ENRICHED trial of preservative-free lubricants in dry eye disease Authors: Andres Bustamante-Arias, Jorge Eduardo Satizabal-García, Dayana Pino-Murcia, Carlos Cifuentes-González Published: 2026 ABSTRACT We evaluated the effectiveness, adherence, and safety of preservative-free (PF) lubricants in adults with dry eye disease (DED) under real-world clinical conditions. The ENRICHED-PF trial (NCT06162442) was a single-arm, prospective interventional study conducted in Cali, Colombia. A total of 120 adults with mild-to-moderate DED initiated PF propylene glycol 0.3% / polyethylene glycol 0.4% drops for six months. Longitudinal analyses were performed using linear mixed-effects models, with multiple imputation applied under a Missing at Random assumption to handle missing data. At baseline, the median Ocular Surface Disease Index (OSDI) score was 31.2 [19.2–47.4]. Longitudinal model-based estimates showed an OSDI estimated marginal mean (EMM) of 33.9 at baseline, 23.1 at 3 months, and 16.8 at 6 months in the mITT population, corresponding to an estimated change of − 17.1 points at month six (p < 0.001), with 65.4% of the modified intention-to-treat (mITT) population and 58% of the per-protocol (PP) population achieving a clinically meaningful response (≥ 10-point improvement). Tear break-up time increased and corneal staining scores decreased (both p < 0.001), while Schirmer I values remained stable. Additionally, NEI VFQ-25 composite scores improved by 8.9±3.1 points (p<0.001). These findings suggest clinically meaningful improvements in symptoms, tear film stability, and ocular surface integrity over time, which were associated with treatment using PF lubricants. However, given the single-arm design, the absence of a comparator group, and loss to follow-up during the study period, the magnitude of these effects should be interpreted with caution and confirmed in controlled comparative studies.Clinical Trial Registration: ClinicalTrials.gov Identifier: NCT06162442.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-50779-6. Introduction Dry eye disease (DED) is one of the most prevalent ocular surface disorders and a leading cause of ophthalmic consultations globally. It is characterized by tear film instability, hyperosmolarity, and ocular surface inflammation, resulting in fluctuating vision, discomfort, and decreased quality of life1,2. Despite advancements in diagnostic technologies and standardized definitions, clinical management remains challenging due to the imperfect correlation between symptoms and signs, which complicates treatment assessment in routine practice3. The TFOS DEWS III Executive Summary reaffirmed this diagnostic gap and emphasized the necessity of integrated approaches combining patient-reported outcomes and objective clinical measures to guide therapy4. Symptom assessment tools such as the Ocular Surface Disease Index (OSDI), Dry Eye Questionnaire-5 (DEQ-5), and the Symptom Assessment iN Dry Eye (SANDE) questionnaire are widely used in clinical and research settings to quantify symptom burden. These instruments are complemented by objective measures, including the Schirmer I test, tear break-up time (TBUT), and corneal and conjunctival staining, which evaluate tear film function and ocular surface integrity5. According to the TFOS DEWS III Management and Therapy Report, DED should be classified into aqueous-deficient, evaporative, or mixed subtypes, each requiring tailored management based on the predominant tear film abnormality6. This stratified approach supports the design of clinical studies assessing therapeutic efficacy across pathophysiologically distinct disease phenotypes. Lubricants remain the cornerstone of DED management. However, chronic use of preserved formulations—particularly those containing benzalkonium chloride (BAK)—has been associated with ocular surface intolerance and decreased adherence7,8. To address these limitations, preservative-free (PF) lubricants have been developed to enhance tolerability and preserve ocular surface homeostasis over long-term use. The TFOS DEWS III Management and Therapy Report recommends PF formulations as a preferred first-line option for patients with chronic lubrication needs or clinical signs of ocular surface damage or inflammation6. Although PF lubricants are widely prescribed, prospective evidence integrating patient-reported outcomes, clinical signs, and vision-related quality of life remains limited. In addition, real-world longitudinal data on adherence and safety are scarce. The ENRICHED-PF trial (Effectiveness aNd PROMs of pReservative-free lubrIcants in Clinical Hydration of Eye Disease) was designed to address these evidence gaps by prospectively evaluating the effectiveness, adherence, and safety of PF lubricants in adults with mild-to-moderate DED. By integrating validated patient-reported questionnaires and objective clinical parameters over six months, this study aims to generate comprehensive and practice-oriented evidence to guide individualized DED management in routine care. Results A total of 120 participants were enrolled in the ENRICHED-PF trial. Of these, 118 initiated treatment (safety set), 67 contributed to the modified intention-to-treat (mITT) population (≥ 1 post-baseline assessment), and 53 comprised the per-protocol (PP) population (completing at least 80% of planned visits through month six without major protocol deviations). This reduction in sample size across analytic populations reflects substantial loss to follow-up over the study period. To explore potential attrition bias, baseline demographic and clinical characteristics were compared between participants who completed the 6 month visit and those who did not complete follow-up (Supplementary Material 3, Table S1). No clear baseline imbalances were observed between groups. The mean age was 35.2 ± 7.8 years in the mITT group and 36.2 ± 7.4 years in the PP group, with women representing 68.7% and 69.8%, respectively. Median OSDI™ and DEQ-5™ scores indicated a moderate level of symptomatic burden, and the combination of preserved Schirmer I values (median 13.0 mm) with short TBUT (median 5.0 s) was consistent with a predominantly evaporative phenotype. According to the DEWS II classification, 77.6% of participants were categorized as evaporative and 17.9% as mixed, whereas no subject met criteria for aqueous-deficient disease, reflecting a lipid-layer–deficient profile typical of younger adults (Table 1, Fig. 1).Table 1Baseline demographic and clinical characteristics of the study population.Variable mITT (n = 67)PP (n = 53)Age (years), mean ± SD35.2 ± 7.836.2 ± 7.4Female, n (%)46 (68.7%)37 (69.8%)OSDI™ (baseline), median [IQR]31.2 [19.2–47.4]31.2 [16.7–42.2]DEQ-5™ (baseline), median [IQR]12.0 [8.0–15.0]11.5 [7.8–15.0]SANDE – Frequency, median [IQR]58.0 [23.0–86.0]49.5 [16.0–84.5]SANDE – Severity, median [IQR]53.0 [21.0–76.0]45.5 [17.5–76.2]Schirmer I (mm, worst eye), median [IQR]13.0 [7.0–17.8]13.0 [7.0–16.5]TBUT (s, worst eye), median [IQR]5.0 [5.0–7.0]5.0 [5.0–7.0]SICCA (worst eye), median [IQR]2.0 [1.0–3.0]2.0 [1.0–3.2]DEWS classificationAqueous-deficient, n (%)0 (0.0%)0 (0.0%) Evaporative, n (%)52 (77.6%)43 (81.1%) Mixed, n (%)12 (17.9%)8 (15.1%) Discordant, n (%)3 (4.5%)2 (3.8%)Values are expressed as mean ± SD or median [IQR] according to distribution (Shapiro–Wilk test). Between-group comparisons were assessed with Student’s t or Mann–Whitney U tests. Abbreviations: mITT = modified intention-to-treat; PP = per-protocol; DEWS = Dry Eye Workshop II; TBUT = tear break-up time; OSDI = Ocular Surface Disease Index; DEQ-5 = Dry Eye Questionnaire-5; SANDE = Symptom Assessment in Dry Eye. Continuous variables are presented as median [interquartile range] unless otherwise specified. Ocular surface parameters (Schirmer I, TBUT, and SICCA) are reported for the worst eye at baseline.Fig. 1CONSORT flow diagram showing participant enrollment, inclusion in mITT and PP populations, and study completion. CONSORT flow diagram showing participant enrollment, inclusion in mITT and PP populations, and study completion across the 6-month follow-up. Percentages are calculated relative to the safety sett population (n = 118). Table 1Baseline demographic and clinical characteristics of the study population.Variable mITT (n = 67)PP (n = 53)Age (years), mean ± SD35.2 ± 7.836.2 ± 7.4Female, n (%)46 (68.7%)37 (69.8%)OSDI™ (baseline), median [IQR]31.2 [19.2–47.4]31.2 [16.7–42.2]DEQ-5™ (baseline), median [IQR]12.0 [8.0–15.0]11.5 [7.8–15.0]SANDE – Frequency, median [IQR]58.0 [23.0–86.0]49.5 [16.0–84.5]SANDE – Severity, median [IQR]53.0 [21.0–76.0]45.5 [17.5–76.2]Schirmer I (mm, worst eye), median [IQR]13.0 [7.0–17.8]13.0 [7.0–16.5]TBUT (s, worst eye), median [IQR]5.0 [5.0–7.0]5.0 [5.0–7.0]SICCA (worst eye), median [IQR]2.0 [1.0–3.0]2.0 [1.0–3.2]DEWS classificationAqueous-deficient, n (%)0 (0.0%)0 (0.0%) Evaporative, n (%)52 (77.6%)43 (81.1%) Mixed, n (%)12 (17.9%)8 (15.1%) Discordant, n (%)3 (4.5%)2 (3.8%)Values are expressed as mean ± SD or median [IQR] according to distribution (Shapiro–Wilk test). Between-group comparisons were assessed with Student’s t or Mann–Whitney U tests. Abbreviations: mITT = modified intention-to-treat; PP = per-protocol; DEWS = Dry Eye Workshop II; TBUT = tear break-up time; OSDI = Ocular Surface Disease Index; DEQ-5 = Dry Eye Questionnaire-5; SANDE = Symptom Assessment in Dry Eye. Continuous variables are presented as median [interquartile range] unless otherwise specified. Ocular surface parameters (Schirmer I, TBUT, and SICCA) are reported for the worst eye at baseline. Baseline demographic and clinical characteristics of the study population. Values are expressed as mean ± SD or median [IQR] according to distribution (Shapiro–Wilk test). Between-group comparisons were assessed with Student’s t or Mann–Whitney U tests. Abbreviations: mITT = modified intention-to-treat; PP = per-protocol; DEWS = Dry Eye Workshop II; TBUT = tear break-up time; OSDI = Ocular Surface Disease Index; DEQ-5 = Dry Eye Questionnaire-5; SANDE = Symptom Assessment in Dry Eye. Continuous variables are presented as median [interquartile range] unless otherwise specified. Ocular surface parameters (Schirmer I, TBUT, and SICCA) are reported for the worst eye at baseline. CONSORT flow diagram showing participant enrollment, inclusion in mITT and PP populations, and study completion. CONSORT flow diagram showing participant enrollment, inclusion in mITT and PP populations, and study completion across the 6-month follow-up. Percentages are calculated relative to the safety sett population (n = 118). Primary endpoint The primary endpoint was the change in OSDI™ score from baseline to six months. A progressive and statistically significant improvement was observed in both analytic populations. In the mITT set, OSDI™ decreased over time, with model-based estimated marginal means (EMMs) indicating a reduction from 3.3 (95% CI 2.9–3.8) at baseline to 2.3 (95% CI 1.9–2.6) at month 6, and in the PP set from 33.5 (95% CI 28.2–38.8) at baseline to 16.7 (95% CI 13.0–20.5) at month 6 (p < 0.001). A clinically meaningful response, defined as a reduction ≥ 10 points, was achieved in 65.4% of the mITT and 58% of the PP participants. Improvements were evident by month 3 and remained present at month 6 (Table 2, Fig. 2).Table 2Patient-reported outcomes (OSDI™, DEQ-5™, SANDE) at baseline, 3 months, and 6 months.QuestionnaireBaseline3 Months6 Monthsp-valueResponders (%)OSDI™ (mITT)33.9 (29.2–38.6)23.1 (17.9–28.3)16.8 (13.0–20.5) < 0.00165.4OSDI™ (PP)33.5 (28.2–38.8)18.8 (14.5–23.2)16.7 (13.0–20.5) < 0.00158DEQ-5™ (mITT)11.6 (10.4–12.7)8.4 (7.4–9.4)6.9 (5.9–8.0) < 0.00157DEQ-5™ (PP)11.4 (10.2–12.6)7.6 (6.6–8.7)6.2 (5.2–7.2) < 0.00160SANDE (mITT)51.0 (43.5–58.6)32.8 (25.9–39.7)17.9 (12.8–23.1) < 0.00160SANDE (PP)50.3 (42.7–57.9)29.4 (23.0–35.8)15.6 (11.2–20.0) < 0.00165Values represent estimated marginal means (EMMs) with 95% confidence intervals derived from linear mixed-effects models. EMMs estimated marginal means, OSDI™ ocular surface disease index, DEQ-5™ 5-item dry eye questionnaire, SANDE symptom assessment in dry eye, mITT modified Intent-to-Treat, PP per-protocol. p-values were obtained from linear mixed-effects models or Friedman tests for repeated measures, as appropriate. Clinical response definitions followed endpoint-specific criteria: a ≥ 10-point improvement for OSDI™9, a ≥ 6-point reduction for DEQ-5™, and a ≥ 20% reduction from baseline for SANDE scores, as described in the Methods and Supplementary Material 2. Model-fit diagnostics and residual analyses are presented in Supplementary Material 3, confirming model stability, convergence, and appropriate variance structure.Fig. 2Change in OSDI™ score over time, showing EMM trajectories with 95% confidence intervals at baseline, 3 months, and 6 months, and the proportion of participants reaching the minimal clinically important difference (MCID ≥ 10 points). Longitudinal changes in symptom scores and responder rates. (A) Estimated marginal means (EMMs ± 95% CIs) of OSDI™ scores at baseline, 3 months, and 6 months in the mITT (red) and PP (blue) populations, showing sustained symptom improvement over time. (B) Proportion of participants achieving a clinically meaningful response (≥ 10-point OSDI™ improvement). (C,D) Estimated marginal means trends for SANDE and DEQ-5™ scores. (E) Distribution of responders across symptom-based and objective endpoints. Additional details are provided in Table 2 and Supplementary Material 3. Table 2Patient-reported outcomes (OSDI™, DEQ-5™, SANDE) at baseline, 3 months, and 6 months.QuestionnaireBaseline3 Months6 Monthsp-valueResponders (%)OSDI™ (mITT)33.9 (29.2–38.6)23.1 (17.9–28.3)16.8 (13.0–20.5) < 0.00165.4OSDI™ (PP)33.5 (28.2–38.8)18.8 (14.5–23.2)16.7 (13.0–20.5) < 0.00158DEQ-5™ (mITT)11.6 (10.4–12.7)8.4 (7.4–9.4)6.9 (5.9–8.0) < 0.00157DEQ-5™ (PP)11.4 (10.2–12.6)7.6 (6.6–8.7)6.2 (5.2–7.2) < 0.00160SANDE (mITT)51.0 (43.5–58.6)32.8 (25.9–39.7)17.9 (12.8–23.1) < 0.00160SANDE (PP)50.3 (42.7–57.9)29.4 (23.0–35.8)15.6 (11.2–20.0) < 0.00165Values represent estimated marginal means (EMMs) with 95% confidence intervals derived from linear mixed-effects models. EMMs estimated marginal means, OSDI™ ocular surface disease index, DEQ-5™ 5-item dry eye questionnaire, SANDE symptom assessment in dry eye, mITT modified Intent-to-Treat, PP per-protocol. p-values were obtained from linear mixed-effects models or Friedman tests for repeated measures, as appropriate. Clinical response definitions followed endpoint-specific criteria: a ≥ 10-point improvement for OSDI™9, a ≥ 6-point reduction for DEQ-5™, and a ≥ 20% reduction from baseline for SANDE scores, as described in the Methods and Supplementary Material 2. Model-fit diagnostics and residual analyses are presented in Supplementary Material 3, confirming model stability, convergence, and appropriate variance structure. Patient-reported outcomes (OSDI™, DEQ-5™, SANDE) at baseline, 3 months, and 6 months. Values represent estimated marginal means (EMMs) with 95% confidence intervals derived from linear mixed-effects models. EMMs estimated marginal means, OSDI™ ocular surface disease index, DEQ-5™ 5-item dry eye questionnaire, SANDE symptom assessment in dry eye, mITT modified Intent-to-Treat, PP per-protocol. p-values were obtained from linear mixed-effects models or Friedman tests for repeated measures, as appropriate. Clinical response definitions followed endpoint-specific criteria: a ≥ 10-point improvement for OSDI™9, a ≥ 6-point reduction for DEQ-5™, and a ≥ 20% reduction from baseline for SANDE scores, as described in the Methods and Supplementary Material 2. Model-fit diagnostics and residual analyses are presented in Supplementary Material 3, confirming model stability, convergence, and appropriate variance structure. Change in OSDI™ score over time, showing EMM trajectories with 95% confidence intervals at baseline, 3 months, and 6 months, and the proportion of participants reaching the minimal clinically important difference (MCID ≥ 10 points). Longitudinal changes in symptom scores and responder rates. (A) Estimated marginal means (EMMs ± 95% CIs) of OSDI™ scores at baseline, 3 months, and 6 months in the mITT (red) and PP (blue) populations, showing sustained symptom improvement over time. (B) Proportion of participants achieving a clinically meaningful response (≥ 10-point OSDI™ improvement). (C,D) Estimated marginal means trends for SANDE and DEQ-5™ scores. (E) Distribution of responders across symptom-based and objective endpoints. Additional details are provided in Table 2 and Supplementary Material 3. Parallel trends were observed for the complementary symptom scales DEQ-5™ and SANDE, which also showed significant reductions over time (all p < 0.001). DEQ-5™ scores decreased over time, with model-based EMMs indicating a progressive reduction (see Table 2 for 95% confidence intervals). SANDE scores showed a similar pattern, with EMM-based analyses indicating consistent improvement across visits (Table 2, Fig. 2). Secondary outcomes Objective ocular surface parameters also changed during follow-up. TBUT increased significantly over time, with model-based EMMs indicating an increase from 5.9 (95% CI 5.4–6.3) at baseline to 9.1 (95% CI 8.4–9.7) at month 6 in the mITT population, with similar improvements observed in the PP analysis (5.8 [95% CI 5.3–6.2] to 9.1 [95% CI 8.5–9.7]; p < 0.001). Schirmer I values remained stable over time (13.7 to 14.0 mm; ns). SICCA scores improved, with EMM-based analyses showing a reduction from 3.3 (95% CI 2.9–3.8) at baseline to 2.3 (95% CI 1.9–2.6) at month 6 in the mITT population, with consistent findings in the PP set (3.4 [95% CI 2.9–3.9] to 2.2 [95% CI 1.9–2.6]; p < 0.01) (Table 3, Figs. 2 and 3).Table 3Objective ocular-surface parameters across study visits.MeasureBaseline3 Months6 Monthsp-valueResponders (%)Schirmer I (mm, mITT)13.7 (11.7–15.7)14.0 (12.1–15.9)14.0 (12.0–16.0)0.41–Schirmer I (mm, PP)13.5 (11.5–15.5)14.3 (12.3–16.3)14.5 (12.5–16.6)0.38–TBUT (s, mITT)5.8 (5.4–6.3)7.1 (6.7–7.5)9.1 (8.5–9.7) < 0.00133TBUT (s, PP)5.9 (5.4–6.4)7.4 (7.0–7.9)9.5 (8.8–10.1) < 0.00138SICCA Composite (mITT)3.3 (2.9–3.8)2.7 (2.3–3.0)2.3 (1.9–2.6) < 0.0169SICCA Composite (PP)3.2 (2.8–3.7)2.5 (2.2–2.9)2.2 (1.8–2.6) < 0.0173EMMs estimated marginal means; Schirmer I (mm) Schirmer test I (mm wetting); TBUT (s) tear break-up time (seconds); SICCA composite composite score of dry eye signs; mITT modified Intent-to-Treat; PP per-protocol. Values represent EMMs with 95% confidence intervals derived from linear mixed-effects models. Continuous variables were analyzed using linear mixed-effects models with visit as a fixed factor and participant as a random intercept; sensitivity analyses were confirmed with Friedman non-parametric tests. Clinical responder definitions followed endpoint-specific criteria: ≥ 2 mm increase from baseline for Schirmer I, ≥ 3 s increase for tear break-up time (TBUT), and ≥ 1 grade reduction for the SICCA ocular staining score, as described in the Methods and Supplementary Material 2. Extended diagnostics for these models are available in Supplementary Material 3 (Supplementary Table S2 and Supplementary Fig. S2), including residual distributions and Bland–Altman agreement plots confirming model adequacy.Fig. 3Longitudinal evolution of objective ocular surface parameters in the mITT and PP populations from baseline to 6months.br (A) Estimated marginal means (EMMs ± 95% CIs) of SICCA composite scores over time.br (B) Estimatedmarginal means (EMMs ± 95% CIs) of tear break-up time (TBUT) over time.br (C) Estimated marginal means (EMMs ±95% CIs) of Schirmer I test values over time.br /These results demonstrate improvements in tear film stability and ocularsurface staining, while tear production remained stable. Additional details are provided in Table 3 and SupplementaryMaterial 3 Table 3Objective ocular-surface parameters across study visits.MeasureBaseline3 Months6 Monthsp-valueResponders (%)Schirmer I (mm, mITT)13.7 (11.7–15.7)14.0 (12.1–15.9)14.0 (12.0–16.0)0.41–Schirmer I (mm, PP)13.5 (11.5–15.5)14.3 (12.3–16.3)14.5 (12.5–16.6)0.38–TBUT (s, mITT)5.8 (5.4–6.3)7.1 (6.7–7.5)9.1 (8.5–9.7) < 0.00133TBUT (s, PP)5.9 (5.4–6.4)7.4 (7.0–7.9)9.5 (8.8–10.1) < 0.00138SICCA Composite (mITT)3.3 (2.9–3.8)2.7 (2.3–3.0)2.3 (1.9–2.6) < 0.0169SICCA Composite (PP)3.2 (2.8–3.7)2.5 (2.2–2.9)2.2 (1.8–2.6) < 0.0173EMMs estimated marginal means; Schirmer I (mm) Schirmer test I (mm wetting); TBUT (s) tear break-up time (seconds); SICCA composite composite score of dry eye signs; mITT modified Intent-to-Treat; PP per-protocol. Values represent EMMs with 95% confidence intervals derived from linear mixed-effects models. Continuous variables were analyzed using linear mixed-effects models with visit as a fixed factor and participant as a random intercept; sensitivity analyses were confirmed with Friedman non-parametric tests. Clinical responder definitions followed endpoint-specific criteria: ≥ 2 mm increase from baseline for Schirmer I, ≥ 3 s increase for tear break-up time (TBUT), and ≥ 1 grade reduction for the SICCA ocular staining score, as described in the Methods and Supplementary Material 2. Extended diagnostics for these models are available in Supplementary Material 3 (Supplementary Table S2 and Supplementary Fig. S2), including residual distributions and Bland–Altman agreement plots confirming model adequacy. Objective ocular-surface parameters across study visits. EMMs estimated marginal means; Schirmer I (mm) Schirmer test I (mm wetting); TBUT (s) tear break-up time (seconds); SICCA composite composite score of dry eye signs; mITT modified Intent-to-Treat; PP per-protocol. Values represent EMMs with 95% confidence intervals derived from linear mixed-effects models. Continuous variables were analyzed using linear mixed-effects models with visit as a fixed factor and participant as a random intercept; sensitivity analyses were confirmed with Friedman non-parametric tests. Clinical responder definitions followed endpoint-specific criteria: ≥ 2 mm increase from baseline for Schirmer I, ≥ 3 s increase for tear break-up time (TBUT), and ≥ 1 grade reduction for the SICCA ocular staining score, as described in the Methods and Supplementary Material 2. Extended diagnostics for these models are available in Supplementary Material 3 (Supplementary Table S2 and Supplementary Fig. S2), including residual distributions and Bland–Altman agreement plots confirming model adequacy. Longitudinal evolution of objective ocular surface parameters in the mITT and PP populations from baseline to 6months.br (A) Estimated marginal means (EMMs ± 95% CIs) of SICCA composite scores over time.br (B) Estimatedmarginal means (EMMs ± 95% CIs) of tear break-up time (TBUT) over time.br (C) Estimated marginal means (EMMs ±95% CIs) of Schirmer I test values over time.br /These results demonstrate improvements in tear film stability and ocularsurface staining, while tear production remained stable. Additional details are provided in Table 3 and SupplementaryMaterial 3 Quality-of-life outcomes Vision-related quality of life, assessed by the NEI VFQ-25 composite score, improved significantly in both analytic populations after PF lubricant therapy. In the mITT analysis, VFQ-25 increased from 82.1 (80.1–84.1) at baseline to 89.6 (87.4–91.9) at month 6 (p < 0.001), while in the PP population it rose from 82.4 (80.2–84.5) to 90.2 (88.0–92.4) (p < 0.001). A positive change (Δ > 0) was achieved by 54% and 62% of participants in the mITT and PP sets, respectively (Fig. 4A,B).Fig. 4Vision-Related Quality-of-Life and Correlation Analyses (mITT vs PP Populations). (A) Proportion of participants with improvement in NEI VFQ-25 scores (Δ > 0). (B) EMMs (±95% CIs) showinglongitudinal improvement in VFQ-25 scores.br (C–D) Associations between changes in VFQ-25 and OSDI™ or SICCA scores.Correlation coefficients (Spearman ρ) summarize these relationships (See Supplementary Material 3). Vision-Related Quality-of-Life and Correlation Analyses (mITT vs PP Populations). (A) Proportion of participants with improvement in NEI VFQ-25 scores (Δ > 0). (B) EMMs (±95% CIs) showinglongitudinal improvement in VFQ-25 scores.br (C–D) Associations between changes in VFQ-25 and OSDI™ or SICCA scores.Correlation coefficients (Spearman ρ) summarize these relationships (See Supplementary Material 3). Correlation analyses showed that improvements in vision-related quality of life paralleled symptomatic and clinical gains. A positive correlation trend was observed between ΔVFQ-25 and ΔOSDI (ρ = 0.27), although this association did not reach statistical significance (p = 0.056). In contrast, ΔVFQ-25 was significantly associated with ΔSICCA (ρ = 0.28, p = 0.045). Correlations with ΔDEQ-5 (ρ = 0.16, p = 0.25), ΔSANDE (ρ = 0.22, p = 0.15), ΔSchirmer (ρ = –0.21, p = 0.14), and ΔTBUT (ρ = –0.05, p = 0.73) were not significant (Fig. 4C,D). Adherence and safety outcomes Adherence estimates varied across analytic datasets. Mean adherence was 65.7 ± 31.1% in the raw dataset, 63.9 ± 28.4% in the truncated dataset, and 85.1 ± 13.8% in the cleaned dataset, which excluded implausible or unverified entries. Adherence was defined as the proportion of prescribed doses effectively administered during the 6 month treatment period, verified through patient diaries and vial returns. In the cleaned adherence dataset, 61.3% of participants achieved adherence ≥ 80%, although these estimates should be interpreted in light of data cleaning and the reduced sample size. No statistically significant associations were detected between adherence level and clinical improvement across the evaluated datasets. Spearman’s correlation coefficients between adherence and change in OSDI™ (ρ = 0.10–0.11, p > 0.40) or TBUT (ρ = 0.11–0.21, p > 0.25) were not statistically significant across datasets. Response rates were also similar between participants with adherence < 80% and ≥ 80%, although these comparisons should be interpreted cautiously given the reduced size of the cleaned adherence dataset (Table 4).Table 4Adherence metrics across analytic datasets .DefinitionnMean % (SD)ρ ΔOSDI (p)ρ ΔTBUT (p)Responder rates (< 80% vs ≥ 80%)Raw6865.7 (31.1)0.10 (0.47)0.11 (0.43)50.0% vs 52.2%Truncated6863.9 (28.4)0.11 (0.44)0.11 (0.44)50.0% vs 52.2%Clean3885.1 (13.8)0.01 (0.98)0.21 (0.27)50.0% vs 52.2%Adherence was quantified as the proportion of prescribed doses effectively administered during the 6 months treatment period. The raw dataset included all adherence entries as reported; the truncated dataset censored unverified or incomplete entries; and the clean dataset excluded implausible or inconsistent values after data validation. Spearman’s ρ coefficients show non-significant associations between adherence and changes in OSDI™ or TBUT. Responder rates were comparable between participants with adherence < 80% and ≥ 80%; these findings should be interpreted cautiously given the variability, missingness, and reduced sample size of the validated adherence dataset. Table 4Adherence metrics across analytic datasets .DefinitionnMean % (SD)ρ ΔOSDI (p)ρ ΔTBUT (p)Responder rates (< 80% vs ≥ 80%)Raw6865.7 (31.1)0.10 (0.47)0.11 (0.43)50.0% vs 52.2%Truncated6863.9 (28.4)0.11 (0.44)0.11 (0.44)50.0% vs 52.2%Clean3885.1 (13.8)0.01 (0.98)0.21 (0.27)50.0% vs 52.2%Adherence was quantified as the proportion of prescribed doses effectively administered during the 6 months treatment period. The raw dataset included all adherence entries as reported; the truncated dataset censored unverified or incomplete entries; and the clean dataset excluded implausible or inconsistent values after data validation. Spearman’s ρ coefficients show non-significant associations between adherence and changes in OSDI™ or TBUT. Responder rates were comparable between participants with adherence < 80% and ≥ 80%; these findings should be interpreted cautiously given the variability, missingness, and reduced sample size of the validated adherence dataset. Adherence metrics across analytic datasets . Adherence was quantified as the proportion of prescribed doses effectively administered during the 6 months treatment period. The raw dataset included all adherence entries as reported; the truncated dataset censored unverified or incomplete entries; and the clean dataset excluded implausible or inconsistent values after data validation. Spearman’s ρ coefficients show non-significant associations between adherence and changes in OSDI™ or TBUT. Responder rates were comparable between participants with adherence < 80% and ≥ 80%; these findings should be interpreted cautiously given the variability, missingness, and reduced sample size of the validated adherence dataset. Treatment persistence was further analyzed using Kaplan–Meier survival estimates, representing time to treatment discontinuation among participants with available follow-up data. The survival probability remained above 0.90 throughout the 6 month period, indicating high persistence among those retained in follow-up (Fig. 5A). This analysis should be interpreted separately from overall study attrition, which reflects loss to follow-up for reasons unrelated to treatment discontinuation.Fig. 5Treatment persistence and safety outcomes. (A) Kaplan–Meier curve showing time to treatment discontinuation over 6 months. This analysis describes time totreatment discontinuation among participants with recorded follow-up and should be interpreted separately from overallstudy attrition, which includes loss to follow-up unrelated to treatment persistence. (B) Distribution of adverse events (AEs)by severity. Treatment persistence and safety outcomes. (A) Kaplan–Meier curve showing time to treatment discontinuation over 6 months. This analysis describes time totreatment discontinuation among participants with recorded follow-up and should be interpreted separately from overallstudy attrition, which includes loss to follow-up unrelated to treatment persistence. (B) Distribution of adverse events (AEs)by severity. No vision-threatening adverse events occurred during the study . A total of 22 participants (18.5%) reported at least one adverse event (AE), 4 (3.4%) experienced a serious AE, and 11 (8.9%) had medication-related events. Ten (8.1%) participants discontinued the study, including five (4.1%) due to mild tolerability or adherence-related issues. All AEs were transient and self-limited, most commonly mild ocular burning, irritation, or blurred vision after instillation. The AE distribution confirmed that most events were non-serious (Fig. 5B). Discussion This prospective interventional study observed clinically meaningful and sustained improvements over time in both symptoms and objective signs of dry DED. OSDI, DEQ-5, and SANDE scores decreased significantly over time, accompanied by increases in TBUT and reductions in SICCA. VFQ-25 scores also improved during follow-up. No vision-threatening adverse events were observed.. These findings suggest that treatment with PF lubricants was associated with improvements over time in symptoms and ocular surface parameters in patients previously exposed to preserved formulations. The magnitude of OSDI improvement observed in the ENRICHED-PF trial aligns with the anchor-based MCID of 7–10 points established by Miller et al.9, suggesting that the reductions observed here reached thresholds generally considered clinically meaningful. Our responder definition (≥ 10 points) also parallels the severity-stratified MCIDs (4.5–7.3 for mild/moderate and 7.3–13.4 for severe DED) derived from the RESTORE registry9. Comparable symptomatic and objective improvements have been reported in controlled studies using other PF formulations. Aragona et al. observed similar OSDI reductions and TBUT gains with carboxymethylcellulose–hyaluronic acid eye drops10, whereas Konstas et al. showed that switching glaucoma patients from preserved to PF therapy restored tear-film stability and reduced ocular staining11. In Latin America, Belalcázar-Rey et al. demonstrated significant OSDI and TBUT improvements with sodium-hyaluronate/chondroitin-sulfate PF drops12, consistent with our findings and suggesting that polymer type is less critical than preservative withdrawal. Earlier systematic reviews such as Rezende Ribeiro et al. found heterogeneous evidence and short follow-up durations, precluding firm conclusions on PF superiority7. The JAMA Ophthalmology umbrella review by McCann et al. emphasized the need for longitudinal studies integrating patient-reported and clinical outcomes13. Our results, collected under TFOS DEWS III diagnostic and management criteria4, provide prospective real-world data from a Latin American cohort that may inform the role of PF lubricants in routine DED management. The parallel improvement in TBUT and ocular-surface staining without a corresponding increase in Schirmer values suggests that the observed changes were more evident in tear-film stability and ocular surface staining than in aqueous tear production. However, mechanistic interpretation remains limited because no inflammatory biomarkers, tear osmolarity, goblet cell density, or MMP-9 measurements were included in this study. The concordance between the observed OSDI change and the validated MCID framework9 suggests that the symptomatic improvements were both statistically significant and likely to be clinically meaningful. The absence of a statistically significant association between adherence level and outcome magnitude should be interpreted cautiously. Adherence measurements showed substantial variability, missingness, and implausible entries, and the cleaned adherence dataset was reduced to 38 participants. Therefore, these analyses were likely underpowered to detect modest adherence–outcome relationships, and no firm conclusions should be drawn regarding the absence of an association between adherence and clinical outcomes. Accordingly, the adherence analyses should be considered exploratory. Major strengths of this study include its prospective design, standardized application of validated PROMs, adherence to TFOS DEWS III criteria4, and use of linear mixed-effects and multiple-imputation models to handle missing data. The adoption of a responder threshold grounded in validated OSDI MCID methodology9 further enhances interpretability and clinical relevance. Loss to follow-up over the six-month period was substantial, with 53 of 118 treated participants completing the final study visit. To evaluate the potential for attrition bias, baseline characteristics were compared between participants who completed the 6-month visit and those who did not complete follow-up (Supplementary Material 3, Table S1). No relevant imbalances were observed across demographic variables, baseline symptom severity, ocular surface parameters, or DEWS phenotypes. Nevertheless, the risk of attrition bias cannot be excluded, particularly if unmeasured factors associated with symptom evolution, treatment experience, or follow-up adherence influenced study completion. In addition, the absence of a control or comparator group limits causal inference, as part of the observed improvement may reflect non-specific effects such as regression to the mean, placebo response, or natural symptom variability, all of which are well documented inDED studies. Because the study was funded by the manufacturer of the evaluated formulation, transparency regarding the role of the sponsor is important. The sponsor had no role in the study design, statistical analysis, interpretation of the results, or manuscript preparation, and all analyses were conducted independently by the investigators. Importantly, treatment persistence estimated using Kaplan–Meier methods should not be interpreted as equivalent to overall study retention, as attrition may occur independently of treatment discontinuation. The study was not specifically powered to detect associations between adherence and clinical outcomes, and therefore such analyses should be interpreted as exploratory. Therefore, the findings should be interpreted as prospective observational evidence describing clinical evolution after transition to preservative-free lubricants rather than as definitive evidence of comparative treatment efficacy. These findings should also be interpreted in the absence of direct mechanistic biomarkers. Nevertheless, the high consistency of findings across analytic populations and alignment with multicenter regional trials strengthen the external validity of the observed patterns10,12. In addition, the study population was relatively young and predominantly composed of patients with evaporative DED, with no cases fulfilling criteria for aqueous-deficient disease. Consequently, the findings may not be directly generalizable to older populations, postmenopausal women, or patients with severe aqueous-deficient DED. Over six months, patients showed consistent improvement in symptoms, meaningful gains in vision-related quality of life, and improvements in tear stability and ocular surface staining, while tear production remained unchanged. These effects were observed in both the mITT and PP populations and were concordant across patient-reported outcomes and objective clinical measures. Adherence estimated from bottle weight suggests substantial variability, with frequent missing or implausible values; although no statistically significant association with clinical change was detected after adjustment for baseline severity, these exploratory analyses were limited by the reduced size of the validated adherence dataset. Safety outcomes were acceptable, with few serious adverse events and a limited number of discontinuations attributable to adverse events or adherence issues. Overall, these findings suggest that PF lubricant therapy may be associated with improvements in symptoms, ocular surface parameters, and vision-related quality of life in routine clinical practice. However, given the single-arm design and the loss to follow-up observed in this study, controlled comparative trials will be important to confirm the magnitude and generalizability of these findings. Methods This study, the ENRICHED-PF trial, was a single-arm, prospective interventional trial conducted at SIGMA Clinic in Cali, Colombia, and registered at ClinicalTrials.gov (NCT06162442; registered on 29 November 2023) ( https://clinicaltrials.gov/study/NCT06162442). The protocol was reviewed and approved by the institutional ethics committee (approval CEI-001071222), and all procedures adhered to the principles of the Declaration of Helsinki. Written informed consent was obtained from all participants before enrollment. Reporting followed the CONSORT extension for non-randomized trials and the TREND statement to ensure transparency in design and analysis14,15 (see Supplementary Material 1). The study was designed to prospectively evaluate clinical evolution after switching from preserved lubricants to a preservative-free formulation under routine practice conditions rather than to establish comparative treatment efficacy. Study population Eligible participants were adults aged ≥ 18 years with a clinical diagnosis of mild-to-moderate dry eye disease (DED) according to the TFOS DEWS II criteria5. All had previously been treated with lubricants containing preservatives. Exclusion criteria comprised severe DED, prior ocular surface surgery, active ocular allergy or infection, systemic therapies known to interfere with tear-film stability, pregnancy, lactation, and hypersensitivity to fluorescein, lissamine green, or tetracaine. Patients with comorbid conditions such as glaucoma, cicatrizing conjunctivitis, or lacrimal duct obstruction were also excluded. Baseline assessments included age, sex, occupation, systemic comorbidities, etiology of DED, disease severity, and meibography findings (see Supplementary Material 1). Intervention and follow-up Following a seven-day wash-out period without any preserved lubricants, subjects received Glicolub Ultra® (polyethylene glycol 0.4% and propylene glycol 0.3%), a preservative-free ophthalmic solution manufactured by Tecnoquímicas S.A., delivered in a Novelia® multi-dose container system (Nemera, La Verpillière, France). The prescribed regimen consisted of one drop instilled in each eye every six hours. Follow-up visits occurred at baseline (post-wash-out), three months, and six months. Participant disposition is summarized in Fig. 1, which shows the number of screened, enrolled, and analyzed participants. A total of 252 subjects were screened, 120 met eligibility criteria and were enrolled, and 118 initiated treatment after completing wash-out. Sixty-seven participants attended at least one post-baseline visit (mITT), and 53 completed the 6-month visit (PP). Reasons for discontinuation included transport (n = 25), healthcare-related (n = 22), loss to follow-up (n = 11), adverse effects (n = 5), treatment-related (n = 2), and other causes (n = 3) (see Supplementary Material 1). Outcomes and assessments At each visit, patients completed validated questionnaires to capture symptom severity and quality of life. These included the Ocular Surface Disease Index (OSDI™, primary endpoint)9, the Dry Eye Questionnaire-5 (DEQ-5™)16, the Symptom Assessment in Dry Eye (SANDE™)17, and the National Eye Institute Visual Function Questionnaire (NEI VFQ-25™)18. Clinical measures included tear break-up time (TBUT; seconds), Schirmer I test without anesthesia (mm/5 min), corneal and conjunctival staining graded according to the Sjögren’s International Collaborative Clinical Alliance Ocular Staining Score (SICCA) criteria19, and intraocular pressure measured with a Goldmann tonometer. For ocular surface parameters (Schirmer I, TBUT, and SICCA), analyses were based on the worst eye per visit, defined as the eye with lower tear production (Schirmer), shorter tear break-up time (TBUT), or higher staining score (SICCA). Dry eye disease phenotypes were classified according to TFOS DEWS II criteria5, while clinical interpretation and management were informed by updated TFOS DEWS III recommendations. Based on these parameters, patients were subclassified according to the predominant pathophysiologic phenotype of dry eye disease: aqueous-deficient (Schirmer ≤ 10 mm/5 min and TBUT > 10 s), evaporative/lipid-deficient (TBUT ≤ 10 s with normal tear production), mixed (both Schirmer ≤ 10 mm/5 min and TBUT ≤ 10 s). This categorization followed the conceptual framework proposed by the TFOS DEWS II diagnostic report2. A final category was defined as discordant (symptom–sign mismatch not fulfilling the above definitions).20–22. Adherence to treatment was assessed by weighing and counting returned vials at each visit. The expected consumption was based on eight drops per day (one drop every six hours in each eye), with a maximum use period of 28 days per vial, assuming that 1 mL = 29 drops and the solution has a density of 1 g/mL. Safety monitoring included documentation of all adverse events (AEs), classified as serious or non-serious, and reasons for treatment discontinuation (see Supplementary Material 2). Analysis populations Three populations were defined for analysis. The safety set included all patients who received at least one dose of the study medication. The modified intention-to-treat (mITT) set comprised those with at least one post-baseline assessment and served as the main analysis group. The per-protocol (PP) set was defined as patients completing at least 80% of planned visits through month six without major protocol violations and was used for confirmatory sensitivity analyses (see Supplementary Material 2). The primary endpoint was the change in OSDI score between baseline and six months. Secondary endpoints included changes in Schirmer I, TBUT, DEQ-5, SANDE, and SICCA staining, as well as quality-of-life measures derived from the NEI VFQ-25. Clinical responders were defined according to established thresholds: a reduction of at least 10 points in OSDI, a reduction of at least six points in DEQ-5, a reduction of at least 20% in SANDE, an increase of at least two millimeters in Schirmer I, an increase of at least three seconds in TBUT, and a reduction of at least one grade in SICCA staining (more details see Supplementary Material 2). Statistical analysis Continuous outcomes were analyzed primarily using linear mixed-effects models (LMMs), with visit as a categorical fixed effect and subject as a random intercept, allowing inclusion of all available observations under a Missing At Random (MAR) assumption. This assumption cannot be formally verified and should be interpreted with caution. Estimated marginal means (EMMs) with 95% confidence intervals were derived from fitted models to quantify longitudinal changes.23 As a sensitivity analysis, robust ordinary least squares (OLS) regression with heteroskedasticity-consistent HC3 standard errors was performed to confirm the robustness of the findings. Responder proportions were calculated based on predefined minimal clinically important differences and reported with exact binomial confidence intervals. Correlations between changes in quality-of-life scores and clinical measures were estimated with Spearman’s rho.24 Predictors of NEI VFQ-25 improvement were explored with multivariable linear regression. Adherence was evaluated as a continuous percentage of expected versus observed drop use, as well as categorically by defining high adherence as ≥ 80%. Associations between adherence and changes in OSDI or TBUT were examined with Spearman correlations and Wilcoxon rank-sum tests. Adverse events were summarized descriptively by severity and seriousness, and time-to-discontinuation was analyzed with Kaplan–Meier methods25. Missing data were handled under the missing-at-random assumption using multiple imputation. Sensitivity analyses included complete-case and per-protocol models. Multiplicity was addressed hierarchically, with the OSDI as the primary endpoint and secondary outcomes adjusted using the Holm–Bonferroni procedure26 (see Supplementary Material 2). All analyses were performed in Python 3.12 (pandas, statsmodels, scipy, and matplotlib libraries). TFOS DEWS II and III diagnostic and management framework concepts guided phenotype classification and interpretation of treatment response4–6. (Supplementary Material 2). The statistical analysis plan was developed by the investigators and is provided in Supplementary Material 2. Some sensitivity analyses, including multiple-imputation procedures and complementary analytic populations, were specified after observing the extent of loss to follow-up during the study period. The study sponsor had no role in the statistical analysis, interpretation of the results, or preparation of the manuscript. Supplementary Information Supplementary Information 1. Supplementary Information 2. Supplementary Information 3. Supplementary Information 1. Supplementary Information 2. Supplementary Information 3.