Article Text

Download PDFPDF

Visual outcomes following intraophthalmic artery melphalan for patients with refractory retinoblastoma and age appropriate vision
  1. Maria Tsimpida1,
  2. Dorothy A Thompson2,
  3. Alki Liasis2,
  4. Vicki Smith1,
  5. Judith E Kingston1,3,
  6. Mandeep S Sagoo1,4,5,
  7. M Ashwin Reddy1,2,4,5
  1. 1Retinoblastoma Unit, Royal London Hospital, London, UK
  2. 2Clinical and Academic Department of Ophthalmology, Great Ormond Street Hospital for Children, London, UK
  3. 3Paediatric Oncology, Great Ormond Street Hospital, London, UK
  4. 4Moorfields Eye Hospital, London, UK
  5. 5UCL Institute of Ophthalmology, London, UK
  1. Correspondence to M Ashwin Reddy, Retinoblastoma Unit, Royal London Hospital, Whitechapel Road, London E1 2LL, UK; ashwin.reddy{at}


Background/aims To determine the frequency and cause of visual loss following intra-arterial melphalan (IAM) in patients with retinoblastoma with age appropriate vision.

Methods Assessment of patients with refractory retinoblastoma that had undergone systemic chemotherapy, with or without local treatment, and were subsequently treated with IAM. Eyes of patients with a healthy foveola were assessed. The main outcome measures included visual, macular (including Pattern Visual Evoked Potentials and Fundus Fluorescein Angiography) and retinal functions (Electroretinograms).

Results Five of twelve eyes (42%) demonstrated severe visual loss following IAM at last follow-up (median 21 months). This was due to either retinal detachment (1 eye, 20%) or choroidal ischaemia involving the foveola (4 eyes, 80%). All 3 eyes that had technical difficulties or vasospasm during catheterisation suffered visual loss. 8 out of 10 eyes that had a non-age adjusted dose of melphalan suffered visual loss. Electroretinograms post-IAM deteriorated in 4 of 8 eyes (50%) and Pattern Visual Evoked Potentials deteriorated in 3 (37%), though only one of these 3 showed concomitant visual acuity loss.

Conclusions Structural and vascular damage to the foveola limited visual acuity. Complications associated with catheterisation and high doses of melphalan may be contributory factors to visual morbidity. Although visual loss is described, no patient developed metastases and most retained good vision.

  • Retina
  • Vision
  • Neoplasia
  • Electrophysiology
View Full Text

Statistics from


Super-selective intra-arterial melphalan (IAM) treatment for retinoblastoma has recently gained popularity. Following the initial report in 2008 by Abramson et al,1 several studies have confirmed successful tumour control.2–7 However, IAM is not without the risk of iatrogenic effects, such as chorioretinal atrophy4 ,5 ,8 ,9 and the broader question has been raised whether sparing the globe represents success when the eyes have guarded visual potential.10

It is essential to appraise the safety profile of any new treatment, and while tumour control has been documented,3 ,5 relatively little attention has been paid to the effects on vision. Munier et al8 reported final visual acuity (VA) of 20/50 in 31%, but did not describe the proportion of eyes starting with good visual potential, that is, foveola free of tumour. The results of electroretinograms (ERGs) under anaesthesia and fundus fluorescein angiography (FFA) have been used as proxies for vision in systemic11 and intra-arterial chemotherapy.12 ,13

We present our results on the effect of IAM on visual potential in eyes with refractory retinoblastoma where the foveolae were spared. We gathered behavioural assessments pretreatment and post-treatment. Additionally we performed visual evoked potentials (VEPs), flash ERGs (where possible) and FFAs to document and investigate visual loss.13

Materials and methods

A prospective study of patients with refractory retinoblastoma receiving IAM from May 2009 to July 2012 was conducted. Eyes with tumours involving the foveola or causing subretinal fluid extending to the foveola were excluded from this study. Approval for use of IAM was obtained from the Great Ormond Street Hospital Children Drugs and Therapeutics Committee and Barts Health Clinical Effectiveness Unit (#761) within tenets of the Declaration of Helsinki. Informed consent was obtained from the parents or legal guardians, after discussion of the findings, potential risks and benefits of the procedure. IAM was considered in cases where the tumours failed to respond adequately to previous treatments or there was a new recurrence. All patients had received systemic chemotherapy in the form of six cycles of carboplatin, vincristine and etoposide . Local treatments including external beam radiotherapy (EBRT), plaque brachytherapy, cryotherapy and laser were used as deemed necessary by the senior ophthalmologists (MSS and MAR).

Our method of intraophthalmic artery melphalan has been previously reported.5 IAM was injected directly into the ophthalmic artery in only one eye in all cases at each session. The dose and number of cycles (spaced at 4 weeks) varied as our concern regarding the toxicity of melphalan increased. All patients had an examination under anaesthesia 3 weeks after each treatment. FFAs were performed in patients who had a visible abnormality of the fundus following treatment. ERGs and VEPs before and after IAM infusion were scheduled to coincide with attendance for the procedure and were obtained whenever possible.

Pattern and flash VEPs were recorded according to ISCEV (International Society for Clinical Electrophysiology of Vision) standards14 from three occipital electrodes; O1, Oz and O2 referred to FpZ. Pattern reversal VEPs (PrVEPs) were elicited to high contrast checkerboards. If PrVEPs were affected by nystagmus, we also acquired pattern onset VEPs to the appearance of the checkerboard for 230 ms followed by a uniform blank screen of mean luminance of 300 ms. Data from the midline Oz were analysed and reported in this paper. Pattern VEPs (PVEPs) and flash VEPs (fVEPs) provide a qualitative estimate of the vision level the pathways to the striate cortex may support, for example vision levels are deemed ‘good’ if PrVEPs are evident to 50′ or smaller checks, ‘moderate’ if present to 100′ or 200′, ‘poor’ if only present to checks larger than this and rudimentary if flash VEPs, but no PVEPs, are evident.

ERGs were recorded from periorbital skin electrodes to flash stimuli from a handheld Grass strobe under photopic and scotopic conditions, which are physiologically akin to the diagnostic ISCEV standard protocol.15 Standard flashes (gr4) scotopically elicit mixed rod cone ERGs and photopically cone ERGs.16

As part of our protocol, patients had orthoptic examinations before and 3 weeks after each IAM treatment. This included VA assessment, cover testing at near (1/3 m) and distance (6 m), ocular motility examination and investigation of binocular vision. Visual acuities were assessed using Cardiff Cards, Keeler Cards, Kays picture tests, Single Sheridan Gardner Tests, linear Snellen acuity and Crowded LogMar, depending upon the age of the child. When possible VA was assessed uniocularly, otherwise binocular VA was measured. If quantitative assessment was not possible qualitative methods were used, that is, fixing and following on a target and/or fixation preference.

Information regarding age at the time of first IAM treatment, laterality, fellow eye status, previous local treatments, reason for IAM treatment, number of IAM sessions, dose of melphalan infused, follow-up period and postmelphalan complications in the form of retinal detachment and/or vasculopathy (retinal or choroidal) changes were collected. The distance of the tumour from the foveola was also noted. Prior subretinal fluid was assessed clinically, and where possible with Spectralis OCT imaging (Heidelberg Engineering, Carlsbad, California, USA). VA levels were defined as AAN (age appropriate normal) or below normal. For patients who could not cooperate with Snellen graded tests, (ie, Kays, Sheridan Gardner and Snellen), a loss of at least two preferential looking cards was taken as a significant change.17 ,18


From May 2009 to July 2012, 31 eyes of 28 patients with refractory retinoblastoma were treated with IAM in our department. Nineteen eyes with tumours involving the foveola were excluded. Table 1 lists the baseline patient and ocular features of the 12 included eyes from 12 patients. The median age at the time of the first IAM treatment was 18 months (range 9–82 months).The mean distance of the tumour to the foveola was 2 mm (median 2 mm; range 0.2–11 mm). The absence of subfoveal fluid on funduscopy was noted and was confirmed in three patients using OCT imaging.

Table 1

Visual outcomes following intraophthalmic artery melphalan for retinoblastoma: summary of patient and ocular features

All patients were alive at last follow-up with no indication of metastases. Tumour control was achieved in nine eyes (75%) in this group and the other three eyes (25%) required EBRT (two eyes) and cryotherapy (one eye). The side effects of IAM to seven patients have previously been reported.5


The indications for using IAM were tumour edge relapse in 10 patients and persistent vitreous seeding in two eyes. Table 2 lists the number, dose and complications of IAM, with tumour and visual outcomes. The mean number of IAM cycles was 2.4 (median 2; range 1–5). The standard melphalan dose was 5 mg in 30 mL solution. This was increased to 7.5 mg in two eyes in an attempt to control tumour activity. A lower dose of 3 mg was delivered to one eye (patient 10), due to concerns regarding toxicity. Eight of ten patients given a relatively high dose dependent on age suffered structural or vascular damage. Median follow-up, from the first IAM to last visit was 21 months (range 8–35 months).

Table 2

Visual outcomes following intraophthalmic artery melphalan for retinoblastoma: Dose, complications and results

Three patients had technical difficulties from catheterisation of the ophthalmic artery. One eye (patient 8) had ophthalmic artery vasospasm during cannulation for the attempted fourth dose of IAM. The procedure was abandoned without injection of melphalan and the child volunteered that he could not see with that eye. This example indicates that difficulty with catheterisation can be associated with severe visual loss. In the other two instances (patients 1 and 5) anatomical variation was overcome successfully without vasospasm. However, these eyes sustained loss of visual function. During ophthalmic artery catheterisation, a physiological reaction with fall in tidal volume and hypotension (low blood pressure (BP)) occurred in five cases (patients 2, 4, 6, 11, 12) with visual loss in one (patient 2).

Visual acuity

VA measurements before and after each IAM session are depicted in table 3 and summarised in table 4. Monocular VA assessment was achieved in seven eyes. Before IAM, all 12 eyes had AAN VA.16 ,17 None of the seven eyes with stable VA at the last visit had received radiotherapy prior to IAM, and two eyes developed sectoral retinal pigment epithelium (RPE) changes not involving the foveola.

Table 3

Visual outcomes following IAM for retinoblastoma: functional, vascular (FFA) and structural results

Table 4

Visual outcomes following intraophthalmic artery melphalan for retinoblastoma: summary of visual acuity results

At the end of the study period five eyes (42%) had severe VA loss: two eyes (patients 8, 9) lost >6 lines of VA using a Snellen graded test and three younger patients were deemed to have severe VA loss from preferential looking methods or fixation behaviour, for example, patient 1 had both eyes open, a VA of 6/30 (20/600) with Cardiff Cards (AAN) with a left preference (treated eye); post-IAM a right fixation preference developed and she was not able to fix or follow with the treated eye.

Fundus changes

Before IAM treatment there were no eyes that exhibited choroidal, retinal or RPE changes attributable to prior systemic chemotherapy or radiotherapy. All patients with fundus changes potentially due to toxicity after IAM had a FFA. Reasons for vision loss included diffuse retinal detachment (RD) in one eye (patient 1), diffuse choroidal ischaemia in two eyes (patients 2, 8) and foveola involving sectoral choroidal ischaemia in two eyes (patients 5, 9). The median time interval between the first IAM treatment and the development of chorioretinal atrophy or retinal detachment was 7 months (ranging from 3 months to 21 months). FFAs showed delayed choroidal filling and segmental window defects secondary to RPE loss that evolved to chorioretinal atrophy over time in all cases. No retinal vascular abnormalities such as artery occlusions were noted.

Four of the five eyes that lost vision had been previously exposed to radiation either EBRT (1 eye, patient 2) or plaque brachytherapy (3 eyes, patients 1, 8, 9). There were two eyes (patients 3, 11) with macula-sparing choroidal ischaemia that had no prior radiation, suggesting that the IAM procedure itself can cause vasculopathy.

Electrodiagnostic tests

The ERG and VEP results before and after IAM were available for 8 of 12 eyes, (patients 3, 5, 6, 7, 8, 9, 11, 12), before in one eye (patient 4) and after in another (patient 10) (table 3). The baseline ERGs of affected eyes tended to be smaller than fellow, normal eyes. The ERG summates electrical activity from intact retinal areas. While amplitude reduction reflects loss of retina, and in retinoblastoma most likely reflects the extent of tumour, time to peak delays in the ERG typically reflect dysfunctional retina and are independent of macular involvement, as exemplified in figure 1. Deterioration of the ERG was regarded as an amplitude reduction of 30% and time to peak increases <3 ms to standard flashes.

Figure 1

Composite showing pretreatment extent of retinal damage from tumour in LE of patient 8, and the pretreatment lower amplitude baseline LE electroretinogram (ERG) cf fellow normal eye. The rod driven and cone b-waves are reduced in amplitude, reflecting loss of retinal function and the cone b-wave is delayed indicating dysfunction of the central cone rich regions. A post-treatment OCT image shows preserved foveolar architecture. Access the article online to view this figure in colour.

Following IAM, ERGs deteriorated in the treated eye of four patients, (5, 6, 7 and 12), while ERGs were stable in four other patients (3, 8, 9, 10). In patients 12 and 7 the b-wave, rather than a-wave, is reduced indicating greater dysfunction of the inner retina.

VEPs were stable in five of eight (63%) eyes and deteriorated in three (37%) eyes (patients 3, 5 and 7), that subsequently developed sector RPE changes. In cases 5 and 7, deterioration in PVEP was associated with deterioration in ERGs, but in case 3 the ERG was stable. Five patients had manifest latent nystagmus or binocular deficiency nystagmus syndrome19 due to early disruption of binocularity. PVEPs were obtained to the smallest checks in presence of manifest latent nystagmus for example, patient 11.

Grades of visual potential inferred from the PVEP findings suggested 6/10 patients could support good vision, evidenced by VEPs to 6.25′, the smallest checksize. Patients 3, 5, 7 showed PVEP evidence of poorer macular function post-IAM, though two of these eyes achieved AAN vision at final follow-up.


We present our results on visual function, by means of VA assessment and VEP, in addition to retinal function (ERG testing), following IAM infusion in patients with refractory retinoblastoma. This report is unique in monitoring VEP changes secondary to macular pathway dysfunction following any treatment for retinoblastoma. We were careful to include only eyes with a healthy foveola and found 5 of 12 eyes (42%) sustained a reduction in VA at last follow-up. All five eyes that lost vision had either diffuse retinal detachment or choroidal ischaemia involving the foveola. A further two eyes showed deterioration of the pattern VEP without alteration of AAN vision; these eyes developed sectorial ischaemia that spared the foveola.

Understanding the effects of intra-arterial chemotherapy on visual outcome is essential in order to counsel families about this treatment. Munier et al8 reported that 4 of 13 patients retained vision of 20/50 or better, but did not state the proportion of patients who had good acuity or tumour-free foveolae pretreatment. Post-treatment they found 3 of 13 (23%) had choroidal ischaemia or retinal arteriolar embolism. Shields et al4 found 6 of 17 eyes (35%) had choroidal vasculopathy or retinal artery occlusion.

This study looked at a cohort of patients who had good vision before IAM treatment. Brodie et al12 found that retinal function, as assessed by ERG testing, improves in patients with retinal detachments. This has been extrapolated to suggest that vision improves. By limiting our study to patients with normal pretreatment vision, we can provide information for this group. Patients with poor vision before IAM should be considered separately.

We assessed the impact of previous treatments on the final VA and none of the patients who retained good VA following melphalan treatment had been exposed to previous radiotherapy treatment. However two patients, who had not been exposed to radiation, developed chorioretinal changes confined to non-foveal areas. There is a suggestion from our results that any form of radiation before use of IAM is associated with worse complications. Also the distance of the tumour from the foveola was not a factor in final VA.20

Three eyes developed visual loss following difficulty with catheterisation. A method of delivering melphalan without direct catheterisation of the ophthalmic artery (via a balloon obstructing the carotid artery) might be expected to have a lower risk of visual loss. However, only 51% of eyes without a macular tumour achieved 20/40 vision or better.6 In this study, catheterisation of the ophthalmic artery was performed to improve stability and to increase the dose available to the eye. An increased risk of ophthalmic artery trauma may have contributed to complications and visual loss. There will be a learning curve for any procedure and this may be reflected in the complications from this cohort.

We started treatment in 2009 when the standard dose of melphalan was 5 mg.4 Gobin et al21 reduced this dose according to age in order to minimise toxicity and 8 of 10 children who were given a standard rather than reduced dose (table 2) suffered structural/vascular damage. Two patients who were given an age appropriate dose did not. The dose of melphalan is a compelling factor.

ERGs under anaesthesia have been used as a proxy for vision in systemic11 and intra-arterial chemotherapy.12 The ERG outcome measure reported mostly is the 30 Hz flicker: an ERG index of generalised cone function. This shows mainly stability, though deterioration or recovery can also occur.11 ,12 ,22 ,23 Assessing retinal macular function directly with pattern ERGs or multifocal ERGs in an awake child demands prolonged fixation, or under anaesthesia requires refraction and sophisticated stimulation. Neither is easily achieved, but it is possible to use PVEPs as an indirect measure of function of the central 2–10° of the macula and these can be performed in awake pre-verbal children.24

In contrast to previously reported cases none of our patients showed an improvement of the ERG.11 ,12 ,22 ,23 This may be attributed to the fact that no eyes treated in our series had an ERG extinguished secondary to extensive retinal detachments that resolved following treatment. Four of 8 patients tested showed some deterioration in flash ERG following treatment. Two of these patients (3 and 7) also had PVEP evidence of macular pathway dysfunction, but did not have evidence of visual loss at last follow-up. This emphasises the fact that behavioural measures reflect different attributes of visual function. Another possibility is that the electrodiagnostic results reflect a transient change.

To conclude, VA was reduced in 42% (5/12) of eyes with refractory retinoblastoma and good visual potential, treated with IAM. This rises to 58% (7/12) when PVEP evidence is included. The main reason for vision loss was the development of choroidal ischaemia in the previously unaffected foveolar area. The exact cause is unknown but melphalan toxicity and difficulties in catheterisation of the ophthalmic artery are implicated. ERGs deteriorated in half of children tested suggesting that this treatment can have a negative effect on the entire retina. The predisposition to ocular toxicity after IAM still avoids enucleation of the eye and in our experience many parents would like to avoid enucleation even if ocular toxicity was a risk. Further studies are needed to establish the IAM dose that would achieve maximum tumour control and minimum ocular morbidity, with the best preservation of the pretreatment visual function.


The authors acknowledge Dr Tanzina Chowdhury for her assistance with the oncology care of patients in this study. We would also like to acknowledge Dr Stefan Brew for his care of these patients in the department of radiology at Great Ormond Street Hospital for Children.


View Abstract


  • Contributors MAR takes full responsibility for the accuracy of this paper. All authors were involved in the conception and design, analysis and interpretation of data and drafting the article for important intellectual content and final approval of the version to be published.

  • Competing interests None.

  • Ethics approval Clinical Effectiveness Unit, Barts Health (#761).

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

  • Data sharing statement DAT has the original data from electrodiagnostic tests. MT and MAR have clinical data available on these patients.

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.