Importance
Cyclin-dependent kinase 4/6 (CDK4/6) inhibitors are standard therapy for hormone receptor–positive (HR+), HER2-negative breast cancer. By blocking the G1/S cell-cycle transition, these agents may impair renewal of the corneal epithelium. No controlled study has systematically evaluated corneal epithelial changes in patients receiving CDK4/6 inhibitors.
Objective
To determine whether CDK4/6 inhibitor–based therapy is associated with cornealepithelial alterations independent of aromatase inhibitor exposure and tear film dysfunction.
Design, Setting, and Participants
Retrospective comparative cross-sectional study at a tertiary ophthalmology center. A total of 132 women were enrolled: 45 receiving a CDK4/6 inhibitor plus an aromatase inhibitor (CDKAI group), 44 receiving aromatase inhibitor monotherapy (AI group), and 43 age-matched postmenopausal controls without systemic oncologic therapy.
Exposures
CDK4/6 inhibitor (ribociclib, palbociclib, or abemaciclib) combined with an aromatase inhibitor; aromatase inhibitor alone; or no systemic oncologic therapy.
Main Outcomes and Measures
Prevalence and severity of punctate epitheliopathy and vortex keratopathy, assessed by a masked ophthalmologist. Secondary outcomes included Schirmer I test, tear film break-up time, and Ocular Surface Disease Index (OSDI).
Results
PE was present in 44.4% of eyes in the CDKAI group vs 4.7% in the AI group and 2.3% in controls ( χ ² = 34.31; P <.001). All moderate (13.3%) and severe/complicated (8.9%) PE cases occurred exclusively in the CDKAI group. Vortex keratopathy was observed in 13.3% of CDKAI patients and in none of the other groups ( P =.025). Schirmer values, tear film break-up time, and OSDI scores did not differ among groups (all P >.05). Within the CDKAI group, PE was not associated with treatment duration ( P =.963) or tear film parameters.
Conclusions and Relevance
In this comparative study, CDK4/6 inhibitor–based therapy was associated with significantly higher prevalence and severity of PE and vortex keratopathy, independent of aromatase inhibitor exposure and in the absence of measurable tear film dysfunction. These findings suggest a direct cytostatic effect on the corneal epithelium. Symptom scores were low, although OSDI interpretation was limited by incomplete responses. Proactive corneal surface evaluation with fluorescein staining may be warranted during CDK4/6 inhibitor treatment.
INTRODUCTION
C yclin-dependent kinase 4 and 6 (CDK4/6) inhibitors—palbociclib, ribociclib, and abemaciclib—have become the backbone of first-line treatment for hormone receptor–positive (HR+), human epidermal growth factor receptor 2–negative (HER2−) advanced and metastatic breast cancer. ,, By selectively blocking the CDK4/6–cyclin D1 complex, these agents prevent phosphorylation of the retinoblastoma protein (Rb), thereby arresting the cell cycle at the G1/S transition and suppressing tumor cell proliferation. Combined with aromatase inhibitors or other endocrine therapies, CDK4/6 inhibitors have demonstrated significant improvements in progression-free and overall survival in the PALOMA, MONALEESA, and MONARCH trial, establishing their role as a standard-of-care regimen recommended by major clinical practice guidelines. ,,,,,
The pharmacologic rationale that makes CDK4/6 inhibitors effective against cancer—G1/S checkpoint blockade—also raises a predictable concern for nonmalignant tissues whose homeostasis depends on continuous cell division. The corneal epithelium is among the most rapidly renewing tissues in the human body, completing a full turnover cycle in approximately 7 to 10 days. , This process is driven by the orderly progression of limbal stem cell–derived transit-amplifying (TA) cells through the cell cycle; at each division, TA cells must traverse the CDK4/6-dependent G1/S checkpoint before entering S-phase. ,, Pharmacologic inhibition of CDK4/6 can therefore be expected to reduce the proliferative output of the limbal niche, slow centripetal epithelial migration, and compromise the replacement of terminally differentiated surface cells. , The downstream consequences would include delayed epithelial wound healing, increased susceptibility to punctate epitheliopathy (PE), and potentially disrupted migration patterns manifesting as vortex keratopathy.
Clinically, PE may range from asymptomatic fluorescein staining to foreign body sensation, burning, tearing, photophobia, fluctuating vision, recurrent epithelial erosion, and, in severe cases, persistent epithelial defect or infectious keratitis. Vortex keratopathy is often visually silent, but dense or centrally located epithelial deposits may occasionally be associated with glare, halos, blurred vision, or epithelial irregularity. Because PE is commonly interpreted within the dry eye spectrum, evaluating its relationship with established ocular surface parameters such as tear film break-up time (TBUT), Schirmer testing, and Ocular Surface Disease Index (OSDI) is important for distinguishing tear film–mediated disease from direct epithelial toxicity.
The systemic safety profile of CDK4/6 inhibitors—neutropenia, fatigue, hepatotoxicity, gastrointestinal disturbances—is well established. ,,,,, In contrast, their potential effects on the ocular surface have received comparatively little attention. Isolated case reports and small series have described vortex keratopathy and corneal epithelial deposits in patients treated with abemaciclib, palbociclib, and ribociclib. ,,,, Pharmacovigilance analyses of the FDA Adverse Event Reporting System and the European EudraVigilance database have generated safety signals for ocular disorders, including blurred vision and corneal changes, , although such analyses are constrained by under-reporting, the absence of denominator data, and variable case definitions. No controlled comparative study has systematically evaluated the corneal epithelial surface in CDK4/6 inhibitor–treated patients using standardized clinical grading.
A further interpretive challenge arises from the obligatory coadministration of aromatase inhibitors with CDK4/6 inhibitors. Estrogen receptors are expressed in both the corneal epithelium and the lacrimal gland, and estrogen deprivation has been linked to tear film instability, reduced aqueous production, and ocular surface dysfunction in postmenopausal women. ,, The vortex keratopathy pattern has also been reported with tamoxifen, another endocrine agent used in breast cancer, though through a distinct deposition mechanism. , Without a comparator group receiving aromatase inhibitor monotherapy, any corneal findings in patients on combination therapy cannot be attributed to CDK4/6 inhibition specifically. Disentangling the contributions of the two drug classes requires a study design that includes both combination-treated and aromatase inhibitor–only cohorts, as well as a drug-free control group.
We therefore designed the present study to address three questions. First, is CDK4/6 inhibitor plus aromatase inhibitor therapy associated with a higher prevalence and severity of PE and vortex keratopathy compared with aromatase inhibitor monotherapy or no systemic oncologic treatment? Second, are any such epithelial changes accompanied by measurable differences in tear film function or subjective symptoms, or do they represent a distinct, subclinical form of corneal surface toxicity? Third, within the CDK4/6 inhibitor–treated group, are conventional ocular surface parameters predictive of PE status? We hypothesized that CDK4/6 inhibitor exposure would be associated with corneal epithelial disruption independent of tear film abnormalities, a pattern more consistent with a direct cytostatic mechanism than with secondary dry eye disease.
METHODS
Study design and ethical approval
This retrospective comparative study was conducted at a tertiary ophthalmology center affiliated with a university hospital. The study protocol adhered to the tenets of the Declaration of Helsinki and was approved by the institutional scientific research ethics committee (approval number: 2025-066; decision no: 3/26; date: February 13, 2025). Written informed consent was obtained from all participants prior to enrollment.
Study population
A total of 132 female participants were enrolled and assigned to three groups on the basis of their current systemic treatment ( Table 1 ). The CDKAI group ( n = 45) comprised patients with histologically confirmed HR+/HER2− breast cancer who had been receiving a CDK4/6 inhibitor (palbociclib, ribociclib, or abemaciclib) in combination with an aromatase inhibitor (letrozole, anastrozole, or exemestane) for a minimum of 3 months. The AI group ( n = 44) included patients with the same tumor profile who had been receiving aromatase inhibitor monotherapy for at least 3 months. The control group ( n = 43) consisted of age-matched postmenopausal women with natural menopause, with no history of malignancy or systemic oncologic therapy, recruited from the general ophthalmology outpatient clinic. In this study, the terms “female” and “women” refer to sex as recorded in the participants’ medical records; gender identity was not assessed.
TABLE 1
Baseline Demographic and Clinical Characteristics of the Study Population.
| Characteristic | CDKAI ( n = 45) | AI ( n = 44) | Control ( n = 43) | P |
|---|---|---|---|---|
| Age (y), mean ± SD | 54.62 ± 10.13 | 59.73 ± 7.74 | 60.00 ± 6.89 | .973 |
| Age (y), median (range) | 54 (32-74) | 59.5 (45-75) | 60 (47-80) | |
| Menopausal status, n (%) | ||||
| Natural menopause | NR | NR | 43 (100%) | |
| Treatment-induced menopause | NR | NR | 0 (0%) | |
| Menopause duration (y), mean ± SD | NA b | NA b | 12.35 ± 7.97 | — |
| Menopause duration, median (range) | NA b | NA b | 10 (2-30) | |
| Treatment duration (mo), mean ± SD | 16.64 ± 14.31 | 30.95 ± 23.39 | — | .001 a |
| Treatment duration, median (range) | 12 (3-60) | 24 (4-87) | — | |
| CDK4/6 inhibitor type, n (%) | ||||
| Ribociclib | 35 (77.8%) | — | — | |
| Palbociclib | 8 (17.8%) | — | — | |
| Abemaciclib | 2 (4.4%) | — | — | |
| Aromatase inhibitor type, n (%) | ||||
| Letrozole | 35 (77.8%) | 35 (79.5%) | — | |
| Anastrozole | 9 (20.0%) | 9 (20.5%) | — | |
| Exemestane | 1 (2.2%) | 0 (0%) | — |
Age compared by Kruskal–Wallis test; treatment duration by Mann–Whitney U test.
NA = not applicable; NR = not reliably reported.
a P <.05.
b Not ascertainable: menopause duration in the CDKAI and AI groups could not be reliably determined because a proportion of patients experienced treatment-induced menopause.
Bold values indicate statistically significant results ( P <.05).
In the CDKAI and AI groups, precise duration of menopause could not be reliably ascertained because a proportion of patients had entered menopause as a consequence of their oncologic treatment (chemotherapy-induced or treatment-related menopause) rather than through natural processes. Therefore, menopause duration was recorded and reported only for the control group, all of whom had undergone natural menopause.
Exclusion criteria were as follows: ocular surgery within the preceding 6 months; active ocular infection or inflammation; use of topical ophthalmic medications other than preservative-free artificial tears; prior diagnosis of ocular surface disease requiring chronic treatment; corneal dystrophy, ectasia, or scarring; contact lens wear; concurrent use of other systemic chemotherapeutic agents or targeted therapies with documented ocular toxicity; and inability or unwillingness to provide informed consent.
Ophthalmologic examination
All participants underwent a standardized bilateral ophthalmologic examination performed by a single experienced ophthalmologist who was masked to participants’ group assignment. The ophthalmologist was unaware of treatment status at the time of examination, and all assessments were performed using standardized techniques and automated devices where applicable. Fluorescein staining photographs were graded at a later date without access to clinical or treatment data. The examination included best-corrected visual acuity, slit-lamp biomicroscopy, and the following assessments.
PE. PE was evaluated by instilling fluorescein sodium dye and examining the corneal surface under cobalt blue light illumination. Erosion severity was graded on a four-tier ordinal scale: absent (grade 0, no staining); mild (grade 1, scattered punctate staining involving less than one-third of the corneal surface); moderate (grade 2, punctate staining involving one-third to two-thirds of the corneal surface); or severe/complicated (grade 3, confluent staining, frank epithelial defects, or filamentary changes involving more than two-thirds of the corneal surface).
For fluorescein staining, a sterile fluorescein sodium strip was moistened with a single drop of sterile saline and gently applied to the inferior conjunctival fornix. Participants were asked to blink several times, and the corneal surface was examined under cobalt blue illumination approximately 1 to 2 minutes after instillation. Excess pooling was avoided, and staining was graded before Schirmer testing to minimize test-related ocular surface alteration.
Vortex keratopathy. Defined as a whorl-shaped (cornea verticillata) pattern of corneal epithelial or subepithelial deposits identified on retroillumination and direct slit-lamp examination. ,
Tear film assessment. Aqueous tear production was measured using the Schirmer I test (without topical anesthesia; standardized filter-paper strip placed in the lower conjunctival fornix for 5 minutes; wetting length recorded in millimeters). Tear film stability was assessed by TBUT, defined as the interval in seconds between the last complete blink and the first appearance of a dry spot on the fluorescein-stained tear film; three consecutive measurements were obtained and the mean was recorded.
Anterior segment biometry. Central corneal thickness (CCT) was obtained by optical biometry. Anterior chamber depth (ACD) and axial length (AL) were obtained by optical biometry. Corneal endothelial cell density (ECD) was determined by noncontact specular microscopy.
Subjective symptom assessment. Ocular surface symptoms were quantified using the validated 12-item OSDI (DEWS III) questionnaire (score range 0-100; higher scores indicate greater symptom burden). The OSDI was administered to all participants; however, due to logistical constraints inherent to the oncologic treatment setting and incomplete responses, fully completed questionnaires were available for 81 of 132 participants (61.4%). The completion rate was lower in the CDKAI group (16 of 45; 35.6%), possibly reflecting the higher treatment burden or fatigue associated with active combination oncologic therapy.
Data quality and outlier handling
All data were reviewed for plausibility prior to analysis. Values exceeding ±3 SDs from the group mean for continuous biometric variables were flagged for verification against source records. Any excluded values are noted in the relevant table footnotes.
Statistical analysis
Analyses were performed using IBM SPSS Statistics (IBM Corp.). Distribution normality was tested with the Shapiro–Wilk test. Continuous variables are reported as mean ± SD and median (minimum-maximum); categorical variables as frequency (percentage). Three-group comparisons of categorical variables used Pearson’s chi-square test or, where expected cell counts were insufficient, the Fisher–Freeman–Halton test. Continuous variables meeting the normality assumption were compared by one-way ANOVA; otherwise, the Kruskal–Wallis H test was used. Posthoc pairwise comparisons employed the independent-samples t test or Mann–Whitney U test as appropriate. Within the CDKAI group, a prespecified subgroup analysis compared all measured parameters between participants with and without PE in the right eye.
The primary outcome was the between-group difference in PE prevalence, for which no multiplicity adjustment was required. All secondary analyses—including vortex keratopathy, tear film parameters, biometric measurements, and the CDKAI subgroup analysis—should be considered exploratory; a Bonferroni-corrected threshold was applied to secondary comparisons where applicable. Findings with uncorrected P values between.01 and.05 should be interpreted with caution and are flagged in the tables. Subgroup analysis by individual CDK4/6 inhibitor was not performed due to the small sample sizes for palbociclib ( n = 8) and abemaciclib ( n = 2); all CDK4/6 inhibitors were therefore analyzed collectively as a single exposure group. All tests were two-sided, and P <.05 was considered nominally significant.
RESULTS
Baseline characteristics
A total of 132 participants were enrolled: 45 in the CDKAI group, 44 in the AI group, and 43 in the control group. Baseline demographic and clinical characteristics are summarized in Table 1 . The three groups were well matched for age (Kruskal–Wallis H = 0.00, P =.973), with median ages of 54 years (range 32-74), 59.5 years (range 45-75), and 60 years (range 47-80), respectively. Within the CDKAI group, the majority of patients received ribociclib (35/45, 77.8%), followed by palbociclib (8/45, 17.8%) and abemaciclib (2/45, 4.4%). The distribution of aromatase inhibitors was similar between the CDKAI and AI groups, with letrozole being the most frequently used agent in both (77.8% and 79.5%, respectively). All participants in the control group had undergone natural menopause, with a mean duration of 12.35 ± 7.97 years (median 10; range 2-30). In the CDKAI and AI groups, menopause duration could not be reliably determined due to the variable contribution of treatment-induced menopause. The median treatment duration was significantly shorter in the CDKAI group (12 months; range 3-60) compared with the AI group (24 months; range 4-87; Mann–Whitney U = 587.5, P =.001), reflecting the more recent introduction of CDK4/6 inhibitors into clinical practice.
Corneal epithelial findings
PE prevalence differed markedly among groups for both the right and left eyes ( χ ² = 34.31, P <.001 for each; Table 2 ). In the CDKAI group, PE was observed in 20 of 45 participants (44.4%), all of whom exhibited bilateral involvement. In the AI group, PE was present in 2 of 43 evaluable eyes (4.7%), and in the control group, in 1 of 43 eyes (2.3%). Posthoc pairwise comparisons confirmed that the CDKAI group had a significantly higher prevalence than both the AI and control groups, which did not differ from each other.
TABLE 2
Comparison of Categorical and Continuous Ocular Variables Among the CDKAI, AI, and Control Groups.
| Variable | n | CDKAI ( n = 45) | AI ( n = 44) | Control ( n = 43) | ||||
|---|---|---|---|---|---|---|---|---|
| %/Mean ± SD | n | %/Mean ± SD | n | %/Mean ± SD | Test | P | ||
| Vortex: absent | 39 | 86.7% | 43 | 100% | 43 | 100% | 6.60 | .025 a,b |
| Present | 6 | 13.3% | 0 | 0% | 0 | 0% | ||
| PE RE: absent | 25 | 55.6% | 41 | 95.3% | 42 | 97.7% | 34.31 | <.001 a |
| Present | 20 | 44.4% | 2 | 4.7% | 1 | 2.3% | ||
| Mild | 10 | 22.2% | 2 | 4.7% | 1 | 2.3% | ||
| Moderate | 6 | 13.3% | 0 | 0% | 0 | 0% | ||
| Severe/complicated | 4 | 8.9% | 0 | 0% | 0 | 0% | ||
| PE LE: absent | 25 | 55.6% | 41 | 95.3% | 42 | 97.7% | 34.31 | <.001 a |
| Present | 20 | 44.4% | 2 | 4.7% | 1 | 2.3% | ||
| Schirmer RE (mm) | 45 | 14.16 ± 6.85 | 44 | 12.02 ± 7.40 | 43 | 9.84 ± 5.83 | 1.74 | .187 |
| Schirmer LE (mm) | 45 | 14.18 ± 6.76 | 44 | 12.20 ± 7.93 | 43 | 10.60 ± 5.73 | 0.25 | .618 |
| TBUT RE (s) | 44 | 5.00 ± 3.07 | 44 | 5.86 ± 3.46 | 43 | 6.72 ± 4.58 | 0.75 | .386 |
| TBUT LE (s) | 44 | 4.98 ± 3.22 | 44 | 5.82 ± 3.67 | 43 | 7.14 ± 4.83 | 2.34 | .126 |
| CCT RE (µm) | 45 | 528.1 ± 35.8 | 44 | 539.2 ± 32.9 | 40 | 533.3 ± 34.4 | 1.16 | .316 |
| CCT LE (µm) | 45 | 529.8 ± 36.8 | 44 | 539.6 ± 34.3 | 40 | 533.3 ± 34.5 | 0.87 | .420 |
| ACD RE (mm) c | 43 | 3.22 ± 0.38 | 41 | 3.19 ± 0.47 | 39 | 3.13 ± 0.39 | 0.44 | .644 |
| ACD LE (mm) | 43 | 3.27 ± 0.36 | 41 | 3.14 ± 0.40 | 40 | 3.11 ± 0.37 | 0.08 | .777 |
| OSDI d | 16 | 5.13 ± 6.82 | 25 | 3.64 ± 3.23 | 40 | 5.30 ± 6.09 | 1.32 | .251 |
| ECD RE (cells/mm²) | 43 | 2537 ± 244 | 44 | 2468 ± 262 | 40 | 2466 ± 269 | 0.03 | .872 |
| ECD LE (cells/mm²) | 43 | 2518 ± 256 | 44 | 2467 ± 277 | 40 | 2497 ± 216 | 0.03 | .858 |
| AL RE (mm) | 45 | 23 ± 1 | 43 | 23 ± 1 | 39 | 23 ± 1 | 1.19 | .275 |
| AL LE (mm) | 45 | 23.2 ± 1.03 | 43 | 23.2 ± 0.83 | 39 | 23.05 ± 0.67 | 1.26 | .261 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
