Purpose
To compare safety and functional outcomes of vitrectomy with or without preoperative intravitreal bevacizumab (IVB) for proliferative diabetic retinopathy (PDR).
Design
A meta-analysis of randomized controlled trials.
Methods
PubMed, Medline, EMBASE, and Cochrane Controlled Trials Register were searched to identify potentially relevant randomized controlled trials. A total of 394 participants with 414 eyes in 8 trials were analyzed using RevMan 5.1 software. The primary measures included intraoperative bleeding, total surgical time, and early and late recurrent hemorrhage.
Results
Vitrectomy with IVB pretreatment achieved shorter overall surgical time (mean difference = −26.89 minutes, 95% confidence interval [CI] −31.38 to −22.39, P < .00001) and smaller number of endodiathermy applications (mean difference = −3.46, 95% CI −6.43 to −0.49, P = .02) compared to vitrectomy alone. The IVB group was also associated with less intraoperative bleeding (odds ratio [OR] = 0.10; 95% CI 0.02 to 0.46; P = .003) and recurrent vitreous hemorrhage within first month (OR = 0.35; 95% CI 0.21 to 0.58; P < .0001), but the proportion of recurrent vitreous hemorrhage after the first month was comparable between both groups. There were no significant differences in other complication rates between the 2 groups, with the exception of iatrogenic retinal break, which was more likely with the vitrectomy-alone group (OR = 0.27, 95% CI 0.12 to 0.63, P = .003). Results were robust to sensitivity analyses.
Conclusions
Adjuvant intravitreal injection of bevacizumab prior to vitrectomy in PDR patients significantly eased the procedure, diminished intraoperative complications, and reduced early postoperative hemorrhage without increasing the risk of vision-threatening complications. Further trials should determine the optimal interval and dosage for IVB injection.
Proliferative diabetic retinopathy (PDR) featuring neovascularization and fibrous proliferation is one of the foremost diseases that can lead to blindness if not properly treated. Patients with nonabsorbable vitreous hemorrhage, tractional retinal detachment (RD), or extensive fibrovascular proliferations are generally candidates for pars plana vitrectomy. However, a guarded prognosis should be expected because of the relatively high rate of complications during and after surgery. One of the most challenging complications is intraoperative bleeding, which makes the surgeons’ maneuvers very difficult because of poor visualization and requires the use of more equipment. All of these factors can prolong the total surgical time and hinder surgical outcomes. Moreover, massive bleeding during surgery may even be uncontrollable, leading to surgical failure.
Bevacizumab (Avastin; Genentech, Inc, South San Francisco, California, USA), a full-length recombinant humanized monoclonal antibody directed against vascular endothelial growth factor (VEGF), is one of the most potent anti-VEGF agents. The antiangiogenic properties of intravitreal bevacizumab (IVB) have been researched in patients with age-related macular degeneration, myopic choroidal neovascularization, and neovascular glaucoma, and all results have indicated its apparent efficacy and safety. Intravitreal bevacizumab injections have also been reported to ease diabetic epiretinal membrane dissections. For vitrectomy in patients with PDR, especially in patients with ample, active neovascularization and/or extensive or multiple layers of fibrovascular proliferation, preoperative IVB injections are thus expected to be helpful. Zhao and associates conducted a meta-analysis of 4 randomized controlled trials and 2 comparative case studies, and the results showed that IVB before vitrectomy can facilitate the surgery and reduce intraoperative and postoperative complications. However, these results have not been supported by more recent reports. Thus, we conducted a meta-analysis including newly published randomized controlled trials (RCTs) to reevaluate the effectiveness and safety of bevacizumab injections as a preoperative adjunct.
Methods
Data Sources
This study was a meta-analysis of randomized controlled trials. Articles limited to randomized controlled trials were searched using PubMed, Medline, EMBASE, and the Cochrane Central Register of Controlled Trials. “Vitrectomy,” “bevacizumab,” and “proliferative diabetic retinopathy” comprised the terms for the sensitive search. The reference lists of every primary article and previous systematic review were scrutinized for information about additional trials. We performed the final search on September 5, 2012. This study adhered to the tenets of the Declaration of Helsinki and relevant laws in China, Mexico, Italy, and Iran.
Study Selection
First, 2 reviewers (Z.H.Z., K.L.) independently assessed studies for possible eligibility at the title and/or abstract level. They reviewed the corresponding articles in full text. Trials were considered for inclusion if they met the following criteria: (1) randomized controlled trials that compared vitrectomy alone vs vitrectomy with IVB pretreatment for proliferative diabetic retinopathy; and (2) trials reporting at least 1 of the primary outcome measures (intraoperative bleeding, surgical time, and postoperative vitreous hemorrhage) or secondary outcome measures (frequency of endodiathermy, visual acuity, and other intraoperative and postoperative complications). For duplicated publications, only the data from the longest period of follow-up were used in the analysis.
Data Extraction
Two independent reviewers (Z.H.Z., X.X.) extracted data from the included trials with a customized form. Pertinent data included study population characteristics (age, number of patients and eyes in study, and location), intervention groups, comparison groups, outcome variables, and duration and completeness of follow-up. The corresponding authors of the individual trials were also contacted for unpublished information.
Quality Assessment
Jadad scores on a scale of 0 to 5 were used to evaluate the quality of each trial. Each trial was assessed for 3 main aspects of its study design: randomization, masking, and participant withdrawals/dropouts. Trials with a score higher than 3 were considered to be high quality.
Outcome Variables
The following variables were used for primary outcome comparisons: (1) intraoperative bleeding; (2) total surgical time; and (3) early (within 1 month) and late (after 1 month) postoperative vitreous hemorrhage.
Number of endodiathermy applications, postoperative best-corrected visual acuity (BCVA), and other intraoperative and postoperative complications were considered as secondary outcomes.
Statistical Analysis
All statistical analyses were performed with Review Manager Version 5.1 (The Cochrane Collaboration, Oxford, England) using 2-tailed P values and a 95% confidence interval (CI). A P value less than .05 was considered to be statistically significant. The dichotomous outcomes were calculated by pooled odds ratios (OR). For continuous outcomes, analysis was performed using the mean difference. The fixed-effects model was used to pool the data. When there was substantial heterogeneity, the results using the random-effects model were presented.
Heterogeneity
Heterogeneity was explored using the Q test with calculated I 2 , indicating the percentage of variability in effect estimates that were attributable to heterogeneity rather than to chance. I 2 values of 50% or more were considered to indicate substantial heterogeneity; in these cases, sensitivity analyses were performed to identify the largest group of studies that passed the test of heterogeneity.
Sensitivity Analysis
Sensitivity analyses were undertaken using the following subgroups to assess reliability: (1) studies of high quality (Jadad score ≥3), (2) studies with a year of publication later than 2008.
Publication Bias
Publication bias was explored by searching for asymmetry in the funnel plot.
Results
Results of Search
The flow chart of studies from initial results to final inclusions is shown in Figure 1 . Of the 98 potentially relevant studies yielded by electronic searches, 9 met all of the predefined inclusion criteria. Of these studies, 1 article was excluded from the analysis because its data could not be pooled in any comparison. Eight randomized controlled trials published between 2008 and 2011 were included in this meta-analysis.
Characteristics and Quality of Trials
A total of 394 participants with 414 eyes in the 8 included trials were enrolled in this meta-analysis. Table 1 shows the main characteristics and quality scores of the included trials. Four out of 8 trials graded the surgical complexity of the injection and control groups, and the complexity scores were statistically similar for both groups. Seven out of 8 trials had groups matched for preoperative visual acuity. Four of the 8 trials contained groups matched for age and sex, and 2 of the other 4 trials matched groups for age. Masking surgeons is impossible with this type of trial. Four trials reported masking outcome assessors, and 1 trial reported masking patients. Regarding attrition rate, 6 trials had 100% completeness of follow-up; 1 trial lost 4 eyes (6%) to follow-up, and the other trial lost 5 eyes (5%) to follow-up. Overall, 4 out of the 8 trials were deemed to be high quality (Jadad score ≥3).
| Study a /Year | Country | FU (mo) | Pts/Eyes (n) | Mean Age (Y) ± SD (Range) | Type of PPV | Time Interval Between IVB Injection and Vitrectomy | IVB Injection Dosage | Allocation Concealment | Masking of Pts | Masking of Outcome Assessor | Loss to FU (Eyes) | Quality Score | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Vitrectomy With IVB | Vitrectomy Without IVB | ||||||||||||
| Farahvash/2011 | Iran | 3 | 35/635 | 58.3 (49-73) | 58.7 (37-72) | 20-gauge | 7 days | 1.25 mg | NA | NA | NA | 0 | 3 |
| Ahn/2011 | Korea | 6 | 91/107 | 51.0 ± 9.5 | 55 ± 11.4 | 23-gauge | 1-14 days | 1.25 mg/0.05 mL | NA | NA | Yes | 5 | 3 |
| Hernández-Da Mota /2010 | Mexico | 6 | 40/40 | 55.7 ± 9.9 | 55.7 ± 7.4 | 23-gauge | 2 days | 1.25 mg/0.05 mL | NA | NA | Yes | 0 | 2 |
| Rizzo/2008 | Italy | 6 | 22/22 | 52 (24-63) b | 20-gauge | 5-7 days | 1.25 mg | NA | NA | NA | 0 | 3 | |
| di Lauro/2010 | Italy | 6 | 68/72 | NA b | 20-gauge | 7 days or 20 days | 1.25 mg/0.05 mL | NA | NA | NA | 0 | 2 | |
| Modarres/2009 | Iran | 3 | 40/40 | 55.8 ± 11.3 | 53.2 ± 11.7 | 20-gauge | 3-5 days | 2.5 mg/0.1 mL | NA | NA | Yes | 0 | 2 |
| Ahmadieh/2009 | Iran | 1 | 68/68 | 55.2 (21-76) b | NA | 7 days | 1.25 mg/0.05 mL | Yes | Yes | Yes | 4 | 3 | |
| El-Batarny/2008 | Oman | 6 | 30/30 | 44 ± 11 | 46 ± 12 | NA | 5-7 days | 1.25 mg/0.05 mL | NA | NA | NA | 0 | 2 |
b Represents mean age (age range) of all the patients from both groups.
Intraoperative Bleeding
Data for intraoperative bleeding were collected from 5 trials. In 1 trial (Hernández-Da Mota and Nuñez-Solorio ), intraoperative bleeding occurred in all eyes of both groups. The 4 remaining trials favored the IVB injection group, and the meta-analysis of pooled data showed statistically significant differences between the groups (OR = 0.10; 95% CI 0.02 to 0.46; P = .003) ( Figure 2 ). Because heterogeneity was evident for the outcome (I 2 = 68%), a random-effects model was used to combine the data. Sensitivity analysis was performed, and the result pooled from the largest group of homogeneous studies showed that the difference between the 2 groups was still significant ( P < .00001) ( Figure 2 ).
Number of Endodiathermy Applications
Data for this outcome were available from 4 trials. Examination of the forest plot showed that endodiathermy was used more often in the control group than in the IVB injection group, and the result was statistically significant (mean difference = −3.46, 95% CI −6.43 to −0.49, P = .02) ( Figure 3 ). Because of substantial heterogeneity (I 2 = 97%), the result including the largest group of homogeneous studies was also shown in Figure 3 ( P < .00001).
Surgical Time
Five trials were included for the comparison. The forest plot showed that the control group was associated with longer surgical time than the IVB injection group. Analysis of these data revealed that the difference between groups was statistically significant (mean difference = −26.89 minutes, 95% CI −31.38 to −22.39, P < .00001) ( Figure 4 ).
Early Postoperative Vitreous Hemorrhage
Seven trials reported data for the proportion of postoperative vitreous hemorrhage during the first month. The forest plot showed that all 7 trials favored the IVB injection group. Analysis of these data revealed that the differences between groups in favor of IVB were statistically significant (OR = 0.35; 95% CI 0.21 to 0.58; P < .0001) ( Figure 5 ).
Late Postoperative Vitreous Hemorrhage
Five trials reported data for the proportion of postoperative vitreous hemorrhage after the first month. The forest plot illustrated that the control group had a higher proportion of eyes suffering from late postoperative vitreous hemorrhage than the IVB injection group, with the exception of the results from Farahvash and associates. Analysis of these data did not show a difference between the 2 groups (OR = 0.55; 95% CI 0.25 to 1.21; P = .14) ( Figure 6 ). A sensitivity analysis was performed to examine the effect of excluding the data from Farahvash and associates, whereby the statistical result did not change (OR = 0.45; 95% CI 0.19 to 1.07; P = .07).
Postoperative Best-Corrected Visual Acuity
Five trials reported postoperative BCVA (mean ± standard deviation) as a logarithm of the minimal angle of resolution (logMAR). Examination of the forest plot demonstrated that, with the exception of the results shown by Ahn and associates the IVB injection group achieved better BCVA at a follow-up of ≥3 months (range, 3-6 months). Meta-analysis of these data revealed no statistically significant differences between the 2 groups (mean difference = −0.36 logMAR, 95% CI −0.84 to 0.13, P = .15) ( Figure 7 ). Because heterogeneity (I 2 = 86%) was substantial, a random-effects model was used to present the data, and sensitivity analyses including the largest group of homogeneous studies were performed. However, the statistical result did not change (mean difference = −0.04 logMAR, 95% CI −0.22 to 0.14, P = .67) ( Figure 7 ).
