Diabetic Macular Edema (DME) is a significant complication of diabetes mellitus, particularly prevalent among the elderly, and is a leading cause of vision impairment globally. The prevalence of diabetes is increasing, with projections indicating it will affect 592 million individuals worldwide within the next two decades, thereby escalating the burden of DME (1). DME leads to fluid accumulation in the macula, resulting in vision loss and impaired daily functioning (2, 3). The pathophysiology of DME involves chronic hyperglycemia, which triggers microvascular damage and inflammatory pathways, leading to the breakdown of the blood-retinal barrier (BRB) and subsequent fluid leakage into the retinal layers (3, 4). Key mediators such as vascular endothelial growth factor (VEGF) and inflammatory cytokines play crucial roles in macular swelling (2, 5).

Current management of DME primarily involves anti-VEGF therapies, which have revolutionized treatment by effectively reducing macular thickness and improving visual acuity. Agents like ranibizumab and aflibercept are well-established in clinical practice (2, 6). Conbercept, another anti-VEGF agent, acts by binding multiple VEGF isoforms and placental growth factor (PlGF), showing promise in clinical trials, although specific data on its efficacy in the elderly population remains limited (6). Despite the effectiveness of these treatments, there are notable limitations, including inter-patient variability in response and the need for frequent injections (4, 7). This necessitates ongoing research into extended durability drugs and alternative therapeutic strategies to improve patient outcomes and reduce treatment burdens (5, 6).

The assessment of aqueous humor biomarkers in DME is crucial for understanding local disease activity and predicting treatment outcomes. Intraocular markers, such as cytokines in the aqueous humor, are significant because they directly reflect the local inflammatory and angiogenic processes occurring in the eye, unlike systemic markers which may not accurately represent ocular conditions (8, 9). Elevated levels of cytokines like VEGF, IL-6, and MCP-1 in the aqueous humor have been consistently associated with DME, indicating their potential as biomarkers for disease severity and treatment response (8, 9). Specifically, higher baseline levels of VEGF and IL-6 have been linked to less favorable anatomical and visual outcomes following anti-VEGF therapy (10, 11). Real-world studies are essential as they provide insights into the effectiveness of treatments in diverse patient populations (12, 13). Identifying factors that predict which elderly patients will benefit most from anti-VEGF therapy is crucial (13, 14). Therefore, real-world data can help tailor more effective and personalized treatment strategies, improving outcomes for patients with DME across various demographics (14, 15).

Despite the growing application of anti-VEGF therapies for DME, a critical gap remains in understanding conbercept’s real-world efficacy specifically among older patients, particularly with respect to intraocular biomarker profiles. Although elevated VEGF and IL−6 levels are consistently associated with greater baseline macular edema, published reports diverge on whether these biomarkers ultimately portend a favorable or unfavorable response to anti−VEGF treatment. Some studies indicate that eyes with higher VEGF experience larger CST reductions once angiogenic drive is suppressed, whereas persistent IL−6–mediated inflammation may limit visual recovery. The present study therefore prospectively evaluates how baseline VEGF and IL−6 together influence both anatomical and functional outcomes in an elderly DME population. By focusing on these older individuals, our findings could refine patient selection and guide more personalized conbercept regimens, ultimately improving both clinical outcomes and resource allocation.

A total of 150 elderly patients with diabetic macular edema (DME) were enrolled. As shown in Table 1 , the mean age was 73.4 ± 5.2 years (range, 65–89), and 50.7% were male. The average duration of diabetes was 12.8 ± 5.6 years, with a mean HbA1c of 7.9 ± 1.2%. Most participants (62.7%) were phakic, and 62.7% had coexisting hypertension. Baseline best-corrected visual acuity (BCVA) was 50.3 ± 8.2 letters, and the central subfield thickness (CST) was 445.6 ± 65.3 µm.

Prior to the first conbercept injection, aqueous humor samples were collected and analyzed for VEGF, IL-6, IL-8, TNF-α, and bFGF. Table 2 summarizes these baseline biomarker levels, with mean ± SD values of 90.5 ± 35.2 pg/mL for VEGF and 45.8 ± 18.1 pg/mL for IL-6. The baseline levels of IL-8, TNF-α, and bFGF were 30.3 pg/mL, 8.7 pg/mL and 18.9 pg/mL, respectively.

All participants received a three-injection loading phase over the first 3 months. Subsequent treatments were administered pro re nata (PRN). As detailed in Table 3 , the mean number of injections at 6 months was 4.2 ± 1.1, with 62 patients (41.3%) requiring one additional post-loading injection, and 21.3% needing two or more. Adjunct focal laser was used in 7.3% of cases, mainly for persistent or focal edema.

Marked improvements in BCVA and CST were observed. At 6 months, BCVA increased by +8.6 ± 4.6 letters from baseline (p<0.001) and CST decreased by -85.1 ± 40.2 µm (p<0.001) ( Table 4 ). Approximately 30.7% of eyes achieved a ≥10-letter gain, and 20.7% reached a ≥15-letter improvement.

Baseline VEGF remained the only cytokine that retained statistical significance across all anatomical metrics after multiple−comparison correction: it correlated positively with absolute CST change (r = +0.33, FDR−adjusted p = 0.020) and, importantly, with percentage CST reduction (r = +0.31, FDR−adjusted p = 0.025) ( Table 5 ). By contrast, IL−6 lost significance when baseline thickness was normalized (r = +0.15, FDR−adjusted p = 0.220), suggesting its previously observed association was largely driven by greater initial oedema rather than intrinsic sensitivity to therapy. Neither IL−8, TNF−α, nor bFGF showed significant relations with anatomical outcomes once FDR correction was applied ( Table 5 ). Functionally, higher VEGF and IL−6 levels remained negatively correlated with BCVA gain at the unadjusted level (r = –0.37 and –0.26, respectively), but only VEGF showed an FDR−adjusted p−value below 0.05 (0.020) ( Table 5 ). Collectively, these findings indicate that VEGF provides independent, baseline−adjusted predictive information for anatomical response, whereas IL−6 primarily reflects disease severity rather than superior therapeutic efficacy.

Serious ocular adverse events were rare (1 case of endophthalmitis, 0.7%), and IOP elevations requiring medication occurred in 5.3% of patients ( Table 6 ). Mild anterior chamber inflammation, subconjunctival hemorrhage, and cataract progression were also documented but did not necessitate study discontinuation in most cases.

Participants were stratified into mild vs. moderate/severe DME based on baseline CST (<400 vs. ≥400 µm). Patients with moderate/severe DME had a higher baseline CST (470.5 ± 55.2 vs. 373.6 ± 21.4 µm, p<0.001), required more injections (4.5 ± 1.1 vs. 3.8 ± 1.2, p=0.005), and exhibited a slightly larger CST reduction (–86.2 ± 36.7 vs. –61.2 ± 28.5 µm, p=0.001) ( Table 7 ). Despite these differences, final BCVA at 6 months was comparable across subgroups ( Table 7 ).

A logistic model was employed to identify factors predicting a robust visual improvement (≥15 letters gained). Baseline CST (OR=1.08 per 10 µm, p=0.008), VEGF (OR=1.11 per 10 pg/mL, p<0.001), and IL-6 (OR=1.16 per 5 pg/mL, p=0.014) emerged as independent predictors ( Table 8 ). Conversely, neither age nor diabetes duration significantly influenced the odds of a ≥15-letter gain.

VEGF displayed the highest area under the curve (AUC=0.74, p<0.001). IL-6 had an AUC of 0.68, whereas IL-8, TNF-α, and bFGF demonstrated comparatively weaker predictive capabilities ( Table 9 ).

A linear mixed-effects model incorporating multiple time points (baseline, 1, 3, 6 months) confirmed a significant average monthly improvement in BCVA (+1.26 letters/month, p<0.001). Patients with “high VEGF” (>80 pg/mL) started with a lower initial BCVA but gained letters faster over time, reflected in a positive time × VEGF interaction (p=0.031) ( Table 10 ).

Our findings indicate that conbercept therapy produced marked anatomical (CST) and functional (BCVA) improvements in this elderly DME cohort. These gains occurred alongside biomarker patterns—higher baseline VEGF and IL−6 linked to larger CST reductions—that refine our understanding of treatment response. Our findings indicated a notable improvement in BCVA and a significant reduction in CST over six months of conbercept therapy, aligning with results from both randomized trials and observational studies where BCVA improvements of 0.41 ± 0.39 to 0.23 ± 0.20 logMAR and considerable CMT reductions have been documented (16, 17). Although one major trial did not observe a statistically significant CMT decrease (p = 0.385) (16), other long-term observational reports have shown meaningful reductions (e.g., from 510.9 ± 186.1 µm to 277.1 ± 122.9 µm) alongside a favorable safety profile and no severe adverse events (17). Our average injection frequency during the study period was lower than the 10.6 ± 2.0 injections reported over two years in a broader DME cohort (17), possibly reflecting differences in treatment protocols or comorbidity burdens in our older population. Regarding aqueous humor biomarkers, our correlation patterns for VEGF and IL-6 largely concur with prior studies indicating that these cytokines correlate with disease activity and thickness measures (18, 19), though variations in assay methods, sample sizes, and patient demographics—such as those with advanced non-proliferative diabetic retinopathy or prior panretinal photocoagulation—can lead to heterogeneous results (9, 20–22). Furthermore, our finding that older patients can exhibit elevated baseline cytokine levels is consistent with studies attributing age-related rises to both systemic and ocular factors (8, 21, 23). When CST reduction was expressed as a percentage of baseline thickness and baseline CST was entered as a covariate, VEGF—but not IL−6—remained an independent predictor, suggesting that VEGF carries predictive information beyond initial edema burden, whereas IL−6 predominantly reflects baseline severity. This distinction reconciles the paradox of robust anatomical response yet modest functional gain. Although baseline VEGF yielded the highest discriminative value in our cohort (AUC = 0.74), this level of accuracy is considered moderate. Accordingly, VEGF alone is unlikely to serve as a definitive gatekeeper for treatment intensification. Previous DME studies have reported similar single−marker AUCs (≈0.65–0.78), underscoring the inherent complexity of predicting therapeutic response from one molecule. Future work should evaluate combined models—including VEGF, IL−6, and structural OCT parameters—that may achieve higher composite AUCs and more clinically actionable thresholds.

Mechanistically, our data support the concept that elevated VEGF and IL-6 drive both angiogenic and inflammatory pathways, respectively, which can yield strong anatomical responses while producing variable visual outcomes (24–27). It is plausible that VEGF blockade successfully mitigates neovascular leakage and edema, but persistent inflammation modulated by IL-6 may limit functional gains, a notion consistent with reports describing robust anatomical improvements yet incomplete visual recovery (28, 29). Age-related physiological changes and systemic comorbidities may alter pharmacokinetics or ocular tissue susceptibility, potentially leading to delayed or atypical responses relative to younger cohorts (28, 29). This underscores the multifactorial nature of DME pathophysiology and the importance of integrating biomarker data into individualized treatment regimens, especially as age, metabolic status, and inflammatory burden converge to shape outcomes in this heterogeneous population.

Advanced age significantly impacts pharmacodynamics and clinical responses, primarily due to physiological changes that alter drug efficacy and safety. As individuals age, there is a decline in organ function, leading to decreased renal and hepatic clearance, which affects drug metabolism and elimination (30, 31). This results in an increased volume of distribution for lipid-soluble drugs and prolonged half-lives (30). Furthermore, older adults often exhibit heightened sensitivity to certain medications, particularly those affecting the central nervous system, such as benzodiazepines and opioids, due to age-related declines in CNS function (31, 32). Cardiovascular pharmacodynamics also show variability, with some agents demonstrating reduced effectiveness, while others may have increased effects in treatment-naive elderly patients (32). The complexity of managing medications in this population is compounded by polypharmacy and the potential for drug-drug interactions, particularly with direct oral anticoagulants used for venous thromboembolism (33). Overall, understanding these age-related changes is crucial for optimizing pharmacotherapy in geriatric patients (34).

The six-month follow-up may be insufficient for detecting long-term efficacy or potential late-onset complications, and the single-center setting could introduce selection bias or limit the applicability of our conclusions. Another limitation is the absence of randomization and a formal control group. Although this design enhances external validity by reflecting everyday practice, it precludes direct causal inference relative to untreated or alternative−treatment cohorts. We partially mitigated this limitation by adjusting for baseline covariates in multivariable mixed−effects models. We also acknowledge that biomarker levels can fluctuate based on subclinical inflammation, sample-handling variations, or collection timing, making consistent measurement crucial for robust analysis. Aqueous VEGF and IL−6 measured only at baseline cannot reflect the changes before and after treatment, serial sampling could have clarified whether conbercept directly modulates these cytokines over time or whether reductions correlate with anatomical response. Ethical and practical constraints precluded additional paracenteses in this elderly cohort, but future studies incorporating micro−volume sampling or tear−film surrogates are warranted. Although we employed strategies such as electronic record checks and phone reminders to minimize dropouts, missing data could still affect the generalizability of the results. Clinically, baseline cytokine levels may guide patient selection and individualized anti-VEGF regimens—particularly in those with elevated VEGF or IL-6 who might require more frequent injections or adjunctive anti-inflammatory agents—and future work should consider longer follow-up, comparative trials against other agents or implants, and biomarker-guided thresholds for therapeutic intensification. Further molecular and genetic studies could refine this biomarker-driven model, ultimately supporting more personalized and durable treatment strategies for elderly patients with DME.

In conclusion, our findings underscore the strong efficacy of conbercept in improving visual and anatomical outcomes among older diabetic macular edema patients, while also highlighting the pivotal role that baseline aqueous humor biomarker levels—particularly VEGF and IL-6—can play in identifying those most likely to benefit. These results collectively emphasize that incorporating biomarker profiling into routine assessment could enhance therapeutic decision-making, enabling clinicians to optimize treatment frequencies and potentially integrate adjunct anti-inflammatory strategies.