Introduction

Wet age-related macular degeneration (AMD) is a chronic, progressive disease of the central retina and a major cause of irreversible vision loss worldwide.1, 2 Central to the pathogenesis of the disease is the overexpression of vascular endothelial growth factor (VEGF), which stimulates choroidal neovascularization and causes blood and fluid to leak into the macula. The last decade has seen the introduction of intravitreal anti-VEGF agents, which have revolutionized the treatment of wet AMD, offering patients previously unachievable improvements in vision.

Ranibizumab (Lucentis), a humanised monoclonal antibody fragment, was the first anti-VEGF agent shown to improve visual acuity (VA) in patients with wet AMD.3, 4, 5 Regulatory approval for the use of ranibizumab was granted on the basis of the MARINA4, 6, 7 and ANCHOR3, 8 study data, which showed mean gains of 7–11 letters over 12 months with monthly dosing. Up to 40% of patients gained more than 15 letters during this time and few lost vision. Moreover, improvements were largely maintained at the 24-month follow-up.

The early clinical trial data clearly demonstrated the benefits that ranibizumab could offer; however, the requirement for monthly intravitreal injections places a high burden on patients and healthcare systems. Subsequent studies, therefore, investigated whether equivalent VA gains could be achieved with less frequent injections. In the PIER study, patients received three monthly loading doses followed by quarterly injections of 0.5 mg ranibizumab.9 However, there was a mean loss of 0.2 letters by month 12. Other studies used pro re nata (PRN) regimens, where the decision to administer the drug is dependent on the disease status assessed by the physician at regular monitoring visits (eg, vision worsening or increase in macula thickness). Outcomes of these PRN regimens are variable with mean gains of 2.3–9.3 letters.10, 11, 12, 13, 14, 15, 16 For example, in the SAILOR study, which utilised a PRN regimen based around quarterly monitoring visits following three initial monthly doses, there was a mean gain of only 2.3 letters.10 By contrast, the results from the PrONTo13, 14 study (PRN regimen with monthly monitoring visits following three initial monthly doses) were more promising with VA gains approaching MARINA and ANCHOR results (9.3 letters over 12 months).3, 4, 6, 7, 8 It seems that only regimens with frequent monitoring and strict retreatment criteria can achieve the visual outcomes anywhere near those seen with fixed monthly dosing.

In the meantime, ranibizumab became available for use in routine clinical practice and its effectiveness in real life started to be evaluated. The last few years saw several studies report that outcomes achieved with ranibizumab in clinical practice failed to match the efficacy observed in the early ranibizumab clinical trial programme upon which its license was granted. These include WAVE,17 HELIOS,18 LUMIERE,19 AURA,20 and the MEDISOFT database.21 Interestingly, these studies showed VA gains of 3.8–6.7 letters during the most intensive treatment period (loading period), but unlike MARINA and ANCHOR, these initial visual outcomes were not maintained over time. In fact, most of these studies saw a change in mean VA at 12 months of −1 to 3.2 letters.17, 18, 19, 20, 21 These poor visual outcomes may stem from an inability to adhere to strict a PRN regimen (which is required for PRN to be effective) in routine clinical practice; this hypothesis is supported by the low injection frequency reported by these studies (a mean number of 4.3–5.1 injections over 12 months).17, 19, 20, 21

To better understand the overall picture concerning the real-world effectiveness of ranibizumab, a review of the literature was conducted to identify studies of ranibizumab in clinical practice. This review provides both a summary of the design and methodological quality of the studies and a basic analysis of VA outcomes from the studies.

Materials and methods

Search criteria

We conducted a systematic search of the PubMed database for English language peer-reviewed papers published before 1 December 2014. The search protocol and structure of the review was based on that used by the Cochrane Collaboration. Search terms were as follows: (1) (long-term OR real-life OR longitudinal OR cohort OR clinical experience OR open-label OR real-world OR database OR non-interventional OR non-interventional OR observational) NOT (randomised[ti] OR randomized[ti]); and (2) (wet AMD OR AMD OR exudative AMD OR neovascular AMD); and (3) (ranibizumab[ti] OR Lucentis[ti]).

Further studies to be considered for inclusion were identified by reviewing the references lists from the studies identified during the PubMed search and by assessing relevant papers already known to the authors.

Inclusion and exclusion criteria

Studies considered for inclusion were screened for eligibility in three sequential stages, by the review of: (1) the title; (2) the abstract, and (3) the full manuscript (Figure 1). At each stage, the rationale for retention or rejection of the study was recorded.

Figure 1
figure 1

Study selection flow chart.

Inclusion criteria were as follows: non-randomised controlled studies where treatment was given according to the licensed posology (monthly injections until there are no signs of disease activity and/or maximum VA is achieved, followed by PRN), and where baseline VA and change in VA at 12 months were reported. There were no restrictions on geographic region, clinical characteristics, baseline VA or previous treatment.

Exclusion criteria were as follows: studies that were primarily focused on specific sub-types of wet AMD (eg, polypoidal choroidal vasculopathy or retinal angiomatous proliferation), and studies specifically evaluating switch from other agents (bevacizumab and aflibercept). Follow-up and extensions to Phase III studies were not included as these were not considered real-world studies.

Data analysis

Key outcomes were mean change in VA from baseline and the proportion of patients gaining ≥15 letters or losing ≤15 letters at 12 months. Simple descriptive analyses of these data, including grand means (mean of the means of the several studies) and weighted means (mean weighted against number of eyes per study), were used to pool VA outcomes across the total population. Additional sub-analyses were used to compare visual outcomes according to patient previous treatment history (naive vs non-naive patients), type of study (prospective vs retrospective), and approach to missing data (last observation carried forward). For the purpose of our analysis conversion to Early Treatment Diabetic Retinopathy Study letters (ETDRS) letters was carried out in studies providing other measures of VA (eg, logMAR).

Results

Study selection

Figure 1 shows how the studies were selected for inclusion in our analysis. The PubMed search identified a total of 197 citations. Upon screening using our pre-defined inclusion criteria, 143 were discarded after review of the publication titles, 27 were discarded after the review of publication abstracts, and a further 8 were discarded after the review of full text. No additional studies were identified through the review of reference lists within the publications we identified; however, we were aware of an important study (AURA)20 that was not identified by the pre-defined search terms but was relevant to this review. Therefore, 20 studies in total were included in our evaluation.

Study characteristics

The characteristics of the 20 real-world studies of ranibizumab included in this review are summarised in Table 1. Most studies were conducted specifically to evaluate the effectiveness of intravitreal ranibizumab for the treatment of wet AMD in a real-world setting; however, the design and endpoints employed to achieve this aim were variable (Table 1 and Supplementary Table 1). Retrospective studies were the most common, with only six having a prospective design. The strengths and weaknesses of the individual studies are summarised in Table 2.

Table 1 Study characteristics: aims, design, and treatment protocols in real-world studies of ranibizumab for the treatment of wet AMD
Table 2 Strengths, weaknesses and approaches to missing data in real-world studies of ranibizumab in wet AMD

The size of the studies varied widely, with the number of eyes ranging from 5422 to 12 951.21 In total, 23 261 eyes were included, of which 18 358 had 12-month follow-up data. Approximately half of the studies considered only one eye per patient (generally the first treated eye).

Mean VA at baseline ranged from 48.8 to 61.6 letters, excluding the studies by Shona et al,32 Pushpoth et al,26 and Nomura et al,28 who reported VA outcomes for groups of patients stratified according to different subcategories (eg, baseline VA) but did not report a study mean. The grand mean (mean of the means of the several subsamples) was 54.6±7.3, and when weighted according to the number of eyes in the study was 54.1.

Two of the 20 studies did not provide any information regarding previous treatment history.20, 23 Eleven studies enroled only treatment-naive patients (Table 1), while seven could have included patients who received previous treatment for wet AMD. Previous treatments included photo dynamic therapy (PDT) (n=3)22, 24, 25 and anti-VEGF intravitreal injections (n=4).17, 18, 26, 27

Most studies (n=18) were conducted in a single country, of which 15 were European, 1 was Australian,24 and 2 were Japanese.22, 28 The country with the largest number of studies (n=7) was the UK. The remaining two studies included in the analysis were multinational.

Change in VA

The grand mean (±SD) VA change from baseline to 12 months reported in the 20 studies was 2.9±3.2 letters; however, when weighted according to the number of eyes evaluated, the change was 1.95 letters (Figure 2). Fourteen studies reported an improvement in VA, with a maximum mean gain in VA recorded in a single study of 5.5 letters.29 Two studies reported a decline in visual outcomes with the largest decline in VA of 2.0 letters.30 In the remaining four studies the VA at study end remained similar to baseline values (−0.8 to 0.97 letters).

Figure 2
figure 2

Mean change from baseline in visual acuity and number of injections in real-world studies and pivotal randomised controlled studies after 12 months of ranibizumab treatment in patients with wet AMD. *Represents average data for the 20 studies included in the current review (see Table 3).

VA gains in the 12 studies of anti-VEGF-naive patients were slightly greater than that for the full study population, with a grand mean of 3.5±3.9 letters and a weighted mean of 3.5 letters.

Retrospective studies found greater mean gains than prospective studies (grand mean of 3.5±3.5 vs 1.3±1.5 letters; weighted mean of 2.5 letters vs 0 letters). Studies that used a last observation carried forward (LOCF) approach to their data analysis (n=16) reported lower VA gains on average than the full study population (grand mean gain at 12 months of 3.1±3.5 letters; weighted mean 1.9 letters).

Patients gaining 15 or more letters at 12 months

Eleven studies (including 3869 eyes with 12-month data) reported the percentage of patients gaining ≥15 letters. At 12 months, the means was 19±7.5%. Weighting for the number of eyes available for analysis had little effect (Figure 3a).

Figure 3
figure 3

(a) Percentage of patients gaining ≥15 letters. *Represents average data for the 11 studies included in the current review which reported this outcome (Table 3). (b) Percentage of patients losing <15 letters. *Represents average data for the 12 studies included in the current review, which reported this outcome (Table 3).

Similar results were observed for the studies of anti-VEGF-naive patients. On average 19±8.4% of these patients gained ≥15 letters (13–24%), with a weighted mean of 20%.

No noteworthy changes were seen in studies reporting LOCF, when compared with the overall analysis.

In retrospective studies, a greater percentage of patients gained ≥15 letters than in prospective studies (grand mean of 19±8.3% and 16±1.4%). Weighting for the number of eyes had little effect with mean values changing to 20% and 15.5%, respectively.

Patients losing 15 letters or less at 12 months

Twelve studies (including 12 467 eyes with 12-month data) reported the percentage of patients losing ≤15 letters. The means±SD at 12 months was 89±6.5% (74.4–97.4%) and was similar to the weighted mean (Figure 3b).

On average, 91±5.9% of anti-VEGF-naive patients lost ≤15 letters. Again, weighting for the number of eyes had little effect (89%).

No noteworthy results were observed in studies reporting LOCF.

Similar results were observed between retrospective and prospective studies regarding the percentage (grand mean) of patients losing ≤15 letters (90±5.2% vs 92±7.4%). No noteworthy differences were found when these changes were weighted for the number of eyes (at 12 months).

Number of injections over 12 months

The mean number of injections ranged from 4.2 to 7.5 (Table 3), with a grand mean (±SD) of 5.5±0.8. Anti-VEGF-naive patients received a similar number of injections to the full study population (grand mean of 5.4±0.7). Weighting for the number of eyes at 12 months had little effect on the outcomes of these analyses.

Table 3 Effects of ranibizumab on key visual acuity outcomes in real-world studies of wet AMD

Similar number of injections were received by patients in the retrospective and prospective studies (grand mean number of injections of 5.6±0.7 and 5.1±1, respectively). In this analysis, weighting according to the number of eyes assessed at 12 months gave different results, showing that patients treated in prospective studies received one less injection on average than those evaluated by retrospective studies (4.5 vs 5.7 injections).

Other outcomes

Ten studies reported mean changes in central retinal thickness, six reported safety data and three reported vision-related quality of life. A detailed overview of these and other outcomes is provided for reference in Supplementary Table 1.

Discussion

This review was performed in order to determine whether, in patients with wet AMD, real-world outcomes achieved with ranibizumab match the positive outcomes reported in the pivotal ranibizumab trials,3, 4, 6, 7, 8, 9 and subsequent studies of 0.5 mg PRN regimens.10, 11, 12, 13, 14, 15, 16

The grand mean (+2.9 letters) and weighted mean (+1.95 letters) calculated from the changes in VA reported at 12 months in the trials included in this review are considerably lower than those reported in the pivotal studies of ranibizumab (fixed monthly dosing),3, 4, 6, 7, 8 with the exception of the PIER study9, 31 (fixed quarterly dosing) that showed a mean loss of 0.2 letters. Specifically, the recently published AURA study conducted in multiple countries and considered a true and current depiction of the real-world use of ranibizumab for the treatment of AMD, clearly demostrated that the VA outcomes are far worse than the ones reported in the landmark trials (MARINA and ANCHOR).3, 4, 6, 7, 8 Rather, they are more in line with those observed in PIER9, 31 and SAILOR,10 which used less intensive treatment schedules (fixed quarterly dosing and PRN dosing based on a quarterly monitoring schedule, respectively), and compare unfavourably with MARINA4, 6, 7 and ANCHOR.3, 8

What are the factors that underlie the reduced effectiveness of ranibizumab in clinical practice? A key consideration may be the reduced number of injections that patients receive in clinical practice. Across the 20 real-world studies included in the present review, the mean number of injections ranged from 4.2–7.5; the grand mean at 12 months was 5.5±0.8 injections and the weighted mean was 5.4 injections, notably less than the 12 injections received in ANCHOR/MARINA3, 4, 6, 7, 8 (fixed monthly dosing). A reduced number of injections in the real-world setting is not unexpected as adherence to treatment regimens tends to be harder to maintain owing to the challenges of implementing intensive treatment regimens outside of the clinical trial setting. However, the reduced efficacy of ranibizumab in real-world practice observed in our study cannot be explained simply by the reduced number of injections, as the pooled mean values were: (1) less than the 0.5 mg PRN arms of CATT11, 12 and HARBOUR (6.9 and 7.7 injections, respectively),15, 16 which saw greater increases in VA; (2) similar to PrONTO (5.6 injections),13, 14 which demonstrated a greater increase in VA; and (3) greater than in SAILOR (4.6 injections),10 which showed a similar increase in VA as observed in the present analysis (see Figure 2 for change in VA from baseline).

Besides the reduced number of injections, the schedule on which they are given as well as the monitoring frequency and retreatment criteria might also be a reason for the limited efficacy of ranibizumab in the clinic. The studies in this and previous analyses typically employed PRN regimens as a means of appropriately managing treatment burden. By definition, a PRN regimen only treats patient with symptomatic disease; hence, it is possible that recurring fluid may cause progressive damage. Results from CATT11, 12 and HARBOUR15, 16 show that it is possible to achieve good visual outcomes with a PRN regimen when frequent regimented monitoring visits and strict retreatment criteria are in place. However, with less intensive monitoring, as seen in the SAILOR study (quarterly monitoring schedule)10 and in routine clinical practice as per the ranibizumab label, visual outcomes using PRN are poor.

Recently, the treat-and-extend regimen has emerged as a potential alternative method of delivering the best-possible visual outcomes, with the appropriate injection frequency, while minimising the treatment burden. With treat-and-extend, the physician can identify the appropriate injection frequency for each patient, proactively injecting at each visit and deciding when to treat next rather than whether to treat at that time. Thus, treatment remains ahead of the disease. In the future, it will be interesting to evaluate whether this newer treatment approach can improve outcomes in the real-life situation.

Baseline VA is an additional determinant of outcome in terms of VA gains (ceiling effect), with patients with poorer vision tending to show more gain. However, the mean baseline acuity of the studies included in the present analysis is comparable to that reported in the clinical trials (respective ranges 48.8–61.6 vs 47.1–61.5) with the grand mean (54.6) and weighted mean (54.1) lying in the centre of these values. Therefore, it is difficult to ascribe the poorer outcomes observed in the present analysis to baseline acuity alone.

As with any review of this type, it is important to consider the limitations of the current analysis. By their nature, real-world studies (non-randomised and not controlled studies) are not as rigorously designed as clinical trials and therefore require careful interpretation. In particular, 13 out of the 20 included studies, including the largest study by Tufail et al,21 were retrospective. Data may be incomplete, with patients being lost to follow-up or not having outcomes recorded at appropriate time points. Some studies employed an LOCF method to account for these missing data; although a standard and valid technique, LOCF analysis may skew study results particularly in conditions that change over time, for example, wet AMD. Morevover, differences in VA measurements in real-lide studies vs clinical trials (Snellen vs ETDRS) could also be considered as a limitation.

In addition, the inclusion criteria vary across studies included in the present analysis, with many including patients irrespective of their treatment history or baseline VA. Such variability can affect outcomes. For example, patients with good baseline acuity demonstrated a smaller improvement in response to ranibizumab than patients with poor baseline acuity,29, 32 while treatment-naive patients showed a better response than previously treated individuals.26 Differences such as these may obscure the effects of treatment if the numbers of patients from a particular group are unrepresentative of the population as a whole. However, with nearly 20 000 patients analysed, the real-life data reviewed here could be considered as representative of current practice.

From an analytical perspective, the validity of data pooling across studies with different designs is inherently limited. The strengths and weaknesses of each of the 20 real-world studies included in the review, along with a description of the approaches adopted to missing values in the studies, are summarised in Table 2.

Finally, only peer-reviewed studies indexed on PubMed were included. It is possible that additional relevant studies not listed on PubMed are not captured here.

In conclusion, the current review strongly suggests that, in patients with wet AMD, VA outcomes achieved with ranibizumab in clinical practice do not match those reported in pivotal studies. Although factors such as reduced injection frequency and patient heterogeneity are likely to have a part in this phenomenon, further research is required to determine the reasons for this discrepancy. Nonetheless, clinicians should be aware of this discrepancy when consenting patients for receiving ranibizumab for wet AMD.