Article Text

Radiation-induced optic neuropathy: a review
  1. Andrew R Carey1,
  2. Brandi R Page2,
  3. Neil Miller1
  1. 1 Ophthalmology, Johns Hopkins University School of Medicine, Wilmer Eye Institute, Baltimore, Maryland, USA
  2. 2 Dept. of Radiation Oncology, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
  1. Correspondence to Dr Andrew R Carey, Ophthalmology, Johns Hopkins Wilmer Eye Institute, Baltimore, Maryland, USA; drcarey06{at}


Radiation is a commonly used treatment modality for head and neck as well as CNS tumours, both benign and malignant. As newer oncology treatments such as immunotherapies allow for longer survival, complications from radiation therapy are becoming more common. Radiation-induced optic neuropathy is a feared complication due to rapid onset and potential for severe and bilateral vision loss. Careful monitoring of high-risk patients and early recognition are crucial for initiating treatment to prevent severe vision loss due to a narrow therapeutic window. This review discusses presentation, aetiology, recent advances in diagnosis using innovative MRI techniques and best practice treatment options based on the most recent evidence-based medicine.

  • Optic Nerve
  • Imaging
  • Diagnostic tests/Investigation
  • Drugs
  • Degeneration

Data availability statement

Data sharing not applicable as no datasets generated and/or analysed for this study.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Radiation-induced optic neuropathy (RION) is a late-onset toxicity often described as a catastrophic condition owing to its predilection for rapid, severe and permanent bilateral vision loss leading to blindness. With the advent of new cancer treatments, patients are living longer and patients who previously would not have been expected to survive long enough to develop RION may present with new vision loss, worsening their already compromised quality of life.

Avoidance of radiation toxicity and treatment of the tumour can become competing forces. On the one hand, avoidance of normal structures at risk is important in maintaining quality of life; on the other, a marginal failure leading to local recurrence and potential progression to distant metastases may threaten life. Therefore, in the event of local recurrence, ‘salvage’ therapy to avoid these consequences is often undertaken. At this stage, surgery, chemotherapy and radiation are all less successful than early therapy.1 Therefore, a balanced thought process and a collaborative approach with a multidisciplinary team is appropriate, taking into account competing factors and also the patient’s goals. In neuro-ophthalmic practice balancing the risk of incomplete tumour coverage versus damage to, or loss of function of, surrounding optic nerve, chiasm and retina is an important and often difficult discussion.

Although many primary tumours arise from or involve the eye and orbital structures (eg, melanoma), primary brain tumours, metastatic tumours and local head and neck cancers may also lead to optic structures becoming ‘organs at risk’ (OAR). When a radiation treatment plan is designed, OARs are contoured using data from CT with or without MRI. Most often, images are obtained of the optic nerves, optic chiasm, brainstem, globes and lenses, and, where relevant, the lacrimal gland. The posterior globe may be used as a surrogate for the retina. Contours are used to calculate partial and full volume doses and, where possible, as avoidance structures in the dosimetric planning calculations to reduce risk of toxicity.

Tumours adjacent to the ocular OARs are diverse in histology and prognosis. Some are best treated by resection followed by radiation with or without chemotherapy, others by radiation alone. Radiation techniques employed include three-dimensional (3D) conformal radiation therapy, intensity-modulated radiation therapy (IMRT), electron therapy, brachytherapy and proton therapy. Both photons and protons can be administered using fractionated radiotherapy, single-fraction radiotherapy or stereotactic radiosurgery (SRS). Online supplemental table 1 lists some of the most common tumours in and around ocular OARs and the radiation techniques most commonly employed.

Supplemental material

Radiation courses can range from 1 to 30 or more treatments depending on the lesion being treated, its location with respect to OARs, and the technique used. For example, conventional radiation therapy or IMRT often consists of about 25–35 doses of 1.8–2 Gy per fraction to the target volume in the brain or head and neck. In an IMRT plan, there is the ability to have rapid dose fall-off near OARs, such that, for example, the optic nerve may be exposed to much lower than 1.8 Gy per fraction and a lower total dose. During SRS, which is widely used for brain metastases, only 1–5 doses (fractions) with very high doses of radiation treatment per fraction (hypofractionation) are delivered. In this technique, high dose to the centre of the target is directed with a sharp gradient to the tumour leading to less radiation spillage to surrounding OARs. Due to higher doses in the target volume, the incidence of radiation necrosis is higher in SRS than in conventional fractionated or IMRT.

With few exceptions (eg, plaque brachytherapy, ocular melanoma therapy, SRS), radiation therapy plans in and around the optic apparatus are fractionated and delivered daily over 5–7 weeks. Decisions regarding the schedule are based on the tumour volume and guidance from prior randomised controlled clinical trial data (when available). In cases where the optic nerve may be included in the field, or when reirradiation is undertaken, a course of hyperfractionated radiotherapy (defined as doses <1.8 Gy per fraction, often delivered twice daily, >6 hours apart) is often considered to help spare long-term effects on late-responding tissues.2 Optic nerve and brain structures are considered ‘late responding’ tissues as toxicity generally appears several months to years after therapy.

Palliative radiation often is delivered hypofractionated, at a high dose per fraction in fewer fractions to allow for prompt relief of symptoms and completion of treatment at the cost of greater toxicity to late-responding structures. A limited life expectancy and the prompt need to complete palliative treatment prior to starting systemic therapy for metastatic or progressive disease may be the overriding priority. In the modern era of successful immunotherapy, targeted agents and increasing availability of improved outcomes of surgery and radiation, patients with metastatic disease are living longer. Therefore, late toxic damage such as optic neuropathy and cognitive decline are more common. For example, therapies such as whole brain radiotherapy (WBRT), which was designed for use at a time when life expectancy often was limited, may now result in long-term toxicity. A typical WBRT course is delivered in 3 Gy fractions, to 30 Gy.3 To illustrate the effect of 30 Gy in 3 Gy fractions on the optic nerve, the linear quadratic equation for biologically effective dose (BED) can be used.4 Employing the BED equation ( -bed-calculator) for the optic nerve and using an alpha beta ratio of 1.6, the BED equivalent is 59.2 Gy.

Indeed, in the past, it was common for radiation oncologists to offer whole brain radiation therapy (WBRT) to patients with even a small number of brain metastases.5 However, as patients are living longer such patients may be treated in a standard fashion using SRS, to reduce the risk of impaired cognition and radiation toxicity. The risk of offering a limited field therapy such as SRS may outweigh benefit if a significant number of brain metastases are not included in the radiation field.6 However, in WBRT structures important for cognition such as the hippocampi are exposed to significant doses. For this reason, efforts have been made to generate hippocampal avoidant-WBRT (HA-WBRT)7; however, this does not avoid the anterior visual pathway, which may lead to visual loss. As HA-WBRT requires lengthy, intensive planning with IMRT it often starts 1–2 weeks after the simulation with a potential knock-on delay in starting systemic therapy. Thus, HA-WBRT is not appropriate for urgently required therapy. With respect to other hypofractionated and palliative regimens, specifically for cancers of the head and neck that may overlap ocular OARs, many of these are delivered in an emergency setting using 3D conformal therapy that is amenable to rapid calculation and delivery. These regimens are for patients with severe tumour symptoms (eg, bleeding, pain) needing urgent intervention and who have a very limited life expectancy. Those who are stable enough can be treated with an IMRT hypofractionated plan in the head and neck, which may be necessary when treating prior irradiated fields. Depending on the clinical goals and the functional status of the patient, palliative regimens range from a 1 fraction course, 3 fraction course ‘the 0-7-21 regimen’ or 4 fractions delivered twice daily and repeated (‘Quad Shot regimen’), or palliative regimens ranging from 5 to 10 treatments delivered in succession (see online supplemental table 1).

Clinical presentations of RION

Patients who experience RION usually present with acute to subacute, painless, vision loss in one or both eyes; when both eyes are involved, the visual loss may be simultaneous or sequential. Similar to other optic neuropathies, the term Anterior RION (A-RION) is used to refer to RION characterised by optic disc swelling, often with peripapillary nerve fibre layer haemorrhages (although the location of injury may extend from the optic disc all the way back to the pre-chiasmal optic nerve as identified on MRI (figure 1)). Whereas RION associated with a normal optic disc is called posterior or retrobulbar RION (P-RION) and that also may involve the optic chiasm and optic tracts. A-RION versus P-RION presentations mostly depend on the tumour location: tumours in or adjacent to ocular/orbital tissues and the paranasal sinuses are the highest risk for both A-RION, whereas cavernous sinus and intracranial tumours are highest risk for P-RION; WBRT can result in either A-RION or P-RION.

Figure 1

Fundus photographs and MRI in a patient with bilateral anterior radiation-induced optic neuropathy and radiation retinopathy following WBRT. (A) Optos colour fundus image of the right eye shows optic disc swelling, a peripapillary nerve fibre layer haemorrhage at 9 o’clock, and dot haemorrhages in the macula. (B) Optos colour fundus photo of the left eye shows optic disc swelling, a nerve fibre layer haemorrhage at 6 o’clock and nasal parafoveal macular oedema. (C) axial T1 postcontrast fat saturation MRI shows bilateral optic disc elevation and enhancement (arrows) with no retrobulbar enhancement. (D) Axial diffusion-weighted image shows restricted diffusion in both optic nerve heads (arrows) and in the orbital segments of both retrobulbar optic nerves (arrowheads). WBRT, whole brain radiotherapy.

It should be emphasised that some patients with RION may have optic disc pallor at the time of presentation. In this setting, it is impossible to know if the patient initially had optic disc swelling (A-RION) that has resolved into pallor or if the patient initially had a normal-appearing disc that has become pale (P-RION).

There are three stages of radiation toxicity: acute (during treatment or within 3 weeks of completion); early/intermediate (3 weeks to 6 months after completion), and late (more than 6 months after completion). Radiation necrosis is most apt to occur in the late stage, with a mean time to onset 2 years after completion of radiation but ranging from 6 months up to 7 years. Interestingly, children may present earlier than adults, with 12% presenting in the acute phase and 72% in the early phase, with no significant difference in time of onset between those who have received initial or a repeat treatment.8

Differential diagnosis of RION

The most important differential in patients with suspected RION is tumour progression or recurrence involving the optic nerve(s) because this obviously has a substantially different prognosis and treatment direction. In most cases, the tumour either compresses or infiltrates the nerve. In addition, because of the presence of a swollen optic disc associated with peripapillary superficial (flame-shaped) haemorrhages, A-RION may be mistaken for non-arteritic or arteritic anterior ischaemic optic neuropathy, central retinal vein occlusion, or, in patients with diabetes, diabetic retinopathy.

Radiation not only can cause direct damage to the optic nerve; it also can lead to secondary optic nerve damage. For example, during radiation or shortly thereafter, radiated tissue adjacent to the optic apparatus can become oedematous, causing expansion of the tissue and compression of nearby structures that may include the optic nerve. Patients who receive radiation to the head or neck may develop vascular occlusive disease that can result in emboli to the optic nerve. Indeed, internal carotid artery stenosis develops in ~30% of patients who undergo head and neck radiotherapy, with significant stenosis occurring with an increased relative risk of 8.7.9

In cases of visual failure following radiation, there are causes other than RION or involvement by progressive or recurrent tumour to consider. For example, patients with prolonged or recurrent cancer who are on long-term immunosuppressive chemotherapies are predisposed to infectious complications such as fungal, parasitic and viral disease that can damage the nerve. In addition, patients on immunotherapies such as checkpoint inhibitors are at increased risk of autoimmune/inflammatory optic nerve damage.10

Aetiology of RION

Multiple mechanisms have been proposed for RION, including direct neural injury and vascular mechanisms. Radiation is known to cause DNA damage, with the greatest impact being on the most rapidly dividing cells.11 In the central nervous system (), this represents tumour cells, followed by vascular pericytes and endothelial cells. Pathological studies of brain radiation necrosis demonstrate a loss of pericytes and vascular smooth muscle cells.12 Similarly, in radiation retinopathy, there are marked vascular changes mimicking diabetic retinopathy capillary non-perfusion13 as well as reduced total retinal blood flow and increased oxygen saturation suggesting reduced gas exchange at the capillary level.14 This vascular injury leads to ischaemia as well as vascular incompetence, and these are compounded by the response in ischaemic tissue of increased production of vascular endothelial growth factor that exacerbates vascular hyperpermeability and tissue oedema. In volume restricted tissue such as the optic nerve, this oedema can worsen perfusion, leading to an ischaemic feedback loop.15

Intrinsic factors

The literature is conflicting regarding intrinsic risk factors for RION. For example, some studies report increased risk with known vascular risk factors such as older age, hypertension and smoking,16–18 whereas other studies have not confirmed smoking, diabetes, or hypertension as risk factors,19 and studies on choroidal melanoma suggest that younger age may be a risk factor.20

Extrinsic factors

Certain extrinsic risk factors do increase the risk of RION the greatest being the radiation dose to which optic pathways are exposed, which is dictated by the tumour type (which determines the total effective radiation dose), proximity to optic pathways, number of fractions and dose per fraction. The total dose seems to be more predictive than the mean dose per fraction; a maximum dose of >60 Gy to the optic pathways significantly increases the risk of RION.17 21–24

Certain radiation protocols, such as SRS and SRS+WBRT may increase the risk for CNS, and particularly optic pathway, radiation complications because of the low tolerance of the optic tissues to radiation. Thus, these techniques generally are avoided when dealing with tumours adjacent to the optic structures.25 26

It is difficult to determine the precise risk at any specific radiation dose due to heterogeneous reporting of radiation doses: total prescribed dose; mean dose to optic structures; maximum dose at various volumes to optic structures; actual dose; dose per fraction and relative biological effectiveness dose in addition to confounding risk factors discussed above. The generally accepted ‘safe’ total dose for the optic pathways is<55 Gy,27 with preferable fraction limits of <1.8 Gy if possible and definitely <2.0 Gy. However, both unilateral and bilateral RION have been reported at doses given in 1.8–2 Gy fractions with the total dose to the optic nerve <50 Gy24 28 as well as with WBRT at 30 Gy in 10 fractions (noting the BED is ~59.2 Gy as previously discussed).18 29–31 Three case series consisting of a total of 46 patients treated with 54 Gy of radiation for optic nerve sheath meningiomas reported RION occurring in 0 to 8%, with a mean of 4.4%.32–34

In the pre-IMRT era, the risk of RION at doses less than 60 Gy to the optic nerve was reported to range from 0–10%, with a mean of 3.3%. At doses of 60 Gy, the range was 10%–34%, with a mean of 20%.21 22 28 35–41 With IMRT, rates of RION between 50 and 60 Gy exposure to the optic nerve have been estimated at 7% and above 60 Gy, 13%.24 Harris and Levene reported an 18% incidence of RION among patients with suprasellar tumours treated with conventional radiation therapy with fractions of 2.5 Gy and above.42

Concurrent chemotherapy has been reported to increase the risk of RION. Although many chemotherapeutic agents are designed to increase the efficacy of radiation when given concurrently, many of these agents also potentiate toxicity and others are known to have intrinsic risk for optic neuropathy (eg, PL1 inhibitors).43

Finally, patients receiving reirradiation at or around the optic nerve are at a higher risk for RION than initial radiation, particularly if given within 10 years of initial radiation.


It appears that the potential for RION in children and in adults who undergo radiation therapy is similar. In one series, in children who received proton beam treatment for paediatric esthesioneuroblastomas, one of six optic nerves that received a maximum dose above 55 Gy Equivalent (GyE) developed optic neuropathy.44 In a series of paediatric craniopharyngiomas, optic neuropathy was attributed to radiation in one patient who received 55 Gy at fractions of 2 Gy to the tumour.38 A review of craniopharyngioma treatment in children found that toxicity was very rare at cumulative doses of 54 Gy and fractions of 1.8 Gy or less, similar to adults.45


The diagnosis of RION is based on a comprehensive eye exam including: best-corrected visual acuity; colour vision; quantitative perimetry; pupil light responses; and a dilated fundus examination to rule out mimickers and other radiation complications. Perimetry, preferably automated aids diagnosis, monitoring of progression and response to treatment. For severe vision loss, a Goldmann size V stimulus protocol is recommended. Next, a high-resolution MRI with thin slices (preferably <1 mm) with and without contrast and with fat saturation sequences is crucial as contrast-enhanced T1-weighted images typically show enhancement of a discrete segment of the intracranial prechiasmatic optic nerve (figure 2), often with accompanying expansion and T2 hyperintensity.46 The enhancement is time limited (2–17 months in one study), at which time visual function usually stabilises. Occasionally, enhancement precedes visual loss. High-resolution diffusion-weighted (figures 1 and 3) and perfusion imaging can also be helpful.47–49 Discussing the clinical scenario and concerns with a radiologist can help optimise protocols, particularly as the radiologist may not consider RION in a patient with known cancer and visual failure.

Figure 2

T1 gadolinium-enhanced MRI study in a patient with bilateral posterior radiation-induced optic neuropathy. (A) Axial, (B) sagittal, (C) coronal postcontrast MRIs show enhancement of the prechiasmatic optic nerves (white arrows).

Figure 3

MRI in a patient with left posterior radiation-induced optic neuropathy. (A) Axial diffusion-weighted imaging. (B) Corresponding axial T1 postcontrast MRI also shows restricted diffusion without contrast enhancement of the prechiasmal and canalicular left optic nerve (white arrows).

Treatment options for RION


Several randomised controlled trials have shown safety and efficacy of intravenous bevacizumab compared with corticosteroids in patients with CNS radiation toxicity.50 51 Single-arm studies have shown improvement in patients both corticosteroid dependent and refractory as well as those refractory to anticoagulation or hyperbaric oxygen.52–54 The most commonly studied and used protocol for intravenous bevacizumab is a single dose of 5 mg/kg every 2 weeks weeks for four rounds, but doses used range from 1 mg/kg to 10 mg/kg given every 2–4 weeks.50 51 55 A systematic review of patients treated with bevacizumab for CNS radiation toxicity found 93% had evidence of improvement on MRI, with 40% having complete resolution of neurologic deficits, 8% having improvement in deficits and 10% experiencing stabilisation of deficits, respectively.56 In addition, 97% of patients were able to reduce or stop corticosteroids.

To date, there have been no prospective clinical trials assessing the effectiveness of intravenous bevacizumab on RION; however, several case reports and series have documented improvement or stabilisation of vision.57 58 Gondo et al described a patient whose A-RION and maculopathy following radiation for maxillary sinus cancer completely resolved with improvement of visual acuity from 20/50 to 20/20 after a single intravitreal injection of bevacizumab.59 One of us (ARC) has observed stabilisation of vision with intravitreal bevacizumab in several patients with A-RION when limited to the intraocular optic nerve (papillopathy). On the other hand, a large case series of patients with RION following proton therapy for choroidal melanoma failed to show any benefit from intravitreal injections.60 It is unclear if these data can be extrapolated to other treatment protocols, as the radiation for proton therapy is hypofractionated (usually 4–5 fractions with cumulative dose ranging 56–70 GyE for individual fractions of 14–15 GyE).61 Because of the time-sensitive nature of RION, the authors do not recommend withholding bevacizumab while trying alternative treatments.

A number of treatments for RION other than bevacizumab have been proposed based on individual cases or retrospective assessments, including corticosteroids, heparin, pentoxifylline and hyperbaric oxygen. Although a few case reports show efficacy with each, none has been shown to provide consistent benefit. For example, there are a few case reports of patients with improved vision after treatment with hyperbaric oxygen. In all cases, initial visual acuity was 20/70 or better, or the visual deficits were hemianopias not RION, and treatment was initiated within 72 hours of vision impairment in the eye that improved.62–64 A retrospective comparison of patients treated with corticosteroids in China identified 68 eyes with RION that responded to corticosteroids and 64 eyes that did not; the only differentiating factor identified by the authors was blood urea nitrogen level although the difference was small (4.8 mmol/L vs 4.1 in responders vs non-responders, respectively, with reference range of 2.1–8.5 mmol/L, p=0.002).65 Although a multifactorial model resulted in an area under the receiver operator curve of 0.876, this was not validated with an external data set. In addition, the authors did not report a number of important data, including radiation maximum dose to the optic pathways, time from onset of vision loss to initiation of treatment, magnitude of vision loss or degree of vision improvement.

Pentoxifylline has been advocated for use in RION based solely on its benefit in radiation-associated soft-tissue toxicity66; however, the one randomised control trial on which the recommendation was based demonstrated a delayed effect of several weeks, which is likely beyond the therapeutic window for RION based on hyperbaric oxygen data.67 Thus, there remains no convincing support for the use of this drug in CNS disease, let alone RION.

Prevention of RION

Although there are no proven or accepted preventative treatments for RION, there are some data that can be extrapolated from early stage animal research. For example, models of brain necrosis from SRS have evaluated prophylactic vs treatment benefits of ACE-1 inhibitors, mainly ramipril, as well as bevacizumab. Both agents showed efficacy in preventing brain toxicity compared with their effects once brain necrosis had already begun.68 In addition, the prophylactic effects of bevacizumab have been studied in patients undergoing radiation for uveal melanoma both via plaque brachytherapy and proton therapy in a single-centre retrospective study using intravitreal bevacizumab every 4 months.69 The investigators compared the results in 292 patients who received bevacizumab with 126 controls. Although the authors found an absolute risk reduction in the development of cystoid macular oedema in bevacizumab-treated patients from 40% to 26% (p=0.004, number needed to treat (NNT)=7) and a reduction in clinically evident radiation maculopathy from 31% to 16% (p=0.001, NNT=7), with a concomitant reduction in moderate and severe vision loss, the study failed to show any benefit for ‘clinically evident papillopathy’ as diagnosed by ocular oncologists with a reduction from 10% to 7%, (p=0.4). As discussed above in the treatment section, it is unclear if these data can be extrapolated to other radiation protocols in view of the high dose per fraction in proton protocols and the continuous radiation given during plaque brachytherapy, with an average dose of 70–80 Gy which results in a ~40% rate of radiation maculopathy. A meta-analysis of retrospective studies looking at prophylaxis of RION in patients treated for melanoma concluded that there was evidence of protective effects from intravitreal bevacizumab,70 although two of the three studies cited by the authors included repeated data, so it is unclear if and how this may have impacted their conclusions.


Patients with cancer and oncologists are accustomed to a team-based approach. This strategy can potentially help identify patients at high risk of developing RION and establish an effective monitoring schedule that could lead to early diagnosis.

Patients at highest risk for RION are those with choroidal melanoma and those with head and/or neck squamous cell carcinomas due to the high dose needed for tumour control and proximity of the lesions to the orbits. Radiation therapy for primary and metastatic tumours in the sellar and parasellar regions and those receiving WBRT is also high risk. Predicting an exact risk based on radiation dose alone is problematic as other factors increase risk, including: prior optic neuropathy; concurrent chemotherapy; age and vascular risk factors.

Although no consistently beneficial treatments for RION are known, prompt treatment with either intravenous (P-RION) or intravitreal (A-RIOP) bevacizumab may have the potential to improve or at least stabilise vision. Further research is needed to refine preventive therapies and treatment options. Improvement in risk determination will enable better monitoring of patients at high risk permitting early diagnosis and treatment before irreversible and severe vision loss occurs.

Data availability statement

Data sharing not applicable as no datasets generated and/or analysed for this study.

Ethics statements

Patient consent for publication


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Twitter @drewcareyMD

  • Contributors NM had the idea for the article, ARC and BRP performed the literature search, ARC and BRP wrote the article, all contributors had final review and approval, ARC is the guarantor.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; internally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.