Introduction

Central serous chorioretinopathy (CSC) is a common retinopathy with an uncertain pathology, characterized by serous detachment of the neurosensory retina.1, 2, 3, 4 The disorder is usually self-limited, although some patients are left with permanent visual impairment because of pigment epithelium and photoreceptor damage, especially in chronic CSC.1, 2, 5, 6 Hypotheses include abnormal alterations at the retinal pigment epithelium (RPE) level2, 3, 7 and choroidal vascular hyperpermeability, as demonstrated on indocyanine green angiography.7, 8, 9

CSC has a high spontaneous remission rate, but there is evidence of the benefit of early treatment.10, 11, 12 CSC with single, extrafoveal leaking point can be treated using focal photocoagulation to shorten the duration of symptoms without altering the final visual outcomes and the recurrent rate.13, 14, 15 This method, however, has a significant adverse effect such as symptomatic scotomas, secondary choroidal neovascularization (CNV), and so on.16, 17 Recently, photodynamic therapy (PDT) with verteporfin has been tried as an alternative treatment to reduce underlying choroidal hyperpermeability and congestion.18, 19 The effect of the vascular modulation was successful with visual improvement in most of patients. However, there is a risk of complications, including RPE atrophy, choriocapillary hypoperfusion, and the development of CNV, especially with standard-dose PDT.9, 20 Half-dose PDT seems to be effective and safe, but its long-term efficacy is unknown.

Bevacizumab (Avastin, Roche, Basel, Switzerland), a monoclonal antibody to vascular endothelial growth factor (VEGF), is a new treatment that exerts antipermeability effects in diabetic macular edema and CNV.2, 21 There have been several off-label clinical trials of intravitreal bevacizumab injection (IVB) in CSC.1, 2, 13, 22, 23, 24, 25 Most showed positive results, with improved visual acuity and reduced subretinal fluid. However, these findings should be interpreted cautiously because of the self-limiting characteristics of CSC, which can show spontaneous improvement within months.1, 2, 12, 22, 23, 24, 25

Therefore, we performed a meta-analysis of the efficacy of IVB in terms of visual acuity and macular thickness to gain a better perspective regarding the therapeutic options in CSC.

Materials and methods

Search method

Three databases (PubMed, EMBASE, and Cochrane) were last searched on 20 August. ‘Central serous chorioretinopathy’, ‘bevacizumab’, and ‘avastin’ comprised the terms for the sensitive search. There was no restriction on study design but the eligible studies only covered those that were written in English. Duplicate articles were manually removed.

Inclusion and exclusion criteria

Published studies, regardless of sample size or study design, were included if the changes in the means and SDs from baseline to 6 months after injection were available for the best-corrected visual acuity (BCVA) in logMAR and central macular thickness (CMT) in μm. The follow-up period varied across the studies, and we chose to analyze the results at 6 months as this period was the most common to all of the included studies. The results of subjects who received IVB were extracted, and the treatment of controls was assigned as simple observation, PDT, or subthreshold laser. The primary outcomes were the change in BCVA and CMT from baseline after IVB. The mean difference and SD at the 6-month follow-up were calculated from the data in the included studies. Secondary outcomes were any reported complication of IVB in eyes with CSC. Case reports, interventional case series, and comments were reviewed but not subjected to analysis, and conference abstracts that had not been published were excluded (Table 1).

Table 1 Characteristics of the excluded studies
Full size table

Quality assessment

The quality of the included randomized clinical trials (RCTs) was evaluated using the Delphi list.26 Determined items were the following: randomization, allocation concealment, baseline similarity within groups, specified eligibility criteria, blinded outcome assessment, blinded care provider, blinded patient, point estimates and measure of variability presented, and intention-to-treat analysis.26 Quality scores were calculated, with a response of ‘yes’ given one point for each item and ‘not available’ or ‘no’ given zero points. Total scores are presented in Table 2.

Table 2 Quality assessment of included randomized clinical trial studies based on Delphi list
Full size table

Statistical analysis

The meta-analysis was conducted using RevMan 5.1,27 with application of a random effect model. Heterogeneity was examined using the I2 statistic and was defined as significant at I2 <50%. Squares indicate mean difference estimates, and lines extending from the squares reveal the associated 95% intervals in the forest plot display. Confidence intervals that do not intersect the vertical line at 0 indicate statistical significance at the 0.05 level.

Sensitivity analysis

In order to assess the influence of two non-RCTs, and Snellen and ETDRS to logMAR conversion included in the meta-analysis, these studies were excluded and the analysis was performed. Sensitivity analyses were undertaken using the following subgroups to assess reliability: (1) all studies including two RCTs and two non-RCTs, and (2) two RCTs with logMAR visual acuity.

Results

Result of search

The literature search yielded 155 articles, 50 from PubMed, 102 from EMBASE, and 3 from Cochrane Library. After excluding 138 ineligible articles because of various reasons such as duplicates from multiple databases, articles not matching the current topic, single case description, and different treatment modality, 17 studies were identified. There were 6 comparative studies, including 2 RCTs and 4 non-RCT studies (Table 3). The remaining 11 studies were interventional case series. After excluding 2 comparative studies because of the lack of data at 6 months of follow-up and the lack of observation group,28, 29 in total 112 patients were included in four comparative studies. Among those, 50 patients were in the treatment group with IVB as compared with 62 patients in the observation group.

Table 3 Characteristics of the reviewed comparative studies
Full size table

Quality assessment

Nine criteria were used to judge the quality of two RCTs included in the meta-analysis. The two RTCs were of reasonable to good quality, and the mean quality score was 55.6%.

Characteristics of studies

The differences in the baseline BCVA, CMT, and demographics between the IVB and control groups were not significant in each of the included studies. However, the clinical heterogeneity among the included studies was considerable in several areas. As for the inclusion criteria, the two studies by Lim et al30 and Aydin31 recruited patients with acute CSC of <3-month duration, whereas the other studies included those with CSC of >3 months. The included studies used the different control groups: the three studies by Artunay et al,32 Lim et al,30 and Aydin31 used patients with observation, whereas the other studies by Semeraro et al29 and Lee et al28 used only those treated with PDT. Both the observation group and a group treated with a subthreshold laser were used as controls in the study by Koss et al.33 Moreover, four studies had consistent treatment arms comparing 1.25 mg/0.05 ml of IVB to the control group, whereas Artunay et al32 performed IVB with 2.5 mg/1.0 ml. Furthermore, Artunay et al,32 Lim et al,30 and Aydin31 simply evaluated the effect of IVB after a single injection, whereas the others allowed repeated treatment with the same modality, guided by BCVA, OCT, and fluorescein angiography (FA).

Visual acuity

Figure 1 shows the forest plot of the BCVA results of IVB and the observation group as a control. Examination of the forest plot demonstrated that, with the exception of the results shown by Aydin,31 the IVB injection group achieved better BCVA at a follow-up of 6 months. Meta-analysis of these data revealed no statistically significant differences between the two groups (−0.02 logMAR, 95% CI −0.14 to 0.11, P=0.80). Because heterogeneity (I2=68%) was substantial, a random effects model was used to present the data.

Figure 1
figure 1

Forest plot showing the mean differences in BCVA in logMAR, with 95% confidence intervals, of experimental (IVB group) compared with the control group (observation) at 6 months. The differences were not significant.

Full size image

Central macular thickness

The forest plot showed that the IVB group was associated with more reduction in CMT than the control group. However, analysis of these data revealed that the difference between groups was not statistically significant (−8.37 μm, 95% CI −97.26 to 80.52, P=0.85; Figure 2).

Figure 2
figure 2

Forest plot of the mean differences in CMT in μm, with 95% confidence intervals, of experimental (IVB group) compared with the control group (observation) at 6 months. The differences were not significant.

Full size image

Comparison among different treatment modalities was not possible because of the small number of available data on the subject.

Sensitivity analysis

Sensitivity analyses of primary outcomes excluding two non-RCTs resulted in similar outcomes. The two non-RCTs included in this meta-analysis did not have an important impact: (1) (−0.02 logMAR, 95% CI −0.14 to 0.11, P=0.80) and (2) (−0.09 logMAR, 95% CI −0.20 to 0.01, P=0.09).

Adverse effects

There was no report of complications or side effects of IVB in patients with CSC in either the included or excluded studies except mild subconjuctival hemorrhage at the injection site.

Discussion

CSC is a self-limiting disease, but it can cause permanent photoreceptor damage or RPE atrophy that leads to visual impairment.34 The greater the understanding of the pathophysiology of CSC with advances in imaging, the more doubts about the so-called ‘benign’ nature of CSC. Several treatment modalities have been tried. Treatment with acetazolamide might shorten the duration of the disease without influencing the final visual acuity or recurrence rate.35 Other medical treatments such as mifepristone, aspirin, finasteride, and propanolol have been reported with good results but require well-designed randomized controlled study to evaluate their efficacy and safety.36, 37, 38, 39 One of the traditional treatments is focal laser photocoagulation, although its application is limited because of possible scotoma or secondary CNV when the RPE leak is close to the fovea.14 Some authors reported that laser photocoagulation was effective in shortening the duration of CSC and reducing the recurrence rate, whereas others reported no definite advantage in terms of the final vision or recurrence.2 PDT is another option for treatment of CSC, as choroidal hyperpermeability has been suggested as the underlying pathology, but foveal atrophy was reported as a possible complication and PDT is expensive.19, 40

Recently, anti-VEGF antibody has been used in CSC as an off-label method.1, 2, 3, 22, 23, 24, 28, 30, 32, 33, 41, 42, 43 Bevacizumab is a recombinant humanized full-length monoclonal antibody of VEGF that penetrates the retina and is transported into the photoreceptor outer segments, RPE, and choroid after intravitreal injection.44 The half-life of bevacizumab in the vitreous is 5.6 days, and it remains in the eye for 4 to 6 weeks.45 Many studies of other macular diseases have reported that the CMT starts decreasing or stabilizes in most eyes at 3–6 weeks.45, 46, 47, 48 The mechanism of IVB in the resolution of subretinal fluid in CSC is unknown, but might involve effects on choroidal vascular hyperpermeability.2, 3, 13

Considering the self-limiting characteristics of CSC, usually within 3–6 months, many of the clinical case studies selected patients with CSC of >3 months as subjects for IVB. Most case series reported no specific complications, although there is always some risk of postinjection endophthalmitis.49 Most studies reported that IVB might be effective in terms of improving visual acuity and reducing CMT without significant complications and can be considered an optional intervention,1, 2, 23, 24, 28, 32, 33 but these conclusions should be interpreted cautiously owing to the self-limiting nature of acute and even in chronic CSC. Moreover, it should be noted that the aqueous humor and plasma levels of VEGF were not significantly increased in patients with CSC compared with the healthy control group.50

Four studies were included in this meta-analysis. They compared the effects of IVB with simple observation. Nonrandomized studies were included, although it is usually better to exclude these from meta-analyses, because we believe that BCVA and CMT are not influenced by randomization. In this meta-analysis, the visual acuity at 6 months after injection was not significantly improved in the IVB group compared with simple observation, and the CMT was also not significantly reduced in the IVB group. Thus, a beneficial role of IVB in treating chronic or acute CSC can be discouraging.

Comparison with other treatment modalities was less valuable because of the relative lack of studies, and the role of IVB remains elusive. Semeraro et al29 reported that IVB group showed greater improvement in the mean BCVA compared with low-fluence PDT (300 mW/cm2 for 83 s, light dose 25 J/cm2) group at 9-month follow-up, although it was not statistically different. In contrast, it was reported by Lee et al28 that the mean BCVA was increased significantly more in standard PDT-treated group. Foveal thinning occurred more frequently in PDT group than in IVB group in both studies. Subthreshold laser photocoagulation (810 nm infrared diode) resulted superior to the IVB for resolution of SRF and BCVA improvement at 6 and 10 months after treatment in the study by Koss et al.33

Although our study confirmed the lack of efficacy of IVB for CSC over a 6-month period, the outcome of this treatment is still unknown. Despite the considerable available literature providing the possible efficacy of IVB for CSC, the majority of studies was simple interventional case series, and was not randomized. As a result, small sample sizes made it difficult to detect differences between the treatments and the observation. The second limitation is that the clinical heterogeneity among comparative studies was considerable in several areas, including inconsistency in inclusion criteria, and treatment protocols. The acute and chronic CSC were all subjected to the meta-analysis together. It would have been more meaningful meta-analysis to differentiate them, considering there might be inherent differences in the nature of acute and chronic CSC. However, chroidal vascular hyperpermeability is considered to be fundamental mechanisms for acute and chronic CSC, and the current treatment concept such as PDT or IVB is focused on modulating hyperpermeability in choroid. There has been language bias as the eligible studies only covered those that were written in English. Despite these limitations, this analysis makes an important contribution to the field because this study was the first meta-analysis to explore the use of intravitreal bevacizumab for CSC.

In conclusion, meta-analysis failed to verify the positive effect of IVB in CSC based on the epidemiological literature published to date. In the future, more studies using larger scales and better methodologies will help to clarify the uncertain relationship between CSC and IVB.