Article Text

Download PDFPDF
Severe dysfunctional tear syndrome patients and resolution of central corneal staining: retrospective cohort study
  1. Zhen Ling Teo1,2,
  2. Clarisse Chu1,
  3. Louis Tong1,2,3,4
  1. 1 Ocular Surface Research Group, Singapore Eye Research Institute, Singapore
  2. 2 Department of Ophthalmology, Singapore National Eye Centre, Singapore
  3. 3 Department of Ophthalmology, Yong Loo Lin School of Medicine, National University of Singapore, Singapore
  4. 4 Ophthalmology & Visual Sciences Academic Clinical Program, Duke-NUS Medical School, Singapore
  1. Correspondence to Louis Tong, Singapore National Eye Centre, Singapore 168751, Singapore; louis.tong.h.t{at}


Purpose Severe dry eye is widely prevalent yet difficult to treat. This study aims to evaluate for improvement in epithelial status and the risk factors for lack of improvement in a cohort of patients in Singapore with severe dry eye.

Methods We retrospectively identified 1712 patients with severe dry eye (≥grade 3 Delphi) in at least one eye, referred to a tertiary centre dry eye clinic from 2006 to 2017. We included patients with central corneal staining grade of ≥2 at referral and minimum follow-up duration of 6 months (n=407). An epithelial staining grade of <2 at the last visit was considered a significantly improved outcome.

Results The mean follow-up duration was 4.0±2.4 years, with 88.0% (358/407) of patients achieving significant improvement. Various treatment modalities including topical corticosteroids (32.4%), cyclosporine (52.8%) and punctal plugs (24.1%) were used. Risk factors for non-improvement of staining grade include autoimmune disease (OR 3.2, 95% CI: 1.7 to 6.1), rheumatoid arthritis (RA) (OR 3.4 (1.8 to 6.6)), graft-versus-host disease (GVHD) (OR 3.4 (1.0 to 11.7)), reduced baseline Schirmer’s test (OR 2.1 (1.2 to 3.9)) and reduced tear break up time (OR 2.0 (1.0 to 3.8)). On multivariate analyses, RA and GVHD were still significant risk factors. Gender, age and meibum viscosity were not significantly associated with epithelial staining grade improvement.

Conclusions Overall, a high rate of corneal epithelial improvement was achieved. Nevertheless, there is an unmet need for more effective measures to reduce epitheliopathy in severe dry eye, especially in patients with systemic immune-mediated disease.

  • ocular surface
  • cornea
  • treatment other
  • tears

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Dry eye disease (DED) is one of the most common ocular surface conditions globally, affecting 5%–50% of the adult population, particularly in Asia, women and older age groups.1 2 The prevalence in Southeast Asia is high at 20%–52.4% and 14.5% in the USA.2 It can significantly impede work productivity, sleep quality and physical activity, generating an estimated $55.4 billion per year in socioeconomic cost in the USA.1

The underlying pathophysiology of DED is complex.3 Anatomical abnormalities, meibomian gland dysfunction (MGD), inflammatory disease and environmental factors predispose to ocular surface inflammation and subsequent loss of tear film homeostasis.4 Treatment options aim to increase tear volume and quality, reduce inflammation and treat lid disease;5 however, each is met with varying clinical response.6 7 Prolonged use of prescription eyedrops is associated with significant adverse effects, for example, corticosteroids may induce glaucoma, cataracts and the activation of latent ocular infection. Cyclosporine eyedrops can induce stinging that may be sufficient to induce a loss of compliance. The lack of sustainable and effective treatment modalities has led patients to seek alternative treatments such as traditional acupuncture or herbal medical therapy. However, these have also been shown to provide limited symptomatic improvement beyond 1 month.8

The management of dry eye is further complicated by the varying treatment end points used by clinicians. Many parameters, including patient’s symptoms, corneal staining, tear break up time (TBUT), Schirmer’s test and tear film inflammatory markers, have previously been used for assessing treatment outcome.

Since treatment for dry eye is often chronic, extensive and not free of adverse effects, it is important to determine if treatment objectives have been met and if healthcare resources have been used appropriately. A longitudinal British study found that patients with dry eye improved after chronic management.9 However, this study was performed before cyclosporine was widely used and it was not performed in an Asian setting. While a previous paper4 has identified graft versus host disease (GVHD), autoimmune disease and aqueous deficiency as risk factors for the diagnosis of dry eye, there is limited information on whether these conditions also limit improvement after management with existing treatment modalities.

Our study aims to identify the proportion of patients with severe dry eye who achieved objective improvement after a period of treatment. We further examine the risk factors for failure to improve with the available treatment options. The knowledge of these risk factors will then allow for the further development of targeted treatment for dry eye, and the devotion of resources to these at-risk groups.


Study design and participants

This was a retrospective cohort study conducted at the Singapore Eye Research Institute. Questionnaire data collected at first visit were obtained with written informed consent while anonymised data for subsequent visits were used with assumed consent.


We identified 1712 adult patients (21 years old and above), who were referred to the dry eye clinic in the Singapore National Eye Centre, a tertiary referral centre, between 2006 and 2017, and found to have severe dry eye (defined as at least grade 3 in Delphi classification).10 The cornea was divided into five zones and Baylor staining score in each zone was evaluated by counting the number of fluorescein spots, any confluence of spots and presence of filaments, after instillation of a minimal volume of fluorescein dye using a Floret strip. A higher score indicated more severe staining. In this study, the central zone score was selected as entry criterion and outcome, as the central zone is the most visually significant and has been emphasised as necessary for Delphi grade 3 disease. While there are other cornea grading scales that measure staining over the entire cornea (cumulative zone score), we did not use a total score as adding up all zones has been previously shown to result in the loss of important information.11 For each individual, only one eye was selected for inclusion in the study to avoid bias and over-representation of certain groups of participants such as those with systemic autoimmune disease. The eye with greater initial central staining score was selected as the study eye. In situations where staining was the same in both eyes, the eye with a more discernible change from baseline was used for analysis.

Participants who had less than 6 months of follow-up duration or did not have cornea data recorded at follow-up visits were excluded.

Study procedures

Baseline visit evaluation

All participants underwent a standardised examination and documentation by the same dry eye specialist to exclude key differential diagnoses such as ocular allergies. Key examination findings such as Schirmer’s test, TBUT and corneal epithelial staining grade were recorded.12 Whenever necessary, a standardised transparent overlay was used to isolate the central zone for recording of staining score on the slit lamp biomicroscope.

Cornea staining was chosen as the treatment outcome as previous European multicenter randomised trials (SANSIKA 2016 and 2017)13 14 showed that corneal fluorescein staining was the main objective parameter that showed objective improvement with topical cyclosporine use. In addition, the relationship between symptom severity and clinical signs is inconsistent due to disease variability and subjectivity in symptom reporting;15 staining can be documented more objectively than symptoms.

The detailed methodology in figure 1 shows other data collected at baseline visit, including history of all systemic and rheumatological diseases previously diagnosed by qualified physicians such as rheumatologists and haematologists. The methodology of documentation has been reported elsewhere.16 17

Figure 1

Study selection process and methodology.

Outcome evaluation at last clinic visit

All patients were routinely examined with the slit lamp biomicroscope by trained ophthalmologists and drawings of staining spots were made in medical records. All patients also underwent a brief symptom enquiry on whether symptoms were better, worse or had no change from the second last visit and this was noted down in the clinical notes. In cases where staining was more marked, photographic evidence was performed. Patients’ case notes and electronic medical records were retrieved, and two trained investigators collected data consistently with a standardised data collection form. A significant improvement was defined as central zone Baylor staining score <2 at the last visit.

Statistical analysis

The proportion of the sample with epithelial staining improvement as outcome was calculated and the 95% interval of the binomial proportion calculated using an exact method ( The proportions of participants with each potential baseline risk factor were computed. Normality for each scale variable was explored. For univariate analyses, chi square tests were used to compare whether categorical risk factors affect corneal epithelial improvement (dichotomous variable). We used a TBUT cut-off of 3 seconds in the chi-square tests because it was close to the median and mean of the sample. In binary logistic regression (multivariate analyses), the outcome or dependent variable was the significant improvement of the corneal epithelial status (i.e., central epithelial staining grade <2 at last visit), and the odds-ratio and 95% CI for each independent variable was determined. The level of alpha was set at 0.05 for statistical significance. Microsoft Excel and SPSS for Macintosh V.24 were used in the analyses, including the tests of normality.


Clinical and demographic characteristics of participants

In brief, out of 1712 patients with severe dry eye, 407 eyes from 407 patients with central Baylor staining score ≥2 were studied (figure 1). The clinical and demographic characteristics of the study participants are shown in table 1.

Table 1

Key baseline independent variables (univariate) n=407

The mean age was 56.5±14.6 with majority of patients being of Chinese ethnicity (89.9%, 366/407). Majority were women (80.1% (326/407)) and only 19.9% were men. The test of normality (Shapiro-Wilk) showed that age was not normally distributed, and it was not significantly associated with non-improvement in staining grade (Mann-Whitney U test).

Notably, 17.2% (70/407) had significant autoimmune disease; 8.8% (36/407) had primary Sjogren syndrome (pSS), 9.3% (38/407) had rheumatoid arthritis (RA), 1.0% (4/407) had systemic lupus erythematosus (SLE) and 0.5% (2/407) had ankylosing spondylitis (AS). GVHD was diagnosed in 3.2% (13/407) of patients. A Schirmer’s test of ≤5 mm was observed in 44.7% (182/407) and TBUT <3 s in 58.0% (236/407).

Outcome at the end of follow-up

The mean duration of follow-up was 4.0±2.4 years (median of 43.9 months (IQR 23.1–68.4)), with 88.0% (95% CI: 84.8 to 91.2) (358/407) of patients achieving significant improvement to central Baylor staining score of <2 at last visit (table 2).

Table 2

Change in outcomes at final follow-up visit (univariate) n=407

Out of the 12.0% of patients who did not achieve significant improvement, we noted that all of them had insignificant improvement or the same Baylor staining grade as at baseline visit. No patients had any worsening of central Baylor staining score as compared with baseline (table 2 and online supplementary table S1). Further evaluation of the three patients who had no change in Baylor staining scores revealed that two patients had significant comorbidities that may have contributed to a lack of improvement; one patient had RA and glaucoma with trabeculectomy failure; one patient had GVHD with cytomegalovirus and herpes zoster ocular infections. Interestingly, one patient with persistent central staining grade 4 had no significant comorbidities and subsequently defaulted follow-up after 1 year.

In this study, we evaluated risk factors for the lack of improvement of corneal staining to grade <2 but could not analyse the risk factors for no change in grading versus any improvement of staining grade as the number of patients with no change was too small (3/407) (data not shown).

Effect of autoimmune disease on non-improvement

The presence of autoimmune disease was positively correlated with non-improvement in cornea staining (table 3). The diagnosis of any one of the following four autoimmune diseases (RA, pSS, SLE or AS) was a risk factor for non-improvement in cornea staining grade (OR 3.4, 95% CI: 1.8 to 6.6) (table 4). RA was an independent risk factor for non-improvement in staining grade (OR 3.0, 95% CI: 1.4 to 6.7).

Table 3

Baseline risk factors associated with non-improvement of cornea staining at last follow-up visit (bivariate) n=407

Table 4

Baseline risk factors associated with non-improvement of cornea staining at last follow-up visit (multivariate) n=407

Gender was potentially confounding as chi square analysis showed a greater proportion of autoimmune disease, especially pSS, in female participants (online supplementary table S2). However, the presence of autoimmune disease (RA, SLE, pSS and AS) was still a significant risk factor for non-improvement after adjustment for age and gender (OR 3.4, 95% CI: 1.8 to 6.7); adjustment for Schirmer’s test and TBUT (OR 3.1, 95% CI: 1.5 to 6.3) (table 4).

Of interest, 85.7% (60/70) patients with autoimmune disease were treated with topical cyclosporine/glucocorticoid for some duration in the follow-up period. However, despite these treatments, autoimmune disease was still a risk factor for non-improvement.

Effect of aqueous deficiency on non-improvement

Low Schirmer’s test (OR 2.1, 95% CI: 1.2 to 3.9) and reduced TBUT (OR 2.0, 95% CI: 1.0 to 3.8) were risk factors for non-improvement in staining grade (table 4). However, both were no longer significant after adjustment for autoimmune disease and other clinical factors.

Graft versus host disease as a risk factor for non-improvement

Three percent (13/407) of patients had GVHD. Patients with GVHD were at risk of non-improvement in staining grade (OR 3.4, 95% CI: 1.0 to 11.7) (table 4). Although GVHD was no longer a statistically significant risk factor after adjustment for multiple clinical factors, this could be explained by the small number of patients with GVHD resulting in a lack of statistical power. Unsurprisingly, 84.6% (11/13) with GVHD used topical cyclosporine and/or glucocorticoid in the study period. Out of these, 10 patients used both while one used only cyclosporine. The use of immunosuppressants by GVHD patients was higher than the 56.6% of total sample population who used cyclosporine or glucocorticoid.

Role of glucocorticoid and cyclosporine

A variety of treatment modalities were used, including topical corticosteroids, topical cyclosporine and punctal plugs (online supplementary table S3 and S4). Patients often had more than one modality of treatment. Up to 56.6% (230/407) participants used topical cyclosporine and/or glucocorticoid (online supplementary table S3). There was a positive association between cyclosporine use and non-improvement in staining grade (OR 3.6, 95% CI: 1.8 to 7.1) (table 5). Patients on glucocorticoids were also at risk of non-improvement (OR 2.7, 95% CI: 1.5 to 4.9) (table 5). After adjustment for autoimmune disease status, GVHD and lower Schirmer’s reading, glucocorticoid treatment was no longer significant for non-improvement in staining grade. However, cyclosporine use was still associated with a lack of improvement (OR 2.4, 95% CI: 1.1 to 5.1).

Table 5

Treatment associated with non-improvement of cornea staining at last follow-up visit (multivariate) n=407

Age, gender, MGD, pterygium and ocular surgery were not significantly associated with staining improvement (online supplementary table S5).

Out of 407 participants, we identified 55 participants who were already on glucocorticoid or cyclosporine formulations at baseline visit. We further analysed the data excluding these 55 patients and yielded similar results (online supplementary tables S6–8). The presence of autoimmune disease, GVHD and cyclosporine use remained significant clinical risk factors for non-improvement.

While evaluating symptom change was not a study aim, patients are usually concerned with symptom improvement. We analysed baseline risk factors that may be associated with perceived symptomatic improvement at the last visit (online supplementary table S9). The presence of itch at baseline was associated with a lack of symptomatic improvement by the last visit even after adjusting for MGD, autoimmune disease and gender. Chi square analysis showed a significant association between the presence of itch and a history of systemic allergies (p=0.012, data not shown).

Discussion and conclusion

In this study, we retrospectively identified the proportion of severe dry eye patients who had objective clinical improvement after a medium to long duration of treatment, ranging from 0.5 to 10.3 years. We also identified risk factors for non-improvement despite treatment.

The following risk factors were found to limit improvement in severe dry eye: autoimmune disease (RA, pSS, SLE and AS), GVHD, Schirmer’s test ≤5 mm and TBUT <3 s. In addition, RA and GVHD remained significant predictors for poor response despite the use of immunosuppressant eyedrops.

Lower Schirmer’s test and reduced TBUT at baseline were associated with non-improvement in corneal staining. In aqueous deficient patients, previous studies have found high levels of tear cytokines and matrix metalloproteinase,18 19 which can cause epitheliopathy. This can explain the association between a lower Schirmer’s test, reduced TBUT and non-improvement in staining grade. In the Dry Eye Workshop two recommendations, an abnormal Schirmer’s test is not necessary for diagnosis, but it is recommended for dry eye classification.20 In this study, a low baseline Schirmer’s reading may be a predictor of poor response to treatment.

In this cohort, despite adjustment for systemic autoimmune disease, patients treated with cyclosporine were still at higher risk of non-improvement. A possible reason may be that patients on cyclosporine might have a longer duration of persistent epithelial staining before entering the study. These patients could have more indolent inflammatory processes that were harder to reverse. It is also difficult to ascertain compliance of treatment in a retrospective design. Patients often had a variety of treatment throughout the follow-up period and had varied treatment compliance and duration of treatment. The use of cyclosporine formulations can cause stinging that is sufficient to induce a loss of compliance and corticosteroid treatments could be ceased due to complications. These factors may have confounded the results and thus our results do not suggest a lack of efficacy of these treatments.

There was no association between punctal plug treatment and improvement in corneal staining. Previous studies have shown that punctal plugs had a minimal effect on tear cytokines and MMP-9 levels.19 This may explain the limited effect of punctal plug occlusion especially on autoimmune-related dry eye patients who were included in this study. Our practice in this dry eye clinic is to insert only one punctal plug at the lower punctum as previous studies have shown a lack of additional benefit with dual punctal plug insertion.21 22 It is also difficult to obtain robust data on punctal plug occlusion as patients may not be aware when punctal plugs become dislodged and many patients do not want replacement of dislodged punctal plugs. There is a variable duration of insertion and a variable number of times each eye underwent punctal plug insertion. These factors are likely to have limited the efficacy of punctal plugs in improving corneal epithelial staining outcomes in this study.

This study also found an association between itch and a lack of symptomatic improvement. Chi-square analysis showed an association between itch and systemic allergies. The association between greater symptoms and a known allergy was also previously shown in a British study.23 These patients may have generalised hypersensitivity and occult ocular surface allergy. A previous paper has shown that there was a correlation between allergic symptoms, dry eyes and MGD via the postulated mechanism of inflammation.24 However, it is important to note that the last visit was quite long after the baseline visit and it was not possible for patients to accurately recall whether symptoms had improved in comparison to the initial visit. Symptom improvement over a long period is difficult to study, especially in a retrospective design, since patient recall is highly subjective and likely biassed towards recent times.

The strength of this study is the identification of risk factors associated with persistence of marked central cornea staining in a fairly large cohort of patients with severe dry eye. Another strength was the utilisation of modern treatment options such as topical cyclosporine and steroid formulations. However, further new treatments such as lifitegrast, tacrolimus, azithromycin, diquafosol and rebapimide were not evaluated as these were not available during the study period. If these treatment formulations were available, they could potentially alter the results of our cohort. The use of autologous serum was also not evaluated in the current study.25 Finally, another strength includes the uniform documentation of patients in a single clinic.

A limitation of this study is the lack of analysis of previous systemic treatment, and the duration of pre-existing corneal staining. We did not use composite endpoints of symptom and/or epithelial response, thus it is possible that patients still experience significant symptoms despite a reduction in corneal staining. Hence, an improvement in central corneal staining alone may not be associated with patient satisfaction. We also did not quantify the proportion of patients who may have ceased cyclosporine for various reasons (burning sensation, financial concerns). Our results are not comparable with clinical trials like SANSIKA, as in this observational study, patients are freely treated with immunosuppressants, punctal plugs, artificial tears and viscous gels depending on the practising clinician. The dose of steroids and other drops are also tailored as deemed appropriate by the clinician.

It is interesting to note that majority of patients with GVHD and autoimmune disease were treated with immunosuppressants. However, they still have a higher risk for non-improvement in corneal staining. This suggests that there is a need for more efficacious medications for treatment of severe dry eye, especially in autoimmune disease and GVHD patients.

In conclusion, this study found that after an average of 4 years of treatment in a specialised centre, 88% of patients achieved significant objective improvement in a major sign of DED. Patients with autoimmune disease and GVHD are at risk for non-improvement. Although 87% of autoimmune patients had either prescription medications or punctal plug insertions, not all patients had improved epitheliopathy. This suggests an unmet need for more effective treatment options for dry eye especially in patients with GVHD and autoimmune disease.


We would like to thank Ms Sharon Yeo and Ms Cynthia Boo from the Singapore Eye Research Institute for their administrative assistance throughout this study period.



  • Funding This study was funded by National Medical Research Council Singapore (NMRC/CSA/017/2017).

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Ethics approval The study was approved by the institutional review board of Singapore Health Services and complied with the Tenets of Declaration of Helsinki for human research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.

Linked Articles

  • At a glance
    Keith Barton James Chodosh Jost B Jonas