The Risks of Corneal Surface Damage in Aqueous-Deficient Dry Eye Disease: A 17-Year Population-Based Study in Taiwan


  • A 17-year population-based study examines the risk of corneal surface damage and epidemiology of aqueous-deficient dry eye disease in Taiwan.

  • Aqueous-deficient dry eye disease increases the risk of corneal surface damage, such as recurrent corneal erosion, corneal ulcers, or corneal scars.

  • The risk factors for corneal surface damage in aqueous-deficient dry eye disease include younger age (<18 years), female sex, diabetes mellitus, and autoimmune diseases.

  • The prevalence of aqueous-deficient dry eye disease increases during the study period, especially in patients older than 65 years.


To investigate the epidemiologic characteristics and risk of corneal surface damage in patients with aqueous-deficient dry eye disease (DED) in Taiwan.


Retrospective, population-based cohort study.


We used claims data in the Taiwan National Health Insurance Research Database from 1997 to 2013 of patients with DED, defined according to diagnoses, drug codes, and clinical follow-up. A comparison cohort without DED was selected through propensity score matching. The main outcome measures were corneal surface damage, including corneal erosion, corneal ulcers, or corneal scars.


Patients with DED had a significantly higher rate of corneal surface damage (hazard ratio [HR]: 2.70; 95% confidence interval [CI] 2.38-3.06, P < .001), especially higher in patients aged <18 years (HR 6.66; 95% CI 3.58-12.41) than in older patients and in women (HR 2.98; 95% CI 2.57-3.46) than in men (HR 2.22; 95% CI 1.78-2.77), compared to those in the non-DED cohort. DED with diabetes mellitus ( P = .002), rheumatoid arthritis ( P = .029), or systemic lupus erythematosus ( P = .005) was positively associated with corneal surface damage. The overall prevalence of DED was 7.85%, higher among women (10.49%) than men (4.92%), and increased with age (0.53%, 3.94%, 10.08%, and 20.72% for ages <18, 18-39, 40-64, and >65 years, respectively). The prevalence increased gradually during the study period.


The younger age group (<18 years) had the highest risk of corneal surface damage in aqueous-deficient DED. Other predisposing factors included female sex, diabetes, and autoimmune diseases. To improve clinical care, special attention is required for patients with DED with these risk factors.

T he increasing prevalence of dry eye disease (DED) has become a global health concern. The prevalence estimates for DED range from 5% to 50%, which vary with different definitions of DED and the characteristics of the study population. The prevalence appears higher in individuals of Asian ethnicity, and there is now an increased awareness of DED among young adults. , In Taiwan, the prevalence rate in patients older than 65 years is 33.7%. In a population-based study, the overall rate was found to be 4.6%, although patients younger than 20 years were excluded. DED is a multifactorial disease that leads to damaged ocular surface and visual disturbance, thereby affecting quality of life, and is the most common reason for seeking medical eye care services. DED imposes a significant economic burden on the patient and the society, directly through medical cost and indirectly by reducing work productivity.

The 2 main categories of DED classification are aqueous-deficient and evaporative DED, in which aqueous-deficient DED is mainly associated with decreased lacrimal secretion, while evaporative DED is owing to meibomian gland dysfunction, though there are overlaps between the 2 groups. The International Dry Eye Workshop II (DEWS II) has described tear film instability, hyperosmolarity, ocular surface inflammation and damage, and neurosensory abnormalities as the key pathologic mechanisms in the development of DED. An unstable tear film does not provide sufficient protection of the ocular surface. DED leads to characteristic punctate epithelial keratitis, filamentary keratitis, superior limbic keratoconjunctivitis, lid parallel conjunctival folds, and lid wiper epitheliopathy. The disrupted ocular surface can lead to more severe corneal surface abnormalities, such as recurrent corneal erosion, corneal ulcers, and corneal scars. However, studies have reported that DED is associated with corneal erosion, sterile corneal ulcers, or microbial corneal ulcers only when it is concomitant with systemic diseases such as diabetes mellitus (DM) and autoimmune diseases.

In the clinic, patients with DED could present with no to severe corneal surface damage. Several studies have provided information on the risk factors of DED. However, none have investigated the risk of corneal surface damage in DED. Therefore, we conducted this population-based study to identify the risk of corneal surface damage in patients with aqueous-deficient DED in Taiwan. Furthermore, we identified the epidemiology of aqueous-deficient DED in the population.


Data Source

The National Health Insurance Research Database (NHIRD) is maintained by the National Health Research Institutes of Taiwan. The NHIRD contains registration files and original claims data for reimbursement. All data from the NHIRD are encrypted and de-identified. The Longitudinal Health Insurance Database, a subset of the NHIRD, contains original claims data of 1 million randomly sampled beneficiaries. We accessed data from January 1, 1997, to December 31, 2013, for this population-based study. This study adhered to the tenets of the Declaration of Helsinki and was approved by the Ethics Institutional Review Board of Chang Gung Memorial Hospital (IRB: 102-2147B), which waived requirements for informed consent because the medical data were de-identified.

Patient Enrollment and Exclusion Criteria

We identified patients with DED according to the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes 370.33, 372.53, 375.15, and 710.2. To ensure the accuracy of DED diagnoses, we selected outpatients who had at least 2 ophthalmology clinic visits with the diagnoses of DED, and had been prescribed eye lubricants. The Taiwan National Health Insurance program regulates the reimbursement of eye lubricants to patients with a Schirmer test score of less than 5 mm; therefore the selected patients in the cohort are mainly patients with aqueous-deficient DED. To simplify, the cohort is termed “DED group” or “DED cohort” throughout the manuscript. We excluded those with a clinical follow-up of less than 1 year or a diagnosis of corneal diseases, including corneal opacity and other disorders of the cornea (371); corneal ulcers (370.0); interstitial and deep keratitis (370.5); corneal neovascularization (370.6); eye burn (940); or open globe injury (871) before the diagnosis of DED. Moreover, we excluded those who died during the study to exclude serious health conditions.

Outcome Measurements and Risk Factors

Corneal surface damage was defined as diagnosis of recurrent corneal erosion (371.42), corneal ulcers (370.0), or corneal scars (371.0) during the follow-up visit of patients with DED.

Risk factors for corneal surface damage were determined before the diagnosis of DED. Diseases were collected according to the ICD-9-CM codes and drugs were collected according to the Anatomical Therapeutic Chemical codes. The codes are shown in Supplemental Table 1 (Supplemental Material available at

Statistical Analysis

Each patient in the DED cohort was matched by using 1:1 propensity score matching with a counterpart in the non-DED cohort to achieve minimal bias. , In propensity score matching, different weighted scores were given for each factor, and the patients were matched according to the sum of the scores. The patients were matched by the following factors: age, sex, number of outpatient clinic visits, Charlson comorbidity index score, and follow-up years. Continuous variables were presented as the mean ± standard deviation. An absolute standardized mean difference of <0.1 indicated a negligible difference between the 2 study cohorts. Subgroup and risk analyses were performed using a Cox regression model. The prevalence rate was the number of patients identified with DED divided by the alive population in the year. All statistical analyses were performed using the SAS EG software package (SAS Institute, Inc, Cary, North Carolina, USA).


Study Population

From January 1, 1997, to December 31, 2013, a total of 738,839 patients visited the outpatient ophthalmology clinic. After applying the exclusion criteria, we selected 682,388 patients, of which 51,107 had been diagnosed as having DED and 631,281 had not. Finally, after propensity score matching, 23,667 patients were analyzed in each cohort ( Figure 1 ).

Figure 1

Selection of study cohorts. DED = dry eye disease; NHIRD = National Health Insurance Research Database.

Characteristics of the Study Cohort

Table 1 presents the demographics and clinical characteristics of the non-DED and DED cohorts after propensity score matching. The absolute standardized mean difference values were less than 0.1 for mean age (47.13 years vs 46.57 years), male sex (7,720 [32.62%] vs 9,049 [34.01%]), and Charlson comorbidity index score (0.99 vs 0.96), indicating that the study population was well balanced between the non-DED and DED cohorts. The non-DED group contained more patients younger than 18 years (2,390 [10.10%] vs 527 [2.23%]) and older than 65 years (4,465 [18.87%] vs 3,259 [13.77%]). Patients in the DED cohort had more outpatient clinical follow-ups (3.17 years vs 2.70 years) and longer follow-up periods (6.80 years vs 6.15 years), which accords with the clinical behavior of patients with DED. The characteristics before propensity score matching are shown in Supplemental Table 2 (Supplemental Material available at

Table 1

Demographic Information of the Study Population After Propensity Score Matching

Non-DED (N = 23,667) DED (N = 23,667) ASMD
Age, years a 47.13 ± 19.79 46.57 ± 15.93 0.03
Group age, n (%)
0-17 years 2,390 (10.10) 527 (2.23) 0.30
18-39 years 5,354 (22.62) 7,898 (33.37) 0.21
40-64 years 11,458 (48.41) 11,983 (50.63) 0.04
≥65 years 4,465 (18.87) 3,259 (13.77) 0.12
Male, n (%) 7,720 (32.62) 8,049 (34.01) 0.03
OPD visit within 1 year, n (%) 2.70 (3.43) 3.17 (2.64) 0.15
CCI score a 0.99 ± 1.49 0.96 ± 1.27 0.02
Follow-up, years a 6.15 ± 4.3 6.80 ± 4.11 0.15
Comorbidities, n (%)
Chronic pulmonary disease 4,762 (20.12) 4,866 (20.56) 0.02
Diabetes mellitus 2,835 (11.98) 2,779 (11.74) 0.07
Cerebrovascular disease 1,565 (6.61) 1,450 (6.13) 0.02
Any malignancy 1,139 (4.81) 779 (3.29) 0.08
Heart disease 1,007 (4.25) 673 (2.84) 0.08
Renal disease 665 (2.81) 570 (2.41) 0.03
Peripheral vascular disease 515 (2.18) 453 (1.91) 0.02
Autoimmune disease 374 (1.58) 679 (2.87) 0.09
Dermatologic disease 285 (1.20) 487 (2.06) 0.07
Dementia 172 (0.73) 143 (0.60) 0.02
Liver disease 71 (0.30) 33 (0.14) 0.03

ASMD = absolute standardized mean difference; CCI = Charlson comorbidity index; DED = dry eye disease; OPD = out-patient department.

a Continuous variate is presented as mean ± standard deviation.

Study Outcomes

The outcomes are listed in Table 2 , and the results of subgroup analyses are listed in Tables 3 and 4 . The cumulative prevalence of patients with DED according to sex and age are presented in Figure 2 .

Table 2

Forms of Corneal Surface Damage Between the 2 Study Cohorts

Outcome Non-DED (N = 23,667) DED (N = 23,667) HR (95% CI) P Value
Corneal surface damage, n (%) 338 (1.43) 995 (3.50) 2.70 (2.38-3.06) <.0001
Recurrent corneal erosion, n (%) 46 (0.19) 395 (1.39) 7.64 (5.62-10.39) <.0001
Corneal ulcer, n (%) 167 (0.71) 360 (1.27) 1.97 (1.64-2.38) <.0001
Corneal scar, n (%) 125 (0.53) 240 (0.84) 1.83 (1.47-2.28) <.0001

Only gold members can continue reading. Log In or Register to continue

Stay updated, free articles. Join our Telegram channel

Jul 10, 2021 | Posted by in OPHTHALMOLOGY | Comments Off on The Risks of Corneal Surface Damage in Aqueous-Deficient Dry Eye Disease: A 17-Year Population-Based Study in Taiwan

Full access? Get Clinical Tree

Get Clinical Tree app for offline access