Anti‐vascular endothelial growth factor for neovascular age‐related macular degeneration

Summary of findings for the main comparison. Summary of findings: bevacizumab versus ranibizumab

Bevacizumab compared with ranibizumab for neovascular age‐related macular degeneration
Participant or population: people with neovascular age‐related macular degeneration Settings: clinical centers Intervention: intravitreal injections of bevacizumab Comparison: intravitreal injections of ranibizumab
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No. of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Ranibizumab	Bevacizumab
Gain of 15 letters or more visual acuity at one year	257 per 1000	231 per 1000 (188 to 285)	RR 0.90 (0.73 to 1.11)	2446 (6)	⊕⊕⊕⊕ high
Loss of fewer than 15 letters visual acuity at one year	942 per 1000	942 per 1000 (923 to 960)	RR 1.00 (0.98 to 1.02)	2446 (6)	⊕⊕⊕⊕ high
Mean change in visual acuity at one year (number of letters)	The mean change across ranibizumab groups ranged from gains of 3 to 8 letters	The mean change in visual acuity in the bevacizumab groups was on average 0.51 fewer letters gained (95% CI 1.64 fewer letters to 0.62 more letters)	MD ‐0.51 (‐1.64 to 0.62)	2446 (6)	⊕⊕⊕⊕ high
Reduction in central retinal thickness at one year	The mean reduction in central retinal thickness across ranibizumab groups ranged from 30 to 182 μm	The mean reduction in central retinal thickness in the bevacizumab groups was on average 13.97 μm less (95% CI 26.52 less to 1.41 less)	MD ‐13.97 (‐26.52 to ‐1.41)	1995 (4)	⊕⊕⊕⊕ high	Two additional trials reported no difference between groups for this outcome; however, these data were not reported in formats that could be included in meta‐analysis
No problems in quality of life domains at one year	Range of 591 per 1000 to 861 per 1000 across five quality of life domains	Range of 608 per 1000 to 828 per 1000 across five quality of life domains	Range of RRs 0.96 (0.90 to 1.04) to 1.02 (0.89 to 1.17)	548 (1)	⊕⊕⊕⊝ moderate¹	Quality of life domains included: mobility, self care, usual, activities, pain/discomfort, anxiety/depression
Serious systemic adverse events at one year²	139 per 1000 with at least one serious systemic adverse event	177 per 1000 (148 to 212)	RR 1.27 (1.06 to 1.52)	2597 (4)	⊕⊕⊕⊝ moderate¹
Serious ocular adverse events at one year	< 5 per 1000	< 5 per 1000	Range of RRs 0.51 (0.05 to 5.62) to 7.05 (0.36 to 136.28)	Range 1670 to 2280 (2 to 3)	⊕⊕⊕⊝ moderate¹	Studies reported different ocular adverse events
The basis for the assumed risk* is estimated by the proportion with the event in the ranibizumab group. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; MD: mean difference
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group grades of evidence: High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. ¹ Quality of life and adverse event outcomes downgraded to moderate quality as not all eligible trials reported these outcomes and numbers of some adverse events were small (< 1%). ² A new Cochrane review on systemic safety of bevacizumab versus ranibizumab (Moja 2014) includes more up‐to‐date data for this finding, from trials listed as ongoing in this review. Please refer to Moja 2014 for the most up‐to‐date information on systemic safety for bevacizumab versus ranibizumab.

Summary of findings 2. Summary of findings: pegaptanib versus control

Pegaptanib compared with control for neovascular age‐related macular degeneration
Participant or population: people with neovascular age‐related macular degeneration Settings: clinical centers Intervention: intravitreal injections of pegaptanib Comparison: sham injections
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No. of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Pegaptanib
Gain of 15 letters or more visual acuity at one year	20 per 1000	57 per 1000 (25 to 132)	RR 2.83 (1.23 to 6.52)	1186 (1)	⊕⊕⊕⊕ high
Loss of fewer than 15 letters visual acuity at one year	554 per 1000	687 per 1000 (615 to 770)	RR 1.24 (1.11 to 1.39)	1186 (1)	⊕⊕⊕⊕ high
Mean change in visual acuity at one year (number of letters)	The mean change in the control group was a loss of 15 letters	The mean change in visual acuity in the pegaptanib groups was on average 6.72 more letters gained (95% CI 4.43 letters to 9.01 letters)	MD 6.72 (4.43 to 9.01)	1186 (1)	⊕⊕⊕⊕ high
Reduction in central retinal thickness at one year	‐	‐	‐	‐	‐	Outcome not assessed by this trial.
No problems in quality of life domains at one year	‐	‐	‐	‐	‐	Treatment with pegaptanib was associated with better scores on the NEI‐VFQ questionnaire, specifically for distance vision and role limitation domains; however, standard deviations for scores were not reported.
Serious systemic adverse events at one year	151 per 1000 with at least one serious systemic adverse event	189 per 1000 (140 to 257)	RR 1.25 (0.93 to 1.70)	1190 (1)	⊕⊕⊕⊝ moderate¹
Serious ocular adverse events at one year	7 per 1000 with any eye disorder	26 per 1000 (6 to 109)	RR 3.84 (0.91 to 16.20)	1190 (1)	⊕⊕⊕⊝ moderate¹
The basis for the assumed risk* is estimated by the proportion with the event in the control group. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; MD: mean difference
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. ¹ Adverse events downgraded to moderate quality as the numbers of events were small (< 1%) for many specific adverse events.

Summary of findings 3. Summary of findings: ranibizumab versus control

Ranibizumab compared with control for neovascular age‐related macular degeneration
Participant or population: people with neovascular age‐related macular degeneration Settings: clinical centers Intervention: intravitreal injections of ranibizumab Comparison: sham injections with or without verteporfin photodynamic therapy
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No. of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Ranibizumab
Gain of 15 letters or more visual acuity at one year	59 per 1000	230 per 1000 (93 to 566)	see comment	1322 (3)	⊕⊕⊕⊝ moderate¹	Meta‐analysis not performed due to high I² (80%).
Loss of fewer than 15 letters visual acuity at one year	610 per 1000	934 per 1000 (861 to 1000)	RR 1.53 (1.41 to 1.64)	1322 (3)	⊕⊕⊕⊕ high
Mean change in visual acuity at one year (number of letters)	The mean change across control groups ranged from a loss of 10 to 16 letters	The mean change in visual acuity in the ranibizumab groups was on average 17.80 more letters gained (95% CI 15.95 letters to 19.65 letters)	MD 17.80 (15.95 to 19.65)	1322 (3)	⊕⊕⊕⊕ high
Reduction in central retinal thickness at one year	‐	‐	‐	‐	‐	We were unable to find data on central retinal thickness in reports from any of the three included trials comparing ranibizumab with control interventions.
Mean change in vision‐related quality of life	The mean change across control groups in vision‐related quality of life scores ranged from ‐3 to 2 points	The mean change across control groups in vision‐related quality of life scores ranged from 5 to 7 points	MD 6.69 (3.38 to 9.99)	1134 (2)	⊕⊕⊕⊝ moderate²	Using the NEI‐VFQ questionnaire with a 10‐point difference considered as being clinically meaningful.
Serious systemic adverse events at one year	Range of 5 to 83 per 1000 for various systemic adverse events	Range of 0 to 55 per 1000 for various systemic adverse events	Range of RRs 0.17 (0.01 to 4.24) to 2.08 (0.23 to 18.45)	603 (2)	⊕⊕⊕⊝ moderate³
Serious ocular adverse events at one year	Range of 0 to 68 per 1000 for various ocular adverse events	Range of 3 to 118 per 1000 for various ocular adverse events	Range of RRs 0.52 (0.03 to 8.25) to 2.71 (1.36 to 5.42)	603 (2)	⊕⊕⊕⊝ moderate³
The basis for the assumed risk* is estimated by the proportion with the event in the control group. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; MD: mean difference
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. ¹ Gain of vision outcome downgraded due to high statistical heterogeneity. ² Quality of life outcomes downgraded due to not all studies reporting this outcome and non‐clinically significant results. ³ Adverse events downgraded to moderate quality as not all eligible trials reported all types of adverse events and numbers were small (<1%) for many specific adverse events.

Summary of findings 4. Summary of findings: bevacizumab versus control

Bevacizumab compared with control for neovascular age‐related macular degeneration
Participant or population: people with neovascular age‐related macular degeneration Settings: clinical centers Intervention: intravitreal injections of bevacizumab Comparison: standard therapy (intravitreal injections of pegaptanib, verteporfin photodynamic therapy with or without triamcinolone acetonide, or sham injections)
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No. of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Bevacizumab
Gain of 15 letters or more visual acuity at one year	38 per 1000	293 per 1000 (92 to 937)	RR 7.80 (2.44 to 24.98)	159 (2)	⊕⊕⊕⊝ moderate¹
Loss of fewer than 15 letters visual acuity at one year	700 per 1000	896 per 1000 (763 to 1000)	RR 1.28 (1.09 to 1.50)	159 (2)	⊕⊕⊕⊝ moderate¹
Mean change in visual acuity at one year (number of letters)	‐	‐	‐	‐	‐	The mean change from baseline in visual acuity was 7.0 letters in the bevacizumab group and ‐9.4 letters in the control group in one study. The second study reported participants in the bevacizumab group gained 8 letters on average and participants in the control group lost 3 letters on average.
Reduction in central retinal thickness at one year	‐	‐	‐	‐	‐	The mean change was ‐91 μm in the bevacizumab group and ‐55 μm in the control group in one study and ‐113 μm in the bevacizumab group and ‐72 μm in the control group in the other study.
Mean change in vision‐related quality of life	‐	‐	‐	‐	‐	Outcome not assessed by these trials.
Serious systemic adverse events at one year	15 per 1000 experienced any systemic adverse event	31 per 1000 (3 to 331)	RR 2.03 (0.19 to 21.85)	131 (1)	⊕⊕⊝⊝ low²
Serious ocular adverse events at one year	91 per 1000 experienced any ocular adverse event	169 per 1000 (66 to 431)	RR 1.86 (0.73 to 4.74)	131 (1)	⊕⊕⊝⊝ low²
The basis for the assumed risk* is estimated by the proportion with the event in the control group. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; MD: mean difference
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. ¹ Vision outcomes downgraded to moderate quality due to small sample sizes. ² Adverse events downgraded to low quality as sample sizes were small and numbers of events were small (< 1%) for many specific adverse events.

Background

Description of the condition

Introduction

Age‐related macular degeneration (AMD) is a progressive, degenerative disease of the retina that occurs with increasing frequency with age. There are two major types of AMD, commonly referred to as non‐neovascular ('dry') and neovascular ('wet') AMD. The non‐neovascular type is characterized by drusen (yellow spots under the retina), pigmentary changes (re‐distribution of melanin within the retinal pigment epithelium (RPE) under the retina and migration of melanin into the retina), and geographic atrophy (loss of the RPE and choriocapillaris).

This review concerns neovascular AMD and its treatment. The hallmark of neovascular AMD is choroidal neovascularization (CNV). Breaks in the RPE and Bruch’s membrane allow naturally occurring vessels in the choroid to grow aberrantly into the subretinal space. These choroidal neovascular vessels typically leak and bleed, causing exudative or hemorrhagic retinal detachments. Without treatment, the process usually evolves into a fibrous scar, which replaces the outer layers of the retina, the RPE, and the choriocapillaris. The scarred retina has greatly diminished visual capacity.

Epidemiology

AMD is a leading cause of irreversible vision loss in the elderly in developed countries (Bourne 2014; Bunce 2006; Congdon 2004; Ghafour 1983; Hyman 1987; Leibowitz 1980; Tielsch 1994). While the non‐neovascular type is much more common, the neovascular form of AMD is responsible for most cases of severe vision loss. The incidence of progression from non‐neovascular AMD to neovascular AMD is increased by the presence of numerous, large and confluent drusen in the macula as well as by the presence of pigment in the macula. Neovascular AMD occurs in only 10% of people with AMD, yet 80% of those with severe visual loss (worse than 20/200 Snellen acuity) have the neovascular form (Leibowitz 1980). Once neovascular disease develops in one eye, the risk of developing neovascular disease in the other eye in the same person is approximately 40% by 5 years (AREDS 2001; SST 20).

The overall prevalence of AMD, in a meta‐analysis of studies from Australia, Europe, and the United States, has been estimated at 1.47% (95% confidence interval (CI) 1.38% to 1.55%) (Friedman 2004); however, AMD increases in prevalence with age, with a low incidence among individuals aged less than 50 years. Thus, the burden of disease is greatest in regions where life expectancy is highest. Among those aged 80 years or older, the prevalence of neovascular AMD is estimated to be 5.79% (95% CI 4.72 to 7.01%) in the UK (Owen 2003) and 8.18% (95% CI 7.07% to 9.29%) in the United States (Friedman 2004).

There is no consistent evidence that modifiable factors such as lipid levels, blood pressure, light exposure, or alcohol intake put people at greater risk of developing AMD. One notable exception is smoking (Klein 2008; Mitchell 2002; Smith 1996). Elevated baseline levels of inflammatory biomarkers such as C‐reactive protein have been found to be associated with the development of early and late AMD in a large population‐based cohort (Boekhoorn 2007). Furthermore, several studies have shown gene‐environment interactions of complement factor H with smoking and C‐reactive protein (Deangelis 2007; Haddad 2006; Schaumberg 2007; Seddon 2006). High doses of vitamins C and E, beta‐carotene, and zinc provide a modest protective effect against the progression to advanced AMD in individuals with extensive drusen or in initially unaffected fellow eyes with neovascular AMD (AREDS 2001; AREDS2 2013).

As the population continues to age, a higher prevalence of this disease is expected in the future, at least in certain populations. A population‐based survey estimated AMD, as a contributing cause of blindness, increased worldwide from 4.4% (95% CI 4.0 to 5.1) in 1990 to 6.6% (95% CI 5.9 to 7.9) in 2010 (Bourne 2014).

Presentation and diagnosis

Neovascular AMD may affect one eye or both eyes at the same time or sequentially. The symptoms of neovascular AMD are metamorphopsia (distortions while looking at objects), scotomata (black or gray spots), and blurry vision. Depending upon the location of CNV and the vision in the fellow eye, individuals with AMD may be unaware of the change in visual acuity or may note difficulty with performing normal activities that require good central vision, such as reading and writing, watching television, driving and recognizing faces. When AMD affects only one eye, visual loss may go undetected until monocular testing at a routine eye examination or chance occlusion of the better eye. Frequently, people are unaware that their disturbed binocular vision is caused by changes in only one eye.

A diagnosis of neovascular AMD is made clinically and with the help of imaging such as optical coherence tomography (OCT) and fluorescein angiography, which may be necessary to detect subtle exudation in some individuals who have experienced a recent change in visual acuity. At the onset of symptoms, fundus examination often reveals subretinal exudation of fluid, lipid, or blood. OCT, a non‐invasive imaging modality, shows cross‐sectional images of the retina, RPE, and choroid. Some studies have defined the characteristic appearance of the different stages of the disease process on OCT (Ting 2002; Van Kerckhoven 2001). The most characteristic findings on OCT corresponding to a CNV lesion include areas of hyporeflectivity under the retina that, in turn, correspond to subretinal fluid, cystic hyporeflective changes consistent with macular edema, and attenuation of the photoreceptor/chorio‐capillaris layer. CNV has several characteristic patterns on fluorescein angiography. Classic CNV is defined as an area of early hyperfluorescence with increasing fluorescein leakage on late frames of the angiogram (MPSG 1991). Occult CNV occurs in two different patterns: fibrovascular pigment epithelial detachment and late leakage from an undetermined source. Classic CNV typically has well‐demarcated borders, whereas occult CNV usually has poorly demarcated borders.

Another test, indocyanine green (ICG) angiography, may aid in evaluating individuals with neovascular AMD, as it images the choroidal circulation better than fluorescein angiography and may show 'hot' spots under the RPE that are amenable to treatment. ICG angiography is particularly useful in the diagnosis of polypoidal choroidal vasculopathy, a form of AMD most common in Asian populations.

Description of the intervention

Until the advent of anti‐VEGF agents, treatments most frequently used for neovascular AMD included thermal laser photocoagulation and verteporfin photodynamic therapy (PDT). A Cochrane systematic review concluded that laser photocoagulation effectively slowed the progression of neovascularization in non‐subfoveal lesions compared with observation alone (Virgili 2007). A Cochrane review of verteporfin PDT concluded that PDT was effective in preventing clinically significant vision loss (Wormald 2007). However, neither laser photocoagulation or PDT offered any significant chance for vision improvement.

Over the past two decades, researchers have developed new drugs for the treatment of neovascular AMD. These drugs target a protein in the body known as vascular endothelial growth factor (VEGF) that stimulates the growth of the abnormal blood vessels in neovascular AMD in a process called angiogenesis; the drugs block VEGF leading to regression of the abnormal blood vessels. Antiangiogenic therapy is the most commonly used treatment for neovascular AMD, particularly subfoveal neovascular lesions.

An example of an anti‐VEGF antagonist is pegaptanib (Macugen^®, a trademark of Eyetech/Pfizer, Inc.). Pegaptanib is a chemically synthesized 28‐base ribonucleic acid molecule. It is an aptamer (foldable single‐strand nucleic acid) and has a capability to change its three‐dimensional structure to fit a target protein, in this case VEGF. By binding to VEGF, pegaptanib blocks and inactivates VEGF, thus, halting the process of neovascularization. Pegaptanib was approved for the treatment of neovascular AMD by the Food and Drug Administration (FDA) in the United States in December 2004.

Ranibizumab, previously known as rhuFab‐VEGF (Lucentis^®, a trademark of Genentech, Inc.), is another example of an anti‐VEGF medication developed for ocular administration. It is a humanized antibody fragment capable of binding to the VEGF protein to prevent it from binding to its receptor, thus inhibiting angiogenic activity. Ranibizumab was the first treatment for neovascular AMD that offered a realistic hope for vision improvement; it was approved by the FDA in 2007.

Bevacizumab is another anti‐VEGF agent that is used to treat CNV secondary to neovascular AMD. Bevacizumab (Avastin^®, a trademark of Genentech, Inc.) is a humanized monoclonal antibody against VEGF. It is the larger parent molecule from which ranibizumab was derived. Bevacizumab is currently approved for the treatment of conditions such as colorectal cancer, but it is widely used by ophthalmologists as an off‐label drug for neovascular AMD.

Aflibercept, previously known as VEGF Trap (Eylea^®, a trademark of Regeneron Pharmaceuticals, Inc.), is another anti‐VEGF agent; the molecule serves as a VEGF decoy to inhibit the growth of new blood vessels. Aflibercept was approved for the treatment of neovascular AMD by the FDA in 2011. Because its mechanism of action is slightly different than those of the drugs listed above (pegaptanib, ranibizumab, and bevacizumab) and it was introduced after the protocol for this review was written, we have not evaluated aflibercept in this review.

How the intervention might work

Angiogenesis is a complex process whereby interactions between stimulatory and inhibitory factors result in new blood vessel formation. These factors have been identified in CNV formation in animal models and human tissue (Aiello 1994; Kvanta 1996; Lopez 1996). Antiangiogenic therapies work either by blocking stimulatory factors or by promoting inhibitory factors, thus disrupting the formation of new vessels. Agents that block the activity of VEGF (anti‐VEGFs), a polypeptide with mitogenic effects on endothelial blood vessels, form one type of antiangiogenic therapy. VEGF antagonists have been shown to inhibit CNV in animal models.

In the past, the primary goal of both laser photocoagulation and PDT was to prevent or delay further loss of visual acuity in the treated eye. With the development of agents to counteract VEGF, together known as anti‐VEGF agents, the primary goal of the intravitreal injection of these agents is to retain or improve visual acuity. Currently, anti‐VEGF agents are administered most commonly via monthly intravitreal injections or as needed after three consecutive monthly injections.

Why it is important to do this review

The previous version of this Cochrane review documented the effectiveness of anti‐VEGF agents in halting the loss of visual acuity in a substantial fraction of treated eyes (Vedula 2008). Further, intravitreal injections with ranibizumab led to improvements in vision not previously observed with other AMD treatments. Since this Cochrane review was first published in 2008, numerous studies have been conducted to evaluate the safety and effectiveness of various anti‐VEGF agents, treatment modalities, and combination therapies for the treatment of neovascular AMD (Table 1). This review is restricted to: (1) primary RCTs of anti‐VEGF agents versus no anti‐VEGF treatment; and (2) head‐to‐head (comparative effectiveness) RCTs of one anti‐VEGF agent versus another. Studies of dosage, different treatment strategies, and the combination of anti‐VEGF agents with other treatments are outside the scope of this review. The emphasis of this updated review is the stabilization of or improvement in visual acuity with treatment.

Table 1. Table of Study Acronyms

Acronym	Details
*Included studies*
ABC	Avastin^® (Bevacizumab) in Choroidal Neovascularization Trial
ANCHOR	Anti‐VEGF Antibody for the Treatment of Predominantly Classic Choroidal Neovascularization in Age‐related Macular Degeneration
CATT	Comparison of Age‐related macular degeneration Treatment Trials
GEFAL	French Evaluation Group Avastin® Versus Lucentis^®
IVAN	A randomized controlled trial of alternative treatments to Inhibit VEGF in Age‐related choroidal Neovascularisation
MANTA	A Randomized Observer and Subject Masked Trial Comparing the Visual Outcome After Treatment With Ranibizumab or Bevacizumab in Patients With Neovascular Age‐related Macular Degeneration Multicenter Anti VEGF Trial in Austria
MARINA	Minimally Classic/Occult Trial of the Anti‐VEGF Antibody Ranibizumab in the Treatment of Neovascular Age‐Related Macular Degeneration
PIER	A Phase IIIb, Multicenter, Randomized, Double‐Masked, Sham Injection‐Controlled Study of the Efficacy and Safety of Ranibizumab in Subjects with Subfoveal Choroidal Neovascularization with or without Classic CNV Secondary to Age‐Related Macular Degeneration
VISION	VEGF Inhibition Study in Ocular Neovascularization
*Ongoing studies*
BRAMD	Comparison of Bevacizumab (Avastin^®) and Ranibizumab (Lucentis®) in Exudative Age‐related Macular Degeneration
LUCAS	Lucentis® Compared to Avastin^® Study
MAAM	Avastin® and Macugen® Versus Avastin^® Versus Macugen^®
RATE	Ranibizumab and the Risk of Arterial Thromboembolic Events
VIBERA	Prevention of Vision Loss in Patients With Age‐Related Macular Degeneration by Intravitreal Injection of Bevacizumab and Ranibizumab
Other studies evaluating anti‐VEGF therapies for AMD*
ADVANCE	Safety and Efficacy of Oral PTK787 in Patients With Subfoveal Choroidal Neovascularization Secondary to Age‐Related Macular Degeneration (NCT00138632)
ARMAST	Photodynamic Therapy Combined With Bevacizumab vs Bevacizumab Alone for Neovascular Age‐Related Macular Degeneration (NCT00696592)
ATLAS	Repeated Eye Injections of Aflibercept for Treatment of Wet Age Related Macular Degeneration (NCT01773954)
BEAT‐AMD	Systemic Avastin Therapy in Age‐Related Macular Degeneration (NCT00531024)
BeMOC	Randomised controlled trial of bevacizumab in choroidal neovascularisation secondary to age related macular degeneration (ISRCTN12980412)
CARBON	Safety & Efficacy Study Evaluating the Combination of Bevasiranib & Lucentis Therapy in Wet AMD (NCT00557791)
CLOVER	Combination Lucentis and Ocular Photodynamic Therapy With Visudyne, With Evaluation‐based Retreatment (NCT00680498)
COBALT	Safety & Efficacy Study Evaluating the Combination of Bevasiranib & Lucentis Therapy in Wet AMD (NCT00499590)
DENALI	Efficacy/Safety of Verteporfin Photodynamic Therapy and Ranibizumab Compared With Ranibizumab in Patients With Subfoveal Choroidal Neovascularization (NCT00436553)
EVEREST	Efficacy and Safety of Verteporfin Added to Ranibizumab in the Treatment of Symptomatic Macular Polypoidal Choroidal Vasculopathy (NCT00674323)
EXCITE	Efficacy and Safety of Ranibizumab in Patients With Subfoveal Choroidal Neovascularization Secondary to Age‐related Macular Degeneration (NCT00275821)
EXTEND‐I/II/III	Efficacy and Safety of Ranibizumab in Patients With Subfoveal Choroidal Neovascularization Secondary to Age‐related Macular Degeneration (NCT00826371; NCT00470678)
FOCUS	RhuFab V2 Ocular Treatment Combining the Use of Visudyne^® to Evaluate Safety (NCT00056823)
GMAN	Greater Manchester Avastin® for choroidal Neovascularisation trial (ISRCTN34221234)
HARBOR	A Study of Ranibizumab Administered Monthly or on an As‐needed Basis in Patients With Subfoveal Neovascular Age‐related Macular Degeneration (NCT00891735)
HORIZON	An Open‐Label Extension Trial of Ranibizumab for Choroidal Neovascularization Secondary to Age‐Related Macular Degeneration (cohort of participants who completed the MARINA, ANCHOR, or FOCUS trials)
LAST	A Pilot Study to evaLuate the Role of High‐dose rAnbizumab (2.0mg) in the Management of AMD in Patients With perSistent/recurrenT Macular Fluid Less Than 30 Days Following Treatment With Intravitreal Anti‐VEGF Therapy (NCT01115556)
LOW‐VISION	Intravitreal Bevacizumab for Low Vision in Neovascular Age‐related Macular Degeneration (NCT01327222)
LUV	Lucentis Utilizing Visudyne Combination Therapy in the Treatment of Age‐Related Macular Degeneration (NCT00423189)
MERLOT	Macular EpiRetinal Brachytherapy Versus Lucentis® Only Treatment (NCT01006538)
MONET	Phase II Open Label Multicenter Study For Age Related Macular Degeneration Comparing PF‐04523655 Versus Lucentis In The Treatment Of Subjects With CNV (NCT00713518)
MONT BLANC	Verteporfin Photodynamic Therapy Administered in Conjunction With Ranibizumab in Patients With Subfoveal Choroidal Neovascularization Secondary to Age‐related Macular Degeneration (NCT00433017)
NEXUS	Efficacy and Safety Study of iSONEP With and Without Lucentis/Avastin to Treat Age‐related Macular Degeneration (NCT01414153)
PERSPECTIVES	An Open Label Trial to Investigate Macugen for the Preservation of Visual Function in Subjects With Neovascular AMD (NCT00327470)
PrONTO	Prospective Optical coherence tomography imaging of patients with Neovascular AMD Treated with intra‐Ocular ranibizumab (NCT00344227)
RADICAL	Reduced Fluence Visudyne‐Anti‐VEGF‐Dexamethasone In Combination for AMD Lesions (NCT00492284)
SAILOR	Safety Assessment of Intravitreal Lucentis® for Age‐Related Macular Degeneration (NCT00251459)
SALUTE	Comparison of Safety, Effectiveness and Quality of Life Outcomes Between Labeled Versus "Treat and Extend" Regimen in Turkish Patients With Choroidal Neovascularisation Due to AMD (NCT01148511)
SUMMIT	Unclear (clinical trial program including the DENALI, EVEREST, and MONT BLANC trials)
SUSTAIN	Study of Ranibizumab in Patients with Subfoveal Choroidal Neovascularization Secondary to Age‐Related Macular Degeneration (NCT00331864)
VERITAS	A Safety and Efficacy Study Comparing the Combination Treatments of Verteporfin Therapy Plus One of Two Different Doses of Intravitreal Triamcinolone Acetonide and the Verteporfin Therapy Plus Intravitreal Pegaptanib (NCT00242580)
VIEW‐1/2	Vascular Endothelial Growth Factor (VEGF) Trap‐Eye: Investigation of Efficacy and Safety in Wet Age‐Related Macular Degeneration (NCT00509795; NCT00637377)
WALTZ	Wet Age‐Related Macular Degeneration AL‐39324 Treatment Examination (NCT00992563)

*List of studies that may or may not be listed as excluded studies. Clinical trial identifiers are shown in parentheses.

Objectives

Methods

Criteria for considering studies for this review

Types of studies

We included RCTs only.

Types of participants

We included trials in which the participants had neovascular AMD as defined by study investigators.

Types of interventions

We included studies in which anti‐VEGF treatment was compared with another treatment, sham treatment, or no treatment. We did not include studies in which different doses of one anti‐VEGF treatment were compared with each other, with no control or comparator group. We did not include studies of aflibercept (VEGF Trap‐Eye/EYLEA^® solution) or studies that used anti‐VEGF agents in combination with other treatments.

Types of outcome measures

Primary outcomes

The primary outcome for this review was best‐corrected visual acuity (BCVA) at one year of follow up. As all the included RCTs randomized only one eye per participant (i.e., the study eye), we defined the primary outcome for the comparison of treatments as the proportion of participants who gained 15 letters or more (3 lines) of BCVA in the study eye when BCVA was measured on a visual acuity chart with a LogMAR scale.

Secondary outcomes

(1) Visual acuity outcomes

(a) Proportion of participants who gained 15 letters or more of BCVA in the study eye at two years of follow up
(b) Proportion of participants who lost fewer than 15 letters of visual acuity
(c) Proportion of participants who lost fewer than 30 letters of visual acuity
(d) Proportion of participants in whom blindness was prevented in the study eye, defined as those eyes with visual acuity better than 20/200
(e) Proportion of participants maintaining visual acuity, defined as gain of 0 letters or more (i.e., no loss of BCVA from baseline)
(f) Mean change in visual acuity

In addition to visual acuity outcomes, the following secondary outcomes were considered
(2) Contrast sensitivity, reading speed, or any other validated measure of visual function as measured in the included studies
(3) Assessment of morphological characteristics by fluorescein angiography or OCT, including mean change in size of CNV, mean change in size of total lesion, and mean change in central retinal thickness (CRT)
(4) Quality‐of‐life measures, as assessed with any validated measurement scale
(5) Economic data, such as comparative cost analyses
(6) Ocular or systemic adverse outcomes

Follow up

We included trials in which participants were followed for at least one year. We also included outcomes at two years of follow up when data were available.

Search methods for identification of studies

Electronic searches

We searched CENTRAL (which contains the Cochrane Eyes and Vision Group Trials Register) (2014, Issue 3), Ovid MEDLINE, Ovid MEDLINE In‐Process and Other Non‐Indexed Citations, Ovid MEDLINE Daily, Ovid OLDMEDLINE (January 1946 to March 2014), EMBASE (January 1980 to March 2014), Latin American and Caribbean Health Sciences Literature Database (LILACS) (January 1982 to March 2014), the metaRegister of Controlled Trials (mRCT) (www.controlled‐trials.com), ClinicalTrials.gov (www.clinicaltrials.gov) and the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) (www.who.int/ictrp/search/en). We did not use any date or language restrictions in the electronic searches for trials. We last searched the electronic databases on 27 March 2014.

See: Appendices for details of search strategies for CENTRAL (Appendix 1), MEDLINE (Appendix 2), EMBASE (Appendix 3), LILACS (Appendix 4), mRCT (Appendix 5), ClinicalTrials.gov (Appendix 6) and the ICTRP (Appendix 7).

Searching other resources

We reviewed the reference lists of included trial reports and related systematic reviews to identify additional potentially relevant trials.

We contacted pharmaceutical companies conducting studies on anti‐VEGF drugs for information about any ongoing or completed clinical trials not published. One author (SSV) handsearched abstracts from the annual meetings of the Association for Research in Vision & Ophthalmology (ARVO) for the years 2006 and 2007 for ongoing trials (http://files.abstractsonline.com/SUPT/163/1807/PresentationTitle.htm; http://files.abstractsonline.com/SUPT/163/1601/Presentation_Title_PDF_wlinks.htm accessed November 24, 2007). After 2007, the Cochrane Eyes and Vision Group began handsearching conference abstracts reporting clinical trials and the identified trial records are listed in CENTRAL. Another author (KL) handsearched abstracts from the 2006 annual meeting of the European VitreoRetinal Society (http://www.evrs.eu/2006‐evrs‐congress‐cannes/ accessed November 27, 2012). For future updates of this review, we will consider handsearching abstracts for the following conferences when they have not been searched by the Cochrane Eyes and Vision Group: ARVO; Macula Society; Retina Society; subspecialty meetings from the American Academy of Ophthalmology meeting; American Society of Retinal Surgeons; and European VitreoRetinal Society.

Data collection and analysis

Selection of studies

Two review authors independently evaluated the titles and abstracts resulting from the electronic searches. We classified each record as 'definitely relevant', 'possibly relevant', or 'definitely not relevant'; a third review author resolved discrepancies. We obtained full‐text reports for all records assessed as 'definitely relevant' or 'possibly relevant'. Two review authors independently assessed the full‐text reports and classified each study as 'include', 'exclude', 'awaiting classification', or 'ongoing'; a third review author resolved discrepancies. For trials identified by handsearching conference abstracts, a second author verified eligibility based on the stated criteria. We contacted authors to clarify any details necessary to make a complete assessment of the relevance of the study. We documented studies excluded after review of the full‐text report and noted the reasons for exclusion.

Data extraction and management

Two review authors independently extracted study characteristics, including details of study methods, participants, interventions, outcomes, and funding resources, using data collection forms developed specifically for this purpose. We contacted the trial authors for data on primary and secondary outcomes in the individual trials when the information was not clearly available from published reports. We extracted data on visual acuity, adverse events, and other outcomes for the two trials forming part of the VISION 2004 study from documents available on the FDA website. We also extracted data from figures published in the trial reports and communicated with the authors to verify extracted data. One author entered data into Review Manager (RevMan 2012), and a second author verified the data entry.

Assessment of risk of bias in included studies

Two review authors assessed potential sources of bias in trials according to methods set out in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). The following parameters were considered: random sequence generation and method of allocation concealment (selection bias), masking of participants and researchers (performance bias), masking of outcome assessors (detection bias), rates of losses to follow up and non‐compliance as well as failure to include analysis of all participants after randomization (attrition bias), reporting bias, and other potential sources of bias. We judged each potential source of bias as low risk, unclear risk, or high risk. We contacted authors of trials for additional information when descriptions of study methods needed to assess bias domains were unclear or not reported.

Measures of treatment effect

Data analysis was guided by Chapter 9 of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2011). The primary outcome and some secondary outcomes for this review related to BCVA in the study eye. We analyzed visual acuity, measured on LogMAR charts, as both dichotomous and continuous outcomes. We calculated the risk ratios (RRs) with 95% confidence intervals (CIs) for dichotomous outcomes. Dichotomous visual acuity outcomes included: proportion of participants who gained 15 letters or more (same as a gain of 3 lines or more) of visual acuity; proportion of participants who lost fewer than 15 letters (same as fewer than 3 lines) of visual acuity; proportion of participants who lost fewer than 30 letters (same as fewer than 6 lines) of visual acuity; proportion of participants not blind (defined as visual acuity better than 20/200); and proportion of participants maintaining visual acuity (same as gain of 0 letters or more). We calculated the mean difference (MD) in mean change of visual acuity from baseline as a continuous visual acuity outcome.

Secondary outcomes relating to visual function and morphology of CNV also included both dichotomous and continuous outcomes. We calculated RRs with 95% CIs for dichotomous outcomes and MDs with 95% CIs for continuous outcomes. Contrast sensitivity outcomes, measured by Pelli‐Robson charts, were reported both dichotomously (proportion of participants with a gain of 15 letters or more of contrast sensitivity) and continuously (mean number of letters of contrast sensitivity). We calculated MDs with 95% CIs for near visual acuity and reading speed outcomes when sufficient data were available.

Continuous morphological outcomes included mean change in size of CNV, mean change in size of lesion, and mean change in CRT. We included one dichotomous morphological outcome, which was the resolution of subretinal or intraretinal fluid based on OCT evaluation.

We analyzed quality‐of‐life scores as continuous outcomes. Because the trials that reported quality‐of‐life outcomes included in meta‐analyses used the same scale, we did not need to calculate standardized mean differences.

We reported adverse events as RRs with 95% CIs when sufficient data were available. Otherwise we reported the numbers of participants experiencing adverse events in narrative and tabular form.

Unit of analysis issues

The unit of analysis was the individual (one study eye per participant).

Dealing with missing data

We used multiple sources to identify relevant data for this review, such as journal publications, conference abstracts, FDA documents, and clinical trial registries. When data were unclear (e.g., data were extracted from graphs or derived from percentages), we contacted study investigators for verification. When data were missing, we contacted study investigators for additional information. If no response was received within two weeks, we attempted to contact them again. Whenever no response was received by six weeks after the first attempt, we used the data as available.

For outcome data, we used the data as reported in the trial reports or as supplied by the primary investigators. We noted the number of participants with missing data and the statistical methods used in the individual studies to analyze data (e.g., available case analysis, last‐observation‐carried‐forward, etc.). We did not impute missing outcome data for our analyses.

Assessment of heterogeneity

We assessed statistical heterogeneity based on the Chi² test, I² statistic and the overlap of CIs in the forest plots. We considered a Chi² P value of < 0.10 to represent significant statistical heterogeneity and an I² statistic of 60% or more to represent substantial statistical heterogeneity. We assessed clinical and methodological heterogeneity among studies by comparing the study populations, interventions, and methods of each study.

Assessment of reporting biases

We assessed selective outcome reporting for each study by comparing the outcomes specified in a protocol, research plan, or clinical trial registry with the results reported. When protocols, research plans, or clinical trial registry records were not available, we assessed selective outcome reporting based on the outcomes specified in the methods section of the study report and the data that were collected as described in the study design. In further updates of this review, whenever 10 or more studies are included in a meta‐analysis, we will use a funnel plot to judge publication bias.

Data synthesis

Statistical analyses were performed using RevMan 2012. We did not combine studies in meta‐analysis when we identified clinical or methodological heterogeneity (e.g., different anti‐VEGF agents or outcome time points); instead we either analyzed data by type of anti‐VEGF agent and time point or, when data were not sufficient for meta‐analysis, we reported a narrative summary. We used a random‐effects model for all analyses. When the I² statistic was 60% or greater, suggesting substantial statistical heterogeneity, we assessed the direction of treatment effects across studies and the overlap of the CIs to determine whether meta‐analysis was appropriate. We did not adjust estimates of treatment effects to account for comparisons of different doses of an anti‐VEGF agent to a single control group, as observed in several studies.

Subgroup analysis and investigation of heterogeneity

In the previously published version of this review we conducted subgroup analyses of the primary outcome, as specified in the protocol, by stratifying the data according to the angiographic subtype of CNV using the definitions adopted in the included trials (Vedula 2008). Because we changed the primary outcome to a gain of 15 letters or more of visual acuity for this version of the review, we did not conduct these subgroup analyses as data were insufficient. If data by angiographic subtype of CNV are available for inclusion in future updates to this review, we will include these subgroup analyses.

Sensitivity analysis

In an earlier published version of this review we conducted sensitivity analyses to examine potential bias caused by missing data from participants excluded after randomization or lost to follow up in analyses for the primary outcome. We did this by analyzing the primary outcome assuming that: 1) participants lost to follow up had lost 15 letters or more of visual acuity (worst‐case analysis); and 2) participants did not lose 15 letters or more of visual acuity at one year follow up (best‐case analysis) (Vedula 2008). Because these analyses did not alter the conclusions of the review, we did not conduct these sensitivity analyses for this version of the review and do not plan to conduct them in future updates.

We planned to conduct sensitivity analyses to assess the impact of studies graded as having a high risk of bias on any parameter, unpublished data only, or industry funding. After assessing the data collected, we determined these analyses were not needed because studies within each meta‐analysis did not differ based on these factors.

Results

Description of studies

Results of the search

The electronic searches for the first published version of this review (conducted in August 2005, October 2006, June 2007 and February 2008) resulted in the identification of a total of 1407 titles and abstracts (Vedula 2008). We selected 36 records for full‐text review, and identified five trials described in 10 reports for inclusion in the review (ANCHOR 2006; EOP 1003; EOP 1004; FOCUS 2006; MARINA 2006). We excluded 16 studies (24 reports) and listed two additional studies identified through the handsearching of abstracts as awaiting classification. Acronyms used to refer to the studies in this review are listed in Table 1.

Two concurrent, randomized trials that used individual participant data meta‐analyses under the acronym VISION were identified (Gragoudas 2004), an international trial (EOP 1003) and a North American trial (EOP 1004). In the first published version of this review, we assessed the data from these two trials separately and analyzed them according to the original protocol of the review. We obtained the data for the primary and secondary outcomes for the two trials from the information available on the FDA website and by contacting the authors. For this update we considered the two trials as one study, VISION 2004, and collected new data from published articles as available. The characteristics of the two individual trials are summarized in Appendix 8 and Appendix 9.

For this update, we also refined the eligibility criteria to exclude studies in which anti‐VEGF treatment was given in combination with other AMD treatments and to include trials in which two anti‐VEGF agents had been compared (i.e., head‐to‐head trials). Combination therapies for AMD will be covered in a separate Cochrane review. Thus, we did not include the FOCUS 2006 trial, which compared ranibizumab with PDT versus PDT alone and was included in the first version of this review, in this update of the review.

We conducted updated electronic searches in September 2008, April 2011, February 2013, and March 2014. Because we modified the eligibility criteria and new authors joined the review team, we combined all search results and assessed the records as a new review. In all, there were 4827 unique records from electronic searches of bibliographic databases, 403 clinical trial registrations, and 19 additional records identified by the handsearching of conference abstracts (Figure 1). From the bibliographic databases, we identified 153 records for full‐text review. Of these 153 records, we included 12 RCTs (reported in 108 records) and excluded 39 studies (reported in 45 records). We excluded two additional studies from three records identified by handsearching. We list the reasons for exclusion of each of the 41 studies in the 'Characteristics of excluded studies' table. We included the remaining 16 records identified by handsearching as additional reports of the included studies. We identified seven additional studies from the search of clinical trial registries, one of which is awaiting classification due to insufficient information to determine eligibility and six are ongoing (or completed with results not yet published). Descriptions of studies awaiting classification and those that are ongoing are available in the Characteristics of studies awaiting classification section and the Characteristics of ongoing studies section, respectively.

Figure 1

Study flow diagram. Results of searches as of 27 March 2014.

Included studies

Types of participants

This review included a total of 5496 participants from 12 RCTs; the number of participants per trial ranged from 28 to 1208. In all 12 trials, one eye per participant was randomized. The countries in which the trials were conducted spanned the globe: two studies were international (ANCHOR 2006; VISION 2004), four were conducted in the United States only (CATT 2011; MARINA 2006; PIER 2008; Subramanian 2010), two each in Austria (MANTA 2013; Sacu 2009) and the United Kingdom (ABC 2010; IVAN 2013), and one each in France (GEFAL 2013) and India (Biswas 2011). The 12 trials were similar in that they all enrolled both men and women 50 years of age or older who had subfoveal CNV secondary to AMD. Among the included trials, there were variations in the types of eligible neovascular lesions (e.g., predominantly classic CNV, minimally classic CNV, or occult CNV), lesion sizes, and baseline visual acuities of participants. Although the majority of participants in most trials were women, all but one of the enrollees in one trial were men (Subramanian 2010).

All trials predefined visual acuity eligibility criteria for the study eye of each participant. The most common criterion was a BCVA of 20/40 to 20/320 (Snellen equivalent) in the study eye, which was specified in six studies (ABC 2010; ANCHOR 2006; MANTA 2013; MARINA 2006; PIER 2008; VISION 2004). BCVA eligibility ranges included participants with somewhat better visual acuity in the CATT 2011 (20/25 to 20/320), GEFAL 2013 (20/32 to 20/320), and IVAN 2013 (20/320 or better) trials, but potentially worse visual acuity in the Sacu 2009 (20/40 to 20/800) and Subramanian 2010 (20/400 or better) studies. In Biswas 2011, participants with a BCVA between 35 and 70 Early Treatment Diabetic Retinopathy Study (ETDRS) letters were eligible; however, the test distance was not reported.

Five trials included only participants with no previous treatment for CNV or AMD (Biswas 2011; CATT 2011; IVAN 2013; MANTA 2013; Sacu 2009). The remaining seven trials allowed participants to have received previous therapy for AMD, with certain restrictions as to the type (e.g., verteporfin PDT, intravitreal injections, or surgery), location, and time interval since last treatment. Five trials enrolled participants with either primary or recurrent CNV in the study eye (ANCHOR 2006; MARINA 2006; PIER 2008; Subramanian 2010; VISION 2004) and one enrolled participants with primary CNV only (ABC 2010).

Of the six studies that reported the type of neovascular lesion, the ANCHOR 2006 study had the highest proportion of participants with predominantly classic CNV (410/423, 97%). The other five studies had fewer participants with predominantly classic CNV. In the ABC 2010 study, 25% of 131 participants had predominantly classic CNV; the remaining 75% had either minimally classic or occult CNV. In the VISION 2004 study, 26% of 1208 participants had predominantly classic CNV, 36% had minimally classic CNV, and 38% had occult CNV. The PIER 2008 study had similar proportions as the VISION 2004 study, with 19% of 184 participants having predominantly classic CNV, 38% having minimally classic CNV, and 43% having occult CNV at baseline. Forty‐four percent of 120 participants had occult CNV in Biswas 2011.The MARINA 2006 study was limited to participants with only minimally classic or occult CNV and, thus, had the greatest proportion of participants with occult CNV (451/716, 63%).

Two studies that did not report neovascular lesion type described the subfoveal component of the CNV lesion in the study population. In the CATT 2011 study (1208 participants), 58% had CNV in the foveal center, 27% had fluid in the foveal center, 8% had hemorrhage in the foveal center, and 6% had other foveal center involvement. The distribution was similar in the IVAN 2013 study (628 participants) in which 54% participants had CNV in the foveal center, 29% had hemorrhage in the foveal center, and 13% had other foveal center involvement. The two smallest studies (Sacu 2009; Subramanian 2010), with 28 participants in each, and the GEFAL 2013 (501 participants) and MANTA 2013 (321 participants) studies did not describe the type of neovascularization or subfoveal component of the CNV lesion in the study population.

Five trials specified size of the lesion as an inclusion criterion. Four trials (ABC 2010; GEFAL 2013; MARINA 2006; PIER 2008) included participants with lesions of 12 disc areas (DA) or smaller (1 DA = 2.54 mm², i.e., standard DA) and one study (Sacu 2009) set 4 DAs as the maximum lesion size.

Additional details about each trial included in this review are summarized in the Characteristics of included studies table.

Types of interventions

Comparisons of interventions evaluated in the trials included in this review are listed in Table 2 and are summarized here. Among the 12 trials, there were four comparisons of interventions: one study evaluated three doses of pegaptanib versus sham injection (VISION 2004), three studies compared two doses of ranibizumab with sham injections or PDT (ANCHOR 2006; MARINA 2006; PIER 2008), two studies compared bevacizumab with other treatments for AMD (ABC 2010; Sacu 2009), and six studies were head‐to‐head trials of bevacizumab versus ranibizumab (Biswas 2011; CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013; Subramanian 2010).

See: Summary of findings for the main comparison Summary of findings: bevacizumab versus ranibizumab; Summary of findings 2 Summary of findings: pegaptanib versus control; Summary of findings 3 Summary of findings: ranibizumab versus control; Summary of findings 4 Summary of findings: bevacizumab versus control

Table 2. Treatment groups in included trials

Study Treatment period	Intervention 1	Intervention 2	Intervention 3	Intervention 4
Pegaptanib versus control
VISION 2004 2 years; re‐randomized at end of first year	0.3 mg pegaptanib every 6 weeks	1.0 mg pegaptanib every 6 weeks	3.0 mg pegaptanib every 6 weeks	Sham every 6 weeks
Ranibizumab versus control
ANCHOR 2006 2 years	0.3 mg ranibizumab monthly plus sham verteporfin PDT	0.5 mg ranibizumab monthly plus sham verteporfin PDT	Sham intravitreal injection plus verteporfin PDT	‐
MARINA 2006 2 years	0.3 mg ranibizumab monthly	0.5 mg ranibizumab monthly	Sham intravitreal injection monthly	‐
PIER 2008 2 years	0.3 mg ranibizumab monthly for 3 months, then every 3 months	0.5 mg ranibizumab monthly for 3 months, then every 3 months	Sham intravitreal injection monthly for 3 months, then every 3 months	‐
Bevacizumab versus control
ABC 2010 1 year	1.25 mg bevacizumab given first three injections every 6 weeks, then as needed	Standard therapy (0.3 mg pegaptanib every six weeks, verteporfin PDT, or sham injection)	‐	‐
Sacu 2009 1 year	1.0 mg bevacizumab monthly for 3 months, then as needed	Verteporfin PDT plus same day 4 mg triamcinolone acetonide	‐	‐
Bevacizumab versus ranibizumab
CATT 2011 2 years; re‐randomized at end of first year	1.25 mg bevacizumab monthly for 1 year; at 1 year, re‐randomization to ranibizumab monthly or variable dosing	0.5 mg ranibizumab monthly for 1 year; at 1 year, re‐randomization to ranibizumab monthly or variable dosing	1.25 mg bevacizumab as needed after first injection for 2 years	0.5 mg ranibizumab as needed after first injection for 2 years
IVAN 2013 2 years; ongoing	1.25 mg bevacizumab monthly for 2 years	0.5 mg ranibizumab monthly for 2 years	1.25 mg bevacizumab monthly for 3 months, then as needed in 3 month cycles	0.5 mg ranibizumab monthly for 3 months, then as needed in 3 month cycles
Biswas 2011 18 months	1.25 mg bevacizumab monthly for 3 months, then as needed	0.5 mg ranibizumab monthly for 3 months, then as needed	‐	‐
GEFAL 2013 1 year	1.25 mg bevacizumab; maximum of one injection per month	0.5 mg ranibizumab; maximum of one injection per month	‐	‐
MANTA 2013 1 year	1.25 mg bevacizumab monthly for 3 months, then as needed	0.5 mg ranibizumab monthly for 3 months, then as needed	‐	‐
Subramanian 2010 1 year	0.05 ml bevacizumab monthly for 3 months, then as needed	0.05 ml ranibizumab monthly for 3 months, then as needed	‐	‐

PDT: photodynamic therapy

The VISION 2004 investigators compared sham injections with intravitreal injection of pegaptanib at dosages of 0.3 mg, 1.0 mg, or 3.0 mg given every 6 weeks over a 48‐week period.

Two different doses of ranibizumab (0.3 mg and 0.5 mg) were evaluated in three trials (ANCHOR 2006; MARINA 2006; PIER 2008). The control groups and the dosing schedule for ranibizumab differed among the three trials. Monthly intravitreal injection of ranibizumab (for 12 months) was compared with sham intravitreal injections in MARINA 2006. For participants assigned to receive sham intravitreal injections in MARINA 2006, verteporfin PDT was allowed whenever the CNV lesion in their eyes became predominantly classic CNV. Monthly injection of ranibizumab combined with sham PDT (for 24 months) was compared with verteporfin PDT and sham intravitreal ranibizumab injections in ANCHOR 2006. A regimen of monthly injection of ranibizumab for three months followed by an injection every three months was compared with sham intravitreal injections in PIER 2008.

Bevacizumab was evaluated in eight trials. In the ABC 2010 trial, a 1.25 mg dose of bevacizumab was compared with standard therapy. Standard therapy was determined by clinical evaluation and included 0.3 mg pegaptanib, verteporfin PDT, or sham injection. In the small Sacu 2009 trial, a 1 mg dose of bevacizumab was compared with verteporfin PDT combined with intravitreal triamcinolone. In six trials, bevacizumab was compared for non‐inferiority with ranibizumab. In both CATT 2011 and IVAN 2013, in addition to the primary comparison of the two agents, monthly injections of the anti‐VEGF agents were compared with an 'as needed' regimen following three initial injections of the assigned agent. The latter treatment regimen, using a 0.5 mg dose of ranibizumab and a 1.25 mg dose of bevacizumab, was used to compare the two anti‐VEGF agents in the Biswas 2011, GEFAL 2013, MANTA 2013, and Subramanian 2010 trials.

Types of outcome measures

Visual acuity

BCVA was the basis of the primary outcome for all the included studies. The primary outcome for this review, the proportion of participants who gained 15 letters or more of BCVA at one year of follow up, was the primary outcome for one study (ABC 2010) and a secondary outcome for the remaining 11 studies. The proportion of participants losing fewer than 15 letters at one year was the primary outcome for the three earliest studies (ANCHOR 2006; MARINA 2006; VISION 2004) and a secondary outcome for the remaining nine studies. The primary outcome was mean change in visual acuity at one year for six studies (CATT 2011; GEFAL 2013; MANTA 2013; PIER 2008; Sacu 2009; Subramanian 2010), and the mean change in visual acuity at 18 months for one study (Biswas 2011). The remaining five studies reported mean change in visual acuity as a secondary outcome. The primary outcome for one study (IVAN 2013) was best‐corrected distance visual acuity at two years of follow up; BCVA (as opposed to mean change from baseline) was not an outcome considered in this review.

Other visual acuity outcomes relevant to this review were also reported by some of the included studies. Loss of fewer than 30 letters of visual acuity was reported in five studies (ABC 2010; ANCHOR 2006; MARINA 2006; Subramanian 2010; VISION 2004); BCVA better than 20/200 was reported for eight studies (ANCHOR 2006; CATT 2011; GEFAL 2013; IVAN 2013; MARINA 2006; PIER 2008; Subramanian 2010; VISION 2004); and maintenance of visual acuity (defined as a gain of 0 letters or more) was reported in four studies (ANCHOR 2006; Sacu 2009; Subramanian 2010; VISION 2004). Investigators of included studies reported a number of other visual acuity outcomes that we did not consider in this review.

In all studies, visual acuity was measured using the ETDRS chart, which has a LogMAR scale. Each line on the ETDRS chart consists of 5 letters; thus, a change of 15 letters approximates a 3‐line change in visual acuity. The outcome for visual acuity of 20/200 or better was reported as the Snellen equivalent.

Visual function

Visual function outcomes were assessed in five studies. In the ABC 2010 trial, contrast sensitivity and reading ability were specified as secondary outcomes. In the IVAN 2013 trial, contrast sensitivity, near visual acuity, and reading index outcomes were specified as secondary outcomes. We identified one conference abstract in which contrast sensitivity outcomes were reported for the ANCHOR 2006, MARINA 2006, and PIER 2008 trials.

Visual function outcomes were not reported by six studies (Biswas 2011; CATT 2011; MANTA 2013; Sacu 2009; Subramanian 2010; VISION 2004).

Morphological outcomes

All studies included at least one measure relating to the morphological characteristics of neovascular lesions in study eyes. In many cases, sufficient data were not available in publications or conference abstracts to analyze these outcomes informatively. Whenever possible we used data provided by primary investigators or asked primary investigators to confirm data extracted from graphs. We have not reported data derived from graphs in study reports unless confirmation of the data was received from study investigators.

All studies used fluorescein angiography. Fundus photography also was used in five studies (ANCHOR 2006; GEFAL 2013; MARINA 2006; PIER 2008; VISION 2004) and ICG angiography was used in two studies (GEFAL 2013; Sacu 2009). Mean change in the size of CNV was evaluated by fluorescein angiography in six studies (ABC 2010; ANCHOR 2006; GEFAL 2013; MARINA 2006; PIER 2008; VISION 2004) and mean change in the size of neovascular lesions was evaluated by fluorescein angiography in seven studies (ABC 2010; ANCHOR 2006; CATT 2011; IVAN 2013; MARINA 2006; PIER 2008; VISION 2004).

OCT for the assessment of subretinal characteristics of eyes with neovascular AMD was not used in the earliest study included in the review (VISION 2004). The next three studies conducted chronologically (ANCHOR 2006; MARINA 2006; PIER 2008) used OCT to assess a subset of study participants. In the eight studies most recently conducted (ABC 2010; Biswas 2011; CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013; Sacu 2009; Subramanian 2010), OCT was used in all study participants and at least one OCT measure was specified as a primary or secondary outcome. Mean change in CRT was assessed in the 11 studies that used OCT. We considered central macular thickness, central foveal thickness, and center point thickness to be interchangeable terms for CRT.

Other morphological outcomes, such as area of CNV leakage and subretinal fluid, were reported by individual studies, but we did not include these outcomes in this review.

Quality‐of‐life outcomes

Vision‐specific quality of life was evaluated in four studies (ANCHOR 2006; MARINA 2006; PIER 2008; VISION 2004) using the 25‐item National Eye Institute‐Visual Functioning Questionnaire (NEI‐VFQ). The NEI‐VFQ, administered by an interviewer, relies on patient‐reported responses to specific visual function questions in order to calculate overall and subscale scores, which can range from 0 to 100, with higher values representing better visual function.

One study (IVAN 2013) evaluated general quality of life using the EuroQoL health‐related quality of life assessment (EQ‐5D). For the EQ‐5D, participant responses to specific health questions are converted to scales of 1 to 3, where 1 represents no health problems, 2 represents moderate health problems, and 3 represents extreme health problems. The scores for each of the five subscale domains (mobility, self‐care, usual activities, pain/discomfort and anxiety/depression) are then summarized into a single index score ranging from ‐0.59 to 1.00, with 1.00 representing no health problems. Both the NEI‐VFQ and EQ‐5D are validated tools by which to assess quality‐of‐life outcomes.

Quality‐of‐life outcomes were been reported by seven studies (ABC 2010; Biswas 2011; CATT 2011; GEFAL 2013; MANTA 2013; Sacu 2009; Subramanian 2010).

Economic outcomes

Two studies included economic‐related outcomes as prespecified secondary outcomes. In the CATT 2011 study, the annual costs associated with each treatment group were evaluated. In the IVAN 2013 study, cumulative resource use and costs for each treatment group were evaluated.

Adverse events

Ocular and non‐ocular adverse events up to one year of follow up were reported by studies (ABC 2010; GEFAL 2013; MANTA 2013; Sacu 2009; Subramanian 2010), up to 18 months of follow up by one study (Biswas 2011), up to two years of follow up by five studies (ANCHOR 2006; CATT 2011; IVAN 2013; MARINA 2006; PIER 2008), and up to four years of follow up by one study (VISION 2004).

Excluded studies

We excluded 41 studies after full‐text assessment: 18 studies were not RCTs; 9 studies followed participants for less than one year; 7 studies were dose‐response studies in which no control or comparator arm was part of the study; 5 studies compared combination therapies in which treatment groups received the same anti‐VEGF therapy; 1 study did not include participants with neovascular AMD; and 1 study evaluated intravitreal aflibercept for the treatment of AMD, which is covered in a separate Cochrane review.

See: Characteristics of excluded studies

Risk of bias in included studies

Assessments of risks of bias for each included study are given at the end of each respective 'Characteristics of included studies' table. When unpublished information was needed to assess the risk of bias for any given parameter, we contacted primary investigators for additional information. We have documented these instances together with the investigators' responses in the 'Characteristics of included studies' tables. Figure 2 summarizes the 'Risk of bias' assessments for all 12 studies.

Figure 2

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Allocation

Overall the included studies were at low risk of selection bias. Reports from 11 of the 12 studies described methods of random sequence generation that we judged to confer a low risk of bias; the method used in the Subramanian 2010 trial was not described in sufficient detail for us to assess its risk of bias. The most common method used for random sequence generation was dynamic randomization, used in five studies (ABC 2010; ANCHOR 2006; MARINA 2006; PIER 2008; VISION 2004). Three studies used permuted block randomization designs (CATT 2011; IVAN 2013; MANTA 2013), two studies used random number tables or lists (Biswas 2011; GEFAL 2013), and one study only reported using a computer‐randomized schema (Sacu 2009).

Investigators of 10 of the 12 trials reported adequate allocation concealment. In Biswas 2011, it was unclear whether the randomization sequence, determined by random number tables generated prior to study enrollment, was concealed or made available to the study investigators. Reports from the PIER 2008 study did not describe how assignments were allocated; we were unable to make an assessment using available information. In seven studies a third party or central coordinating center was employed (ABC 2010; ANCHOR 2006; GEFAL 2013; MANTA 2013; MARINA 2006; Sacu 2009; Subramanian 2010) and in three studies (CATT 2011; IVAN 2013; VISION 2004) a computer‐based portal was used for allocation concealment.

Communication with investigators from the Biswas 2011, Subramanian 2010, and PIER 2008 studies did not yield additional information about the methods used to assess these risks of bias (email communications).

Masking (performance bias and detection bias)

Most of the included studies we judged to be at low risk of performance bias and detection bias. Only one study (Sacu 2009) was an open‐label study in which no form of masking was employed. Participants in the CATT 2011 study initially were masked to the drug (not to the dosing schedule), but may have become aware of the treatment assignments due to billing records. The authors of Biswas 2011 did not report whether study participants were masked. In both the Biswas 2011 and CATT 2011 studies, personnel and outcome assessors were masked. In the remaining nine studies, study participants, personnel (other than personnel directly administering treatment), and outcome assessors were masked; thus, we assessed these studies as being at low risk of performance bias and detection bias. The most common method used to mask participants in studies in which intravitreal injections were compared to no injections was the use of sham injections when participants were not assigned or did not require an injection. In the head‐to‐head studies of ranibizumab versus bevacizumab, participants were masked as to which treatment group they were assigned. To minimize detection bias, study investigators who were involved with assessing outcomes were separate from the treating physicians and masked to treatment groups, with the exception of the Sacu 2009 study in which no masking was done.

Incomplete outcome data

In all 12 trials, few participants missed the follow‐up examination specified as the primary time for assessing the study's primary outcome or were not treated in accord with the randomized treatment assignment. In nine trials, the rates of loss to follow up at the primary follow‐up visits were less than 15%; the GEFAL 2013, MANTA 2013, and Subramanian 2010 studies had 19%, 23%, and 21% missing data, respectively. Losses to follow up were evenly balanced across treatment groups among the included studies.

Seven trials included in this review analyzed the data using methods designed to overcome, in part, loss of information due to missed follow‐up examinations. Six of these seven trials used the last‐observation‐carried‐forward method to impute missing data (ABC 2010; ANCHOR 2006; MANTA 2013; MARINA 2006; PIER 2008; VISION 2004) and the seventh trial (Sacu 2009) did not report the method for imputing data for one participant with missing data. The remaining five trials reported the available‐case data in which only participants with data were included in the analyses: 91.5% in CATT 2011, 81% in GEFAL 2013, 89% in IVAN 2013, 87% in Biswas 2011, and 79% in Subramanian 2010. The investigators of all trials reported that they had analyzed data for participants by the treatment arms to which they had been assigned. However, analyses using single imputation methods or available‐case data assume that participants are lost to follow up at random and may introduce bias if this assumption is not true.

Selective reporting

With the exception of the Biswas 2011 study, we identified protocols or clinical trial registrations for 11 of the included studies. We judged seven of these 11 trials to be free of reporting bias based on the consistency between study outcomes defined in the protocols and clinical trial registrations and those reported in the study results papers. Although quality‐of‐life outcomes were specified, we identified no report of quality‐of‐life findings from the Subramanian 2010 trial. We also found no reports of reading ability outcomes, which were specified as secondary outcomes in the ABC 2010 trial. Results for three outcomes specified in the IVAN 2013 trial protocol were not reported in published articles of the one‐year and two‐year results: treatment satisfaction, survival free from treatment failure, and exploratory (serum) analysis. Differences in outcomes between the trial registration and the published one‐year results paper of the GEFAL 2013 study included differences in details of outcome specification (e.g., efficacy of treatments versus proportion of participants with a gain of 15 letters or more of visual acuity), outcomes specified in the trial registration not reported in the paper, and a newly added outcome in the paper that was not listed in the trial registration.

Other potential sources of bias

Various other aspects of trial design, reporting, trial sponsorship, and financial interests of investigators were considered as other potential sources of bias.

ANCHOR 2006, MARINA 2006, PIER 2008, and VISION 2004 were sponsored by pharmaceutical companies marketing the study drugs under investigation; data from these trials were submitted to the FDA to obtain approval of ranibizumab and pegaptanib. In addition, the pharmaceutical company sponsors had important roles in the trial design, analysis, and reporting. Some investigators from other trials reported that they received financial support from pharmaceutical companies; however, because the companies did not have direct sponsorship of the trials we did not judge these trials to be at risk of bias for this domain (CATT 2011; GEFAL 2013; IVAN 2013). We observed no other potential sources of bias in the remaining five studies.

Effects of interventions

We conducted meta‐analyses of results by drug, combining different doses and regimens of the same drug evaluated in the individual trials, as evident from Table 2. In the forest plots reported in this review, it should be noted that, for all visual acuity outcomes, effect estimates to the right of the vertical line of the forest plots (i.e., risk ratios > 1 and mean differences > 0) favor the test treatment.

Pegaptanib versus sham

One study, comprising two individual RCTs, compared three doses of intravitreal pegaptanib (0.3 mg, 1.0 mg, and 3.0 mg) with a sham injection control group (VISION 2004). The study was conducted at 117 international centers and enrolled 1208 adult participants (50 years of age or older) with subfoveal CNV lesions secondary to AMD. Findings from the two trials were combined in study reports. There were 904 participants in the pegaptanib groups and 304 in the sham injection group. At one year of follow up, 1186 (98%) participants were included in the primary analyses and 1053 (87%) remained in the study and were re‐randomized according to their original treatment assignment. Participants in the pegaptanib groups were re‐randomized to continue current treatment or discontinue treatment and participants in the sham group were re‐randomized to continue with sham injections, discontinue sham injections, or receive one of the three study doses of pegaptanib. Study follow up continued for one year after the re‐randomization and participants were analyzed in three cohorts: those who continued with their original assignments, those who discontinued treatment, and those who received sham injections during the first year then pegaptanib during the second year. In total, 1053 (87%) participants were included in the two‐year analysis; however, we did not analyze the two‐year data since results were reported for changes from year 1 to year 2 rather than from baseline to year 2.

The VISION 2004 study, sponsored by Eyetech Pharmaceuticals and Pfizer, was assessed as having a low risk of bias in most domains. The study had high retention rate (90% of participants followed for one year and 85% for two years).

(1) Visual acuity

(a) Gain of 15 letters or more of visual acuity

More participants in the combined pegaptanib groups than in the sham group had a gain of 15 letters or more of visual acuity at year 1. Since gaining vision is a positive outcome, an RR greater than 1 favors treatment with pegaptanib. The RR for the combined pegaptanib versus sham groups was 2.83 (95% CI 1.23 to 6.52); that is, eyes treated with pegaptanib were 2.83 times more likely to gain 15 letters or more of vision than eyes treated with sham injections (Analysis 1.1; Figure 3).

Figure 3

Forest plot of comparison: 1 Pegaptanib versus control, outcome: 1.1 Gain of 15 letters or more visual acuity at one year.

(b) Loss of fewer than 15 letters of visual acuity

At one year, the RR for loss of fewer than 15 letters of visual acuity when comparing the combined pegaptanib groups with the sham group favored treatment with pegaptanib and was 1.24 (95% CI 1.11 to 1.39) (Analysis 1.2).

(c) Loss of fewer than 30 letters of visual acuity

At one year, the RR for the combined pegaptanib groups versus the sham group was 1.15 (95% CI 1.08 to 1.23), indicating that eyes treated with pegaptanib were 15% (95% CI 8% to 23%) less likely to have lost 30 letters or more of vision compared with eyes treated with sham therapy (Analysis 1.3).

(d) Prevention of blindness (visual acuity better than 20/200)

Pegaptanib resulted in fewer blind participants across the three treatment groups at one year of follow up. The pooled RR for having visual acuity better than 20/200 was 1.33 (95% CI 1.15 to 1.52) compared with sham therapy (Analysis 1.4).

(e) Maintenance of visual acuity

A greater number of participants treated with pegaptanib maintained visual acuity at one year. The pooled RR for pegaptanib versus sham therapy was 1.49 (95% CI 1.19 to 1.88) (Analysis 1.5).

(f) Mean change in visual acuity

The study investigators provided us with the data required to analyze the mean change in visual acuity at one year of follow up. Visual acuity was measured using the ETDRS chart placed at 2 m from the participant. The mean change from baseline in number of letters read on the ETDRS chart was measured. The final mean visual acuity in the three active treatment groups was consistently greater than in the sham group.

The MD in mean change in visual acuity from baseline between the combined pegaptanib groups versus the sham group was 6.72 letters (95% CI 4.43 to 9.01), meaning that eyes treated with pegaptanib lost on average 6.72 letters fewer than eyes in the sham‐treated group (Analysis 1.6). On the logMAR scale, 0.10 logMAR units correspond to 1 line (5 letters) on the visual acuity chart. Thus, the MD between the pegaptanib and sham groups was equivalent to 0.13 logMAR units; that is, the mean change in visual acuity was less in the pegaptanib groups than in the sham group by 0.13 logMAR units.

(2) Visual function

Visual function outcomes were not reported in VISION 2004.

(3) Morphological outcomes

Mean change in size of CNV

Insufficient information was available to analyze the mean change in size of CNV; however, study investigators provided data allowing us to evaluate mean CNV size at one year of follow up. Given that baseline CNV sizes were comparable among all study participants, the difference in the mean size of CNV between study groups at one year may be used to estimate the treatment effect. Total CNV sizes were measured as numbers of standard DA. Pegaptanib treatment, across all doses studied, resulted in a lower final mean CNV size compared to the sham group at one year of follow up (MD 0.92 DAs; 95% CI 0.42 to 1.42) (Analysis 1.7). We considered a difference in size of CNV of one DA or more as a clinically meaningful difference.

Mean change in size of lesion

Insufficient information was available to analyze the mean change in size of total subfoveal lesion; however, study investigators provided data allowing us to analyze the mean size of lesion at one year of follow up. Mean size of lesion is described in terms of standard DA. Pegaptanib treatment resulted in a lower mean size of lesion at one year of follow up compared to sham treatment (MD 0.86 DAs; 95% CI 0.35 to 1.37) (Analysis 1.8).

Mean change in CRT

OCT was not used in the VISION 2004 study; CRT outcomes were not measured.

(4) Quality‐of‐life outcomes

Vision‐related quality of life was measured in one of the two trials from the VISION 2004 study using the NEI‐VFQ questionnaire (EOP 1004). Since it was validated only for United States English, this questionnaire was administered only to participants from the United States and Canada, 578 (48%) of 1208 total study participants, at baseline and at weeks 30 and 54 of follow up. At one year, data were available for 569 (98%) of the 578 study participants who completed the questionnaire at baseline. Treatment with pegaptanib was associated with better scores on the NEI‐VFQ questionnaire, specifically for distance vision and role limitation domains. However, standard deviations for scores were not reported.

(5) Economic outcomes

We did not find any report of economic outcomes comparing pegaptanib with sham treatment from the VISION 2004 study.

(6) Adverse events

Ocular and systemic adverse events were reported in the VISION 2004 study. Participants in the pegaptanib groups experienced an ocular adverse event nearly four times more often (RR 3.84; 95% CI 0.91 to 16.20) and were 1.25 times more likely to have a serious systemic adverse event (RR 1.25; 95% CI 0.93 to 1.70) than participants in the control group. Results for the most frequent adverse events are shown in Table 3. Although uncommon, 12 eyes treated with pegaptanib injections for one year developed endophthalmitis, compared to no cases in control eyes. Because of the small number of events, risk estimates for individual adverse events are imprecise.

Table 3. Adverse events up to one year: pegaptanib vs. control

Ocular adverse event*	0.3 mg pegaptanib n = 295	1.0 mg pegaptanib n = 301	3.0 mg pegaptanib n = 296	All doses pegaptanib n = 892	Control n = 298	RR [95% CI] All doses vs. control
Any eye disorder	9 (3%)	4 (1%)	10 (3%)	23 (3%)	2 (< 1%)	3.84 [0.91, 16.20]
Endophthalmitis	6 (2%)	3 (1%)	3 (1%)	12 (1%)	0	8.37 [0.50, 140.95]
Retinal detachment	1 (< 1%)	2 (< 1%)	2 (< 1%)	5 (< 1%)	0	3.68 [0.20, 66.41]
Traumatic cataract	1 (< 1%)	2 (< 1%)	2 (< 1%)	5 (< 1%)	0	3.68 [0.20, 66.41]
Retinal hemorrhage	1 (< 1%)	0	1 (< 1%)	2 (< 1%)	0	1.67 [0.08, 34.77]
Vitreous hemorrhage	0	0	1 (< 1%)	1 (< 1%)	0	1.00 [0.04, 24.59]
Uveitis	0	0	1 (< 1%)	1 (< 1%)	0	1.00 [0.04, 24.59]
Elevated intraocular pressure	1 (< 1%)	0	0	1 (< 1%)	0	1.00 [0.04, 24.59]
Papilledema	0	0	0	0	1 (< 1%)	0.11 [0.00, 2.73]
Non‐ocular adverse event*	0.3 mg pegaptanib n = 295	1.0 mg pegaptanib n = 301	3.0 mg pegaptanib n = 296	All doses pegaptanib n = 892	Control n = 298	RR [95% CI] All doses vs. control
At least 1 serious adverse event	55 (19%)	50 (17%)	64 (22%)	169 (19%)	45 (15%)	1.25 [0.93, 1.70]
Cardiac disorders	11 (4%)	4 (1%)	10 (3%)	25 (3%)	14 (5%)	0.60 [0.31, 1.13]
Neoplasms (benign, malignant, unspecified)	11 (4%)	7 (2%)	8 (3%)	26 (3%)	12 (4%)	0.72 [0.37, 1.42]
Injury and procedural complications, such as fractures (also includes traumatic cataracts)	10 (3%)	9 (3%)	8 (3%)	27 (3%)	3 (1%)	3.01 [0.92, 9.84]
Nervous system disorders	10 (3%)	5 (2%)	10 (3%)	25 (3%)	7 (2%)	1.19 [0.52, 2.73]
Infections and infestations	2 (<1%)	7 (2%)	11 (4%)	20 (2%)	5 (2%)	1.34 [0.51, 3.53]
Gastrointestinal disorders	3 (1%)	6 (2%)	5 (2%)	14 (2%)	4 (1%)	1.17 [0.39, 3.52]
Respiratory, thoracic, mediastinal disorders	2 (< 1%)	5 (2%)	5 (2%)	12 (1%)	4 (1%)	1.00 [0.33, 3.08]
Musculoskeletal and connective tissue	1 (< 1%)	5 (2%)	3 (1%)	9 (1%)	2 (<1%)	1.50 [0.33, 6.92]
Renal and urinary disorders	2 (< 1%)	3 (1%)	2 (<1%)	7 (<1%)	3 (1%)	0.78 [0.20, 3.00]
Vascular disorders	3 (1%)	2 (< 1%)	2 (< 1%)	7 (< 1%)	3 (1%)	0.78 [0.20, 3.00]

CI: confidence interval

RR: risk ratio

*Most frequent serious adverse events experienced by 1190 participants in the VISION 2004 study

Ranibizumab versus control

Three studies comprising a total of 1323 participants compared two doses of intravitreal ranibizumab (0.3 mg or 0.5 mg) with sham or control treatment. In the ANCHOR 2006 study, 280 participants received ranibizumab and 143 participants received verteporfin PDT therapy. Injections were administered monthly and verteporfin PDT therapy was administered on day 0 and as needed at visits at months 3, 6, 9, and 12. In the MARINA 2006 study, 478 participants received ranibizumab and 238 participants received sham injections. All injections were administered on a monthly basis. In the PIER 2008 study, 121 participants received ranibizumab and 63 participants received sham injections. Injections were administered monthly for the first three months, then every three months. During the second year of the PIER 2008 study, participants in the 0.3 mg ranibizumab and sham‐treated groups crossed over to receive 0.5 mg ranibizumab.

Overall, we rated the risk of bias as low among the three studies. Participant masking in all studies was achieved by the use of sham injections and by sham PDT therapy in the ANCHOR 2006 study. Although the clinicians administering treatment were not masked, those assessing outcomes were masked to treatment groups. At one year of follow up, two participants, one participant in the ANCHOR 2006 study and one in the PIER 2008 study, were excluded from the analyses. The remaining study participants were analyzed and missing data were imputed using the last‐observation‐carried‐forward method. All three studies were funded by pharmaceutical companies (Genentech, USA, and Novartis Pharma, Switzerland).

(1) Visual acuity

(a) Gain of 15 letters or more of visual acuity

At one year of follow up, we observed substantial statistical heterogeneity in the effect of ranibizumab compared with sham therapy for a gain of 15 letters or more of visual acuity; the I² statistic was 80% and the Chi² test for heterogeneity was statistically significant (P value < 0.01) (Analysis 2.1; Figure 4). Therefore, we did not combine study results in meta‐analysis. The RR for a gain of 15 letters or more of visual acuity comparing the combined ranibizumab groups with sham control was 6.79 (95% CI 3.41 to 13.54) in ANCHOR 2006, 5.81 (95% CI 3.29 to 10.26) in MARINA 2006, and 1.30 (95% CI 0.53 to 3.19) in PIER 2008. Although the direction of treatment effect in all three trials included in this analysis favored ranibizumab, the magnitude of effect observed in PIER 2008 was smaller and not statistically significant compared with that observed in MARINA 2006 and ANCHOR 2006. This difference may have been attributable to differences among the studies with respect to dosing schedules (monthly injections in ANCHOR 2006 and MARINA 2006 versus injections monthly for the first three months, then every three months in PIER 2008).

Figure 4

Forest plot of comparison: 2 Ranibizumab versus control, outcome: 2.1 Gain of 15 letters or more visual acuity at one year.

At two years of follow up, there was less statistical heterogeneity for this outcome, as indicated by the I² statistic (30%) and the Chi² test for heterogeneity (Analysis 2.2), although the PIER 2008 findings still deviated somewhat from those of the other two trials. The proportion of participants who were treated with ranibizumab and gained 15 letters or more at two years was 5.77 times the proportion of participants treated with control interventions who gained 15 letters or more (RR 5.77; 95% CI 3.38 to 9.84) (Analysis 2.2).

(b) Loss of fewer than 15 letters of visual acuity

A greater proportion of participants treated with ranibizumab lost fewer than 15 letters of visual acuity at one year of follow up compared with those treated with sham or control therapy. Participants were 1.53 times more likely to not lose 15 letters or more of visual acuity when treated with ranibizumab compared with sham or control therapy (RR 1.53; 95% CI 1.41 to 1.64) (Analysis 2.3). We observed no statistical heterogeneity among the three trials (I² = 0) and the CIs of the individual trials overlapped one another.

At two‐year follow‐up, the beneficial effect of ranibizumab persisted at a similar magnitude and was statistically significant when compared with sham or control therapy. Nearly twice as many participants treated with ranibizumab lost fewer than 15 letters of visual acuity at two years of follow up than those in control groups (RR 1.62; 95% CI 1.32 to 1.98) (Analysis 2.4). We observed substantial statistical heterogeneity in the analysis comparing ranibizumab with control (I² = 78%; P value for Chi² test of heterogeneity = 0.01); however, the CIs among the individual studies overlapped and the effect estimates were in the same direction. This difference may have been attributable to the control group in the ANCHOR 2006 study receiving an active treatment (verteporfin PDT therapy) compared with sham injections in MARINA 2006 and PIER 2008.

(c) Loss of fewer than 30 letters of visual acuity

Data for this outcome were available from only two of the three trials that compared ranibizumab with control interventions (ANCHOR 2006; MARINA 2006). At one‐year follow up, fewer than 1% of participants treated with ranibizumab lost 30 letters or more of visual acuity (5/757) compared with 14% in the control groups (53/381). Comparing both ranibizumab groups combined with controls, we observed a 15% benefit of ranibizumab with respect to the loss of fewer than 30 letters of visual acuity (RR 1.15; 95% CI 1.11 to 1.20) (Analysis 2.5). The meta‐analysis for this outcome revealed no statistical heterogeneity (I² = 0%) and the point estimates and CIs of the two studies analyzed overlapped.

The treatment effect persisted through two years, with fewer participants treated with ranibizumab losing 30 letters or more (16/757, 2%) than participants in the control groups (77/381, 20%). Comparing both ranibizumab groups combined versus controls, we observed a 22% benefit of ranibizumab with respect to the loss of fewer than 30 letters of visual acuity after two years (RR 1.22; 95% CI 1.15 to 1.29) (Analysis 2.6).

(d) Prevention of blindness (visual acuity better than 20/200)

Ranibizumab resulted in fewer cases of blindness at both one and two years of follow up compared with control interventions. In a meta‐analysis comparing the combined ranibizumab groups with the control intervention groups, a greater proportion of participants in the ranibizumab groups had visual acuity better than 20/200 than participants in the control group at one year (RR 1.69; 95% CI 1.41 to 2.03) (Analysis 2.7) and two years (RR 1.73; 95% CI 1.52 to 1.98) (Analysis 2.8). Although the point estimates and CIs of the individual studies overlapped one another, some degree of statistical heterogeneity was observed for this outcome (I² = 62% at one year and 26% at two years).

(e) Maintenance of visual acuity

Data on the maintenance of visual acuity were available for only one of the three trials comparing ranibizumab with a control intervention (ANCHOR 2006). At both one and two years of follow up, a greater proportion of participants treated with ranibizumab maintained visual acuity (i.e., visual acuity at follow up was the same as or better than at baseline), compared with participants in the control group. At one year, the RR for the maintenance of visual acuity comparing ranibizumab with control was 2.53 (95% CI 1.95 to 3.27) (Analysis 2.9). At two years, the corresponding effect estimate was 2.71 (95% CI 2.08 to 3.54) (Analysis 2.10).

(f) Mean change in visual acuity

On average, at both one and two years of follow up, participants treated with ranibizumab read more letters on ETDRS charts placed at 4 m than participants in the control groups. Participants treated with ranibizumab were able to read 18 letters more at the one‐year follow up (MD 17.80, 95% CI 15.95 to 19.65) (Analysis 2.11) and 20 letters more at the two‐year follow up (MD 20.11, 95% CI 18.08 to 22.15) (Analysis 2.12) than participants in the control groups.

(2) Visual function

Visual function outcomes were not specified as outcomes of interest by any of the three trials; however, we identified one conference abstract that discussed contrast sensitivity outcomes in participants from these trials (see Korobelnik 2006 under ANCHOR 2006). No between‐group comparisons were reported for contrast sensitivity as measured using Pelli‐Robson charts, but the abstract author reported that participants in the ranibizumab groups had statistically significant increases of 2 to 4 letters (i.e., approximately one contrast level) after one year. Participants in the control groups lost an average of 3 letters (i.e., one contrast level) at one year. The MD comparing ranibizumab with control would be 6 letters (i.e., two contrast levels on the Pelli‐Robson chart), based on data extracted from the abstract.

(3) Morphological outcomes

Mean change in size of CNV

We were unable to identify and extract any data on mean change in size of the CNV from any of the three included trials comparing ranibizumab with control interventions.

Mean change in size of lesion

Data on the mean change in size of the total subfoveal lesion were available from two of the three included trials comparing ranibizumab with control interventions (ANCHOR 2006; PIER 2008). The mean reduction in the size of the lesion was greater by 2.34 DAs (95% CI 1.88 to 2.81) among participants treated with ranibizumab compared with participants treated with control interventions after one year (Analysis 2.13). At two years, this effect persisted in the ANCHOR 2006 study (MD 2.44, 95% CI 1.87 to 3.00), but was muted in the PIER 2008 study (MD 0.59, 95% CI ‐0.55 to 1.73) (Analysis 2.14). Due to substantial statistical heterogeneity (I² = 88%) and differences in control groups in the two trials during the second year of follow up, we did not combine these studies in a meta‐analysis.

Mean change in CRT

We were unable to find data on CRT in reports from any of the three included trials comparing ranibizumab with control interventions.

(4) Quality‐of‐life outcomes

Vision‐related quality of life was measured in all three trials using an interviewer‐administered NEI‐VFQ questionnaire at baseline, and after one and two years of follow up. Two trials provided sufficient data to include in meta‐analysis (ANCHOR 2006; MARINA 2006). The investigators of both studies considered a 10‐point change in scores as clinically meaningful.

At one year, overall vision‐related quality of life improved more often among participants in ranibizumab groups compared with participants in control groups (MD 6.69; 95% CI 3.38 to 9.99). The MD was greater in the MARINA 2006 study (MD 8.20; 95% CI 6.05 to 10.35) than in the ANCHOR 2006 study (MD 4.81; 95% CI 1.74 to 7.87). This difference between the two trials may be because participants in the control group of the ANCHOR 2006 study received an active treatment, verteporfin PDT therapy. Subscale domains of the NEI‐VFQ questionnaire in which participants in ranibizumab groups showed greater improvement at one‐year of follow up than participants in control groups included near‐vision activities, distance‐vision activities, vision‐related dependency, driving ability, general health, role difficulties, mental health, general vision, social functioning, color vision, and peripheral vision (Analysis 2.15). The I² statistic for subscale analyses ranged from 0 to 91%, which may have been due to differences in control groups. No combined MDs differed by more than 10 points between ranibizumab and control groups.

At two years, overall vision‐related quality of life improved more often among participants in ranibizumab groups compared with participants in control groups (MD 8.63; 95% CI 3.31 to 13.95). Similar to one‐year results, the MD was greater in the MARINA 2006 study (MD 11.15; 95% CI 8.81 to 13.48) than in the ANCHOR 2006 study (MD 5.70; 95% CI 1.96 to 9.44). Subscale domains of the NEI‐VFQ questionnaire in which participants in ranibizumab groups showed greater improvement at two years of follow up compared with participants in control groups were consistent with those identified at one‐year (Analysis 2.16). The I² statistic for subscale analyses ranged from 0 to 87%, reflecting greater comparative differences between the treatment and sham control groups in the MARINA 2006 study than between the treatment and active control groups in the ANCHOR 2006 study. For five subscales, MDs differed by more than 10 points between the ranibizumab and control groups: near vision activities, distance vision activities, vision‐related dependency, driving ability, and mental health (Analysis 2.16).

We did not extract the limited data available from the third trial (PIER 2008) because the data in the full‐text articles were presented only as graphs and the information contained in the conference abstracts was insufficient for inclusion in our analysis. Correspondence with the trial investigators did not yield additional information for data analysis.

(5) Economic outcomes

We did not identify data on economic outcomes comparing ranibizumab with controls directly from the ANCHOR 2006, MARINA 2006, and PIER 2008 studies. Estimates of the cost of treatment with ranibizumab were reported to be USD 27,004 for the first year and USD 26,417 for the second year, based on data from the MARINA 2006 study; data were not reported for the control group (Brown 2008).

(6) Adverse events

Ocular and systemic adverse events were reported in the ANCHOR 2006 and PIER 2008 studies at one‐year follow up (Table 4) and in all three studies at two‐year follow up (Table 5). At both the one‐ and two‐year follow ups, there were small numbers of participants who experienced ocular adverse events, such as endophthalmitis, uveitis, retinal detachment, and retinal or vitreous hemorrhage, and non‐ocular adverse events, such as myocardial infarction, stroke or cerebral infarction, ischemic cardiomyopathy, and death (< 1% of total participants). Because of the small number of events, risk estimates for these adverse events are imprecise.

Table 4. Adverse events up to one year: ranibizumab vs. control

Ocular adverse event*	0.3 mg ranibizumab n = 196	0.5 mg ranibizumab n = 201	All doses ranibizumab n = 397	Control n = 206	RR [95% CI] All doses vs. control
Endophthalmitis	0	2 (< 1%)	2 (< 1%)	0	2.60 [0.13, 53.92]
Retinal detachment	1 (< 1%)	0	1 (< 1%)	1 (< 1%)	0.52 [0.03, 8.25]
Traumatic cataract	18 (9%)	22 (11%)	40 (10%)	14 (7%)	1.48 [0.83, 2.66]
Retinal hemorrhage	2 (1%)	0	2 (< 1%)	2 (< 1%)	0.52 [0.07, 3.66]
Vitreous hemorrhage	1 (< 1%)	0	1 (< 1%)	0	1.56 [0.06, 38.13]
Uveitis	0	1 (< 1%)	1 (< 1%)	0	1.56 [0.06, 38.13]
Elevated intraocular pressure (30 mmHg or more increase)	13 (7%)	17 (8%)	30 (8%)	7 (3%)	2.22 [0.99, 4.98]
Ocular inflammation (trace to 4+)	21 (11%)	26 (13%)	47 (12%)	9 (4%)	2.71 [1.36, 5.42]
Non‐ocular adverse event*	0.3 mg ranibizumab n = 196	0.5 mg ranibizumab n = 201	All doses ranibizumab n = 397	Control n = 206	RR [95% CI] All doses vs. control
Death	3 (2%)	2 (<1%)	5 (1%)	2 (< 1%)	1.30 [0.25, 6.63]
Myocardial infarction	1 (< 1%)	3 (1%)	4 (1%)	1 (< 1%)	2.08 [0.23, 18.45]
Stroke or cerebral infarction	1 (< 1%)	1 (< 1%)	2 (< 1%)	1 (< 1%)	1.04 [0.09, 11.38]
Ischemic cardiomyopathy	0	0	0	1 (< 1%)	0.17 [0.01, 4.24]
Treatment‐emergent hypertension	7 (4%)	15 (7%)	22 (6%)	17 (8%)	0.67 [0.36, 1.24]
Non‐ocular hemorrhage	9 (5%)	13 (6%)	22 (6%)	6 (3%)	1.90 [0.78, 4.62]

CI: confidence interval

RR: risk ratio

*Adverse events experienced by 420 participants in the ANCHOR 2006 study and 183 participants in the PIER 2008 study. Adverse events at one‐year follow up not reported in MARINA 2006.

Table 5. Adverse events up to two years: ranibizumab vs. control

Ocular adverse event*	0.3 mg ranibizumab n = 434	0.5 mg ranibizumab n = 440	All doses ranibizumab n = 874	Control n = 441	RR [95% CI] All doses vs. control
Endophthalmitis	2 (< 1%)	6 (1%)	8 (< 1%)	0	8.59 [0.50, 148.44]
Retinal detachment	2 (< 1%)	0	2 (< 1%)	2 (< 1%)	0.50 [0.07, 3.57]
Traumatic cataract	65 (15%)	76 (17%)	141 (16%)	57 (13%)	1.25 [0.94, 1.66]
Retinal hemorrhage	1 (< 1%)	0	1 (< 1%)	1 (< 1%)	0.50 [0.03, 8.05]
Vitreous hemorrhage	3 (< 1%)	1 (< 1%)	4 (< 1%)	2 (< 1%)	1.01 [0.19, 5.49]
Uveitis	3 (< 1%)	4 (< 1%)	7 (<1%)	0	7.58 [0.43, 132.36]
Elevated intraocular pressure (30 mmHg or more increase)**	45 (15%)	61 (20%)	106 (18%)	11 (4%)	4.81 [2.63, 8.81]
Ocular inflammation (1+ to 4+)	32 (7%)	30 (7%)	62 (7%)	8 (2%)	3.91 [1.89, 8.09]
Non‐ocular adverse event*	0.3 mg ranibizumab n = 434	0.5 mg ranibizumab n = 440	All doses ranibizumab n = 874	Control n = 441	RR [95% CI] All doses vs. control
Death	12 (3%)	9 (2%)	21 (2%)	13 (3%)	0.82 [0.41, 1.61]
Myocardial infarction	7 (2%)	8 (2%)	15 (2%)	7 (2%)	1.08 [0.44, 2.63]
Stroke or cerebral infarction	6 (1%)	6 (1%)	12 (1%)	5 (1%)	1.21 [0.43, 3.42]
Ischemic cardiomyopathy	0	0	0	1 (< 1%)	0.17 [0.01, 4.12]
Treatment‐emergent hypertension	60 (14%)	69 (16%)	129 (15%)	68 (15%)	0.96 [0.73, 1.25]
Nonocular hemorrhage	38 (9%)	40 (9%)	78 (9%)	24 (5%)	1.64 [1.05, 2.55]

CI: confidence interval

RR: risk ratio

*Adverse events experienced by 420 participants in the ANCHOR 2006 study; 713 participants in the MARINA 2006 study; and 182 participants in the PIER 2008 study.
**Adverse events for elevated intraocular pressure not reported in the ANCHOR 2006 study at two‐year follow up (n = 297 in 0.3 mg ranibizumab group, n = 300 in 0.5 mg ranibizumab group, and n = 298 in 0.3 mg control group).

With respect to ocular adverse events, eyes treated with ranibizumab more often developed cataracts compared with eyes in the control groups at both the one year (RR 1.48; 95% CI 0.83 to 2.66) and two year follow ups (RR 1.25; 95% CI 0.94 to 1.66). Elevated intraocular pressure (IOP), defined as a 30 mmHg or more increase, occurred more often in eyes in the ranibizumab groups than eyes in the control groups at both one‐year (RR 2.22; 95% CI 0.99 to 4.98) and two‐year follow up (RR 4.81; 95% CI 2.63 to 8.81). Ocular inflammation, graded from trace (1+) to 4+, also occurred more often in eyes in the ranibizumab groups than in eyes in the control groups at both the one year (RR 2.71; 95% CI 1.36 to 5.42) and two year follow ups (RR 3.91; 95% CI 1.89 to 8.09). Two eyes during the first year of ranibizumab injections and six more during the second year developed endophthalmitis, compared to no cases in the control eyes.

With respect to non‐ocular adverse events, participants in the ranibizumab groups less often experienced treatment‐emergent hypertension than participants in the control groups at one‐year follow up (RR 0.67; 95% CI 0.36 to 1.24); however, at two‐year follow up the risk was the same between the ranibizumab and control groups (RR 0.96; 95% CI 0.73 to 1.25). Non‐ocular hemorrhage occurred more often in participants in the ranibizumab groups than in participants in the control groups at both one‐year (RR 1.90; 95% CI 0.78 to 4.62) and two‐year follow up (RR 1.64; 95% CI 1.05 to 2.55).

Bevacizumab versus control

Two studies of 159 total participants compared intravitreal bevacizumab injections with control treatment. In the ABC 2010 study, 131 participants received either 1.25 mg intravitreal bevacizumab (65 participants) or standard therapy consisting of pegaptanib injections (38 participants), verteporfin PDT (16 participants), or sham injections (12 participants). In Sacu 2009, 28 participants received either 1.0 mg intravitreal bevacizumab (14 participants) or verteporfin PDT with 4 mg intravitreal triamcinolone acetonide (14 participants). In both studies intravitreal bevacizumab was administered as needed following the first three scheduled injections. We assessed both studies as having a low risk of bias overall.

(1) Visual acuity

(a) Gain of 15 letters or more of visual acuity

At one year of follow up, the overall RR for a gain in 15 letters or more of visual acuity for bevacizumab versus control treatment was 7.80 (95% CI 2.44 to 24.98) (Analysis 3.1; Figure 5); that is, nearly eight times as many people treated with bevacizumab gained 15 letters or more of visual acuity after one year of treatment compared with control.

Figure 5

Forest plot of comparison: 3 Bevacizumab versus control, outcome: 3.1 Gain of 15 letters or more visual acuity at one year.

(b) Loss of fewer than 15 letters of visual acuity

More participants treated with bevacizumab lost fewer than 15 letters of visual acuity at one year of follow up compared with participants in the control group. The overall RR for bevacizumab versus control treatment was 1.28 (95% CI 1.09 to 1.50) (Analysis 3.2).

(c) Loss of fewer than 30 letters of visual acuity

In the ABC 2010 study, 64/65 participants in the bevacizumab group did not lose 30 letters or more visual acuity at one year of follow‐up compared with 63/66 participants in the control group (RR 1.03; 95% CI 0.97 to 1.10). The authors of Sacu 2009 did not report results for this outcome.

(d) Prevention of blindness (visual acuity better than 20/200)

Blindness was not reported in Sacu 2009. The authors of the ABC 2010 trial noted that more participants in the bevacizumab group than in the control group had visual acuities of 20/200 or better at one year.

(e) Maintenance of visual acuity

Maintenance of visual acuity was not reported in the ABC 2010 study. In Sacu 2009, a greater proportion of participants in the bevacizumab group (11/14) maintained visual acuity at one year compared with participants in the control group (5/14) (RR 2.20; 95% CI 1.03 to 4.68).

(f) Mean change in visual acuity

Insufficient data were available to analyze the difference in mean changes in visual acuity between treatment groups. In the ABC 2010 trial, the mean change from baseline in visual acuity was 7.0 letters in the bevacizumab group and ‐9.4 letters in the control group at one year of follow up. This equates to a MD of 16.4 letters (more than 3 lines of visual acuity); however, we were unable to compute the standard error (SE) using information available. The authors of Sacu 2009 reported a statistically significant difference between groups at one year, when participants in the bevacizumab group had gained 8 letters on average and participants in the control group had lost 3 letters on average.

(2) Visual function

Visual function outcomes were not reported in Sacu 2009. Outcomes for contrast sensitivity, measured with Pelli‐Robson charts, were reported from the ABC 2010 trial. Although the published protocol for the ABC 2010 trial lists reading ability (including a maximum reading speed, critical print size, and reading acuity), measured with Minnesota Reading charts, as a secondary outcome for the study, we did not identify reports with results for this outcome.

At one year, no statistical difference was observed between bevacizumab and control groups in terms of a gain of 15 letters or more (i.e., five levels of contrast) of contrast sensitivity (RR 2.03; 95% CI 0.39 to 10.71); however, a greater proportion of participants in the bevacizumab group (23/65) compared with the control group (10/66) gained 6 letters or more (i.e., two levels of contrast) of contrast sensitivity (RR 2.34; 95% CI 1.21 to 4.51). Also, participants in the control group more often lost 6 letters or more (two contrast levels) of contrast sensitivity compared to participants in the bevacizumab group (RR 0.22; 95% CI 0.07 to 0.72).

(3) Morphological outcomes

Mean change in size of CNV

The median change in the size of CNV was reported in the ABC 2010 study. At 54 weeks, the size of CNV regressed by 0.88 mm² (interquartile range (IQR), reduction of 4.08 mm² to increase of 0.40 mm²) in the bevacizumab‐treated group compared with 0.27 mm² (IQR, reduction of 2.58 mm² to increase of 1.24 mm²) in the control group. The authors of Sacu 2009 did not report results for this outcome.

Mean change in size of lesion

The median change in the size of the subfoveal lesion was reported only for the ABC 2010 study. At 54 weeks, the size of the total subfoveal lesion regressed by 0.03 mm² (IQR, reduction of 1.88 mm² to increase of 2.63 mm²) in the bevacizumab‐treated group and increased by 2.33 mm² (IQR, reduction of 0.06 mm² to increase of 6.44 mm²) in the control group. The authors of Sacu 2009 did not report results for this outcome.

Mean change in CRT

In the ABC 2010 study, the mean change in CRT at 54 weeks was ‐91 μm in the bevacizumab group and ‐55 μm in the control group (P value = 0.08). In Sacu 2009, the mean change in CRT at 12 months was ‐113 μm in the bevacizumab group and ‐72 μm in the control group (P value = 0.8; analysis of variance, ANOVA). No measures of variability were reported for these outcomes, precluding meta‐analysis.

(4) Quality‐of‐life outcomes

Quality‐of‐life outcomes were not assessed by the investigators of the ABC 2010 or Sacu 2009 trials.

(5) Economic outcomes

Economic outcomes were not assessed by the investigators of the ABC 2010 or Sacu 2009 trials.

(6) Adverse events

Serious ocular and non‐ocular adverse events occurring among 65 bevacizumab‐treated participants and 66 control participants were reported in the ABC 2010 study. Serious ocular events affecting at least one study participant included uveitis (two bevacizumab participants; one control participant), rhegmatogenous retinal detachment (no bevacizumab participants; one control participant), vitreous hemorrhage (one bevacizumab participant; no control participants), and ocular inflammation (eight bevacizumab participants; four control participants). No instances of presumed endophthalmitis, retinal tear, or lens damage were reported in either group. Three participants experienced a non‐ocular adverse event: myocardial infarction (bevacizumab group), death due to vascular cause (bevacizumab group), and non‐ocular hemorrhage reported as serious (control group).

The authors of Sacu 2009 reported no occurrences of severe ocular or systemic events during the study period.

Bevacizumab versus ranibizumab

Six non‐inferiority trials directly compared intravitreal bevacizumab with intravitreal ranibizumab. There were 2806 total participants in the six studies. In the largest study, 1208 participants were randomized in a 2 x 2 factorial design (two drugs administered in two dosing schedules) to receive 1.25 mg intravitreal bevacizumab or 0.5 mg intravitreal ranibizumab on a monthly or as‐needed basis (CATT 2011). Participants in the as‐needed dosing groups received the first three injections monthly; they then received an injection whenever treatment was needed based on monthly examinations. After one year of treatment, participants in the groups treated monthly were re‐randomized to continue treatment on a monthly basis or to change to treatment as needed. Participants in the as‐needed dosing groups remained on their original assignments and all participants were followed for another year. The IVAN 2013 study, with 628 participants, had four treatment groups similar to those in the CATT study: 1.25 mg intravitreal bevacizumab monthly, 0.5 mg intravitreal ranibizumab monthly, 1.25 mg intravitreal bevacizumab as needed, and 0.5 mg intravitreal ranibizumab as needed. Participants in the as‐needed dosing groups received the first three injections monthly; they then received three consecutive monthly treatments whenever treatment was needed. The treatment period was two years. In the four smaller studies, participants were randomized to receive 1.25 mg intravitreal bevacizumab or 0.5 mg intravitreal ranibizumab on an as‐needed basis for one year (GEFAL 2013; MANTA 2013; Subramanian 2010) or 18 months (Biswas 2011). We included the 18‐month data with the 12‐month data.

For the data analyses in this section, we combined groups of the same drug type regardless of dosing regimen. Thus the bevacizumab and ranibizumab groups include both monthly and as‐needed dosing schedules. Risk of bias in most domains was low among the studies and none of these studies was funded by pharmaceutical companies. At one‐year follow up, data for the primary outcome were reported for 2446 (87%) of 2806 participants.

(1) Visual acuity

(a) Gain of 15 letters or more of visual acuity

Overall, the proportion of participants who gained 15 letters or more of visual acuity at one year did not differ statistically significantly between bevacizumab‐ and ranibizumab‐treated groups (RR 0.90; 95% CI 0.73 to 1.11) (Analysis 4.1; Figure 6). Individual results for five of the six trials crossed unity and the I² statistic was 34%.

Figure 6

Forest plot of comparison: 4 Bevacizumab versus ranibizumab, outcome: 4.1 Gain of 15 letters or more visual acuity at one year.

At two years, data were available for 1030 (85%) of 1208 participants in the CATT 2011 trial and 517 (82%) of 628 participants in the IVAN 2013 trial. Results were consistent with one‐year outcomes in terms of the effect estimate and CIs when comparing the proportion of participants who gained 15 letters or more of visual acuity between ranibizumab‐ and bevacizumab‐treated groups (RR 0.84; 95% CI 0.64 to 1.11) (Analysis 4.2). When analyzing only the 778 participants who remained in their originally randomized groups in the CATT 2011 trial (i.e., excluding participants who were switched to a different treatment regimen after one year), summary results were unchanged (RR 0.84; 95% CI 0.64 to 1.11). The I² statistics for these analyses were at or about 50% indicating a difference in treatment effect between the CATT 2011 and IVAN 2013 trials.

(b) Loss of fewer than 15 letters of visual acuity

At one year of follow up, the overall RR for the loss of fewer than 15 letters of visual acuity was 1.00 (95% CI 0.98 to 1.02) when comparing participants treated with bevacizumab and those treated with ranibizumab (Analysis 4.3). The CIs for all six individual studies also crossed the line of unity. These results suggest that there was no clinical or statistical difference between the two drugs in terms of the loss of fewer than 15 letters of visual acuity after one year of treatment.

At two years of follow up, the relative treatment effect between the two drugs was almost identical to the relative effect at one year when analyzing participants based on their original randomization (RR 0.97; 95% CI 0.94 to 1.00) or participants who remained in their originally randomized groups (RR 0.98; 95% CI 0.94 to 1.01) (Analysis 4.4).

(c) Loss of fewer than 30 letters of visual acuity

No participant in the Subramanian 2010 study lost 30 letters or more of visual acuity during the one‐year study period. This outcome was not reported by the other five trials.

(d) Prevention of blindness (visual acuity better than 20/200)

Four trials reported the proportion of participants with visual acuity better than 20/200 as an outcome (CATT 2011; GEFAL 2013; IVAN 2013; Subramanian 2010) and two did not (Biswas 2011; MANTA 2013).

At one year of follow up, the proportion of participants with visual acuity better than 20/200 was neither clinically nor statistically significantly different when comparing participants treated with bevacizumab and participants treated with ranibizumab (RR 0.98; 95% CI 0.96 to 1.01) (Analysis 4.5). There was no statistical heterogeneity among studies (I² = 0%) and the CIs for these four individual studies all crossed the line of unity.

At two years, results were consistent with one‐year outcomes in that no significant difference in the proportion of participants with visual acuity better than 20/200 was observed between bevacizumab‐ and ranibizumab‐treated groups (RR 1.00; 95% CI 0.95 to 1.06) (Analysis 4.6). When analyzing only the 778 participants in the CATT 2011 trial who remained in their originally randomized groups, results were similar (RR 1.01; 95% CI 0.95 to 1.06). There was moderate statistical heterogeneity between the CATT 2011 and IVAN 2013 trials for these analyses (I² > 40%), which could be an artifact of the precision of the individual study effect estimates.

(e) Maintenance of visual acuity

Maintenance of visual acuity was not reported in the CATT 2011, GEFAL 2013, IVAN 2013, or MANTA 2013 trials. In Subramanian 2010, 10/15 (67%) participants maintained baseline visual acuity after one year of treatment with bevacizumab and 6/7 (86%) participants maintained baseline visual acuity after one year of treatment with ranibizumab (RR 0.78; 95% CI 0.49 to 1.24).

The investigators of Biswas 2011 reported different cut‐points for the change in visual acuity at 18 months follow up. In the bevacizumab group, 16 (32%) participants gained more than 5 letters, 30 (60%) participants did not change more than 5 letters, and 4 (8%) participants lost more than 5 letters of visual acuity. In the ranibizumab group, 18 (33%) participants gained more than 5 letters, 30 (56%) participants did not change more than 5 letters, and 6 (11%) participants lost more than 5 letters of visual acuity.

(f) Mean change in visual acuity

At one year, the mean difference in mean change in visual acuity between bevacizumab and ranibizumab groups was less than 1 ETDRS letter (MD ‐0.51, 95% CI ‐1.64 to 0.62) (Analysis 4.7). The CIs for all six individual studies crossed the line of no difference and the I² statistic was 0%.

Data for the mean change from baseline in visual acuity at two years were reported in the IVAN 2013 trial and for only the 778 participants who remained in their originally randomized groups in the CATT 2011 trial. The mean difference between bevacizumab and ranibizumab groups was less than 2 ETDRS letters (MD ‐1.15, 95% CI ‐2.82 to 0.51) (Analysis 4.8).

(2) Visual function

Only one of the six trials comparing bevacizumab with ranibizumab reported visual function outcomes (IVAN 2013). At one year, participants in the ranibizumab and bevacizumab groups were comparable in regard to mean letters of contrast sensitivity (adjusted MD 0.20; 95% CI ‐0.47 to 0.87) and reading index (MD ‐5.53; 95% CI ‐14.59 to 3.54). Participants in the ranibizumab groups had slightly better (8%) near LogMAR visual acuity compared with participants in the bevacizumab group (adjusted geometric mean ratio 0.92; 95% CI 0.84 to 1.00; P value = 0.058).

At two years of follow up, results for visual function outcomes were similar to those at one year. Participants in the ranibizumab and bevacizumab groups were comparable in regard to mean letters of contrast sensitivity (adjusted MD 0.21; 95% CI ‐0.62 to 1.04) and reading index (MD ‐1.34; 95% CI ‐8.29 to 5.61). Participants in the ranibizumab groups had slightly better (6%) near LogMAR visual acuity compared with participants in the bevacizumab group (adjusted geometric mean ratio 0.94; 95% CI 0.85 to 1.04).

(3) Morphological outcomes

Mean change in size of CNV

One study reported mean change in size of CNV from baseline (GEFAL 2013). At one year, there was no difference observed between bevacizumab (156 participants) and ranibizumab (144 participants) groups (MD 0.00 DAs; 95% CI ‐0.32 to 0.32).

Mean change in size of lesion

In two of the six studies, the outcome of change in size of total lesion was reported. We considered a difference of one or more DAs as a clinically meaningful difference.

In the CATT 2011 study, the mean change in size of lesion was similar in both the bevacizumab (479 participants) and ranibizumab (509 participants) groups at one year (MD 0.20 optic DAs; 95% CI ‐0.09 to 0.49). Among the 778 participants who remained in their originally randomized groups through two years, participants in the bevacizumab groups (341 participants) showed larger increases in lesion size compared with those in the ranibizumab groups (360 participants) (MD 1.37 mm²; 95% CI 0.39 to 2.36).

In the IVAN 2013 study, the median change in lesion size after one year of treatment was similar in both the bevacizumab (median ‐1.79 DAs; IQR ‐5.18 to 0.00) and ranibizumab (median ‐1.92 DAs; IQR ‐4.81 to ‐0.01) groups. After two years, the median change in size of lesion was ‐1.86 DAs (IQR ‐5.51 to 0.16) in the bevacizumab group and ‐0.96 DAs (IQR ‐4.29 to 0.39) in the ranibizumab group.

Mean change in size of lesion was not reported in Biswas 2011, GEFAL 2013, MANTA 2013, or Subramanian 2010. The authors of MANTA 2013 reported that "no significant difference was observed in terms of lesion size between the two groups (P = 0.55)".

Mean change in CRT

Five of the six trials reported mean change in CRT at one year. Participants treated with bevacizumab showed less reduction in CRT compared with participants treated with ranibizumab in four trials (MD ‐13.97 μm; 95% CI ‐26.52 to ‐1.41) (Analysis 4.9). This difference is not considered to be clinically meaningful as it falls within the typical range of measurement error. The authors of Subramanian 2010 reported a mean change of ‐50 μm in the bevacizumab group and ‐91 μm in the ranibizumab group at one year. Mean change in CRT was not reported as an outcome in the MANTA 2013 study reports; however, the investigators reported that "differences were not significant between the groups (P = 0.81)".

At two years, the trend was similar among the study participants who remained in their originally randomized groups in the CATT 2011 and IVAN 2013 trials. Participants in the bevacizumab groups showed less reduction in CRT compared with participants in the ranibizumab groups (MD ‐12.40 μm; 95% CI ‐33.83 to ‐9.04) (Analysis 4.10).

(4) Quality‐of‐life outcomes

One study (IVAN 2013) evaluated quality of life using the EQ‐5D.

At one‐year of follow up, the median (IQR) EQ‐5D summary score was the same for both the bevacizumab‐ and ranibizumab‐treated groups (median 0.85; IQR 0.73 to 1.00). The number of participants who reported "no health problems" for each of the five subscale domains was similar between groups (Analysis 4.11).

At two‐year follow up, the median (IQR) EQ‐5D summary score was the same as at the one year follow up (median 0.85; IQR 0.73 to 1.00) in both the bevacizumab‐ and ranibizumab‐treated groups. The number of participants who reported "no health problems" for each of the five subscale domains was similar in the two groups (Analysis 4.12).

Quality‐of‐life outcomes have not been reported from the remaining five studies (Biswas 2011; CATT 2011; GEFAL 2013; MANTA 2013; Subramanian 2010).

(5) Economic outcomes

Three studies included economic‐related outcomes as prespecified secondary outcomes. In the CATT 2011 study, the annual costs associated with each treatment group were evaluated in USD. In the IVAN 2013 study, cumulative resource use and costs for each treatment group were evaluated in GBP. In the GEFAL 2013 study, medicoeconomic outcomes were prespecified as secondary outcomes of interest; however, no results for economic outcomes were published with the one‐year results.

The average annual cost of treatment per participant was USD 490 in the bevacizumab groups (USD 595 when treated monthly and USD 385 when treated as needed) compared with USD 18,590 (USD 23,400 when treated monthly and USD 13,800 when treated as needed) in the ranibizumab groups in the first year of the CATT 2011 study. For the 778 participants who remained in their originally randomized groups, the average cost of two years of treatment was USD 860 per participant in the bevacizumab groups (USD 1170 when treated monthly and USD 705 when treated as needed) and USD 31,805 per participant in the ranibizumab groups (USD 44,800 when treated monthly and USD 25,200 when treated as needed).

In the IVAN 2013 study, the average total cost of treatment per participant for the first year was GBP 1580 in the bevacizumab groups (GBP 1654 when treated monthly and GBP 1509 when treated as needed) compared with GBP 8035 in the ranibizumab groups (GBP 9656 when treated monthly and GBP 6398 when treated as needed). These values corresponded to approximately USD 2500 and USD 12,700 for the bevacizumab and ranibizumab groups, respectively (based on an average exchange rate of 1.58 for years 2010 to 2011). The mean difference was GBP 8001 (SE 113) when comparing monthly treatment with ranibizumab versus bevacizumab, and GBP 4889 (SE 184) when comparing as‐needed treatment with ranibizumab versus bevacizumab. Economic outcomes at two years of follow up have not been reported for the IVAN 2013 study.

(6) Adverse events

Although all six trials reported information related to adverse events, there was variation in the types of adverse events reported among studies.

At one year, no serious ocular events were reported in three trials (Biswas 2011; MANTA 2013; Subramanian 2010). Minor adverse events reported in these three trials included subconjunctival hemorrhage, increased IOP, transient post‐injection pain, and mild ocular inflammation; the numbers of participants who experienced these adverse events were not reported. There were no cases of endophthalmitis or retinal detachment in these three trials. In the CATT 2011, GEFAL 2013, and IVAN 2013 studies, less than 1% of participants had endophthalmitis, retinal detachment, retinal pigment epithelial tear, traumatic cataract, or uveitis (Table 6). Because of the small number of events, risk estimates for these adverse events are imprecise.

Table 6. Adverse events up to one year: bevacizumab vs. ranibizumab

Serious ocular adverse event	Studies reporting adverse event*	Bevacizumab		Ranibizumab		RR [95% CI] Bevacizumab vs. ranibizumab
Serious ocular adverse event	Studies reporting adverse event*	Number with event	Total participants	Number with event	Total participants	RR [95% CI] Bevacizumab vs. ranibizumab
Endophthalmitis	CATT 2011; GEFAL 2013	4 (< 1%)	832	3 (< 1%)	838	1.34 [0.30, 5.98]
Retinal detachment	CATT 2011; GEFAL 2013	3 (< 1%)	832	0	838	7.05 [0.36, 136.28]
Retinal pigment epithelial tear	CATT 2011; IVAN 2013	3 (< 1%)	882	3 (< 1%)	913	1.04 [0.21, 5.11]
Traumatic cataract	CATT 2011; GEFAL 2013; IVAN 2013	1 (< 1%)	1128	2 (< 1%)	1152	0.51 [0.05, 5.62]
Severe uveitis	CATT 2011; IVAN 2013	4 (< 1%)	882	1 (< 1%)	913	4.14 [0.46, 36.97]
Non‐ocular adverse event	Studies reporting adverse event**	Bevacizumab		Ranibizumab		RR [95% CI] Bevacizumab vs. ranibizumab
Non‐ocular adverse event	Studies reporting adverse event**	Number with event	Total participants	Number with event	Total participants	RR [95% CI] Bevacizumab vs. ranibizumab
At least 1 serious adverse event	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	227 (18%)	1282	183 (14%)	1315	1.27 [1.06, 1.52]
Death	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	25 (2%)	1282	20 (2%)	1315	1.28 [0.72, 2.30]
Myocardial infarction	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	8 (< 1%)	1282	10 (< 1%)	1315	0.82 [0.32, 2.07]
Stroke or cerebral infarction	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	5 (< 1%)	1282	8 (< 1%)	1315	0.64 [0.21, 1.95]
Transient ischemic attack	CATT 2011; GEFAL 2013; IVAN 2013	4 (< 1%)	1128	4 (< 1%)	1152	1.02 [0.26, 4.07]
Venous thrombotic event	CATT 2011; GEFAL 2013; IVAN 2013	8 (< 1%)	1128	2 (< 1%)	1152	4.09 [0.87, 19.20]
Cardiac disorders	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	37 (3%)	1282	36 (3%)	1315	1.05 [0.67, 1.66]
Gastrointestinal disorders	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	24 (2%)	1282	11 (< 1%)	1315	2.24 [1.10, 4.55]
Infections	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	42 (3%)	1282	27 (2%)	1315	1.60 [0.99, 2.57]
Injury and procedural complications	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	30 (2%)	1282	21 (2%)	1315	1.47 [0.84, 2.55]
Neoplasms (benign, malignant, unspecified)	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	20 (2%)	1282	21 (2%)	1315	0.98 [0.53, 1.79]
Nervous system disorders	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	25 (2%)	1282	24 (2%)	1315	1.07 [0.61, 1.86]
Surgical or medical procedure	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	26 (2%)	1282	13 (1%)	1315	2.05 [1.06, 3.97]

CI: confidence interval

RR: risk ratio

* CATT 2011 (n = 586 in bevacizumab group; n = 599 in ranibizumab group); GEFAL 2013 (n = 246 in bevacizumab group; n = 239 in ranibizumab group); IVAN 2013 (n = 296 in bevacizumab group; n = 314 in ranibizumab group)
**CATT 2011 (n = 586 in bevacizumab group; n = 599 in ranibizumab group); GEFAL 2013 (n = 246 in bevacizumab group; n = 239 in ranibizumab group); IVAN 2013 (n = 296 in bevacizumab group; n = 314 in ranibizumab group); MANTA 2013 (n = 154 in bevacizumab group; n = 163 in ranibizumab group)

At one year, no serious systemic adverse events were reported in Subramanian 2010. Systemic adverse events were not assessed in Biswas 2011. In the remaining four trials (CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013), 18% of participants in the bevacizumab groups versus 14% of participants in the ranibizumab groups experienced at least one serious adverse event (RR 1.27; 95% CI 1.06 to 1.52). Mortality from any cause was approximately 2% in both the bevacizumab and ranibizumab groups in the first year of follow up (RR 1.28; 95% CI 0.72 to 2.30). Less than 1% of participants had a myocardial infarction, stroke or cerebral infarction, transient ischemic attack, or venous thrombotic event (Table 6). Rates were comparable between bevacizumab and ranibizumab groups with respect to cardiac disorders (RR 1.05; 95% CI 0.67 to 1.66), neoplasms (RR 0.98; 95% CI 0.53 to 1.79), and nervous system disorders (RR 1.07; 95% CI 0.61 to 1.86). There were more gastrointestinal disorders (RR 2.24; 95% CI 1.10 to 4.55), infections (RR 1.60; 95% CI 0.99 to 2.57), injuries and procedural complications (RR 1.47; 95% CI 0.84 to 2.55), and surgical or medical procedures (RR 2.05; 95% CI 1.06 to 3.97) reported in the bevacizumab groups compared with ranibizumab groups at one year.

At two years, data for ocular and systemic adverse events were available for the CATT 2011 and IVAN 2013 trials. Less than 1% of participants had endophthalmitis, retinal detachment, retinal pigment epithelial tear, traumatic cataract, or uveitis (Table 7). Because of the small number of events, risk estimates for these adverse events are imprecise. In the bevacizumab groups, 36% of participants had at least one serious adverse event compared with 30% in the ranibizumab groups (RR 1.20; 95% CI 1.05 to 1.37). Mortality from any cause was 6% and 5% in the bevacizumab and ranibizumab groups, respectively (RR 1.12; 95% CI 0.76 to 1.65). There were 2% or fewer participants with myocardial infarction, stroke or cerebral infarction, venous thrombotic event, or transient ischemic attack (Table 7). As with one‐year outcomes, more gastrointestinal disorders (RR 2.74, 95% CI 1.49 to 5.02), infections (RR 1.37, 95% CI 0.96 to 1.95), and injuries and procedural complications (RR 1.33, 95% CI 0.86 to 2.05) were reported in the bevacizumab groups compared with ranibizumab groups. More cardiac disorders were also reported in the bevacizumab groups compared with ranibizumab groups at two years (RR 1.25; 95% CI 0.92 to 1.71). Rates were comparable between bevacizumab and ranibizumab groups with respect to neoplasms (RR 0.98; 95% CI 0.63 to 1.53), nervous system disorders (RR 1.06; 95% CI 0.70 to 1.60), and surgical or medical procedures (RR 0.91; 95% CI 0.44 to 1.84).

Table 7. Adverse events up to two years: bevacizumab vs. ranibizumab

Ocular adverse event (CATT trial)*	Bevacizumab n = 586	Ranibizumab n = 599	RR [95% CI] Bevacizumab vs. ranibizumab
Endophthalmitis	7 (1%)	4 (< 1%)	1.79 [0.53, 6.08]
Ocular adverse event (IVAN trial)**	Bevacizumab n = 296	Ranibizumab n = 314	RR [95% CI] Bevacizumab vs. ranibizumab
Traumatic cataract	1 (< 1%)	1 (< 1%)	1.06 [0.07, 16.88]
Severe uveitis	1 (< 1%)	0	3.18 [0.13, 77.80]
Retinal detachment	0	1 (< 1%)	0.35 [0.01, 8.64]
Retinal pigment epithelial tear	1 (< 1%)	3 (< 1%)	0.35 [0.04, 3.38]
Non‐ocular adverse event†	Bevacizumab n = 882	Ranibizumab n = 913	RR [95% CI] Bevacizumab vs. ranibizumab
At least 1 serious adverse event	314 (36%)	271 (30%)	1.20 [1.05, 1.37]
Death	51 (6%)	47 (5%)	1.12 [0.76, 1.65]
Myocardial infarction	11 (1%)	13 (1%)	0.88 [0.39, 1.94]
Stroke or cerebral infarction	11 (1%)	14 (2%)	0.81 [0.37, 1.78]
Venous thrombotic event	14 (2%)	6 (< 1%)	2.42 [0.93, 6.26]
Transient ischemic attack**	1 (< 1%)	1 (< 1%)	1.04 [0.06, 16.52]
Cardiac disorders	81 (9%)	67 (7%)	1.25 [0.92, 1.71]
Gastrointestinal disorders	37 (4%)	14 (2%)	2.74 [1.49, 5.02]
Infections	66 (7%)	50 (5%)	1.37 [0.96, 1.95]
Injury and procedural complications	45 (5%)	35 (4%)	1.33 [0.86, 2.05]
Neoplasms (benign, malignant, unspecified)	36 (4%)	38 (4%)	0.98 [0.63, 1.53]
Nervous system disorders	44 (5%)	43 (5%)	1.06 [0.70, 1.60]
Surgical or medical procedure**	14 (5%)	16 (5%)	0.91 [0.44, 1.84]

CI: confidence interval

RR: risk ratio

*Adverse events for endophthalmitis not reported in the IVAN 2013 study; data for CATT 2011 study only
**Adverse events for traumatic cataract, uveitis, retinal detachment, retinal pigment epithelial tear, transient ischemic attack, and surgical or medical procedure not reported in the CATT 2011 study; data for IVAN 2013 study only
†Adverse events experienced by 1185 participants in the CATT 2011 study and 610 participants in the IVAN 2013 study

Discussion

Summary of main results

All twelve trials included in this systematic review were of good methodological quality and demonstrated the beneficial effect of anti‐VEGF therapy on visual acuity in the management of neovascular AMD. Participants treated with any one of the anti‐VEGF agents featured in these trials, pegaptanib (one trial), ranibizumab (three trials), or bevacizumab (two trials), more often maintained visual acuity at one year and less often lost visual acuity, compared with participants who received no anti‐VEGF agent. Stability of visual acuity at one year was more often achieved in an anti‐VEGF treatment group than in a control group not treated with anti‐VEGFs. The safety profile of anti‐VEGFs was acceptable based on the information reported in the included studies.

Functional outcomes (e.g., visual acuity) correlated with quality‐of‐life outcomes, when reported, and anatomic outcomes (e.g., lesion size and retinal thickening) across trials. Participants treated with pegaptanib showed a decrease in size of the choroidal neovascular complex with less leakage observed on fluorescein angiograms compared with participants treated with sham injections. In bevacizumab‐treated participants, there was a reduction in CRT on OCT compared with participants in the control groups.

Improvement in vision‐specific quality of life was reported more often in the anti‐VEGF‐treated groups compared to the control groups. Improved scores on the NEI‐VFQ scale were reported with both pegaptanib and ranibizumab compared to controls. Cost utility analysis, based on data from one trial and with standardized utilities of degree of visual loss, which compared ranibizumab with pegaptanib found ranibizumab to be associated with a better quality of life when compared with pegaptanib (Brown 2008). Data on visual function (e.g., contrast sensitivity) costs were sparse in these trials.

We found no trial in which pegaptanib had been compared with another anti‐VEGF agent head‐to‐head. Six head‐to‐head trials compared bevacizumab versus ranibizumab. At one and two years of follow up, differences between bevacizumab and ranibizumab for visual acuity outcomes were comparable clinically and statistically, although CIs for some outcomes reported by individual studies indicate some uncertainty in the true effects. In terms of visual function, one trial showed better near LogMAR visual acuity among participants in the ranibizumab groups than among participants in the bevacizumab groups at one‐year follow up; this effect had diminished at two‐years follow up. At one and two years of follow up, there were no clinically meaningful differences in the reduction of CRT between bevacizumab‐treated participants and ranibizumab‐treated participants. Participant responses to quality‐of‐life questionnaires were comparable between the two treatment groups. A small number of ocular adverse events were reported for both bevacizumab or ranibizumab (fewer than 1%) across all trials. However, endophthalmitis rates were higher with injection of anti‐VEGF agents than with intravitreal surgery unless estimates were based on number of injections given rather than number of eyes treated. It is important that individuals with AMD and their ophthalmologists be aware of this small, but serious risk. At both one and two‐year follow ups, fewer participants in the ranibizumab groups experienced any serious systemic adverse event compared with participants in the bevacizumab groups.

Overall completeness and applicability of evidence

The aim of this review was to investigate both the effects and quality of life associated with intravitreally injected anti‐VEGF agents for the treatment of neovascular AMD when compared to either sham treatment or a different anti‐VEGF treatment administered at comparable dosages and regimens. Only RCTs were included in this review, each with a minimum follow up of one year. The primary outcome for this review was the proportion of participants who gained 15 letters or more of BCVA by the one‐year follow‐up examination. Secondary outcomes included other visual acuity outcomes at one and two years of follow up, visual function outcomes, morphological characteristics assessed by fluorescein angiography or OCT, ocular and systemic adverse outcomes, cost outcomes, and quality‐of‐life measures. Multiple sources were used to identify relevant data for this review, not only journal publications, but also conference abstracts, FDA documents, and clinical trial registries. When data were unclear or missing, study investigators were contacted for clarification or information.

This review ultimately included representative and applicable outcomes data on 5496 participants from 12 trials conducted in various countries that included both men and women aged 50 years or older with subfoveal CNV secondary to AMD. Approximately half of the trials reported the type of neovascular lesion, with all lesion types (predominantly classic CNV, minimally classic CNV, and occult CNV only) represented among these trials. All studies included at least one measure related to the morphological characteristics of study eyes, with fluorescein angiography used in all studies and OCT used in all but the earliest of these 12 RCTs.

The initial RCTs of anti‐VEGF agents incorporated in this review, both individually and collectively, established a new paradigm for the management of neovascular AMD, particularly for lesions under or near the central fovea, and validated the administration of intravitreal anti‐VEGF therapy in affected individuals with clinical profiles similar to those of the participants enrolled in these trials. Reported outcomes related to visual acuity gains, the stability of visual acuity at one year, the decreased risk of significant visual acuity loss, and the low rates of ocular and systemic adverse events are mirrored in real‐life clinical encounters when anti‐VEGF agents are used to manage neovascular AMD in the retina specialist's office (Carneiro 2012; Gillies 2014; Holz 2013; Rasmussen 2014). As observed in the clinical trials incorporated into this review, morphologic changes in the CNV lesion complex, with regard to decreased size on fluorescein angiography, decreased leakage on fluorescein angiography, and decreased CRT on OCT, are also observed to occur in‐office in individuals receiving anti‐VEGF therapy for neovascular AMD (Carneiro 2012).

With completion of head‐to‐head trials of bevacizumab versus ranibizumab, and the finding of little or no difference in outcomes between the two drugs, a major challenge for the ophthalmologist and individual with AMD has been the choice of anti‐VEGF agent. Issues considered have been costs, availability, and quality control of the preparation of bevacizumab for intravitreal injection. Issues as yet unresolved are the optimal frequency with which anti‐VEGF agents should be injected in most affected eyes, the length of calendar time over which anti‐VEGF agents must be injected to maintain the benefits seen with two‐year outcomes, and the long‐term ocular and systemic effects of these treatments.

Quality of the evidence

In addition to the inclusion of only RCTs in this review, two review authors assessed potential sources of bias in these trials according to methods established by The Cochrane Collaboration. Parameters considered included selection bias, performance bias, detection bias, attrition bias, and reporting bias; each potential source of bias was graded as low risk, unclear risk, or high risk. Overall, the included studies were found to be at low risk for all categories of bias. In all 12 trials, few participants missed the primary outcome visit or were not treated per‐protocol assignment. In nine trials, the rates of loss to follow up at the primary follow‐up visits were less than 15%. Athough not the best method to account for missing data, six trials used the last‐observation‐carried–forward method to impute missing data. Protocols or clinical trial registrations were identified for 11 of the 12 included studies. Seven of these 11 trials were judged to be free of reporting bias based on the consistency between study outcomes defined in the protocols and clinical trial registers and those reported in study publications to date.

Various other aspects of trial design, reporting, and financial support were considered as potential sources of bias. Four of 12 trials, one study of pegaptanib and three studies comparing ranibizumab with controls, were sponsored by pharmaceutical companies that marketed the study drugs under investigation. In addition, the pharmaceutical company sponsors had important roles in the design, analysis, and reporting of these trials and some of the investigators reported that they had financial relationships with the company that manufactured the study drug.

Potential biases in the review process

For this review, we conducted broad electronic searches for studies and imposed no date or language restrictions in the searches in order to minimize potential biases in the study selection process. We followed standard Cochrane review methodology.

Agreements and disagreements with other studies or reviews

Whether assessed by systematic, comprehensive reviews, like this one, or by more traditional, clinical reviews, treatments for neovascular AMD with anti‐VEGF compounds appear to be efficacious and safe (Ip 2008; Mitchell 2011; Schmucker 2010; Schmucker 2012). Beneficial effects with pegaptanib, ranibizumab, and bevacizumab are evident in terms of the proportion of participants with stabilization or small losses of BCVA. Ranibizumab and bevacizumab additionally resulted in a greater proportion of participants with improved BCVA after one and two years of injections. In independent studies and comprehensive reviews, visual acuity effects have been consistent with morphologic changes in the size and composition of the CNV lesion complex as well as with the observed change in CRT on OCT following treatment with these agents. In general, considerations of costs were limited in the trials included in this systematic review; additional analyses indicating a favorable cost utility ratio for anti‐VEGF agents versus control or no treatment were cited in research using RCT and observational data (Cohen 2008; Earnshaw 2007; Fletcher 2008; Hernandez‐Pastor 2008; Javitt 2008; Wolowacz 2007). Economic analyses have documented the lower cost of bevacizumab compared with ranibizumab to achieve the same benefits (Raftery 2007; Stein 2014). A separate Cochrane review specifically evaluating the systemic safety of bevacizumab versus ranibizumab also concluded no significant differences between intravitreal injection of the two drugs after two years of follow up with respect to deaths or overall serious systemic adverse events (Moja 2014).

Figure 1

Study flow diagram. Results of searches as of 27 March 2014.

Figure 2

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Figure 3

Forest plot of comparison: 1 Pegaptanib versus control, outcome: 1.1 Gain of 15 letters or more visual acuity at one year.

Figure 4

Forest plot of comparison: 2 Ranibizumab versus control, outcome: 2.1 Gain of 15 letters or more visual acuity at one year.

Figure 5

Forest plot of comparison: 3 Bevacizumab versus control, outcome: 3.1 Gain of 15 letters or more visual acuity at one year.

Figure 6

Forest plot of comparison: 4 Bevacizumab versus ranibizumab, outcome: 4.1 Gain of 15 letters or more visual acuity at one year.

Analysis 1.1

Comparison 1 Pegaptanib versus control, Outcome 1 Gain of 15 letters or more visual acuity at one year.

Analysis 1.2

Comparison 1 Pegaptanib versus control, Outcome 2 Loss of fewer than 15 letters visual acuity at one year.

Analysis 1.3

Comparison 1 Pegaptanib versus control, Outcome 3 Loss of fewer than 30 letters visual acuity at one year.

Analysis 1.4

Comparison 1 Pegaptanib versus control, Outcome 4 Visual acuity better than 20/200 at one year.

Analysis 1.5

Comparison 1 Pegaptanib versus control, Outcome 5 Maintenance of visual acuity at one year.

Analysis 1.6

Comparison 1 Pegaptanib versus control, Outcome 6 Mean change in visual acuity at one year (number of letters).

Analysis 1.7

Comparison 1 Pegaptanib versus control, Outcome 7 Reduction in size of CNV at one year (Mean number of disc areas).

Analysis 1.8

Comparison 1 Pegaptanib versus control, Outcome 8 Reduction in size of lesion at one year (Mean number of disc areas).

Analysis 2.1

Comparison 2 Ranibizumab versus control, Outcome 1 Gain of 15 letters or more visual acuity at one year.

Analysis 2.2

Comparison 2 Ranibizumab versus control, Outcome 2 Gain of 15 letters or more visual acuity at two years.

Analysis 2.3

Comparison 2 Ranibizumab versus control, Outcome 3 Loss of fewer than 15 letters visual acuity at one year.

Analysis 2.4

Comparison 2 Ranibizumab versus control, Outcome 4 Loss of fewer than 15 letters visual acuity at two years.

Analysis 2.5

Comparison 2 Ranibizumab versus control, Outcome 5 Loss of fewer than 30 letters visual acuity at one year.

Analysis 2.6

Comparison 2 Ranibizumab versus control, Outcome 6 Loss of fewer than 30 letters visual acuity at two years.

Analysis 2.7

Comparison 2 Ranibizumab versus control, Outcome 7 Visual acuity better than 20/200 at one year.

Analysis 2.8

Comparison 2 Ranibizumab versus control, Outcome 8 Visual acuity better than 20/200 at two years.

Analysis 2.9

Comparison 2 Ranibizumab versus control, Outcome 9 Maintenance of visual acuity at one year.

Analysis 2.10

Comparison 2 Ranibizumab versus control, Outcome 10 Maintenance of visual acuity at two years.

Analysis 2.11

Comparison 2 Ranibizumab versus control, Outcome 11 Mean change in visual acuity at one year (number of letters).

Analysis 2.12

Comparison 2 Ranibizumab versus control, Outcome 12 Mean change in visual acuity at two years (number of letters).

Analysis 2.13

Comparison 2 Ranibizumab versus control, Outcome 13 Reduction in size of lesion at one year (Mean number of disc areas).

Analysis 2.14

Comparison 2 Ranibizumab versus control, Outcome 14 Reduction in size of lesion at two years (Mean number of disc areas).

Analysis 2.15

Comparison 2 Ranibizumab versus control, Outcome 15 Mean change in quality of life scores at one year.

Analysis 2.16

Comparison 2 Ranibizumab versus control, Outcome 16 Mean change in quality of life scores at two years.

Analysis 3.1

Comparison 3 Bevacizumab versus control, Outcome 1 Gain of 15 letters or more visual acuity at one year.

Analysis 3.2

Comparison 3 Bevacizumab versus control, Outcome 2 Loss of fewer than 15 letters visual acuity at one year.

Analysis 4.1

Comparison 4 Bevacizumab versus ranibizumab, Outcome 1 Gain of 15 letters or more visual acuity at one year.

Analysis 4.2

Comparison 4 Bevacizumab versus ranibizumab, Outcome 2 Gain of 15 letters or more visual acuity at two years.

Analysis 4.3

Comparison 4 Bevacizumab versus ranibizumab, Outcome 3 Loss of fewer than 15 letters visual acuity at one year.

Analysis 4.4

Comparison 4 Bevacizumab versus ranibizumab, Outcome 4 Loss of fewer than 15 letters visual acuity at two years.

Analysis 4.5

Comparison 4 Bevacizumab versus ranibizumab, Outcome 5 Visual acuity better than 20/200 at one year.

Analysis 4.6

Comparison 4 Bevacizumab versus ranibizumab, Outcome 6 Visual acuity better than 20/200 at two years.

Analysis 4.7

Comparison 4 Bevacizumab versus ranibizumab, Outcome 7 Mean change in visual acuity at one year (number of letters).

Analysis 4.8

Comparison 4 Bevacizumab versus ranibizumab, Outcome 8 Mean change in visual acuity at two years (number of letters).

Analysis 4.9

Comparison 4 Bevacizumab versus ranibizumab, Outcome 9 Reduction in central retinal thickness at one year.

Analysis 4.10

Comparison 4 Bevacizumab versus ranibizumab, Outcome 10 Reduction in central retinal thickness at two years.

Analysis 4.11

Comparison 4 Bevacizumab versus ranibizumab, Outcome 11 No problems in quality of life domain at one year.

Analysis 4.12

Comparison 4 Bevacizumab versus ranibizumab, Outcome 12 No problems in quality of life domain at two years.

Summary of findings for the main comparison. Summary of findings: bevacizumab versus ranibizumab

Bevacizumab compared with ranibizumab for neovascular age‐related macular degeneration
Participant or population: people with neovascular age‐related macular degeneration Settings: clinical centers Intervention: intravitreal injections of bevacizumab Comparison: intravitreal injections of ranibizumab
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No. of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Ranibizumab	Bevacizumab
Gain of 15 letters or more visual acuity at one year	257 per 1000	231 per 1000 (188 to 285)	RR 0.90 (0.73 to 1.11)	2446 (6)	⊕⊕⊕⊕ high
Loss of fewer than 15 letters visual acuity at one year	942 per 1000	942 per 1000 (923 to 960)	RR 1.00 (0.98 to 1.02)	2446 (6)	⊕⊕⊕⊕ high
Mean change in visual acuity at one year (number of letters)	The mean change across ranibizumab groups ranged from gains of 3 to 8 letters	The mean change in visual acuity in the bevacizumab groups was on average 0.51 fewer letters gained (95% CI 1.64 fewer letters to 0.62 more letters)	MD ‐0.51 (‐1.64 to 0.62)	2446 (6)	⊕⊕⊕⊕ high
Reduction in central retinal thickness at one year	The mean reduction in central retinal thickness across ranibizumab groups ranged from 30 to 182 μm	The mean reduction in central retinal thickness in the bevacizumab groups was on average 13.97 μm less (95% CI 26.52 less to 1.41 less)	MD ‐13.97 (‐26.52 to ‐1.41)	1995 (4)	⊕⊕⊕⊕ high	Two additional trials reported no difference between groups for this outcome; however, these data were not reported in formats that could be included in meta‐analysis
No problems in quality of life domains at one year	Range of 591 per 1000 to 861 per 1000 across five quality of life domains	Range of 608 per 1000 to 828 per 1000 across five quality of life domains	Range of RRs 0.96 (0.90 to 1.04) to 1.02 (0.89 to 1.17)	548 (1)	⊕⊕⊕⊝ moderate¹	Quality of life domains included: mobility, self care, usual, activities, pain/discomfort, anxiety/depression
Serious systemic adverse events at one year²	139 per 1000 with at least one serious systemic adverse event	177 per 1000 (148 to 212)	RR 1.27 (1.06 to 1.52)	2597 (4)	⊕⊕⊕⊝ moderate¹
Serious ocular adverse events at one year	< 5 per 1000	< 5 per 1000	Range of RRs 0.51 (0.05 to 5.62) to 7.05 (0.36 to 136.28)	Range 1670 to 2280 (2 to 3)	⊕⊕⊕⊝ moderate¹	Studies reported different ocular adverse events
The basis for the assumed risk* is estimated by the proportion with the event in the ranibizumab group. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; MD: mean difference
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group grades of evidence: High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. ¹ Quality of life and adverse event outcomes downgraded to moderate quality as not all eligible trials reported these outcomes and numbers of some adverse events were small (< 1%). ² A new Cochrane review on systemic safety of bevacizumab versus ranibizumab (Moja 2014) includes more up‐to‐date data for this finding, from trials listed as ongoing in this review. Please refer to Moja 2014 for the most up‐to‐date information on systemic safety for bevacizumab versus ranibizumab.

Summary of findings for the main comparison. Summary of findings: bevacizumab versus ranibizumab

Summary of findings 2. Summary of findings: pegaptanib versus control

Pegaptanib compared with control for neovascular age‐related macular degeneration
Participant or population: people with neovascular age‐related macular degeneration Settings: clinical centers Intervention: intravitreal injections of pegaptanib Comparison: sham injections
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No. of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Pegaptanib
Gain of 15 letters or more visual acuity at one year	20 per 1000	57 per 1000 (25 to 132)	RR 2.83 (1.23 to 6.52)	1186 (1)	⊕⊕⊕⊕ high
Loss of fewer than 15 letters visual acuity at one year	554 per 1000	687 per 1000 (615 to 770)	RR 1.24 (1.11 to 1.39)	1186 (1)	⊕⊕⊕⊕ high
Mean change in visual acuity at one year (number of letters)	The mean change in the control group was a loss of 15 letters	The mean change in visual acuity in the pegaptanib groups was on average 6.72 more letters gained (95% CI 4.43 letters to 9.01 letters)	MD 6.72 (4.43 to 9.01)	1186 (1)	⊕⊕⊕⊕ high
Reduction in central retinal thickness at one year	‐	‐	‐	‐	‐	Outcome not assessed by this trial.
No problems in quality of life domains at one year	‐	‐	‐	‐	‐	Treatment with pegaptanib was associated with better scores on the NEI‐VFQ questionnaire, specifically for distance vision and role limitation domains; however, standard deviations for scores were not reported.
Serious systemic adverse events at one year	151 per 1000 with at least one serious systemic adverse event	189 per 1000 (140 to 257)	RR 1.25 (0.93 to 1.70)	1190 (1)	⊕⊕⊕⊝ moderate¹
Serious ocular adverse events at one year	7 per 1000 with any eye disorder	26 per 1000 (6 to 109)	RR 3.84 (0.91 to 16.20)	1190 (1)	⊕⊕⊕⊝ moderate¹
The basis for the assumed risk* is estimated by the proportion with the event in the control group. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; MD: mean difference
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. ¹ Adverse events downgraded to moderate quality as the numbers of events were small (< 1%) for many specific adverse events.

Summary of findings 2. Summary of findings: pegaptanib versus control

Summary of findings 3. Summary of findings: ranibizumab versus control

Ranibizumab compared with control for neovascular age‐related macular degeneration
Participant or population: people with neovascular age‐related macular degeneration Settings: clinical centers Intervention: intravitreal injections of ranibizumab Comparison: sham injections with or without verteporfin photodynamic therapy
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No. of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Ranibizumab
Gain of 15 letters or more visual acuity at one year	59 per 1000	230 per 1000 (93 to 566)	see comment	1322 (3)	⊕⊕⊕⊝ moderate¹	Meta‐analysis not performed due to high I² (80%).
Loss of fewer than 15 letters visual acuity at one year	610 per 1000	934 per 1000 (861 to 1000)	RR 1.53 (1.41 to 1.64)	1322 (3)	⊕⊕⊕⊕ high
Mean change in visual acuity at one year (number of letters)	The mean change across control groups ranged from a loss of 10 to 16 letters	The mean change in visual acuity in the ranibizumab groups was on average 17.80 more letters gained (95% CI 15.95 letters to 19.65 letters)	MD 17.80 (15.95 to 19.65)	1322 (3)	⊕⊕⊕⊕ high
Reduction in central retinal thickness at one year	‐	‐	‐	‐	‐	We were unable to find data on central retinal thickness in reports from any of the three included trials comparing ranibizumab with control interventions.
Mean change in vision‐related quality of life	The mean change across control groups in vision‐related quality of life scores ranged from ‐3 to 2 points	The mean change across control groups in vision‐related quality of life scores ranged from 5 to 7 points	MD 6.69 (3.38 to 9.99)	1134 (2)	⊕⊕⊕⊝ moderate²	Using the NEI‐VFQ questionnaire with a 10‐point difference considered as being clinically meaningful.
Serious systemic adverse events at one year	Range of 5 to 83 per 1000 for various systemic adverse events	Range of 0 to 55 per 1000 for various systemic adverse events	Range of RRs 0.17 (0.01 to 4.24) to 2.08 (0.23 to 18.45)	603 (2)	⊕⊕⊕⊝ moderate³
Serious ocular adverse events at one year	Range of 0 to 68 per 1000 for various ocular adverse events	Range of 3 to 118 per 1000 for various ocular adverse events	Range of RRs 0.52 (0.03 to 8.25) to 2.71 (1.36 to 5.42)	603 (2)	⊕⊕⊕⊝ moderate³
The basis for the assumed risk* is estimated by the proportion with the event in the control group. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; MD: mean difference
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. ¹ Gain of vision outcome downgraded due to high statistical heterogeneity. ² Quality of life outcomes downgraded due to not all studies reporting this outcome and non‐clinically significant results. ³ Adverse events downgraded to moderate quality as not all eligible trials reported all types of adverse events and numbers were small (<1%) for many specific adverse events.

Summary of findings 3. Summary of findings: ranibizumab versus control

Summary of findings 4. Summary of findings: bevacizumab versus control

Bevacizumab compared with control for neovascular age‐related macular degeneration
Participant or population: people with neovascular age‐related macular degeneration Settings: clinical centers Intervention: intravitreal injections of bevacizumab Comparison: standard therapy (intravitreal injections of pegaptanib, verteporfin photodynamic therapy with or without triamcinolone acetonide, or sham injections)
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No. of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Bevacizumab
Gain of 15 letters or more visual acuity at one year	38 per 1000	293 per 1000 (92 to 937)	RR 7.80 (2.44 to 24.98)	159 (2)	⊕⊕⊕⊝ moderate¹
Loss of fewer than 15 letters visual acuity at one year	700 per 1000	896 per 1000 (763 to 1000)	RR 1.28 (1.09 to 1.50)	159 (2)	⊕⊕⊕⊝ moderate¹
Mean change in visual acuity at one year (number of letters)	‐	‐	‐	‐	‐	The mean change from baseline in visual acuity was 7.0 letters in the bevacizumab group and ‐9.4 letters in the control group in one study. The second study reported participants in the bevacizumab group gained 8 letters on average and participants in the control group lost 3 letters on average.
Reduction in central retinal thickness at one year	‐	‐	‐	‐	‐	The mean change was ‐91 μm in the bevacizumab group and ‐55 μm in the control group in one study and ‐113 μm in the bevacizumab group and ‐72 μm in the control group in the other study.
Mean change in vision‐related quality of life	‐	‐	‐	‐	‐	Outcome not assessed by these trials.
Serious systemic adverse events at one year	15 per 1000 experienced any systemic adverse event	31 per 1000 (3 to 331)	RR 2.03 (0.19 to 21.85)	131 (1)	⊕⊕⊝⊝ low²
Serious ocular adverse events at one year	91 per 1000 experienced any ocular adverse event	169 per 1000 (66 to 431)	RR 1.86 (0.73 to 4.74)	131 (1)	⊕⊕⊝⊝ low²
The basis for the assumed risk* is estimated by the proportion with the event in the control group. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; MD: mean difference
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. ¹ Vision outcomes downgraded to moderate quality due to small sample sizes. ² Adverse events downgraded to low quality as sample sizes were small and numbers of events were small (< 1%) for many specific adverse events.

Summary of findings 4. Summary of findings: bevacizumab versus control

Table 1. Table of Study Acronyms

Acronym	Details
*Included studies*
ABC	Avastin^® (Bevacizumab) in Choroidal Neovascularization Trial
ANCHOR	Anti‐VEGF Antibody for the Treatment of Predominantly Classic Choroidal Neovascularization in Age‐related Macular Degeneration
CATT	Comparison of Age‐related macular degeneration Treatment Trials
GEFAL	French Evaluation Group Avastin® Versus Lucentis^®
IVAN	A randomized controlled trial of alternative treatments to Inhibit VEGF in Age‐related choroidal Neovascularisation
MANTA	A Randomized Observer and Subject Masked Trial Comparing the Visual Outcome After Treatment With Ranibizumab or Bevacizumab in Patients With Neovascular Age‐related Macular Degeneration Multicenter Anti VEGF Trial in Austria
MARINA	Minimally Classic/Occult Trial of the Anti‐VEGF Antibody Ranibizumab in the Treatment of Neovascular Age‐Related Macular Degeneration
PIER	A Phase IIIb, Multicenter, Randomized, Double‐Masked, Sham Injection‐Controlled Study of the Efficacy and Safety of Ranibizumab in Subjects with Subfoveal Choroidal Neovascularization with or without Classic CNV Secondary to Age‐Related Macular Degeneration
VISION	VEGF Inhibition Study in Ocular Neovascularization
*Ongoing studies*
BRAMD	Comparison of Bevacizumab (Avastin^®) and Ranibizumab (Lucentis®) in Exudative Age‐related Macular Degeneration
LUCAS	Lucentis® Compared to Avastin^® Study
MAAM	Avastin® and Macugen® Versus Avastin^® Versus Macugen^®
RATE	Ranibizumab and the Risk of Arterial Thromboembolic Events
VIBERA	Prevention of Vision Loss in Patients With Age‐Related Macular Degeneration by Intravitreal Injection of Bevacizumab and Ranibizumab
Other studies evaluating anti‐VEGF therapies for AMD*
ADVANCE	Safety and Efficacy of Oral PTK787 in Patients With Subfoveal Choroidal Neovascularization Secondary to Age‐Related Macular Degeneration (NCT00138632)
ARMAST	Photodynamic Therapy Combined With Bevacizumab vs Bevacizumab Alone for Neovascular Age‐Related Macular Degeneration (NCT00696592)
ATLAS	Repeated Eye Injections of Aflibercept for Treatment of Wet Age Related Macular Degeneration (NCT01773954)
BEAT‐AMD	Systemic Avastin Therapy in Age‐Related Macular Degeneration (NCT00531024)
BeMOC	Randomised controlled trial of bevacizumab in choroidal neovascularisation secondary to age related macular degeneration (ISRCTN12980412)
CARBON	Safety & Efficacy Study Evaluating the Combination of Bevasiranib & Lucentis Therapy in Wet AMD (NCT00557791)
CLOVER	Combination Lucentis and Ocular Photodynamic Therapy With Visudyne, With Evaluation‐based Retreatment (NCT00680498)
COBALT	Safety & Efficacy Study Evaluating the Combination of Bevasiranib & Lucentis Therapy in Wet AMD (NCT00499590)
DENALI	Efficacy/Safety of Verteporfin Photodynamic Therapy and Ranibizumab Compared With Ranibizumab in Patients With Subfoveal Choroidal Neovascularization (NCT00436553)
EVEREST	Efficacy and Safety of Verteporfin Added to Ranibizumab in the Treatment of Symptomatic Macular Polypoidal Choroidal Vasculopathy (NCT00674323)
EXCITE	Efficacy and Safety of Ranibizumab in Patients With Subfoveal Choroidal Neovascularization Secondary to Age‐related Macular Degeneration (NCT00275821)
EXTEND‐I/II/III	Efficacy and Safety of Ranibizumab in Patients With Subfoveal Choroidal Neovascularization Secondary to Age‐related Macular Degeneration (NCT00826371; NCT00470678)
FOCUS	RhuFab V2 Ocular Treatment Combining the Use of Visudyne^® to Evaluate Safety (NCT00056823)
GMAN	Greater Manchester Avastin® for choroidal Neovascularisation trial (ISRCTN34221234)
HARBOR	A Study of Ranibizumab Administered Monthly or on an As‐needed Basis in Patients With Subfoveal Neovascular Age‐related Macular Degeneration (NCT00891735)
HORIZON	An Open‐Label Extension Trial of Ranibizumab for Choroidal Neovascularization Secondary to Age‐Related Macular Degeneration (cohort of participants who completed the MARINA, ANCHOR, or FOCUS trials)
LAST	A Pilot Study to evaLuate the Role of High‐dose rAnbizumab (2.0mg) in the Management of AMD in Patients With perSistent/recurrenT Macular Fluid Less Than 30 Days Following Treatment With Intravitreal Anti‐VEGF Therapy (NCT01115556)
LOW‐VISION	Intravitreal Bevacizumab for Low Vision in Neovascular Age‐related Macular Degeneration (NCT01327222)
LUV	Lucentis Utilizing Visudyne Combination Therapy in the Treatment of Age‐Related Macular Degeneration (NCT00423189)
MERLOT	Macular EpiRetinal Brachytherapy Versus Lucentis® Only Treatment (NCT01006538)
MONET	Phase II Open Label Multicenter Study For Age Related Macular Degeneration Comparing PF‐04523655 Versus Lucentis In The Treatment Of Subjects With CNV (NCT00713518)
MONT BLANC	Verteporfin Photodynamic Therapy Administered in Conjunction With Ranibizumab in Patients With Subfoveal Choroidal Neovascularization Secondary to Age‐related Macular Degeneration (NCT00433017)
NEXUS	Efficacy and Safety Study of iSONEP With and Without Lucentis/Avastin to Treat Age‐related Macular Degeneration (NCT01414153)
PERSPECTIVES	An Open Label Trial to Investigate Macugen for the Preservation of Visual Function in Subjects With Neovascular AMD (NCT00327470)
PrONTO	Prospective Optical coherence tomography imaging of patients with Neovascular AMD Treated with intra‐Ocular ranibizumab (NCT00344227)
RADICAL	Reduced Fluence Visudyne‐Anti‐VEGF‐Dexamethasone In Combination for AMD Lesions (NCT00492284)
SAILOR	Safety Assessment of Intravitreal Lucentis® for Age‐Related Macular Degeneration (NCT00251459)
SALUTE	Comparison of Safety, Effectiveness and Quality of Life Outcomes Between Labeled Versus "Treat and Extend" Regimen in Turkish Patients With Choroidal Neovascularisation Due to AMD (NCT01148511)
SUMMIT	Unclear (clinical trial program including the DENALI, EVEREST, and MONT BLANC trials)
SUSTAIN	Study of Ranibizumab in Patients with Subfoveal Choroidal Neovascularization Secondary to Age‐Related Macular Degeneration (NCT00331864)
VERITAS	A Safety and Efficacy Study Comparing the Combination Treatments of Verteporfin Therapy Plus One of Two Different Doses of Intravitreal Triamcinolone Acetonide and the Verteporfin Therapy Plus Intravitreal Pegaptanib (NCT00242580)
VIEW‐1/2	Vascular Endothelial Growth Factor (VEGF) Trap‐Eye: Investigation of Efficacy and Safety in Wet Age‐Related Macular Degeneration (NCT00509795; NCT00637377)
WALTZ	Wet Age‐Related Macular Degeneration AL‐39324 Treatment Examination (NCT00992563)
*List of studies that may or may not be listed as excluded studies. Clinical trial identifiers are shown in parentheses.

Table 1. Table of Study Acronyms

Table 2. Treatment groups in included trials

Study Treatment period	Intervention 1	Intervention 2	Intervention 3	Intervention 4
Pegaptanib versus control
VISION 2004 2 years; re‐randomized at end of first year	0.3 mg pegaptanib every 6 weeks	1.0 mg pegaptanib every 6 weeks	3.0 mg pegaptanib every 6 weeks	Sham every 6 weeks
Ranibizumab versus control
ANCHOR 2006 2 years	0.3 mg ranibizumab monthly plus sham verteporfin PDT	0.5 mg ranibizumab monthly plus sham verteporfin PDT	Sham intravitreal injection plus verteporfin PDT	‐
MARINA 2006 2 years	0.3 mg ranibizumab monthly	0.5 mg ranibizumab monthly	Sham intravitreal injection monthly	‐
PIER 2008 2 years	0.3 mg ranibizumab monthly for 3 months, then every 3 months	0.5 mg ranibizumab monthly for 3 months, then every 3 months	Sham intravitreal injection monthly for 3 months, then every 3 months	‐
Bevacizumab versus control
ABC 2010 1 year	1.25 mg bevacizumab given first three injections every 6 weeks, then as needed	Standard therapy (0.3 mg pegaptanib every six weeks, verteporfin PDT, or sham injection)	‐	‐
Sacu 2009 1 year	1.0 mg bevacizumab monthly for 3 months, then as needed	Verteporfin PDT plus same day 4 mg triamcinolone acetonide	‐	‐
Bevacizumab versus ranibizumab
CATT 2011 2 years; re‐randomized at end of first year	1.25 mg bevacizumab monthly for 1 year; at 1 year, re‐randomization to ranibizumab monthly or variable dosing	0.5 mg ranibizumab monthly for 1 year; at 1 year, re‐randomization to ranibizumab monthly or variable dosing	1.25 mg bevacizumab as needed after first injection for 2 years	0.5 mg ranibizumab as needed after first injection for 2 years
IVAN 2013 2 years; ongoing	1.25 mg bevacizumab monthly for 2 years	0.5 mg ranibizumab monthly for 2 years	1.25 mg bevacizumab monthly for 3 months, then as needed in 3 month cycles	0.5 mg ranibizumab monthly for 3 months, then as needed in 3 month cycles
Biswas 2011 18 months	1.25 mg bevacizumab monthly for 3 months, then as needed	0.5 mg ranibizumab monthly for 3 months, then as needed	‐	‐
GEFAL 2013 1 year	1.25 mg bevacizumab; maximum of one injection per month	0.5 mg ranibizumab; maximum of one injection per month	‐	‐
MANTA 2013 1 year	1.25 mg bevacizumab monthly for 3 months, then as needed	0.5 mg ranibizumab monthly for 3 months, then as needed	‐	‐
Subramanian 2010 1 year	0.05 ml bevacizumab monthly for 3 months, then as needed	0.05 ml ranibizumab monthly for 3 months, then as needed	‐	‐
PDT: photodynamic therapy

Table 2. Treatment groups in included trials

Table 3. Adverse events up to one year: pegaptanib vs. control

Ocular adverse event*	0.3 mg pegaptanib n = 295	1.0 mg pegaptanib n = 301	3.0 mg pegaptanib n = 296	All doses pegaptanib n = 892	Control n = 298	RR [95% CI] All doses vs. control
Any eye disorder	9 (3%)	4 (1%)	10 (3%)	23 (3%)	2 (< 1%)	3.84 [0.91, 16.20]
Endophthalmitis	6 (2%)	3 (1%)	3 (1%)	12 (1%)	0	8.37 [0.50, 140.95]
Retinal detachment	1 (< 1%)	2 (< 1%)	2 (< 1%)	5 (< 1%)	0	3.68 [0.20, 66.41]
Traumatic cataract	1 (< 1%)	2 (< 1%)	2 (< 1%)	5 (< 1%)	0	3.68 [0.20, 66.41]
Retinal hemorrhage	1 (< 1%)	0	1 (< 1%)	2 (< 1%)	0	1.67 [0.08, 34.77]
Vitreous hemorrhage	0	0	1 (< 1%)	1 (< 1%)	0	1.00 [0.04, 24.59]
Uveitis	0	0	1 (< 1%)	1 (< 1%)	0	1.00 [0.04, 24.59]
Elevated intraocular pressure	1 (< 1%)	0	0	1 (< 1%)	0	1.00 [0.04, 24.59]
Papilledema	0	0	0	0	1 (< 1%)	0.11 [0.00, 2.73]
Non‐ocular adverse event*	0.3 mg pegaptanib n = 295	1.0 mg pegaptanib n = 301	3.0 mg pegaptanib n = 296	All doses pegaptanib n = 892	Control n = 298	RR [95% CI] All doses vs. control
At least 1 serious adverse event	55 (19%)	50 (17%)	64 (22%)	169 (19%)	45 (15%)	1.25 [0.93, 1.70]
Cardiac disorders	11 (4%)	4 (1%)	10 (3%)	25 (3%)	14 (5%)	0.60 [0.31, 1.13]
Neoplasms (benign, malignant, unspecified)	11 (4%)	7 (2%)	8 (3%)	26 (3%)	12 (4%)	0.72 [0.37, 1.42]
Injury and procedural complications, such as fractures (also includes traumatic cataracts)	10 (3%)	9 (3%)	8 (3%)	27 (3%)	3 (1%)	3.01 [0.92, 9.84]
Nervous system disorders	10 (3%)	5 (2%)	10 (3%)	25 (3%)	7 (2%)	1.19 [0.52, 2.73]
Infections and infestations	2 (<1%)	7 (2%)	11 (4%)	20 (2%)	5 (2%)	1.34 [0.51, 3.53]
Gastrointestinal disorders	3 (1%)	6 (2%)	5 (2%)	14 (2%)	4 (1%)	1.17 [0.39, 3.52]
Respiratory, thoracic, mediastinal disorders	2 (< 1%)	5 (2%)	5 (2%)	12 (1%)	4 (1%)	1.00 [0.33, 3.08]
Musculoskeletal and connective tissue	1 (< 1%)	5 (2%)	3 (1%)	9 (1%)	2 (<1%)	1.50 [0.33, 6.92]
Renal and urinary disorders	2 (< 1%)	3 (1%)	2 (<1%)	7 (<1%)	3 (1%)	0.78 [0.20, 3.00]
Vascular disorders	3 (1%)	2 (< 1%)	2 (< 1%)	7 (< 1%)	3 (1%)	0.78 [0.20, 3.00]
CI: confidence interval RR: risk ratio *Most frequent serious adverse events experienced by 1190 participants in the VISION 2004 study

Table 3. Adverse events up to one year: pegaptanib vs. control

Table 4. Adverse events up to one year: ranibizumab vs. control

Ocular adverse event*	0.3 mg ranibizumab n = 196	0.5 mg ranibizumab n = 201	All doses ranibizumab n = 397	Control n = 206	RR [95% CI] All doses vs. control
Endophthalmitis	0	2 (< 1%)	2 (< 1%)	0	2.60 [0.13, 53.92]
Retinal detachment	1 (< 1%)	0	1 (< 1%)	1 (< 1%)	0.52 [0.03, 8.25]
Traumatic cataract	18 (9%)	22 (11%)	40 (10%)	14 (7%)	1.48 [0.83, 2.66]
Retinal hemorrhage	2 (1%)	0	2 (< 1%)	2 (< 1%)	0.52 [0.07, 3.66]
Vitreous hemorrhage	1 (< 1%)	0	1 (< 1%)	0	1.56 [0.06, 38.13]
Uveitis	0	1 (< 1%)	1 (< 1%)	0	1.56 [0.06, 38.13]
Elevated intraocular pressure (30 mmHg or more increase)	13 (7%)	17 (8%)	30 (8%)	7 (3%)	2.22 [0.99, 4.98]
Ocular inflammation (trace to 4+)	21 (11%)	26 (13%)	47 (12%)	9 (4%)	2.71 [1.36, 5.42]
Non‐ocular adverse event*	0.3 mg ranibizumab n = 196	0.5 mg ranibizumab n = 201	All doses ranibizumab n = 397	Control n = 206	RR [95% CI] All doses vs. control
Death	3 (2%)	2 (<1%)	5 (1%)	2 (< 1%)	1.30 [0.25, 6.63]
Myocardial infarction	1 (< 1%)	3 (1%)	4 (1%)	1 (< 1%)	2.08 [0.23, 18.45]
Stroke or cerebral infarction	1 (< 1%)	1 (< 1%)	2 (< 1%)	1 (< 1%)	1.04 [0.09, 11.38]
Ischemic cardiomyopathy	0	0	0	1 (< 1%)	0.17 [0.01, 4.24]
Treatment‐emergent hypertension	7 (4%)	15 (7%)	22 (6%)	17 (8%)	0.67 [0.36, 1.24]
Non‐ocular hemorrhage	9 (5%)	13 (6%)	22 (6%)	6 (3%)	1.90 [0.78, 4.62]
CI: confidence interval RR: risk ratio *Adverse events experienced by 420 participants in the ANCHOR 2006 study and 183 participants in the PIER 2008 study. Adverse events at one‐year follow up not reported in MARINA 2006.

Table 4. Adverse events up to one year: ranibizumab vs. control

Table 5. Adverse events up to two years: ranibizumab vs. control

Ocular adverse event*	0.3 mg ranibizumab n = 434	0.5 mg ranibizumab n = 440	All doses ranibizumab n = 874	Control n = 441	RR [95% CI] All doses vs. control
Endophthalmitis	2 (< 1%)	6 (1%)	8 (< 1%)	0	8.59 [0.50, 148.44]
Retinal detachment	2 (< 1%)	0	2 (< 1%)	2 (< 1%)	0.50 [0.07, 3.57]
Traumatic cataract	65 (15%)	76 (17%)	141 (16%)	57 (13%)	1.25 [0.94, 1.66]
Retinal hemorrhage	1 (< 1%)	0	1 (< 1%)	1 (< 1%)	0.50 [0.03, 8.05]
Vitreous hemorrhage	3 (< 1%)	1 (< 1%)	4 (< 1%)	2 (< 1%)	1.01 [0.19, 5.49]
Uveitis	3 (< 1%)	4 (< 1%)	7 (<1%)	0	7.58 [0.43, 132.36]
Elevated intraocular pressure (30 mmHg or more increase)**	45 (15%)	61 (20%)	106 (18%)	11 (4%)	4.81 [2.63, 8.81]
Ocular inflammation (1+ to 4+)	32 (7%)	30 (7%)	62 (7%)	8 (2%)	3.91 [1.89, 8.09]
Non‐ocular adverse event*	0.3 mg ranibizumab n = 434	0.5 mg ranibizumab n = 440	All doses ranibizumab n = 874	Control n = 441	RR [95% CI] All doses vs. control
Death	12 (3%)	9 (2%)	21 (2%)	13 (3%)	0.82 [0.41, 1.61]
Myocardial infarction	7 (2%)	8 (2%)	15 (2%)	7 (2%)	1.08 [0.44, 2.63]
Stroke or cerebral infarction	6 (1%)	6 (1%)	12 (1%)	5 (1%)	1.21 [0.43, 3.42]
Ischemic cardiomyopathy	0	0	0	1 (< 1%)	0.17 [0.01, 4.12]
Treatment‐emergent hypertension	60 (14%)	69 (16%)	129 (15%)	68 (15%)	0.96 [0.73, 1.25]
Nonocular hemorrhage	38 (9%)	40 (9%)	78 (9%)	24 (5%)	1.64 [1.05, 2.55]
CI: confidence interval RR: risk ratio Adverse events experienced by 420 participants in the ANCHOR 2006 study; 713 participants in the MARINA 2006 study; and 182 participants in the PIER 2008 study. *Adverse events for elevated intraocular pressure not reported in the ANCHOR 2006 study at two‐year follow up (n = 297 in 0.3 mg ranibizumab group, n = 300 in 0.5 mg ranibizumab group, and n = 298 in 0.3 mg control group).

Table 5. Adverse events up to two years: ranibizumab vs. control

Table 6. Adverse events up to one year: bevacizumab vs. ranibizumab

Serious ocular adverse event	Studies reporting adverse event*	Bevacizumab		Ranibizumab		RR [95% CI] Bevacizumab vs. ranibizumab
Serious ocular adverse event	Studies reporting adverse event*	Number with event	Total participants	Number with event	Total participants	RR [95% CI] Bevacizumab vs. ranibizumab
Endophthalmitis	CATT 2011; GEFAL 2013	4 (< 1%)	832	3 (< 1%)	838	1.34 [0.30, 5.98]
Retinal detachment	CATT 2011; GEFAL 2013	3 (< 1%)	832	0	838	7.05 [0.36, 136.28]
Retinal pigment epithelial tear	CATT 2011; IVAN 2013	3 (< 1%)	882	3 (< 1%)	913	1.04 [0.21, 5.11]
Traumatic cataract	CATT 2011; GEFAL 2013; IVAN 2013	1 (< 1%)	1128	2 (< 1%)	1152	0.51 [0.05, 5.62]
Severe uveitis	CATT 2011; IVAN 2013	4 (< 1%)	882	1 (< 1%)	913	4.14 [0.46, 36.97]
Non‐ocular adverse event	Studies reporting adverse event**	Bevacizumab		Ranibizumab		RR [95% CI] Bevacizumab vs. ranibizumab
Non‐ocular adverse event	Studies reporting adverse event**	Number with event	Total participants	Number with event	Total participants	RR [95% CI] Bevacizumab vs. ranibizumab
At least 1 serious adverse event	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	227 (18%)	1282	183 (14%)	1315	1.27 [1.06, 1.52]
Death	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	25 (2%)	1282	20 (2%)	1315	1.28 [0.72, 2.30]
Myocardial infarction	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	8 (< 1%)	1282	10 (< 1%)	1315	0.82 [0.32, 2.07]
Stroke or cerebral infarction	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	5 (< 1%)	1282	8 (< 1%)	1315	0.64 [0.21, 1.95]
Transient ischemic attack	CATT 2011; GEFAL 2013; IVAN 2013	4 (< 1%)	1128	4 (< 1%)	1152	1.02 [0.26, 4.07]
Venous thrombotic event	CATT 2011; GEFAL 2013; IVAN 2013	8 (< 1%)	1128	2 (< 1%)	1152	4.09 [0.87, 19.20]
Cardiac disorders	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	37 (3%)	1282	36 (3%)	1315	1.05 [0.67, 1.66]
Gastrointestinal disorders	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	24 (2%)	1282	11 (< 1%)	1315	2.24 [1.10, 4.55]
Infections	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	42 (3%)	1282	27 (2%)	1315	1.60 [0.99, 2.57]
Injury and procedural complications	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	30 (2%)	1282	21 (2%)	1315	1.47 [0.84, 2.55]
Neoplasms (benign, malignant, unspecified)	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	20 (2%)	1282	21 (2%)	1315	0.98 [0.53, 1.79]
Nervous system disorders	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	25 (2%)	1282	24 (2%)	1315	1.07 [0.61, 1.86]
Surgical or medical procedure	CATT 2011; GEFAL 2013; IVAN 2013; MANTA 2013	26 (2%)	1282	13 (1%)	1315	2.05 [1.06, 3.97]
CI: confidence interval RR: risk ratio * CATT 2011 (n = 586 in bevacizumab group; n = 599 in ranibizumab group); GEFAL 2013 (n = 246 in bevacizumab group; n = 239 in ranibizumab group); IVAN 2013 (n = 296 in bevacizumab group; n = 314 in ranibizumab group) **CATT 2011 (n = 586 in bevacizumab group; n = 599 in ranibizumab group); GEFAL 2013 (n = 246 in bevacizumab group; n = 239 in ranibizumab group); IVAN 2013 (n = 296 in bevacizumab group; n = 314 in ranibizumab group); MANTA 2013 (n = 154 in bevacizumab group; n = 163 in ranibizumab group)

Table 6. Adverse events up to one year: bevacizumab vs. ranibizumab

Table 7. Adverse events up to two years: bevacizumab vs. ranibizumab

Ocular adverse event (CATT trial)*	Bevacizumab n = 586	Ranibizumab n = 599	RR [95% CI] Bevacizumab vs. ranibizumab
Endophthalmitis	7 (1%)	4 (< 1%)	1.79 [0.53, 6.08]
Ocular adverse event (IVAN trial)**	Bevacizumab n = 296	Ranibizumab n = 314	RR [95% CI] Bevacizumab vs. ranibizumab
Traumatic cataract	1 (< 1%)	1 (< 1%)	1.06 [0.07, 16.88]
Severe uveitis	1 (< 1%)	0	3.18 [0.13, 77.80]
Retinal detachment	0	1 (< 1%)	0.35 [0.01, 8.64]
Retinal pigment epithelial tear	1 (< 1%)	3 (< 1%)	0.35 [0.04, 3.38]
Non‐ocular adverse event†	Bevacizumab n = 882	Ranibizumab n = 913	RR [95% CI] Bevacizumab vs. ranibizumab
At least 1 serious adverse event	314 (36%)	271 (30%)	1.20 [1.05, 1.37]
Death	51 (6%)	47 (5%)	1.12 [0.76, 1.65]
Myocardial infarction	11 (1%)	13 (1%)	0.88 [0.39, 1.94]
Stroke or cerebral infarction	11 (1%)	14 (2%)	0.81 [0.37, 1.78]
Venous thrombotic event	14 (2%)	6 (< 1%)	2.42 [0.93, 6.26]
Transient ischemic attack**	1 (< 1%)	1 (< 1%)	1.04 [0.06, 16.52]
Cardiac disorders	81 (9%)	67 (7%)	1.25 [0.92, 1.71]
Gastrointestinal disorders	37 (4%)	14 (2%)	2.74 [1.49, 5.02]
Infections	66 (7%)	50 (5%)	1.37 [0.96, 1.95]
Injury and procedural complications	45 (5%)	35 (4%)	1.33 [0.86, 2.05]
Neoplasms (benign, malignant, unspecified)	36 (4%)	38 (4%)	0.98 [0.63, 1.53]
Nervous system disorders	44 (5%)	43 (5%)	1.06 [0.70, 1.60]
Surgical or medical procedure**	14 (5%)	16 (5%)	0.91 [0.44, 1.84]
CI: confidence interval RR: risk ratio Adverse events for endophthalmitis not reported in the IVAN 2013 study; data for CATT 2011 study only *Adverse events for traumatic cataract, uveitis, retinal detachment, retinal pigment epithelial tear, transient ischemic attack, and surgical or medical procedure not reported in the CATT 2011 study; data for IVAN 2013 study only †Adverse events experienced by 1185 participants in the CATT 2011 study and 610 participants in the IVAN 2013 study

Table 7. Adverse events up to two years: bevacizumab vs. ranibizumab

Comparison 1. Pegaptanib versus control

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gain of 15 letters or more visual acuity at one year Show forest plot	1		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected

2 Loss of fewer than 15 letters visual acuity at one year Show forest plot	1		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected

3 Loss of fewer than 30 letters visual acuity at one year Show forest plot	1		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected

4 Visual acuity better than 20/200 at one year Show forest plot	1		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected

5 Maintenance of visual acuity at one year Show forest plot	1		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected

6 Mean change in visual acuity at one year (number of letters) Show forest plot	1		Mean Difference (IV, Random, 95% CI)	Totals not selected

7 Reduction in size of CNV at one year (Mean number of disc areas) Show forest plot	1		Mean Difference (Random, 95% CI)	Totals not selected

8 Reduction in size of lesion at one year (Mean number of disc areas) Show forest plot	1		Mean Difference (Random, 95% CI)	Totals not selected

Comparison 1. Pegaptanib versus control

Comparison 2. Ranibizumab versus control

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gain of 15 letters or more visual acuity at one year Show forest plot	3		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected

2 Gain of 15 letters or more visual acuity at two years Show forest plot	3	1322	Risk Ratio (M‐H, Random, 95% CI)	5.77 [3.38, 9.84]

3 Loss of fewer than 15 letters visual acuity at one year Show forest plot	3	1322	Risk Ratio (M‐H, Random, 95% CI)	1.53 [1.41, 1.64]

4 Loss of fewer than 15 letters visual acuity at two years Show forest plot	3	1322	Risk Ratio (M‐H, Random, 95% CI)	1.62 [1.32, 1.98]

5 Loss of fewer than 30 letters visual acuity at one year Show forest plot	2	1138	Risk Ratio (M‐H, Random, 95% CI)	1.15 [1.11, 1.20]

6 Loss of fewer than 30 letters visual acuity at two years Show forest plot	2	1138	Risk Ratio (M‐H, Random, 95% CI)	1.22 [1.15, 1.29]

7 Visual acuity better than 20/200 at one year Show forest plot	3	1322	Risk Ratio (M‐H, Random, 95% CI)	1.69 [1.41, 2.03]

8 Visual acuity better than 20/200 at two years Show forest plot	3	1322	Risk Ratio (M‐H, Random, 95% CI)	1.73 [1.52, 1.98]

9 Maintenance of visual acuity at one year Show forest plot	1		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected

10 Maintenance of visual acuity at two years Show forest plot	1		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected

11 Mean change in visual acuity at one year (number of letters) Show forest plot	3	1322	Mean Difference (IV, Random, 95% CI)	17.80 [15.95, 19.65]

12 Mean change in visual acuity at two years (number of letters) Show forest plot	3	1322	Mean Difference (IV, Random, 95% CI)	20.11 [18.08, 22.15]

13 Reduction in size of lesion at one year (Mean number of disc areas) Show forest plot	2	606	Mean Difference (IV, Random, 95% CI)	2.34 [1.88, 2.81]

14 Reduction in size of lesion at two years (Mean number of disc areas) Show forest plot	2		Mean Difference (IV, Random, 95% CI)	Totals not selected

15 Mean change in quality of life scores at one year Show forest plot	2		Mean Difference (IV, Random, 95% CI)	Subtotals only

15.1 Overall vision‐related quality of life	2	1134	Mean Difference (IV, Random, 95% CI)	6.69 [3.38, 9.99]
15.2 Near vision activities	2	1134	Mean Difference (IV, Random, 95% CI)	8.45 [0.28, 16.62]
15.3 Distance vision activities	2	1134	Mean Difference (IV, Random, 95% CI)	9.65 [3.20, 16.09]
15.4 Dependency	2	1134	Mean Difference (IV, Random, 95% CI)	9.82 [6.86, 12.77]
15.5 Driving ability	2	1080	Mean Difference (IV, Random, 95% CI)	9.85 [6.34, 13.36]
15.6 General health	2	1134	Mean Difference (IV, Random, 95% CI)	3.18 [0.54, 5.82]
15.7 Role difficulties	2	1134	Mean Difference (IV, Random, 95% CI)	6.99 [0.76, 13.23]
15.8 Mental health	2	1134	Mean Difference (IV, Random, 95% CI)	8.42 [5.75, 11.10]
15.9 General vision	2	1134	Mean Difference (IV, Random, 95% CI)	8.20 [5.90, 10.50]
15.10 Social functioning	2	1134	Mean Difference (IV, Random, 95% CI)	8.03 [5.36, 10.69]
15.11 Color vision	2	1127	Mean Difference (IV, Random, 95% CI)	2.51 [‐0.02, 5.05]
15.12 Peripheral vision	2	1133	Mean Difference (IV, Random, 95% CI)	5.20 [0.37, 10.03]
15.13 Ocular pain	2	1134	Mean Difference (IV, Random, 95% CI)	‐1.78 [‐3.67, 0.11]
16 Mean change in quality of life scores at two years Show forest plot	2		Mean Difference (IV, Random, 95% CI)	Subtotals only

16.1 Overall vision‐related quality of life	2	1134	Mean Difference (IV, Random, 95% CI)	8.63 [3.31, 13.95]
16.2 Near vision activities	2	1134	Mean Difference (IV, Random, 95% CI)	11.52 [3.49, 19.55]
16.3 Distance vision activities	2	1134	Mean Difference (IV, Random, 95% CI)	10.86 [3.82, 17.90]
16.4 Dependency	2	1134	Mean Difference (IV, Random, 95% CI)	11.06 [3.29, 18.83]
16.5 Driving ability	2	1080	Mean Difference (IV, Random, 95% CI)	13.53 [9.51, 17.55]
16.6 General health	2	1134	Mean Difference (IV, Random, 95% CI)	2.58 [‐0.18, 5.35]
16.7 Role difficulties	2	1134	Mean Difference (IV, Random, 95% CI)	9.44 [1.34, 17.54]
16.8 Mental health	2	1134	Mean Difference (IV, Random, 95% CI)	10.07 [3.98, 16.17]
16.9 General vision	2	1134	Mean Difference (IV, Random, 95% CI)	9.61 [5.49, 13.72]
16.10 Social functioning	2	1134	Mean Difference (IV, Random, 95% CI)	8.12 [1.77, 14.47]
16.11 Color vision	2	1127	Mean Difference (IV, Random, 95% CI)	5.70 [2.89, 8.51]
16.12 Peripheral vision	2	1133	Mean Difference (IV, Random, 95% CI)	6.79 [1.48, 12.09]
16.13 Ocular pain	2	1134	Mean Difference (IV, Random, 95% CI)	‐1.10 [‐3.13, 0.92]

Comparison 2. Ranibizumab versus control

Comparison 3. Bevacizumab versus control

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gain of 15 letters or more visual acuity at one year Show forest plot	2	159	Risk Ratio (M‐H, Random, 95% CI)	7.80 [2.44, 24.98]

2 Loss of fewer than 15 letters visual acuity at one year Show forest plot	2	159	Risk Ratio (M‐H, Random, 95% CI)	1.28 [1.09, 1.50]

Comparison 3. Bevacizumab versus control

Comparison 4. Bevacizumab versus ranibizumab

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gain of 15 letters or more visual acuity at one year Show forest plot	6	2446	Risk Ratio (M‐H, Random, 95% CI)	0.90 [0.73, 1.11]

2 Gain of 15 letters or more visual acuity at two years Show forest plot	2		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

2.1 Participants in groups as randomized at baseline	2	1547	Risk Ratio (M‐H, Random, 95% CI)	0.84 [0.64, 1.11]
2.2 Participants remaining in same groups after re‐randomization	2	1295	Risk Ratio (M‐H, Random, 95% CI)	0.84 [0.64, 1.11]
3 Loss of fewer than 15 letters visual acuity at one year Show forest plot	6	2446	Risk Ratio (M‐H, Random, 95% CI)	1.00 [0.98, 1.02]

4 Loss of fewer than 15 letters visual acuity at two years Show forest plot	2		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

4.1 Participants in groups as randomized at baseline	2	1547	Risk Ratio (M‐H, Random, 95% CI)	0.97 [0.94, 1.00]
4.2 Participants remaining in same groups after re‐randomization	2	1295	Risk Ratio (M‐H, Random, 95% CI)	0.98 [0.94, 1.01]
5 Visual acuity better than 20/200 at one year Show forest plot	4	2026	Risk Ratio (M‐H, Random, 95% CI)	0.98 [0.96, 1.01]

6 Visual acuity better than 20/200 at two years Show forest plot	2		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

6.1 Participants in groups as randomized at baseline	2	1547	Risk Ratio (M‐H, Random, 95% CI)	1.00 [0.95, 1.06]
6.2 Participants remaining in same groups after re‐randomization	2	1295	Risk Ratio (M‐H, Random, 95% CI)	1.01 [0.95, 1.06]
7 Mean change in visual acuity at one year (number of letters) Show forest plot	6	2446	Mean Difference (IV, Random, 95% CI)	‐0.51 [‐1.64, 0.62]

8 Mean change in visual acuity at two years (number of letters) Show forest plot	2	1295	Mean Difference (IV, Random, 95% CI)	‐1.15 [‐2.82, 0.51]

9 Reduction in central retinal thickness at one year Show forest plot	4	1995	Mean Difference (IV, Random, 95% CI)	‐13.97 [‐26.52, ‐1.41]

10 Reduction in central retinal thickness at two years Show forest plot	2	1199	Mean Difference (IV, Random, 95% CI)	‐12.40 [‐33.83, 9.04]

11 No problems in quality of life domain at one year Show forest plot	1		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected

11.1 Mobility	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
11.2 Self care	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
11.3 Usual activities	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
11.4 Pain/discomfort	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
11.5 Anxiety/depression	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
12 No problems in quality of life domain at two years Show forest plot	1		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected

12.1 Mobility	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
12.2 Self care	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
12.3 Usual activities	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
12.4 Pain/discomfort	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
12.5 Anxiety/depression	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]

Comparison 4. Bevacizumab versus ranibizumab