Ranibizumab versus laser therapy for the treatment of very low birthweight infants with retinopathy of prematurity (RAINBOW): an open-label randomised controlled trial

Summary

Background

Despite increasing worldwide use of anti-vascular endothelial growth factor agents for treatment of retinopathy of prematurity (ROP), there are few data on their ocular efficacy, the appropriate drug and dose, the need for retreatment, and the possibility of long-term systemic effects. We evaluated the efficacy and safety of intravitreal ranibizumab compared with laser therapy in treatment of ROP.

Methods

This randomised, open-label, superiority multicentre, three-arm, parallel group trial was done in 87 neonatal and ophthalmic centres in 26 countries. We screened infants with birthweight less than 1500 g who met criteria for treatment for retinopathy, and randomised patients equally (1:1:1) to receive a single bilateral intravitreal dose of ranibizumab 0·2 mg or ranibizumab 0·1 mg, or laser therapy. Individuals were stratified by disease zone and geographical region using computer interactive response technology. The primary outcome was survival with no active retinopathy, no unfavourable structural outcomes, or need for a different treatment modality at or before 24 weeks (two-sided α=0·05 for superiority of ranibizumab 0·2 mg against laser therapy). Analysis was by intention to treat. This trial is registered with ClinicalTrials.gov, NCT02375971.

Interpretation

Between Dec 31, 2015, and June 29, 2017, 225 participants (ranibizumab 0·2 mg n=74, ranibizumab 0·1 mg n=77, laser therapy n=74) were randomly assigned. Seven were withdrawn before treatment (n=1, n=1, n=5, respectively) and 17 did not complete follow-up to 24 weeks, including four deaths in each group. 214 infants were assessed for the primary outcome (n=70, n=76, n=68, respectively). Treatment success occurred in 56 (80%) of 70 infants receiving ranibizumab 0·2 mg compared with 57 (75%) of 76 infants receiving ranibizumab 0·1 mg and 45 (66%) of 68 infants after laser therapy. Using a hierarchical testing strategy, compared with laser therapy the odds ratio (OR) of treatment success following ranibizumab 0·2 mg was 2·19 (95% Cl 0·99−4·82, p=0·051), and following ranibizumab 0·1 mg was 1·57 (95% Cl 0·76−3·26); for ranibizumab 0·2 mg compared with 0·1 mg the OR was 1·35 (95% Cl 0·61−2·98). One infant had an unfavourable structural outcome following ranibizumab 0·2 mg, compared with five following ranibizumab 0·1 mg and seven after laser therapy. Death, serious and non-serious systemic adverse events, and ocular adverse events were evenly distributed between the three groups.

Findings

In the treatment of ROP, ranibizumab 0·2 mg might be superior to laser therapy, with fewer unfavourable ocular outcomes than laser therapy and with an acceptable 24-week safety profile.

Introduction

Retinopathy of prematurity (ROP) is a disease of the developing retina. Mild forms of ROP resolve spontaneously with few sequelae, but severe ROP can progress to retinal detachment, vision impairment, and blindness. Ablation of the peripheral retina with cryotherapy¹ or confluent laser photocoagulation² reduces but does not eliminate ocular morbidity. Annually, 28 300−45 600 infants worldwide are diagnosed with irreversible visual impairment from ROP.³

Antivascular endothelial growth factor (anti-VEGF) agents injected into the vitreous are widely used for vasoproliferative disease and diseases associated with hyperpermeabilty in adult ophthalmic practice. They present a new opportunity for ROP treatment. To date, the largest clinical trial of anti-VEGF in preterm infants compared the efficacy of intravitreal bevacizumab with laser therapy in a defined US population.⁴ Fewer retreatments occurred in the bevacizumab group, but the difference was significant only for ROP located in retina zone I (the most posterior region of the retina).⁴ This study was preceded and followed by several case series,^5–7 small trials,^8–10 and two small dosing studies.^11,12 Despite the low level of evidence cited in a Cochrane review in 2018,¹³ use of anti-VEGF agents for treatment of ROP in clinical practice is increasing worldwide.^14,15

Several aspects of anti-VEGF therapy in ROP remain unanswered: ocular efficacy, the appropriate drug and dose, the need for retreatment, and the possibility of long-term systemic effects.^13–17 In light of these unanswered questions and the widespread off-label use of anti-VEGF treatment in ROP, assessment of anti-VEGF agents under defined trial conditions is urgently needed, not only in the USA and Europe but also in other countries where neonatal care can be more variable. The need is especially pressing in low-resource settings where an estimated 60% of the world’s ROP occurs,³ and anti-VEGF agents are being widely used as the primary treatment. This choice of treatment is not surprising for several reasons. Compared with laser therapy, intravitreal injection is relatively simple to administer, has substantially shorter administration time, requires fewer resources and less user expertise, and often no anaesthesia is given. However, intravitreal anti-VEGF treatment introduces a risk of intraocular infection,¹⁸ the need for longer and more frequent follow-up,¹⁹ retinal abnormalities after treatment,²⁰ and potential local and systemic developmental effects.¹⁷

The ranibizumab compared with laser therapy for the treatment of infants born prematurely with retinopathy of prematurity (RAINBOW) study was designed to evaluate the efficacy and safety of two doses of ranibizumab versus conventional laser therapy in a randomised open-label study that recruited infants worldwide, with a broad scope of ROP stages. Here, we report the outcomes of RAINBOW from enrolment until 24 weeks after treatment.

Methods

Study design and participants

RAINBOW is a randomised, open-label, superiority trial to assess the efficacy and safety of intravitreal ranibizumab compared with laser therapy for the treatment of ROP done in 87 neonatal and ophthalmic centres in 26 countries. Eligible infants had a birthweight less than 1500 g and a diagnosis of bilateral ROP zone I stage 1+, 2+, 3, or 3+, or zone II stage 3+, or aggressive posterior ROP (AP-ROP).²¹ Zone II stage 2+ was not included because treatment is controversial in some countries.²² Exclusion criteria included ocular and neurological comorbidities that might result in confounding visual impairment, and active ocular infection within 5 days before investigational treatment (appendix p 7). Unilateral cases in which only one eye met treatment criteria were not included.

The follow-up phase was until day 169 (24 weeks) when the infant exited the study and became eligible to join the RAINBOW extension study in which long-term ophthalmological and paediatric efficacy and safety will be studied to 5 years of age. The study was done in accordance with the Declaration of Helsinki. The study protocol was reviewed and approved by an independent ethics committee or institutional review board at each contributing centre. Parents or guardians provided written informed consent. An independent data and safety monitoring board oversaw the safety of study participants.

Randomisation and masking

Eligible infants were randomly assigned 1:1:1 to receive a single bilateral intravitreal dose of ranibizumab 0·2 mg, a single bilateral intravitreal dose of ranibizumab 0·1 mg, or laser therapy at baseline (day 1, treatment phase). Random allocation was done by local investigators by using web-based interactive response technology, stratified by ROP zone of the worst eye, because this characteristic might affect treatment outcome, and with blocks pre-allocated to fixed sections, combined with dynamic block allocation for the stratification of geographical region (at the level of country or site) to reflect variations in practice. Region 1 comprised countries with a neonatal mortality of less than five per 1000 livebirths in 2012, and region 2 countries with neonatal mortality rates five per 1000 or greater.²³ Neither treatment nor outcomes were masked.

Procedures

The allocated intravitreal treatment was a single dose of 0·2 mg or 0·1 mg ranibizumab applied to both eyes. Signs of disease persistence or recurrence were used by investigators to determine whether additional treatment was needed (appendix p 8). Laser treatment was administered per local practice (standard treatment). Under the protocol, in the laser therapy group supplementary treatment to skip lesions was allowed up to day 11. In the ranibizumab groups, up to two additional treatments with ranibizumab were allowed in each eye at a minimum of 28-day intervals. A switch to a different treatment method was considered as treatment failure. The retina was assessed at baseline and at each study visit up to day 169 (24 weeks). Examinations were done using indirect ophthalmoscopy and, where available, by wide-field retinal imaging. 75 (86%) of 87 study centres did retinal imaging. These images were transmitted to a central reading centre for documentation.

Patient data were captured via electronic case report forms by the investigation sites. Data validation, handling, and management were done on validated systems and in compliance with Good Clinical Practice and other applicable regulatory and national requirements.

Outcomes

The primary objective was to investigate whether intravitreal ranibizumab 0·2 mg had superior efficacy to laser therapy in the treatment of ROP, as defined by survival without active ROP, unfavourable structural outcomes, or the need for a treatment modality other than that assigned (treatment switch), in both eyes, up to 24 weeks after starting investigational treatment. Active ROP was defined as vessel dilatation (plus disease) in at least two quadrants, or extra-retinal vessels extending from the retina into the vitreous, representing potentially sight-threatening disease. Unfavourable structural outcomes included structural abnormalities that have potential effects on visual acuity: retrolental membrane obscuring the view of the posterior pole, substantial temporal retinal vessel dragging causing abnormal structural features or macular ectopia, posterior retinal fold involving the macula, or retinal detachment involving the macula.

The principal secondary objective was to investigate the efficacy of ranibizumab 0·1 mg relative to 0·2 mg or to laser therapy. Other secondary objectives were to assess time to intervention with a second modality for ROP or development of unfavourable outcome or death, recurrence of ROP requiring any post-baseline intervention up to 24 weeks, ocular and systemic safety of intravitreal ranibizumab, pharmacokinetics of intravitreal ranibizumab, and systemic VEGF levels.

Pharmacology

We adopted a sparse sampling approach to ranibizumab pharmacokinetics and VEGF pharmacodynamics, sampling within 24 h of baseline treatment, and around 14 (7–21) and 28 (22–28) days after treatment. The concentration of ranibizumab in serum samples was determined using enzyme-linked immunosorbent assay. The concentration of VEGF was determined in K3-EDTA plasma samples by an electrochemiluminescence sandwich immunoassay on samples taken before treatment and around day 14 and 28.

Statistical analysis

We assumed a failure event rate for ranibizumab 0·2 mg of 6·5% and laser therapy 42·4% in zone I, and 5·1% and 12·5% in zone II, respectively, which was similar to that observed in the BEAT-ROP trial⁴ for the recurrence of ROP at 54 weeks’ postmenstrual age, and that each patient had a 45% and 55% probability of ROP in zones I and II, respectively. At least 80 evaluable patients per treatment group (100 enrolled patients with an estimated dropout rate of 20%) were expected to provide more than 90% power to show superiority of ranibizumab 0·2 mg against laser therapy. Slow enrolment and recruitment challenges led the sponsor to change this target after consultation with the European Medicine Agency. 60 patients per treatment group (48 evaluable patients assuming a dropout rate of 20%) were expected to provide more than 80% power to show superiority of ranibizumab 0·2 mg against laser therapy. This minimum target replaced the original target but it was agreed that additional patients above this new target should be enrolled, until preplanned closure of enrolment on June 30, 2017, which was achieved.

The Cochran-Mantel Haenszel test was used to test the primary outcome because recruitment was stratified by zone of ROP. The prespecified level of significance was two-sided α=0·05 for superiority of ranibizumab 0·2 mg against laser therapy. A sequential test strategy was used to minimise the risk of type 1 errors caused by multiple comparisons. Significance testing of secondary outcomes was done only if the test of the primary comparison (ranibizumab 0·2 mg vs laser therapy) was significant. The Mantel-Haenszel odds ratio (OR) with 95% CI is presented for each comparison. Efficacy analysis was by intention to treat per patient. This trial is registered with ClinicalTrials.gov, NCT02375971.

Role of the funding source

The funder of the study had full access to data collection, analysis, and interpretation and was involved in the writing of the manuscript, as well as the decision to submit. The authors had full access to the data, which are owned by Novartis, and had final responsibility for the decision to submit for publication.

Results

Between Dec 31, 2015, and June 29, 2017, 87 centres in 26 countries enrolled infants (appendix p 10). 225 infants were enrolled into RAINBOW and were randomly allocated to one of the three study groups. Of these infants, 218 received baseline treatment (day 1), and 201 completed the study (figure 1).

Figure 1: CONSORT diagram detailing enrolment, randomisation, and follow-up of infants in the RAINBOW trial.

*Includes four deaths. †Includes four deaths and one treatment switch before withdrawal.

The median gestational age for participants at birth was 26 weeks (range 23−32). Gestational age was slightly lower in the ranibizumab 0·2 mg group than in the other two groups and lower in region 1 than in region 2. Most other baseline characteristics were well balanced between study groups (table 1).

Table 1:

Demographics and baseline characteristics

	Ranibizumab 0·2 mg (n=74)	Ranibizumab 0·1 mg (n=77)	Laser therapy (n=74)
Geographical region *
Region 1	45 (61%)	45 (58%)	44 (60%)
Region 2	29 (39%)	32 (42%)	30 (41%)
Sex
Male	33 (45%)	37 (48%)	37 (50%)
Female	41 (55%)	40 (52%)	37 (50%)
Mother’s race or ethnic origin
White	43 (58%)	45 (58%)	45 (61%)
Black	0	4 (5%)	3 (4%)
Asian	27 (37%)	22 (29%)	23 (31%)
Other	4 (5%)	6 (8%)	3 (4%)
Gestational age (weeks)
Overall	25 (23–32)	26 (23–32)	26 (23–32)
Region 1	24 (23–30)	25 (23–29)	24 (23–31)
Region 2	28 (24–32)	28 (24–32)	28 (24–32)
Gestational age category (weeks)
≤24	32 (43%)	22 (29%)	29 (39%)
>24 to <27	18 (24%)	21 (27%)	17 (23%)
≥27	24 (32%)	34 (44%)	28 (37%)
Birthweight (g)
Mean (SD)	791 (244)	886 (299)	831 (284)
Z score mean (SD)	−0·49 (1·18)	−0·52 (1·04)	−0·56 (1·14)
Plurality
Singleton	52 (70%)	55 (71%)	48 (65%)
Multiple	18 (24%)	18 (23%)	18 (24%)
Postmenstrual age at baseline treatment (weeks)
Age	36·7 (30·3–51·9)	36·9 (31·9–54·9)	36·6 (30·6–55·3)

Data are n (%) or median (range), unless otherwise indicated. Plurality data were missing for 4 (5%) patients who received ranibizumab, 4 (5%) patients who received ranibizumab, and 8 (11%) patients who received laser therapy.

^*Neonatal mortality in geographical region was <5 per 1000 livebirths in region 1 and >5 per 1000 livebirths in region 2.²³

Categorisation of ROP was similarly distributed across the treatment groups at baseline (appendix p 11). ROP was present in zone I in 86 (38%) infants and in zone II in 138 (61%) infants. Most infants had stage 3+ disease, 135 (60%) in zone II and 37 (16%) in zone I. AP-ROP was present in 30 (13%) of 225 infants, 29 of whom had zone I disease.

The primary efficacy outcome was known for 214 infants (figure 1). Treatment success occurred in 45 (66%) of 68 infants who received laser therapy (table 2) and in 56 (80%) of 70 infants who received ranibizumab 0·2 mg (OR compared with laser therapy 2·19, 95% Cl 0·99−4·82, p=0·051); hence, other significance testing was not undertaken. Treatment success occurred in 57 (75%) of 76 infants who received ranibizumab 0·1 mg (OR compared with laser therapy 1·57, 95% Cl 0·76−3·26; OR compared with ranibizumab 0·2 mg 1·35, 95% Cl 0·61−2·98).

Table 2:

Primary efficacy outcome and prespecified subgroup analysis

	Ranibizumab 0·2 mg	Ranibizumab 0·1 mg	Laser therapy
Patients entered	74	77	74
Patients with known primary outcome*	70	76	68
Treatment success†	56 (80%)	57 (75%)	45 (66%)
Reason for not meeting primary objective‡
Active ROP present§	0	3	0
Unfavourable structural outcome¶	1	5	7
Treatment switch||	11	13	18
Death of infant	4	4	4
Prespecified subgroup analyses
Primary outcome by ROP zone
Zone 1	19/28 (68%)	21/30 (70%)	14/23 (61%)
Zone 2	37/42 (88%)	36/46 (78%)	31/45 (70%)
Primary outcome by gestation
≤24 weeks	22/29 (76%)	17/22 (77%)	12/27 (44%)
>24 to <27 weeks	16/18 (89%)	18/21 (86%)	9/15 (60%)
≥27 weeks	18/23 (78%)	22/33 (67%)	24/26 (92%)
Primary outcome by region**
Region 1	36/42 (86%)	37/45 (82%)	26/42 (62%)
Region 2	20/28 (71%)	20/31 (64%)	19/26 (73%)
Primary outcome by infant sex
Male	28/33 (85%)	26/36 (72%)	19/35 (54%)
Female	28/37 (76%)	31/40 (78%)	26/33 (79%)

Data are n (%) or n/N (%), unless otherwise indicated. ROP=retinopathy of prematurity.

^*See figure 1.

^†Alive and without treatment switch and unfavourable structural outcome or active ROP at day 169.

^‡Infants can have multiple events.

^§Three infants with active ROP at day 169 had vessel dilatation (plus disease) but no extraretinal vessels.

^¶Unfavourable structural outcome in either eye, any one of: retrolental membrane obscuring the view of the posterior pole; substantial temporal retinal vessel dragging causing abnormal structural features or macular ectopia; posterior retinal fold involving the macula; and retinal detachment involving the macula.

^||Treatment switch is the intervention for ROP with a treatment modality other than the modality of the first investigational treatment (this includes additional standard of care treatment such as vitrectomy).

^**Neonatal mortality was <5 per 1000 livebirths in region 1 and >5 per 1000 livebirths in region 2.²³

Three infants in the ranibizumab 0·1 mg group had active ROP at week 24, comprising plus disease in two or more quadrants but with no extra-retinal vessels extending from the retina. No patients in the other groups had active disease. One infant had an unfavourable structural outcome following ranibizumab 0·2 mg, compared with five following ranibizumab 0·1 mg and seven after laser therapy (appendix p 12).

Because of minor baseline differences in key variables, a single confirmatory post-hoc exploratory logistic regression analysis for the primary outcome adjusting for three potential confounders—gestational age, geographical region, and infant sex—was done; the OR in favour of ranibizumab 0·2 mg versus laser treatment was 2·32 (95% CI 1·04−5·16).

Overall, treatment success was higher for all three groups in zone II than in zone I, and the gradient of effect remained in favour of ranibizumab (table 2). Among the two ranibizumab groups, full peripheral vascularisation assessed by indirect ophthalmoscopy occurred by day 169 in 28 (38%) infants in the ranibizumab 0·2 mg group and 21 (27%) infants in the ranibizumab 0·1 mg group.

61 infants received additional post-baseline treatments: 23 (31%) in the ranibizumab 0·2 mg group, 24 (31%) in the ranibizumab 0·1 mg group, and 14 (19%) in the laser therapy group (figure 2). In the laser group this was in addition to the treatment of skip lesions in 11 infants before day 11 but including one infant who received laser therapy on day 12. Including treatment to skip lesions, 22 (30%) infants in the laser group needed additional post-baseline treatment (laser therapy or switch to ranibizumab). 12 infants in each of the ranibizumab groups had one or more ranibizumab retreatments and achieved a successful primary outcome. The median interval between baseline and first retreatment was 55 days (range 29−111) in the ranibizumab 0·2 mg group and 57 days (30−128) in the ranibizumab 0·1 mg group.

Figure 2: Progress of infants with retinopathy of prematurity through additional post-baseline study treatments.

Flow chart shows the number of infants receiving at least one additional treatment (ranibizumab or laser therapy), either allowed in the protocol (up to 29 days when ranibizumab retreatment was first allowed) or up to 169 days, and treatment switches, which contributed to the primary outcome (shaded pink). Infants might have received one or more additional treatments in one or both eyes. Laser treatments to skip lesions were allowed as part of the baseline treatment but are included because they represent extra treatment episodes. *Allowed in protocol.

Death, serious adverse events (AEs), and non-serious systemic AEs were evenly distributed among the three groups (appendix pp 12−14). All events were from complications of preterm birth. Four deaths occurred in each group, including one infant who was treated with ranibizumab 0·1 mg in whom the investigator suspected that death was related to the study drug or procedure. The infant died at home following a feed within 24 h of treatment and was classified as fatal respiratory failure (appendix p 12). Serious and non-serious ocular AEs were similarly distributed between the three groups (appendix pp 14−15).

One infant (ranibizumab 0·2 mg) had a moderate cataract in one eye noted on day 28, thought to be lens damage by the injecting needle. Cataract surgery was not done during the study.

One infant (ranibizumab 0·1 mg) developed unilateral endophthalmitis and orbital infection 6 days after intravitreal injection. This infant was described as having bilateral conjunctivitis and left periocular staphylococcal infection 11 days before ranibizumab injection and treated with antibiotics. Infection was deemed resolved by the investigator before study entry, and prophylactic topical antibiotic was given to the left eye for 3 days immediately before intravitreal injection. Ipsilateral conjunctivitis and orbital infection presented 4 days after injection and endophthalmitis was evident at day 6. Vitrectomy was done at 8 weeks. The retinal structural outcome was unfavourable.

There were no significant differences between the study groups in body length, head circumference, knee to heel length, weight, or blood pressure from baseline to days 85 or 169 (appendix p 16).

Median serum ranibizumab and plasma VEGF levels within 24 h of, and 14 days and 28 days after, intravitreal injection are shown in the appendix p 16. Serum ranibizumab levels decayed from peak levels immediately after baseline to much lower levels after 28 days. Plasma VEGF levels were variable and there was no clear evidence of plasma VEGF suppression or of differences between the three treatment groups (figure 3). Detailed population pharmacokinetics of ranibizumab and related VEGF pharmacodynamics will be reported elsewhere.

Figure 3: Vascular endothelial growth factor levels following study entry to 35 days for each treatment group.

(A) Ranibizumab 0·2 mg. (B) Ranibizumab 0·1 mg. (C) Laser therapy.

Discussion

In RAINBOW, ranibizumab was an effective and well tolerated treatment for ROP in a range of settings around the world. The proportion of infants with successful outcome was greater following intravitreal ranibizumab 0·2 mg compared with laser therapy and does not exclude superiority of ranibizumab over laser therapy.²⁴ In particular, ranibizumab 0·2 mg resulted in fewer eyes with unfavourable structural outcomes than did ranibizumab 0·1 mg or laser therapy. A ranibizumab dose of 0·1 mg offered no advantage over the 0·2 mg dose. Further safety data and functional outcomes to 5 years of age will be ascertained during the ongoing extension study. The observed systemic safety profile was as expected in a preterm population, and ocular adverse events were consistent with the established profile for ranibizumab in adults.

RAINBOW included ROP in all zone II locations, including mid and anterior zone II. The efficacy of ranibizumab was similar to laser in both zone I and zone II disease. In all three treatment groups, the highest risk of unfavourable outcome occurred in zone I disease, as has been reported in other trials.^1,4,25 Switching to alternative forms of treatment was frequent in all three groups. This might have been partly caused by the study design because the rules for additional treatments were set at a low threshold to minimise the risk of visual impairment and applied to both laser therapy and ranibizumab groups.

Our trial had several limitations. Care was provided by clinicians from a wide range of settings and experience to mimic routine clinical practice. Instructions for the administration of ranibizumab were provided but not training in the use of fundoscopy to determine the primary outcome, and not all centres had access to retinal photography. The trial was open label, not masked, and used no placebo. Decisions on retreatment were made on an individual basis and retreatment with ranibizumab restricted to intervals of 28 days. Clinician preference for one treatment could lead to biased decisions to re-treat. The study was limited by slow enrolment, which led to a modified recruitment target and reduced power. It is unclear why the significance of the primary comparison was marginal. Retrospectively, it is difficult to ascribe this result either to treatment effects being similar or to an inadequate sample size.

Although RAINBOW and the BEAT-ROP trial⁴ both evaluated anti-VEGF agents, direct comparisons should be approached with caution as the two trials have very different protocols. RAINBOW included all forms of ETROP type 1 ROP² apart from zone II stage 2+ from an international group of hospitals. By contrast, the BEAT-ROP trial included only stage 3+ disease in zone I and posterior zone II in a single region. In RAINBOW, outcomes were evaluated per patient 24 weeks after first treatment (at a mean age of 60 weeks’ [SD 4·4] postmenstrual age), whereas in BEAT-ROP ascertainment was by each eye to 54 weeks’ postmenstrual age (and occurred between 50 to 70 weeks’ postmenstrual age). RAINBOW used a composite primary outcome that included survival without active ROP, unfavourable structural outcomes, or treatment switch. By contrast BEAT-ROP used recurrence of retinal neovascularisation requiring retreatment. Therefore, the two studies are not directly comparable.

In the ETROP randomised trial,²⁵ 9·1% of eyes had unfavourable structural outcomes after laser treatment, which compares to 1·43% of infants in RAINBOW following ranibizumab 0·2 mg and 10% of infants following laser therapy. The efficacy of ranibizumab with fewer unfavourable structural outcomes compared with laser therapy needs to be weighed against the two serious ocular complications—cataract and endophthalmitis—that each occurred in one eye in this study. To minimise risk, any form of periocular infection just before a planned intravitreal injection, even if treated, should be regarded as an exclusion criterion.

In RAINBOW, additional treatments were administered throughout the follow-up period. Regular follow-up eye examinations are therefore required after ranibizumab treatment. Median time to recurrence requiring additional treatment was 8 weeks in this study, indicating that this time frame is particularly important during follow-up examinations, but much later recurrences after anti-VEGF therapy can also occur.²⁶

Ranibizumab is a fully humanised monoclonal antigen-binding 48kDa Fab antibody fragment.²⁷ In adults, the serum half-life of ranibizumab is approximately 9 days after intravitreal injection, probably due to slow egress from the vitreous.²⁸ RAINBOW is the first study to measure ranibizumab pharmacokinetics in preterm infants. Serum ranibizumab levels fell slowly with detectable but much reduced levels at day 29. By contrast, bevacizumab, a 149 kDa monoclonal antibody, has a serum half-life of 21 days after intravitreal injection in preterm infants.²⁹

VEGF-mediated angiogenesis is essential for the development of a number of tissues,^30,31 but little is known about the effects of VEGF suppression in preterm infants. We measured plasma rather than serum VEGF levels because this approach avoids VEGF release from platelets during thrombolysis and better reflects true free VEGF.³² There was no clear evidence of plasma VEGF suppression compared with infants in the laser therapy group or of differences over time between the three treatment groups in RAINBOW, confirming findings in CARE-ROP.¹¹ By contrast, following intravitreal bevacizumab in preterm infants serum VEGF levels were substantially reduced for many weeks compared with baseline or those treated with laser therapy.^29,33

In summary, ranibizumab 0·2 mg was as effective and safe in the treatment of active ROP as laser therapy, might be superior, and was associated with better short-term ocular outcomes. Compared with laser treatment, the potential for procedural complications and the need for regular clinical follow-up after ranibizumab treatment must be balanced against fewer adverse structural outcomes. On Sept 4, 2019, ranibizumab received approval for retinopathy of prematurity in the EU from the European Commission (European Medicines Agency).

Research in context.

Evidence before this study

Retinopathy of prematurity (ROP) is the primary cause of blindness in 28 300–45 600 premature infants per year worldwide. Over the past 25 years, treatment with laser therapy has been able to halt the disease and prevent blindness. However, this does not eliminate ocular morbidity, which remains of concern. Uncontrolled retinal neovascularisation, which drives the clinical progression of the disease, is provoked by local secretion of vascular endothelial growth factor (VEGF) in ischaemic retinal tissue. Anti-VEGF agents could therefore be used to target uncontrolled retinal neovascularisation in ROP.

A Cochrane review published in 2018 identified five trials of bevacizumab and one trial of pegaptanib, grading the quality of the evidence as low or very low over eight outcomes followed for up to 55 weeks and mortality at 30 months of age, whether alone or in combination with laser and cryotherapy. A further search for English language publications between Jan 1, 2011, and March 31, 2018, was done using PubMed, MEDLINE, and Web of Science using the search terms “anti-VEGF”, “individual agents”, and “retinopathy of prematurity”. Two small trials of ranibizumab were identified comprising 50 infants (in progress) and 19 infants (published).

Without research evidence of either long-term benefit or risk, and with few attempts at finding the appropriate dose, off-label anti-VEGF use has become widespread and is the first and sometimes only treatment option in many settings across the world. Several aspects of the use of anti-VEGF agents remain unanswered: ocular efficacy, the appropriate drug and dose, the need for retreatment, and the possibility of long-term systemic effects from sustained suppression of VEGF in a developing infant.

Added value of this study

RAINBOW shows that ranibizumab, an established anti-VEGF agent in adult practice, might be superior to laser therapy in infants with ROP. The use of ranibizumab seems to be associated with better eye outcomes and it has an acceptable short-term safety profile. Follow-up is underway to determine vision outcomes to 5 years of age. Systemic VEGF levels appear to be unchanged after ranibizumab. RAINBOW investigated two ranibizumab doses to collect data on anti-VEGF dose finding in ROP.

Implications of all the available evidence

Intravitreal injection of ranibizumab can be considered as a novel treatment for ROP. However, disease recurrence must be monitored closely because it is common with anti-VEGF agents and can occur later compared with laser therapy. No detectable effects on systemic VEGF levels seem advantageous for ranibizumab over other anti-VEGF agents. It remains too early to determine definitive vision outcomes or to confirm long-term systemic safety. Both will be addressed in the ongoing RAINBOW extension trial.

Data sharing

Study results have been disclosed as required by corresponding regulations and are publicly available on ClinicalTrials.gov, the EU clinical trial registry, and the Novartis clinical trial results websites.