Abstract and Introduction
Purpose of Review: The purpose of this review is to summarize complications of treatment for retinopathy of prematurity (ROP) and to compare complications of laser and intravitreal antivascular endothelial growth factor (VEGF) injections.
Recent Findings: Poor structural outcomes and myopia are more common with laser for severe ROP than with anti-VEGF. Clinical trial data show unfavourable outcomes in 9.1–9.5% of laser treated, and 1.4–3.6% of anti-VEGF treated eyes. Additional randomized trial data show risk for very high myopia (≥-8.00D) to be 3.8 and 51.4% for zone I eyes treated with bevacizumab and laser, respectively. However, anti-VEGF may be complicated by late recurrence and is more likely to require retreatment than laser. Laser often necessitates general anaesthesia with its attendant risks, including worse short-term respiratory outcomes. Neurodevelopmental complications have been reported with anti-VEGF, but existing studies are subject to bias.
Summary: Treatment complications are substantially different for the two modalities in common use today. In more severe cases, risk of poor structural outcome and myopia favour treatment with anti-VEGF. In less severe ROP, risk of recurrence and the need for additional treatments may favour laser. Additional data are needed to establish comparative risks of neurodevelopmental complications.
New complications of retinopathy of prematurity (ROP) treatment have naturally emerged as treatment for ROP has evolved. Today, ophthalmologists caring for infants who need treatment for ROP have two primary options to consider: laser ablation of peripheral avascular retina and injection of antivascular endothelial growth factor (anti-VEGF) agents. The choice between these two modalities is, in many ways, driven by considerations of potential complications for each. In this article, we will review, with an emphasis on prospectively collected data, the major complications of ROP treatment and will compare laser ablation to anti-VEGF injections.
We divide complications of ROP treatment into two sections: one that considers local ocular complications and another that considers systemic complications. This section considers local ocular complications.
Unfavourable Structural Outcome. Clinical trials in ROP have traditionally used 'unfavorable structural outcome' as a major study endpoint. This endpoint is generally defined as the presence of posterior retinal fold, retinal detachment or opacified cicatricial tissue behind the lens.[1,2] The pivotal Early Treatment for Retinopathy of Prematurity (ETROP) trial utilized this endpoint and, in 2003, established the rate of this treatment complication for laser. ETROP randomized 401 infants to either early or conventional laser treatment and found unfavourable structural outcomes at 9 months were significantly reduced with early laser [9.1 vs. 15.6% (P < .001)]. This rate of this complication with laser was essentially duplicated years later in another randomized trial of ROP published in 2019 (RAINBOW). In RAINBOW, or Ranibizumab Versus Laser Therapy for the Treatment of Very Low Birthweight Infants with Retinopathy of Prematurity: an Open-Label Randomized Controlled Trial, seven of 74 eyes (9.5%) treated with laser met the endpoint of unfavourable structural outcome. Results of these two prospective trials firmly established the rate of unfavourable structural outcome for laser as approximately 9.1–9.5%.
For anti-VEGF therapy, there are also prospective clinical trial data on unfavourable structural outcomes. RAINBOW compared two doses of ranibizumab (0.1 and 0.2 mg) to laser. The primary outcome was survival with no active ROP, no unfavourable structural outcomes and no need for different treatment modality at or before 24 weeks. Unfavourable structural outcomes for ranibizumab were found in five of 77 eyes (6.5%) treated with 0.1 mg, and one of 70 (1.4%) eyes treated with 0.2 mg. For bevacizumab, the most robust data comes from the Dosing Study of Bevacizumab for ROP sponsored by the Pediatric Eye Disease Investigator Group (PEDIG).[4,5] In this prospective trial studying de-escalation of bevacizumab doses, unfavourable structural outcomes at 6 months were found in four of 112 eyes (3.6%). Doses in this study ranged from 0.25 mg down to 0.031 mg. No association between dose and risk for unfavourable outcome was detected, but the number of eyes studied was too small to detect a mild association.
A large comparison trial recently examined short-term retinal detachment rates 8 weeks after treatment with either laser or anti-VEGF, an endpoint closely related to unfavourable structural outcome. This trial was a secondary analysis of data from the retrospective G-ROP-1 study (2006–2012) and prospective G-ROP-2 study (2015–2017) and reported 1167 eyes receiving either laser (1003 eyes) or anti-VEGF (164 eyes). The authors found a significant difference in retinal detachment rate between laser and anti-VEGF among eyes needing treatment before 36 weeks post menstrual age (PMA): 7.9% (29/368) compared with 0% (0/90) for laser and anti-VEGF respectively (P < 0.001). However, for eyes requiring treatment at 36 weeks PMA or later, there was no significant difference in retinal detachment rate between laser (20/635, 3.1%) and anti-VEGF (1/74, 1.4%) (P = 0.27).
Myopia. Myopia's association with ROP has been known for many years. More severe ROP is associated with more severe myopia. Whether treatment of ROP has any influence on myopia was uncertain until the anti-VEGF era of ROP treatment. Data from Cryo-ROP suggested that peripheral ablation with cryotherapy did not necessarily exacerbate the tendency for myopia, but rather preserved the integrity of more severe eyes that would have otherwise developed retinal detachment. Severely affected eyes were thought to be more susceptible to high myopia resulting from severe ROP and not from the treatment. However, studies comparing anti-VEGF to laser for ROP have demonstrated that myopia is less severe with anti-VEGF treatment than it is with peripheral ablation. As such, severe myopia may be thought of as a complication of peripheral retinal ablation for ROP.
In the Bevacizumab Eliminates the Angiogenic Threat for ROP Study (BEAT-ROP), refractive outcomes were compared between patients treated with bevacizumab monotherapy (0.625 mg) and near confluent laser therapy. At age 2.5 years, the mean spherical equivalent results were -1.51 D in bevacizumab treated vs. -8.44 D in laser treated (P < .001) for zone I ROP, -0.58 D in bevacizumab treated vs. -5.83 D in laser treated (P < .001) for zone II ROP. Very high myopia (≥-8.00 D) occurred in 3.8 and 1.7% of zone I and zone II eyes, respectively, treated with bevacizumab compared with 51.4 and 36.4% of eyes treated with laser (P < 0.001). Authors have speculated that improved retinal vascular development with anti-VEGF treatment allows for more normal levels of the local growth factors required for proper signalling cascades in anterior segment development. Most,[12–16] but not all,[17,18] of the other published studies on refractive changes in ROP treatment report reduced myopia with anti-VEGF compared with laser, and similar refractions in anti-VEGF treated eyes compared to eyes with spontaneous regression of ROP.
Recurrence of Retinopathy of Prematurity. The possibility of recurrence after treatment has received increased attention since the introduction of anti-VEGF for ROP. With laser peripheral ablation, a single treatment session was considered definitive, permanent treatment in most cases. There was some debate about the density of the laser pattern required to achieve permanent regression,[20,21] and there was an understanding that inadvertent skip areas could result in persistent activity, but, when proper technique was followed, laser was considered to be an effective onetime treatment.
With the introduction of anti-VEGF therapy for ROP, reports emerged describing severe recurrence with dismal outcomes after what often appeared to be successful initial treatment.[23–25] There are now multiple reports of late ROP recurrence developing months or even years after anti-VEGF injection initially showing a good response.[26–30] Given these late recurrences, the appropriate endpoint for ROP monitoring after anti-VEGF therapy remains unknown. As a result, some authors recommend prophylactic laser treatment to persistent avascular retina after 60 weeks PMA to prevent late recurrence.[27,31] Moreover, prophylactic laser to persistent avascular retina does not seem to negate the refractive benefits of primary anti-VEGF.
Although late and very late recurrence after successful initial response seems to be a distinguishing feature of anti-VEGF for ROP, recurrence and need for retreatment do occur with laser as well. For instance, in ETROP retreatment was performed in 11 and 14% of conventionally managed and early treated eyes, respectively. However, when comparing the ETROP retreatment rates to the recent PEDIG dose de-escalation study of bevacizumab for ROP, quite a contrast is seen. In the PEDIG trial, bevacizumab-treated eyes needed additional treatment 41% of the time (25/61 eyes). Granted, some of these retreatments were for persistent avascular retina after 50 weeks PMA (11/61 eyes) and were not for recurrent ROP. However, 14 out of 61 eyes (23%) had either early failure of treatment (before 4 weeks) or late recurrence of ROP (after 4 weeks and out to 6 months). This prospective PEDIG trial used de-escalating doses bevacizumab down to 0.031 mg. Would higher doses yield lower retreatment/recurrence rates? The PEDIG trial detected no relationship between lower doses and need for additional treatment. The sample size was, however, too small for a definitive conclusion about a relationship between dose and recurrence. The BEAT-ROP trial found a recurrence rate of 4% (6/140 eyes) of eyes treated with 0.625 mg bevacizumab. However, these eyes were only followed to 54 weeks PMA.
The RAINBOW trial reported recurrence/retreatment rates for ranibizumab. In this trial, the ranibizumab groups were allowed two additional injections at 28-day intervals. A switch to a different treatment method was considered a treatment failure and, for our purposes, will serve as a proxy for recurrent of ROP. Treatment switch occurred in 11 out of 70 eyes (15.7%) with 0.2 mg, 13 out of 76 eyes (17.1%) with 0.1 mg and 18 out of 68 eyes (26.5%) with laser. Eyes in this trial were followed for 24 weeks. With longer follow-up, recurrence in ranibizumab-treated eyes may be higher.
Since BEAT-ROP was published, numerous retrospective studies and one small prospective study have reported recurrence and retreatment rates after various doses of bevacizumab and ranibizumab for ROP in patients followed for varying lengths of time. Reported recurrence/retreatment rates range from 0 to 64%.[12,13,15,17,33–41] Some studies have found a higher recurrence rate for ranibizumab than for bevacizumab.[13,42] One study reported a comparison of 0.625 to 0.0625 mg of bevacizumab with 45 patients in each group. The plus resolved more quickly in the high-dose group, but progression of vascularization into the peripheral retina was better in the low-dose group. Two patients (4.4%) in each group required retreatment.
Endophthalmitis. The incidence of endophthalmitis after intravitreal injection in adult eyes is less than one per 1000.[44,45] For infants with ROP, we do not have good data on risk of endophthalmitis. There have, however, been reports of endophthalmitis and outcomes can be very poor.[46,47] We know that endophthalmitis in children generally is associated with a much worse prognosis than adults. So, the potential complication of endophthalmitis after intravitreal injection for ROP is an important consideration even though it may be very rare. For laser, there is essentially no risk of endophthalmitis.
Cataract. Cataract is a serious potentially blinding condition when it occurs in infancy. Cataract can occur after treatment of ROP with either laser or after anti-VEGF injections. In the ETROP study of laser for ROP, cataract occurred in 1.2% of eyes. The mechanism for cataract development after laser is thought to be related to heating of lens protein. Secondary heating of the lens can occur when there is uptake of laser energy by the iris or by the vascularized tunica vasculosa lentis. The wavelength of laser is important for ROP treatment. Infrared 810 nm laser energy has minimal uptake by vascular structures so the tunica vasculosa lentis is not affected by this wavelength. In contrast, 532 nm green laser is efficiently taken up by vascular tissue such as the tunica vasculosa lentis. This difference likely explains why the reported incidence of cataract is higher with green laser compared to infrared laser.
Cataract can also occur with intravitreal injection for ROP. In this setting, cataract can result from needle induced trauma to the lens. In BEAT-ROP, there were no cataracts among the 70 eyes treated with bevacizumab injection and three cataracts out of 73 (4.1%) treated with laser. In the RAINBOW Trial, cataract was reported in one of 151 eyes (0.7%) injected with ranibizumab and 0 of 74 eyes treated with laser. In the PEDIG dosing study, one out of 112 eyes (0.9%) treated with injection developed cataract.
The potential for systemic complications is an important consideration when choosing a treatment for infants with ROP. In this section, we review the literature on potential systemic complications of laser ablation and anti-VEGF injections for ROP.
Mechanical Ventilation and Other Systemic Complications During or Shortly After Treatment. In ETROP, general anaesthesia with intubation was used for 36.6% of early treatment eyes and in 30.9% of conventionally managed eyes. In addition, patients in this study were reintubated within 10 days of treatment after stopping mechanical ventilation in a substantial proportion of cases (11.1% for early treatment and 5.1% for conventional treatment). Apnoea, bradycardia or arrhythmia were seen with laser in 8.6 and 4.2% of early-treated and conventionally treated patients, respectively. Likewise, acquired or increased cyanosis were seen in these groups in 3.6 and 1.7%, respectively. For anti-VEGF injections in infants, mechanical ventilation is not required. Worsening of apnoea, bradycardia and cyanosis do not seem to be issues during treatment because the injection procedure is so brief relative to laser. In prospective trials of anti-VEGF for ROP (BEAT-ROP, RAINBOW, PEDIG Dose De-escalation Trial), these endpoints are not reported or discussed.[3,5,32]
A recent retrospective study examining respiratory outcomes in ROP treatment compared status at baseline to several time points after treatment in 119 infants treated with laser and 19 infants treated with intravitreal bevacizumab (IVB). For laser, general anaesthesia with intubation in the operating room was used. For IVB, intravenous sedation with no additional respiratory support was used. Return to respiratory baseline was significantly less common in eyes treated with laser compared with those treated with IVB at 24 h (40 vs. 74%; P = 0.0115), 48 h (53 vs. 79%; P = 0.0453) and 7 days (79 vs. 100%; P = 0.0242). At 28 days, no difference was found between the groups (laser, 97%; IVB, 100%; P > 0.99).
Neurodevelopment. Whether neurodevelopmental outcomes are worse after treatment of ROP with IVB compared with laser is a controversial subject that has been the subject of numerous publications in recent years. VEGF is important for normal vascular development in the brain and other organs. Bevacizumab is measurable in systemic circulation in infants after IVB for up to 60 days.[52,53] After intravitreal anti-VEGF injection, serum VEGF levels are suppressed.[54,55] As such, there is sound rationale for concern about adverse effects of systemic VEGF suppression after intravitreal injection.
To understand the controversy surrounding reported neurodevelopmental impairment (NDI) with ROP treatment, one must understand the tendency for selection bias in comparing groups treated with IVB vs. laser. Concerns about safety have been present since IVB was first reported as treatment for ROP in 2007.[56,57] Most ophthalmologists have reserved IVB for particularly severe ROP or for infants who were too severely ill to undergo general anaesthesia for laser ablation. In addition, in 2011, BEAT-ROP reported better outcomes for zone I ROP treated with IVB compared with laser, and no difference in outcomes for zone II ROP. Nonrandomized studies comparing IVB to laser therefore have more patients with severe ROP, and more patients with severe comorbidities, in the IVB groups. More severe ROP is more likely to occur in smaller more preterm infants who are at higher baseline risk for NDI. Moreover, severe ROP, particularly in zone I, is an independent risk factor for NDI aside from other systemic factors.
The three studies reporting worse NDI after IVB treatment of ROP all suffer from selection bias.[59–61] The largest of these was reported by Natarajan et al. and includes data from a large US National Neonatal Research Network. It compared 181 infants receiving IVB and 224 infants undergoing laser or cryotherapy, all treated over a 4-year period after publication of the BEAT-ROP results. The study found no difference between groups in the primary outcome of severe NDI or death. However, when examined separately, the authors found a higher rate of death but not of NDI. Odds of a poor cognitive score were also higher with IVB. As the authors acknowledge, infants in the IVB group had lower median birth weight (P = 0.02), lower gestational age (P = 0.05), longer median time on ventilation (P = 0.04) and more supplemental oxygen (P = 0.01). None of these factors were considered for statistical adjustment and there was no adjustment for zone of ROP. In contrast, there are at least eight studies that report no association between IVB and developmental harm.[62–69] These studies are smaller than the study by Natarajan et al., but many have less inherent bias. A prospective case–control study reported by Fan et al. found no difference in NDI at 1.5 years comparing 38 patients with severe ROP treated with IVB to 31 untreated controls with similar baseline characteristics. As reported by the authors, this study is limited by its sample size, and may have lacked sufficient power to identify small but clinically significant differences. A Canadian study by Raghuram et al. compared 34 infants (60 eyes) receiving IVB and 30 infants (51 eyes) receiving laser and found no difference in NDI at 18–24 months. This study calculated adjusted odds ratios, controlling for gestational age, sex, ROP severity and other baseline comorbidities. A before-and-after study was recently published by Rodriguez et al. comparing infants treated with laser (before 2011) with those treated with IVB (after 2011), with no differences in the baseline characteristics between the two groups (n = 46 and 40, respectively). The authors found no difference in rates of death, cerebral palsy, sensorineural hearing loss or Bayley Scales of Infant Development scores. Finally, the single randomized controlled trial comparing IVB to laser, which followed 16 patients from the BEAT-ROP study, found no link between IVB and NDI.
To summarize the data on NDI after treatment of ROP, there are conflicting reports on whether IVB increases risk of poor outcome. Larger studies tend to be more biased with more severe ROP, and sicker infants, in the anti-VEGF groups. Smaller studies, with less inherent bias, are more limited by sample size but have shown no difference in outcomes. Authors on both sides of the controversy agree that rigorous prospective clinical trial data are required for an accurate appraisal of neurodevelopment after ROP treatment. However, given the profound impact of blindness on neurodevelopment, and as severe ROP treated with anti-VEGF has superior structural and refractive outcomes, commentators have suggested that existing evidence does not support withholding anti-VEGF treatment for severe ROP.[70,71]
Curr Opin Ophthalmol. 2021;32(5):475-481. © 2021 Lippincott Williams & Wilkins