Introduction

Retinal vein occlusion is a frequent vascular disease that often leads to severe visual impairment. One of the main reasons for visual loss is the development of macular edema (ME). Treatment options include grid laser treatment (for branch retinal vein occlusion), intravitreal injection of steroids, surgical procedures and, most recently, off-label treatment with intravitreal anti-VEGF agents [14]. Vascular endothelial growth factor (VEGF) has been shown to play an important role in the pathophysiology of ME by inducing blood–retinal barrier breakdown and increased vascular permeability. Intravitreal levels of VEGF are significantly correlated with the severity of ME [5, 6]. During recent years, intravitreal anti-VEGF treatment with bevacizumab has been shown to efficiently reduce ME and improve visual acuity (VA) in numerous case series and prospective or retrospective studies. As treatment effect wears off after a certain time interval, a high percentage of patients suffer recurrences of macular edema and need retreatment. This study aims to evaluate the outcome of patients with a follow-up of 6 months to 2 years who were treated with an OCT-guided reinjection scheme.

Subjects and methods

Sixty-one patients with ME secondary to retinal vein occlusion were enrolled in the study (61 eyes: 27 with central retinal vein occlusion, 34 with branch retinal vein occlusion). Twenty-eight patients were female and 33 were male. There was an even gender distribution in branch retinal vein occlusion (BRVO) patients, whereas in central retinal vein occlusion (CRVO) patients 59% were male. All patients had a minimum follow-up of 25 weeks, with a mean follow-up of 60 ± 29 weeks (range: 25–128). Mean VA at baseline was 0.18 (0.75 logMAR ± 0.38) in patients with CRVO and 0.32 (0.50 logMAR ± 0.29) in patients with BRVO. Mean age was 70.3 ± 8.0 and 66.5 ± 12.3 years in CRVO and BRVO patients respectively. Duration of retinal vein occlusion symptoms varied considerably between the patients, with an average of 12 ± 10 weeks in CRVO and 37 ± 77 weeks in BRVO patients. Main inclusion criteria were ME involving the centre of the fovea with a minimum central retinal thickness (CRT) at baseline of ≥250 µm. All patients were naïve to treatment of ME. No previous vitreoretinal surgery, intravitreal injections or laser treatment had been performed. Patients with neovascularisation at baseline were not included. Table 1 summarises the demographic data of CRVO and BRVO patients.

Table 1 Baseline characteristics of patients with central retinal vein occlusion (CRVO) and branch retinal vein occlusion (BRVO)

Intravitreal injections of bevacizumab (2.5 mg/0.1 ml) were performed under sterile conditions. Written informed consent was obtained, and patients were informed about the off-label use of bevacizumab and potential side-effects. This investigation followed the tenets of the Declaration of Helsinki. Follow-up examinations were carried out every 6–8 weeks and included ophthalmologic examination, best-corrected visual acuity (BCVA) testing (ETDRS chart) and optical coherence tomography (OCT). OCT analysis was performed with Stratus OCT 3 (Carl Zeiss Meditec, Jena, Germany), using the cross-sectional scan modus of six individual scans. Criteria for reinjection were increasing or persistent ME in at least one of the scans. If none of the scans revealed any sub- or intraretinal fluid, no reinjection was performed. The criterion for reinjection was not central retinal thickness, but morphologically visible edema in terms of intra- or subretinal fluid. If ME decreased but was still present, a reinjection was given. The minimum time interval between two injections was 6-8 weeks.

Main outcome measures were VA and CRT, which was measured manually using the calliper tool of the OCT software. For statistical analysis, CRVO and BRVO patients were divided into three subgroups according to treatment response as judged by OCT. Group 1 was defined as patients who had dry OCT findings at last visit, without any recurrence of ME within the last 25 weeks. Group 2 included patients who responded to treatment, with complete resolution of fluid at one or more follow-up visits, but had experienced a recurrence of ME within the last 25 weeks or at last visit. Group 3 included all patients who had never shown a complete resolution of ME at any study visit.

Statistical analysis was performed with SAS 9.1 (SAS Institute GmbH, Heidelberg, Germany). Follow-up and baseline data were compared using the paired t-test. Comparison of baseline parameters between subgroups was done with the Chi-square test for categorical variables and analysis of variance (ANOVA) for continuous variables. For multivariate analysis, a logistic regression model was utilized. Pearson correlation coefficients were calculated to determine the influence of baseline variables on final VA. Continuous variables are presented as mean values ± standard deviation (SD). The level of statistical significance was set at p < 0.05.

Results

Bevacizumab treatment resulted in significant reduction of ME and improvement of VA in CRVO and BRVO at last visit (Fig. 1a,b).

Fig. 1
figure 1

Central retinal thickness (a) and visual acuity (b) at baseline (white) and last visit (grey) in patients with CRVO and BRVO. Significant changes are marked with asterisks (**p < 0.01, ***p < 0.001)

Patients with CRVO (n = 27) received an average of 4.1 ± 2.2 injections. Mean CRT decreased from 748 ± 265 µm to 373 ± 224 µm (p < 0.001), and VA improved by 1.9 ± 3.2 lines, from 0.18 (0.75 logMAR ± 0.38) to 0.27 (0.57 logMAR ± 0,48), (p < 0.01). At last visit, 44.4% of patients with CRVO gained three lines or more.

Among CRVO patients with complete resolution of ME at one or more study visits, 47% showed a complete resolution after the initial injection. Another 32% had a complete first-time resolution of ME after the second injection. Only 21% of patients who developed dry OCT findings during follow-up needed more than two injections for an initial resolution of fluid. These four patients achieved a first-time resolution of fluid after three, four, five or seven injections respectively. If a complete resolution of fluid occurred after the first injection, mean time to second injection was 14.4 ± 4.8 weeks.

Patients with BRVO (n = 34) received an average of 4.9 ± 2.9 injections. Mean CRT decreased from 602 ± 207 µm to 386 ± 178 µm (p < 0.001), and VA improved by 1.8 ± 2.6 lines from 0.32 (0.50 logMAR ± 0.29) to 0.48 (0.32 logMAR ± 0.21), (p < 0.001). At last visit, 38.3% of patients with BRVO gained at least three lines. All BRVO patients with complete resolution of ME at one or more study visits showed a complete first-time resolution after the first injection (77%) or after the second injection (23%). No BRVO patient without complete resolution of ME after the first or second injection achieved dry OCT after subsequent injections. If a complete resolution of fluid occurred after the first injection, mean time to second injection was 20.7 ± 18.4 weeks.

As response to therapy with bevacizumab differed individually between the patients, we divided them into three subgroups (Fig. 2a,b) according to treatment response as judged by OCT (see subjects and methods section).

Fig. 2
figure 2

Central retinal thickness (a) and visual acuity (b) at baseline (white) and last visit (grey) of CRVO and BRVO subgroups. Significant changes are marked with asterisks (*p < 0.05, **p < 0.01, ***p < 0.001)

Thirty-three percent (n = 9) of CRVO patients did not suffer a recurrence of ME within the last 25 weeks, and were categorized group 1. These patients received an average of 2.6 ± 2.0 injections. Follow-up after last injection without recurrence of ME was 67.6 ± 27.1 weeks (range: 25–101). CRT declined significantly from 590 ± 270 µm to 248 ± 69 µm and VA improved by 4.2 ± 2.6 lines. At last visit, 44.4% of group 1 CRVO patients had a VA of 20/25 or better.

Thirty-seven percent (n = 10) of CRVO patients had a recurrence of ME at last visit or within the last 25 weeks (group 2). They received an average of 5.1 ± 2.4 injections. Although 50.0% of these patients showed a recurrence of ME at last visit, CRT had reduced significantly from 771 ± 213 to 278 ± 132 µm and VA improved by 1.5 ± 3.3 lines.

Thirty percent (n = 8) of CRVO patients did not show a complete resolution of ME at any follow-up visit (group 3). These patients received a minimum of three injections. The average number of injections was 4.5 ± 1.4 (3–7). A reduction of ME from 878 ± 268 µm to 633 ± 222 µm was achieved, but VA did not change significantly, by −0.4 ± 1.9 lines. Only 12.5% of these patients had a VA of 20/40 or better at last visit, and 37.5% were worse than 20/200.

In BRVO patients, 15% (n = 5) were classified group 1. They received an average of 2.0 ± 2.1 injections. Follow-up after last injection without recurrence of ME was 50.4 ± 20.5 weeks (range: 28–71). CRT declined significantly from 493 ± 125 µm to 217 ± 46 µm, and VA improved by 1.6 ± 2.1 lines. At last visit, 60.0% of the patients had a VA of 20/25 or better.

Fifty percent (n = 17) of BRVO patients were categorized group 2. They received an average of 4.6 ± 3.1 injections. Of these patients, 70.6% showed a recurrence of ME at last visit. Nevertheless, CRT had reduced significantly from 624 ± 225 to 382 ± 155 µm (BRVO), and VA improved by 2.5 ± 2.3 lines.

Thirty-five percent (n = 12) of BRVO patients did not show a complete resolution of ME at any follow-up visit (group 3). The average number of injections was 6.5 ± 1.7 (range: 4–11). Although a significant reduction of ME from 615 ± 206 µm to 463 ± 200 µm was achieved, VA did not change significantly by 0.8 ± 0.8 lines. In contrast to the unfavourable visual outcome of group 3 CRVO patients, 50% of BRVO patients in this subgroup had a VA of 20/40 or better, and no eyes were worse than 20/200.

When therapy was stopped in BRVO and CRVO patients of subgroup 3, patients were discharged from the study. At the time of this analysis, the last injection was no longer than 8 weeks ago. No patient developed neovascularisation during therapy.

Comparison of baseline parameters between the three subgroups

Further analysis was carried out in order to determine differences in baseline parameters between the three groups.

There was an equal gender distribution in all subgroups of CRVO and BRVO with only one exception: nine of ten patients of CRVO group 2 were male (p = 0.043). Although significant, this was attributed to the small number of patients in this subgroup, and not regarded as a true difference.

Univariate analysis (ANOVA) was performed to determine differences in continuous baseline parameters like patients’ age, baseline VA and baseline CRT between the three groups. CRVO patients of group 1 had a thinner CRT at baseline than patients of groups 2 or 3 (p = 0.059). When just group 1 and 3 were compared, the difference was statistically significant (p = 0.021). There was no difference in age or baseline VA between the three groups.

Multivariate analysis by logistic regression analysis was performed to investigate the joint effect of patients’ age, gender, baseline VA and baseline CRT on treatment outcome as defined by the three subgroups. Analysis revealed a significant influence of age (p = 0.049) and especially of baseline CRT (p = 0.0036) in CRVO. Old age and a high CRT at baseline were associated with a bad prognosis. Gender and baseline VA did not have a significant effect on treatment outcome in patients with CRVO.

In BRVO patients, the investigated baseline parameters did not differ between the three subgroups, and did not show any significant association with treatment outcome.

Prognostic value of response to first injection

In order to evaluate the prognostic value of the response to the first injection, we compared the number of patients with a complete resolution of ME after the initial injection in groups 1 and 2. Patients of group 3 by definition never had a complete resolution of fluid, and were not included in this analysis. In patients with CRVO, response to the initial injection seems to be predictive of long-term outcome. Of group 1 patients with CRVO, 77.8% had a resolution of fluid after the first injection, compared to only 20.0% of group 2 patients (p = 0.01). In patients with BRVO, 100.0% of group 1 had dry OCT findings, compared to 70.6% of group 2. This difference was not statistically significant.

Correlation of final VA with baseline parameters and final CRT

In patients with CRVO, final VA was strongly correlated with baseline VA (r = 0.743, p < 0.0001). Age and final CRT were also significantly correlated with final VA. Old age and bad VA at baseline can be regarded as risk factors for a bad final VA. These two parameters are not completely independent factors, as age and baseline VA significantly correlate with each other (0.441, p = 0.021).

In BRVO correlation of final VA with initial VA was less pronounced than in CRVO. The strongest relationships with final VA showed the patients´ age and final CRT.

In both groups, CRVO and BRVO, baseline CRT did not significantly correlate with final VA (Table 2).

Table 2 Correlation of final visual acuity (VA) with baseline factors and final central retinal thickness (CRT). Table shows Pearson correlation coefficients (bold) and p-values

Correlation of change of VA with change of CRT

Change of central retinal thickness ΔCRT (CRTbaselinel - CRTfinal) and visual acuity ΔVA (VAbaseline - VAfinal) was calculated for CRVO and BRVO. A decrease in CRT was significantly correlated with an improvement of VA in BRVO (r = 0.541, p = 0.001). There was no correlation in patients with CRVO (r = 0.171, p = 0.39).

Discussion

Patients with ME due to retinal vein occlusion benefit from treatment with 2.5 mg bevacizumab. In BRVO and CRVO treatment leads to a highly significant reduction of CRT and improvement of VA. The beneficial effect of treatment with bevacizumab has been shown in numerous retrospective and prospective studies [712].

Not all patients benefit equally from treatment with bevacizumab. Whereas some patients show a longstanding improvement after a certain number of injections, others suffer regular recurrences of ME or fail to show significant improvement. To determine the proportion of patients who benefit from treatment, we categorized patients with CRVO or BRVO into three subgroups according to treatment response as judged by OCT.

Thirty-three percent of CRVO and 15% of BRVO patients show a longstanding effect of treatment after an average of 2.4 ± 1.7 injections without any recurrence of ME during the last 25 weeks (group 1). Thirty-seven percent of CRVO and 50% of BRVO patients respond to treatment with a complete resolution of ME at a minimum of one follow-up visit but show only short periods of improvement and suffer frequent relapses (group 2). In 30% of CRVO and 35% of BRVO eyes, only a partial reduction of ME can be achieved without complete resolution of fluid and no significant improvement of VA occurred (group 3). As all patients were scheduled every 6–8 weeks, some patients of group 3 might have experienced a temporary improvement of VA before the next follow-up visit, which was not considered in this analysis.

Multiple regression analysis revealed that CRVO patients with favourable response to treatment were younger and especially had less severe ME at baseline. There was no difference between the 3 subgroups regarding gender and baseline VA. Interestingly, in BRVO patients none of the baseline parameters investigated such as age, gender, CRT and VA differed significantly between the subgroups.

In eyes with CRVO response to first injection seems to be predictive of long-term treatment result. Seventy-eight percent of patients with CRVO who did not suffer a recurrence of ME during the previous 6 months (group 1) showed a complete resolution of ME after the initial injection, compared to 20% of patients with short-term improvement (group 2). In BRVO response to first injection was very high in both group 1 and group 2, but did not differ significantly between the two groups, maybe due to the small number of patients in group 1. On the other hand, persisting fluid after the first injection might be regarded as a bad prognostic sign in BRVO. The prognostic importance of the initial injection is also supported by the fact that only 21% of CRVO and none of the BRVO patients with dry OCT findings during their clinical course did not already respond to the first or second injection with a complete resolution of fluid. Hence, only a small proportion of patients without complete resolution of fluid after the first two injections will sufficiently respond to further injections. The prognostic value of the response to the initial injection is also supported by findings of other authors, who reported a correlation of CRT at 1 week or 1 month after the initial injection with CRT at 1 year in CRVO patients treated with ranibizumab [13].

The prognostic importance of a favourable response to the initial bevacizumab injection in patients with BRVO has also been described by Chung et al. concerning visual function. They categorized patients into two groups according to change in VA at last visit. Of patients with 5 or more ETDRS letters gain at last visit, 78.6% showed a significant improvement in VA after the initial injection, whereas 72.7% patients with less than 5 letters gain in VA did not improve significantly after the initial injection [7].

The definition of subgroups in our study was done by OCT findings according to anatomical results of treatment. But does morphological improvement reflect functional improvement?

In our study population, patients of group 1 had a better final VA than patients of group 2, who in turn had a better VA than group 3. Several authors described the close relationship between the CRT and VA in CRVO or BRVO with simultaneous improvement or worsening of retinal thickness and VA during injection therapy with bevacizumab at least in a part of the patients [8, 12, 14]. Our results support these findings by demonstrating a moderate correlation between final VA and final CRT in CRVO and BRVO. There was also a moderate correlation of a decrease in CRT and an improvement of VA in BRVO. Similar correlations between changes in VA and CRT in BRVO patients have been shown by other authors [15, 16]. Surprisingly, in our CRVO patients changes of these parameters did not correlate with each other, although final VA and final CRT were correlated. The relation between VA and CRT seems to be much weaker in eyes with CRVO. This is also supported by findings of other authors who could not observe a correlation of VA and CRT at 1-year follow-up in CRVO patients treated with ranibizumab [13]. It can be speculated that other factors like macular ischemia play a more predominant role than ME. Nevertheless, an angiographic evaluation of areas of nonperfusion has not been performed, due to extensive bleeding in a significant number of CRVO patients at baseline.

In at least one third of our patients, a reduction of ME was not accompanied by a significant improvement of VA. Patients of subgroup 3 showed a reduction of ME of approximately 250 µm (CRVO) or 150 µm (BRVO), without improvement of VA. Few authors report on the number of patients who do not respond well to treatment. Chung et al. retrospectively analysed 50 patients with BRVO treated with 1.25 mg bevacizumab. Forty-four percent of these patients did not show a significant VA improvement, although ME had significantly decreased, by approximately 250 µm at last visit [7]. Pai et al. reported no improvement in VA in 23.8% of patients with retinal vein occlusion, although a reduction of ME was achieved after a single injection with 1.25 mg bevacizumab [17]. These numbers appear to be consistent with our findings, although a direct comparison with our study is difficult because we categorized our patients primarily by OCT and not by VA. It seems reasonable to consider discontinuation of treatment if no improvement occurs after two to three injections. Further research is needed to determine factors responsible for the reduced responsiveness of these patients to therapy. Maybe a comparison of intravitreal VEGF levels of patients with good or poor response to bevacizumab might give further clues.

In CRVO eyes, VA at baseline is a strong predictor of final VA (r = 0.743, p < 0.0001). No other baseline findings investigated showed a comparably high correlation with final VA. This is in agreement with findings of other authors in untreated CRVO patients [18, 19]. The central vein occlusion study group reported that 80% of patients with a VA worse than 20/200 remained at that level or worsened, whereas in patients with an initial VA of 20/40 or better, likelihood of retaining good acuity was high at the 3-year follow-up [18]. Moderate correlations of final VA in our CRVO patients were also seen with age and final CRT.

In BRVO eyes, final VA was correlated with VA at baseline, age and final CRT. A moderate correlation of initial and final VA in BRVO eyes has also been shown by other authors in patients treated with bevacizumab [15], with oral or instillation drug therapy or photocoagulation of capillary non-perfusion areas, as well as in untreated eyes under observation [19, 20]. Interestingly, in CRVO and BRVO patients final VA was not correlated with baseline CRT, suggesting that even patients with severe ME at baseline can benefit from therapy with bevacizumab.

This is the first study with a high number of patients with a mean follow-up of 1 year evaluating the long-term results of bevacizumab treatment for ME secondary to retinal vein occlusion. Reinjections were performed only in the case of persistent or recurring ME. This treatment scheme allowed us to keep the total number of injections very low, thus minimizing the risk of endophthalmitis. Nevertheless, this strategy resulted in a highly significant improvement in CRT and VA. Some patients do not benefit from treatment. Discontinuation of injections and alternative treatment should be considered in this subgroup.