y .72y

Abbreviations: COMS Z Collaborative Ocular Melanoma Studies.; Gr2þ Z grade 2þ; HLD Z Hyperlipidemia; HTN Z Hypertension; T2DM Z Type 2 Diabetes Mellitus. * c2 test. y Fisher exact test.

independent predictive utility of DVH parameters relative to the strongest clinical risk factor.

Results Patients Corneal DVH parameters are listed in Table 2, subdivided by tumor location. Eighty-seven patients (94.6%) were treated to 50 Gy (RBE) in 5 fractions, and the remaining 5 patients (5.4%) received 56 Gy (RBE) in 4 fractions. Twelve patients had tumors involving the anterior chamber, and thus the iris or ciliary body. Twelve patients had tumors originating in the choroid or posteriorly but extending past the equator of the eye. The remaining 68 patients had posterior tumors. No patients reported significant vision loss or required enucleation by 6 months after radiation. No patients had a local recurrence. One patient was found to have systemic disease in the liver at 6-month follow-up.

Corneal toxicity Ten of 92 patients (10.9%) experienced grade 2þ (GR2þ) corneal toxicity requiring intervention within 6 months of treatment. Of these 10 patients, 5 experienced grade 2 toxicity and were treated with steroid eye drops, and 5 experienced grade 3 toxicity requiring steroid eye drops

Volume -  Number -  2019

as well as contact bandage lens and/or amniotic membrane transplant. There were no grade 4 toxicities. Toxicities were verified by the patients’ ophthalmologists to be secondary to radiation therapy rather than other factors such as abrasion. Of the 10 patients experiencing toxicity, 7 had tumors within the anterior chamber, all of which involved the iris or ciliary body. Three had posteriorly originating tumors extending past the equator of the eye. Stratified by tumor location, the GR2þ toxicity incidence was 58.3% for anterior chamber tumors, 25.0% for posterior tumors extending past the equator, and 0% for tumors that were wholly posterior. Stratified by COMS size criteria, the GR2þ toxicity incidence was 50.0% for large tumors, 11.3% for medium tumors, and 3.0% for small tumors. Characteristics of patients and corresponding toxicity rates by category are listed in Table 2, with Kaplan-Meier depictions of toxicity incidence shown in Figure 1. Patients experienced 1 or more corneal toxicity subtypes, including abrasion (n Z 1), discomfort (n Z 5), and persistent epithelial defects (n Z 7).

Persistent epithelial defects Seven patients had persistent epithelial defects noted at a median follow-up of 3 months after radiation therapy. All 7 events occurred in patients whose tumors were anteriorly positioned, with 4 involving either the iris or the ciliary body. The remaining 3 tumors originated posteriorly but extended anterior to the equator of the globe. Patients with persistent epithelial defects reported varying decline in visual acuity of the affected eye, some of which were severe (Table E1; available online at https://doi. org/10.1016/j.ijrobp.2019.02.005). All were treated initially with steroid eye drops. In those with persistent inflammatory responses and incomplete epithelialization, contact bandage lens and amniotic membrane grafts were used. For defects unresponsive to these measures, referrals for potential corneal limbal stem cell transplant were made. Two patients were actively considering the procedure by 6-month follow-up, but none had undergone limbal stem cell transplant to date.

Patient and tumor factors predictive of corneal toxicity Table 2 and Figure 1 A and 1B summarize the incidence of GR2þ corneal toxicity by various patient and tumor factors. Significant predictors for increased toxicity incidence included increased COMS size (P Z .004) and anterior positioning of the tumor (P < .0001). However, PTV dichotomized by the median did not correlate with corneal toxicity (Fisher P Z .16). Median PTV D95% was 50.9 Gy (RBE) for both groups (P Z .89).

Corneal toxicity after proton therapy for uveal melanoma

5

DVH parameters predictive of corneal toxicity Table 3 summarizes that DVH parameters including corneal mean dose, maximum dose, V5, V25, V45, and V50 were significant predictors for GR2þ corneal toxicity (P < .0001). Receiver operating characteristic analysis showed high area under the curve (AUC) for all factors, including V45 (AUC, 0.98), V25, and mean dose (AUC, 0.97; Table 4). Candidate DVH cutoff parameters as dosimetric constraints to mitigate corneal toxicity were corneal V25 <58%, corneal V45 <32%, mean corneal dose <32 Gy (RBE), and maximum corneal dose <51.8 Gy (RBE; Table 4). Figure 1C and 1D summarize the incidence of GR2þ corneal toxicity by maximum and mean dose to the cornea, respectively, with patients stratified by respective cutoffs of 51.8 Gy (RBE) and 32 Gy (RBE). Figures1E and 1F summarize incidence of toxicity by volume of cornea receiving 25 Gy (RBE) and 45 Gy (RBE), stratified by respective cutoffs of 58% and 32%. In all cases, sensitivity for detecting toxicity incidence was 90%, and specificity was 96%.

Bivariate analysis To investigate whether DVH parameters add predictive power for toxicity incidence beyond the most important clinical/tumor related factor (tumor location, P < .0001), a bivariate logistic regression model including both DVH parameters and tumor location was created (Table E2; available online at https://doi.org/10.1016/j.ijrobp.2019.02. 005). The analysis showed that V25 (P Z .04), V45 (P Z .04), and mean dose (P Z .03) could independently predict for toxicity after adjusting for location.

Discussion This study details the rate of GR2þ acute corneal toxicity in our cohort of patients with uveal melanoma treated with PBT, with an overall rate of 10.9%. Given the lack of published data rigorously evaluating corneal toxicity rates after PBT for uveal melanoma, it is unclear whether this rate of toxicity is high or typical. Although our center is among newer commercial centers that have recently been built, our beam characteristics, including dose falloff and penumbra, have been noted to be similar to those of dedicated eye lines that have historically treated patients for ocular melanoma, such as the Harvard Cyclotron.17 The penumbra of our facility’s 26-mm range fully modulated beam with a diameter of 25 mm is 2.8 mm. The penumbra of the Harvard cyclotron’s modulated beam is 1.6 mm, and that of the Centre Antoine Lacassagne in Nice, France, is 2.4 mm. These comparative measurements are available in an unpublished but widely distributed review of proton ophthalmological treatment facilities operational in 1995 under the Clatterbridge Centre for Oncology by A.

6

International Journal of Radiation Oncology  Biology  Physics

Lee et al.

A 80

Tumor Location 100

Small Medium Large

Incidence (%) of GR2+ Toxicity

Incidence (%) of GR2+ Toxicity

B

COMS Category

100

60 40 20

Posterior Anterior to Equator Anterior Chamber

80 60 40 20

0

0 0

1

2

3

4

5

6

0

1

Months from Treatment

C

33 53 6

32 50 4

32 49 3

32 48 3

32 48 3

32 Posterior 47 Anterior to Equator 3 Anterior Chamber

Cornea Max Dose

100

Incidence (%) of GR2+ Toxicity

33 51 5

D

≤ 51.8 Gy (RBE) > 51.8 Gy (RBE)

80 60 40 20

68 12 12

0

68 12 9

5

6

68 11 7

68 10 6

68 9 6

68 9 6

68 9 5

≤ 32.0 Gy (RBE) > 32.0 Gy (RBE)

80 60 40 20

1

2

3

4

5

6

0

1

Months from Treatment ≤ 51.8 Gy (RBE) 80 > 51.8 Gy (RBE) 12

80 9

79 7

79 5

79 4

2

3

4

5

6

79 4

79 4

79 3

4

5

6

81 2

81 1

Months from Treatment 79 4

79 3

Cornea V25

100

≤ 32.0 Gy (RBE) 80 > 32.0 Gy (RBE) 12

F

≤ 58% > 58%

80 60 40 20

80 9

79 7

79 5

Cornea V45

100

Incidence (%) of GR2+ Toxicity

Incidence (%) of GR2+ Toxicity

4

0 0

E

3

Cornea Mean Dose

100

Incidence (%) of GR2+ Toxicity

Small Medium Large

2

Months from Treatment

≤ 32% > 32%

80 60 40 20 0

0 0

1

2

3

4

5

6

0

1

Months from Treatment ≤ 58% 80

> 58% 12

80 9

79 7

79 5

79 4

2

3

Months from Treatment 79 4

79 3

≤ 32% 82 > 32% 10

80 7

81 5

81 3

81 2

Fig. 1. Kaplan-Meier curves of grade 2þ corneal toxicity incidence after proton beam therapy as a function of dose and clinical factors. Dotted lines indicate 95% confidence intervals. The number of patients remaining at risk for toxicity by subgroup for every 1-month interval is listed beneath the respective Kaplan-Meier curves. Abbreviations: COMS Z Collaborative Ocular Melanoma Studies; GR2þ Z grade 2þ.

Volume -  Number -  2019 Table 3 cornea

Corneal toxicity after proton therapy for uveal melanoma

7

Six-month rates of corneal toxicity for patients receiving proton beam therapy, by dose-volume histogram parameters for the

Parameters 25 Gy (RBE) (V25) % Volume 0%-58% >58% 45 Gy (RBE) (V45) % Volume 0%-32% >32% Mean dose 32 Gy (RBE) >32 Gy (RBE) Maximum dose 51.8 Gy (RBE) >51.8 Gy (RBE)

n

6-mo GR2þ corneal toxicity rate (%) by proton beam dose received

P value <.0001*

80 12

1.3 75.0 <.0001*

82 10

1.2 75.0

80 12

1.3 75.0

80 12

1.3 75.0

<.0001* <.0001*

Abbreviations: Gr2þ Z grade 2þ; RBE Z relative biological effectiveness. * Mann-Whitney test.

Kacperek, PhD. A larger penumbra may reduce nearby OAR sparing, depending on the location of the lesion. In the case of anteriorly positioned tumors, this could reduce corneal sparing and be a theoretical contributor to a higher rate of toxicity. The observed toxicity rate was significantly higher (58.3%) among anterior chamber tumors alone and 25% for tumors anterior to the equator but not involving the anterior chamber. It would be reasonable to counsel all patients with iris or ciliary body tumors to expect further intervention for corneal toxicities because the cornea uniformly receives a high dose generally above 30 Gy (RBE). Given that large reductions of dose to the cornea may be difficult in cases of iris/ciliary body tumors where corneal dose is already very high because of tumor and OAR proximity (Fig. 2A), the treatment-modifying value of the DVH parameter cutoffs we propose may be greatest in cases of tumors that are anterior to the equator but not within the anterior chamber (Fig. 2B, 2C). These cases with intermediate doses to the cornea may be the optimal candidates for modifying treatments to reduce GR2þ toxicity incidence. Corneal toxicities remain poorly characterized in the literature, and most reports are from plaque brachytherapy Table 4

studies. The most common side effects of plaque brachytherapy for anterior uveal melanomas within 1 month of treatment include keratopathy and uveitis, whereas by 6 months the most common side effect tends to be cataract formation.18 However, no persistent epithelial defects have been described. In another study, 125I plaque brachytherapy of uveal melanomas anterior to the equator (62 of 133, including ciliary body involvement) resulted in significant rates of cataract formation (50%) with minor rates of keratitis (<3%) by 5 years.19 Again, no persistent epithelial defects were described. Furthermore, a Canadian investigation of ciliary body melanoma treatment with 125 I plaque brachytherapy noted only a 5% uveitis rate and no persistent corneal toxicities.20 Significant corneal toxicities appear to be more common in patients treated with external beam radiotherapy for anterior tumors. In a study of 107 patients with iris melanomas receiving PBT, corneal toxicities evaluated after a median follow-up of 49.5 months included transient dry corneal syndrome with associated discomfort in 35 patients, treated with lubricating eye drops. Also noted were uveitis in 4 patients, 2 of whom had anterior granulomatous uveitis and were treated successfully with

Predictors of corneal toxicity: Univariate analyses and Youden cutoff values

Cornea DVH parameter

Toxicity median (n Z 10)

No toxicity median (n Z 82)

Youden cutoff

Sensitivity (%)

Specificity (%)

ROC AUC

P value

V5% V25% V45% V50% Mean dose Gy (RBE) Maximum dose Gy (RBE)

100 96.1 51.3 27.8 42.9 52.8

24.9 0.3 0 0 4.0 27.4

92 58 32 20 32 51.8

90 90 90 90 90 90

89 96 99 99 96 96

0.93 0.97 0.98 0.94 0.97 0.96

<.0001 <.0001 <.0001 <.0001 <.0001 <.0001

Abbreviations: AUC Z area under the curve; DVH Z dose-volume histogram; RBE Z relative biological effectiveness; ROC Z receiver operating characteristic.

8

Lee et al.

International Journal of Radiation Oncology  Biology  Physics

Fig. 2. Dosimetric diagrams for 3 patients on T2-weighted magnetic resonance images, 1 with an anterior chamber iris tumor (A) who experienced GR2þ corneal toxicity, and 2 patients with anterior or anteriorly extending tumors with different levels of corneal dose who did not experience GR2þ toxicity (B, C). Representative corneal contours are outlined in green, and gross tumor volumes are outlined in red. (A) Most patients experiencing GR2þ corneal toxicity had anterior chamber tumors involving either the iris or ciliary body. These patients uniformly received high doses to the cornea and limbus. (B, C) Patients with tumors extending anteriorly past the equator had lower but nonzero rates of GR2þ corneal toxicity, given their intermediate dose profiles. No patients with posterior tumors demonstrated persistent corneal toxicity because they received very little dose to the cornea. Isodose lines: red Z 50 Gy (RBE), fuchsia Z 45 Gy (RBE), yellow Z 40 Gy (RBE), orange Z 30 Gy (RBE), light blue Z 20 Gy (RBE), and blue Z 10 Gy (RBE). Abbreviations: GR2þ Z grade 2þ; RBE Z relative biological effectiveness. (A color version of this figure is available at https://doi.org/10.1016/j.ijrobp.2019.02.005.) local steroids, and scleral necrosis in 1 patient. Chronic corneal complications occurred in 6 patients (5.6%) within 40 months of treatment, including 2 with recurring corneal ulcers requiring bandage contact lens and healing eyedrops. The other chronic toxicities seen included band keratopathy, recurrent superficial punctate keratitis, recurrent keratalgia, and transient peripheral corneal edema.21 Patients with posterior choroidal melanomas unsuitable for brachytherapy and treated with stereotactic photon radiotherapy experienced corneal epithelial defects at a rate of 0.9% and 4.3% at 3 and 12 months after radiation therapy, respectively. By 5 and 10 years after treatment, these rates were 22.4% and 23.5%, respectively.22 It is worth noting that the scoring system used by these authors for corneal epithelial defects centered on corneal ulcers, and it is unclear whether these were persistent or transient changes leading to vision deterioration. Furthermore, this report focused on patients with only posterior tumors (and thus presumably low doses to the cornea), whereas all patients in our cohort experiencing corneal toxicity had anteriorly positioned tumors (and thus much higher doses to the cornea). Thus, a true comparison with our cohort’s toxicity rates is difficult. Our cohort’s corneal toxicity rates may nevertheless increase over time, and we will report these in future works as our follow-up data mature. Despite these corneal complications, a recent study demonstrated preservation of excellent vision in 45% of patients treated with PBT for uveal melanoma. This study identified 28-Gy exposure of the macula and optic nerve as independent DVH predictors of post-PBT vision loss among those with initially favorable visual acuity and overall

corroborates the general consensus that proton therapy for ocular melanomas is associated with a favorable radiogenic toxicity profile.23 Although proton therapy is a historically well-known pillar of ocular melanoma treatment, new studies show continually improving toxicity outcomes, such as better lens-sparing approaches allowing effective tumor control while reducing the incidence of cataracts, as well as reduced incidence of neovascular glaucoma when treating uveal melanomas with certain dosimetric constraints.24,25 With proton therapy centers becoming more common worldwide, the number of patients being treated with PBT for ocular melanomas is expected to rise. Better characterizing side effects of PBT, especially with advanced planning technologies that allow detailed DVH analyses, will allow physicians to better inform their patients during the consent for treatment process and to predict who may be more at risk for needing procedures such as corneal transplant. Thus, the present study could be expected to affect clinical care in 2 main ways. First, with corneal contouring and dose-volume characterization, corneal dose may be reduced using a lateral beam or a blended lateral and anterior approach to spare portions of the cornea. Second, in patients in whom DVH parameters unavoidably exceed our proposed constraints, aggressive monitoring and early (or even prophylactic) intervention with steroid eye drops or a contact bandage lens may be able to mitigate treatment-related toxicity. Limitations of this study include the patient selection bias for those treated at a single institution and referred by ocular oncologists with access to both brachytherapy plaques and PBT. There could also be a selection bias for those patients who sought access to PBT. Another limitation is that the

Volume -  Number -  2019

specific DVH predictors depend on the definition of the corneal contour. Because contours were defined using MRI scans, which can have significant geometrical distortions in the eyes, differences between MRI scanners and scan protocols may affect definition of this volume and the accuracy of the DVH predictor. A final limitation is that the follow-up period of the present patient cohort is short, whereas late radiation-induced toxicities, such as corneal epithelial defects and the need for enucleation because of neovascular glaucoma, are known to develop many months to years after radiation therapy.22,25 Thus, the present study with 6 months of follow-up may not capture all corneal defects or toxicity events that patients experience post-PBT. However, because radiation keratitiserelated corneal toxicities do require intervention, often occur within the acute or subacute posttreatment windows, and can in some cases profoundly affect quality of life via either corneal pain or decreased vision, we believe that the present study is an important addition to the literature. We provide clinical and dosimetric factors by which to predict which patients are most at risk of toxicity, with the hope that this will allow treating physicians to provide relevant information during the informed consent process. This approach may also assist physicians in identifying patients who may benefit from early intervention with steroid medications, although the association between corneal dose and major toxicities such as persistent epithelial defect requiring corneal limbal stem cell transplantation or neovascular glaucoma requiring enucleation, have yet to be established. Future analysis will include continued patient accrual with longer median follow-up. The larger patient numbers will allow us to also assess late treatment toxicities such as persistent epithelial defects and enucleation among patients treated with PBT for uveal melanoma at our institution.

Conclusions

Corneal toxicity after proton therapy for uveal melanoma

3. 4. 5.

6.

7. 8.

9.

10.

11.

12.

13.

14.

15.

16. 17.

18. 19.

Corneal toxicity is a notable side effect of PBT for anteriorly positioned uveal melanoma. Anteriorly located tumors necessitate high corneal coverage and dose, and toxicities are unsurprisingly most common among these patients. However, use of multibeam and OAR-optimized treatment planning to respect dosimetric constraints may reduce the likelihood of this toxicity and aid in prediction of patients with posterior tumors extending past the equator of the globe who are most at risk for toxicity and may benefit from early intervention.

References 1. McLaughlin CC, Wu XC, Jemal A, et al. Incidence of noncutaneous melanomas in the U.S. Cancer 2005;103:1000-1007. 2. Collaborative Ocular Melanoma Study Group. The COMS randomized trial of iodine 125 brachytherapy for choroidal melanoma,

20.

21. 22.

23.

24.

25.

9

III: Initial mortality findings. COMS Report No. 18. Arch Ophthalmol 2001;119:969-982. Gragoudas ES, Goitein M, Koehler AM, et al. Proton irradiation of small choroidal malignant melanomas. Am J Ophthalmol 1977;83:665-673. Gragoudas ES. The Bragg peak of proton beams for treatment of uveal melanoma. Int Ophthalmol Clin 1980;20:123-133. Gragoudas ES, Lane AM, Munzenrider J, et al. Long-term risk of local failure after proton therapy for choroidal/ciliary body melanoma. Trans Am Ophthalmol Soc 2002;100:43-48; [discussion: 48-49]. Gragoudas ES, Lane AM, Regan S, et al. A randomized controlled trial of varying radiation doses in the treatment of choroidal melanoma. Arch Ophthalmol 2000;118:773-778. Gragoudas ES, Seddon J, Goitein M, et al. Current results of proton beam irradiation of uveal melanomas. Ophthalmology 1985;92:284-291. Lane AM, Kim IK, Gragoudas ES. Long-term risk of melanoma-related mortality for patients with uveal melanoma treated with proton beam therapy. JAMA Ophthalmol 2015;133:792-796. Mishra KK, Quivey JM, Daftari IK, et al. Long-term results of the UCSF-LBNL randomized trial: Charged particle with helium ion versus iodine-125 plaque therapy for choroidal and ciliary body melanoma. Int J Radiat Oncol Biol Phys 2015;92:376-383. Jensen AW, Petersen IA, Kline RW, et al. Radiation complications and tumor control after 125I plaque brachytherapy for ocular melanoma. Int J Radiat Oncol Biol Phys 2005;63:101-108. Badiyan SN, Rao RC, Apicelli AJ, et al. Outcomes of iodine-125 plaque brachytherapy for uveal melanoma with intraoperative ultrasonography and supplemental transpupillary thermotherapy. Int J Radiat Oncol Biol Phys 2014;88:801-805. Perez BA, Mettu P, Vajzovic L, et al. Uveal melanoma treated with iodine-125 episcleral plaque: An analysis of dose on disease control and visual outcomes. Int J Radiat Oncol Biol Phys 2014;89:127-136. Papakostas TD, Lane AM, Morrison M, et al. Long-term outcomes after proton beam irradiation in patients with large choroidal melanomas. JAMA Ophthalmol 2017;135:1191-1196. Seddon JM, Gragoudas ES, Polivogianis L, et al. Visual outcome after proton beam irradiation of uveal melanoma. Ophthalmology 1986;93: 666-674. Hartsell WF, Kapur R, Hartsell SOC, et al. Feasibility of proton beam therapy for ocular melanoma using a novel 3D treatment planning technique. Int J Radiat Oncol Biol Phys 2016;95:353-359. Van Buskirk EM. The anatomy of the limbus. Eye 1989;3:101-108. Slopsema RL, Mamalui M, Zhao T, et al. Dosimetric properties of a proton beamline dedicated to the treatment of ocular disease. Med Phys 2014;41:1-11. Saakyan SV, Amiryan AG, Valskiy VV, et al. [Plaque radiotherapy for anterior uveal melanomas]. Vestn Oftalmol 2015;131:5-11. Lumbroso-Le Rouic L, Charif Chefchaouni M, Levy C, et al. 125l plaque brachytherapy for anterior uveal melanomas. Eye 2004;18:911-916. Krema H, Simpson ER, Pavlin CJ, et al. Management of ciliary body melanoma with iodine-125 plaque brachytherapy. Can J Ophthalmol 2009;44:395-400. Thariat J, Rahmi A, Salleron J, et al. Proton beam therapy for iris melanomas in 107 patients. Ophthalmology 2017;125:606-614. Dunavoelgyi R, Dieckmann K, Gleiss A, et al. Radiogenic side effects after hypofractionated stereotactic photon radiotherapy of choroidal melanoma in 212 patients treated between 1997 and 2007. Int J Radiat Oncol Biol Phys 2012;83:121-128. Polishchuk AL, Mishra KK, Weinberg V, et al. Temporal evolution and dose-volume histogram predictors of visual acuity after proton beam radiation therapy of uveal melanoma. Int J Radiat Oncol Biol Phys 2017;97:91-97. Thariat J, Jacob S, Caujolle JP, et al. Cataract avoidance with proton therapy in ocular melanomas. Investig Ophthalmol Vis Sci 2017;58: 5378-5386. Mishra KK, Daftari IK, Weinberg V, et al. Risk factors for neovascular glaucoma after proton beam therapy of uveal melanoma: A detailed analysis of tumor and dose-volume parameters. Int J Radiat Oncol Biol Phys 2013;87:330-336.