Conbercept for patients with age-related macular degeneration: a systematic review

Pubmed, Embase, and Cochrane Central Register of Controlled Trials (CENTRAL) published in Cochran Library were searched using the search strategies detailed in Additional file 1: Table S2, from their earliest records to June 2017. Clinicaltrials.​gov was searched with the terms “age-related macular degeneration” and “conbercept”. The China National Knowledge Infrastructure(CNKI), VIP database, and Wanfang database were also searched with Chinese terms.

All included studies met the following criteria: (1) randomized controlled studies (RCTs); (2) participants with wet age-related macular degeneration aged more than 50 years old; (3) the intervention was conbercept irrespective of dosage and schedule; (4) the comparisons included conservative treatment, ranibizumab, transpupillary thermotherapy, and triamcinolone acetonide; (5) studies included at least one of the following outcomes: the primary outcome was the mean change in best-corrected visual acuity (BCVA, which was measured by the logarithmic visual acuity chart) from baseline to the third month after the first treatment; secondary outcomes included the mean change in central retinal thickness (CRT, which was measured by optical coherence tomography) from baseline to the last visit, the mean change in plasma level of VEGF from baseline to the last visit, and the incidence of adverse events (AEs); (6) publication written in English or Chinese. We excluded the patients with glaucoma, cataracts, or retinopathy caused by diabetes or hypertension and studies without available raw data.

Two investigators independently screened the titles and abstracts of the articles identified by literature searching (Additional file 1: Table S2 shows the searching strategy), and assessed the studies using predetermined inclusion criteria. The full texts of all potentially relevant articles were retrieved for detailed review. Any disagreement in the process of selection was resolved by discussion. Two authors independently extracted following data from included articles: (1) authors; (2) year of publication; (3) country or region where the study was conducted; (4) study design and use of control; (5) number of participants randomized into each group; (6) gender, age, and disease duration of participants; (7) treatment regimens (dose and schedule); (8) outcomes of each study and their definitions; (9) numerical data for outcomes assessment; (10) sources of funding.

Two authors independently assessed the bias risk of each included study using the checklist developed by Cochrane Collaboration [14, 15]. Any disagreements about the risk of bias was resolved by discussion. The items included random sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting, and other bias. We categorized the judgments as low, high or unclear risk of bias and created plots of bias risk assessment in Review Manager Software (RevMan 5.3).

We calculated a kappa statistic for measuring the agreement level between two authors regarding to the decisions made on study selection. The value of kappa (K) 0.40–0.59 was considered fair agreement, 0.60–0.74 as good and 0.75 or more as excellent [16].

If more than one study reported the same outcome, a pairwise meta-analysis was conducted. We analyzed RCTs using risk ratios (RRs) for the incidence of AEs and mean differences (MDs) for BCVA, CRT, and VEGF level, with corresponding 95% confidence intervals (CIs) to compare differences between conbercept and control groups. We pooled RRs with the Mantel-Haenszel method, and MDs with the inverse variance method using RevMan 5.3, respectively. Statistical heterogeneity among studies was examined by the Chi-square test and quantified by the I ² statistic [14]. We applied a fixed-effects model to synthesize data when heterogeneity was not significant (P>0.1 and I ² <50%). When heterogeneity was significant (P ≤ 0.1 and I ²  ≥ 50%) and could not be explained by subgroup analyses or in terms of clinical or methodological features of the trials, a random-effects model was used. We explored sources of heterogeneity based on the following subgroup analyses: type of control groups (e.g. conservative treatment, triamcinolone acetonide, transpupillary thermotherapy, or ranibizumab). We carried out sensitivity analyses by using alternative pooling methods (Peto vs. Mantel-Haenszel method), and statistical models regarding to heterogeneity (random-effects vs. fixed-effect).

A total of 780 citations were obtained from the literature search and the selection process is shown in Fig. 1. Eighteen RCTs (1285 participants) [17–34] were included in this systematic review. Agreement on study selection between two reviewers was excellent (K = 0.83). All the RCTs were single-center studies conducted in China. As shown in Table 1, the comparisons were conservative treatment (3 RCTs [17–19], 232 participants), ranibizumab (6 RCTs [20–25], 395 participants), transpupillary thermotherapy (4 RCTs [26–29], 326 participants), and triamcinolone acetonide (5 RCTs [30–34], 332 participants). The follow-up time ranged from 1 to 12 months after the first treatment, except one study [18] without reporting this information.

As shown in Table 2, all participants were aged 51–87 years old with disease duration of 7 days to 10 years (10 studies [17, 19, 23–26, 28–31] did not report the disease duration). The reported dose of conbercept ranged from 0.5 to 1.5 mg, except 2 studies [21, 30] which failed to report dosing information.

As shown in Fig. 2, the random sequences of 11 studies [19, 21, 23, 24, 28–31, 33, 34] were generated by a random number table or simulation, while all the studies failed to describe the method of allocation concealment. Therefore, the risk of selection bias related to allocation concealment was unable to be assessed. The risk of performance bias of all studies was uncertain, as the blinding method was not reported. All studies had low risk of attrition bias, as there was no loss to follow-up. No studies contained information related to registration information nor had protocols available, so it was unknown whether all the pre-designed outcomes in these studies had been reported. Since none of the studies were reported being supported by pharmaceutical industry funding, the bias caused by conflict of interest was low. Due to the limited number of the included studies for the same outcome, publication bias investigation was not performed.

The mean change in BCVA from baseline to the third month after the first treatment was reported in 6 studies [23–25, 30–32] (435 participants). Subgroup analyses were performed and stratified by control group selection (Fig. 3). The heterogeneity of each subgroup was not statistically significant (I ² <50%, P>0.1), so the MDs of the mean changes of BCVA were pooled with a fixed-effects model. The difference between the pooled results of the two subgroups was significant (P < 0.00001). Compared to triamcinolone acetonide, conbercept significantly improved the BCVA in the third month after the first treatment [MD = 0.11, 95%CI (0.08, 0.15)]. While the mean change in BCVA from baseline in conbercept group was similar to that in ranibizumab group [MD = 0.00, 95% CI (− 0.03, 0.04)].

The mean change in CRT from baseline to the last visit was reported in all 18 studies [17–34] (1285 participants). Subgroup analyses were performed and stratified by control group selection (Fig. 4). The MDs of the mean changes of CRT were pooled with a fixed-effects model because the heterogeneity of each subgroup was not statistically significant (I ² <50%, P>0.1). There were significant differences among the pooled results of the four subgroups (P < 0.00001). Compared with the other four therapies (conservative treatment, ranibizumab, transpupillary thermotherapy, triamcinolone acetonide), conbercept significantly reduced the CRT at the last visit [MD = − 49.51, 95% CI (− 67.45, − 31.58); MD = − 9.96, 95% CI (− 17.61, − 2.32); MD = − 60.51, 95% CI (− 92.14, − 28.89); MD = − 79.17, 95% CI (− 96.34, − 61.99), respectively].

Two studies [21, 22] (110 participants) reported the plasma level of VEGF. According to Fig. 5, the heterogeneity was not statistically significant (I ²  = 0%, P = 0.62). The pooled results with a fixed-effects model indicated that conbercept significantly lowered the plasma level of VEGF [MD = − 15.86, 95% CI (− 23.17, − 8.55)] compared to ranibizumab.

Eight studies [17, 19, 21, 24, 28, 30–32] (566 participants) reported the incidence of any AEs. Due to the significant heterogeneity in conbercept vs. conservative treatment subgroup (I ²  = 75%, P = 0.05), the RRs were pooled with a random-effects model (Fig. 6). The incidence of AEs in the conbercept group was similar to the conservative treatment [RR = 8.81, 95% CI (0.20, 388.62)], ranibizumab [RR = 1.25, 95% CI (0.38, 4.12)] [21], and transpupillary thermotherapy groups [RR = 5.00, 95% CI (0.25, 101.81)] [28], but significantly lower than triamcinolone acetonide group [RR = 0.25, 95% CI (0.09, 0.72)]. None of the studies reported serious AEs, and the most common AEs were increased IOP and ophthalmecchymosis. The incidences of increased IOP (Fig. 7) and ophthalmecchymosis (Fig. 8) in the conbercept group were similar to ranibizumab [RR = 0.20, 95% CI (0.01, 3.97); RR = 1.50, 95% CI (0.27, 8.22), respectively] [21], transpupillary thermotherapy [RR = 3.00, 95% CI (0.13, 71.65); RR = 3.00, 95% CI (0.13, 71.65), respectively] [28], and triamcinolone acetonide [RR = 0.50, 95% CI (0.13, 1.94); RR = 3.00, 95% CI (0.13, 71.22) [32], respectively] groups, but were significantly higher than conservative treatment group [RR = 14.00, 95% CI (1.88, 104.25); RR = 20.00, 95% CI (2.74, 145.96), respectively].

The sensitivity analyses were performed by pooling methods and statistical models regarding to test heterogeneity, and the results (BCVA, CRT, Plasma level of VEGF, and AEs) were robust.