**Background:** Diabetic macular edema (DME) is a serious ocular complication of diabetes that can occur at any stage of diabetic retinopathy (DR) and is defined as retinal thickening in the macula involving or close to the center. Anti-vascular endothelial growth factor (anti-VEGF) therapy is the first-line treatment, but approximately 44–68% of DME eyes experience persistent DME after at least 24 weeks of therapy, and 32% have visual loss. The pathomechanism of DME is complex and multifactorial, involving VEGF, chronic inflammatory factors, and other biochemical pathways. Bibliometric analysis is a quantitative method to detect research trends, hotspots, and contributions of authors, institutions, and countries. This study aimed to conduct a bibliometric and visualized analysis of DME over a 20-year period (2003–2022) to disclose research topics and ascertain emerging focuses.
**Methods:** Data were collected from the Web of Science Core Collection (WoSCC) on January 9, 2023, using the search term TS = (“Diabetic Macular Edema” OR “Diabetic Macular Oedema”) for the timespan 2003-01-01 to 2022-12-31, with document types limited to “article” and “review” and language restricted to English. A total of 4482 publications were retrieved. Microsoft Excel 2010 was used to analyze publication trends and relative research interest (RRI), and a prediction model f(x) = ax^3 + bx^2 + cx + d was applied to estimate future cumulative publications. VOSviewer version 1.6.17 was used for coauthorship analysis (authors, institutions, countries) and co-occurrence analysis of the top 100 keywords, generating network and overlay visualization maps. The average appearing year (AAY) of keywords was calculated to identify recent hotspots.
**Key Results:** The annual global publications increased steadily from 36 in 2003 to 390 in 2022. The United States was the most productive country (1339 publications, 71,754 citations, H-index 121), followed by China (437 publications, 11,188 citations) and England (367 publications, 14,405 citations). Johns Hopkins University was the most productive institution (176 publications, 17,015 citations), and Bressler NM was the most productive author (76 publications, 9621 citations). The top 100 keywords were classified into five clusters: (1) therapy and adverse effects of DME (e.g., injection, vitrectomy, dexamethasone, corticosteroids); (2) clinical biomarkers of DME (e.g., optical coherence tomography, visual acuity, retinal thickness, hyperreflective foci); (3) mechanistic research on DME (e.g., endothelial growth factor, VEGF, inflammation, oxidative stress); (4) improving bioavailability and efficacy (e.g., ranibizumab, bevacizumab, aflibercept, brolucizumab, faricimab); and (5) early diagnosis of diabetic complications (e.g., diabetic retinopathy, prevalence, risk factors, deep learning). The most frequent keywords were “diabetic macular edema” (2137 occurrences), “retinopathy” (1365), “ranibizumab” (974), and “optical coherence tomography angiography” (56). Keywords with the highest AAY (most recent) included “deep learning” (AAY: 2020.83), “optical coherence tomography angiography” (AAY: 2019.59), “intravitreal aflibercept” (AAY: 2019.29), and “dexamethasone implant” (AAY: 2019.20). The model fitting curves predicted continued growth in publications globally and in the top five countries (United States, China, England, Japan, Germany), with China and England showing exponential growth.
**Clinical Implications:** This bibliometric analysis highlights that DME research has been dominated by anti-VEGF therapy over the past two decades, but emerging hotspots such as deep learning, optical coherence tomography angiography, intravitreal aflibercept, and dexamethasone implant indicate a shift toward more personalized and precise management. The identification of five keyword clusters underscores the multifaceted nature of DME research, encompassing therapy, biomarkers, mechanisms, drug delivery innovations, and early diagnosis. The growing interest in deep learning (AAY: 2020.83) for automated classification and early diagnosis of DME, with reported diagnostic accuracy of 93.78 ± 5.32% using 3D convolutional neural networks, suggests that artificial intelligence will play an increasing role in clinical practice. The focus on improving bioavailability and efficacy through novel agents like brolucizumab and faricimab, as well as sustained-release implants, addresses the limitations of frequent intravitreal injections and variable patient responses. These trends point toward individualized treatment strategies and multidisciplinary care involving ophthalmology, endocrinology, and nutrition to optimize patient outcomes.