Objectives To assess the safety of intravitreal bevacizumab (IVB) as a monotherapy and to evaluate the relationship between quality of treatment and adverse events.
Data sources Cochrane Library, Ovid MEDLINE, MEDLINE in-process, Ovid EMBASE and Toxicology Literature Online (TOXLINE) from January 2009 to May 2012. Studies included in an earlier systematic review were also assessed for inclusion.
Study eligibility criteria, participants and interventions Randomised controlled trials (RCTs), controlled trials or observational studies including ≥10 participants reporting adverse events data following IVB monotherapy as a primary treatment in patients (aged 18 years or more) with any eye condition were included.
Study appraisal and synthesis methods Study selection was undertaken independently by a minimum of two reviewers using pre-defined criteria. Data abstraction and quality assessment were performed by one reviewer, and then checked by a second reviewer. Study quality was assessed for only RCTs in accordance to the Cochrane Risk of Bias Tool. Additional items relating to safety data were also assessed. Results were tabulated or meta-analysed as appropriate.
Results 22 RCTs and 67 observational studies were included. Only two RCTs reported valid safety data. Rates of serious adverse events following treatment were low. There was insufficient data to explore the relationship between the incidence of adverse events and quality of IVB injection.
Limitations A majority of relevant existing studies were characterised by small sample sizes, unclear diagnostic criteria and reporting of safety outcomes.
Conclusions and implications of key findings Available evidence demonstrates low rates of serious local and systemic adverse events following treatment. However, the role of IVB quality in the incidence of adverse events remains unclear. Robust evidence is needed to examine the relationship between the incidence of adverse events and variables such as injection techniques, pre-existing risk factors (eg, immunosuppression, cross-contamination) and quality of IVB treatment.
This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/
Statistics from Altmetric.com
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.
Strengths and limitations of this study
Eighty-nine studies of bevacizumab monotherapy in patients with diverse ophthalmic conditions were included.
A majority of relevant existing studies were characterised by small sample sizes, unclear diagnostic criteria and reporting of safety outcomes.
The relationship between the incidence of adverse events and variables such as injection techniques, pre-existing risk factors (eg, immunosuppression, cross-contamination) and quality of bevacizumab could not be explored due to limited data.
Age-related macular degeneration (AMD) and diabetic retinopathy (DR) have been identified as two of the three most common causes of age-specific visual impairment in England and Wales.1 More recently, effective treatment options have included anti-vascular endothelial growth factors (anti-VEGFs), which have been shown to both delay deterioration in vision as well as improve vision.2 Ranibizumab (Lucentis, Novartis) is licensed for the treatment of wet AMD and diabetic macular oedema (DMO) and costs £742.17 per injection (0.23 mL vial). Pegaptanib (Macugen; Pfizer) for treatment of AMD is available at a price of £514 per injection (300 μg vial). However, bevacizumab, which cost £242.66 for 4 mL/100 mg vial, is used as an unlicensed intervention in ophthalmic conditions. Although many doses for intravitreal administration can be produced from a single bevacizumab vial and therefore can be supplied for a much lower cost, the actual cost of dispensing smaller doses is uncertain. However, annual cost savings have been estimated if bevacizumab is used as standard treatment instead of ranibizumab in patients with AMD.3
Bevacizumab remains an unlicensed ophthalmic treatment for a number of reasons. There is an on-going debate with regard to intravitreal bevacizumab (IVB) use and its quality in clinical practice. One major concern relates to the risks associated with the reformulating the drug for intravitreal injections as well as possible adverse events (AEs) associated with systemically administered anti-VEGFs. Bevacizumab is reformulated for intravitreal use to deliver a smaller volume. However, the resulting reformulated product is considered by the Medicines and Healthcare products Regulatory Agency, a UK regulatory body for medicines and medical devices, as an unlicensed product. Another concern has centred predominantly on the possible risk of serious AEs such as endophthalmitis. To date, IVB safety evidence have been inconclusive.4 ,5 The aim of this review was to assess the safety, in terms of rates of specific serious AEs, of IVB monotherapy in ophthalmic conditions.
We updated an existing systematic review on AEs of intravitreal anti-VEGF reported by van der Reis et al,5 which searched reports from 1948 to 2009. We adapted the search strategy by including specific AE terms and omitting selected terms because the previous search strategy:
included fewer AE terms
used broad terms such as ‘cause’ and ‘response’ and
applied specific study design filters, for example, in vitro studies.
Free text and subject headings or thesaurus terms relating to the intervention (eg, bevacizumab, avastin) were combined with AE floating subheadings or specific AE terms. We searched the Cochrane Library; Ovid MEDLINE; MEDLINE in-process; Ovid EMBASE; and Toxicology Literature Online (TOXLINE) from January 2009 to May 2012 because this review was part of a project commissioned by the National Institute for Health and Care Excellence (NICE) through its Decision Support Unit (DSU) between April to August 2012. We did not search clinical trial registers. No experimental and functional study design filter or language restrictions were used. Our MEDLINE search strategy presented as an on-line supplementary file 1 was translated across different databases. Reference lists of all relevant studies and systematic reviews were checked and a citation search of relevant articles was also undertaken.
Study selection was undertaken independently by a minimum of two reviewers using pre-defined criteria. Any disagreements in the selection process were resolved by consensus or referral to a third reviewer. All published or unpublished randomised controlled trials (RCTs), controlled trials or observational studies including ≥10 participants reporting AE data following IVB monotherapy as a primary treatment in patients (aged 18 years or more) with any eye condition were included. Relevant comparators were limited to monotherapies for RCTs only. Articles were excluded if patients had received prior treatment or received IVB as an adjunctive treatment. Non-English reports, narrative reviews, editorials, letters or publications relating to preclinical and biological studies were also excluded.
Data extraction and quality assessment
Data extraction and quality assessment were performed independently by one reviewer. Disagreements were resolved by discussion with a second reviewer and if agreement could not be reached, a third reviewer was consulted. Where multiple publications of the same study were identified, data were extracted and reported as a single study. Information abstracted included study characteristics, participant details (eg, number of patients, eye condition, mean age and baseline comparability), intervention and comparator details (eg, source, dose, injection quality and frequency of treatment) and outcomes. Outcomes of interest were limited to important and serious ocular and systemic AEs as listed below:
Non ocular haemorrhage (gastrointestinal, pulmonary, other non-ocular bleeds)
Cerebrovascular accident (stroke)
Transient ischaemic attack
Infectious endophthalmitis (infection of the eye)
Retinal (pigment epithelium) tear
Anterior chamber reaction (including acute intraocular inflammation; uveitis; inflammation of the anterior chamber and hypopyon)
Lens damage/injury (including cataract, clouding of the lens)
Ocular hypertension (raised intraocular pressure >21 mm Hg)
For RCTs, study quality was assessed in accordance to the Cochrane Risk of Bias Tool. Additionally, we assessed items relating to safety data for RCTs; these included follow-up period greater than 6 months, definition of AE and description of method of ascertaining AE. A formal quality assessment was not undertaken for observational studies. While checklists exist for evaluating the methodological quality of a range of non-randomised studies, there is no consensus on how to incorporate a single tool to appraise different study types in a review.6 It was anticipated that a variety of non-randomised study designs would be identified, so criteria assessed were limited to study design (eg, prospective or retrospective), length of follow-up and baseline comparability when appropriate.
A pooled analysis was undertaken using the Cochrane Review Manager software where appropriate. The relative risk was calculated for dichotomous outcomes using a fixed effects model (Mantel-Haenszel method). Otherwise, descriptive statistics were tabulated. Estimates of AE rates were calculated by dividing the number of events by the number of patients who received IVB (event rate per patient) or the number of eyes treated (event rate per treated eye).
A flow chart of the study selection is shown in figure 1. Eighty-nine full text articles were included (n=22 RCTs, n=67 non-randomised studies). Of these 20 studies, including 1 RCT, were identified from the previous review. A total of 293 full text articles were excluded. Reasons for exclusion were wrong population, intervention or study type (n=162), unsuitable publication type (reviews, commentaries or editorials; n=34) and absence of usable data (n=97). A full list of excluded studies with reasons for exclusion is available on request.
A total of 22 RCTs comparing IVB with a variety of interventions as well as an observational control group with safety data were included, as presented in table 1.2 ,7–27 Study populations were patients with AMD (n=7 studies); DMO (n=8 studies); retinal vein occlusion (RVO) (n=4 studies) and other ophthalmic conditions (n=3). Assessment of study quality is presented as an on-line supplementary file 2. Study quality was considered to be moderate to low with only two RCTs16 ,19 meeting the criteria for valid safety data.
Sixty-seven observational studies were included as summarised in table 2. Most studies included patients with a single condition with fewer studies including a population with multiple conditions. Study quality was, generally, difficult to assess due to the quality of reporting. Approximately 65% of studies (n=44/67) were retrospective in design with follow-up periods of more than 6 months reported in less than a third (n=18/67) of included studies. Baseline characteristics of participants were comparable in two non-randomised studies28 ,29 and three case-control studies.30–32
Treatment schedule and source
Administration of 1.25 mg/0.05 mL was the most commonly reported dosage of IVB. Frequency of dosing and follow-up schedules varied across studies. Information relating to the source of IVB was reported in 35% (n=14/22) of RCTs but less than a fifth (19%; n=13/67) of observational studies.28 ,42–44 ,47 ,48 ,59 ,62 ,66 ,68 ,74 ,81 ,91 IVB was mostly provided by a local dispensing service such as the hospital's pharmacy. There were limited data to assess quality of administered IVB.
Reporting of AEs
Ascertainment of AEs was presented more objectively in RCTs compared to observational studies. Non-RCT evidence was unclear because several studies reported absence of events as ‘no serious complications’; or ‘no ocular complications’, or ‘no adverse events were observed’, thereby providing limited information on diagnostic techniques or criteria for reported AEs. Furthermore, for AEs such as visual loss, ocular haemorrhage, hypertension and hospitalisation, the relationship between the outcomes and treatment schedule or setting remained largely unclear.
AEs reported in RCTs
Pooled 1-year data16 ,19 indicated that the risk of death (RR 1.38; 95% CI 0.71 to 2.68) or arteriothrombotic events (RR 0.81; 95% CI 0.42 to 1.59) were not significantly different between patients with AMD who received IVB or intravitreal ranibizumab (IVR). Furthermore, no significant difference in death between the IVB and IVR arms was observed when the CATT16 (2 year data) and IVAN19 (1 year preliminary data) clinical trials were pooled to provide long-term data analyses as shown in figure 2. Cardiac disorders, transient ischaemic attack and hospitalisation for angina were not significantly different between patients with AMD treated with IVB and IVR.19 However, serious systemic AE rates remained significantly lower in the IVR group (n=1795, RR 1.27 CI 1.09 to 1.47).
Two smaller studies, Biswas et al15and Gharbiya et al,18 with safety data for patients with AMD reported no significant AEs. No significant differences were found for death and myocardial infarction (MI) in studies that compared IVB to pegaptinib26or sham injection7 (n=232 patients, RR 0.30; 95% CI 0.01 to 7.18).
Rates for endophthalmitis were not significantly different between IVB and IVR treatment groups for patients with AMD (1 RCT; RR 1.79; 95% CI 0.53 to 6.08).16 There were no reports of endophthalmitis,9 ,11 ocular hypertension,9 ,11 retinal detachment2 ,9 or vitreous haemorrhage9 ,11 in treatment groups comparing IVB with laser therapy in patients with DMO (n=269). In patients with DMO,21 ,22 ,25 ocular hypertension (IOP>21 mm Hg) was significantly higher in the IVT group (n=183; RR 0.13; CI 0.02 to 0.69) compared with the IVB group. A similar but non-significant trend was demonstrated in patients with RVO (n=32 patients; RR 0.08; CI 0.00 to 1.25).17
One short-term study at 3 months20 showed that posterior vitreous detachment was significantly higher in the IVB group compared with laser therapy (n=110; RR 17.00; CI 1.01 to 287.50). However, the rates of uveitis, vitreous haemorrhage, pigment epithelial tears and cataract progression were low and indicated no significant differences between IVB and laser therapy. No significant differences in rates of foveal haemorrhage13 (n=81; RR 0.62; 95% CI 0.28 to 1.35) or hyphema27 (n=26; RR 7.8; 95% CI 0.46 to 131.62) were found in patients with RVO who had IVB or sham injection.
AEs reported in observational studies
Table 3, summarising safety data reported in observational studies, displays extensive variation in the detail of reporting of AEs with most studies not reporting or observing AEs of interest. While high event rates were reported for hospitalisation, hypertension, anterior chamber reaction and visual loss, these rates need to be interpreted with caution due to previously mentioned issues with reporting along with likely confounders.
Systemic AEs reported included death (0.4–3.8%),40 ,48 ,54 ,94 arterial thromboembolism (0–1.4%),81 hypertension (0–15.6%),67 ,77 ,83 ,94 MI (0–8.2%),28 ,45 ,81 ,94 cerebrovascular accident (0–8.7%)39 ,45 ,54 ,83 ,94 and transient ischaemic attack (TIA) (0.4–1.0%).40 ,81
Visual loss was the most commonly reported ocular event,28 ,32 ,35 ,50 ,53 ,56 ,69 ,89 the definition of visual loss was often unclear and occasionally associated with AEs such as anterior chamber inflammation, severe intraocular inflammation or retinal detachment. Consequently, it is uncertain whether visual loss occurred as an AE from treatment or progression of the patient's condition. Infectious endophthalmitis was reported in 10 studies (range 0–1.0%). Three of the 13 studies28 ,42–44 ,47 ,48 ,59 ,62 ,66 ,68 ,74 ,81 ,91 in which patients received locally prepared IVB reported cases of infectious endophthalmitis. Reported rates were 0.02% (n=3/12 585 injections),47 0.2% (n=1/625)62and 0.8% (n=1/112).42 A higher rate of 0.9% (n=1/109) was reported in a study with IVB supplied by a compounding pharmacy.48 Positive cultures of microorganisms were reported in a study from Fong et al,48 and another study from Wu et al.94
Eighty-nine studies were included in this systematic review of AEs, 22 of which were RCTs. Trials compared IVB with a number of different therapies and eye conditions, though most were in AMD, DMO and RVO. Most ocular and systemic safety measures had zero events in treatment groups or were not significantly different between groups. The quality of reporting of studies made it impossible to evaluate the impact of both known and unknown confounding factors (eg, the use of prophylactic antibiotic eye drops) on the incidence of AEs.
The most robust data for safety are from the CATT16 and IVAN19 trials which were large trials that reported longer term data. The results of these trials when meta-analysed revealed a statistically significantly higher rate of 1 or more serious systematic AE (RR 1.27; 95% CI 1.09 to 1.47) in the IVB group. In this analysis, the IVAN study19 alone did not show a statistically significant difference while event rates were higher in the CATT. The recently published 2-year results of the IVAN study, which was not included in this review, has reported relatively worse safety outcomes for patients on discontinuous treatment compared to continuous treatment.97 In addition, there were no observed differences in mortality, frequency of thrombotic events or hospitalisation due to cardiac failure between groups of patients treated with IVB or IVR. Reported pooled analysis of the 2 years results of the CATT and IVAN studies tends to demonstrate that IVB and IVR are comparable in terms of safety. It is also important to note that AEs were more common in those patients who received discontinuous rather than patients on continuous treatment, that is, those with lower exposure to the drug experienced higher AE rates. An explanation for this observation is the possible role of immunological processes in drug interactions.97 It is also important to note that the CATT study demonstrated some imbalances at baseline between randomised patients which may need further exploration. More patients randomised to IVB had had a previous TIA compared to those in the IVR arms. Similarly, more IVB patients had a history of MI.16 Despite these caveats, these trial designs offer the most robust assessment of AEs to date.
Overall, the evidence on IVB safety from observational studies was uncertain. This has previously been reported elsewhere.4 ,5 Included studies were frequently associated with methodological weaknesses that limited the validity of the reported findings. The majority of studies were retrospective in design with small study samples or inadequate follow-up periods (less than 6 months). With respect to larger studies, observational data from Curtis et al.45 suggest no difference in the risk of AEs between IVB and IVR once socioeconomic confounders are accounted for. On the other hand, results of an unpublished study of Medicare patients funded by Genentech54 found an increased risk of stroke and death in IVB patients. The available abstract, however, did not provide sufficient information to allow an in-depth analysis of the results of this study. A recently published population-based, nested case-control study reported by Campbell et al98 (n=91 378) found no relationship between the risk of MI, venous thromboembolism, stroke or congestive heart failure and the administration of IVR or IVB. While the risk of systemic AEs was similar for both treatment groups, there was an increased risk of acute MI for a subgroup of patients with diabetes who received IVB.
This review highlighted the challenges of assessing the safety of IVB especially due to limited opportunities for in-depth detailed analyses of the relationship between IVB preparation and reported rates of infectious endophthalmitis. In the past, case reports have suggested contaminated batches of IVB as the primary source of infection; a published review of patient safety information held by the National Patient Safety Agency in England and Wales99 reported an increased risk of serious AEs including endophthalmitis following IVB treatment. The authors acknowledged that identifying the source of infection (ie, contaminated injection procedure or infected anti-VEGF) could be complex. However, Jonas et al,63 reporting on AE rates in a study population which included patients who had received IVB and IVT, suggested that event rates were statistically independent of drug injected (p=0.45), operating surgeon (p=0.18) and patient's age (p=0.87).
It is also important to highlight limitations of this review. By relying on the previous systematic review5 as a source of evidence, it is possible that studies that were not identified in that review may have been missed in this review. Our searches were undertaken up to 2012. An updated electronic literature search was conducted up to 23 May 2014, retrieving a total of 1300 records. A preliminary shift of titles resulted in 333 potentially relevant abstracts for further detailed examination. We would prefer to have undertaken a full update. Unfortunately, this is not possible for us at present due to lack of the extensive time and resources required. Although comprehensive and up-to-date systematic reviews are desirable, a recent analysis of a sample of systematic reviews showed that the median duration of survival indicating a requirement for an update was 5.5 years (95% CI 4.36 to7.67) in systematic reviews of randomised trials of procedures or conventional drugs.100 Furthermore, many RCTs randomised small numbers of participants and these may have been underpowered to detect differences in AEs.4 ,97 Generalisability of findings may also be limited due to differences between study participants and patients seen in routine practice. In addition, there were concerns relating to ascertainment of exposure particularly in observational studies.98 The influence of excluding non-English publications in this review is unclear. Additionally, adopting a narrow focus in the definition of AEs implies that data on less serious or rare events were not presented.
Overall, rates of serious AEs following IVB were low when compared to other intravitreal treatments, sham injection and laser therapy with relatively higher rates being reported in head-to-head studies of IVB versus ranibizumab. Most outcomes were, however, not significantly different between treatment groups. Current evidence from observational data still remains limited due to relatively small sample sizes, unclear definition, evaluation and reporting of safety outcomes as well as adequate follow-up periods. However, an opportunity to explore the relationship between the incidence of AEs and other variables such as injection techniques, pre-existing risk factors (eg, immunosuppression, cross-contamination) and quality of IVB could offer cost-saving options in providing treatment for certain ophthalmic conditions.
We would like to thank Jennifer Dunn for administrative support throughout the project.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Contributors EP coordinated the systematic review and was responsible for the conception and design, acquisition of data, analysis and interpretation of data and drafting and revision of the final manuscript. JR was responsible for the conception and design, acquisition of data, analysis and interpretation of data and drafting and revision of the final manuscript. RW was responsible for the developing and undertaking the electronic literature searches and drafting and revision of the final manuscript. EE-H and ME were responsible for acquisition of data, analysis and interpretation of data and drafting and revision of the final manuscript. AP contributed to the conception and design of the study, assisted with data collection analysis, interpretation of data and drafting and revision of the final manuscript. AW contributed to the conception and design of the study, assisted with analysis and interpretation of data and drafting and revision of the final manuscript.
Funding This manuscript is based on a report which was funded by the National Institute for Health and Care Excellence (‘NICE’) through its Decision Support Unit.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Extra data relating to references of excluded full-text articles is available by emailing Edith Poku (email@example.com).