Article Text

Intravitreal bevacizumab (Avastin) versus ranibizumab (Lucentis) for the treatment of age-related macular degeneration: a safety review
1. Christine Schmucker1,
2. Yoon K Loke2,
3. Christoph Ehlken3,
4. Hansjuergen T Agostini3,
5. Lutz L Hansen3,
6. Gerd Antes1,
7. Monika Lelgemann4
1. 1German Cochrane Centre, University Medical Centre Freiburg, Freiburg, Germany
2. 2School of Medicine, University of East Anglia, Norwich, UK
3. 3University Eye Hospital, University Medical Centre Freiburg, Freiburg, Germany
4. 4Health Technology Assessment Centre, University Bremen, Germany
1. Correspondence to Christine Schmucker, German Cochrane Centre, Institute of Medical Biometry and Medical Informatics, Department of Medical Biometry and Statistics, University Medical Centre Freiburg, Berliner Allee 29, Freiburg 79110, Germany; schmucker{at}cochrane.de

## Abstract

Aim To conduct a systematic review in order to compare adverse effects (AE) and the reporting of harm in randomised controlled trials (RCTs) and non-RCTs evaluating intravitreal ranibizumab and bevacizumab in age-related macular degeneration.

Methods Medline, Embase and the Cochrane Library were searched with no limitations of language and year of publication. Studies which compared bevacizumab or ranibizumab as monotherapy with any other control group were included. Case series were included if they met predefined quality standards.

Results The 2 year results of phase III trials evaluating ranibizumab show that the rates of serious ocular AE were low (≤2.1%) but indicate major safety concerns (RR 3.13, 95% CI 1.10 to 8.92). A possible signal with regard to thromboembolic events (RR 1.35, 95% CI 0.66 to 2.77) and a significant increase in non-ocular haemorrhage (RR 1.62, 95% CI 1.03 to 2.55) were also noted. In contrast to ranibizumab trials, the RCTs evaluating bevacizumab are of limited value. The main shortcomings are small sample sizes and an apparent lack of rigorous monitoring for AE. A critical assessment of the large number of published case series evaluating bevacizumab also shows that no reliable conclusions on safety can be drawn using this study design. Therefore, any perception that intravitreal bevacizumab injections are not associated with major ocular or systemic AE are not supported by reliable data.

Conclusion The bevacizumab studies show too many methodological limitations to rule out any major safety concerns. Higher evidence from ranibizumab trials suggests signals for an increased ocular and systemic vascular and haemorrhagic risk which warrants further investigation.

• systematic review
• bevacizumab
• ranibizumab
• safety
• retina
• drugs
• medical education

## Statistics from Altmetric.com

Age-related macular degeneration (AMD) is the leading cause of irreversible blindness in people over the age of 50 in the developed world.1 Although an estimated 80% of patients with AMD have the non-neovascular form,2 the neovascular (wet or exudative) form is responsible for almost 90% of severe visual loss resulting from AMD.3

Anti-angiogenic therapy, for example anti-vascular endothelial growth factors (anti-VEGF), which aims to prevent further neovascularisation rather than only destroy it, is the latest approach to the treatment of exudative AMD. Currently, the most commonly used VEGF antagonists are ranibizumab (Lucentis; Genentech, San Francisco, California, USA) and bevacizumab (Avastin; Genentech). Ranibizumab, which is an antibody fragment from the bevacizumab molecule with an increased binding affinity for all forms of VEGF, has been approved for the treatment of patients with neovascular AMD by the Food and Drug Administration and by the European Medicines Agency since 2006 and 2007, respectively. The approval was based on three randomised controlled trials (RCTs).4 Two of these studies showed that approximately 95% of the patients treated with monthly ranibizumab injections lost fewer than 15 letters in 12 months, compared to 64% of patients receiving photodynamic therapy (PDT)5 and 62% receiving sham treatment.6 In addition, approximately every third patient showed improvements in visual acuity under ranibizumab treatment. The costs of ranibizumab, however, are immense. Using monthly injections with a dose of 0.5 mg, the annual costs come to more than US$23 000 per patient.7 In contrast to ranibizumab, bevacizumab was not developed for the treatment of AMD and consequently has no approval for this use. Bevacizumab is approved for the treatment of specific cancers, for example, metastatic colon and rectum cancer. Even before ranibizumab was licensed, bevacizumab had been used as an off-label treatment for AMD. The first report of intravitreal bevacizumab administration for neovascular AMD was published in 2005.8 After this initial report, numerous case series which (apparently) support the efficacy and safety of bevacizumab were published. The costs of intravitreal bevacizumab are much less than for ranibizumab. Small aliquots in syringes for intraocular injections can be prepared for about US$17–50 a month (≤US\$600 annually).7

Despite lacking evidence, most published reviews agree that bevacizumab seems to be similar to ranibizumab, effective in maintaining visual acuity. However, the safety and tolerability of bevacizumab in comparison to ranibizumab have not been adequately assessed and hence, the crucial question whether the existing safety data justify the widespread intravitreal off-label use of bevacizumab has not yet been answered. Moreover, it is essential to assess differences in reporting of harm between well controlled phase III studies, RCTs which do not fulfil phase III study requirements, and non-RCTs. In particular, a critical assessment of the large number of published case series evaluating bevacizumab will show to what extent they can be used as a substitute for high quality trials.

## Materials and methods

### Systematic literature search

We searched Medline (Ovid), Embase and the Cochrane Library from inception until March 2008. An update search focusing on RCTs was carried out in August 2009. The search strategy was based on combinations of medical subject headings (MeSH) and keywords and was not restricted to specific languages or years of publication. The search strategy used in Medline is presented in appendix A. Search strategies for other databases were modified to meet the requirements of each database. The searches were supplemented by handsearching the bibliographies of included studies and reviews and by contacting the pharmaceutical manufacturer (Genentech) of ranibizumab and bevacizumab. Currently conducted RCTs comparing Avastin and Lucentis were searched both in the register for clinical trials (http://clinicaltrials.gov/) and in the WHO International Clinical Trials Registry Platform (http://www.who.int/ictrp/en/).

### Study selection

Included were RCTs and non-RCTs which compared intravitreal bevacizumab or ranibizumab as monotherapy with any other treatment in patients with neovascular AMD. Case series were included if they enrolled a minimum of 10 patients and met predefined quality standards; that is, the publication had to provide adequate information regarding patient selection criteria or the selection of patients had to be consecutive. Studies which included patients with indications other than exudative AMD, patients previously treated with VEGF inhibitors or patients receiving systemic anti-VEGF therapy were excluded.

### Data extraction and quality assessment

Titles and abstracts were reviewed using the above mentioned selection criteria. Full papers of appropriate studies were obtained for detailed evaluation. Data extraction and quality assessment was carried out after a modified evaluation tool of the Centre for Reviews and Dissemination.9 Information on the number of participants, ascertainment of exposure (eg, dosage and frequency of drug administered), follow-up time, definition of expected adverse effects, method used to collect adverse effects data, ascertainment of outcomes (ocular and systemic adverse effects) and transparency of patient flow were abstracted. All stages of study selection, data extraction and quality assessment were done independently by two reviewers (CS and CE, or CS and ML). Any disagreement was resolved by discussion and consensus.

### Statistical analysis

Ranibizumab data were analysed using the R software.10 This program was used to compute statistics and generate forest plots to compare safety outcomes in different treatment arms. A χ2 test (p<0.05) and an I2 test were used to test for statistical heterogeneity between studies. We used the fixed effects model (Mantel–Haenszel method) in the meta-analysis of rare events as it has been shown to be the more appropriate and less biased approach compared to the random effects model.11 A narrative summary was provided for data that were unsuitable for pooling (studies evaluating bevacizumab).

## Results

### Results of the search and selection process

Figure 1 shows the numbers of studies identified at each stage of the systematic review. After removing duplicate references, the searches identified 3628 citations. The inclusion criteria were met by four RCTs 6 12–14 (11 publications 5 6 12–20) evaluating ranibizumab versus PDT, sham or usual care with a total of 1392 patients, and four RCTs 21–24 (five publications 21–25) evaluating bevacizumab versus PDT with or without triamcinolone with a total of 287 patients. In addition, 17 case series 26–42 examining bevacizumab, including a total of 1790 patients, were analysed.

Figure 1

Flow diagram.

### Study characteristics

#### Ranibizumab (RCTs)

Characteristics of the RCTs evaluating intravitreal ranibizumab are presented in table 1. The ANCHOR study, which compared monthly ranibizumab injections with PDT, enrolled 423 patients with predominantly classic subfoveal choroidal neovascularisation (CNV).12 Follow-up time was 24 months. The MARINA study enrolled 716 patients with minimally classic and occult subfoveal CNV and compared monthly intravitreal ranibizumab with sham injections over 24 months.6 The PIER study also used sham as a comparator and enrolled 184 patients with occult or classic subfoveal CNV.13 In contrast to the MARINA study, treated patients received ranibizumab injections once monthly for three consecutive months, followed by a dose administered once every three months (follow-up time: 12 months). Heier randomised 64 patients to monthly intravitreal ranibizumab with varying doses for 3 months or usual care (ie, PDT in predominantly classic lesions and observation in all other lesions).14 In the second part of the study, patients could continue their regimen for three additional months or cross over to the alternative treatment. All RCTs evaluating ranibizumab were multicentre trials. The ANCHOR,12 MARINA6 and PIER13 studies were sponsored by pharmaceutical companies. The source of funding was not reported in the study of Heier.14

Table 1

Characteristics of RCTs evaluating ranibizumab and bevacizumab

#### Bevacizumab (RCTs)

Table 1 shows study characteristics of RCTs evaluating intravitreal bevacizumab. Bashshur randomised 64 patients with predominantly classic CNV to PDT or intravitreal bevacizumab injections pro re nata.21 The patients were followed-up for 6 months. The 3 month study of Lazic enrolled 165 patients with minimally classic or occult CNV and compared a single bevacizumab injection with a single PDT session or a combination therapy.22 Hahn used PDT in combination with triamcinolone as comparator and enrolled 30 patients with occult or (minimally) classic subfoveal CNV.23 The patients in this study received monthly bevacizumab injections for three consecutive months. Saccu also used PDT in combination with triamcinolone and enrolled 28 patients with occult or (minimally) classic subfoveal CNV.24 Treated patients received bevacizumab injections once monthly for three consecutive months, followed by a dose administered pro re nata. Follow-up time was 12 months. All trials evaluating bevacizumab were monocentre studies. Two studies reported that no pharmaceutical sponsor was involved,21 23 and two did not provide data on the source of funding.22 24

#### Bevacizumab (case series)

In total 11 prospective 26–36 and six retrospective 37–42 case series evaluating bevacizumab were analysed (table 2). The number of included patients ranged between 13 and 625 (median 48). Minimum follow-up time was one month and maximum follow-up 12 months. The patients received between one and four injections and the applied dosage of bevacizumab varied between 1.0 mg and 2.5 mg. In 10 case series 27 30 33 35 36 38–42 patients were injected pro re nata, in three case series 26 31 37 a single injection of bevacizumab was given, and in four case series 28 29 32 34 injections at intervals of 4 or 6 weeks were administered. Funding sources were specified in nine publications.29–31 33–35 38 39 41 A pharmaceutical sponsor was, however, not reported.

Table 2

Characteristics of case series evaluating bevacizumab

#### Ranibizumab (RCTs)

Intravitreal ranibizumab injections have been associated with endophthalmitis (≤2.1%), uveitis (≤1.3%), retinal detachment (≤1.5%), retinal tear (≤1.9%), vitreous haemorrhage (≤8.0%) and traumatic lens damage (≤0.4%) (table 3).6 12 14 A pooled analysis on serious ocular adverse effects indicated some major safety issues (RR 3.13, p=0.03, figure 2A). In addition, all trials reported a transient increase in intraocular pressure in the study eye after intravitreal injections.

Table 3

Rates of ocular adverse effects under ranibizumab and bevacizumab/RCTs

Figure 2

Forest plots of ranibizumab (any dose vs any control) for different safety outcomes.

#### Bevacizumab (RCTs)

Most RCTs evaluating bevacizumab stated generically that ocular adverse effects were not noted. As the studies did not provide a detailed breakdown on the presence or absence of specific adverse effects, we are left to make the assumption that intravitreal bevacizumab injections were not associated with major problems, such as endophthalmitis, uveitis, retinal detachment, lens damage and vitreous haemorrhage (table 3). Any transient increase in intraocular pressure after the injections was also not reported. However, an increased rate of pigment epithelial tears (5.5% vs 0.0%), posterior vitreous detachment (14.6% vs 0.0%) and cataract progression (7.3% vs 0.0%) was reported in one RCT evaluating bevacizumab.22

#### Bevacizumab (case series)

Four publications reported an increased rate of endophthalmitis (range between 0.2% and 0.9%, table 4).38 40–42 An increased rate of retinal pigment epithelial tears (rips) was observed in five reports (range between 0.9% and 7.5%).34 35 37 38 41 Vitreous detachment was reported in 9.8% of patients in one publication 35 and an increased rate of submacular haemorrhage was observed in two case series (2.7%,38 7.5% 39). Moreover, in the retrospective case series of Wu, an increased rate of uveitis (0.3%), retinal detachment (0.6%) and vitreous haemorrhage (0.08%) was reported.42 Six case series reported minor ocular adverse effects, such as pain,31 conjunctival hyperaemia,31 subconjunctival haemorrhage,26 31 32 42 mild intraocular inflammation,26 transient corneal epitheliopathy 34 and transient blurred vision.35

Table 4

Rates of ocular adverse effects under bevacizumab/case series

#### Ranibizumab (RCTs)

The rate of non-ocular adverse effects of single RCTs are displayed in table 5 and pooled analysis for different systemic outcomes in figure 2B–D. The rate of key arterial non-fatal thromboembolic effects (myocardial infarction and stroke) during the first and second year of the ANCHOR12 and MARINA6 trials was numerically, but not statistically significantly higher in the 0.5 mg arm than in the control arm (3.6%12 and 2.5%,6 respectively vs 1.4% and 0.8%, respectively). However, a pooled analysis indicated that there may be a safety signal (RR 1.35, 95% CI 0.66 to 2.77). In the ANCHOR,12 MARINA6 and PIER13 studies, the incidence of serious non-ocular haemorrhage (such as gastrointestinal haemorrhage, traumatic subdural haematoma and duodenal ulcer haemorrhage) was also consistently higher in the ranibizumab than in the control groups (2.9% (0.3 mg),12 2.1% (0.5 mg)6 and 0.6% (0.5 mg)13 vs 0.7%, 0.8% and 0.0%). A pooled analysis indicated that this risk reached the standard thresholds for statistical significance (RR 1.62, p=0.04). Treatment-emergent hypertension was not more common in the ranibizumab than in the control groups.6 12 13 In the safety study of Heier, non-ocular adverse effects were not specified.14 However, it was reported that the only non-ocular adverse event judged by an investigator to be a possible effect of ranibizumab was a case of metallic taste in the mouth.

Table 5

Rates of systemic adverse effects under ranibizumab and bevacizumab/RCTs

#### Bevacizumab (RCTs)

Intravitreal bevacizumab injections were apparently not associated with any systemic adverse effects in the existing RCTs (table 5). This assumption is based on the following limited details concerning the harms reported within the articles. Two trials mentioned generically that no systemic effects were observed.21 24 One study did not mention systemic complications in the results.23 Another study reported that no thromboembolic events were observed.22 Other non-ocular complications were not mentioned in this study.

#### Bevacizumab (case series)

Non-ocular adverse effects were observed in one retrospective case series (table 6).42 The incidence of cerebrovascular accidents and myocardial infarct was 0.3% and the incidence of acute hypertension was 1.5%. In addition, 0.6% of patients with exudative AMD showed an iliac artery aneurysm. Overall, eight case series mentioned—by using numerical data—that no thromboembolic events occurred.28–30 34–36 38 41 Six publications stated briefly that no systemic adverse effects were observed (except that blood pressure was reported in more detail) 26 27 31–33 39 and two publications did not mention systemic complications.37 40 However, the primary study goal of these two case series was ocular complications.

Table 6

Rates of systemic adverse effects under bevacizumab/case series

### Methodological limitations

#### RCTs

The methodological quality of RCTs evaluating ranibizumab and bevacizumab is presented in table 7. In contrast to the pharmaceutical industry sponsored RCTs evaluating ranibizumab, the results of the RCTs evaluating bevacizumab are of limited value. The main limitations stemmed from the lack of any description as to how adverse effects were rigorously monitored, as well as the inadequate reporting of actual events. For example, none of the RCTs evaluating bevacizumab defined the method used to collect adverse effects data sufficiently or provided an adequate definition of expected adverse effects. In addition, in two of four RCTs evaluating bevacizumab the follow-up time was not sufficient to assess potential negative systemic effects, such as death or thromboembolic events (less than 6 months).22 23 In contrast, RCTs evaluating ranibizumab showed follow-up times of up to 24 months.6 12 Besides the above mentioned shortcomings, the sample size for bevacizumab treated patients was much lower than for ranibizumab treated patients (112 vs 941) and the number of received injections differed greatly (bevacizumab: 1–7 intravitreal injections per patient; ranibizumab: ≤24 injections per patient).

Table 7

Methodological quality of RCTs evaluating ranibizumab and bevacizumab

#### Case series

A critical assessment of the large number of published case series showed that no reliable conclusions on safety can be drawn using this study design (table 8). For example, of 11 prospective case series including 501 patients, only two evaluated more than 100 patients.30 35 However, the loss to follow-up was 60–70% and reasons for drop-outs were not given in these publications. In addition, more than half of the evaluated case series did not describe whether all of the originally included patients were considered in the results (transparency of patient flow not given).26 28 30 34 35 37–41 Similar to RCTs evaluating bevacizumab, only four 31 37 39 40 of 17 case series provided a definition of expected adverse effects, and three 27 39 41 (partly) defined the method used to collect adverse effects data. The currently available safety data from case series are—similar to the data from RCTs—further limited by the low number of received bevacizumab injections and short follow-up times.

Table 8

Methodological quality of case series evaluating bevacizumab

## Discussion

### Principal findings

In contrast to the RCTs evaluating ranibizumab, the trials evaluating bevacizumab were not sponsored by the pharmaceutical industry. However, they show—as described above—common methodological weaknesses (eg, short follow-up times, small sample sizes, and an inadequate reporting of adverse effects). Thus, the findings that intravitreal bevacizumab injections are not associated with major ocular or systemic adverse effects are not supported by reliable data from RCTs.

Beside RCTs, numerous case series evaluating bevacizumab have been published. Not surprisingly, they show major methodological weaknesses and are of limited validity. For example, the often cited retrospective multicentre PACORES study,42 which is the only publication that provides information on the rates of systemic adverse effects of intravitreal bevacizumab, reports ‘self-reported’ harm data. This can lead, as also discussed by the study authors, to an underestimation of adverse effects. Some case series showed a high loss to follow-up without giving reasons for drop-outs.30 35 However, a complete follow-up is necessary in order to determine if those patients who withdrew due to adverse effects are different from those who did not adhere. Taking also into account that intravenous bevacizumab for the management of colorectal cancer is associated with major systemic adverse effects, the low (or zero) rates for intravitreal bevacizumab are questionable—even though the dose of intravitreal bevacizumab is about 0.25% of that used for intravenous treatment.

### Strengths and weaknesses of this review

Both RCTs and non-RCTs (in particularly case series) were considered for the current review. We were interested in data from non-RCTs, because it is assumed that safety data from this study type are more reliable than data from RCTs.44 For example, in RCTs, data on adverse effects could be underestimated mainly due to the inclusion of highly selected (non-representative) patients and/or publication bias.45 In addition, small sample sizes limit the ability to detect rare but serious adverse effects.45 In contrast, non-RCTs often utilise large databases, therefore it is more likely that rare adverse effects for a wide range of patients can be detected with this study type.45 However, the current literature shows that in the case of bevacizumab in AMD, no well conducted non-RCTs with large sample sizes have been published.

### Other reviews

Up to now, ranibizumab and bevacizumab have been evaluated in several systematic reviews. However, the published reviews focused on the beneficial effect or clinical effectiveness of VEGF inhibitors without adequately addressing adverse effects.46–48 The reason is that due to an ongoing methodological debate about the assessment of adverse effects, the conclusions on safety are more complicated and need a very thorough and often time-consuming evaluation. However, it is obvious that the inadequacies of the bevacizumab safety data may potentially lead to situations where bevacizumab is used inappropriately or where patients are not fully informed of possible harm and potentially avoidable adverse consequences49—in contrast to the clear safety information that is available with the licensed substance which has been evaluated in phase III trials.

### Implications for clinical practice

This review highlights that the perceived low rates of adverse effects for bevacizumab are not supported by reliable data. The published RCTs and case series evaluating bevacizumab are of limited value, and therefore they cannot be used as substitutes for high-quality trials. In addition, higher evidence from ranibizumab trials suggests signals for an increased ocular and systemic vascular and haemorrhagic risk which warrants further investigation.

Results of ongoing head-to-head studies such as the IVAN study in Great Britain and the CATT study in the USA are in progress. Besides evaluating efficacy, these studies should have enough power to address major safety issues of bevacizumab compared to ranibizumab. Initial study results are expected to be available by 2011. In the meantime, patients and doctors should be aware of the insufficient safety data regarding intravitreal bevacizumab.

## Acknowledgments

We are grateful to Dr Gerta Ruecker for providing us with statistical support.

## Appendix A

### Search strategy in Medline (Ovid)

 # Search Results 1 Exp macular degeneration/ 8909 2 Exp retinal degeneration/ 19448 3 Exp neovascularisation, pathologic/ 23310 4 Exp eye/ 235449 5 Exp eye diseases/ 341820 6 Exp ophthalmology/ 14674 7 4 or 5 or 6 478892 8 3 and 7 6336 9 Exp retinal neovascularisation/ 1334 10 Exp choroidal neovascularisation/ 2078 11 Exp choroid/(blood supply) 7896 12 Exp Macula Lutea/ 7688 13 ((Macul* or retina* or choroid*) adj5 degener*).tw. 11098 14 ((Macul* or retina* or choroid*) adj5 neovasc*).tw. 4532 15 ((Macul* or retina* or choroid*) adj5 neo-vasc*).tw. 6 16 Maculopath*.tw. 2006 17 (Macul* adj2 lutea).tw. 88 18 amd.tw. 2942 19 cnv.tw. 2083 20 (Subfoveal neovasc* or subfoveal neo-vasc*).tw. 178 21 (Extrafoveal neovasc* or extrafoveal neo-vasc*).tw. 8 22 (Juxtafoveal neovasc* or juxtafoveal neo-vasc*).tw. 8 23 (Occult neovasc* or occult neo-vasc*).tw. 27 24 (Classic neovasc* or classic neo-vasc*).tw. 12 25 (Chor* neovas* or chor* neo-vas*).ot. 44 26 1 or 2 or 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22 or 23 or 24 or 25 41179 27 Exp Antibodies, Monoclonal/ 135039 28 Monoclonal antibod*.mp. 134118 29 Exp Angiogenesis Inhibitors/ 27451 30 Angiogenesis inhibit*.mp. 7033 31 Exp Endothelial Growth Factors/ 7888 32 ((Endothelial adj3 growth factor*) or ECDGF or endo-GF).mp. 24508 33 Exp Vascular Endothelial Growth Factors/ 17886 34 Vegf.mp. 17009 35 Bevacizumab.mp. (mp=title, original title, abstract, name of substance word, subject heading word) 1417 36 Avastin.mp. 289 37 Aanibizumab.mp. 142 38 Rhufab.tw. 4 39 Lucentis.mp. 34 40 Pegaptanib.mp. 171 41 Macugen.mp. 45 42 Or/27-41 236411 43 26 and 42 1893 44 Exp animals/ 13541475 45 Exp humans/ 10254739 46 44 not (44 and 45) 3286736 47 43 not 46 1406 48 Remove duplicates from 47 1379

View Abstract

## Footnotes

• Funding German health insurance fund (Verband der Ersatzkassen e. V. (vdek), Askanischer Platz 1, D-10963 Berlin, Germany).

• Competing interests None.

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

## Request Permissions

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