Risk factors for radiation-induced optic neuropathy: a case-control study

Abstract

Importance

Identifying risk factors for RION could promote a more conservative approach to radiation treatment planning in vulnerable patients.

Background

This study explored possible factors beyond radiation dose associated with the development of radiation-induced optic neuropathy (RION) after external beam radiation therapy.

Design

This was a retrospective case-control study conducted at a university hospital tertiary care center.

Participants

Cases (n=14) meeting criteria for a diagnosis of RION by neuro- ophthalmologic exam were identified from a single-center neuro-ophthalmology database. Controls (n=31) without RION were selected from a single-center radiation oncology database.

Methods

Controls were matched to cases based upon maximum radiation dose to the optic apparatus. Patient characteristics and treatment parameters were interrogated by univariate analysis for attributes predisposing to RION.

Main Outcome Measures

The primary parameter was a significant association of patient characteristics or treatment parameters with RION.

Results

Controlling for radiation dosage, no significant associations for alternative risk factors were identified.

Conclusions and Relevance

These results support the literature suggesting that the primary risk factor for developing RION is radiation dosage and that additional, patient- and tumor-related risk factors may play only a minor role.

INTRODUCTION

Radiation-induced optic neuropathy (RION) is a rare complication after external beam radiation therapy (EBRT) in which exposure of the anterior visual pathway to radiation results in acute, painless vision loss in one or both eyes months to years after treatment¹. Radiation-induced ischemic necrosis from free radical damage to the vascular endothelium has been theorized as the pathogenesis of RION. The severity of visual impairment ranges from subclinical changes in visual evoked potentials² to complete blindness. In many cases, visual function progressively deteriorates soon after the initial development of RION. While this disease-course suggests a possible treatment window, no interventions have definitely demonstrated efficacy stabilizing or reversing the visual decline ¹.

Maximum radiation dose to the anterior visual pathway has been established as a significant determinant for the development of RION^{1, 3,4}. The anterior visual pathway has been shown to tolerate no greater than 50 Gy of cumulative radiation in fractions less than 2 Gy, with the incidence of RION increasing markedly at higher levels^5–9. The incidence of RION following radiation treatment of head and neck or skull-base tumors has been quoted as ranging from 0% for <50Gy to 16% for >70Gy¹⁰. Current EBRT treatment planning protocols limit maximum points doses to the optic pathway to 54–55 Gy in 1.8–2 Gy fractions to minimize this complication¹¹.

While previous studies have been instrumental in defining dose tolerance thresholds, fewer have investigated patient-associated systemic or demographic risk factors for developing RION. Those that do consider patient-associated attributes have been descriptive case reports, case series, and reviews. These have described potentially predisposing factors for RION according to plausible pathogenic mechanisms ^{3, 12–14}. Demizu et al found a significant association for diabetes mellitus¹⁵. Deng et al suggested that compression of the optic nerve by tumor may sensitize the visual pathway to radiation injury¹⁶. Guy et al reported adjuvant chemotherapy as a potential risk factor¹⁷. Gragoudas et al argue that risk may be enhanced by underlying vascular disorders¹⁸ and therefore that hypertension, hyperlipidemia and smoking status may prove significant. Multiple authors indicate age may play a role^{4, 19, 20}. However, given the low incidence of RION under current EBRT planning protocols, a systematic analysis of these risk factors has remained elusive.

Identifying risk factors for developing RION could inform EBRT planning, monitoring and follow-up for high-risk patients. This would allow for a more complete discussion with the patient, modification of such risk factors, reconsideration of EBRT dose constraint to the optic apparatus, and more aggressive neuro-ophthalmologic and neuroradiologic screening after EBRT to help mitigate the incidence and impact of this debilitating outcome.

A case-control study is an ideal method to identify risk factors which may be causally related to a specific outcome. We therefore performed a case-control study to identify potential patient-associated risk factors for RION. Cases (n=14) with a neuro-ophthalmologic diagnosis of RION were compared against controls (n=31) who had experienced maximum point-doses to the optic pathway within an equivalent range but did not develop RION. The groups’ characteristics and treatment parameters were interrogated by univariate analysis for predisposing attributes in an attempt to identify significant attributes associated with the development of RION.

METHODS

Cases were retrospectively identified from a single-center, neuro-ophthalmology database. All 14 RION cases were diagnosed with RION by neuro-ophthalmologic exam and assessment, and received EBRT from May 2008 to June 2014.

Selection criteria included: clinical evidence of optic nerve dysfunction (reduced visual acuity, color impairment, optic disc pallor or swelling, etc); EBRT exposing the optic apparatus to ionizing radiation (including fractionated intensity-modulated radiation therapy, whole brain radiation therapy, or proton beam therapy); loss of vision >3 months post-treatment (consistent with the reported timeline for the development of RION¹); MRI changes consistent with RION (T2 hyperintensity of the optic nerves and chiasm and/or focal enhancement after intravenous gadolinium²¹); and exclusion of alternative disorders or tumor recurrence.

EBRT planning records which contoured the optic apparatus were available for 8 of the cases (57%). RT planning records which did not specifically contour the optic apparatus but did report overall radiation doses to the affected area were available for the remaining 5 cases. A range of values for the maximum radiation doses to the affected optic nerves was extracted from radiation treatment planning records. This range (4479–6320 cGy) set the upper and lower bounds for radiation exposure to the optic nerves in selected controls.

Controls were retrospectively selected from a single-center, radiation oncology database. Criteria included RT for intracranial or skull-base tumors with radiation exposure to optic apparatus and at least three years of follow up and no mention of vision loss prior to or following radiation therapy. A total of 31 controls were matched based exclusively upon total radiation to the optic apparatus. By matching controls and cases based on radiation dosage, which is widely reported in the literature as an independent risk factor for the development of RION, this study design attempts to identify other patient-specific factors predisposing to RION.

Variables of interest including race, age, gender, presence of diabetes mellitus, hypertension or hyperlipidemia, smoking status, adjuvant chemotherapy or surgery, and radiological evidence of tumor impingement on the optic apparatus were assessed by thorough chart review. There was no selection criteria regarding tumor type. There was no analysis of radiation dose as this was selected as matching criteria based on the already widely reported significance of radiation dose as a significant factor for the development of RION^{1, 3, 4}.

For categorical variables, the control group and case group were compared with Chi-square or Fishers exact tests as appropriate. For continuous variables, difference in the mean of the two groups were compared with t-tests or Kruskal-Wallis tests as appropriate. Absent matched pairing necessary for conditional logistic regression, a logistic regression model was used for univariate analyses of predictors for RION. Significance was defined as a p-value < 0.05. Statistical analyses were performed with SAS version 9,3 (SAS Institute, Cary, NC, USA). All statistical tests were two sided.

RESULTS

Patient Characteristics

The relevant demographic and medical characteristics of cases and controls are summarized in Table 1. The case group included 14 patients with a neuro-ophthalmologic diagnosis of RION. The median age of patients in the case group was 58 (5–82). There were 9 men and 5 women. A diversity of tumors were represented, unified by the diagnosis of RION. Of these tumors, 8 (57%) demonstrated radiological evidence of impingement on the optic apparatus either by mass effect, abutment, ensheathment, or perineural invasion. In total, 3 patients (21%) had diabetes mellitus, 9 patients (64%) had hypertension, 5 patients (36%) had hyperlipidemia and 3 patients (21%) had ever smoked tobacco. Regarding adjuvant therapies, 9 patients (64%) had surgery prior to RT and 4 patients (29%) had concurrent chemotherapy.

Table 1.

Patient characteristics and treatment parameters

Characteristics	Cases (n=14) N (%)	Controls (n=31) N (%)
Demographics
Age, years (median (range))	58 (5–82)	63 (36–83)
Age ≤ 60 years	7 (50%)	12 (39%)
Race
African American	3 (21%)	9 (29%)
Other	11 (79%)	22 (71%)
Female gender	5 (36%)	19 (61%)
Tumor characteristics
Adenocarcinoma (metastases, lung primaries)	2 (14%)	0 (0%)
Angiosarcoma/SCC	1 (7%)	2 (6%)
Glioblastoma	2 (14%)	11 (35%)
Meningioma	4 (28%)	7 (23%)
Pituitary macroadenoma / Craniopharyngioma	5 (36%)	11 (25%)
AVP impingement	8 (57%)	12 (31%)
Comorbidities
Diabetes	3 (21%)	4 (13%)
Hypertension	9 (64%)	18 (58%)
Hyperlipidemia	5 (36%)	6 (19%)
Past / present smoker	3 (21%)	14 (45%)
Treatment parameters
Surgery prior to RT	9 (64%)	21 (68%)
Concurrent chemotherapy	4 (28%)	12 (39%)

Abbreviations: SCC = squamous cell carcinoma, RT = radiation therapy

Cases in this study commonly presented complaining of “blurry vision” in the weeks prior to neuro-ophthalmologic evaluation. The mean latency period from last radiation treatment session to diagnosis of RION was 360 days (191 to 969 days).

Best corrected visual acuity at the time of diagnosis ranged from 20/25 to hand motion, and progressed to no light perception in the most severe instances. MRI at diagnosis most commonly showed focal enhancement on T2 weighted images with increased FLAIR signal in the affected optic nerves or chiasm. Patients in the control group were selected based on having experienced maximum point-doses to the optic pathway in an equivalent range as patients in the case group.

Cases and controls underwent EBRT, primarily intensity-modulated radiation therapy (IMRT), which delivers radiation beams along multiple arcs with the shape and intensity of each beam designed to conform the dose around critical structures such as the optic nerve thereby maximizing radiation delivery to the tumor while minimizing delivery to adjacent tissues. The radiation doses and dose volumes for treatments and controls are summarized in Table 2. As per the defined selection criteria, the values of maximum point-dose to the optic nerve between cases and controls are closely approximated.

Table 2.

Maximum radiation dose (cGy) to optic nerve

	Median	Range	25%tile	75%tile
Cases (n=8)	5130	4500–6600	5040	5850
Controls (n=31)	5400	4500–6600	5040	6000

Univariate Analysis

Controlling for maximum radiation dose to the optic apparatus, no significant associations between any attributes of interest and an outcome of RION were found (Figure 1, Table 3). No significant associations were found for established vascular risk factors including age, hypertension, hyperlipidemia, diabetes mellitus, or smoking status. No significant association was found for adjuvant therapies including chemotherapy or surgery. No significant association was found for radiological evidence of tumor impingement on the optic apparatus. Of note, vascular risk factors such as hyperlipidemia, diabetes, older age, and hypertension were associated with higher, albeit non-significant, odd ratios for RION. There being no significant differences in univariate analysis, a multivarite model was not applied.

Figure 1.

Univariate analysis of patient characteristics and treatment parameters

Table 3.

Univariate analysis of patient characteristics and treatment parameters

Characteristic	Odds ratio	95% Confidence Interval
Age ≤60	1.58	0.44–5.65
African American Race	0.68	0.15–2.97
Female gender	0.35	0.10–1.30
AVP impingement	2.1	0.59–7.61
Diabetes	1.84	0.35–9.6
Hypertension	1.3	0.35–4.8
Hyperlipidemia	2.3	0.56–9.9
Smoker	0.33	0.08–1.43
Adjuvant Surgery	0.86	0.23–3.23
Adjuvant Chemo	0.63	0.16–2.5

Abbreviations: AVP = anterior visual pathway

DISCUSSION

Researchers have previously analyzed the radiation-associated predictive factors for developing RION and found that radiation dosage is the most significant determinant for the incidence of RION. The risk increases with increasing radiation dose from 0% for <50Gy to upwards or 16% for >70Gy^{10, 19}. An analysis of 273 patients receiving EBRT with curative intent for head and neck tumors and had radiation fields that included the optic nerve and/or chiasm found the incidence of RION was 9%¹². A retrospective study of 219 patients receiving radiation therapy for cancers of the paranasal sinuses and nasal cavity, found no cases of RION when the total dose was under 50 Gy, and calculated the 10-year actuarial risk for RION as 5% for doses of 50–60 Gy and 30% for doses of 61–78 Gy²². An additional study of 215 optic nerves in 131 patients found no risk of RION for doses less than or equal to 59 Gy¹⁹. For EBRT doses >60Gy, the risk of RION may increase with increasing dose per fraction¹⁹. In this study, patients were treated with standard fractionated EBRT where treatment was delivered in 1.8–2 Gy per fraction. Hyper-fractionation delivers more than one fraction per day but at lower dose per fraction, such as 1.2 Gy twice per day. Previous studies have suggested that hyper-fractionation may reduce the risk of RION by using lower dose per fraction^12,19.

What literature has considered patient-associated predictive factors (i.e. patient demographics and comorbidities) is largely case-based. Bhandare¹², Demizu¹⁵ and Parsons¹⁹ identified potential risk factors by isolating a rare few cases from prospective cohorts of hundreds receiving EBRT. Schreiber¹⁴ and Girkin²¹ suggest potential risk factors from retrospective case reports of individual patients.

Controlling for the significance of radiation dose by selecting the control subjects according to radiation exposures analogous to the case group, this study sought to analyze patient-associated predictive factors for developing RION independent of radiation dosage. In the univariate analysis, no significant associations with patient-specific or vascular factors were found.

RION is a form of late delayed radiation-induced brain injury presenting greater than six months after radiotherapy. The associated white matter necrosis and vascular abnormalities seen on histopathology have variously been attributed to vascular endothelial injury or a reduction in the proliferating capacity of glial cells²³. Neither explanation completely accounts for all clinical and experimental data, and so late delayed radiation induced injury may represent a complex interaction among multiple cell types²⁴.

Previously reported risk factors are consistent with the theorized pathogenesis of RION as a radiation-induced endotheliopathy^{12, 15, 18, 19}. Similarly, this study found vascular risk factors including hyperlipidemia, diabetes, older age, and hypertension to be associated with higher odd ratios for RION. However, no association was significant. The absence of significant findings in this regard is perhaps attributable to the limited sample size and merits further investigation.

Adjuvant chemotherapy has been theorized by others to increase the risk of RION^{25, 26}. While no association was found in this study, the particular chemotherapeutic agent may affect the risk. In particular, vincristine has been implicated in optic neuropathy. In this study, temozolomide was overrepresented as the chemotherapeutic agent for patients receiving adjuvant chemotherapy. Subgroup analysis of adjuvant chemotherapy by chemotherapeutic agent may identify particular agents of concern.

This study sought to explore additional factors predisposing to RION in groups that otherwise received identical radiation with the premise that some factor other than radiation dosage must be contributory. Finding no such additional risk factor, these results support literature which strongly suggests that the primary risk factor for developing RION is radiation dosage and that additional patient-specific and vascular risk factors play a comparably minor role.

The significance of the findings in this preliminary study are constrained by the small sample size. The non-significant trend toward increased odds ratios for various cardiovascular risk factors may suggest that the study was not sufficiently powered. Indeed, for proportions for females gender between the cases and controls (rounded to 0.35 and 0.60) the study would need n=62 per group to have 80% power; for age ≤60, (rounded to 0.50 and 0.40) the study would need n=388 per group to have 80% power.

A multi-site study would allow the researchers to more readily find significant associations and to interrogate multi-factorial predispositions, but such a study would be challenging given the rarity of the condition. A case-control study is the first step in trying to identify potential risk factors for RION. Future case-control studies could include pooled data from multiple institutions, allowing for a larger cohort of cases and controls, which might detect a disease-effect for some of these variables.

The retrospective nature of the study imposes limitations. The majority of cases did not have pre-operative neuro-ophthalmalogic examinations, and so the degree of visual decline from baseline was not able to be assessed. Controls received no neuro-ophthalmologic evaluation, so the possibility of sub-clinical or incipient RION was not excluded from that group, despite setting inclusion criteria at greater than three years with no reported visual decline.

Reasoning from the proposed pathogenesis of RION as vascular endothelial or microglial injury, various treatments have been proposed. Systemic corticosteroids or intraocular corticosteroids, systemic anticoagulants, angiotensin converting enzyme (ACE) inhibitors, hyperbaric oxygen, systemic or intraocular anti-VEGF have all been implemented with mixed results as reported in small case series and animal studies²⁷; however, no treatment to date has been definitely demonstrated to halt or reverse vision loss from RION.

Identifying risk factors for RION could promote a more conservative approach to radiation treatment planning in vulnerable patients. Three-dimensional conformal intensity-modulated radiotherapy could be used to reduce fraction size and minimize radiation dose to the optics for these high-risk patients. Pre- and post-treatment ophthalmologic examination could identify incipient or subclinical cases with the potential to intervene earlier in the disease course. However, given the current ineffectiveness of present interventions, screening for at-risk patients promises the best potential to prevent vision-loss from RION.