Age-related macular degeneration (AMD) is the leading cause of severe blindness in the elderly in Western countries [11, 13]. Exudative macular degeneration is the most severe form with the highest risk of blindness [13]. Vision loss is the result of macular haemorrhage and fluid caused by leakage from subfoveal choroidal neovascularisation (CNV), which is the hallmark of exudative AMD. Vascular endothelial growth factors (VEGF) have a key role in the formation and leakage of CNV [28]. Recently, the anti-VEGFs pegabtanib and ranibizumab became available for the treatment of exudative AMD [12, 36, 38, 43, 45]. In contrast to pegabtanib, intravitreal injections with ranibizumab led to a significant vision improvement. In fact, ranibizumab was the first drug to improve vision in exudative AMD, compared to interventions used earlier which only delayed progression [12, 38]. Major disadvantages of this intervention are the costs and the need to give intravitreal injections repeatedly. Bevacizumab, which is approved for the treatment of colon cancer but not for the treatment of AMD, is closely related to ranibizumab. In contrast to pegaptanib that binds only the VEGF isoforms that are 165 kD (VEGF165) and larger [26], both ranibizumab and bevacizumab bind and inhibit all biologically active forms of VEGF [47]. Bevacizumab is a full-length, humanized monoclonal antibody against VEGF, whereas ranibizumab is a humanized antigen binding fragment against VEGF, and both proteins were genetically engineered from the same murine monoclonal antibody against VEGF. A major advantage of bevacizumab is its price, which is about 1–5% the price of ranibizumab. Mainly for this reason, it is now used worldwide and on a large scale off-label for the treatment of exudative AMD. Its off-label use is obviously controversial, also because the use of anti-VEGF therapy is associated with an increased risk for thromboembolic events. Until now, no systematic review has shown evidence for the therapeutic effect of bevacizumab, or its safety. In this systematic review, the effects of bevacizumab on visual acuity (VA) and central retinal thickness (CRT) are quantified, and the adverse events are reported. Moreover, the various therapeutic strategies are described.

Material and methods


The databases searched were MEDLINE, EMBASE, and the Cochrane database. The reference lists of included articles and personal files were also searched until no new articles were found. Search terms for the application in age-related macular degeneration were “Macula*” or “AMD” or “ARMD” or “intra(-)vitreous” or “intra-vitreal” in any field, and for the intervention the search terms were “Bevacizumab” or “Avastin” in any field. All years were included up to March 2008. Published articles on paper, or papers available electronically before publication on paper were included. Language restrictions were English, German, French or Dutch. Two observers, independently of each other, searched the acquired list to identify relevant articles.


Articles included were randomised controlled trials, non-randomised controlled studies, or before-and-after studies in more than one patient. Studies with systemic or intravitreal therapy were included. Studies were excluded that did not have VA as the primary outcome, had as the primary objective to study differences between subgroups, or included also patients other than patients with exudative AMD. All articles selected by at least one observer were copied or printed. The ultimate selection was based on a discussion between observers.

Data abstraction and validity assessment

Two observers read all the articles and retrieved the data. These observers were not masked for the journal or any other aspect of the article. Data on inclusion criteria, exclusion criteria and adverse events were copied from a HTLM version of the article and edited for the table. One other observer was given only the text of the method section to score quality items. This observer was also given a separate results section to score the baseline and change in ETDRS score and CRT. This observer was blinded for the results of the other observers. Differences between observers were discussed until a consensus was reached. The items scored for quality of the follow-up studies were: prospective follow-up, consecutive case series, loss to follow-up, masked assessment of the outcome, standardized assessment of the outcome and other interventions during the study period. [39] The baseline value of the VA and CRT and these scores at last examination or at the examination with complete follow-up, when possible, were noted.

Study characteristics and outcome

Data included were author and year, type of study and quality issues, inclusion criteria, exclusion criteria, details of the treatment (systemic or intravitreal therapy, dosage, frequency of dosing and criteria for additional injections), number of patients at baseline and at follow-up and follow-up time, baseline ETDRS and CRT and changes in these outcomes and adverse events. The quality of the before-and-after studies was scored according to the following criteria often used to score quality of follow-up studies: prospective study, consecutive cases, loss to follow-up, blind assessment of outcome, standardized assessment of outcome. The criterion “no other intervention” was omitted, since studies that included such intervention were omitted from inclusion.

The quality of the RCTs was scored with the Delphi list, supplemented with criteria from the Dutch Cochrane Centre. The complete list of criteria is shown in Table 2 [44].

Quantitative data analysis

The VA score was converted to the ETDRS score when the ETDRS score was not assessed directly. We used a conversion of 0.1 logmar (one Snellen line) to 5 ETDRS letters (one ETDRS line). The change in ETDRS score was calculated as the difference between baseline value and the value at the last examination, or at the examination when there was no loss to follow-up. A summary for change in ETDRS and CRT was calculated as the weighted mean of the changes. The number of patients at the follow-up examination determined the weight. Calculations were made separately for intravenous injections and intravitreal injections. The RCT arms that included the treatment with only bevacizumab were included in the before-and-after studies. A subgroup analysis was conducted separately for several characteristics for the follow-up studies, which included the following: total quality score, dosage of intravitreal bevacizumab, previous treatment in more than 50% of patients, size of the study, duration of follow-up, and method of assessment for visual acuity (ETDRS or Snellen). The study quality items and results of the RCTs were also presented. The results of the RCTs were presented as described after conversion to ETDRS scores.


The total number of articles retrieved was 561. A large number of articles (n = 535) were not included in the review because of several reasons (Table 1). Ultimately, 26 out of these 561 were suitable and used in the analysis. Three randomised controlled trials were published and presented separately [9, 27, 31]. We also combined in this review the data of the patients of these three RCTs who had received only intravitreal bevacizumab, and the data of patients of 23 before-and-after studies with follow-up of patients who had received bevacizumab for exudative AMD (Tables 2, 3, 4 and 5)[14, 710, 1518, 23, 25, 27, 29, 3035, 37, 40, 41, 48]. A total of 1,435 patients included at baseline in these studies were included in this review.

Table 1 Flow chart for the selection of articles
Table 2 Study quality scores (according to Delphi list supplemented with criteria from the Dutch Cochrane centre) of three randomized controlled clinical trials that included one treatment group with intravitreal bevacizumab
Table 3 Results of three randomized controlled trials that included one group that received intravitreal bevacizumab
Table 4 Study characteristics of 23 studies on the effect of bevacizumab in age-related exudative AMD
Table 5 Summary of exclusion criteria of the 23 studies on the effect of bevacizumab in age-related exudative AMD

The three RCTs all showed that bevacizumab is more effective than PDT (with or without triamcinolone) [9, 27, 31]. The RCTs were of a poor-to-reasonable quality and lack of masking was the main methodological problem.

Two articles showed the results of the same study [34, 35]. Of these two, the article with the largest series of patients and longest follow-up period was included in this review [35]. In this study the effect of intravenous bevacizumab in AMD was studied. Two other studies included patients who had received intravenous bevacizumab [10, 23]. The data of these three studies were included in the summary of the effect of intravenous injections. After intravenous bevacizumab the weighted mean change in ETDRS score was +12.8 (range +11 to +14) and the weighted mean change in CRT was −129 μm (range −100 to −202).

Twenty-three articles, including the patients in the RCTs who received bevacizumab, reported the change in VA after intravitreal bevacizumab (see Table 2) [14, 9, 1518, 23, 25, 2733, 37, 40, 48]. The weighted mean change after intravitreal injections was 8.6 (range 2 to 26), and for the CRT −90 (range −46 to −190). The data are summarised in Tables 5 and 6. Inclusion criteria varied, and were composed of: age range, VA range, recent disease progression, poor response to previous treatment including PDT or pegaptanib, not eligible for PDT, previous treatment, and type of CNV. A summary of the various exclusion criteria in these studies is displayed in Table 5. Mostly, intravitreal injections of 1.25 mg bevacizumab were given. The strategy for additional injections differed and two main strategies could be discerned: 1) monthly injections irrespective of the results, or 2) additional injections when vision loss reoccurred or when macular fluid reoccurred on OCT. The mean number of patients included at baseline who received intravitreal injections per study was 61 (range 10–266). The weighted mean follow-up time was 15 weeks (range 4–48). The ocular or systemic adverse events that were reported in a total of 1,435 patients who had received an intravitreal injection, and a total of several thousand intravitreal injections, are summarized in Table 6. Table 7 summarizes the incidence of reported presumed adverse events.

Table 6 Reported results of 23 studies on the effect of bevacizumab in exudative AMD
Table 7 Reported adverse events in the 23 follow-up studies with intravitreal injections of bevacizumab for exudative AMD, in a total of 1,396 patients who received one or more injections

Studies differed in quality (Tables 4 and 8). The difference in change in VA between subgroups according to differences in quality was no more than 1.9 letters (Table 8). The change in ETDRS score was 2.7 letters higher for the studies with a higher quality. No study had a blinded assessment of the outcome, except for one study (but blinded only for the assessment of the CRT [10]. The highest difference between two groups was for those who had a higher dosage compared to those with a lower dosage (Table 8).

Table 8 Change in ETDRS score and central retinal thickness after intravitreal bevacizumab stratified according to different study characteristics


This systematic review shows that after administration of bevacizumab in patients with exudative age-related macular degeneration, an improvement in both the VA and CRT are reported. Moreover, bevacizumab is more efficacious than PDT. The weighted mean increase in VA after intravitreal bevacizumab injections in 1,396 patients is 8.6 letters on the ETDRS score. The weighted mean decrease in CRT is 90 μm. The incidence of adverse events was low. All studies showed an impressive effect, and the three RCTs showed superiority of bevacizumab to PDT with or without intravitreal triamcinolon.

The observed change in the follow-up studies of bevacizumab is similar to the change observed after monthly intravitreal injections of ranibizumab in exudative age-related macular degeneration. The mean change in ETDRS letters after monthly injections of 0.5 mg ranibizumab varied between +5.9 for occult or minimally classic CNV and +9.8 for classic CNV after 3 months [13, 38]. The observed increase in VA is also closely similar to the increase of 10.8 ETDRS letters after ranibizumab three times and additional injections when necessary based on OCT and VA change [21]. It is interesting to see that in the 1-year follow-up study of Bashur et al., fewer injections were needed than in the 1-year follow-up study on ranibizumab of Fung et al., with almost the same criteria for reinjection, knowing that the intravenous half-life of bevacizumab is 150% that of ranibizumab in rabbit eyes [5, 6, 8, 21]. However, the dosage in Bashur’s study was 2.5 mg bevacizumab.

The similarity in effect between bevacizumab and ranibizumab is also supported by a non-randomised, single centre comparative study in which these two drugs were compared [42]. Baseline patient characteristics were similar in both groups of 44 eyes in the bevacizumab group and the 53 eyes in the ranibizumab group. After 113 days of follow-up and a mean of 1.8 injections in the bevacizumab group the mean increase in VA was 0.8 lines. After 44 days of follow-up and a mean of 2.0 injections of ranibizumab, the mean increase in VA was 0.7 lines (p = 0.87). The improvement in VA is also observed in a study that compared starting with bevacizumab to starting with pegaptanib followed by bevacizumab [19]. In the latter group, no statistically significant difference from baseline was observed.

Adverse events were rare in the 1,396 patients who received a total of several thousand intravitreal injections. Moreover, some adverse events are the result of the procedure (endophthalmitis, retinal detachment, cataract progression), others are mild (vitritis), or an event may also occur as a result of the natural course of AMD itself or with other therapies such as photodynamic therapy and ranibizumab [14, 24]. Although the reported thromboembolic events could have been caused by bevacizumab, the time frame between injection and the event, the presence of additional risk factors, the age of the population, or the increased risk of cardiovascular diseases in patients with AMD make this less likely. On the other hand, patients at risk of a thrombo-embolic event were excluded in some studies.

These results confirm the results of an internet surveillance program in which the follow-up for adverse events in 5,228 patients and 7,113 intravitreal injections with bevacizumab from 70 centres in 12 countries was reported [20]. Adverse events reported were corneal abrasion, lens injury, endophthalmitis, retinal detachment, inflammation or uveitis, cataract progression, acute vision loss, central retinal artery occlusion, subretinal haemorrhage, retinal pigment epithelium tears, blood-pressure elevation, transient ischaemic attack, cerebrovascular accident and death. None of the adverse event rates exceeded 0.21% [22]. These results are also confirmed by a 12-month follow-up study of 1,265 consecutive patients with various diagnoses, including exudative AMD, who received 4,303 intravitreal injections of bevacizumab [46]. Systemic adverse events occurred in eighteen patients (1.5%). These were: acute elevation of blood pressure (0.59%), cerebrovascular accidents (0.5%), myocardial infarction (0.4%), iliac artery aneurysms (0.17%), toe amputation (0.17%) and deaths (0.4%). Ocular adverse events were bacterial endophthalmitis (0.16%), tractional retinal detachment (0.16%), uveitis (0.09%), rhegmatogenous retinal detachment (0.02%) and vitreous haemorrhage. Some of the adverse events may have been the result of the underlying disease. The type and incidence of the reported adverse events in the bevacizumab studies does not seem to be very different from the reported type and incidence from two large RCTs of ranibizumab [12, 38]. The frequency of injections in the two ranibizumab RCTs was monthly for a period of 1 year, while in the studies included in this review only a single intravitreal injection, or three monthly injections with or without repeated injections were given. This treatment will ultimately lead to fewer injection-related adverse events in the latter case.

It is known that Snellen and ETDRS visual acuities don’t match exactly. One shortcoming of this review, therefore, was that Snellen acuities as reported in the various studies were converted to ETDRS scores using a conversion of 0.1 logmar (one Snellen line) to 5 ETDRS letters (one ETDRS line). On the other hand, for patients with macular degeneration in the lower vision range, it was recently shown that visual acuity measurement with Snellen charts may underestimate VA as compared to measurements with ETDRS charts [19]. This implies that the visual acuity results of this systematic review, which included several studies using Snellen charts only, may underestimate visual acuity improvement after bevacizumab as compared to the ETDRS assessments in the ranibizumab trials. [20]

We conclude that bevacizumab is effective in improving VA in exudative AMD, and that its effect is likely to be equivalent to ranibizumab. According to current reports severe adverse events for bevacizumab are rare in the short term.