A comprehensive systematic review and network meta-analysis evaluating topical eye-drop therapies indicates that no single pharmacological regimen demonstrates universally superior efficacy across every ocular surface measure following modern cataract surgery. Published in Ophthalmology and Therapy, the investigation synthesized evidence from forty-seven randomized controlled trials covering 102 study arms and 4,920 assessed eyes to examine how distinct topical medications manage postoperative ocular surface disruption. Although cataract procedures have increasingly transitioned from basic sight-restoring interventions to refractive surgeries requiring optimal visual quality, ocular discomfort and tear stability issues frequently persist. Across four separate network models tracking physiological recovery at one month, individual eye-drop formulations demonstrated isolated, outcome-specific benefits rather than consistent broad-spectrum improvements, leading researchers to conclude that clinical care cannot be guided by simplistic statistical ranking hierarchies.
The study protocol was prospectively registered in the PROSPERO international database under registration number CRD42026138249 and conducted in alignment with the PRISMA 2020 guidelines and its network meta-analysis extension. To assemble the trial corpus, investigators systematically searched nine electronic literature databases from their inception through January 30, 2026, including PubMed, the Cochrane Central Register of Controlled Trials, Embase, Web of Science, SpringerLink, ScienceDirect, and three major Chinese biomedical databases: CNKI, Wanfang Data, and VIP. Eligible trials were restricted exclusively to randomized controlled designs enrolling adult patients undergoing cataract surgery, predominantly phacoemulsification with intraocular lens implantation, who received postoperative or perioperative topical eye-drop regimens. Non-randomized cohorts, animal studies, single-arm studies, and abstract-only conference proceedings were excluded. From an initial retrieval of 441 records, 47 randomized controlled trials published between 2013 and 2026 satisfied all eligibility criteria, with 38 of these trials encompassing 81 arms and 3,898 analysis units contributing directly to at least one quantitative network synthesis.
To rigorously compare diverse commercial products and chemical agents, the researchers categorized treatments into a structured 19-node analytical framework based on active pharmaceutical composition and targeted ocular surface action. These defined treatment nodes included routine control care, saline or vehicle controls, hyaluronic acid monotherapy, non-hyaluronic-acid artificial tears, carboxymethylcellulose combined with hyaluronic acid, hydroxypropyl guar lubricants, lipid-containing lubricants, and trehalose paired with hyaluronic acid. Pharmacological nodes also comprised topical anti-inflammatory immunomodulators and tear secretagogues such as cyclosporine A, diquafosol sodium, and rebamipide, as well as tissue repair factors including deproteinized calf blood extract, recombinant human epidermal growth factor combined with hyaluronic acid, and recombinant bovine basic fibroblast growth factor with hyaluronic acid. Combination regimens such as cyclosporine A paired with carboxymethylcellulose, cyclosporine A with deproteinized calf blood extract, rebamipide with hyaluronic acid, and diquafosol sodium combined with hyaluronic acid were evaluated as distinct therapeutic entities, while standard postoperative background treatments like prophylactic antibiotics and corticosteroids administered uniformly across study arms did not define independent nodes.
Quantitative synthesis was conducted using frequentist random-effects network meta-analysis models in Stata 19, tracking four specific ocular parameters evaluated exactly one month following surgery. The primary clinical outcome was tear-film breakup time, an objective metric of tear-film stability measured in seconds, which synthesized 38 trials and 3,898 analysis units across all nineteen nodes. Secondary endpoints comprised tear secretion measured by the Schirmer test in millimeters across 26 trials and 2,709 analysis units; corneal epithelial damage scored via corneal fluorescein staining on a unified zero-to-five scale across 29 trials and 2,967 analysis units; and subjective symptom burden measured by the Ocular Surface Disease Index across 22 trials and 2,309 analysis units. Methodological consistency was maintained across the networks, as every included tear-film breakup trial utilized fluorescein dye and all Schirmer assessments were conducted without topical anesthetic drops, eliminating procedural discrepancies across trial cohorts.
For the primary endpoint of one-month tear-film breakup time, moderate-confidence evidence established that combination therapy using diquafosol sodium plus hyaluronic acid probably prolonged stability by a mean difference of 4.17 seconds compared with routine postoperative control care, within a 95% confidence interval of 1.99 to 6.35 seconds. Similarly, cyclosporine A paired with carboxymethylcellulose exhibited a moderate-confidence increase of 3.97 seconds, with a 95% confidence interval of 1.60 to 6.34 seconds. A third combination regimen consisting of cyclosporine A plus deproteinized calf blood extract yielded a numerically larger point estimate of 4.98 seconds, with a 95% confidence interval spanning 1.93 to 8.03 seconds; however, this finding was graded as low confidence due to substantial methodological fragility. Additional moderate-confidence improvements ranging between 2.02 and 3.02 seconds over control were identified for single-agent deproteinized calf blood extract, diquafosol sodium alone, recombinant human epidermal growth factor with hyaluronic acid, and cyclosporine A monotherapy, while conventional unsupplemented artificial lubricants yielded only low-confidence or negligible differences.
Analyses of tear production and corneal epithelial integrity demonstrated distinct therapeutic patterns across separate physiological mechanisms. In the Schirmer test network, moderate-confidence results demonstrated that cyclosporine A combined with carboxymethylcellulose produced the largest increase in tear volume over control, yielding a mean difference of 4.47 millimeters with a 95% confidence interval of 0.57 to 8.36 millimeters. Diquafosol sodium plus hyaluronic acid improved secretion by 2.77 millimeters with a 95% confidence interval of 0.16 to 5.38 millimeters, and diquafosol monotherapy generated an increase of 1.62 millimeters. When examining corneal fluorescein staining scores, moderate-confidence evidence showed that recombinant human epidermal growth factor combined with hyaluronic acid significantly lowered epithelial staining by a mean difference of negative 0.89 points, within a 95% confidence interval of negative 1.49 to negative 0.29 points. Lower staining scores were also estimated for cyclosporine A with carboxymethylcellulose and cyclosporine A with deproteinized calf blood extract, but both estimates were downgraded to low confidence due to severe local incoherence and elevated risk of bias.
Patient-reported symptom burden evaluated through the Ocular Surface Disease Index remained strictly exploratory due to data sparsity and trial limitations, precluding formal grading under the Confidence in Network Meta-Analysis framework. Compared with routine control care, lower symptom scores were documented for recombinant human epidermal growth factor paired with hyaluronic acid, which reduced the index by a mean difference of negative 8.69 points, alongside hyaluronic acid with Ginkgo biloba extract at negative 8.20 points, lipid-containing lubricants at negative 7.00 points, and trehalose with hyaluronic acid at negative 5.68 points. Conversely, rebamipide combined with hyaluronic acid was paradoxically linked to higher symptom scores of 11.71 points across two trials comprising 75 analysis units, an anomalous observation that authors noted lacked biological clarity and direct consistency with objective epithelial staining metrics. Adverse event reporting, medication withdrawals, and less common physiological biomarkers such as inflammatory cytokine concentrations and goblet-cell counts were inconsistently described across trials, preventing pooled quantitative network syntheses for surgical safety.
Critical methodological appraisal using the Cochrane Risk of Bias 2 tool revealed pervasive quality limitations across the evidence base. Of the 47 randomized controlled trials evaluated, only 14 were judged to be at an overall low risk of bias, while 32 trials raised some methodological concerns and one trial was rated at high overall risk of bias. Methodological vulnerabilities occurred most frequently in outcome measurement procedures and deviations from intended interventions, driven by incomplete investigator masking for subjective examinations. When the Confidence in Network Meta-Analysis framework was applied to evaluate within-study bias, indirectness, reporting bias, imprecision, heterogeneity, and incoherence, many favorable nodes lost certainty. Notably, top-performing nodes in surface curves of cumulative ranking were frequently underpinned by only one or two isolated clinical trials involving few participants, such as the cyclosporine A combinations which relied on single trials containing 25 and 67 analysis units.
Sensitivity analyses and statistical evaluations further underscored the fragility of several leading point estimates. When researchers removed the single trial presenting a high overall risk of bias from the tear-film stability network, estimates for established nodes remained broadly stable, but the cyclosporine A plus deproteinized calf blood extract node became completely inestimable because its sole supporting trial was eliminated. Furthermore, ninety-five percent prediction intervals calculated for the consistency model crossed zero for 14 out of the 18 active comparisons against control, demonstrating that future clinical replications could easily observe null or divergent effects.
Ultimately, the researchers emphasized that clinical decision-making following cataract surgery must avoid selecting topical eye drops based solely on statistical league table rankings or unverified claims of broad superiority. Because current randomized evidence originates primarily from trials conducted in East Asia and Europe with middle-aged to older participants, racial generalizability remains unconfirmed and regional drug availability varies considerably. The investigators concluded that clinicians must tailor perioperative care to specific clinical presentations, balancing epithelial damage, tear stability, preoperative surface health, and cumulative chemical preservative exposure. Future progress will require adequately powered, multicenter, double-masked head-to-head randomized trials that standardize postoperative anti-inflammatory regimens and provide long-term follow-up to definitively establish evidence-based guidelines for post-surgical ocular surface recovery.
Subject of Research: Topical eye-drop regimens for ocular surface management and dry eye following cataract surgery
Article Title: Comparative Efficacy of Topical Eye-Drop Regimens for Dry Eye After Cataract Surgery: A Systematic Review and Network Meta-analysis
Article References: Song, B., Li, Y., Liu, C., & Du, Z. (2026). Comparative Efficacy of Topical Eye-Drop Regimens for Dry Eye After Cataract Surgery: A Systematic Review and Network Meta-analysis. Ophthalmology and Therapy, 15(10), 3245-3272. https://doi.org/10.1007/s40123-026-01484-0
Image Credits: AI Generated
DOI: 10.1007/s40123-026-01484-0
Keywords: cataract surgery, dry eye disease, topical eye drops, network meta-analysis, tear-film breakup time, Schirmer test, corneal fluorescein staining, Ocular Surface Disease Index, diquafosol sodium, hyaluronic acid, cyclosporine A, carboxymethylcellulose, deproteinized calf blood extract, ocular surface health, postoperative recovery, ophthalmology
News Source: Ophelia Keating. (November 20, 2026). Network Meta-Analysis Evaluates Eye Drops for Dry Eye After Cataract Surgery. Scienmag.



