Description of the condition
Neovascular age-related macular degeneration (AMD) is a progressive and degenerative disease of the retina in which blood vessels grow aberrantly into the subretinal, intraretinal, and intrachoroial space. Without treatment, its natural course will result in a fibrous scar that greatly diminishes central visual capacity. In industrialized countries, late AMD is the leading cause of legal blindness in elderly populations. The estimated prevalence of any stage AMD across numerous population–based studies showed a pooled prevalence to be 8.69% (95% credible interval (CrI) 4.26 to 17.40), and of late AMD to be 0.37% (CrI 0.18 to 0.77), with a higher prevalence in Europeans than in Asians and in Africans (e.g. late AMD: 0.50%, 0.28% and 0.37%, respectively) (Wong 2014). Across ethnic groups, the main risk factor of AMD is age: the late stage ranges from 0.13% (95% CrI 0.08 to 0.18%) among people aged 50 to 59 years to 3.25% (95% CrI 2.21 to 4.60%) among people aged 80 years or older. In the UK, the estimated annual incidence of neovascular AMD is 2.3 (95% CrI 1.4 to 4.0) per 1000 women and 1.4 (95% CrI 0.8 to 2.4) per 1000 men (Owen 2012). In the year 2040, global projected cases of any AMD are 288 million (95% CrI 205 to 399), with the largest number of cases in Asia and Europe (Wong 2014).
Description of the intervention
Vascular endothelial growth factor (VEGF) is an angiogenic cytokine that promotes vascular leakage and growth. Its signalling is over expressed in neovascular AMD as well as in some tumours (e.g., colorectal cancer). Accordingly, VEGF inhibitors such as bevacizumab and ranibizumab have been used to block its pathological angiogenesis. Bevacizumab is approved by drug regulatory authorities for intravenous use as a cancer therapy, whereas ranibizumab is approved for intravitreal use in the treatment of neovascular AMD.
Bevacizumab and ranibizumab derive from the same anti-VEGF mouse monoclonal antibody (Ferrara 2006), but differ in the monoclonal antibody fragment and glycosylation of proteins (Meyer 2011). In intravitreal injections, both drugs can enter systemic circulation, although bevacizumab, a full-length antibody, exhibits a longer half-life (Avery 2014). The shared molecular structure and pharmacological profile of the two drugs, as well as bevacizumab’s therapeutic utility (Braithwaite 2014), have led to the widespread off-label/unlicensed use of bevacizumab to treat neovascular AMD as a less expensive alternative to ranibizumab (Miller 2013).
The approved dosage of ranibizumab as indicated in the “Summary of Product Characteristics” is 0.5 mg. The dosage of bevacizumab most recommended by ophthalmologists for intravitreal injection is 1.25 mg, approximately 400-fold less than the intravenous dose used in cancer therapy (Schmucker 2010). In fact, when administered intravitreally, only a small fraction of bevacizumab enters systemic circulation (Kim 2009; Krohne 2008; Stergiou 2011). Retreatment regimens include continuous (i.e., monthly injections) and discontinuous treatment, including “as needed” (Pro Re Nata (PRN)) and “treat-and-extend” regimens in which the drug is injected less frequently as long as there is no recurrence of neovascular manifestations.
Why it is important to do this review
The use of bevacizumab or ranibizumab relies on evidence of superiority of one drug over the other. A number of factors will influence the decision, including different profiles for effectiveness, ocular and systemic adverse events, resource use, and the feasibility of the intervention in practice. There is evidence that bevacizumab is associated with a relative improvement in best corrected visual acuity (BCVA) similar to that of ranibizumab (-1.15 letters, 95% confidence interval (CI) -2.82 to 0.51). BCVA outcome favoured continuous treatment regimen when compared with discontinuous regimen (-2.23 letters, 95% CI -3.93 to -0.53) (Chakravarthy 2013).
There is limited evidence that bevacizumab might be associated with higher ocular adverse events (risk ratio (RR) 3.1; 95% CI 1.1 to 8.9); however, the absolute rates of serious ocular adverse events (e.g., endophthalmitis, uveitis) were low (≤ 2.1%) (Schmucker 2012), and may have been unrelated to the drugs, but to the procedure (i.e., intravitreal injection). Most discordant judgments between the two drugs focused on differences in their systemic safety.
VEGF has important growth promoting and maintenance roles in a variety of cells and tissues, raising concern that these agents may interfere with normal physiology and worsen conditions such as coronary or peripheral arterial diseases. VEGF inhibition in cancer patients treated systemically with bevacizumab, at higher dosages than are given in neovascular AMD, was found to increase the risk of fatal events due to haemorrhage (RR 2.77, 95% CI 1.07 to 7.16), neutropenia (RR 2.37, 95% CI 0.61 to 9.18), and gastrointestinal perforation (RR 2.45, 95% CIs 0.63 to 9.51) (Ranpura 2011). Accordingly, concerns have been expressed about the systemic safety of anti-VEGF drugs, even at the small doses delivered with intravitreal injection (Lim 2011).
Comparison of AMD Treatments Trials (CATT), the first published large randomised controlled trial (RCT) comparing ranibizumab directly with bevacizumab (CATT 2011), reported a statistically significant finding of an excess of serious systemic adverse events (e.g., life-threatening or resulting in significant patient disability) related to bevacizumab when compared with ranibizumab (RR 1.29, 95% CI 1.01 to 1.66). The publication of results from a second large RCT, Inhibition of VEGF in Age-related Choroidal Neovascularisation (IVAN) (Chakravarthy 2013), and the concomitant meta-analysis of the CATT and IVAN safety results at two years prompted by the IVAN and CATT data monitoring committees, again, showed that the sum of all serious adverse events differed by treatment regimen. However, when serious adverse events were compared by organ system class or by specific adverse events (e.g. arterial thrombotic event), there were no differences between the drugs. These results and their varying interpretations fuelled medical and health-policy debates on the off-label use of bevacizumab as a far less-costly alternative to ranibizumab.
To assess the systemic safety of intravitreal bevacizumab compared with intravitreal ranibizumab in people with neovascular AMD.
Criteria for considering studies for this review
Types of studies
We will include head-to-head RCTs comparing bevacizumab and ranibizumab in people affected by neovascular AMD. We will include trials irrespective of the dosage, whether treatment is continuous or discontinuous, or duration of follow-up.
Types of participants
We will include people affected by neovascular AMD irrespective of age, sex, or progression of the condition.
Types of interventions
We will compare the systemic safety of intravitreal bevacizumab (brand name Avastin®; Genentech/Roche, South San Francisco, CA) with ranibizumab (brand name Lucentis®; Genentech/Roche, South San Francisco, CA). For this review, we will not consider placebo-controlled trials and trials comparing other anti-VEGF agents approved for neovascular AMD (e.g., pegaptanib and aflibercept). These studies can contribute to a large network meta-analysis that simultaneously summarizes direct evidence (which comes from studies directly randomising treatments of interest) and indirect evidence (which comes from studies comparing treatments of interest with placebo) (Salanti 2008). We will be completing the network meta-analysis as a second phase of this project, to address the issue of the relative effectiveness and safety across a network of RCTs testing anti-VEGF agents.
Types of outcome measures
Primary outcome domains will include:
(1) All-cause mortality.
(2) All serious systemic adverse events (hereinafter referred to as All SSAEs), the sum of individuals affected by one or more SSAEs recorded in a trial. The International Conference on Harmonisation Good Clinical Practice (ICH GCP) Guideline defines SSAEs as medical occurrences that result in death, are life threatening, require hospital admission or prolongation of hospital stay, cause persistent or significant disability/incapacity, or are medically important events or reactions (ICH 2014). We will accept the definition of SSAE adopted by the study authors, but recognize that some studies may not have adopted the ICH GCP Guideline.
Secondary outcome domains will include:
(1) Myocardial infarction (MI).
(3) Arteriothrombotic event, defined as any participant who have experienced at least one of the following events: a) myocardial infarction, b) non-haemorrhaging stroke, c) angina, d) ischaemic heart disease, e) thrombosis, or f) death from cardiovascular diseases.
(4) Serious haemorrhage as defined by each study, including, but not limited to cerebral, pulmonary, and gastrointestinal haemorrhage (these are usually defined as a haemorrhage that is associated with anaemia, transfusion, haemostatic intervention, hospitalisation, or fatal bleeding).
(5) Serious neutropenia as defined by each study (these are usually defined as neutropenia of grade 3 and 4 associated with sepsis and life-threatening infections) (National Cancer Institute 2003).
(6) Gastrointestinal perforation.
(7) Serious infection as defined by each study, including, but not limited to, pneumonia, lung abscess, and pyothorax (these are usually defined as an infection associated with the use of intravenous antibiotic, hospitalisation, intubation, or death). We will exclude ocular infections.
(8) Treatment-related drug discontinuation.
(9) SSAEs classified according to the Medical Dictionary for Regulatory Activities System Organ Classes (MedDRA SOCs) (version 17.0) (ICH 2014), including: benign, malignant, or unspecified neoplasms; cardiovascular disorders; gastrointestinal disorders; general disorders and administration site conditions; infections and infestations; nervous system disorders; and respiratory, thoracic and mediastinal disorders.
(10) Serious adverse events previously associated with drugs affecting the VEGF pathway (i.e., arteriothrombotic events, systemic haemorrhage, congestive heart failure, venous thrombotic events, hypertension, and vascular death).
The SSAEs classified by MedDRA SOCs differ from our primary outcome All SSAEs as they explore specific subsets of SSAEs, providing the opportunity to explore the biological plausibility of each.
Search methods for identification of studies
We will systematically search MEDLINE, EMBASE, and The Cochrane Central Register of Controlled Trials (CENTRAL) using and updating the search strategy prepared for an update of a Cochrane systematic review on antiangiogenic drug effectiveness for neovascular AMD (Vedula 2008). RCTs that investigated an anti-VEGF treatment compared to another treatment, sham treatment, or no treatment will be eligible. We will also search clinical trial registers, including the metaRegister of Controlled Trials (mRCT), ClinicalTrials.gov, and the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) to identify other ongoing studies or completed studies that have not yet been published. For unpublished RCTs, we will search the Internet for pre-publication study presentations at conferences or meetings and contact study authors seeking information on safety data.
Searching other resources
We will search the reference lists of relevant studies to identify additional studies.
Data collection and analysis
We will collect and combine outcomes at the maximum follow-up times reported up to a maximum of two years, since this corresponds to the maximal cumulative drug dosage. We will conduct a pre-specified subgroup analysis to explore SSAE outcome at the one-year follow-up.
Selection of studies
Two investigators (EL and GV) will independently screen the titles and abstracts of studies identified through the literature searches and additional sources. We will retrieve and independently assess the full-text or unpublished reports using predefined inclusion criteria. We will resolve discrepancies through discussion and, when necessary, by consulting an additional investigator (LM).
Data extraction and management
Two investigators will independently extract data (EL and GV) on study characteristics and enter data into RevMan (RevMan 2012). Three investigators will extract data on primary and secondary outcomes (EL, GV, and KK).
Assessment of risk of bias in included studies
We will assess the risk of bias in each included study following the criteria outlined in Chapter 8 of theCochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), which addresses the following key domains: randomisation sequence generation; allocation concealment; masking (blinding) of participants, trial personnel, and outcome assessors; incomplete outcome data; selective outcome reporting; and other sources of bias (e.g., early termination of a trial due to benefits or the failure of authors to disclose the study’s source of financial support).
We will evaluate additional risk of bias items specific to adverse events using the following items: (1) adverse event definition: if the definition of adverse events were pre-specified in the protocol and collected based on standard criteria or classification system (e.g., MedDRA SOCs) and (2) method of adverse event assessment: if the researchers actively monitored for adverse events or simply provided spontaneous reporting of adverse events that arose during the study.
If the information in published study reports or unpublished supporting documents (e.g., meeting abstracts or presentations) is partial or inadequate to assess the risk of bias, we will contact the study authors for clarification. If the authors do not respond, we will assess the risk of bias based on the available information.
Measures of treatment effect
We will use the risk ratio (RR) to estimate the relative treatment effect of the two anti-VEGF drugs.
Unit of analysis issues
Individual participants will serve as the unit of analysis. Since repeated SSAEs can occur in the same participant, we will consider the number of individuals with at least one SSAE, rather than the number of SSAEs. However, the individual-level analysis might decrease the statistical power of meta-analyses.
Dealing with missing data
In all studies, we will carry out the analyses, as far as possible, on an intention-to-treat basis. In other words, we will attempt to include all participants randomised to each group in the analyses, irrespective of the treatment received or if the participants completed the study follow up (Akl 2013). As a second option, we will collect data on participants who received at least one dose of study medication, as treated analysis.
If there is a discrepancy between the number randomised and the number analysed in each treatment group, we will calculate and report the percentage lost to follow-up in each group. Where data are inadequate to assess the extent of SSAEs, we will contact the principal investigators of included studies to supply any unreported data. Where it is not possible to obtain information on missing data, we will record this in the data collection form and report it in the 'Risk of bias' table. We will further discuss the extent to which the missing data could alter the results/conclusions of the review.
Assessment of heterogeneity
We anticipate that a low number of RCTs in a pairwise comparison (resulting from either the paucity of studies in the field or incomplete reporting) will prevent the formal assessment of statistical heterogeneity. Nevertheless, we will calculate the Chi
Assessment of reporting biases
We anticipate that a low number of RCTs in a pairwise comparison will also prevent the formal assessment of publication bias.
In this context, we reasoned that there may be true differences across the population of potential studies as they may have enrolled participants at different risk levels for adverse events. For instance, some studies might have included participants at high risk for arteriothrombotic events while others may have excluded such participants. For this reason, we will use a Mantel-Haenszel random-effects model for meta-analyses, which provides a robust estimate when pooling sparse data (Robins 1986). We will not formally adjust for multiplicity of comparisons, but will consider this issue when interpreting the analyses.
Summary of Findings Table
We will summarise the strength of evidence for all-cause deaths, All SSAEs, serious systemic infections, arterial thromboembolic events, myocardial infarctions, and strokes, using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) methodology (Guyatt 2008). An iterative electronic correspondence discussion process will be used to reach a consensus on: a) the factors that affect confidence in the estimate of effects, including risk of bias (i.e., design and study limitations), imprecision, indirectness (directness in the GRADE approach includes generalizability and applicability), inconsistency of results (i.e., heterogeneity), magnitude of effect, and issues of residual plausible confounding and b) the rating of the evidence. Because imprecision of RRs is an expected limitation when investigating deaths and adverse events in RCTs, which might be rare, we will follow the GRADE guidelines for assessing this quality item (Guyatt 2011a; Guyatt 2011b). We will focus on the 95% CI around the absolute effects, considering that a follow-up between one and two years will be sufficient. We will adopt a minimal difference of 1% for deaths and 5% for All SSAEs as clinically relevant. If the 95% CI includes treatment effects above these thresholds and, therefore, does not exclude an absolute detrimental effect appreciably less than 1% and 5%, we will evaluate the precision as insufficient and will downgrade the quality of the overall evidence. We are adopting these thresholds based on rating imprecision guidelines (Guyatt 2011b) as well as our own judgement. We are setting a clinical decision threshold boundary of 1% absolute difference as we reasoned that this difference is important to both patients and health systems. For example, if the absolute difference in death rates between ranibizumab and bevacizumab is very small (absolute difference of 0.3% with a 95% CI ranging from −0.1% to 0.7%), the results of the meta-analysis will exclude an important difference favouring either drugs, and we will not downgrade the quality of evidence for mortality. We will adopt two different imprecision thresholds 1% or 5% because the differing importance of the outcome mortality or serious adverse events is influencing our judgement.
We will present the overall evidence in a Summary of Findings table with summary estimates of absolute and relative effects and their quality according to the GRADE methodology (Guyatt 2013). For each outcome, we will categorize our confidence in the estimate of effect as one of four levels, ranging from very low to high.
Subgroup analysis and investigation of heterogeneity
We do not plan subgroup analysis for the current review.
We will compute a Mantel-Haenszel RR using a fixed-effect model to investigate any influence of small study effects on the pooled RR, since the random-effects model tends to attribute greater weight to small studies with increasing heterogeneity (Sterne 2011).
We will perform a sensitivity analysis, excluding unpublished data, as well as a leave-one-out meta-analysis to assess the independent influence of each study on the summary estimate (Tobias 1999).
The Cochrane Eyes and Vision Group (CEVG) will execute the searches. We thank Sheila Bird, Hugh McIntyre and Tasanee Braithwaite and Jennifer Evans (CEVG Editor) for their comments on this protocol as well as Toby Lasserson and Orla Ni Ogain (from the Cochrane Editorial Unit).
Contributions of authors
All authors contributed to the protocol.
Declarations of interest
Sources of support
- No sources of support supplied
- Emilia Romagna Regional Health Service, Italy.This systematic review has been funded by the Emilia Romagna Regional Health Service to the Italian Cochrane Centre. The funding source had no role in the development of the systematic review or the writing of this manuscript.
- Italian Ministry of Health, Italy.Lorenzo Moja is a recipient of a Research Early Career Award from the Italian Ministry of Health on "Improving appropriateness and transparency of processes to develop guidelines on controversial clinical areas: evidence, values and context preferences to help mitigate disputes and enhance the applicability of recommendations to practice" (GR-2011-02348048).