Objective  To evaluate the efficacy and safety of intravitreal bevacizumab (IVB) as a preoperative adjunct to vitrectomy for proliferative diabetic retinopathy (PDR).
Methods  Meta-analysis of randomized controlled trials (RCTs). A computerized search was conducted in the Medline, Embase, Cochrane Library, Chinese Biomedical Database and Chinese Journal Full-text Database combined with manually searching of bibliographies of pertinent articles, journals and literature reference proceedings. RCTs on IVB as a preoperative adjunct to vitrectomy for PDR were selected according to inclusion and exclusion criteria. After the data extraction, methodological quality assessment of RCTs that were included using the Jadad scale. The statistical analysis was performed by Stata/SE 11.2. The weighted mean difference (WMD) with 95% confidence interval (CI) was calculated for continuous variable, and odds ratio (OR) with 95% CI for dichotomous variable. The outcome measures included the incidence of intraoperative bleeding, frequency of endodiathermy, mean number of endodiathermy, surgical time, the incidence of intraoperative complications, the incidence of postoperative VH, postoperative best-corrected visual acuity (BCVA), the incidence of postoperative retinal attachment and postoperative complications.
Results  In total, eight RCTs that fulfilled the eligibility criteria were included in the meta-analysis involving 181 patients in IVB group and 172 patients in control group. According to the Jadad scale, one RCT scored 5 points, one RCT scored 3 points, one RCT scored 2 points and the other five RCTs scored 1 point. The results suggested that IVB group achieved lower incidence of intraoperative bleeding than control group(OR = 18.54, 95% CI: 6.14 - 55.99, P = 0.00), and both in the preoperative 5-7 days IVB subgroup (OR = 12.46, 95% CI: 2.01 - 77.11, P = 0.01), and in the preoperative 7 days IVB subgroup (OR = 23.30, 95% CI: 5.78 - 93.90, P = 0.00). The results suggested that IVB group achieved less frequency of endodiathermy than control group (OR = 21.41, 95% CI: 6.57 - 69.81, P = 0.00). The results suggested that IVB group achieved less mean number of endodiathermy than control group (WMD = 4.38, 95% CI: 3.76 - 4.99, P = 0.00). The results suggested that IVB group achieved less surgical time than control group (WMD = 24.04, 95% CI: 18.37 - 29.72, P = 0.00). The results suggested that IVB group achieved lower incidence of iatrogenic retinal breaks than control group (OR = 4.10, 95% CI: 1.44 - 11.64, P = 0.01). The results suggested that IVB group achieved lower incidence of postoperative VH than control group both in the postoperative early subgroup (OR = 3.28, 95% CI: 1.58 - 6.82, P = 0.00), and in the postoperative late subgroup (OR = 2.51, 95% CI: 1.21 - 5.22, P = 0.01). The results suggested that IVB group achieved lower incidence of postoperative VH than control group both at postoperative 1 week (OR = 2.38, 95% CI: 1.07 – 5.30, P = 0.03), and at postoperative 1 month (OR = 3.87, 95% CI: 1.40 – 10.68, P = 0.01), but no significant differences were found between IVB and control group either at postoperative 3 months (OR = 2.52, 95% CI: 0.74 – 8.57, P = 0.14)), or at postoperative 6 months (OR = 3.26, 95% CI: 0.50 – 21.45, P = 0.22). The results suggested that IVB group achieved better postoperative BCVA than control group (WMD = 0.29, 95% CI: 0.13 - 0.44, P = 0.00), but no significant differences were found between IVB and control group on the incidence of postoperative retinal reattachment (OR = 0.39, 95% CI: 0.10 - 1.59, P = 0.19), postoperative retinal redetachment (OR = 2.36, 95% CI: 0.74 – 7.56, P = 0.15) and postoperative neovascular glaucoma(OR = 1.47, 95% CI: 0.28 – 7.71, P = 0.65).
Conclusions  IVB as a preoperative adjunct in vitrectomy could achieve lower incidence of intraoperative bleeding, less frequency of endodiathermy and mean number of endodiathermy, shorter surgical time, lower incidence of iatrogenic retinal breaks as well as prevent postoperative VH effectively, and relatively safe. However, a high-quality, multicenter, large sample, long-term RCT is warranted to be further investigated.