Effects of duration and number of symptoms on vision-related anxiety in patients with Inherited Retinal Diseases

ABSTRACT Background Patients with Inherited Retinal Diseases (IRDs) are at increased risk for vision-related anxiety due to progressive and irreversible vision loss, yet little is known about risk factors for anxiety in these patients. Materials and Methods This was a single-center, retrospective cross-sectional study at a large academic center. 128 adults with an IRD and without other significant eye conditions were recruited between December 2016 and March 2020. Participants were asked about the duration and number of symptoms they had in the following vision domains: reading, contrast vision, color vision, glare/light sensitivity, night vision, and peripheral vision. The outcomes of interest were the two domains of the Michigan Vision-Related Anxiety Questionnaire (MVAQ), rod- and cone-function related anxiety. We conducted an adjusted analysis to isolate the independent effect of duration and number of symptoms on vision-related anxiety. Results Of 126 participants had complete data, 62 (49%) were female and 64 (51%) were male, with an average age of 49 years (range: 18–87). Patients with duration of symptoms for greater than 25 years had an adjusted anxiety theta that was one-half standard deviations lower than patients with symptoms for less time. Patients with higher number of symptoms had higher anxiety theta after adjusting for confounding variables (p < 0.0001). Conclusions The number of symptoms but not the duration of symptoms, is an independent risk factor for vision-related anxiety. Patients with more symptoms are at higher risk for vision-related anxiety. Having symptoms for longer than 25 years may reduce this anxiety. KEY POINTS Question: How does the duration and number of symptoms that patients with Inherited Retinal Diseases have affect their vision-related anxiety? Findings: In this cross-sectional study of 126 patients with Inherited Retinal Diseases, the number of symptoms, but not the duration of symptoms, was associated with higher vision-related anxiety. Patients with symptoms for longer than 25 years had less vision-related anxiety. Meaning: Patients with more vision-related symptoms may experience more vision-related anxiety.


Introduction
Inherited Retinal Diseases (IRDs) are a group of genetic conditions that cause progressive, irreversible vision loss (1). Patients with IRDs are at increased risk for anxiety and depression due to their visual impairment (2)(3)(4)(5)(6). There are two known risk factors for vision-related anxiety and depression: worse general health (7) and worse vision, defined by selfreported visual impairment (8), visual acuity (9), or visual field size (10), depending on the disease. Identifying additional risk factors for higher vision-related anxiety may improve screening practices and referrals to targeted psychosocial interventions for patients who may benefit.
The Michigan Vision-Related Anxiety Questionnaire (MVAQ) and Michigan Retinal Degeneration Questionnaire (MRDQ) are two validated patient-reported outcomes (PRO) measures in patients with IRDs (11,12). The MVAQ and MRDQ provide insight into vision-related disability and anxiety that otherwise cannot measured by physical examination or testing.
Using complete data from 126 patients with IRDs who completed both the MVAQ and MRDQ, we conducted an epidemiologic analysis of how the duration and number of symptoms relates to vision-related anxiety.

Methods
This was a retrospective, cross-sectional study of patients with a clinical diagnosis of IRD. Data were previously obtained as part of the study that developed, tested, and validated the MRDQ (12,13). The study was approved by the University of Michigan Institutional Review Board (HUM00115127) and was performed in accordance with the Declaration of Helsinki. Participants had to be proficient in English, over 18 years old, without other eye conditions significantly affecting their vision, and able to provide informed consent, and were recruited during routine clinical visits to the University of Michigan Kellogg Eye Center's Retinal Dystrophy Clinic in Ann Arbor, Michigan, USA, between December 2016 and March 2020. The questionnaires were verbally administered. The MRDQ and MVAQ were developed using item response theory, factor analysis, and graded response models. The MRDQ had 59 items in seven symptom domains: central vision, color vision, contrast sensitivity, scotopic function, photopic peripheral vision, mesopic peripheral vision, and photosensitivity (12). The MVAQ had 14 items in two domains: rod-function related anxiety, which included items like worrying about uneven ground at night, and cone-function related anxiety, with items such as worrying when reading up close or about going out on bright days (11).

Study variables
The exposures of interest were the duration and number of vision symptoms. Participants were asked how long they had trouble, in months or years, with symptoms in six domains: reading (a surrogate for central vision), color vision, contrast, night vision (surrogate for scotopic function), peripheral vision (surrogate for photopic and mesopic peripheral vision), and light sensitivity/glare (surrogate for photosensitivity). If a participant denied ever-having symptoms, zero months were recorded.
The outcomes of interest were rod-and cone-function related anxiety, as measured by MVAQ rod and cone anxiety theta. Theta scores are mean-centered at zero and follow a normal distribution, where one unit represents one standard deviation (SD) from the mean. Higher scores represent greater anxiety.
Potential covariates were demographic (age, sex, race, ethnicity), vision-related (visual acuity, visual field size, MRDQ domain disability scores), psychiatric (PHQ-4 score, GAD-2, taking psychiatric medication), and IRD phenotype-specific (rod-cone, cone/cone-rod, macular). Age was calculated based on the study enrollment date. Sex, race, and ethnicity were determined by self-report. Visual acuity was measured on the Snellen chart and converted to the Logarithm of the Minimum Angle of Resolution (logMAR) continuous scale (min: −0.30, max: 1.00; 20/20 Snellen vision is 0, 20/200 Snellen vision is 1, higher logMAR indicates worse vision) (14). Monocular Goldmann visual fields (GVF) were not part of the study but were instead obtained through retrospective chart review within one year of the PRO administration date. As a result, 27 patients were missing visual fields. GVF size was quantified by previously published methodology using Adobe Photoshop (Adobe systems, Mountain View, CA, USA) (15). For each eye, the area of the GVF obtained using a size III4e stimulus was measured in mm (2). The eye with the larger visual field was considered the better eye.
The PHQ-4 instrument is a validated screening measure for depression and anxiety composed of the PHQ-2 screen for depression and the GAD-2 screen for anxiety (16). Scores of three or more on the PHQ-2 and GAD-2 are positive screens for clinical depression and anxiety, respectively (16), and a GAD-2 positive screen has 86% sensitivity and 83% specificity for generalized anxiety disorder (17). GAD-2 positive screen was only reported in the cohort characteristics; otherwise, PHQ4 score was used as a continuous measure of current anxiety and depression levels. Chart review was used to determine whether patients were taking any psychiatric medication, not just anxiolytic, at the time of enrollment, in order to do sensitivity analyses for including these patients in the analysis. Finally, IRD phenotype was based on clinical history, fundus examination, optical coherence tomography, and electroretinography, and classified as either macular, rod-cone, or cone/ cone-rod, per previously established methodology.

Statistical analysis
All analyses were conducted using SAS software version 9.4 (SAS Institute; Cary, NC). Continuous data were summarized with median values and interquartile ranges because many variables were skewed. Categorical data were summarized as frequencies with percentages. Group comparisons for continuous, normally distributed variables (MRDQ and MVAQ thetas) were assessed with t-tests and analysis of variance (ANOVA) tests. Comparisons for continuous, non-normally distributed variables (all other continuous variables) were assessed with Wilcoxon nonparametric t-tests and Kruskal-Wallis nonparametric ANOVAs. Chi-squared tests were done for group comparisons of categorical variables except when cell counts were less than five (as for race), then Fisher's exact test was used.
Confounders and effect modifiers were determined using a comprehensive literature review. Potential confounders of the relationship between number/duration of symptoms and cone anxiety theta were age, sex, race, PHQ-4, visual acuity, number of cone symptoms, cone dystrophy phenotype, and MRDQ central vision, photosensitivity, contrast, and color vision thetas. Potential confounders for the relationship with rod anxiety theta were age, sex, race, PHQ-4, GVF size, MRDQ scotopic, and mesopic function thetas, number of rod symptoms, and rod/rod-cone dystrophy phenotype. Because of the study's catchment area, there were few non-white participants, so race was adjusted for as white/non-white.
We first looked at mean values of each covariate by quartiles of the outcome of interest to look for non-linear relationships. Linear relationships were summarized with correlation coefficients and p values. A Pearson correlation coefficient was used for normally distributed variables and a Spearman correlation coefficient was used for non-normally distributed variables.
To reconcile the difference between the MRDQ's vision symptom domains (reading, color, contrast, glare/light sensitivity, night vision, and peripheral vision) and the MVAQ's two anxiety scores (cone and rod anxiety theta), we ran leastsquared means models of cone and rod anxiety theta for patients with and without difficulty in each symptom domain. Difficulty was defined as having symptoms for any non-zero period of time. To isolate the independent effect of each anxiety theta, models were adjusted for the opposite anxiety theta (i.e., the model of cone anxiety theta was adjusted for levels of rod anxiety theta, and vice versa). Based on these results, for the purpose of analyzing MVAQ cone and rod anxiety thetas, reading, color, contrast, and glare/light-sensitivity were considered "predominantly-cone" symptoms, and night and peripheral vision were considered "predominantly-rod" symptoms.
We expected effect modification of the relationship between number/duration of symptoms and cone anxiety theta by visual acuity and MRDQ central vision disability, so we stratified the analysis of cone anxiety theta by MRDQ central vision theta. Similarly, we expected effect modification for the rod anxiety theta analysis by visual field size and MRDQ scotopic function disability, so we stratified by MRDQ scotopic function theta. To stratify, we split the two MRDQ thetas at the median, zero. MRDQ theta greater than zero (more disability) was considered poor central vision/scotopic function, and vice versa. MRDQ scotopic function theta was chosen over visual field size because it had complete data and because it was measured at the same time as rod anxiety theta (rather than within a year of survey administration like the visual field data), and MRDQ central vision theta was chosen for consistency. Both were highly correlated with their counterpart options (correlation between MRDQ central theta and visual acuity: 0.73, between MRDQ scotopic function theta and visual field size: −0.66).
We ran correlations between the seven MRDQ domain theta variables to determine the most parsimonious models. Only one of each MRDQ domain theta that was highly correlated with others (r > 0.7) was selected. As a result, only three of the seven MRDQ domains (central vision, photosensitivity, and scotopic function) were included in the models.
The relationship between duration of symptoms and anxiety theta was non-linear so we generated quartiles of duration of cone symptoms, applied the same cutoffs to duration of rod symptoms, and ran least-squares means (LS means) models of cone and rod anxiety theta. To obtain the most parsimonious models, each potential covariate was added to the model separately and its impact on the exposure-outcome relationship was assessed. Variables were included in the final model if they substantially changed (approximately 10%) the LS means estimate of the fourth quartile, since the most extreme values were expected to be in this last quartile and thus a 10% change in that estimate was the most sensitive for confounding.
The relationship between number of symptoms and anxiety theta was linear so we applied linear regression using the same method of adjusting for confounding discussed previously (potential confounders were added individually to the models; all variables that changed the effect of the exposure variable by approximately 10% or more were included in the final models). The model assumptions of linearity, independence, homoscedasticity, and normality were all met. LS means models adjusting for the same confounding variables as the linear regression models were used to generate the figure.

Results
126 patients had complete data for the exposures and outcomes of interest. 29 had cone/cone-rod dystrophy, 68 had rod-cone dystrophy, and 29 had macular dystrophy (Table S1). Consistent with disease characteristics, patients with rod-cone dystrophy had better visual acuity (mean = 0.18 logMAR; Snellen 20/30) than patients with cone/cone-rod and macular dystrophy (mean for both = 0.30; Snellen 20/40), but worse visual field size (5 cm 2 in the better eye versus 115 cm 2 for cone/cone-rod dystrophy, 176 cm 2 for macular). Patients with rod-cone dystrophy also had less central vision and photosensitivity disability, reflected by lower MRDQ thetas, but had more scotopic function disability.
Patients with poor central vision were older and had more disability in MRDQ for photosensitivity and scotopic function compared to those with good central vision (Table S2). Patients with poor scotopic function had higher PHQ-4 scores, worse visual acuity, smaller GVF size, and more disability in central vision and photosensitivity compared to those with good scotopic function.
Age, sex, and race did not correlate with cone and rod anxiety theta ( Table 1). As expected, worse visual acuity was associated with higher cone and rod anxiety theta, and smaller visual field size was associated with higher rod, but not cone, anxiety theta. More disability (higher MRDQ theta) in central vision, photosensitivity, and scotopic function was significantly Table 1. Degree and significance of unadjusted associations between covariates and anxiety thetas. Spearman r is provided for the non-normally distributed variables (age, PHQ-4, visual acuity, and visual field). Pearson r is provided for the normally distributed variables (MRDQ central theta, MRDQ photosensitivity theta, and MRDQ scotopic function theta). The correlations between cone and rod anxiety theta are also provided (Pearson r, normally distributed). associated with higher cone and rod anxiety theta. MRDQ central vision theta had the highest correlation with cone anxiety theta (r = 0.51; p < 0.0001) and MRDQ scotopic function theta had the highest correlation with rod anxiety theta (r = 0.71; p < 0.0001). MRDQ central vision theta was highly correlated with MRDQ contrast and color vision thetas (r = 0.76 and 0.68, respectively) and MRDQ scotopic function theta was highly correlated with MRDQ mesopic and photopic function theta (r = 0.96 and 0.86, respectively) (Table S3). MRDQ photosensitivity theta was not highly correlated with any others.

Differentiating predominantly cone and rod symptoms
Cone and rod anxiety theta correlated with each other (r = 0.56; p < 0.0001). Patients with difficulty in reading, color vision, contrast, and glare/light-sensitivity had higher mean cone anxiety theta levels than those who did not have difficulty and no difference in rod anxiety theta, when adjusting for the opposite anxiety theta (Table S4). Patients with difficulty with night and peripheral vision had no significant difference in mean cone anxiety theta, and higher mean rod anxiety theta, than those without difficulty. Most patients had both "cone symptoms" and "rod symptoms" regardless of their dystrophy phenotype (Table S1).

Effect of duration of symptoms on vision-related anxiety
The relationships between duration of rod and cone symptoms and rod and cone anxiety thetas were non-linear ( Figure 1, Table 2). After adjusting for confounding, there was no significant difference in anxiety for patients who had symptoms for 0-25 years (quartiles 1-3). However, having symptoms for longer than 25 years (quartile 4) was associated with a 0.5 SD lower anxiety theta (0.47 SD for cone anxiety theta, 0.50 SD for rod anxiety theta). Further, when stratifying by central vision disability for cone anxiety theta, and by scotopic function for rod anxiety theta, patients with good central vision or scotopic function had cone and rod anxiety thetas, respectively, that were on average one SD lower than those with poor central vision or scotopic function, across all quartiles of duration.

Effect of number of symptoms on vision-related anxiety
Each additional cone symptom resulted in a 0.45 unit increase in cone anxiety theta unadjusted and a 0.29 unit increase after adjusting for confounding (p < 0.0001) (Table 3, Figure 2). This persisted after stratifying by MRDQ central vision. Each additional rod symptom resulted in a 0.65 unit increase in rod anxiety theta unadjusted (p < 0.0001). This persisted after adjusting for confounding by PHQ4 and rod dystrophy phenotype, but this effect was no longer significant after adjusting for MRDQ scotopic function theta. The analysis was limited by there being no patients with poor scotopic function who had zero rod symptoms. Patients with good scotopic function had lower rod anxiety theta than those with poor scotopic function and there was a linear increase in rod anxiety theta with additional rod symptoms, but this did not reach statistical significance (beta = 0.22, p = 0.08).

Discussion
This cross-sectional study of 126 patients with IRDs identified number of symptoms as an independent risk factor for higher vision-related anxiety. Additionally, this study found that the duration of symptoms for longer than 25 years was independently associated with less anxiety. Consistent with prior research, worse visual acuity and visual fields were associated with higher vision-related anxiety (8)(9)(10).
Our classification of reading, color vision, contrast sensitivity, and photosensitivity as predominantly-cone symptoms, and night vision and peripheral vision as predominantly-rod symptoms, was consistent with previous research that patients with cone-predominant dystrophy tend to have more central vision, photosensitivity, contrast, and color vision symptoms (18,19), while those with rod-predominant dystrophy have more night blindness and peripheral vision loss, until the disease encroaches the macula (20). Most IRDs involve both rods and cones and yield both rod and cone symptoms, but grouping the symptoms in this way allowed for a more specific analysis of cone-and rod-function-related anxiety theta.
Our study had a few limitations, the first being sample size, which limited our power. Likely due to this, we were not able to detect expected differences in anxiety by age, sex, and race. Sample size also limited our analysis of the effect of number of rod symptoms on rod anxiety theta, since there were no patients who had poor scotopic function and zero rod symptoms in our cohort. This limited our power to detect a difference in rod anxiety between patients with symptoms in only one domain versus both domains. Still, the adjusted results among patients with good scotopic function suggested that the relationship between number of symptoms and higher anxiety may be true for rod-function-related anxiety as well. Further research with a larger sample size could help clarify this question. We also did not have visual fields for all patients, so we did not adjust for GVF size in the models. Instead, we used MRDQ scotopic function theta, which was highly correlated with GVF size. Additionally, we did not collect complete psychiatric history, but our sensitivity analyses of patients currently taking psychiatric medication suggested that including these patients in our analysis did not change our results. Since we did not have specific information about anxiety diagnoses or medications, so we used PHQ-4 score to adjust for general anxiety levels instead. Finally, this was a cross-sectional study done with a one-time survey, so we are assuming that the relationships observed between participants parallel what we would see if we followed the same individuals over time. Future studies using repeat surveys can better assess this relationship longitudinally.
Despite these limitations, this study had many strengths. IRDs are rare and this was a relatively large sample size of 126 patients with complete data. Also, few PRO tools for any ophthalmologic conditions ask about duration and number of symptoms. As a result, we were able to adjust for confounding and effect modification and isolate the independent effects of duration and number of symptoms.
We did not find a statistically significant effect of number of rod symptoms on rod anxiety after adjusting for MRDQ scotopic function theta. This was likely because MRDQ scotopic function theta was highly correlated with rod anxiety theta (r = 0.71, Table 1). While some of this may be sample size limitation as discussed previously, if this is a true effect, a possible clinical explanation for this is that because fewer symptom domains related to rod anxiety (two: night vision and peripheral vision), compared to cone anxiety (four: reading, color vision, contrast, and glare/light sensitivity), rod-related anxiety may be more related to the degree of vision loss rather than whether they have symptoms in just one of, or both, domains.
There is no research on the prevalence of anxiety in patients with IRDs, but patients with Retinitis Pigmentosa (one type of IRD) have fivefold higher odds of anxiety and depression compared to the general population (7). In addition to higher risk for clinical anxiety disorders, many more visually impaired patients experience subclinical anxiety. In a study of visually impaired adults aged 60 years and older, 40% had subclinical or clinical anxiety (8). Screening, identifying, and referring IRD patients with vision-related anxiety to appropriate therapies may beneficially impact their mental health, perceptions of their vision condition, and their clinical trajectory (5,10,21). There are many different types of psychotherapy available, including both individual and group cognitive behavioral therapy (CBT), psychodynamic therapy, problemsolving treatment, expressive writing therapy, behavioral activation therapy, humanistic therapy, and family psychotherapy (21). Research on depression in visually impaired patients resulted in more recognition of the need for screening and referral to psychotherapy interventions specific to their ophthalmologic conditions (21,22). A similar approach is needed for vision-related anxiety.  Effect of duration of symptoms on vision-related anxiety. Least-squares adjusted mean models of cone and rod anxiety theta, with standard error in parentheses.  Table 3. Adjusted multivariate linear regression results for the effect of number of cone and rod symptoms on cone and rod anxiety theta, respectively. Linear regression models for the effect of number of cone and rod symptoms on cone and rod anxiety theta, respectively, stratified by central vision and scotopic function, respectively. For the unstratified models (All Patients), two adjusted models are reported, one excluding the relevant effect modifier (MRDQ central vision theta for cone; MRDQ scotopic function theta for rod) and one including it. There were no patients with poor scotopic function and zero rod symptoms. All patients with poor scotopic function had at least one symptom.

Effect of Number of Cone Symptoms on Cone Anxiety Theta
In conclusion, this study found that patients with IRDs may experience more vision-related anxiety from having more symptoms and that patients with symptoms for over 25 years may experience less vision-related anxiety. To date though, few studies of vision psychotherapy have included patients with IRDs and there is no IRD-specific psychotherapy program despite the need. Future research should evaluate the effects of specific screening practices and interventions for vision-related anxiety in patients with IRDs. It should also study the effects of duration and number of symptoms in patients with other ophthalmologic conditions and the relationship between subclinical visionrelated anxiety and clinical anxiety disorders. Clinicians who evaluate patients with IRDs may consider the number and duration of symptoms patients have in assessing risk for vision-related anxiety and referral to psychotherapy services.