Efficacy and Safety of Topical Insulin Eye Drops for Corneal Epithelial Defects: A Systematic Review, Meta-Analysis, and Grading of Recommendations Assessment, Development, and Evaluation Assessment

Abstract

Objectives

The aim of this study was to review and meta-analyze the efficacy and safety of topical insulin eye drops (TIED) in treating corneal epithelial defects (CED).

Methods

We registered the protocol in PROSPERO (CRD420251051879). A systematic literature search on PubMed, Cochrane, ScienceDirect, Scopus, and Google Scholar until May 2025 was done to identify controlled comparative studies. Outcomes of interest include time to complete re-epithelialization, re-epithelialization rate, treatment failure, recurrence, and adverse events. We performed meta-analysis using a random-effects model and assessed the certainty of evidence for each result using Grading of Recommendations Assessment, Development, and Evaluation (GRADE) assessment.

Results

Seven studies involving 238 patients were included in the analysis. TIED significantly shortened re-epithelialization time (mean difference [MD] –1.20 days [−1.71–−0.69], p<0.0001) and accelerated the healing rate (MD +0.26 mm²/h [0.10–0.42], p=0.002). In addition, TIED significantly reduced the risk of treatment failure (risk ratio [RR] 0.30 [0.16–0.57], p=0.003) and recurrence (RR 0.25 [0.11–0.56], p=0.0007) compared to conventional treatments, with no adverse events reported. GRADE assessments indicated very low to low certainty of evidence.

Conclusion

TIED may speed corneal healing, cut failures and recurrences, and is well-tolerated and inexpensive. Robust randomized controlled trials are still needed to nail down the optimal dosing, long-term safety, and its role in CED management.

Introduction

Corneal epithelial defects (CED) – breaks in the cornea’s outermost layer – predispose patients to infection, stromal scarring, persistent epithelial defect (PED), and permanent vision loss (1). The epidemiologic burden is substantial: PED affects around 245,000 patients in 2023 across the United States, Japan, and the five largest European countries (France, Germany, Italy, Spain, and the United Kingdom) (2). In an Iranian population study, almost half of adults ≥60 years had some corneal abnormality, with punctate epithelial defects present in 8.8% (3). Etiologies include trauma, infection, and ocular surgery (incidence after vitrectomy up to 22.4%,(4) neurotrophic keratopathy, and systemic diseases such as diabetes mellitus) (5,6).

Conventional therapy – preservative-free lubricants, topical antibiotics, bandage contact lenses, and autologous serum – often provides incomplete or delayed healing, particularly when corneal innervation or tear stability is compromised (7,8). Consequently, more effective, regenerative treatments are needed. Topical insulin eye drops (TIEDs) have emerged as a promising option, especially for refractory or neurotrophic CED. Beyond glucose regulation, insulin acts as a growth factor that stimulates CE proliferation, migration, and survival and suppresses ocular-surface inflammation (9,10). In diabetic animal models, it shows an effect of up-regulating Ki-67, lowering inflammatory cytokines, and reducing neutrophil infiltration, while also promoting corneal-nerve regrowth and elevating neuropeptides (neuropeptides substance P [SP] and calcitonin gene-related peptide [CGRP]) that modulate inflammation and oxidative stress (11).

Clinical data corroborate experimental work, as several trials showed that TIED reliably shrinks epithelial-defect area, speeds re-epithelialization in refractory PED of varying sizes, and lowers recurrence of recurrent erosions – all with excellent tolerance and safety profiles (12,13). Yet, dosing protocols and long-term safety still need definition, and the published evidence remains fragmented across small, heterogeneous trials (14-17). To resolve these gaps, we performed a systematic review and meta-analysis comparing TIED with conventional therapy for CED, focusing on time to complete healing, overall healing rate, treatment failure, recurrence, and adverse events.

Methods

We registered the protocol for this systematic review and meta-analysis in the International Prospective Register of Systematic Reviews (PROSPERO: CRD420251051879) and adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (18).

Eligibility Criteria

We selected studies through clearly defined inclusion and exclusion criteria. Eligible designs encompassed randomized or non-randomized controlled trials (RCTs) and prospective or retrospective comparative cohort studies. Any study enrolling patients – regardless of age – with CED (including PED, neurotrophic keratopathy, or post-operative epithelial breakdown) qualified for review. The intervention of interest was TIED, whether administered as a stand-alone therapy or as an adjunctive treatment to conventional care. Comparators had to comprise standard treatments such as artificial tears, autologous serum, bandage contact lenses, or other topical agents. To ensure consistency, when a single article contained multiple TIED formulations or control arms that fit our criteria, we combined those arms into one composite group for the primary analysis; we then tested the robustness of this decision by conducting sensitivity analyses in which the arms were re-separated and all possible pairings were analyzed independently (19).

A study needed to report at least one clinically relevant endpoint – time to complete re-epithelialization, rate of re-epithelialization, treatment failure or non-healing, recurrence of epithelial defects, or adverse events attributable to TIED – to be included. Only abstracts in English were considered, and if the full-text was in another language, we used DeepL (DeepL SE, Cologne, Germany) (20) to translate it. We excluded studies without a comparison group; case reports, case series, letters to editors, conference abstracts, expert opinions, and review articles; as well as animal or in vitro investigations. Duplicate publications or data subsets already incorporated into more comprehensive reports were likewise removed from consideration.

Literature Search and Study Selection

We performed a systematic search on PubMed, the Cochrane Central Register of Controlled Trials, ScienceDirect, Scopus, and Google Scholar from database inception to May 5, 2025 (last searched: May 5, 2025). The full search strings for each database are listed in Supplementary Table S1. Search strings combined relevant keywords and MeSH terms for “corneal epithelial defects,” “topical insulin,” “re-epithelialization,” and “recurrence,” joined with Boolean operators (“AND,” “OR”). Reference lists of all included studies were hand-searched, and the first author performed an additional manual search to capture records not indexed in the databases. No limits on publication year or language were applied at the search stage. All records were imported into Rayyan (Qatar Computing Research Institute, Doha, Qatar), (21) where duplicates were removed. Two reviewers (CW and AZ) independently screened titles and abstracts, followed by full-text assessment of potentially eligible studies; disagreements were settled by discussion or, when necessary, adjudication by another reviewer (MA).

Data Extraction and Risk of Bias (RoB) Assessment

Two reviewers (CW and AZ) independently extracted study details – author, year, setting, design, sample size, demographics, defect etiology and size – and key outcomes (time and rate of re-epithelialization, treatment failure, recurrence, and adverse events). They evaluated RoB using the Cochrane RoB 2 tool (22) for RCT, where domains of assessment include randomization, intervention deviations, missing data, outcome measurement, and selective reporting. For non-randomized studies, the Risk Of Bias In Non-randomized Studies – of Interventions (ROBINS-I) tool (23) was used, which assessed bias due to confounding, participant selection, intervention classification, deviations, missing data, outcome measurement, and selective reporting. Disagreements were resolved by consensus, with a third reviewer (MA) adjudicating when required.

Data Synthesis and Analysis

We summarized study characteristics in a descriptive table that listed study design, country, CED etiology, patient groups, age, sex, number of eyes, insulin dose regimen, baseline epithelial-defect area, primary outcomes, and a brief results summary. We synthesized all remaining qualitative information narratively.

Statistical analysis Process

We performed all meta-analyses in RStudio (v 2024.04.2, Posit Software, Boston, MA, USA). For continuous outcomes, time to complete re-epithelialization and re-epithelialization/healing rate, we expected between-study variation in definitions. Therefore, we extracted each study’s measurement techniques and intervals for these two outcomes (Supplementary Table S2) and used these to guide pooling versus narrative synthesis. If the variation is large, we consider synthesizing them narratively rather than pooling. Otherwise, we pooled the data as mean differences (MD) (95% confidence interval [CI]) using inverse-variance weighting in a random-effects model. For dichotomous outcomes, treatment failure and recurrence, we calculated pooled risk ratios (RR) with the Mantel–Haenszel method and a Paule-Mandel random effect, adding a 0.5 continuity correction to zero-event cells.(24)

To quantify heterogeneity, we calculated Cochran’s Q, I2, and τ2 with Q-profile confidence intervals, interpreting I2 values of 25%, 50%, and 75% as low, moderate, and high heterogeneity, respectively.(25) Where applicable, we explored heterogeneity with prespecified subgroup meta-analyses – surgical versus non-surgical etiology, dose-defined insulin concentration, and diabetes status – according to what each trial reported. We fitted random-effects models within each stratum and used a X² test for between-subgroup differences when both strata contained ≥2 studies; otherwise, we reported findings narratively. When a trial included multiple eligible insulin or comparator arms, we combined arms in the primary analysis to avoid double-counting and, in sensitivity analyses, re-separated the arms, and evaluated all valid pairings (19).

To assess design effects, for outcomes that mixed randomized and non-randomized evidence, we re-ran the meta-analysis using RCTs only. We did not pursue Bayesian or quality-weighted models because, with so few trials, posterior inferences would hinge on unverifiable priors for τ2 and the effect, adding assumptions without commensurate information. We, therefore, relied on the RCT-only restriction and discussed residual confounding from observational cohorts in the discussion (24). When at least ten studies are available, we explored publication bias with funnel plots and Egger’s test; (26) if fewer qualify, we reviewed study characteristics qualitatively to uncover selective reporting or design-related bias. We also performed the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) assessment (27) to assess the certainty of evidence from every synthesis.

Ethical Approval

Due to the nature of this study, which uses secondary anonymous data from the published literature, the Health Research Ethics Committee of the Faculty of Medicine at Universitas Indonesia have confirmed that this study was exempted from review for ethical approval. This study also follows the Tenets of the Declaration of Helsinki (2013 version).

Results

Literature Search

We found 554 search records in total, and through the rigorous selection process, we ultimately included seven studies (13-17,28,29). Figure 1 contains the PRISMA flow chart of this study.

Figure 1.

PRISMA flow chart of the study selection process.

Study Characteristics

Seven studies (266 eyes) got included in this systematic review and meta-analysis. The studies were conducted across diverse geographical regions, including Malaysia, Egypt, Spain, Mexico, and the United States, and encompassed a range of study designs: three RCTs, (14,15,28) two retrospective case–control studies, (16,29) one retrospective cohort study, (17) and one prospective non-RCT (13). The most common etiology for CED was post-operative complications following vitreoretinal surgery, reported in five studies. Other etiologies included neurotrophic keratopathy, immune-mediated ocular surface disease, and post-traumatic PED. Patient ages ranged from 25 to 72 years across studies, with varying gender distributions.

All studies administered TIED at concentrations between 0.5 U and 2 U per drop, applied 2–4 times daily. Control arms included routine steroid–antibiotic packs, preservative-free lubricants such as sodium hyaluronate (SH) or cornetears gel, autologous serum, and normal saline. Baseline epithelial areas, when reported, spanned roughly 4.7 mm² in post-traumatic defects to beyond 60 mm² in large post-vitrectomy lesions. Two trials (15,28) have more than two arms design. Fai et al. (15) tested 0.5, 1, and 2 IU/drop; we combined these three arms into one arm in for the meta-analysis. Meanwhile, Quiroz-Mendoza et al. (28) tested 0.5 IU/drop, SH, and the combination of the previous two; we combined all arms who received 0.5 IU/drop for the meta-analysis.

RoB in Included Studies

All three RCTs (14,15,28) were rated as having a low RoB across all RoB two domains. These trials demonstrated appropriate randomization procedures, minimal deviations from intended interventions, low levels of missing outcome data, and objective outcome measurements. In addition, there was no evidence of selective outcome reporting.

Among the four non-randomized studies, one study (29) was rated as low RoB based on the ROBINS-I tool, having clearly defined participant selection and balanced intervention groups, with no major concerns across domains. The remaining three studies (13,16,17) carried a moderate overall RoB, chiefly because they did not adjust for key confounders such as defect duration or etiology, systemic disease, or prior therapy; none employed matching or multivariable adjustment. Figure 2a and b details the domain-specific ratings.

Figure 2.

(a) Risk of bias assessment using the Cochrane Risk of Bias (RoB) 2 tool. (b) RoB assessment using the Risk of Bias in Non-randomized Studies – of Interventions tool.

Overview of Study Results

Across every study, TIED accelerated healing, increased closure rates, and reduced recurrences. Eleiwa et al. (16) halved the median healing time to 10.9 days, whereas Diaz-Valle et al. (17) achieved an 84% epithelialization rate versus 48% in controls and cut recurrences to 11% from 43%. Esmail et al. (13) observed no recurrences with insulin compared with 21.4% in the comparator group, and four studies (14,15,28,29) consistently reported faster re-epithelialization or superior healing rates without compromising safety. Fai et al. (15) also observed 0.5 IU/drop as the dose with the fastest result. A summary of study characteristics is presented in Table 1.

Table 1.

Characteristics of included studies comparing topical insulin eye drops vs. standard treatment

Time to Complete Re-epithelialization (Days)

Four studies (16,17,28,29) reported data on the duration required to achieve full corneal re-epithelialization in days. Because their outcome definitions were sufficiently comparable (Supplementary Table S2), we decided to pool them. The results showed that TIED significantly reduced healing time compared to conventional therapies. The MD was −1.08 days (95% CI: −1.53–−0.62; P < 0.01), indicating both statistically and clinically meaningful acceleration of corneal healing (Fig. 3a). High heterogeneity was detected (I2=84.3%, τ2=0, p<0.01), likely attributable to variations in baseline defect size, underlying pathology, and insulin dosing protocols. Nonetheless, all included studies consistently favored TIED. No adverse effects were reported. Sensitivity analysis using all possible pairings from the original and combined arms shows similar results (Supplementary Appendix 1). We stratified by etiology (surgical vs. non-surgical; Supplementary Appendix 2) to explore heterogeneity. In the surgical subgroup, heterogeneity persisted, indicating that etiology alone did not account for the dispersion. The non-surgical subgroup contained a single study, so meta-analysis was not feasible, and the subgroup-difference test was not interpretable. Overall, etiology stratification did not resolve the between-study variability and should be viewed as exploratory.

Figure 3.

(a-d) Forest plots of meta-analyses comparing topical insulin eye drops versus conventional therapies for corneal epithelial defects.

Re-epithelialization Rate (mm²/hour)

Five studies (14,15,17,28,29) assessed the rate of CE regeneration represented as area of re-epithelialization/hours (mm²/h). Because their outcome definitions were sufficiently comparable (Supplementary Table S2), we decided to pool them. We performed the meta-analysis where the data from Fai et al.(15) and Quiroz-Mendoza et al.(28) are from the combined arms and found that TIED significantly enhanced re-epithelialization rate, with a pooled MD of +0.27 mm²/h (95% CI: 0.10 to 0.44; p<0.01) (Fig. 3b). Substantial heterogeneity was noted (I² = 94%, τ2 = 0.03, p<0.01), likely due to variations in methodological design, insulin concentration, and wound measurement techniques. Despite this, all studies demonstrated a positive effect favoring TIED. Sensitivity analysis using all possible pairings from the original and combined arms shows similar results (Supplementary Appendix 3). We also performed the subgroup analysis to break down the heterogeneity using studies surgical etiology versus non-surgical etiology (Supplementary Appendix 2), and it turned out that the amount of heterogeneity in the pooling result using all studies with surgical etiology is still large. Using another subgroup comparison (studies with insulin concentration of 25 IU/mL [0.5 IU/drop] vs. non-25 IU/mL [non-0.5 IU/drop]), the heterogeneity slightly decreased in the group of studies with insulin concentration of 25 IU/mL (0.5 IU/drop), where the I² = 55%. Although a fixed-effect comparison suggested a difference between subgroups (X² = 43.6, p<0.01), the random-effects test, which accounts for substantial within-subgroup heterogeneity (I² = 95%), showed no significant difference (X² = 0.03, p=0.87). We, therefore, found no convincing evidence that the treatment effect varies between the subgroups. Meanwhile, in the RCT-only sensitivity analysis (Supplementary Appendix 3), the re-epithelialization rate remained higher with TIED (MD 0.30 mm²/h, 95% CI 0.15–0.44). Heterogeneity was moderate (I² 54%), while Cochran’s Q was non-significant (p=0.11), a common discordance with only three trials, because Q has low power; we, therefore, interpret heterogeneity mainly from I²/τ².

Failure of Healing

Two studies (16,17) contributed data on failure of epithelial healing, defined as the persistence of epithelial defects despite intervention. A random-effects model revealed that TIED significantly reduced the treatment failure risk, with a pooled RR of 0.30 (95% CI: 0.16 to 0.57; p=0.0003) (Fig. 3c). We found no heterogeneity (I² = 0%, τ² = 0, p=0.6334), suggesting consistency across studies.

Recurrence of Epithelial Defects

Recurrence rates were reported in two studies (13,17). Pooled analysis indicated that TIED significantly decreased recurrence, with an RR of 0.25 (95% CI: 0.11–0.56; p=0.0007), representing a 75% reduction in recurrence compared to controls (Fig. 3d). There was no detected (I²= 0%, τ² = 0, p=0.6443), further reinforcing the consistency of findings.

Adverse Events

There were no major adverse effects reported in any of the included studies. Across all trials, TIED was well tolerated with no systemic or local safety concerns documented. This seemingly clean profile is reassuring but not definitive. All studies were small and short, leaving them underpowered to detect rare or delayed toxicities such as late corneal neovascularization or epithelial hyperplasia. Sparse follow-up after epithelial closure further limits confidence. Future RCTs should predefine and grade ocular and systemic adverse events, maintain active surveillance for at least 6–12 months, and use masked adjudication so that a robust safety margin for TIED can be established.

Publication Bias

Due to the limited number of studies per comparison (n<10), we did not perform publication bias analysis, as small-study effects tests such as Egger’s test lack statistical power in this context. However, study characteristics were systematically reviewed to identify potential sources of bias.

GRADE Assessment

Using the GRADE framework, we rated the certainty of every pooled estimate in this review as low to very low. We downgraded the evidence chiefly for four reasons. First, several studies carried an appreciable RoB. Second, most analyses rested on small sample sizes, which heightens the chance of undetected publication bias and widens confidence intervals, generating imprecision. Third, high between-study heterogeneity in several outcomes signaled inconsistency. Finally, because only a handful of trials contributed data to each comparison, the results remain fragile and could shift with the addition of new evidence. Table 2 presents the summary-of-findings matrix and the corresponding GRADE ratings for every meta-analytic outcome.

Table 2.

Summary of findings and GRADE assessments.

Topical insulin eye drops versus conventional treatments for corneal epithelial defects
Outcomes	Effect (95%CI)	No. of eyes (studies)	Certainty of the evidence (GRADE)
Time to complete re-epithelialization (days)	MD=–1.08, 95% CI: –1.53––0.62, I2=84.3%	167 (4 studies)	⊕⊖⊖⊖ Very low1,2,3 Due to risk of bias among included studies (principally from the residual confounding in retrospective cohorts), the small sample size, and inconsistency
Re-epithelialization rate (mm2/hour)	MD=0.27,95% CI: 0.10–0.44, I2=94%	200 (5 studies)	⊕⊖⊖⊖ Very low1,2,3 Due to risk of bias among included studies (principally from the residual confounding in retrospective cohorts), the small sample size, and inconsistency
Failure of healing	RR=0.30, 95% CI: 0.16–0.57, I2=0%	121 (2 studies)	⊕⊕⊖⊖ Low1,2 Due to risk of bias among included studies (principally from the residual confounding in retrospective cohorts) and the small sample size
Recurrence of epithelial defects	RR=0.25, 95% CI: 0.11–0.56, I2=0%	117 (2 studies)	⊕⊕⊖⊖ Low1,2 Due to risk of bias among included studies (principally from the residual confounding in retrospective cohorts) and the small sample size

GRADE Working Group grades of evidence. High: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect. 1We decided to downgrade for 1 point due to risk of bias (principally from the residual confounding in retrospective cohorts) among included studies. 2We decided to downgrade for another 1 point due to the small sample size used in the synthesis of the meta-analysis result. 3We decided to downgrade for another 1 point due to inconsistency as shown by high heterogeneity found in the meta-analysis result. GRADE: Grading of recommendations assessment, development and evaluation, RR: Risk ratios, CI: Confidence interval.

Discussion

This comprehensive meta-analysis and systematic review demonstrates that TIED represents a transformative therapeutic intervention that has superior effects in treating CED compared to conventional treatments (12,15,30). Our finding is consistent with individual clinical trials reporting complete corneal re-epithelialization within 3–25 days depending on defect size and patient characteristics (12,17,28). Several systematic reviews without meta-analysis (10,30,31) also conclude the same, revealing superior healing outcomes of TIED.

Pathomechanistic Foundations and Molecular Rationale

The clinical efficacy of TIED derives from its multifaceted molecular mechanisms that address fundamental deficiencies in diabetic corneal wound healing (32,33). Insulin activates the PI3K-Akt signaling pathway in corneal epithelial and stromal cells, promoting cellular proliferation, migration, and survival through receptor-mediated mechanisms involving both insulin receptors and insulin-like growth factor receptors (34). This pathway activation correlates directly with enhanced DNA synthesis in basal epithelial cells within 48 h post-injury, explaining the accelerated healing observed clinically (33). In addition, insulin stimulates epidermal growth factor receptor phosphorylation and extracellular signal-regulated kinase activation, creating synergistic signaling cascades that optimize cellular migration and wound closure (32,34).

The therapeutic mechanism extends to neurotropic regeneration, with TIED promoting corneal nerve recovery and neuropeptide release, including SP and CGRP (16,35). This neurotropic action addresses the fundamental pathophysiology of diabetic keratopathy, where hyperglycemia-induced basement membrane damage and advanced glycation end-product accumulation compromise epithelial-stromal interactions (36). The anti-inflammatory properties of insulin, demonstrated through reduced interleukin-1 beta expression and neutrophil infiltration, create an optimal microenvironment for sustained tissue repair (16).

Heterogeneity Sources and their Impact

The wide dispersion in healing times and rates most likely stems from differences in baseline defect size, underlying pathology, insulin dose, and measurement technique. Subgroup exploration based on etiology included an insufficient number of studies; therefore, it remains uncertain whether etiology truly modifies the effect. Yet, it was certain that etiology was not the source of heterogeneity in the two outcomes: Time to re-epithelialization and re-epithelialization. Healing rates were likewise similar in trials that used approximately 25 IU/mL (0.5 IU/drop) insulin and those that used higher concentrations, but the number of studies were too small to establish a dose-response gradient. A planned comparison of diabetic versus non-diabetic eyes could not be carried out because the individual-patient data needed for that stratification were unavailable. Despite this variability, all studies consistently favored TIED, underscoring its robust therapeutic potential. Notably, TIED was effective in all patients regardless of the diabetic status, addressing a critical need given the impaired corneal healing often seen in diabetes (12,33). To reduce the residual heterogeneity presumably from unreported variation in defect chronicity, concomitant therapy, and wound-measurement methods; larger, well-reported trials that capture these covariates prospectively are required to determine definitively whether etiology, insulin concentration, or diabetes status modifies the treatment effect.

Population-specific Efficacy in the Vitreoretinal Surgery Context

The clinical relevance of TIED becomes particularly pronounced in post-vitrectomy populations, as they represent the majority of subjects included in the meta-analysis, where corneal complications affect 22.4% of patients, with 4.6% developing PED (4,37). Diabetic patients represent a uniquely vulnerable cohort, with diabetes mellitus, perfluoropropane tamponade, and surgical complexity serving as independent risk factors for PED after vitrectomy. In this high-risk population, TIED at 0.5 units/drop administered 4 times daily achieved 100% healing within 72 h, substantially outperforming placebo (62.5%) and higher concentrations (15,16). It might be due to receptor saturation kinetics that favor physiological rather than pharmacological dosing (31). This finding has profound implications for clinical implementation, as lower concentrations reduce preparation costs while maximizing therapeutic efficacy. The consistency of benefits across diabetic and non-diabetic populations, regardless of age, gender, or hypertensive status, underscores the universal applicability of insulin’s regenerative mechanisms (35).

Adherence Profile and Clinical Implementation Advantages

TIED demonstrates exceptional adherence characteristics that address traditional barriers to PED management (35). The formulation’s isotonic properties (280–300 mOsm/L), neutral pH (7–8), and low viscosity ensure optimal tolerability without ocular irritation, factors critical for sustained patient compliance in chronic conditions (38). The 4-times-daily dosing regimen aligns with standard ophthalmic medication schedules, facilitating integration into existing therapeutic routines without additional complexity (15,16).

A microbiological stability study confirms 28-day refrigerated storage capability with maintained insulin potency in the 90–110% range when formulated in normal saline, providing practical advantages for both compounding pharmacies and patient use (38). The absence of systemic absorption or glycemic effects eliminates concerns regarding diabetes management interference, a significant advantage over systemic interventions (15,33). Several previous reviews also demonstrate consistent safety profiles with no specific or major adverse events (10,31,35).

Economic Considerations and Healthcare Resource Optimization

The cost-effectiveness profile of TIED presents compelling healthcare economic advantages, particularly when considered against alternative interventions such as amniotic membrane transplantation or complex surgical procedures (35). Insulin’s widespread availability as a generic medication enables cost-effective compounding, with formulations prepared from commercially available subcutaneous insulin through simple dilution procedures (38). The accelerated healing timeline directly translates to reduced healthcare utilization, with mean healing times of 16.6±10.8 days for compounded preparations compared to conventional therapy timelines exceeding 40 days (4,38).

The prevention of surgical interventions represents substantial cost savings, as evidenced by the elimination of amniotic membrane transplantation requirements in insulin-treated groups compared to 11% (2/18) in control populations (16). Long-term economic benefits extend beyond direct treatment costs to encompass reduced complication management, with 77% of patients achieving complete improvement and significant visual acuity gains, minimizing long-term disability costs (35).

Practical Implications: Integration into Existing Treatment Algorithms and Compounding Logistics

Practical implications from the comparative studies are modest but clear. TIED were used as an adjunct to standard care for difficult or postoperative epithelial defects, rather than as monotherapy (14,15,17). The most common regimen was ~0.5 U/drop 4 times daily, prepared by diluting U-100 regular insulin to ~25 IU/mL; higher concentrations were also studied, but available trials do not establish a dose–response advantage over 0.5 U/drop (14,15,28). Patients were typically reviewed within 24–72 h with fluorescein photography to document the defect area, then followed at least weekly until closure; rescue measures (bandage lens or surgery) were considered when improvement failed over about a week (15,28,29). Compounding in these studies was aseptic and pharmacy-led, using regular insulin in 0.9% saline, dispensed in sterile ophthalmic droppers, refrigerated, and in some protocols replaced every ~3 days – practical details that hospitals or licensed compounders can reproduce while applying local beyond-use dating (15,28). Safety reporting was reassuring but short-term: Across controlled studies, no systemic hypoglycemia or vision-threatening events were attributed to TIED, yet small samples and limited follow-up mean uncommon or delayed harms cannot be excluded (14,15,17). Stability data from compounding research support refrigerated saline formulations, but real-world sterility and potency monitoring remain advisable (38).

Taken together, current evidence supports considering TIED as an early adjunct when post-operative or PED are not responding with standard measures, coupled with early reassessment, pharmacy-standard compounding, and prompt escalation if the epithelial area fails to decrease (14,15,17).

Strengths and Limitations of this Study and Future Research Recommendations

This appears to be the first meta-analysis synthesizing controlled human data on TIED for CED. Evidence is still thin: Only seven studies qualified, and few reported dichotomous outcomes such as recurrence or treatment failure. Several pooled estimates showed substantial heterogeneity, likely driven by variations in insulin formulation, dosing, defect etiology, and follow-up duration. Most trials provided only short-term data, and long-term efficacy and safety remain uncertain. Three studies were non-randomized, leaving residual confounding despite ROBINS-I assessment.

Future research should adopt standardized outcome definitions and uniform insulin preparations, recruit larger cohorts, extend follow-up, and incorporate patient-centered endpoints (visual acuity and quality of life). Well-designed RCTs comparing dose regimens and exploring combination therapies (e.g., insulin plus hyaluronic acid or autologous serum) could also clarify optimal strategies and mechanisms, including effects on corneal nerve regeneration. Concretely, future trials should be multicenter, parallel-group RCTs with concealed allocation and blinding (participants, clinicians, and image graders) using identical vehicles/labels. Use multi-arm or factorial designs for dose or add-on questions, with stratified randomization by etiology (post-surgical vs. neurotrophic/other) and diabetes status. Standardize and report compounding procedures, cold-chain handling, and bottle-replacement schedules. Employ centrally read, image-based outcomes: Time to complete re-epithelialization (no fluorescein staining on two exams ≥24 h apart) and re-epithelialization rate (mm²/h) from calibrated planimetry on a fixed schedule; also report proportion healed by certain timepoints (e.g. day 7/14 or further), need for rescue, recurrence at 1/3 months or further, Ocular Surface Disease Index (OSDI) score, and visual acuity, with ≥6–12 months’ follow-up. Predefine and report adverse-event categories with masked adjudication. Analyze by intention-to-treat, specify handling of two eyes per patient, adjust for baseline area/etiology/diabetes, and power for a clinically meaningful difference. For transparent reporting, register a protocol and publish a prespecified statistical analysis plan.

Conclusion

TIED emerges from this review as a promising, well-tolerated therapy that speeds corneal re-epithelialization, boosts overall healing, and lowers both treatment failures and recurrences across varied patient groups and defect etiologies. Readily available, inexpensive, and mechanistically compelling, TIED may serve as an adjunct – or even an alternative – to current CED treatment. Nonetheless, larger, rigorously designed trials are still needed to clarify the optimal dose and schedule, document long-term safety and durability, and anchor TIED within evidence-based guidelines for managing CED.

Supplementary

https://jag.journalagent.com/beyoglu/abs_files/BEJ-00821/BEJ-00821_(2)_Supplementary_Tables.pdf

Conflict of Interest

None declared.

Funding

The author declared that this study has received no financial support.

Use of AI for Writing Assistance

None declared.

Author contributions

Concept – C.W.; Design – C.W., M.A.; Supervision – M.A.; Resource – M.A.; Data Collection and/or Processing – C.W., A.Z.F.R.; Analysis and/or Interpretation – C.W., M.A.; Literature Search – C.W.; Writing – C.W., A.Z.F.R.; Critical Reviews – M.A.

Peer-review

Externally peer-reviewed.