Association Between Attention Deficit Hyperactivity Disorder and Keratoconus

Abstract

Objective:

To investigate the relationship between attention deficit hyperactivity disorder (ADHD) and keratoconus (KC) in a large, diverse adult cohort.

Design:

Retrospective cross-sectional study.

Participants:

Individuals with and without ADHD from the National Institutes of Health All of Us Research Program.

Methods:

Individuals with ADHD were propensity score-matched (1:3) to controls. Univariable and multivariable conditional logistic regression models were employed to evaluate the association between ADHD and KC, with adjustments for psychiatric comorbidities, atopic conditions, obstructive sleep apnea, ADHD medications, conjunctivitis, and tear film-related disorders. An interaction term analysis was performed to assess potential effect modification by sex.

Results:

A total of 40,360 matched participants were included, comprising 10,090 individuals with ADHD and 30,270 without ADHD (mean age: 42.3 ± 15.6 years; mean BMI: 30.5 ± 8.4 kg/m²; 63.1% female). The prevalence of KC was significantly higher among individuals with ADHD (33 out of 10,090, 0.33%) compared to matched controls without ADHD (42 out of 30,270, 0.14%), yielding an unadjusted odds ratio of 2.36 (95% CI=[1.49, 3.72], p<0.01). Participant sex did not significantly modify this association (p=0.38). After multivariable adjustment, the association between ADHD and KC remained significant (OR=2.43, 95% CI=[1.28, 4.61], p=0.01), with tear film-related disorders emerging as an independent predictor of KC (OR=5.37, 95% CI=[2.62–11.01], p<0.01).

Conclusions:

In this large, diverse adult cohort, individuals with ADHD were more than twice as likely to have KC, independent of multiple confounders. Tear film-related disorders, which may potentially represent a proxy for eye rubbing, also emerged as a significant predictor.

Introduction

Keratoconus (KC) is a bilateral, asymmetric ectatic disorder characterized by progressive thinning and steepening of the cornea, often resulting in irregular astigmatism and reduced visual acuity.¹ Although prevalence estimates are highly variable, the global pooled prevalence of KC is approximately 1.38 per 1,000 individuals.² Established risk factors for the development of KC include a positive family history, as well as personal histories of habitual eye rubbing, atopy, and sleep apnea.^2–4 Management strategies range from the use of spectacles or specialty contact lenses for visual rehabilitation to corneal collagen cross-linking, implantation of intracorneal ring segments, and, in advanced cases, corneal transplantation.⁵

Attention deficit hyperactivity disorder (ADHD) is a common, clinically heterogeneous neurodevelopmental condition characterized by incongruent levels of inattention, hyperactivity, and impulsivity, affecting an estimated 6.8% of adults worldwide.⁶ Some individuals with ADHD may exhibit repetitive bodily movements;⁷ it has been postulated that when these symptoms include habitual eye rubbing, they may serve as a behavioural mechanism contributing to KC.⁸ Given the substantial quality-of-life burden associated with KC,⁹ it is crucial to investigate this potential association in populations with ADHD, who may be at particular risk.⁸ A recent cross-sectional study in Israel found a novel association between ADHD and KC using nationwide screening data from Israeli adolescents and young adults, the vast majority of whom were aged 25 years or younger.⁸ Nonetheless, a significant proportion of patients develop KC beyond this age, and this association warrants further investigation in more diverse cohorts worldwide. To address these gaps, the present study leverages a large, diverse cohort from the All of Us Research Program in the United States to examine whether a relationship exists between ADHD and KC in adults.

Methods

Data Source

This cross-sectional study analyzed data from the Registered Tier Dataset (v8) provided by the National Institutes of Health (NIH) All of Us Research Program. The All of Us data repository integrates demographic information, survey responses, and electronic health record (EHR) data from a diverse cohort across the United States, with intentional recruitment of individuals from populations historically underrepresented in biomedical research. Additionally, the All of Us dataset integrates EHR data from various healthcare providers for the same participant, thereby enhancing data comprehensiveness. Upholding standards of data privacy and confidentiality, participant counts fewer than 20 were not reported, in compliance with the All of Us cell suppression policy. All participant data were fully de-identified, and this research was conducted in accordance with the tenets of the Declaration of Helsinki and the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines. The NIH All of Us Institutional Review Board approved the program’s enrollment, data collection, and deidentification processes, and review by the University of Toronto Research Ethics Board was not required due to our secondary use of anonymized data.

Participant Selection

This analysis included adult participants with available EHR data. Participants’ ADHD status was ascertained using the standardized Observational Medical Outcomes Partnership (OMOP) concept identifier 438409, which encompasses the term ADHD along with a comprehensive set of synonymous clinical terms. Individuals with Down syndrome or Ehlers-Danlos syndrome were excluded due to an elevated baseline risk for KC.^10,11 Similarly, although the evidence is less definitive, individuals with rare connective tissue disorders that may predispose to corneal structural abnormalities like KC, including Marfan syndrome and osteogenesis imperfecta, were not considered.^12,13

Propensity Score Matching

A 1:3 nearest-neighbour propensity score matching procedure without replacement was conducted to match each eligible All of Us participant with a diagnosis of ADHD to three unique individuals in the dataset without an ADHD diagnosis. Propensity scores were estimated via logistic regression, incorporating the following covariates: age (calculated from the participant’s date of birth to the date they completed the All of Us survey on basic personal information), body mass index, sex at birth (i.e., male or female), self-reported race/ethnicity (i.e., Asian, Black or African American, Hispanic, White, or Other), health insurance status, highest level of educational attainment (i.e., high school or less, some college, or college graduate or advanced degree), annual household income in USD (i.e., <$25k, $25k–$50k, $50k–$100k, or ≥$100k), and whether the participant had smoked ≥100 cigarettes in their lifetime. Matching was constrained by a caliper width equal to 0.2 times the standard deviation (SD) of the logit of the propensity score to ensure close alignment of covariate distributions between the ADHD and non-ADHD groups. Participants with missing data for any of the specified covariates, whether due to refusal, skipped responses, or indeterminate answers, were excluded from the matching process. The balance between groups after matching was assessed using standardized mean differences (SMDs), where an absolute SMD value of <0.1 was indicative of negligible differences.

Outcomes & Statistical Analysis

All statistical analyses were conducted on July 31^st, 2025 within the All of Us Researcher Workbench using Python 3 in a secure Jupyter Cloud Environment. Demographic characteristics of the matched groups were summarized as means with SDs for continuous variables, and as counts with proportions for categorical variables. The outcome of interest was the occurrence of KC, defined by diagnostic codes mapped to the OMOP concept identifier 374362. This code captures diagnoses of KC within a participant’s EHR data, including unilateral and bilateral cases, as well as stable and unstable cases. Participants were classified as having KC if their EHRs contained at least one KC diagnosis code.

To preserve the integrity of the propensity score matching process and ensure that comparisons between individuals with and without ADHD were made within their matched strata, a univariable conditional logistic regression model was used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for the association between ADHD and KC. An interaction term analysis between ADHD status and sex was conducted to assess potential effect modification, given that Safir and colleagues reported an association between ADHD and KC in males, but not females.⁸ To evaluate the robustness of univariable findings, a multivariable conditional logistic regression model was performed, adjusting psychiatric comorbidities that frequently co-occur with adult ADHD (i.e., a composite variable comprising depressive disorders, anxiety disorders, bipolar disorders, substance use disorders, personality disorders, and tic disorders),^14,15 the atopic triad (i.e., a composite variable comprising asthma, allergic rhinitis, and eczema),¹⁶ obstructive sleep apnea,³ ADHD medications approved by the Food and Drug Administration (i.e., a composite variable comprising amphetamine, atomoxetine, clonidine, dexmethylphenidate, dextroamphetamine, guanfacine, lisdexamfetamine, methamphetamine, methylphenidate, serdexmethylphenidate, and viloxazine),¹⁷ conjunctivitis,¹⁸ and tear film-related disorders (i.e., a composite variable comprising tear film insufficiency, dry eye, blepharitis, and meibomian gland dysfunction).¹⁹ Each composite variable was coded as binary, indicating the presence of at least one of the component diagnoses versus none. A two-tailed P-value of <0.05 was considered statistically significant.

A post hoc sensitivity analysis was performed on May 2^nd, 2026, requiring participants to have at least two diagnostic codes mapped to OMOP concept identifier 374362 to be classified as having KC, thereby increasing confidence in the diagnosis based on EHR-derived codes. This analysis additionally matched participants based on whether they had seen an eye care professional, such as an ophthalmologist or optometrist, within the past year and, if so, the annual frequency with which they had visited an eye care professional (i.e., once, two to three times, or more than three times within the year). This self-reported matching variable was included to enhance comparability in opportunities for KC diagnosis between participants with and without ADHD, and was considered particularly relevant since prior studies have reported associations between ADHD and various ocular comorbidities, including refractive error, strabismus, and amblyopia, which might increase contact with eye care professionals.^20–22

Results

Participant Characteristics

A total of 13,467 individuals with ADHD were identified in the All of Us Research Program, of whom 10,090 (74.9%) had complete covariate data and were included in the analysis. These individuals with ADHD were matched to 30,270 unique individuals without ADHD, yielding an overall sample size of 40,360 participants (Figure 1).

Figure 1.

Flowchart summarizing All of Us participant selection process

The mean age of participants at the time of the All of Us survey completion was 42.3 ± 15.6 years (interquartile range=29 to 54 years), of whom 25,471 (63.1%) were females and 14,889 (36.9%) were males. The mean BMI of participants was 30.5 ± 8.4 kg/m². Self-reported race and ethnicity were distributed as follows: 29,764 (73.8%) identified as White, 2,492 (6.2%) as Hispanic, 2,406 (6.0%) as Black or African American, 698 (1.7%) as Asian, and 5,000 (12.4%) as another group. Most participants (39,327, 97.4%) had health insurance, although 1,033 (2.6%) did not. Regarding highest educational attainment, 20,667 (51.2%) had completed college or an advanced degree, 12,401 (30.7%) had some college education, and 7,292 (18.1%) had a high school education or less. Average annual household income was <$25k for 11,990 (29.7%) participants, $25k–$50k for 8,035 (19.9%) participants, $50k–$100k for 10,120 (25.1%) participants, and ≥$100k for 10,215 (25.3%) participants. Regarding smoking status, 18,248 (45.2%) reported having smoked at least 100 cigarettes in their lifetime, while 22,112 (54.8%) had not. A comprehensive overview of participant characteristics stratified by ADHD status is provided in Table 1, where all characteristics were balanced between groups following propensity score matching (all SMD values <0.1).

Table 1.

Characteristics of included study participants following propensity-score matching

Variable		ADHD (n=10,090)	Non-ADHD (n=30,270)	Total (n=40,360)	|SMD|*
Age, y	Mean (±SD)	42.2 (15.2)	42.3 (15.7)	42.3 (15.6)	<0.01
BMI, kg/m2	Mean (±SD)	30.6 (8.1)	30.4 (8.5)	30.5 (8.4)	0.02
Sex	Female	6,365 (63.08%)	19,106 (63.12%)	25,471 (63.11%)	<0.01
	Male	3,725 (36.92%)	11,164 (36.88%)	14,889 (36.89%)	<0.01
Race/Ethnicity	Asian	169 (1.67%)	529 (1.75%)	698 (1.73%)	<0.01
	Black or African American	647 (6.41%)	1,759 (5.81%)	2,406 (5.96%)	0.03
	Hispanic	639 (6.33%)	1,853 (6.12%)	2,492 (6.17%)	<0.01
	White	7,372 (73.06%)	22,392 (73.97%)	29,764 (73.75%)	0.02
	Other	1,263 (12.52%)	3,737 (12.35%)	5,000 (12.39%)	<0.01
Insurance	No	257 (2.55%)	776 (2.56%)	1,033 (2.56%)	<0.01
	Yes	9,833 (97.45%)	29,494 (97.44%)	39,327 (97.44%)	<0.01
Education	College Graduate or Advanced Degree	5,024 (49.79%)	15,643 (51.68%)	20,667 (51.21%)	0.04
	Some College	3,178 (31.5%)	9,223 (30.47%)	12,401 (30.73%)	0.02
	High School or Less	1,888 (18.71%)	5,404 (17.85%)	7,292 (18.07%)	0.02
Income	<$25k	3,059 (30.32%)	8,931 (29.5%)	11,990 (29.71%)	0.02
	$25k–$50k	2,055 (20.37%)	5,980 (19.76%)	8,035 (19.91%)	0.02
	$50k–$100k	2,518 (24.96%)	7,602 (25.11%)	10,120 (25.07%)	<0.01
	≥$100k	2,458 (24.36%)	7,757 (25.63%)	10,215 (25.31%)	0.03
Smoking	No	5,487 (54.38%)	16,625 (54.92%)	22,112 (54.79%)	0.01
	Yes	4,603 (45.62%)	13,645 (45.08%)	18,248 (45.21%)	0.01

ADHD = Attention Deficit Hyperactivity Disorder; BMI = Body Mass Index; SMD = Standard Mean Difference

^*|SMD| <0.1 indicates negligible difference between ADHD and non-ADHD groups post-matching

Association Between ADHD & KC

In the matched cohort, KC was documented in 33 of 10,090 (0.33%) individuals with ADHD and in 42 of 30,270 (0.14%) individuals without ADHD. This reflected significantly higher odds of KC among individuals with ADHD compared to those without ADHD in the univariable analysis (OR=2.36, 95% CI=[1.49, 3.72], p<0.01). There was no evidence of effect modification by sex (p=0.38).

The association between ADHD and KC remained significant in a multivariable model adjusted for potential confounding comorbidities (OR=2.43, 95% CI=[1.28, 4.61], p=0.01; Table 2). Tear film-related disorders also emerged as an independent predictor (OR=5.37, 95% CI=[2.62, 11.01], p<0.01); however, psychiatric comorbidities (p=0.74), the atopic triad (p=0.93), obstructive sleep apnea (p=0.77), ADHD medications (p=0.31), and conjunctivitis (p=0.41) did not.

Table 2.

Summary of multivariable conditional logistic regression model

Variable	OR*	95% CI	P-value
ADHD	2.43	[1.28, 4.61]	0.01
Psychiatric Comorbidities	0.89	[0.44, 1.80]	0.74
Systemic Atopic Comorbidities	1.03	[0.56, 1.90]	0.93
Obstructive Sleep Apnea	1.11	[0.55, 2.22]	0.77
ADHD Medications	0.54	[0.16, 1.80]	0.31
Conjunctivitis	1.39	[0.64, 3.02]	0.41
Tear Film-Related Disorders	5.37	[2.62, 11.01]	<0.01

ADHD = Attention Deficit Hyperactivity Disorder; CI = Confidence Interval; OR = Odds Ratio

^*Estimated using conditional logistic regression to account for 1:3 propensity score matching based on age, body mass index, sex, race/ethnicity, health insurance status, highest level of educational attainment, annual household income, and smoking status

Sensitivity Analysis

The sensitivity analysis reported herein employed a more stringent KC case definition, requiring at least two KC-related diagnostic records, and further matched ADHD and non-ADHD participants according to frequency of eye care utilization. Robust balance was achieved for eye care utilization within the past year between participants with and without ADHD: zero visits (40.5% vs. 40.5%, SMD<0.01), one visit (42.2% vs. 42.2%, SMD<0.01), two to three visits (13.4% vs. 13.4%, SMD<0.01), and more than three visits (3.9% vs. 3.9%, SMD<0.01).

Individuals with ADHD continued to have significantly higher odds of KC in the univariable analysis (OR=2.45, 95% CI=[1.32, 4.58], p<0.01), with no evidence of effect modification by sex (p=0.82). Likewise, ADHD remained associated with KC in the multivariable analysis (OR=2.35, 95% CI=[1.10, 5.02], p=0.03; Supplemental Table 1).

Discussion

This retrospective analysis identified a significant independent association between ADHD and KC within a large, diverse cohort of adults, whereby individuals diagnosed with ADHD were approximately twice as likely to have KC compared to matched controls. Tear film-related disorders, which may potentially represent a proxy for eye rubbing, also emerged as a significant predictor in this setting. While these findings underscore the importance of KC risk stratification in individuals with ADHD, it is important to note that the absolute prevalence of KC remained low in both groups. As such, our results should not be interpreted as supporting universal KC screening for all individuals with ADHD; rather, ADHD may be a clinically relevant component of a broader KC risk stratification framework, particularly when accompanied by additional risk factors such as habitual eye rubbing, other medical, structural, or ocular surface risk factors, a family history of KC, or unexplained visual symptoms.

Our findings build upon recent work by Safir and colleagues, who analyzed nationwide screening data from Israeli adolescents and young adults, reporting a significantly higher prevalence of ADHD among individuals with KC (n=243/1,553, 15.9%) compared to those without KC (n=100,061/939,230, 10.7%).⁸ In their multivariable analysis adjusted for age, sex, weight, and refractive error, the authors identified a significant association between ADHD and KC, albeit the severities of ADHD and KC did not correlate.⁸ Our analysis extends upon these findings by adjusting for additional comorbidities, including psychiatric disorders, the atopic triad, obstructive sleep apnea, conjunctivitis, and tear film-related disorders, the latter of which emerged as a significant predictor of KC in this setting. Interestingly, Safir and colleagues found the association between ADHD and KC to be significant among a subgroup of males but not females,⁸ whereas our interaction term analysis revealed no evidence of effect modification by sex. These discrepancies may reflect sample size considerations within sex-stratified subgroups in the study by Safir and colleagues, whose cohort was male-predominant (n=1,130/1,553 [72.8%] males in the KC group and n=556,435/939,320 [59.2%] males in the non-KC group),⁸ in contrast to our female-predominant cohort (n=25,471/40,360 [63.1%] females across ADHD and non-ADHD groups). Alternatively, these differences may stem from broader demographic differences between the two study populations. Importantly, participants in the study by Safir and colleagues were also considerably younger compared to those in our study (median age of 17 years versus 39 years),⁸ and given that KC may present earlier in males than in females,²³ these age differences could have also influenced the observed sex-related findings. Nonetheless, the literature remains inconsistent regarding sex-related KC prevalence, with some studies reporting higher prevalences among males,^24,25 and others among females.^5,26–28 Thus, additional population-based investigations are needed to clarify potential sex-specific variations in this association.

Repetitive bodily movements, which can include eye rubbing, are well-documented behavioral manifestations in some individuals with ADHD.²⁹ Habitual eye rubbing may serve as a coping mechanism or stress-relief strategy, potentially mediated by vagal stimulation via the oculocardiac reflex.³⁰ Some individuals with ADHD have been shown to exhibit heightened sensory responsiveness, which might increase their susceptibility to pruritic sensations or result in a reduced ability to resist eye rubbing urges.^31–33 These concepts may explain the sustained association between ADHD and KC observed in our multivariable analysis. However, direct data on eye rubbing frequency were not available, and future prospective studies should aim to document this more meticulously. Our findings should therefore be interpreted as hypothesis-generating epidemiologic evidence of an association between ADHD and KC, rather than evidence that repetitive eye-rubbing behaviors mediate this relationship, as this pathway could not be directly evaluated in the present study.

Repeated mechanical microtrauma from chronic eye-rubbing in susceptible individuals has been postulated to induce corneal epithelial injury, local cytokine-mediated inflammation, myofibroblast differentiation, and biomechanical destabilization, ultimately leading to corneal thinning and ectatic changes.^34,35 For instance, reversible corneal ectasia secondary to chronic eye-rubbing has been previously documented in pediatric cases.^8,36,37 McMonnies and colleagues described transient corneal distortions in an eight-year-old male who frequently rubbed his eyes due to ADHD symptoms; notably, there was resolution of corneal disruptions on videokeratography following the treatment of his ADHD symptoms.³⁵ Similarly, another case involving a 13-year-old female reported transient ectatic corneal changes that resolved after cessation of eye-rubbing behaviors at one-year of follow-up.³⁸ In contrast, persistent KC changes have been documented in adults with chronic eye-rubbing habits. Assayag and colleagues reported a case of bilateral superior KC in a 49-year-old male with ADHD, whose habitual eye rubbing behaviors correlated with KC severity in each eye.³⁹ Similarly, Bral and Termote described unilateral KC in a 60-year-old male due to chronic unilateral eye-rubbing, though longitudinal corneal imaging demonstrated cone stability over a five-year period.⁴⁰ Early recognition of chronic eye rubbing and appropriate multidisciplinary intervention may play a critical role in addressing this modifiable risk factor. However, these mechanistic and case-based observations provide biologic plausibility for the observed association but should not be interpreted as direct evidence of a causal or mediating pathway in the present study.

Individuals with ADHD may have a greater likelihood of various ocular comorbidities, including refractive error, strabismus, and amblyopia, which may increase contact with eye care professionals and thereby influence opportunities for KC detection.^20–22 Nonetheless, our sensitivity analysis, which balanced the frequency of eye care utilization between participants with and without ADHD, helped mitigate potential ascertainment bias, and the findings remained consistent with the primary analysis. Safir and colleagues proposed that future KC screening strategies could be enhanced by combining an ADHD diagnosis with a thorough history of recurrent eye rubbing and additional established KC risk factors.⁸ Consistent with this recommendation, our multivariable analysis found that tear film-related disorders, which may be associated with eye rubbing, emerged as an independent predictor of KC. An increased prevalence of dry eye has previously been reported among individuals with ADHD, independent of psychostimulant use,^41,42 although these medications themselves may exacerbate dry eye pathogenesis.⁴³ Given that ocular pruritus frequently accompanies dry eye and blepharitis, which may promote habitual eye rubbing and increase the risk of KC,⁴⁴ patient education regarding eye rubbing behaviors should be prioritized. Proactive, multidisciplinary collaboration among psychiatry, primary care, and ophthalmology may be valuable to identify and address eye rubbing behaviors in individuals with ADHD as early as possible.

Although none of the systemic comorbidities evaluated in the multivariable model emerged as significant confounders in the association between ADHD and KC, several have been independently linked to KC in prior literature and should still be considered in broader risk assessments. Within the atopic triad, strong relationships have been demonstrated between allergic rhinitis and asthma with KC, potentially mediated by inflammatory cytokines.⁴⁵ Genetic studies also support possible causal associations between eczema and KC.⁴⁶ Furthermore, obstructive sleep apnea has been linked to KC in a meta-analysis.⁴⁷ The underlying mechanism may involve increased levels and activity of matrix metalloproteinase-9, an enzyme implicated in corneal thinning and ectatic progression,⁴⁸ as well as the generation of reactive oxygen species.⁴⁹ Eye rubbing during sleep could represent another possible mechanism in patients with obstructive sleep apnea.⁵⁰ Psychiatric conditions, such as anxiety and depressive disorders, have likewise shown bidirectional associations with KC, wherein individuals with these conditions may be at increased risk of KC development or progression due to neglected eye care or stress-related compromises in immune function,⁵¹ while individuals with KC may more susceptible to anxiety or depression due to its impact on quality of life.⁵²

Several limitations must be acknowledged. Reliance on EHR data from the All of Us program introduces the potential for misclassification or underdiagnosis of conditions such as ADHD and KC. No independent clinical verification of cases was performed. As with other retrospective analyses of large-scale databases, diagnoses also could not be confirmed through manual review of clinicians’ notes or formal clinical evaluation. Additionally, reliance on self-reported measures for variables such as smoking status, educational attainment, and insurance coverage may introduce recall or reporting bias. Although participants’ eye care utilization was ascertained through self-reported survey data and balanced in the sensitivity analysis to improve comparability in opportunities for KC detection between groups, these data captured eye care use only within the past year and therefore could not account for lifetime eye care utilization frequency. Approximately 25% of participants with ADHD in the All of Us dataset were not eligible for propensity score matching due to missing covariate data, suggesting potential selection bias; however, had these individuals been included, ADHD would have remained significantly associated with KC with a similar effect size in both the univariable and multivariable models. Compared with the original matched cohort, this expanded cohort would also have shown minor differences in characteristics: the proportion with a high school education or less would have been approximately 3% greater, largely offset by a decrease in college graduates, while the proportion identifying as White would have been approximately 3% lower, with small increases across other racial and ethnic groups. While propensity score matching and multivariable adjustment were used to address confounding, the influence of unmeasured variables, including genetic predisposition, family history, environmental exposures, ADHD severity, as well as true eye-rubbing frequency and intensity, could not be ascertained. We were also unable to account for KC severity due to the paucity of data on patients requiring prescriptions for scleral contact lenses or keratoplasty, and were similarly limited in assessing the influence of floppy eyelid syndrome due to the low number of participants with the diagnosis.⁵³ Information on medication use was based on prescription records, which may not necessarily reflect actual treatment adherence. Although the multivariable model accounted for ADHD medication use as a binary composite variable, separate adjustment for stimulant and nonstimulant medication subclasses would have yielded similarly nonsignificant results, albeit with reduced statistical power. Likewise, individual adjustment for the various psychiatric and atopic comorbidities considered in the multivariable model would have yielded nonsignificant findings while increasing the likelihood of overadjustment, which we sought to mitigate in the original multivariable model by using composite variables. It remains unclear whether complementary strategies beyond pharmacological treatment, such as behavioral interventions, may influence the association between ADHD and KC; these could not be robustly analyzed within the dataset. Our cross-sectional design limits the ability to derive causal inferences. While the proportion of KC cases diagnosed before versus after ADHD diagnosis could not be reported due to the All of Us cell suppression policy, it is important to note that EHR-recorded diagnosis dates may not reflect true disease onset, particularly given that ADHD symptoms often emerge in childhood and that the timing of clinical diagnosis can vary substantially.⁵⁴ Similarly, because the proportion of participants diagnosed with tear film-related disorders before versus after KC diagnosis could not be established because of the All of Us cell suppression policy, temporality between these diagnoses should not be inferred, and this proxy variable for eye rubbing should be interpreted with caution. It is estimated that approximately 92% of the United States population has health insurance coverage,⁵⁵ whereas roughly 97% of participants in our sample were insured, suggesting potential selection bias whereby individuals with healthcare access may be overrepresented, albeit this covariate was well balanced between groups. Lastly, although the All of Us Research Program intentionally recruits participants from diverse and historically underrepresented populations, its reliance on convenience sampling may limit representativeness relative to the broader United States population, as participants may differ across various social determinants of health that could influence ADHD or KC ascertainment. As such, the observed association may not be fully generalizable across all sociodemographic subgroups, and future population-based studies are needed to validate these findings, particularly because subgroup-specific analyses could not be robustly performed due to limited sample sizes and the All of Us cell suppression policy. Although propensity score matching accounted for socioeconomic variables, residual confounding related to the ability to pay for healthcare within the United States healthcare context cannot be excluded. Future studies conducted in other healthcare systems, including publicly funded settings, are also needed to validate the generalizability of our findings across different healthcare access and delivery models.

Conclusion

In this retrospective analysis of a large, diverse cohort of adults, individuals diagnosed with ADHD were found to have approximately twice the odds of KC compared to matched controls. Tear film-related disorders, which may be associated with eye rubbing, also emerged as an independent predictor. Future longitudinal studies are needed to validate the temporal relationship among ADHD, eye rubbing-related behaviors, and KC, as well as to determine whether targeted behavioral interventions may reduce KC risk in this population. Further research should also incorporate detailed documentation of eye-rubbing frequency, intensity, duration, and triggers to explore this proposed mechanism, as such data were not available in the present dataset.