Association of Dry Eye Symptoms and Signs in Patients with Dry Eye Disease

ABSTRACT Purpose To determine the correlations among symptoms and signs of dry eye disease (DED) in the Dry Eye Assessment and Management (DREAM) study. Methods A total of 535 patients with moderate-to-severe DED were assessed for symptoms using the Ocular Surface Disease Index (OSDI) and four DED signs in both eyes (conjunctival lissamine green staining, corneal fluorescein staining, Schirmer’s testing, and tear break-up time (TBUT)) following standardized protocols at baseline and follow-up visits (months 3, 6, and 12). Spearman correlation coefficients (rho) were calculated for correlations among symptoms and signs of DED at baseline and among changes in symptoms and signs from baseline at 12 months. The confidence intervals and p-values for correlation coefficients were calculated using a cluster bootstrapping to account for inter-eye correlation. Results At baseline, OSDI total score was not correlated with signs; however, OSDI subscale score of ocular symptoms was weakly correlated with corneal staining score (rho = 0.14, p = .002) and Schirmer test score (rho = 0.11, p = .01). There were statistically significant correlations among the four signs (p < .001), with absolute correlation coefficient ranging from 0.14 (conjunctival staining score vs. TBUT) to 0.33 (conjunctival staining score vs. cornea staining score). The correlations among changes in symptoms and signs were weaker, with the highest correlation between change in conjunctival staining and corneal staining (rho = 0.21, p < .001). Conclusions Consistent with previous studies, among DREAM participants with moderate-to-severe DED at baseline, correlations of DED symptoms with signs were low and correlations among four objective signs were low to moderate. The correlations among changes in symptoms and signs were even weaker.


Introduction
Dry eye disease (DED) is a multifactorial disease of the ocular surface characterized by a loss of homeostasis of the tear film accompanied by ocular symptoms. 1Ocular symptoms of DED are non-specific including redness, burning, stinging, foreign body sensation, photophobia, and pruritus. 2Symptoms can lead to a reduction in patients' quality of life such as driving, reading, and computer use along with decreased workplace productivity. 3,4iagnosis of DED is dependent on reports of ocular symptoms and measurement of signs.The quantity and severity of ocular symptoms can be assessed through the use of the Ocular Surface Disease Index (OSDI; Allergan, Inc., Irvine, CA). 5 Assessment for DED signs often includes lissamine green staining of the interpalpebral conjunctiva, corneal fluorescein staining, Schirmer's test, and tear break-up time (TBUT). 1 Several previous studies have reported inconsistent correlations ranging from no correlations to low correlations between DED symptoms and signs, [6][7][8][9][10] while weak-to-moderate correlations among the DED signs. 6,8,10While the discordance between dry eye signs and symptoms is not well understood, one theory attributes the discordance to psychosomatic comorbidities (anxiety, depression, and stress), leading to hyperalgesia and increased perception of clinical pain. 11,12he variations in results for correlations between DED signs and symptoms in previous studies could be due to the differences in study populations, sample size, or the procedures for evaluating DED symptoms or signs.
None of the previous studies evaluated how the change over time of DED symptoms is correlated with change in signs and correlations among changes in signs.Such evaluations may help improve our diagnosis and clinical care of DED, monitor the progression of DED, and provide useful information to select outcome measures for future clinical studies of DED.The Dry Eye and Assessment and Management (DREAM) study provides a large, diverse sample of dry eye patients with moderate-to-severe dry eye who underwent standardized procedures for assessment of signs and symptoms, providing a unique opportunity to evaluate their correlations at baseline and the correlations among their changes at 12 months.

Methods
The details of the design and main results of the DREAM study have been described in previous publications, 13,14 and only the major features of DREAM relevant to this project are described further.
The DREAM study was designed to include a broad spectrum of symptomatic patients with moderate-tosevere DED.From October 2014 through July 2016, 535 eligible patients were enrolled from 27 clinical centers in the United States and randomized 2:1 to an active oral omega-3 fatty acid supplement group or a placebo group (refined olive oil).To be eligible, the participant had at least two of the following four signs in the same eye at both the screening visit and the baseline visit: (1) a conjunctival lissamine green staining score of 1 or more; (2) corneal fluorescein staining score of 4 or more; (3) TBUT of 7 s or less; (4) a result on Schirmer's test with anesthesia of 1-7 mm in 5 min.The trial protocol was approved by the institutional review board associated with each center, carried out under an Investigational New Drug application for the Food and Drug Administration, and registered on ClinicalTrials.gov(NCT02128763).The trial followed the tenets of the Declaration of Helsinki, and written informed consent was obtained from all patients.

Measures for dry eye symptoms and signs
At screening visit, baseline visit, and months 3, 6, and 12, patients were first assessed for dry eye symptoms and then four signs in both eyes following the order of TBUT, corneal fluorescein staining, lissamine green staining, and Schirmer's tear test as described further.

OSDI questionnaire
The patient self-administered a 12-item OSDI questionnaire to assess DED symptoms.The total OSDI score and scores for three subscales (ocular symptoms, visionrelated function, and environmental triggers) were calculated from the responses to each question.Scores on the OSDI range from 0 to 100, with lower score indicating less severity of dry eye symptoms. 5

TBUT
The TBUT measures time (in seconds) from a complete blink to the appearance of the first dry spot (gap) in the tear film, with shorter times indicating more severe dry eye sign.Using a stopwatch, the TBUT was measured three times during the first minute, beginning 30 s after fluorescein instillation.The average of three repeated TBUT measurements was used for analysis.

Corneal fluorescein staining
Corneal staining was graded using the cobalt blue filter of the slit lamp approximately 2.5 min after fluorescein instillation.Using methods standardized by the National Eye Institute (NEI), 15 five areas of the cornea were graded on a scale of 0-3, with a total possible score of 0-15 per eye.Higher score indicates greater severity.

Lissamine green staining of the interpalpebral conjunctiva
One percent lissamine green was examined using white light and instilled into the lower conjunctival sac.After 1-2 min, the nasal-bulbar and temporal-bulbar conjunctiva were graded for staining on a scale of 0-3, with a total possible score of 18, following a modified version of the NEI/industry-recommended guidelines. 15igher score indicates greater severity.

Schirmer's test
Using Schirmer's test II with anesthesia, the length of wetting of paper strips positioned vaulting the lateral third of the lower eyelid margin over a 5-min period was measured in millimeters, with shorter lengths indicating greater severity of sign.

Evaluation of depression
7][18] The score of Mental Component Summary (MCS) of SF-36 has been seen to be an effective screening tool for depression. 17In one study, the SF-36 MCS score ≤42 yielded a sensitivity of 73.7% and a specificity of 80.6% for identifying clinical depression. 17

Statistical analyses
Data from eligible eyes at baseline were used for evaluating correlations among dry eye symptoms and signs, and data for changes from baseline at 12 months were used for evaluating correlations among changes in signs and symptoms.Spearman correlation coefficients (rho) were calculated due to the non-normality of the signs.The 95% confidence intervals (95% CI) for Spearman correlation coefficients and associated p-values were calculated using a cluster bootstrap to account for the inter-eye correlation. 19,20For the correlation analysis among changes in symptoms and signs at 12 months, the randomized treatment group was adjusted by calculating partial correlation coefficients and with sensitivity analysis for correlation analysis stratified by treatment group.As the correlations between DED symptoms and signs may vary with sex, 21 depression status, [22][23][24] and Sjögren syndrome (SS) status, the stratified correlation analyses were performed by sex, baseline depression status, and baseline SS status, and their differences and 95% CI in correlation coefficients were calculated using a cluster bootstrap.All statistical analyses were performed in SAS version 9.4 (SAS Institute Inc., Cary, NC), and twosided p < .05 was considered statistically significant.

Results
Of 535 DREAM participants who completed both screening and baseline visits, 488 (91%) completed 12-month follow-up visit.Baseline characteristics of all patients are displayed in Table 1, and the scores of dry eye symptoms and signs at baseline and 12-month follow-up visit are provided in Supplementary Table S1.

Correlation between dry eye signs and symptoms at baseline
Table 2 shows the Spearman correlation coefficients between DED signs and symptoms at baseline for each of the four signs with the total OSDI score and each subscale score.Their correlations were all minimal (absolute value of Spearman correlation coefficient rho < 0.15).Total OSDI score and subscale scores of vision-related function and environmental triggers were not significantly correlated with each of the four signs (p ≥ .07).However, OSDI subscale score of ocular symptoms was significantly correlated with corneal staining score (rho = 0.14, p = .002)and Schirmer's test score (rho = 0.11, p = .01).
Table 3 shows the results from comparisons of baseline OSDI scores among three groups of participants defined by a number of signs (e.g., 2, 3, and 4) meeting the thresholds used for determining eligibility for each sign.There was no statistically significant difference in the total OSDI score across the three groups (p = .30).However, participants with more signs fulfilling the criteria had significantly higher baseline OSDI subscale score of ocular symptoms (mean scores of 42.5, 45.4, and 48.2, respectively, p = .03,Table 3).
Table 4 shows the Spearman correlation coefficients between changes in signs and changes in symptoms at 12 months.The correlations between their changes were all minimal (absolute value of rho ≤ 0.16).The change in total OSDI score was significantly correlated with the change in conjunctival staining score (rho = 0.10, p = .03)and TBUT rho = −0.16,p < .001)but was not significantly correlated with the change in corneal staining score (rho = 0.05, p = .22).The correlations in OSDI subscales with change in the signs were generally similar to those with the total OSDI.The correlation analysis by treatment group provided similar correlation coefficients (Supplementary Table S2).

Correlations among dry eye signs
There were statistically significant correlations among four DED signs (all p < .001,Table 5), with increasing severity of one sign correlated with increasing severity   Table 6 shows the correlations among change in four DED signs at 12 months.Only change in conjunctival staining score was significantly correlated with change in corneal staining score (rho = 0.21, p < .001)and change in TBUT (rho = −0.12,p = .002).Change in Schirmer's test score was not significantly correlated with change in other signs (p ≥ .22).Similar correlation coefficients were found from analysis stratified by treatment group (Supplementary Table S3).

Correlation between dry eye signs and symptoms and among signs stratified by sex
As shown in Table 7, the correlations between the total OSDI score and DED signs were all minimal in both females (absolute value of rho ranged from 0.01 to 0.06) and males (absolute value of rho ranged from 0.02 to 0.14), and none of their differences were statistically significant.The absolute value of correlation coefficients among DED signs ranged from 0.14 (TBUT vs. conjunctival staining score) to 0.33 (corneal staining score vs. conjunctival staining score) in females and from 0.08 (TBUT vs. conjunctival staining score) to 0.40 (Schirmer's test score vs. corneal staining score) in males.None of their differences were statistically significant.

Correlation between dry eye signs and symptoms and among signs stratified by baseline depression status
As shown in Table 8, the correlations between the total OSDI score and the DED signs were all minimal in both participants with depression (absolute value of rho: 0.02-0.05)and participants without depression (absolute value of rho: 0.01-0.09),and their differences were not statistically significant.
For the correlation among the DED signs, the highest correlation was between corneal staining score and conjunctival staining score in participants without depression (rho = 0.34, [95% CI: 0.26, 0.41]), while the  Note: *Partial Spearman correlation coefficients (rho) were calculated adjusting for treatment group.95% Confidence intervals (95% CI) and p-values were calculated using cluster bootstrap to account for multiple visits per patient and two eyes from the same patient.Note: SS = Sjögren syndrome Sjögren syndrome was defined by the 2012 American College of Rheumatology classification criteria for Sjögren syndrome based on serology and ocular surface stainings.95% Confidence intervals (95% CI) were calculated using cluster bootstrap to account for correlation from two eyes of the same patient.
correlations among DED signs tended to be higher in participants without depression than in participants with depression.None of their differences were statistically significant.

Correlation between dry eye signs and symptoms and among signs stratified by baseline SS status
As shown in Table 9, the correlation between total OSDI score and TBUT was significantly higher in participants with SS (rho = 0.30) than in participants without SS (rho = −0.09).Similarly, the correlations between the total OSDI score and conjunctival staining score were significantly stronger in patients with SS (rho = −0.26)than in patients without SS (rho = 0.02).

Discussion
6][27] Definitive reasons for their weak correlation are unknown.It could be that comorbidities involving neuropathic pain or other chronic pain conditions increase the perception of ocular symptoms in DED patients. 25,27Among DREAM participants with moderate-to-severe DED who were assessed using standardized procedures in a multicenter clinical trial, we found that symptoms, as measured by total OSDI, were not significantly correlated with DED signs at baseline.However, the OSDI subscale and ocular symptoms had significant weak correlations with corneal staining score and Schirmer's test score at baseline.Their correlations did not differ by sex or depression status but tended to be stronger in patients with SS.Understanding the lack of a correlation between signs and symptoms in DED is a continuing conundrum that has escaped clear mechanisms to date.
Evaluation of the correlation among the four standard dry eye signs showed low correlations, with the highest being between corneal staining and conjunctival staining (rho = 0.33) in all participants and was higher in participants with SS (rho = 0.56).These findings were consistent with the findings reported by Sullivan et al. in a study of 263 DED patients, which found highest correlation between corneal staining and conjunctival staining (rho = 0.36). 10atients' mental health disorders can contribute to levels of symptoms of DED that are inconsistent with severity of DED signs. 11,12Zhou et al. found that DREAM participants with depression reported more severe symptoms of DED than patients without depression. 24Labbe et al. found that depression was significantly more prevalent in patients with DED and was significantly correlated to DED symptoms but not correlated with DED signs. 23However, in our analysis, we found that the correlations between DED signs and symptoms tended to be weaker in participants with depression than in participants without depression.Mental health issues, such as depression, may explain some of the discordance between signs and symptoms, but not likely the major cause of these findings.These data do suggest that other factors beyond the DED eye signs are significant for depression.
Better objective minimally invasive metrics that capture biologic changes in the eye might lead to a better assessment of the severity of ocular surface changes in DED and perhaps a better correlation with symptoms. 28owever, other pain syndromes, such as headache and back pain, have also found it hard to find objective metrics that correlate symptoms with disease signs, pointing out the difficulties in measuring and comparing subjective symptoms of a subject over time and in comparing results among subjects and response to treatment.
There may be many reasons that signs and symptoms in DED do not correlate, including that the "objective" signs we use may not be free of bias; for instance, the Schirmer's test has consistently shown low reproducibility, with large differences seen between subjects and significant variations on different days and in different environments. 29,30Symptoms may vary in severity/frequency with season or throughout the day. 31,32For some DED patients, neuropathic pain may be a key issue and unrelated to signs of ocular surface abnormalities 25,27 or maybe the signs we use do not adequately measure surface changes.
Overall, our study had much strength including a large sample size, multicenter study, and standardized procedures for evaluating DED symptoms and signs.We evaluated both their correlation at baseline and correlation among their changes at 12 months.The large sample size allows us to perform stratified correlation analysis by sex, baseline depression status, and baseline SS status.Despite the various strengths of our study, our DREAM study included only patients with moderate-to-severe DED; future research might benefit by including patients with mild DED and those without DED.
In conclusion, this analysis of DED signs and symptoms in DREAM participants suggested that total OSDI did not show correlations with signs at baseline, though the subscale and ocular symptoms showed weak correlations with signs, especially corneal staining.The correlations among the four signs typically measured suggest that they are all measuring ocular surface disease, with the best correlation between corneal and conjunctival stainings.The absent or weak correlations among DED symptoms and signs point to the need for further research, especially, to develop minimally invasive objective metrics to classify DED and determine severity and response to treatment.The results of this study highlight the clinical challenge of appropriately diagnosing DED, classifying DED severity, and monitoring its progression.Further studies are needed to gain a better understanding of the etiopathogenesis of DED and objective ways to measure signs and symptoms.
Note: § Depression was defined as SF-36 mental component summary score ≤42 at baseline.*Sjögren syndrome was defined by the 2012 American College of Rheumatology classification criteria for Sjögren Syndrome based on serology and ocular surface stainings.**Without serology for antibody test for Sjögren syndrome.

Table 2 .
Correlations between dry eye symptoms and signs at baseline.Partial Spearman correlation coefficients (rho) were calculated with adjustment for treatment group.95% Confidence intervals (95% CI) and p-values were calculated using cluster bootstrap to account for two eyes from the same patient.

Table 3 .
Comparison of symptoms across three groups of patients defined by the number of signs fulfilling the criteria at baseline.
Note: *The enrollment criteria required the presence of at least two of the following four signs in the same eye at both the screening visit and the baseline visit:(1) a conjunctival lissamine green staining score of 1 or more; (2) corneal fluorescein staining score of 4 or more; (3) tear break-up time of 7 s or less; (4) a result on Schirmer's test with anesthesia of 1-7 mm in 5 min.The grouping of patients is based on the eye with more signs fulfilling the criteria for this person-level comparison of OSDI scores.**From analysis of variance.

Table 4 .
Correlations between changes in dry eye symptoms and signs at 12 months from baseline (n = 935 eye visits).
Note: * Partial correlation coefficients (rho) are calculated adjusting for treatment group.95% Confidence intervals (95% CI) and p-values are calculated using cluster bootstrap to account for multiple visits per patient and two eyes from same patient.

Table 5 .
Correlations among four dry eye signs at baseline (n = 1022 eyes).Partial Spearman correlation coefficients (rho) were calculated with adjustment for treatment group.95% Confidence intervals (95% CI) and p-values were calculated using cluster bootstrap to account for multiple visits per patient and two eyes from the same patient.

Table 6 .
Correlations among change in dry eye signs at 12 months from baseline (n = 935 eyes).

Table 7 .
The correlation between dry eye symptoms and signs, and among signs for male and female at baseline (n = 1022 eyes).Confidence intervals and p-values were calculated using cluster bootstrap to account for correlation from two eyes of the same patient.

Table 8 .
The correlation between dry eye symptoms and signs, and among signs for participants with and without depression at baseline (n = 1022 eyes).
Note: Depression was defined as SF-36 mental component summary score ≤42 at baseline.95% Confidence intervals (95% CI) were calculated using cluster bootstrap to account for correlation from two eyes of the same patient.

Table 9 .
The correlation between dry eye symptoms and signs, and among signs for participants with and without Sjögren syndrome at baseline (n = 1022 eyes).