Dry eye symptom severity and visual field reliability metrics

Abstract

Purpose:

Symptoms of dry eye disease are commonly encountered in glaucoma patients and can be exacerbated by topical glaucoma medications. Dry eye disease may influence the reliability of visual field (VF) tests and impact accurate interpretation of the results.

Patients and Methods:

Patients at the Veterans Administration Medical Center San Diego completed the five-item dry eye questionnaire (DEQ5) prior to VF testing between December 2018 and February 2019. VF reliability metrics were recorded for each patient. Standard reliability metrics included fixation losses, false positive, and false negative rates. Gate tracking (GT) metrics included percent of stimuli with gaze deviations between 1 and 2 degrees, 3–5 degrees, 6 degrees or greater, and percent of stimuli with tracking failure (TFF or Tracking Failure Frequency). The use of glaucoma medications and artificial tears was also recorded.

Results:

494 patients completed the DEQ5 and VF testing. There was no association between dry eye symptom severity and standard reliability metrics or most GT metrics. However, TFF increased as dry eye symptom severity increased (p=0.015). TFF was also greater in the left eye, which was tested second (p=0.012); no other reliability metrics were related to laterality. Patients were more likely to use artificial tears with increased dry eye symptom severity (p<0.001), but there was no relationship between symptom severity and glaucoma medication use.

Discussion:

Dry eye symptom severity may influence the acceptable range or threshold of TFF when using GT metrics to determine VF reliability. Likewise, the acceptable range or threshold for TFF may be different between eyes.

Precis:

Tracking failure frequency increases with dry eye symptom severity and in the left eye.

Introduction

Symptoms of dry eye disease are commonly encountered in eye clinics.^1,2 Dry eye disease can have a significant impact on quality of life and activities of daily living.^2,3 Patients may have particular difficulties with tasks requiring extended periods of concentration, such as reading.⁴ This may be because dry eye disease can have significant impact on the optical quality of affected patients.⁵ Common visual complaints include fluctuating vision, blurred vision, glare, and eye fatigue.⁶ Dry eye signs and symptoms are particularly common in patients using glaucoma medications.^2,7–10 These patients often are exposed to the preservative benzalkonium chloride (BAK).¹¹ However, dry eye signs and symptoms are common even in patients using preservative-free medications, so the drugs themselves likely contribute to the disease process.¹²

Glaucoma patients undergo frequent visual field (VF) testing to monitor their disease process, and poorer concentration is associated with more variable testing.¹³ Unreliable VFs tend to be more variable and this makes diagnosis and monitoring of glaucoma more difficult.¹⁴ Standard reliability metrics on the Humphrey Field Analyzer (HFA) include fixation losses (FL), false positive (FP) responses, and false negative (FN) responses. Standard reliability metrics are augmented by gaze tracking (GT) metrics, which monitor gaze deviations and tracking failures during stimulus presentation.^15–17 Several small studies (<50 patients) have shown that dry eye signs and symptoms may influence standard reliability and GT metrics.^18,19 In this study we examine the relationship between dry eye symptom severity, use of glaucoma medications and artificial tears, and standard reliability and GT metrics in a large sample of glaucoma patients.

Materials and Methods

The study protocol was approved by the VA San Diego Medical Center Institutional Review Board and adhered to the tenets of the Declaration of Helsinki. A waiver of consent was obtained to review retrospective VF data.

Study Participants

All glaucoma, glaucoma suspect, and ocular hypertensive patients that underwent VF testing in VA ophthalmology and optometry clinics between December 2018 and February 2019 were given the opportunity to respond to the five-item dry eye questionnaire (DEQ5) prior to testing. The DEQ5 was selected for this study because the questions focus on dry eye symptoms without consideration of visual function. Because many of the patients in the study have glaucoma, there was concern that questionnaires that consider visual function, such as the Ocular Surface Disease Index, could have artifactually elevated scores for reasons other than dry eye.

Patients were included in the study if they completed the DEQ5 on the day of VF testing, were at least 18 years old, had an HFA (Carl Zeiss Meditech Inc, Dublin, CA, USA) 24-2 SITA Fast VF performed on both eyes, and had previously performed at least one VF. There were no exclusion criteria for poor reliability or degree of field loss. Data including patient age and gender, mean deviation (MD), number of glaucoma medications, and use of artificial tears were recorded for all included patients. VF loss was classified as minimal (MD>−6), moderate (−6≥MD>−12), or advanced (−12≥MD). Dry eye symptom severity with was classified as minimal (DEQ5<6), moderate (6≤DEQ5<12), or severe (DEQ5≥12).²⁰

Visual Field Reliability

Standard HFA reliability metrics, including FL, FP, and FN, were recorded for all patients. GT metrics were also recorded from the HFA VF test reports; they quantify any gaze deviation detected during each stimulus presentation. Upward bars represent gaze deviations and increasing bar sizes indicate larger gaze deviations. Downward bars represent failures to capture a signal. VFs were exported as JPEG images from Zeiss Forum (Carl Zeiss Meditech Inc, Dublin, CA, USA) and GT metrics were extracted using ImageJ (ImageJ 1.8.0, National Institutes of Health, Bethesda, MA, USA). Similar to prior gaze tracking studies, GT metrics were calculated as the percent of stimuli with gaze deviations between 1 and 2 degrees (M1), 3–5 degrees (M3), 6 degrees or greater (M6), and percent of stimuli with tracking failure (TFF or Tracking Failure Frequency) (Figure 1).¹⁷

Figure 1:

Gaze tracking parameters.

Statistical Analysis

Statistical analyses were performed using SPSS version 27.0 (SPSS, Inc., Chicago, IL) and R version 3.3.1 (R Foundation for Statistical Computing, Vienna, Austria). Reliability metrics were binarized to reject VF reliability when they exceeded a set threshold. Standard reliability metrics (FL and FP) were binarized according to prior manufacturer guidelines with rejection for FL>20% and FP>15%.²¹ Although there are no manufacturer thresholds for FN, we rejected FN>20% based on prior study limits.²² GT metrics were binarized according to previously published 95% thresholds for rejection (M1 66.7%, M3 67.5%, M6 49.5%, and TFF 79.8%).¹⁷ The statistical significance of relationships between DEQ5 severity and drop use and binarized reliability metrics was determined via chi-square tests, with McNemar’s test used to assess agreement within each variable between eyes.

Results

A total of 628 patients underwent VF testing over the study period. Of those patients, 494 completed the DEQ5 and met the remaining inclusion criteria. Patient characteristics, drop use, and dry eye symptom severity are presented in Table 1. Patients had a mean age of 66.3 years and were largely male (93.9%), as is typical for VA-based studies. A majority of patients were not using any glaucoma medications (65.4%) or artificial tears (75.3%). 53 of the patients using glaucoma medications (31%) were also using artificial tears. Minimal and moderate dry eye symptom severity were more common than severe dry eye symptoms (77.5% vs 22.5%).

Table 1:

Patient demographics, drop use, and dry eye symptom severity.

	N = 494 patients
Demographics	Mean (standard deviation)
Age (years)	66.3 (11.3)
Sex (% female)	6.1%
Glaucoma Medications	Total (percent)
0	323 (65.4%)
1	87 (17.7%)
2	38 (7.7%)
3	32 (6.5%)
4	14 (2.8%)
Artificial Tears	Total (percent)
Using artificial tears	122 (24.7%)
Dry Eye Symptom Severity	Total (percent)
Minimal	185 (37.4%)
Moderate	198 (40.1%)
Severe	111 (22.5%)

The proportion of patients using artificial tears increased as dry eye symptom severity increased (minimal 11.4%, moderate 26.3%, severe 45%; p<0.001). There was no association between dry eye symptom severity and the number of medications used (p=0.626). We also examined the total number of bottles of medications used because some patients were on combination medications (dorzolamide/timolol or brimonidine/timolol). There also was no association between dry eye symptom severity and the number of bottles of medications used (p=0.545).

VF characteristics, including reliability metrics, are presented in Table 2. A majority of patients had minimal VF loss (84.7%). There was not a significant difference in VF severity between eyes (p = 0.442). Based on the previously described binarized limits, 327 eyes (33%) were rejected for high FL, 35 eyes (3.5%) for high FP, and 31 eyes (3.1%) for high FN. There was not a significant difference in rejection rates between left and right eyes for standard reliability metrics (Figure 2, FL not shown due to high rejection rate). The 95^th percentile limit for GT metrics was determined using this study population, so 5% of total eyes were rejected for exceeding the determined thresholds. There was no difference in rejection rates between left and right eyes for M1, M3, and M6. However, left eyes were more likely than right eyes to be rejected for TFF (6.9% vs 4.4%, p=0.012) (Figure 2).

Table 2:

Visual field characteristics.

	N = 988 eyes
Visual Field Loss Severity	Total (percent)
Minimal	837 (84.7%)
Moderate	98 (9.9%)
Severe	53 (5.4%)
Standard Reliability Metrics	Mean (standard deviation)
FL (percent)	16.9 (0.22)
FP (percent)	3.83 (7.19)
FN (percent)	4.59 (7.21)
Gaze Tracking Metrics	Mean (standard deviation)
M1 (percent)	28.6 (20.7)
M3 (percent)	24.6 (21.8)
M6 (percent)	10.5 (17.3)
TFF (percent)	17.0 (27.9)

Figure 2:

Between eye comparison of rejection rates for standard reliability and gaze tracking metrics (FL not shown).

The association between DE symptom severity and reliability metrics is presented in Table 3. There was no association between dry eye symptom severity and standard reliability metrics. Likewise, there was no association between dry eye symptom severity and M1, M3 and M6. However, there was an association between dry eye symptom severity and TFF, with TFF increasing as dry eye symptom severity increased (Figure 3). As previously noted, the proportion of TFF rejection was greater in the left eye, but the association between dry eye symptom severity and TFF held whether considering each eye individually or as a pair.

Table 3:

Dry eye severity by reliability metrics one or both eyes

		Dry Eye Symptom Severity
		Mild (n = 185)	Moderate (n = 198)	Severe (n = 111)	p-value
FL > 20%	One Eye Only	52 (28.1%)	57 (28.8%)	26 (23.4%)	0.704
	Both Eyes	27 (14.6%)	34 (17.2%)	22 (19.8%)
FN > 20%	One Eye Only	10 (5.4%)	5 (2.5%)	4 (3.6%)	0.292
	Both Eyes	1 (0.5%)	2 (1.0%)	3 (2.7%)
FP > 15%	One Eye Only	8 (4.3%)	11 (5.6%)	4 (3.6%)	0.914
	Both Eyes	2 (1.1%)	3 (1.5%)	1 (0.9%)
M1 > 95th Percentile	One Eye Only	23 (12.4%)	16 (8.1%)	7 (6.3%)	0.127
	Both Eyes	2 (1.1%)	0 (0.0%)	0 (0.0%)
M3 > 95th Percentile	One Eye Only	23 (12.4%)	15 (7.6%)	9 (8.1%)	0.469
	Both Eyes	1 (0.5%)	1 (0.5%)	0 (0.0%)
M6 > 95th Percentile	One Eye Only	15 (8.1%)	21 (10.6%)	8 (7.2%)	0.335
	Both Eyes	2 (1.1%)	0 (0.0%)	0 (0.0%)
TFF > 95th Percentile	One Eye Only	9 (4.9%)	17 (8.6%)	15 (13.5%)	0.015
	Both Eyes	0 (0.0%)	2 (1.0%)	3 (2.7%)

Figure 3:

Dry eye symptom severity and tracking failure frequency in both eyes

Discussion

This study examined the relationship between dry eye symptom severity, artificial tear and medicated drop use, and VF reliability metrics in a large sample of VA patients. We found that TFF was the only reliability metric to be influenced by dry eye symptom severity; TFF increased with greater dry eye symptom severity. TFF indicates failure of the gaze tracker in measuring gaze direction and can occur due to blinking or poor corneal light reflex.²³ Patients with dry eye disease have increased blink rates and unstable tear films, which may contribute to increased TFF in patients with more dry eye symptoms.^24,25 This suggests that providers should be cautious when using TFF to determine VF reliability in patients with moderate and severe dry eye symptoms. Dry eye symptom severity may be a significant confounder when using TFF to determine VF reliability, much as FN rates can be confounded by VF severity.²⁶

We also noted an increase in TFF in the left eye, which has not been noted in prior studies of GT metrics. All patients in this study had their right eyes tested first. The difference between left and right eyes may be due to an increase in blink frequency, as prior studies have demonstrated an increase in blink frequency in the second eye tested during VF examination.²⁷ Alternatively, tasks requiring concentration decrease tear film stability, so the second eye tested may have greater rates of failure of gaze tracking acquisition due to a poor corneal light reflex.²⁸ Finally, patients both with and without dry eye report eyestrain, blurred vision, and dry eyes after looking at a computer monitor for an extended period, an activity analogous to VF testing.²⁹ No other reliability metrics were influenced by laterality in this study. This suggests that although a single reliability threshold or range can be used for both eyes for most reliability metrics, for TFF any threshold or range will need to be based on the order of testing.

Two prior studies have examined the relationship between signs and symptoms of dry eye disease and reliability metrics.^18,19 One study showed an association between FN and both signs and symptoms of dry eye disease but did not examine GT metrics.¹⁸ This association was not seen in our study or that of Arai et al. The second study included GT metrics and showed an association between ocular surface measurements and TFF, as in this study, but showed additional associations with M3 and M6.¹⁹ However, the study population was significantly smaller than ours, did not use binarized rejection cutoffs, and examined only signs of dry eye, which may not correlate with symptoms.² Several studies have shown improvement in VF indices and standard reliability metrics in glaucoma patients with dry eye disease after using lubricant eye drops.^30–32 However, none of these studies used patients without dry eye as controls. In our sample, there was no difference in standard reliability metrics between patients with varying levels of dry eye symptom severity, so it is possible that all patients will have improvement in standard reliability metrics following instillation of lubricant eye drops.

Patients with greater dry eye symptom severity were more likely to be using artificial tears in our study. This association is likely stronger in our study than the general population because artificial tears are provided to patients by the VA pharmacy rather than requiring over the counter purchase. Surprisingly, neither the number of glaucoma medications nor bottles used were associated with dry eye symptom severity. Many prior studies have shown dry eye signs and symptoms increase with the number of medications used.^2,7–10 However, many of these studies included only glaucoma patients and excluded patients with history of intraocular surgery. Our study population may differ from prior studies because a majority of the patients were not on any drops. Moreover, we did not exclude patients with history of glaucoma surgery, which may increase dry eye symptoms despite decreasing the number of drops used.³³

This study has several limitations. First, the study sample was drawn from VA hospital outpatient veterans, so the patients enrolled were overwhelmingly male and may not be representative of other populations. Second, this study assessed only symptoms of dry eye disease. Although our study examined only subjective dry eye symptoms, the DEQ-5 has been validated in the discrimination of dry eye severity.³⁴ Furthermore, subjective symptoms and objective signs of dry eye disease may not correlate directly with one another.^2,35 Third, we attribute the increase in TFF in the left eye to the order of testing, but cannot rule out this increase being related to laterality. Finally, potential confounders such as age or VF severity may influence reliability metrics. However, our prior studies on this population showed that age did not influence reliability metrics and VF severity only influenced the rate of FN.¹⁷

In summary, this study demonstrates that only TFF is associated with dry eye symptom severity. Additionally, we show that TFF increases in the left eye, the second eye tested. These factors should be considered when assessing TFF as a marker of VF reliability.