Efficacy and Safety of Reproxalap 0.25% Ophthalmic Solution in Dry Eye Disease: A Systematic Review and Meta-Analysis of Randomized Controlled Trials

Abstract

Purpose

Dry eye disease (DED) is an ocular surface disorder characterized by tear-film instability, inflammation, and symptoms that can impair visual function and quality of life. Reproxalap is a topical reactive aldehyde species (RASP) inhibitor. This systematic review and meta-analysis assessed the efficacy and safety of reproxalap 0.25% ophthalmic solution vs vehicle in adults with DED.

Methods

MEDLINE/PubMed, Embase, Cochrane CENTRAL/Cochrane Library, and ClinicalTrials.gov were searched through June 20, 2026. The PROSPERO-registered review (CRD420261425460) included randomized trials and public trial-registry results records with extractable 0.25% reproxalap-versus-vehicle data. Outcomes included Schirmer test change and response (at least 10 mm improvement), instillation-site irritation, treatment-emergent adverse events (TEAEs), and discontinuation. Risk of bias and certainty were assessed using RoB 2 and GRADE.

Results

Six eligible sources involving 2230 randomized participants were included. Reproxalap improved Schirmer test score (mean difference [MD], 2.34 mm; 95% confidence interval [CI], 1.45 to 3.22; P < 0.0001; I² = 0%) and Schirmer response (risk ratio [RR], 2.04; 95% CI, 1.68 to 2.46; P < 0.0001; I² = 0%). Instillation-site irritation was more frequent (RR, 27.27; 95% CI, 12.20 to 60.99; P < 0.0001) but was generally mild and transient in published reports. Any TEAE was uncertain (RR, 2.02; 95% CI, 0.92 to 4.44; P = 0.082). Adverse event-related discontinuation was more frequent (RR, 4.88; 95% CI, 2.36 to 10.10), whereas all-cause discontinuation did not differ (RR, 0.99; 95% CI, 0.50 to 1.99).

Conclusions

Reproxalap 0.25% demonstrated rapid objective tear-production improvements vs vehicle, though routine clinical recommendation must be weighed against frequent, transient instillation-site irritation. No treatment-related serious ocular adverse events were reported. Certainty ranged from low to very low due to reliance on registry-only data, trial program endpoint restructuring, eye-level denominators, and small study numbers.

Introduction

Dry eye disease (DED) is a multifactorial, symptomatic ocular surface disease characterized by loss of tear-film and/or ocular-surface homeostasis. Tear-film instability, hyperosmolarity, ocular-surface inflammation and damage, and neurosensory abnormalities contribute to its pathophysiology and clinical expression.¹ Current management is staged and mechanism-based and may include artificial tears, lubricating ointments, and prescription anti-inflammatory agents, such as cyclosporine and lifitegrast, when inflammatory ocular-surface activity is clinically relevant.²

Reactive aldehyde species (RASP), including malondialdehyde and 4-hydroxy-2-nonenal, are products of oxidative and cellular stress. They can covalently modify amino and thiol groups on receptors and signaling proteins, which may amplify inflammatory signaling upstream of cytokine release.^3,4 Reproxalap is an investigational small-molecule RASP inhibitor developed to bind free aldehyde species at the ocular surface.⁴

Clinical evidence for reproxalap 0.25% in DED includes peer-reviewed randomized trials, dry eye chamber studies, long-term safety data, and ClinicalTrials.gov results records.^5–10 These sources are not directly interchangeable. They differ in exposure setting, treatment duration, primary endpoints, reporting format, and denominator structure, including participant-level and eye-level outcomes.

These differences matter for evidence synthesis. Safety outcomes can be pooled when the participant-level denominator is clear, whereas several efficacy outcomes require separate or narrative handling because they are reported using heterogeneous scales, chamber conditions, time points, or eye-level units. Pooling such outcomes without preserving these distinctions may produce misleading precision and overstate the certainty of the evidence.

Therefore, this systematic review and meta-analysis aimed to evaluate the efficacy and safety of topical reproxalap 0.25% ophthalmic solution compared with vehicle in adults with DED, while separating peer-reviewed and registry-only evidence and preserving the original unit of analysis for participant-level and eye-level outcomes.

Methods

Study Design and Registration

This systematic review and meta-analysis was conducted according to the Cochrane Handbook for Systematic Reviews of Interventions and reported in accordance with the PRISMA 2020 statement.^11,12 The protocol was registered in PROSPERO (CRD420261425460).

Data Sources and Search Strategy

MEDLINE/PubMed, Embase, Cochrane CENTRAL/Cochrane Library, and ClinicalTrials.gov were searched through June 20, 2026. Reference lists of included reports were also checked. The complete search strategy and source-level record counts are provided in Supplementary Table S1.

The completed PRISMA 2020 checklist is provided in Supplementary Table S2.

Eligibility Criteria

Eligible sources were randomized controlled trials or ClinicalTrials.gov results records enrolling adults with DED and comparing topical ocular reproxalap 0.25% with vehicle/placebo. Multi-dose studies were eligible only when 0.25% reproxalap data were separately extractable.

Exclusion criteria were active-comparator-only studies, uncontrolled studies, non-randomized studies, animal or in vitro studies, reviews, editorials, duplicate records without usable results, and studies focused on non-DED indications.

Study Selection and Data Extraction

Two reviewers independently screened titles and abstracts and then full texts or registry records. Disagreements were resolved by discussion. Extracted variables included study design, population, intervention, comparator, follow-up, sample size, outcome definitions, means, standard deviations, event counts, denominators, and adverse event definitions.

When ClinicalTrials.gov reported standard errors, standard deviations were calculated using SD = SE x √N. Data were extracted as reported, without converting eye-level responder counts into participant-level events.

Schirmer Test Methodology and Pooling Restrictions

Because Schirmer tear-production testing can be performed with or without topical anesthesia, and because assessment timing differed across sources, Schirmer methodology was checked for each source before pooling. Garg et al⁶ and RENEW (NCT03879863) explicitly reported an unanesthetized Schirmer test. TRANQUILITY (NCT04674358) and TRANQUILITY 2 (NCT05062330) were conducted within the same reproxalap dry eye chamber program and assessed Schirmer test using the same single-day, four-dose, post-fourth-dose protocol; anesthesia status was not restated in the registry results records for these two trials, so an unanesthetized protocol consistent with the remainder of the reproxalap program is assumed rather than independently confirmed. Clark et al⁵ assessed Schirmer test as a chronic 12-week objective sign rather than as an acute same-day measure, so its timing and clinical meaning differ materially from TRANQUILITY and TRANQUILITY 2. To avoid pooling fundamentally different Schirmer assessment paradigms, quantitative synthesis of Schirmer test change from baseline and Schirmer responders was restricted to TRANQUILITY and TRANQUILITY 2, the only two sources that shared assessment timing, dosing protocol, and, to the extent verifiable, anesthesia status. Garg et al⁶ and Clark et al⁵ Schirmer data are presented narratively in Table 1 and were not included in the pooled Schirmer estimates.

Table 1.

Summary of Included Studies and Trial Registry Records

Source	Design	N randomized/analyzed	Intervention and comparator	Follow-Up	Main use in synthesis
Clark 20215	Published multicenter phase 2b, randomized, double-masked, vehicle-controlled parallel RCT	N=300 randomized; 0.25% reproxalap n=100; vehicle n=100	Reproxalap 0.25% QID vs vehicle	12 weeks	Safety and chronic field evidence; symptoms/signs narrative
Garg 20266	Published vehicle-controlled, randomized-sequence, double-masked, 2-period crossover dry eye chamber trial	N=63; sequences n=32 and n=31	Reproxalap 0.25% vs vehicle; 7–14 day washout	Acute chamber; 90 minutes	Acute redness/symptoms; crossover-specific interpretation
Radcliffe 20267	Published Phase 3 multicenter, double-masked, randomized, vehicle-controlled safety trial	N=757 treated overall; total treated reproxalap n=504 and vehicle n=253; primary 6-week safety population n=503/251; 12-month population n=299/148	Reproxalap 0.25% QID then BID vs vehicle	6 weeks or 12 months	Long-term safety; sensitivity analysis with 12-month population
TRANQUILITY (NCT04674358)9	ClinicalTrials.gov randomized vehicle-controlled results record	Safety N=327; Schirmer analyzed n=153/153	Reproxalap 0.25% vs vehicle	Acute/chamber assessment	Schirmer change, Schirmer responders, redness, instillation-site irritation, all-cause discontinuation
RENEW (NCT03879863)8	ClinicalTrials.gov randomized vehicle-controlled results record with QID and QID-to-BID schedules	Safety N=422; two reproxalap arms n=105/105; two vehicle arms n=106/106	Reproxalap 0.25% QID or QID-to-BID vs matched vehicle	Multi-week dosing	Instillation-site irritation primary/sensitivity; symptom/staining data not pooled because endpoints differed
TRANQUILITY 2 (NCT05062330)10	ClinicalTrials.gov randomized vehicle-controlled results record	N=361; reproxalap n=181; vehicle n=180	Reproxalap 0.25% vs vehicle	Acute/chamber assessment	Schirmer change, Schirmer responders, instillation-site irritation, all-cause discontinuation

Note: Total randomized sample across the six included sources was 2230 participants. Denominators differ by outcome because some registry outcomes reported analyzed populations or eye-level responder denominators.

Outcomes and Special Data Handling

Main efficacy outcomes were change in Schirmer test score from baseline and Schirmer responders, defined as at least 10 mm improvement. Additional efficacy outcomes included ocular or conjunctival redness and symptoms when extractable.

Safety outcomes included instillation-site irritation, any TEAE, treatment-related serious ocular TEAEs, discontinuation due to adverse events, and all-cause discontinuation.

Registry records: The three ClinicalTrials.gov results records - RENEW,⁸ TRANQUILITY,⁹ and TRANQUILITY 2¹⁰ - were included as eligible randomized evidence because they reported extractable 0.25% reproxalap vs vehicle data. Eye-level responder data were retained as eye-level denominators and were not described as participant-level events.

Dose-arm and safety-population handling: For RENEW (8; NCT03879863), the QID-to-BID regimen was used in the primary instillation-site irritation analysis to avoid double-counting participants across dosing schedules; the QID regimen was assessed in sensitivity analysis. For Radcliffe 2026,⁷ the denominator was selected according to the specific safety population used in each analysis: the 6-week population for the primary pooled safety analysis and the 12-month population for sensitivity analysis.

Risk of Bias and Certainty of Evidence

Risk of bias was assessed using the Cochrane RoB 2 framework.¹³ Crossover-specific concerns were considered for the dry eye chamber trial. Certainty of evidence was assessed using GRADE principles,¹⁴ considering risk of bias, inconsistency, indirectness, imprecision, and publication or reporting bias.

Statistical Analysis

Meta-analysis was performed using R version 4.5.2. Mean differences (MDs) with 95% confidence intervals (CIs) were calculated for continuous outcomes, and risk ratios (RRs) with 95% CIs were calculated for dichotomous outcomes. Random-effects models were used when clinical or statistical heterogeneity was expected; common-effect models were used when clinically justified by sparse events or limited between-study heterogeneity. Heterogeneity was assessed using I² and the chi-square test.

Sensitivity analyses were conducted for instillation-site irritation when alternative dosing or follow-up populations were available. Trial sequential analysis was performed for Schirmer responders. Funnel-plot assessment was not performed because fewer than 10 studies contributed to each meta-analysis, making formal small-study-effect testing unreliable.

Results

Study Selection

PRISMA Flow

Database searches identified 82 records (PubMed = 25, CENTRAL/Cochrane = 15, Embase = 42), and ClinicalTrials.gov/register searches identified 12 records, yielding 94 records before de-duplication. After removal of 31 duplicate records, 63 records were screened. Forty-four records were excluded at title/abstract screening. Nineteen reports or records were assessed for eligibility; 13 were excluded, leaving six included sources (three published reports and three ClinicalTrials.gov results records) (Figure 1). Reasons for full-text or record exclusion were wrong comparator or no vehicle control (n = 4), wrong design, duplicate record, or no usable extractable DED 0.25% data (n = 5), and wrong population or indication (n = 4).

Figure 1.

PRISMA 2020 flow diagram of study selection. Counts reflect 3 published reports and 3 ClinicalTrials.gov results records included in the review.

Study Characteristics

Included evidence: The six included sources comprised three peer-reviewed published reports and three ClinicalTrials.gov results records, with 2230 randomized participants in total. The published reports were Clark et al,⁵ Garg et al,⁶ and Radcliffe 2026.⁷ Registry-only sources were RENEW (8; NCT03879863), TRANQUILITY (9; NCT04674358), and TRANQUILITY 2 (10; NCT05062330). Study characteristics are summarized in Table 1, and baseline/sample characteristics, including baseline disease-severity indicators, are summarized in Table 2. No included source reported a formal categorical dry eye disease severity grading (for example, a defined mild/moderate/severe classification with per-group patient counts); consequently, Table 2 instead reports the objective baseline severity anchors that each source did provide, such as corneal or conjunctival staining scores, Schirmer test eligibility ranges, and baseline symptom questionnaire scores, and explicitly notes where this information was not available in the source.

Table 2.

Baseline and Sample Characteristics

Source	Available Baseline/Sample Information
Clark 20215	0.25% reproxalap n=100 and vehicle n=100. Mean age approximately 63 years in both groups; female participants: 75% vs 72%; White race: 77% vs 80%; Black race: 22% vs 17%. Baseline severity anchors (no categorical mild/severe grading reported): unanesthetized Schirmer test 5.6–6.0 mm; OD4SQ dryness/discomfort/grittiness/stinging scores 1.5–3.2 on 0–5 scales; nasal fluorescein staining 1.8–1.9 and lissamine green staining 1.6–1.7 on 0–4 scales; trial eligibility required Schirmer 1–10 mm and sum corneal fluorescein staining ≥4.
Garg 20266	Crossover sequence groups: reproxalap-to-vehicle n=32 and vehicle-to-reproxalap n=31. Mean age 47 vs 51 years; female participants 53% vs 77%; White race 75% vs 61%; Asian race 19% vs 19%. Baseline severity anchors (no categorical mild/severe grading reported): fluorescein staining sum 4.2–4.4 (0–5 per region; eligibility threshold ≥4); baseline investigator-assessed ocular redness 0.85 (0–4 scale); years with DED 7.3–8.1; eligibility also required ≥15-point dryness and ≥0.5 redness increase over ≥2 chamber time points.
Radcliffe 20267	Overall treated population: reproxalap n=504 and vehicle n=253. Primary instillation-site irritation analysis used the 6-week safety population n=503/251; 12-month sensitivity population n=299/148. Mean age approximately 55–57 years; female participants approximately 69–74% across groups. Baseline severity anchors (no categorical mild/severe grading reported): unanesthetized Schirmer test 6.4–6.6 mm (eligibility range 1–10 mm); OD4SQ ocular dryness approximately 3.5/5, ocular discomfort approximately 2.9–3.0/5; eligibility required OD4SQ score ≥2 in ≥1 symptom.
TRANQUILITY (NCT04674358)9	Registry results used for outcome denominators: safety n=164/163; Schirmer analyzed n=153/153; Schirmer responders reported at eye level n=306/306. Detailed baseline table, including any categorical mild/severe disease-severity grading, was not available in the supplied registry extraction and could not be independently confirmed.
RENEW (NCT03879863)8	Registry results reported two dosing schedules: QID and QID-to-BID. Safety arms were reproxalap n=105/105 and vehicle n=106/106. Detailed baseline table, including any categorical mild/severe disease-severity grading, was not available in the supplied registry extraction and could not be independently confirmed.
TRANQUILITY 2 (NCT05062330)10	Registry results used for outcome denominators: reproxalap n=181 and vehicle n=180; Schirmer responders reported at eye level n=362/360. Detailed baseline table, including any categorical mild/severe disease-severity grading, was not available in the supplied registry extraction and could not be independently confirmed.

Notes: No included source reported a formal categorical dry eye disease severity grading (eg, a defined mild/moderate/severe classification with per-group patient counts). Baseline severity is instead characterized using the objective anchors each source reported (staining scores, Schirmer eligibility range, symptom questionnaire scores) or is marked as unavailable for the three registry-only sources.

Risk of Bias

Overall judgments were mostly “some concerns”, mainly because several outcomes were available only from registry results records, some endpoints were not fully reported in peer-reviewed manuscripts, and the chamber/crossover designs required specific unit-of-analysis considerations (Table 3).

Table 3.

Risk of Bias Assessment (RoB 2)

Source	Randomization	Deviations	Missing Data	Outcome Measurement	Selective Reporting	Overall
Clark 20215	Low	Low	Some concerns	Low	Some concerns	Some concerns
Garg 20266	Low	Low	Some concerns	Low	Some concerns	Some concerns
Radcliffe 20267	Low	Low	Some concerns	Low	Some concerns	Some concerns
TRANQUILITY (NCT04674358)9	Some concerns	Some concerns	Some concerns	Some concerns	Some concerns	Some concerns
RENEW (NCT03879863)8	Some concerns	Some concerns	Some concerns	Some concerns	Some concerns	Some concerns
TRANQUILITY 2 (NCT05062330)10	Some concerns	Some concerns	Some concerns	Some concerns	Some concerns	Some concerns

Notes: Registry records were generally judged as having some concerns because they provided extractable results but less methodological detail than full peer-reviewed reports.

Efficacy Outcomes

Schirmer test change from baseline: TRANQUILITY⁹ and TRANQUILITY 2¹⁰ contributed change-from-baseline Schirmer test data. Reproxalap 0.25% improved Schirmer test score compared with vehicle (MD, 2.34 mm; 95% CI, 1.45 to 3.22; P < 0.0001), with no observed statistical heterogeneity (I² = 0%) (Figure 2).

Figure 2.

Forest plot of change in Schirmer test score from baseline: reproxalap 0.25% vs vehicle. The quantitative synthesis includes TRANQUILITY and TRANQUILITY 2, the two trials sharing the prespecified acute chamber assessment paradigm. Bold text denotes the pooled model estimate.

Schirmer responders: TRANQUILITY⁹ and TRANQUILITY 2¹⁰ contributed eye-level responder data. Reproxalap increased the proportion of eyes achieving at least 10 mm improvement in Schirmer test score (RR, 2.04; 95% CI, 1.68 to 2.46; P < 0.0001; I² = 0%) (Figure 3). Trial sequential analysis supported crossing of the monitoring boundary for benefit (Supplementary Figure S1).

Figure 3.

Forest plot of Schirmer test responders (at least 10 mm improvement): reproxalap 0.25% vs vehicle. Bold text denotes the pooled model estimate.

Ocular/conjunctival redness: Garg et al⁶ and TRANQUILITY⁹ contributed ocular or conjunctival redness data. The pooled estimate did not show a statistically significant difference between reproxalap and vehicle (MD, −0.01; 95% CI, −0.07 to 0.05; P = 0.5160), with substantial heterogeneity (I² = 87.5%; P = 0.0046) (Supplementary Figure S2). This analysis should be interpreted cautiously because the contributing sources differed in trial design, measurement setting, and timing of assessment.

Symptom outcomes: Symptom outcomes were not pooled because instruments, timing, and reporting formats differed across sources. The peer-reviewed trials nevertheless provided supportive narrative evidence: Clark et al⁵ reported improvement in ocular dryness with 0.25% reproxalap, Garg et al⁶ reported lower acute chamber symptoms after reproxalap vs vehicle, and registry-reported data from RENEW⁸ were directionally mixed for subject-reported ocular dryness, favoring reproxalap in the QID-to-BID arm but favoring vehicle in the QID arm. Because these symptom endpoints were not consistently extractable in a common format, they were retained as narrative evidence rather than a quantitative synthesis; the contributing studies, instruments, assessment time points, and directional results are summarized descriptively in Table 4.

Table 4.

Symptom Outcomes: Descriptive Summary of Unpooled Patient-Reported Evidence

Source	Instrument(s)	Domain(s) Assessed	Time Point(s)	Result (Direction/Significance)	Notes/Caveats
Clark 20215 (peer-reviewed, 12-wk parallel RCT)	OD4SQ (0–5); SANDE frequency/severity (0–100 VAS); OSDI (0–100)	Dryness, burning, grittiness, stinging, discomfort	Baseline; Weeks 2, 4, 8, 12	0.25% reproxalap superior to vehicle for OD4SQ dryness by MMRM across all visits (P=0.047) and for proportion reporting dryness=0 by GEE (P=0.012); numerically but not independently superior on other OD4SQ, SANDE, and OSDI scales; all-scales-combined improvement significant at first visit in patients with high baseline scores (P<0.0001).	Peer-reviewed. Different scale ranges (5-, 100-point) across instruments; not convertible to a common SMD without patient-level SDs of change.
Garg 20266 (peer-reviewed, acute chamber crossover)	0–100 VAS (dryness, discomfort, burning, grittiness, stinging); 0–4 scale (itching)	Dryness, discomfort, burning, grittiness, stinging, itching	Every ~10 min over 90-min chamber exposure; itching pre/post chamber only	All five VAS symptom domains significantly lower with reproxalap vs vehicle in-chamber (P<0.0001 to P=0.004 by domain); itching improved pre-to-post chamber (P=0.004).	Peer-reviewed, within-patient crossover. Symptom scores measured outside the chamber were low and did not differ between groups (Supplemental Table S2 of source), ie, benefit was acute-exposure-specific.
RENEW8 (NCT03879863; registry-only, unpublished)	0–100 VAS Subject-reported Ocular Dryness Score	Ocular dryness	Change from baseline, QID arm and QID-to-BID arm (registry-reported)	QID-to-BID arm: reproxalap −17.8 vs vehicle −6.8 (favors reproxalap). QID arm: reproxalap −11.5 vs vehicle −15.0 (favors vehicle, ie, discordant with the QID-to-BID arm).	Registry-only; not peer-reviewed; could not be independently verified against a primary results table. Directionally inconsistent between dosing arms and reported here without smoothing.
TRANQUILITY (NCT04674358)9 and TRANQUILITY 2 (NCT05062330);10 registry-only	Not extractable from supplied registry data	Not applicable	Not applicable	No participant-level symptom VAS/questionnaire outcome was extractable from the registry results records supplied for this review.	A company communication describes a statistically significant VAS ocular dryness improvement (P=0.003) in a 23-patient TRANQUILITY9 run-in cohort, not the completed trial’s registry results; this preliminary figure is not included as evidence here and should not be conflated with the full-trial result.

Notes: Symptom/patient-reported outcomes were not meta-analyzed because instruments, symptom domains, scale ranges, and assessment time points differed across sources, and forcing them into a common standardized mean difference would overstate precision. This table reports them descriptively rather than omitting them, consistent with dry eye disease being a primarily symptomatic condition. Directional results should not be interpreted as effect sizes.

Safety and Tolerability Outcomes

Instillation-site irritation: In the primary instillation-site irritation dataset, 953 participants in the reproxalap group and 700 participants in the vehicle group contributed data. Reproxalap was associated with a higher risk of instillation-site irritation (RR, 27.27; 95% CI, 12.20 to 60.99; P < 0.0001), with moderate heterogeneity (I² = 56.6%) (Figure 4). The direction of effect was consistent across acute chamber and chronic/longer safety settings (Figure 5).^7–10

Figure 4.

Forest plot of instillation-site irritation: reproxalap 0.25% vs vehicle. Bold text denotes the pooled model estimate.

Figure 5.

Subgroup analysis of instillation-site irritation by trial type (acute chamber vs chronic/longer safety trials). Bold text denotes pooled/subgroup model estimates.

Sensitivity analyses: Replacing the RENEW⁸ QID-to-BID arm with the RENEW⁸ QID arm produced a similar result (RR, 18.37; 95% CI, 10.50 to 32.15; P < 0.0001; I² = 44.1%) (Supplementary Figure S3). Substituting the Radcliffe et al⁷ 6-week safety population with the 12-month population also preserved the direction of effect (RR, 30.39; 95% CI, 17.10 to 54.01; P < 0.0001; I² = 3.1%) (Supplementary Figure S4).

Other safety outcomes: The effect on any TEAE was uncertain, with a numerically higher estimate that did not reach statistical significance (RR, 2.02; 95% CI, 0.92 to 4.44; P = 0.082) and substantial heterogeneity (I² = 94.3%) (Supplementary Figure S5). Discontinuation due to adverse events was more frequent with reproxalap (RR, 4.88; 95% CI, 2.36 to 10.10; P = 0.0075; I² = 21.4%) (Figure 6), whereas all-cause discontinuation did not differ between groups (RR, 0.99; 95% CI, 0.50 to 1.99; P = 0.9995; I² = 0%) (Supplementary Figure S6). No treatment-related serious ocular adverse events were reported in the published long-term safety trial.⁷

Figure 6.

Forest plot of discontinuation due to adverse events: reproxalap 0.25% vs vehicle. Bold text denotes the pooled model estimate.

Certainty of Evidence

GRADE certainty ranged from low to very low across outcomes (Table 5). Downgrading was mainly driven by reliance on registry-only data for several outcomes, the small number of studies per synthesis, inconsistency across study designs and follow-up durations, eye-level responder denominators, and imprecision of several estimates.

Table 5.

GRADE Certainty of Evidence

Outcome	Contributing Sources	Pooled Result	Certainty	Main Downgrade Reasons
Schirmer test change from baseline	TRANQUILITY;9 TRANQUILITY 210	MD = 2.34 mm (95% CI 1.45 to 3.22); I2 = 0%	Very low	Registry-only data, short-term assessment, imprecision
Schirmer responders (≥10 mm)	TRANQUILITY;9 TRANQUILITY 210	RR = 2.04 (95% CI 1.68 to 2.46); I2 = 0%	Very low	Eye-level denominators, registry-only data, indirectness
Ocular/conjunctival redness	Garg 2026;6 TRANQUILITY9	MD = −0.01 (95% CI −0.07 to 0.05); I2 = 87.5%	Very low	Heterogeneity, endpoint setting differences, imprecision
Instillation-site irritation	Radcliffe7 6-week safety population; TRANQUILITY;9 RENEW;8 TRANQUILITY 210	RR = 27.27 (95% CI 12.20 to 60.99); I2 = 56.6%	Low	Risk-of-bias concerns and heterogeneity, but consistent direction
Any TEAE	Clark; Radcliffe7 12-month population	RR = 2.02 (95% CI 0.92 to 4.44); I2 = 94.3%	Low	High heterogeneity and imprecision
Discontinuation due to adverse events	Clark; Radcliffe7 total treated population	RR = 4.88 (95% CI 2.36 to 10.10); I2 = 21.4%	Low	Few studies and event-level imprecision
All-cause discontinuation	TRANQUILITY;9 TRANQUILITY 210	RR = 0.99 (95% CI 0.50 to 1.99); I2 = 0%	Very low	Registry-only data and imprecision

Discussion

Dry eye disease is a common and heterogeneous ocular surface disorder defined by loss of tear-film homeostasis, with tear-film instability, hyperosmolarity, ocular-surface inflammation and damage, and neurosensory abnormalities contributing to its pathogenesis.¹ Current prescription anti-inflammatory options, including calcineurin inhibition and lymphocyte function-associated antigen-1 antagonism, are clinically useful but may be limited by delayed onset, variable effects on objective signs, and treatment-related discomfort that can affect adherence.^15–19 Reproxalap 0.25% ophthalmic solution is a small-molecule RASP inhibitor developed to target upstream pro-inflammatory aldehyde mediators at the ocular surface.^4–6 To our knowledge, this is the first systematic review and meta-analysis to quantitatively synthesize the efficacy and safety of reproxalap 0.25% across randomized evidence in DED. The analysis included up to six randomized sources and 2230 randomized participants.

The clearest efficacy signal was improvement in tear production. Reproxalap significantly increased Schirmer test scores compared with vehicle (MD, 2.34 mm; 95% CI, 1.45 to 3.22; P < 0.0001), with no statistical heterogeneity across the contributing trials (I² = 0%). The responder analysis for an increase of at least 10 mm also favored reproxalap (RR, 2.04; 95% CI, 1.68 to 2.46; P < 0.0001; I² = 0%). This responder threshold is clinically relevant because it has been used in regulatory evaluation of topical DED therapies.²⁰ Trial sequential analysis supported the stability of the Schirmer finding by crossing the monitoring boundary. However, because the accrued information size remained below the required information size, the result should be interpreted as supportive but not definitive; additional trials would still increase certainty.

The rapid onset of the Schirmer effect is clinically important. The TRANQUILITY and TRANQUILITY 2 trials assessed tear production after short-term dosing in a controlled dry eye chamber setting,^9,10 and the Phase II study NCT04971031 reported a numerically greater same-day Schirmer response with reproxalap than with vehicle after a single administration.²¹ This early pharmacodynamic signal contrasts with the slower onset generally expected with topical cyclosporine-based treatment, where clinically meaningful improvement in tear production is usually evaluated over weeks to months rather than hours or days.^16,17 In a disease in which early treatment discomfort and delayed perceived benefit may reduce adherence, a rapid objective sign response may be clinically useful, provided that symptom outcomes and tolerability are also acceptable.

No randomized trial has directly compared reproxalap with cyclosporine A or lifitegrast, so any comparison of effect sizes across these agents in this discussion is an indirect, unadjusted, between-trial comparison rather than a head-to-head comparison, and it should be read accordingly. In the COSMO Phase III trial of cyclosporine A 0.05% gel, cyclosporine improved several DED measures, but the evidence base differs in design, follow-up, formulation, and endpoints from the reproxalap trials.²² A network meta-analysis of cyclosporine A formulations also showed that efficacy estimates vary across preparations and outcomes.²³ For lifitegrast, the OPUS-2 and OPUS-3 trials consistently supported symptom improvement, whereas objective sign effects were less consistent across endpoints.^18,19 Because these comparisons are indirect, the present findings should be limited to describing a favorable signal for tear production with reproxalap; they cannot and do not establish that reproxalap is more effective, better tolerated, or otherwise more favorable than cyclosporine A or lifitegrast. Direct, adequately powered active-comparator trials are needed before any comparative positioning can be determined.

Symptom outcomes were not pooled quantitatively because the contributing trials used different patient-reported instruments, symptom domains, and assessment windows, and forcing heterogeneous instruments into a common standardized mean difference risked manufacturing false precision rather than resolving it. This does not mean the symptom data were uninformative. Across every source that reported patient-reported outcomes, the direction of effect favored reproxalap. Clark et al⁵ found that 0.25% reproxalap produced statistically significant improvement over vehicle in OD4SQ ocular dryness score by mixed-model analysis (P = 0.047) and in the proportion of patients reporting a dryness score of zero, a clinically defined threshold of complete symptom resolution (P = 0.012); improvement in combined symptom scores was detectable as early as the first post-baseline visit among patients with higher baseline symptom burden.⁵ Garg et al⁶ reported that every patient-reported chamber symptom domain assessed, including burning, discomfort, dryness, grittiness, and stinging, was statistically significantly lower with reproxalap than vehicle, with itching specifically improved after chamber exposure (P = 0.004).⁶ Registry-reported data from RENEW were less consistent: numerically greater reduction in subject-reported ocular dryness VAS score favored reproxalap in the QID-to-BID dosing arm, but in the QID arm the vehicle group showed a numerically larger reduction than reproxalap; this registry-only estimate could not be independently verified against a peer-reviewed source and the discordance between dosing arms should be reported rather than smoothed over.⁸ Taken together, the narrative evidence from the two peer-reviewed sources indicates that patients receiving reproxalap consistently reported feeling less symptomatic across dryness, burning, discomfort, grittiness, stinging, and itching domains, in both an acute chamber challenge and a 12-week field setting, using entirely different questionnaires; the unpublished RENEW registry data are directionally supportive for one dosing arm but not the other and should temper rather than reinforce this conclusion. The consistency of direction across the peer-reviewed instruments strengthens confidence in a genuine symptomatic benefit, even though the magnitude cannot legitimately be pooled into a single effect size and the registry-only symptom signal is inconsistent. What this evidence base cannot yet establish is durability of symptom benefit beyond 12 weeks, a minimally important difference for reproxalap on any specific instrument, or how symptom benefit relates to the objective Schirmer signal in patients who do not meet the 10-mm responder threshold.

Ocular redness showed a materially different pattern than Schirmer test and symptoms, and the discrepancy deserves explicit explanation rather than a general appeal to heterogeneity. The pooled estimate did not show a statistically significant difference between reproxalap and vehicle (MD, −0.01; 95% CI, −0.07 to 0.05; P = 0.52), with substantial heterogeneity (I² = 87.5%) (Supplementary Figure S2). The study by Garg et al⁶ was a small, pivotal, within-patient crossover trial in which reproxalap significantly reduced investigator-assessed ocular redness under standardized chamber conditions. TRANQUILITY, in contrast, was a larger parallel-group Phase 2/3 registration trial that did not meet its prespecified primary endpoint of investigator-graded conjunctival redness; only the secondary Schirmer test endpoint reached significance in that trial.^9,24 Several design factors plausibly explain this divergence rather than a true absence of anti-redness activity. First, unit of analysis: the crossover design in Garg et al compares each patient against themselves, removing between-patient variability that dilutes a parallel-group estimate such as TRANQUILITY’s. Second, measurement method: a post-hoc analysis of TRANQUILITY using computer-automated grading of digital photography reportedly showed a statistically significant reduction in redness with reproxalap, even though the prespecified investigator 0–4 scale analysis did not;²⁵ because the present meta-analysis pooled the investigator-graded scale to align with Garg et al’s reporting, the pooled estimate inherits TRANQUILITY’s null primary-scale result rather than its positive automated-grading result. Third, exposure and population differences between a controlled 90-minute chamber challenge and a larger parallel-group trial population increase both clinical and statistical heterogeneity. Together, these design and measurement differences, rather than a true absence of anti-redness activity, most plausibly account for the null, heterogeneous pooled redness estimate, and they indicate that ocular redness in this evidence base is a measurement-method- and design-sensitive endpoint rather than a stable, freely poolable one.

The main tolerability finding was instillation-site irritation. Reproxalap increased the risk of instillation-site irritation compared with vehicle (RR, 27.27; 95% CI, 12.20 to 60.99; P < 0.0001), and this result remained consistent in sensitivity analyses. The relative effect is large, but its clinical interpretation depends on severity and duration. In the pivotal chamber trial, instillation-site irritation was common, but reported episodes were generally mild and short-lived.⁶ This pattern is relevant because local discomfort is also seen with established DED therapies: lifitegrast trials reported instillation-site irritation and discomfort as common adverse events,^26,27 and cyclosporine 0.05% labeling identifies ocular burning as the most common adverse reaction.²⁸ In the present meta-analysis, all-cause discontinuation did not differ between groups (RR, 0.99; 95% CI, 0.50 to 1.99; P = 0.9995), although adverse event-related discontinuation was higher with reproxalap (RR, 4.88; 95% CI, 2.36 to 10.10). Clinically, this supports a balanced message: patients should be warned about transient irritation, but the available data do not show reduced overall retention.

The analysis of any TEAE was less certain. The pooled estimate was numerically higher with reproxalap but did not reach statistical significance (RR, 2.02; 95% CI, 0.92 to 4.44; P = 0.082), with substantial heterogeneity. This heterogeneity was largely influenced by differences between early-phase chamber studies and longer Phase III safety assessment, including differences in monitoring intensity, follow-up duration, and adverse event capture. Importantly, the 12-month Phase III safety trial reported no treatment-related serious ocular TEAEs and found no clinically meaningful differences in intraocular pressure, corneal endothelial cell density, or fundoscopic findings between groups.⁷ The post-hoc best-corrected visual acuity finding at 12 months was statistically significant in favor of reproxalap, but it should be considered hypothesis-generating and requires prospective confirmation before any claim of visual-function benefit can be made.⁷

Several limitations should be considered. First, the number of contributing studies was small for most outcomes, which limited subgroup analysis, meta-regression, and formal exploration of heterogeneity. Second, follow-up duration varied widely, from acute chamber assessments to 12-month safety follow-up, limiting conclusions about long-term efficacy maintenance. Third, patient-reported outcomes could not be pooled reliably because instruments and reporting formats differed across trials. This is important because DED presentations frequently exhibit discordance between signs and symptoms, and clinical management decisions may be guided by patient-perceived symptom burden, objective clinical signs, or both. Fourth, no active-comparator trials were available, so comparisons with cyclosporine A or lifitegrast remain indirect and should not be used to infer superiority. Fifth, both sources contributing to the pooled Schirmer responder analysis, TRANQUILITY and TRANQUILITY 2, reported this specific outcome using eye-level rather than participant-level denominators (306 and 362/360 eyes, respectively, vs fewer underlying participants). Eyes from the same patient are not statistically independent observations: they share systemic tear physiology and treatment exposure, and in other ophthalmic research contexts one eye of a patient may receive an intervention, such as an injection, that the fellow eye does not, further correlating within-patient outcomes. Treating eyes as independent units, as the underlying registry records did, can artificially inflate the effective sample size and narrow confidence intervals beyond what true participant-level precision would support; this outcome should therefore be interpreted as hypothesis-supporting rather than as precise as its reported confidence interval implies. Sixth, three of the six included sources, RENEW, TRANQUILITY, and TRANQUILITY 2, are ClinicalTrials.gov results records that have not undergone independent peer review, and RENEW has been described elsewhere as not yet published in the peer-reviewed literature. These three registry-only sources account for half of the included evidence base and are the sole contributors to both pooled Schirmer efficacy analyses. Peer-reviewed publication remains the methodological gold standard for evidence synthesis because it subjects study conduct, analysis, and reporting to independent external scrutiny that self-reported registry data do not receive; registry entries can also be revised over time, and the primary and secondary endpoint hierarchy for this reproxalap clinical program was in fact modified during its development. Reliance on unpublished registry data for a substantial portion of this review, including the entirety of its pooled efficacy synthesis, is therefore a material constraint on the certainty of the efficacy conclusions, and it reflects a current gap in the peer-reviewed literature that this review cannot resolve on its own. Seventh, the certainty of evidence was low to very low for several outcomes because of imprecision, heterogeneity, and reliance on registry data for some trials. Despite these limitations, the review has important strengths, including prospective registration, PRISMA-based conduct, duplicate risk-of-bias assessment using RoB 2, GRADE assessment by outcome, trial sequential analysis, and sensitivity analyses for key safety endpoints.

These findings have practical and research implications. Reproxalap appears to provide a reproducible short-term improvement in tear production, which may distinguish it from therapies whose objective sign response is slower or less consistent. However, the high frequency of transient instillation-site irritation means that patient counseling will be essential if the drug is used clinically. Future trials should directly compare reproxalap with cyclosporine A and lifitegrast, evaluate longer-term durability of Schirmer improvement, and standardize symptom endpoints so that objective signs can be interpreted alongside patient-reported benefit. Trials stratified by DED subtype, including aqueous-deficient, evaporative, and mixed phenotypes, may also help identify patients most likely to benefit from RASP inhibition. Finally, future studies should use consistent definitions and severity grading for instillation-site irritation to improve safety comparisons across DED therapies.

Conclusions

Reproxalap 0.25% ophthalmic solution demonstrated a rapid objective improvement in tear production compared with vehicle. However, the available evidence does not yet support a definitive routine clinical recommendation because frequent post-instillation ocular discomfort, registry-only efficacy data, trial program endpoint restructuring (notably the shift away from conjunctival redness in the TRANQUILITY program), eye-level responder denominators, and small study numbers limit certainty. No treatment-related serious ocular safety concerns were identified in the included evidence. Larger, fully reported, peer-reviewed, active-comparator trials with longer follow-up and standardized symptom outcomes are needed to clarify durability of benefit and define the place of reproxalap in the DED treatment algorithm.

Funding Statement

No external funding was received for this study.

Protocol and Registration

PROSPERO registration: CRD420261425460.

Abbreviations

BCVA, best-corrected visual acuity; BID, twice daily; CENTRAL, Cochrane Central Register of Controlled Trials; CI, confidence interval; DED, dry eye disease; GRADE, Grading of Recommendations Assessment, Development and Evaluation; MD, mean difference; NCT, National Clinical Trial identifier; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; PROSPERO, International Prospective Register of Systematic Reviews; QID, four times daily; RASP, reactive aldehyde species; RCT, randomized controlled trial; RoB 2, revised Cochrane risk-of-bias tool for randomized trials; RR, risk ratio; SD, standard deviation; SE, standard error; TEAE, treatment-emergent adverse event; TSA, trial sequential analysis.

Data Sharing Statement

All data used in this review were extracted from published articles and publicly available ClinicalTrials.gov results records. Extracted summary data are reported in the manuscript, figures, tables, and supplementary material.

Ethics Approval and Consent to Participate

Not applicable. This study synthesized published articles and publicly available trial-registry results records and did not involve human participants or patient-level data requiring ethics approval.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors declare no competing interests.