Topic
This systematic review and meta-analysis evaluated the literature-pooled risk of retinal detachment (RD) following intravitreal injection (IVI) of anti–vascular endothelial growth factor (anti-VEGF) agents.
Clinical Relevance
Intravitreal anti-VEGF injections are the most frequently performed intraocular procedure, with RD remaining a poorly quantified complication. Pooled estimates equip clinicians to contextualize RD risk during informed consent.
Methods
This systematic review and meta-analysis followed PRISMA guidelines (PROSPERO: CRD420251175527). MEDLINE, Embase, and Cochrane CENTRAL were searched from inception to January 18, 2026. Included studies reported RD risk after anti-VEGF IVI for neovascular age-related macular degeneration (nAMD), diabetic macular edema, macular edema from retinal vein occlusion, proliferative diabetic retinopathy, or myopic choroidal neovascularization. Paired reviewers independently extracted data and assessed risk of bias using ROBINS-I and the Joanna Briggs Institute checklist for case series. A random-effects single-arm meta-analysis of proportions using a generalized linear mixed model generated pooled risk estimates per injection and per eye.
Results
Twelve studies comprising 834 814 injections and 119 RD events were included. In the meta-analysis of 9 studies (787 849 injections), there were 92 rhegmatogenous RD (RRD) events, with a pooled risk of 0.012% (95% CI [0.009–0.015%], I²: 14.4%), equivalent to 1 in 8675 injections. An acute ≤90-day postprocedural sensitivity analysis yielded a consistent RRD risk of 0.013% (95% CI [0.010–0.016%], I²: 36.8%) (approximately 1 in 7692 injections). An eye-level sensitivity analysis yielded a per-eye risk of 0.08% (95% CI [0.02–0.31%], I²: 66.7%) (approximately 1 in 1250 eyes). Exploratory subgroup analyses showed no differences in RRD risk by injection number (≥100 000 vs <100 000; P =.520), study setting (single-center vs multicenter; P =.258), injection indication (nAMD only vs mixed; P =.061), injection provider (ophthalmologist vs mixed; P =.259), or risk of bias ( P =.356). One study evaluating severe PDR reported tractional RD risk as 3.6% (95% CI [2.3–5.2%]) (approximately 1 in 28 injections), likely reflecting disease-specific fibrovascular contraction.
Conclusion
Low-certainty evidence indicates RRD following anti-VEGF IVI is uncommon, with a per-injection risk of 0.012% (1 in 8675 injections). Per-eye risk was 0.08% (1 in 1250 eyes), which was subject to wider uncertainty but more relevant to patients undergoing long-term treatment; notably, the pooled mean of 13.8 injections per eye in the contributing studies likely underestimates real-world cumulative exposure. Future work should incorporate eye-level denominators and standardized reporting of RD subtypes, ocular characteristics, and injection technique to enable risk stratification and identification of modifiable procedural contributors.
INTRODUCTION
I ntravitreal injection (IVI) of antivascular endothelial growth factor (anti-VEGF) agents has become the standard of care for managing a range of retinal vascular diseases, including neovascular age-related macular degeneration (nAMD), diabetic macular edema (DME), macular edema from retinal vein occlusion (RVO), proliferative diabetic retinopathy (PDR), and myopic choroidal neovascularization (mCNV). , The estimated number of IVIs administered is approximately 7 million in the United States and over 20 million worldwide, which is projected to rise as the population ages and grows. ,
Despite a favorable safety profile, IVIs are associated with rare but vision-threatening complications, such as endophthalmitis, vitreous hemorrhage, sustained elevated intraocular pressure (IOP), cataract formation, and retinal detachment (RD). , While prior meta-analyses have quantified the risk of several postinjection complications, none have provided a pooled estimate of RD risk. , Given the increasing frequency of intravitreal anti-VEGF administration, accurate estimation of RD risk is essential for informed consent, patient counseling, and identification of procedural- and patient-level risk factors. This systematic review and meta-analysis synthesizes the available literature to determine the pooled risk of RD following IVI of anti-VEGF agents.
METHODS
This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) guidelines. This study was registered with the International Prospective Register of Systematic Reviews (CRD420251175527) and adhered to the tenets of the Declaration of Helsinki. Research ethics board approval was waived since the data were extracted from published primary research.
STUDY SELECTION AND DATA EXTRACTION
MEDLINE, Embase, and Cochrane CENTRAL databases were searched using a librarian-informed search strategy from inception to January 18, 2026 (Supplemental Table 1). Studies were included if they assessed the risk of RD following IVI of anti-VEGF agents for nAMD, RVO, DME, PDR, or mCNV as a primary outcome. If not a primary outcome, studies were included if they evaluated over 10 000 IVIs. Eligible study designs included randomized controlled trials, observational cohort studies, case–control studies, and large case series with over 100 IVIs. Studies were limited to peer-reviewed, English-language articles. No restrictions were imposed on study setting, geographic region, or publication year.
Two independent reviewers (P.Yu and L.C.) screened the titles, abstracts, and full texts using Covidence (Veritas Health Innovation, Melbourne, Australia). Discrepancies were resolved with a third study author (D.M.). Two investigators (P.Yu and L.C.) used a predefined data collection form to independently extract data. Study characteristics consisted of first author, year of publication, journal and country of publication, study design, study setting, indications for injection, injection provider, anti-VEGF agents used, total number of injections, and total number of patients. Overall patient characteristics collected included the mean age, sex, total number of eyes, the distribution of indication for injection, needle gauge, injection quadrant, and follow-up duration. Patient characteristics with RD included number of patients with RD, mean age, sex, lens status, RD type, mean number of IVI before RD, needle gauge, injection quadrant, and macula status and posterior vitreous detachment (PVD) at RD diagnosis. Two investigators (P.Yu and L.C.) evaluated risk of bias using the Risk Of Bias In Nonrandomized Studies- of Interventions (ROBINS-I) tool for observational studies and the Joanna Briggs Institute (JBI) Critical Appraisal Checklist for Case Series. , The certainty of evidence in included studies was appraised using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework, beginning at low certainty for all outcomes given the observational study designs. Any disagreements were resolved through discussion with a third study author (D.M.).
STATISTICAL ANALYSIS
The primary outcome was the pooled proportion of RD after IVI. Because the primary objective was to quantify isolated procedural risk and given the lack of standardized follow-up durations across the included literature, this outcome was modeled as a per-injection proportion (procedural risk) rather than a time-bound incidence rate. Studies reporting exclusively on TRDs were excluded from the primary pooled meta-analysis. In eyes with PDR, anti-VEGF administration poses a pharmacological risk of exacerbating pre-existing traction through rapid fibrovascular contraction. This creates a distinct patient and RD population that was excluded to prevent confounding the procedural risk estimates for RRDs. Data from these studies are instead reported narratively.
A random-effects, single-arm meta-analysis of proportions was conducted using a generalized linear mixed model (GLMM) with binomial distribution, logit transformation, and maximum likelihood estimation to derive a pooled risk estimate with 95% CIs (CIs) (α = 0.05). The model assumes that observed risk proportions follow a binomial distribution, that true study-level effects are normally distributed around an overall mean effect, and that between-study heterogeneity arises from random variation rather than systematic bias. The denominator was the total number of injections.
For studies reporting zero RD events, the GLMM method enabled inclusion without the need for continuity correction. Statistical heterogeneity was evaluated using the I² statistic and τ², interpreted according to Cochrane thresholds: 0% to 40% (might not be important), 30% to 60% (may represent moderate heterogeneity), 50% to 90% (substantial heterogeneity), and 75% to 100% (considerable heterogeneity). Heterogeneity was also visually assessed via forest plots. Prespecified subgroup analyses were performed to explore whether pooled RRD risk differed according to study-level characteristics. Studies were stratified by total number of injections (≥100 000 vs <100 000), study setting (single-center vs multicenter), injection indication (nAMD only vs mixed indications), injection provider (ophthalmologist only vs mixed providers), and risk of bias. Separate random-effects GLMM models were fit within each subgroup to estimate pooled risk proportions. Differences between subgroups were assessed using a test for subgroup differences based on between-subgroup heterogeneity (Cochran’s Q).
Sensitivity analyses were conducted using the leave-one-out method, repeating the meta-analysis on each subset of studies with exactly 1 study excluded to assess the robustness of conclusions. A secondary sensitivity analysis was performed using a random-effects GLMM to estimate the pooled per-eye risk of RRD, restricted to the subset of studies explicitly reporting eye-level denominators. An additional sensitivity analysis was conducted to isolate acute events by restricting the meta-analysis to the subset of studies explicitly defining a short-term postprocedural time to RD detection window (defined as ≤90 days). Given that there were <10 included studies, formal publication bias assessments (eg, Egger’s regression analysis) were not performed. All analyses were conducted in R (version 4.3.0; R Foundation for Statistical Computing, Vienna, Austria), and statistical significance was defined as a two-tailed P <.05.
RESULTS
SEARCH RESULTS AND STUDY CHARACTERISTICS
A total of 1021 studies were identified from MEDLINE (n = 366), Embase (n = 592), and CENTRAL (n = 63). After removing 229 duplicates, 792 articles underwent title and abstract screening using the inclusion and exclusion criteria. Of these, 756 were excluded, leaving 36 articles for full-text review. A total of 12 studies reported the risk of RD following IVI delivery of anti-VEGF and were eligible for inclusion ( Figure 1 ). ,,,,,,,,,,, Studies originated from 9 countries, including the United States (n = 3, 25.0%), United Kingdom (n = 2, 16.7%), Australia (n = 1, 8.3%), Brazil (n = 1, 8.3%), Colombia (n = 1, 8.3%), Denmark (n = 1, 8.3%), Germany (n = 1, 8.3%), Japan (n = 1, 8.3%), and Singapore (n = 1, 8.3%). They were either retrospective cohort studies (n = 7, 58.3%), prospective cohort studies (n = 1, 8.3%), or retrospective case series (n = 4, 33.3%), and study settings were split between single center (n = 8, 66.7%) and multicenter (n = 4, 33.3%).
PRISMA flow diagram.
Included studies examined patients receiving IVIs for several indications, including nAMD (n = 9, 75.0%), macular edema secondary to RVO (n = 7, 58.3%), DME (n = 5, 41.7%), mCNV (n = 4, 33.3%), and PDR (n = 3, 25.0%). All studies reported the total number of injections. Nine (75.0%) studies reported the total number of patients, while three (25.0%) did not report the number of patients. Anti-VEGF agents included bevacizumab, ranibizumab, aflibercept, pegaptanib, and brolucizumab. Two (16.7%) studies did not specify the anti-VEGF agent used. Five (41.7%) studies reported the injection quadrant and mean number of injections per patient, while three (25.0%) reported lens status, though these data were limited to RD cases only. Two (16.7%) studies reported the injection quadrant for all injections. Baba et al. and Meyer et al. reported axial length data in patients who developed RRD, with mean axial lengths of 24.8 ± 1.3 mm (range 22.4–26.5 mm) and 25.1 ± 1.5 mm (range 23.3–27.1 mm), respectively. , Half of studies reported information on age and sex. Characteristics of each study, overall patients, and patients with RD are summarized in Tables 1–3 .
TABLE 1
Study Characteristics
| Study | Journal | Country of Publication | Study Design | Study Setting | Indications for Injection | Injection Provider | Anti-VEGF agent(s) used | Total Number of Injections | Total Number of Patients |
|---|---|---|---|---|---|---|---|---|---|
| Arevalo et al. | Ophthalmology | Colombia | Retrospective cohort | Multicenter (6 centers) | PDR | NR | Bevacizumab | 698 | 698 |
| Baba et al. | Japanese Journal of Ophthalmology | Japan | Retrospective case series | Single center | nAMD, DME, ME secondary to RVO, mCNV | Ophthalmologists | NR | 28 190 | 3040 |
| Bulla et al. | Arquivos Brasileiros de Oftalmologia | Brazil | Retrospective cohort | Single center | DME, AMD, ME secondary to RVO, PDR, myopic neovascular membrane, neovascular glaucoma | Single Ophthalmologist | Aflibercept, ranibizumab, bevacizumab, brolucizumab | 11 377 | 1024 |
| Gabrielle et al. | Ophthalmology Retina | Australia | Retrospective cohort | Multicenter | nAMD, DME, ME secondary to RVO | NR | Bevacizumab, ranibizumab, aflibercept | 264 272 | 13 690 |
| Hasler et al. | Acta Ophthalmologica | Denmark | Retrospective cohort | Single center | nAMD, DME, PDR, CNV associated with CSCR, BRVO, CRVO, mCNV, uveitis with ME | Nurses, ophthalmology residents, vitreoretinal surgeons | Ranibizumab | 38 503 | 3679 |
| Huang et al. | Journal of Clinical Research and Ophthalmology | United States | Prospective cohort | Multicenter (69 sites in 13 countries) | nAMD | Ophthalmologists | Pegaptanib, ranibizumab, bevacizumab | 3754 | 501 |
| Mammo et al. | Ophthalmology Retina | United States | Retrospective case series | Single center | nAMD | Ophthalmologists | NR | 203 000 | NR |
| Meyer et al. | Acta Ophthalmologica | Germany | Retrospective case series | Multicenter (6 centers) | nAMD, DME, ME secondary to RVO, mCNV, retinal hemangioma, subretinal hemorrhage, Irvine–Gass syndrome | Medical retina fellow, experienced ophthalmologist | Ranibizumab, bevacizumab | 35 942 | NR |
| Severn et al. | Eye | United Kingdom | Retrospective cohort | Single center | AMD | NR | Ranibizumab | 42 513 | 4727 |
| Simcock et al. | Eye | United Kingdom | Retrospective cohort | Single center | nAMD | Nurse practitioners | Ranibizumab | 11 893 | NR |
| Storey et al. | Ophthalmology | United States | Retrospective case series | Single center | nAMD, BRVO, CRVO | Ophthalmologists | Ranibizumab, bevacizumab, aflibercept | 180 671 | 12 718 |
| Xu et al. | International Ophthalmology | Singapore | Retrospective cohort | Single center | nAMD, DME, mCNV, CRVO, BRVO, neovascular glaucoma, ROP | Senior ophthalmology residents, medical retina specialists | Bevacizumab, ranibizumab, aflibercept | 14 001 | 2225 |
BRVO Branch Retinal Vein Occlusion; mCNV Myopic Choroidal Neovascularization; CRVO Central Retinal Vein Occlusion; CSCR Central Serous Chorioretinopathy; DME Diabetic Macular Edema; ME Macular Edema; nAMD Neovascular Age-Related Macular Degeneration; NR Not Reported; PDR Proliferative Diabetic Retinopathy; ROP Retinopathy of Prematurity; RVO Retinal Vein Occlusion.
TABLE 2
Overall Patient Characteristics
| Study | Mean Age (SD) | Sex (% Female) | Total Number of Eyes | Indication Distribution (%) | Needle Gauge | Injection Quadrant (%) | Follow-Up Duration |
|---|---|---|---|---|---|---|---|
| Arevalo et al. | NR | 52.0 | 548 | PDR (100.0) | NR | NR | NR |
| Baba et al. | NR | NR | NR | nAMD (71.9), RVO (14.5), DME (12.8), mCNV (0.8) | NR | NR | NR |
| Bulla et al. | 65.9 (9.7) | 54.0 | 1310 | DME (35.7), AMD (31.5), ME secondary to RVO (18.3), PDR (10.3), myopic neovascular membrane (1.7), neovascular glaucoma (1.6) | 30-gauge | IT (100.0) | ≥2 years |
| Gabrielle et al. | 77.0 (10.0) | 58.0 | 16 915 | nAMD (87.1), RVO (0.007), DME (0.006) | NR | NR | Up to 10 years |
| Hasler et al. | NR | NR | 4623 | nAMD (87.1), DME (6.5), PDR (0.2), CNV associated with CSCR (1.7), BRVO (1.8), CRVO (1.5), mCNV (1.1), uveitis with ME (0.1) | 30-gauge | ST (100.0) | 12.2 months (mean) |
| Huang et al. | 73.6 (8.7) | 61.3 | 560 | nAMD (100.0) | NR | NR | NR |
| Mammo et al. | NR | NR | NR | nAMD (100.0) | 30- or 31- gauge | ST (100.0) | 6 months minimum |
| Meyer et al. | NR | NR | NR | NR | NR | NR | NR |
| Severn et al. | NR | NR | NR | AMD (100.0) | NR | NR | 5 years |
| Simcock et al. | NR | NR | NR | nAMD (100.0) | NR | NR | NR |
| Storey et al. | 79.9 (NR) | 61.4 | NR | nAMD (80.9), BRVO (10.2), CRVO (8.9) | NR | IT (>90) |
≥3 months (93% of patients)
≥1 year (83% of patients) |
| Xu et al. | 68.5 (11.8) | 43.1 | NR | NR | NR | NR | 8 years |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
