Risk of Incident Dry Eye Disease Among Patients With Chronic Pain Conditions.

This retrospective cohort study used data from the TriNetX Analytics Network, a federated, de-identified electronic health record network comprising data from academic medical centers, community hospitals, and specialty practices in the United States and internationally. The network includes longitudinal demographic information, diagnoses, procedures, medications, and encounter data. The dataset used for the analysis spanned 20 years: December 5, 2005, through December 5, 2025. Because only de-identified data was used, this study was exempt from institutional review board review and was conducted in accordance with the principles of the Declaration of Helsinki. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines for cohort studies. Adult patients aged 18 years or older were identified in TriNetX using International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnosis codes, Systematized Nomenclature of Medicine (SNOMED) drug concepts, and Current Procedural Terminology (CPT) codes for procedures/encounters. The date of the first qualifying ophthalmology visit meeting cohort criteria served as the index date. To ensure a DED–naïve baseline, patients were required to have no prior diagnosis of dry eye syndrome (H04.12) or keratoconjunctivitis sicca, not specified as Sjögren’s (H16.22) on or before the index date. To reduce ascertainment bias related to differences in access to ophthalmologic care, a secondary validation cohort was constructed by restricting the population to patients with a diagnosis of age-related cataract (ICD-10-CM: H25), a common ophthalmic diagnosis that frequently prompts structured eye care engagement independent of dry eye symptoms. This approach was intended to help assess whether the observed association between CPCs and DED was independent of referral bias or differential access to ophthalmologic evaluation. Within this cohort, the index date was the first qualifying cataract diagnosis. CPC exposure was defined as in the primary analysis, and patients were required to have no prior diagnosis of DED before or on the index date. Incident DED and prescription-requiring DED were assessed during follow-up using the same outcome definitions. Patients with prior or subsequent cataract surgery were not separately excluded from the age-related cataract validation cohort. This cohort was intended to serve as a validation analysis restricted to individuals with a common ophthalmic diagnosis prompting eye care engagement, rather than as an analysis of cataract surgery itself. The exposure cohort (“CPC cohort”) was defined as patients with at least one or more of the recognized CPCs: endometriosis (N80), interstitial cystitis (chronic) (N30.1), vulvodynia (N94.81), chronic tension-type headache (G44.22), low back pain (M54.5), temporomandibular joint disorders (M26.6), chronic migraine without aura (G43.7), fibromyalgia (M79.7), myalgic encephalomyelitis/chronic fatigue syndrome (G93.32), irritable bowel syndrome (K58), or chronic fatigue, unspecified (R53.82), prior to a clinical visit to an ophthalmic provider identified using the CPT codes for a new ophthalmic patient examination (92002, 92004) or established ophthalmic patient examination (92012, 92014). 15 CPC exposure was modeled as a binary variable based on the presence of any qualifying CPC diagnosis. The cumulative number of CPC diagnoses per individual was not modeled. The control cohort (“no CPCs cohort”) required the same qualifying ophthalmic examination CPT codes but without any of the listed pain syndrome diagnoses before or after the index date. Primary outcomes were incident DED at 1, 2, and 3 years after the index date, defined by a new diagnosis of dry eye syndrome (H04.12) or keratoconjunctivitis sicca, not specified as Sjögren’s (H16.22) during follow-up. A secondary outcome for prescription-requiring DED was defined by treatment requiring cyclosporine (RxNorm 3008) or lifitegrast (RxNorm 1801820), identified using RxNorm medication codes. This outcome was used as a proxy for treatment escalation beyond lubricant therapy rather than a direct measure of disease severity. Because lifitegrast became available during the study period, this medication-based outcome may be influenced by temporal changes in medication availability, uptake, and prescribing practices. Therefore, prescription-requiring DED was interpreted as a proxy for escalation to dry eye-specific prescription therapy rather than a direct measure of disease severity across the full study period. To control for baseline differences that could confound the association between exposure and outcome, 1:1 propensity score-matching was performed using the TriNetX built-in analysis platform. Matching was conducted using a nearest-neighbor approach with a caliper of 0.25 standard deviations and was considered adequate when standardized mean differences were less than 0.1 across all covariates. Covariates used for matching included baseline demographics such as age at index date, sex, race, and ethnicity as documented in the electronic health record. Additional medical comorbidities were selected to control for other systemic conditions that may influence DED risk, including essential (primary) hypertension (I10), hyperlipidemia, unspecified (E78.5), diabetes mellitus (E08–E13), other anxiety disorders (F41), depressive episode (F32), major depressive disorder, recurrent (F33), other hypothyroidism (E03), hyperthyroidism [thyrotoxicosis] (E05), other rheumatoid arthritis (M06), rheumatoid arthritis with rheumatoid factor (M05), systemic lupus erythematosus (M32), and Sjögren syndrome (M35.0). The same matching strategy and covariate set were applied in the age-related cataract validation cohort. Primary analyses were conducted in the propensity-matched cohorts. For each discrete follow-up window (1-, 2-, and 3-year), risk differences, risk ratios, 95% confidence intervals, and p-values were obtained. In addition, time-to-event outcomes were assessed using Kaplan–Meier methods to generate cumulative incidence curves for incident DED (and for DED requiring cyclosporine or lifitegrast), with between-cohort differences evaluated using the log-rank test. Multivariable Cox proportional hazards regression models were used to evaluate the association between CPC status and time to incident DED within three years of follow-up. CPC status was entered into models as a binary exposure variable defined by the presence of any CPC. Covariates included in the Cox models were prespecified and selected to account for demographic factors and medical comorbidities that may influence DED risk. Differences between variables included in the propensity score-matching and Cox regression reflect the distinct purposes of cohort balancing and time to event risk estimation. Because the primary objective was to evaluate CPC status as a composite exposure, analyses were not stratified by individual CPC subtype. Such stratification would have required separate TriNetX queries, each with distinct exposed and control cohorts, propensity score matching, baseline characteristics, and outcome estimates. These queries would also not be fully comparable, because each reflects the participating organizations and data available within the federated network at the time of analysis, such that denominators and cohort sizes may vary across data pulls.

Prior to matching, the CPC cohort included 659,892 patients and the no CPC cohort included 1,497,893 patients. For follow-up analyses, patients with an index event more than 20 years prior were excluded (12,125 from the CPC cohort and 46,660 from the no CPC cohort), leaving 647,767 and 1,451,233 patients before matching, and 538,364 patients per cohort after matching ( Figure 1 ). Baseline characteristics were well balanced after matching, with standardized mean differences below 0.1 for all covariates ( Table 1 ). The matched cohorts had a mean age of approximately 53 years, were predominantly female (approximately 59%), and had similar distributions of race, ethnicity, and major medical comorbidities, supporting comparability between cohorts. Across all follow-up intervals, patients with at least one CPC demonstrated a substantially higher incidence of newly diagnosed DED compared with matched controls ( Table 2 ). Differences between groups were present at 1-year and increased at year 2 and year 3. In year 1, incident DED occurred in 3.34% of patients with one or more CPC compared with 0.72% of controls, corresponding to a more than fourfold relative increase in risk (RR 4.64; 95% CI, 4.49–4.81; p<0.0001). This difference persisted and increased over time, with incident DED occurring in 5.64% versus 1.15% at 2 years (RR 4.89; 95% CI 4.76–5.02; p<0.0001) and 7.16% versus 1.50% at 3 years (RR 4.78; 95% CI, 4.66–4.89; p<0.0001), respectively. Kaplan–Meier analyses were consistent with these findings, demonstrating early separation of cumulative incidence curves within the first year that continued to diverge throughout follow-up ( Figure 2 ). The CPC cohort demonstrated a consistently higher cumulative incidence of DED than the control cohort at one year (3.65% vs 0.83%), two years (6.57% vs 1.44%), and three years (8.82% vs 2.02%; all p<0.0001). In addition to overall incident DED, patients with CPCs were significantly more likely to develop DED requiring cyclosporine or lifitegrast, used as a proxy for treatment escalation beyond lubricant therapy rather than a direct measure of disease severity. Absolute event rates were lower than for overall DED and consistent differences were observed at each time point ( Table 2 ). At one year, prescription-requiring DED occurred in 0.44% of patients with CPCs compared with 0.17% of controls. At two years, incidence increased to 0.65% versus 0.25%, and by three years to 0.79% versus 0.30%, respectively (all p<0.0001). Time-to-event analysis demonstrated early and sustained separation between groups ( Figure 3 ). The cumulative incidence of prescription-requiring DED was higher in the CPC cohort at one year (0.48% vs 0.20%), two years (0.74% vs 0.31%), and three years (0.95% vs 0.39%; all p<0.0001). In multivariable Cox proportional hazards modeling, the presence of a CPC was independently associated with incident DED over the three-year study period (hazard ratio 4.85; 95% CI, 4.76–4.94; p<0.0001) after adjustment for demographics and other relevant medical comorbidities ( Table 3 ). While several covariates, including age, sex, autoimmune disease, and psychiatric comorbidities, were also associated with DED risk, CPC status demonstrated the largest effect size among variables included in the multivariable model. In the validation cohort restricted to patients with age-related cataract, CPC status remained associated with increased risk of both incident DED and prescription-requiring DED across all follow-up intervals ( Supplemental Table 1 ). Although relative risks were modestly attenuated compared with the primary analysis, the pattern and direction of association were consistent. Multivariable time-to-event modeling further demonstrated that CPC status remained independently associated with incident DED ( Supplemental Table 2 ).

In this large, propensity-matched retrospective cohort study, individuals with at least one CPC demonstrated a substantially higher risk of developing incident DED over one, two, and three years of follow-up compared with matched controls without a CPC. This association persisted across all analytic approaches, including time-to-event modeling and a validation analysis restricted to patients with age-related cataract, a population with a common ophthalmic diagnosis that frequently prompts structured eye care engagement independent of DED symptoms. Notably, the magnitude of risk was large, with patients with existing CPCs exhibiting approximately a four- to five-fold increased hazard of incident DED and a two- to three-fold increased risk of requiring prescription immunomodulatory therapy, which we used as a proxy for escalation to dry eye-specific prescription therapy. Collectively, these findings support the concept that CPCs are not only comorbid with DED but may precede and meaningfully predict its development and possibly its severity. Prior literature has consistently demonstrated clinical overlap between DED symptoms and systemic pain conditions, although most studies have been cross-sectional, clinic-based, or focused on symptom severity rather than disease incidence. 16 , 17 In a Miami Veterans Affairs cohort of 154 patients with DED symptoms, study participants with a higher burden of CPCs reported more severe symptoms and ocular pain than the group of participants with low chronic pain burden, while measures of ocular surface signs (tear osmolarity, tear breakup time, corneal staining, and Schirmer testing) were equivalent between the groups. 18 Similar findings have been noted in other parts of the world. A cross-sectional study of 425 patients in the Netherlands reported that 17% of tertiary DED patients carried a diagnosis of at least one CPC and that these patients had significantly worse Ocular Surface Disease Index (OSDI) scores across all subscales despite no differences in Schirmer testing, tear breakup time, or corneal staining. 19 Population-level data further support overlap between CPCs and DED. A systematic review and meta-analysis including over three million individuals from 11 large studies found that migraine was associated with higher odds of DED, with a pooled odds ratio of 1.59 (95% CI, 1.41–1.79; I 2 = 89.3%; p < .001). 20 Another meta-analysis found that individuals with CPCs had higher odds of DED symptoms, with an odds ratio of 3.51 (95% CI, 3.45-3.57; I 2 = 100 %; p < 0.001), while objective tear parameters showed small or inconsistent differences between cohorts with and without CPCs. 17 These studies collectively suggest an overlap between CPCs and DED, but have not clarified a temporal association. 17 , 19 , 21 The incidence of dry eye disease observed in the present analysis is lower than prevalence estimates reported in many prior studies because the cohort was restricted to individuals without a prior dry eye diagnosis at baseline, thereby capturing incident rather than prevalent disease. This study specifically addressed this gap by demonstrating a temporal association between CPCs and incident DED. Several biologic pathways may explain why the presence of CPCs increases subsequent DED risk, including shared nociplastic mechanisms such as ascending facilitation of nociceptive signaling, reduced endogenous descending inhibition, and other altered trigeminal somatosensory processing mechanisms, as potential factors linking the two entities. 22 , 23 Recent experimental work supports the existence of pain-dominant ocular phenotypes independent of ocular surface damage. In a mouse model, unilateral constriction of the long ciliary nerves, which carry afferent sensory fibers from the cornea to the trigeminal system, produced persistent corneal hyperalgesia and anxiety-like behavior without increased corneal staining or loss of corneal sensation, demonstrating sustained nociceptive signaling in the absence of surface pathology. 24 This pattern parallels clinical observations in which individuals with systemic pain conditions report severe ocular symptoms disproportionate to objective tear film findings. 25 This mechanistic framework has treatment implications because pain-dominant phenotypes of DED may require therapies beyond surface directed topical approaches. In individuals with nociplastic or centralized pain, treatment in other pain syndromes often incorporates therapies that target altered pain processing within the central nervous system such as alpha 2 delta (α2δ) ligands (e.g., gabapentin) and/or serotonin norepinephrine reuptake inhibitors (e.g., duloxetine). 26 , 27 Consistent with this framework, a retrospective study of 101 adults with chronic ocular surface pain with neuropathic features, defined by symptoms out of proportion to ocular surface findings, sensitivity to wind or light, cutaneous allodynia, abnormal corneal sensitivity, or persistent pain despite topical anesthetic, oral therapies such as α2δ ligands reduced ocular pain in most post traumatic (81.2%), migraine like (73.0%), and unilateral (72.7%) phenotypes, but less often in postsurgical (38.5%) ocular pain phenotypes. 28 Similarly, in a randomized clinical trial of 72 individuals with DED with a suspected neuropathic pain component based on the painDETECT questionnaire, the addition of oral gabapentin to artificial tears and topical cyclosporine treatment resulted in significantly greater improvement in OSDI scores, tear breakup time, and Schirmer testing compared with topical therapy alone. 29 Together, these findings reinforce the importance of identifying pain-dominant phenotypes in DED, as aligning treatment strategy with underlying mechanisms may improve symptom control more efficiently and improve quality of life outcomes. Several limitations must be considered when interpreting results from this retrospective analysis. First, large database studies are susceptible to inaccuracies in ICD-10 coding due to inconsistent coding practices across providers, incomplete data entry, or human error, potentially resulting in misclassification of both exposures and outcomes. Second, the electronic health record dataset includes only individuals who accessed medical and eye care, which may limit generalizability to populations with reduced access to health services. Third, symptom severity, ocular surface findings, and pain phenotyping could not be directly assessed within the electronic health record, precluding distinction between tear-deficient, inflammatory, and pain-dominant DED subtypes. Fourth, although propensity matching and multivariable adjustment were performed, residual confounding from unmeasured factors may remain. For example, differences in healthcare utilization patterns, including frequency of ophthalmic encounters, may have influenced the likelihood of DED detection. However, healthcare utilization could not be directly adjusted for because the TriNetX platform does not provide a standardized or validated measure of visit frequency, intensity of care, or indication for evaluation across contributing organizations. Ophthalmic encounter codes indicate only that a qualifying visit was documented; adjusting for such codes would therefore incompletely capture utilization and could itself introduce bias, particularly if ophthalmic visits occurred as a consequence of ocular symptoms or dry eye evaluation. Medication exposure could also not be reliably incorporated into the analyses across the TriNetX network. Medications commonly used to treat CPCs, including antidepressants, antihistamines, and anticholinergic agents, may contribute to ocular surface symptoms and could partially mediate the observed association between CPCs and DED. Fifth, the initiation of cyclosporine or lifitegrast does not necessarily indicate more severe dry eye disease, as prescribing patterns may vary by clinician and institution. This secondary outcome should therefore be interpreted as reflecting escalation to dry eye-specific prescription therapy rather than definitive disease severity. Temporal changes in DED treatment availability and prescribing practices, including the introduction and uptake of lifitegrast during the study period, may also have influenced the secondary prescription-requiring DED outcome and it is important to note that no formal calendar-period adjustment or sensitivity analysis stratified by medication availability was performed. In addition, topical corticosteroids and antibiotics were not included as treatment escalation medications because they are commonly prescribed for a wide range of ocular surface conditions and therefore are less specific indicators of DED treatment escalation. Sixth, in the age-related cataract validation cohort, patients with prior or subsequent cataract surgery were not separately excluded. Because cataract evaluation and perioperative cataract care often involve structured preoperative and postoperative ophthalmic assessment, including evaluation of the ocular surface, this may increase detection of dry eye disease. Cataract surgery itself may also influence ocular surface symptoms, treatment exposure, and dry eye disease diagnosis. Therefore, these factors may represent an additional source of residual confounding within the validation cohort. Seventh, CPC exposure was analyzed as a composite binary variable, and we did not evaluate variation in DED risk across individual CPC subtypes. As a result, we could not determine whether specific CPCs were associated with greater risk of incident DED or prescription-requiring DED. Future studies examining individual CPC phenotypes and differences between diagnosis-only and treatment-requiring DED may provide further mechanistic insight. Finally, because this was a retrospective observational study, the observed associations should not be interpreted as causal. In conclusion, we observed a strong association between CPCs and subsequent development of DED, with individuals carrying one or more CPC diagnoses more than four times as likely to receive a DED diagnosis within three years compared to a matched cohort without CPCs. These results support the growing recognition that, in a subset of patients, DED may co-occur with systemic chronic pain conditions and may reflect shared pain-processing or neurosensory mechanisms in addition to ocular surface pathology. Ultimately, incorporating screening for pain comorbidities into DED evaluation may improve risk stratification, guide earlier recognition of pain-associated phenotypes, and support more individualized management strategies for patients with underlying CPCs.

Dry eye disease (DED) is a highly prevalent condition that can cause significant reductions in one’s quality of life. 1 , 2 Contemporary consensus definitions recognize DED as a multifactorial disease characterized by loss of tear film homeostasis, with inflammation and neurosensory abnormalities playing key etiologic roles. 3 Despite this expanded framework, clinical management often continues to emphasize tear deficiency and ocular surface findings, frequently relying on topical therapies alone. 4 However, a substantial subset of patients report severe symptoms that are disproportionate to clinical signs, demonstrate limited response to standard treatments that affect peripheral or local targets, and accumulate escalating treatment burden over time. 5 – 7 These observations reinforce that DED is not a single entity, but a heterogeneous syndrome with multiple biological drivers and clinically distinct phenotypes. One framework that may explain this heterogeneity is that ocular symptoms in some individuals reflect altered pain processing in the central nervous system rather than isolated ocular surface pathology. 8 Chronic pain conditions (CPCs) are a group of commonly cooccurring disorders, including fibromyalgia, chronic fatigue syndrome, migraine, temporomandibular disorders, irritable bowel syndrome, interstitial cystitis, endometriosis, and vulvodynia, among others, that share altered pain processing, most often involving central mechanisms, though peripheral contributions may also be present. 9 , 10 These conditions share features of central sensitization, altered nociceptive modulation, and high rates of comorbid non-pain CNS-mediated symptoms, such as mood and sleep disorders. 11 , 12 Emerging evidence suggests a clinically meaningful overlap exists between ocular discomfort and systemic pain conditions. In a large U.S.-based case–control study of 72,969 adults, migraine was associated with significantly higher odds of a DED diagnosis, indicating a greater prevalence of DED among individuals with migraine compared with the general population. 13 Extending these findings beyond headache disorders, a Turkish cross-sectional study comparing adults with fibromyalgia syndrome to healthy controls demonstrated markedly higher rates of DED (defined by the presence of corneal staining, 60.6% vs 21.4%) and meibomian gland (MG) dysfunction (defined by abnormalities in various MG parameters, 36.6% vs 4.3%) in fibromyalgia, while other aspects of eye health, including visual acuity and central corneal thickness, were similar across the groups. 14 Together, these data support the concept that ocular symptoms and signs may arise within a broader pain-prone phenotype rather than solely from localized ocular surface pathology. Although prior studies have reported cross-sectional associations between CPCs and DED, few investigations have evaluated whether CPCs precede and predict the development of incident DED at the population level. Clarifying this relationship is clinically important, as it would support a risk-stratified approach to the DED evaluation and justify earlier recognition of pain-associated disease phenotypes. Accordingly, the objective of the current study was to evaluate the association between CPCs and subsequent risk of incident DED and prescription-requiring DED, using a large federated electronic health record network.