Association between serum vitamin D level and rates of structural and functional glaucomatous progression

Abstract

Purpose:

To investigate the association between serum vitamin D level and rates of functional and structural glaucomatous loss over time.

Methods:

This study included 826 eyes of 536 glaucoma or suspect patients with an average follow-up of 4.8 ± 1.9 years. All patients had at least 1 serum vitamin D measurement, and all eyes had at least 2 reliable standard automated perimetry (SAP) tests and 2 spectral domain optical coherence tomography (SD OCT) tests with a minimum follow-up of 6 months. Multivariable linear mixed effects models were used to estimate the association of vitamin D level with rates of change in SAP mean deviation (MD) and OCT retinal nerve fiber layer (RNFL) thickness over time while adjusting for potential confounding factors.

Results:

Patients had an average of 3.4 ± 1.7 SAP tests, 4.8 ± 1.9 SD OCT tests, and 2.3 ± 1.9 vitamin D measurements. Average serum vitamin D level was 33.9 ± 13.2 ng/mL. Mean rates of MD and RNFL change were −0.03 ± 0.08 dB/year and −0.68 ± 0.64 μm/year, respectively. After controlling for confounding factors, there was no statistically significant association between mean vitamin D level and rates of MD (β = 0.038, 95% CI: [−0.006, 0.082], p = 0.09) or RNFL loss over time (β = −0.018, 95% CI: [−0.092, 0.055], p = 0.62).

Conclusions:

We did not find a significant association between vitamin D level and rates of visual field or RNFL loss over time in individuals with glaucoma and glaucoma suspect patients.

PRECIS

In a retrospective cohort study, serum vitamin D levels were not associated with rates of structural or functional loss in glaucoma patients, suggesting that low vitamin D level is not a risk factor for progression.

INTRODUCTION

Glaucoma is a class of chronic neurodegenerative diseases of the optic nerve that is the leading cause of irreversible blindness in the world.^1–4 Glaucoma had an estimated prevalence of 80 million worldwide in 2020, which is projected to increase to 110 million people by 2040, and will disproportionately impact underserved areas of the world, including parts of Africa and Asia.^1,3,4 Further, the prevalence of blindness due to glaucoma has been estimated at 3.6 million globally in 2020.⁵ As such, much effort has been taken to find and understand modifiable risk factors.

The pathophysiology of optic nerve damage and retinal ganglion cell (RGC) loss in glaucoma is not completely understood,² and it is likely that a host of different mechanisms contribute to the neural tissue loss seen in the disease. One proposed mechanism involves direct and indirect oxidative stress damage to the RGCs,⁶ including activation of apoptotic pathways.⁷ Abnormal immune responses and glial cell dysfunction may also lead to increased oxidative stress.⁸ Given its roles as an antioxidant and in modulating inflammation, vitamin D level has been implicated in degenerative diseases of the nervous system, and there has been speculation that it could play a role in glaucoma.^9,10

Previous studies have shown mixed evidence on the potential association between glaucoma and serum vitamin D levels. While some studies found lower serum vitamin D levels in glaucoma patients compared to healthy controls^11,12 others found no statistically significant difference.¹³ Results of studies examining serum vitamin D level as a risk factor for glaucoma also do not show a clear pattern of association.^13,14 These previous studies have used cross sectional or case-control designs.¹⁵ Evidence of the association of serum vitamin D level and longitudinal rates of progression using structural and functional measures is lacking.

In the present study, we analyzed how serum vitamin D levels were associated with rates of glaucomatous structural and functional progression over time in a large cohort of subjects followed under routine clinical care.

METHODS

This was a retrospective cohort study using data extracted from the Duke Glaucoma Registry (DGR), a database of electronic medical records developed by the Vision, Imaging and Performance Laboratory.¹⁶ The database consisted of adults at least 18 years of age with glaucoma or glaucoma suspect diagnoses who were evaluated at the Duke Eye Center or its satellite clinics between 2009 and 2019. The database included medical and medication history, laboratory test values, best-corrected visual acuity, slit-lamp biomicroscopy, intraocular pressure (IOP) measured with Goldmann applanation tonometry (GAT; Haag-Streit, Konig, Switzerland), ophthalmic examination, gonioscopy, stereoscopic optic disc photographs, standard automated perimetry (SAP) tests acquired with the Humphrey Field Analyzer (HFA, versions II and III; Carl Zeiss Meditec, Inc., Dublin, CA), and Spectralis spectral domain optical coherence tomography (SD OCT, Heidelberg Engineering, GmbH, Dossenheim, Germany). Current Procedural Terminology (CPT) and International Classification of Diseases (ICD) codes were also included.

The study was approved by the Duke University Health System’s Institutional Review Board, with a waiver of informed consent due to the retrospective nature of this work. This study was conducted in accordance with the Health Insurance Portability and Accountability act and adhered to the tenets of the Declaration of Helsinki.

Patient selection

Patients were included if they had glaucoma or suspicion of glaucoma based on ICD codes from clinical visits. Eligible patients were required to have at least 1 serum vitamin D level during follow-up, and each included eye was required to have at least 2 reliable SAP tests and 2 reliable peripapillary SD OCT scans during a minimum follow-up period of at least 6 months. All of the SAP, OCT, and serum vitamin D data from 6 months prior to the first available SD OCT scan to 6 months after the last SD OCT scan were used. Tests performed after diagnosis of non-glaucoma pathologies that could have affected SAP or OCT results were excluded: retinal detachment, retinal or malignant choroidal tumors, non-glaucomatous disorders of the optical nerve and visual pathways, uveitis, amblyopia, age-related macular degeneration (wet, atrophic, or late stage), and venous or arterial retinal occlusion. In addition, tests performed after treatment with panretinal photocoagulation or incisional glaucoma surgery (i.e., trabeculectomy or tube shunt implantation), according to CPT codes, were also excluded. ICD and CPT codes used for inclusion and exclusion in the study have been described previously.¹⁶

Visual field tests were conducted with 24-2 and 30-2 pattern Swedish Interactive Threshold Algorithm tests with size III white stimulus. Visual field tests were excluded if they had greater than 33% fixation losses or 15% false positives. SAP mean deviation (MD) was used to assess rates of change in visual fields over time.

Spectralis SD OCT was used to acquire peripapillary retinal nerve fiber layer (RNFL) thickness measurements as described previously.¹⁷ The global average RNFL was calculated as the average thickness of all 768 points distributed equidistantly around the optic nerve head. Only good quality scans, defined as a quality score greater than or equal to 15, were included in the analyses. Since manual review of all tests was impractical, scans with implausible average global RNFL thickness measurements (i.e., less than 20 or greater than 150 μm) were also excluded. These cutoffs represent measurements above the upper range of reported RNFL thickness for healthy controls and below the lower range for glaucoma subjects^18–20 and may indicate the errors in acquisition or segmentation of scans with otherwise good quality scores.²¹ When more than one good quality test was available for the same date, the mean global RNFL thickness of all tests from that date were used in the analysis.

Included eyes were classified into glaucoma or glaucoma suspects based on an objective reference standard previously described,²² applied to the tests at baseline. In brief, the objective criteria for glaucomatous optic neuropathy accounts for the presence and correspondence of structural and functional defects, defined from parameters from SD OCT and SAP. To be considered glaucoma, an eye had to meet the criteria for global loss (i.e., global RNFL thickness from SD OCT outside normal limits and abnormal SAP, as defined by Glaucoma Hemifield Test [GHT] outside normal limits or pattern standard deviation [PSD] with P < 5%); or localized loss (at least one sector in the RNFL thickness outside normal limits with a corresponding abnormality on the opposite SAP hemifield, defined as a hemifield MD with P < 5%). Eyes with normal SAP and SD OCT parameters were considered normal and were not included in the analysis. SDOCT-SAP pairs that did not meet criteria for glaucoma or normal were considered suspects (i.e., only structural or only functional damage).

Data Analyses

Rates of change of SAP MD and global RNFL thickness were estimated using longitudinal linear mixed effects models. While this standard technique has been described in detail elsewhere,²³ in brief, mixed effects models take into account the natural correlation of data over time, as well as the correlation of data from 2 eyes from the same patient. Random slopes and intercepts were introduced to account for differences in rates of change between eyes and subjects. Best linear unbiased predictions (BLUP) were used to estimate individual slopes of change for each eye,²⁴ which have been shown to be more precise than those obtained by ordinary least squares linear regression especially for eyes with a small number of tests over time.^25,26 As the number of tests increase, BLUP estimates converge with the estimates from ordinary least squares regression.

Mean serum vitamin D levels over the study period were calculated for each patient. Separate analyses were performed to evaluate the association of serum vitamin D level and the rate of change in MD and RNFL thickness over time. In multivariable linear mixed effects models, we also controlled for various confounding factors: age at baseline, race, gender, mean IOP, central corneal thickness (CCT), baseline MD or RNFL thickness, time of follow-up, and diagnosis (i.e., glaucoma v. suspect). Subgroup analyses were also conducted separately for glaucoma and suspect eyes, as well as eyes with at least 5 or more OCT or SAP tests. Baseline demographics and clinical characteristics were tabulated from the date of first valid SD OCT.

All statistical analyses were conducted in Stata (version 16, StataCorp LP, College Station, TX) within the Protected Analytics Computing Environment (PACE), a highly protected virtual network space developed by Duke University for analysis of identifiable protected health information. The alpha level (type 1 error) was set at 0.05 for all analyses.

RESULTS

The study included 2773 SAP tests, 3946 SD OCT tests, and 1235 serum vitamin D measurements of 826 eyes from 536 patients. 528 eyes were deemed glaucoma suspects, and 298 were deemed glaucomatous at baseline. The average (± SD) number of SAP tests and SD OCT scans for each eye were 3.4 ± 1.7 and 4.8 ± 1.9, respectively, and each patient had an average of 2.3 ± 1.9 serum vitamin D measurements. The mean time of follow-up was 4.8 ± 1.9 years. The median baseline MD on SAP was −2.2 dB (IQR: [−4.7, −0.9]), and the median baseline RNFL was 81.0 μm (IQR: [72.0, 90.0]). The mean serum vitamin D level throughout follow-up was 33.9 ± 13.2 ng/mL (Figure 1). Detailed demographics and clinical characteristics for the overall cohort as well as subgroups of suspects and glaucoma eyes are shown in Table 1.

Figure 1.

Distributions of mean serum vitamin D levels in the overall study population and the glaucoma and suspect subgroups.

Table 1.

Demographic and clinical characteristics of glaucoma and suspect patients

Variables	Overall	Glaucoma	Suspect
Patient-level characteristics
Number of patients	536	192	344
Age at baseline, years
Mean ± SD	64.3 ± 11.4	66.0 ± 11.1	63.3 ± 11.0
Sex, % female	74.6%	73.4%	75.3%
Race, % African American	32.6%	32.3%	32.8%
Vitamin D
Number of vitamin D measurements
Mean ± SD	2.3 ± 1.9	2.4 ± 2.4	2.3 ± 1.7
Mean vitamin D level, ng/mL
Mean ± SD	33.9 ± 13.2	33.4 ± 14.4	34.2 ± 12.5
Eye-level characteristics
Number of eyes	826	298	528
Follow-up time, years
Mean ± SD	4.8 ± 1.9	4.7 ± 2.0	4.8 ± 1.9
SAP
Number of SAP tests, Mean ± SD	3.4 ± 1.7	3.7 ± 2.0	3.2 ±1.4
Baseline MD, dB
Mean ± SD	−3.7 ± 4.7	−6.4 ± 5.7	−2.2 ± 3.2
Median [IQR]	−2.2 [−4.7, −0.9]	−4.5 [−8.3, −2.5]	−1.5 [−2.8, −0.5]
Slope, dB/year
Mean ± SD	−0.03 ± 0.08	−0.08 ± 0.10	0.00 ± 0.05
Median [IQR]	0.00 ± [−0.04, 0.02]	−0.04 [−0.10, 0.03]	0.01 [−0.01, 0.03]
OCT
Number of OCT scans, Mean ± SD	4.8 ± 1.9	4.8 ± 2.0	4.7 ± 1.8
Baseline RNFL, μm
Mean ± SD	80.4 ± 14.7	66.7 ± 11.1	88.2 ± 10.1
Slope, μm/year
Mean ± SD	−0.68 ± 0.64	−0.71 ± 0.67	−0.66 ± 0.61
Median [IQR]	−0.65 [−1.00, −0.36]	−0.74 [−1.06, −0.43]	−0.60 [−0.95, −0.31]
IOP
Mean IOP, mmHg
Mean ± SD	15.8 ± 3.2	15.4 ± 3.2	16.0 ± 3.1

IOP: intraocular pressure, IQR: interquartile range, MD: mean deviation, OCT: optical coherence tomography, RNFL: retinal nerve fiber layer, SAP: standard automated perimetry, SD: standard deviation

Table 2 shows the results of univariable linear mixed effects regressions investigating the effect of each predictive factor on rates of MD and RNFL change over time. In univariable analyses, mean serum vitamin D level was not associated with rates of change in MD (β = 0.008, 95% CI: [−0.030, 0.046], p = 0.69) or RNFL thickness (β = −0.025, 95% CI: [−0.086, 0.036], p = 0.42). Figure 2 shows scatterplots of mean serum vitamin D level and rates of change in MD and RNFL for the overall cohort as well as for subgroups of glaucoma and suspect eyes. The average rate of change in MD and RNFL over time were −0.06 ± 0.08 dB/year and −0.60 ± 0.61 μm/year, respectively.

Table 2.

Univariable linear mixed-effects regression analyses assessing the effect of various putative factors on rates of change in standard automated perimetry mean deviation and spectral-domain optical coherence tomography retinal nerve fiber layer thickness in the overall cohort.

	Effect on OCT RNFL, μm/year		Effect on SAP MD, dB/year
Parameter	Coefficient	P value	Coefficient	P value
Mean vitamin D, per 10 ng/mL higher	−0.025	0.42	0.008	0.69
Sex, female	0.118	0.21	0.020	0.72
Race, African American	0.056	0.53	0.044	0.41
Baseline Age, per 10 years older	−0.030	0.43	−0.068	<0.01
Diagnosis at baseline, glaucoma	0.050	0.56	−0.105	0.03
Mean IOP, per 1 mmHg higher	−0.039	<0.01	0.008	0.94
CCT, per 40 μm thinner	−0.031	0.50	−0.022	0.45
Follow-up time, per 1 year longer	0.031	0.21	−0.014	0.38
Baseline MD, per 1 dB lower	---	---	0.008	0.24
Baseline RNFL, per 10 μm thinner	0.054	0.07	---	---

CCT: central corneal thickness, IOP: intraocular pressure, MD: mean deviation, OCT: optical coherence tomography, RNFL: retinal nerve fiber layer, SAP: standard automated perimetry

Figure 2.

Scatterplots of the associations between mean serum vitamin D levels and rates of change in standard automated perimetry mean deviation (MD) and spectral-domain optical coherence tomography retinal nerve fiber layer (RNFL) thickness.

We further evaluated the association of mean serum vitamin D levels and rates of MD and RNFL thickness change in our cohort using multivariable analyses, controlling for various potential confounding factors: sex, race, age, diagnosis at baseline, mean IOP, CCT, follow-up time, and baseline MD or RNFL thickness (Table 3). Mean serum vitamin D level was not significantly associated with either the rate of MD change over time (β = 0.038, 95% CI: [−0.006, 0.082], p = 0.09) or rate of RNFL thickness change over time (β = −0.018, 95% CI: [−0.092, 0.055], p = 0.62).

Table 3.

Multivariable linear mixed-effects regression analyses assessing the effect of serum vitamin D level on rates of change in standard automated perimetry mean deviation and spectral-domain optical coherence tomography retinal nerve fiber layer thickness in the overall cohort, controlling for potential confounding factors.

	Effect on OCT RNFL, μm/year		Effect on SAP MD, dB/year
Parameter	Coefficient	P value	Coefficient	P value
Mean vitamin D, per 10 ng/mL higher	−0.018	0.62	0.038	0.09
Sex, female	0.148	0.17	−0.012	0.86
Race, African American	0.094	0.38	0.058	0.37
Baseline Age, per 10 years older	−0.022	0.63	−0.070	0.01
Diagnosis at baseline, glaucoma	−0.070	0.60	−0.090	0.15
Mean IOP, per 1 mmHg higher	−0.049	<0.01	−0.009	0.34
CCT, per 40 μm lower	−0.076	0.13	−0.029	0.33
Follow-up time, per 1 year longer	0.023	0.44	−0.026	0.17
Baseline MD, per 1 dB lower	---	---	0.002	0.78
Baseline RNFL, per 10 μm thinner	0.071	0.12	---	---

CCT: central corneal thickness, IOP: intraocular pressure, MD: mean deviation, OCT: optical coherence tomography, RNFL: retinal nerve fiber layer, SAP: standard automated perimetry

We also performed separate subgroup analyses on groups of glaucoma eyes and suspect eyes to further assess whether any associations of serum vitamin D level and glaucoma progression depended on stage of disease (Table 4, available at http://www.ophthalmologyglaucoma.org). Controlling for the same confounding factors in multivariable analyses, mean serum vitamin D was not significantly associated with rates of change in MD (β = 0.044, 95% CI: [−0.016, 0.104], p = 0.15) or RNFL thickness (β = 0.033, 95% CI: [−0.087, 0.152], p = 0.59) in the glaucoma subgroup. In the suspect subgroup, mean serum vitamin D levels were also not associated with rates of change in MD (β = 0.026, 95% CI: [−0.043, 0.096], p = 0.46) or RNFL thickness (β = −0.032, 95% CI: [−0.125, 0.061], p = 0.50) over time. Finally, we performed multivariable analyses on eyes with at least 5 or more SAP or OCT tests (Table 5, available at http://www.ophthalmologyglaucoma.org). After controlling for confounding variables, mean serum vitamin D levels were also not associated with faster glaucomatous change over time in this subgroup, as measured by SAP MD (β = 0.026, 95% CI: [−0.041, 0.092], p = 0.45) or OCT RNFL thickness (β = −0.042, 95% CI: [−0.125, 0.041], p = 0.32).

Table 4.

Multivariable linear mixed-effects regression analyses assessing the effect of serum vitamin D level on rates of change in standard automated perimetry mean deviation and spectral-domain optical coherence tomography retinal nerve fiber layer thickness in subgroups of glaucoma and suspect eyes, controlling for confounding factors.

	Effect on OCT RNFL, μm/year		Effect on SAP MD, dB/year
Parameter	Coefficient	P value	Coefficient	P value
Glaucoma eyes only (N = 298)
Mean vitamin D, per 10 ng/mL higher	0.033	0.59	0.044	0.15
Sex, female	−0.274	0.16	0.010	0.92
Race, African American	0.284	0.15	−0.085	0.36
Baseline Age, per 10 years older	−0.026	0.76	−0.036	0.38
Mean IOP, per 1 mmHg higher	−0.069	0.02	−0.019	0.19
CCT, per 40 μm lower	−0.176	0.04	−0.000	1.00
Follow-up time, per 1 year longer	−0.003	0.96	−0.015	0.58
Baseline MD, per 1 dB lower	---	---	−0.005	0.53
Baseline RNFL, per 10 μm thinner	0.172	0.03	---	---
Suspect eyes only (N = 528)
Mean vitamin D, per 10 ng/mL higher	−0.032	0.50	0.026	0.46
Sex, female	0.353	<0.01	−0.049	0.60
Race, African American	−0.029	0.82	0.157	0.10
Baseline Age, per 10 years older	−0.034	0.52	−0.086	0.03
Mean IOP, per 1 mmHg higher	−0.034	0.07	0.004	0.81
CCT, per 40 μm lower	−0.003	0.97	−0.056	0.22
Follow-up time, per 1 year longer	0.046	0.18	−0.036	0.17
Baseline MD, per 1 dB lower	---	---	0.009	0.55
Baseline RNFL, per 10 μm thinner	−0.000	1.00	---	---

CCT: central corneal thickness, IOP: intraocular pressure, MD: mean deviation, OCT: optical coherence tomography, RNFL: retinal nerve fiber layer, SAP: standard automated perimetry

Table 5.

Results of the multivariable linear mixed-effects regression analyses assessing the effect of serum vitamin D level on rates of change in standard automated perimetry mean deviation and spectral domain optical coherence tomography retinal nerve fiber layer thickness in eyes with five or more optical coherence tomography and standard automated perimetry tests.

	Effect on OCT RNFL, μm/year N = 425 eyes of 271 patients		Effect on SAP MD, dB/year N = 152 eyes of 104 patients
Parameter	Coefficient	P value	Coefficient	P value
Mean vitamin D, per 10 ng/mL higher	−0.042	0.32	0.026	0.45
Sex, female	0.230	0.06	0.024	0.77
Race, African American	0.017	0.88	−0.004	0.96
Baseline Age, per 10 years older	0.022	0.67	−0.034	0.37
Diagnosis at baseline, glaucoma	−0.078	0.60	−0.130	0.10
Mean IOP, per 1 mmHg higher	−0.037	0.04	0.004	0.78
CCT, per 40 μm lower	−0.153	<0.01	−0.002	0.96
Follow-up time, per 1 year longer	0.030	0.45	0.013	0.70
Baseline MD, per 1 dB lower	---	---	0.007	0.47
Baseline RNFL, per 10 μm thinner	0.034	0.53	---	---

CCT: central corneal thickness, IOP: intraocular pressure, MD: mean deviation, OCT: optical coherence tomography, RNFL: retinal nerve fiber layer, SAP: standard automated perimetry

DISCUSSION

In our study we found no association between mean serum vitamin D levels and rates of progression of SAP MD or global RNFL thickness over time. We found this to be the case in univariable analyses, multivariable analyses controlling for known confounding factors, as well as in separate analyses of subgroups of glaucoma eyes and suspect eyes.

Given the various roles vitamin D plays in human physiology and its implications in other systemic diseases,^9,10 it has been thought of as a potentially useful modifiable risk factor for glaucoma. Previous studies on the association of glaucoma and serum vitamin D levels have found mixed evidence. Some cross sectional case-control studies comparing serum vitamin D levels between glaucoma patients and healthy controls found lower serum vitamin D levels in glaucoma patients.^11,12 One case control study found that while there was a difference in serum vitamin D levels between glaucoma patients and healthy controls, serum vitamin D level did not correlate with the severity of glaucoma as defined by MD.¹² In contrast, another study found that while serum vitamin D levels were lower on average in patients with severe glaucoma compared to early glaucoma, it did not differ between early glaucoma patients and healthy controls.²⁷ And yet, a meta-analysis of three studies found no difference in serum vitamin D level between glaucoma patients and controls.²⁸

In our study design, we addressed several shortcomings of previous studies. Previous works were cross sectional or case-control in nature and either compared serum vitamin D levels between glaucoma and healthy subjects or assessed the risk of having glaucoma based on serum vitamin D levels.¹⁵ A large retrospective cross sectional study of 123,331 South Korean participants found that risk of glaucoma—as determined by clinician grading of a single fundus photo—may be associated with serum vitamin D level in females but not males.¹⁴ Another cross sectional study of 6094 participants reported a J-shape association between serum vitamin D level and risk of glaucoma: they divided patients into 5 quintiles based on serum vitamin D level and found that the 2^nd quintile (i.e., group with the second highest serum vitamin D level) had the lowest risk of glaucoma, compared to the other quintiles, which had progressively increasing risk.¹³ However, only the risk of glaucoma in the 5^th quintile (i.e., lowest level of serum vitamin D, less than 13.08 ng/mL) yielded a statistically significant difference when compared to the 2^nd quintile,¹³ which calls this potential pattern of association into question. One of the limitations to these cross sectional studies as well as several case control studies^11,12,29 is that they lack longitudinal assessments of glaucoma progression. Given that glaucoma is a chronic disease that progresses slowly over years, a study that accounts for longitudinal changes in disease progression would better characterize the potential relationship between serum vitamin D level and glaucoma. Thus, in our study, we included all available data during follow-up from eyes under routine clinical care to evaluate the association of mean serum vitamin D levels and the rates of change in MD and RNFL. Even though our cohort consisted of a large number of eyes with an average follow-up of 4.8 years, we were not able to find a significant association of serum vitamin D levels and rates of glaucomatous change. This suggests that if vitamin D should exert some metabolic protection to glaucoma progression, these changes do not seem to be significantly manifested clinically.

The fact that we assessed progression using both structural and functional metrics is another advantage of our study design. By examining rates of RNFL and MD separately, we were able to assess whether serum vitamin D had an association with either functional or structural progression of glaucoma dependent on the disease stage, given the evidence that eyes may sometimes progress by one metric and not the other³⁰ and the complicated relationship between structural and functional loss in glaucoma.^31,32 In order to further facilitate this examination, we independently assessed these associations in subgroups of suspect eyes and glaucoma eyes, but still found no statistically significant associations in either subgroup.

To define subgroups of glaucoma and suspect eyes, we used a recently developed objective criteria for glaucomatous optic neuropathy based on both visual field and OCT data.²² In contrast, the aforementioned large cross sectional studies defined glaucoma based on clinician grading of a single fundus photo.^13,14 This is problematic given the poor interrater reliability of subjective optic disc gradings that has been widely documented in the literature.^33–35 Others also used RNFL thicknesses from OCT to guide diagnosis of glaucoma, but without visual field data.^12,29 While some others did incorporate both perimetry and structural data in classifying glaucoma patients,^11,27 they were still subjective in nature. By using a clearly outlined objective criteria²² to define glaucoma and suspect eyes in our study, our design is more reproducible and may be free from biases arising from subjective gradings. However, it is possible that some of the eyes may have been classified differently had other proposed classification schemes been used. Nevertheless, given the overall lack of relationship between serum vitamin D level and rates of change in MD and RNFL in the overall group and subgroups, it does not seem likely that this would have made a significant difference in our main outcomes.

The median rate of SAP MD change in our treated glaucomatous population was slow: −0.04 dB/year. Rates ranged from −0.45 to 0.05 dB/year. These rates of change are compatible with rates described for other large clinical populations of glaucoma patients under treatment. In a previous study of 2324 patients, Chauhan and colleagues reported a median rate of only −0.05 dB/year,³⁶ though other studies have reported faster rates of progression as well.³⁷ This underscores the fact that some cohorts of glaucoma patients show slow rates of change under treatment. Importantly, our study likely had enough power to detect a statistically significant association between mean serum vitamin D levels and rates of SAP MD change if one existed in our sample. With 80% power, our sample size would be enough to detect a minimum detectable correlation of r = 0.20, assuming alpha of 0.05. For 90% power, the minimum detectable correlation would be r = 0.23. Regardless, given the slow rate of progression in our cohort, care should be taken when extrapolating our results to populations with faster progression.

Vitamin D deficiency is defined as less than 20 ng/mL and vitamin D insufficiency as between 20 and 30 ng/mL, per the Endocrine Society and the International Osteoporosis Society.^38,39 While the mean serum vitamin D level of our cohort (33.9 ng/mL) was within normal range, 13% of our cohort had vitamin D deficiency and 27% had insufficiency per the above guidelines. When we compared rates of progression in eyes of patients with vitamin D deficiency versus those with normal levels of serum vitamin D, we still did not find any significant difference (results not shown).

It should be noted, however, that since routine serum vitamin D measurements are not part of standard of care for glaucoma or glaucoma suspect patients, patients included in our cohort likely had other reasons for obtaining these measurements. Therefore, our sample may be biased towards inclusion of patients with concurrent diseases that motivated obtaining serum vitamin D levels, such as patients with renal disease and post-menopausal women with osteoporosis or bone mass loss. For example, baseline demographic data in Table 1 shows that about 75% of the patients in our cohort were female, a sex distribution we would not necessarily expect in an average glaucoma or glaucoma suspect cohort. However, it should be noted that previous findings associating low serum vitamin D levels and glaucoma diagnosis were significant in females but not males.¹⁴ Therefore, the unbalanced sex distribution in our cohort would more likely have captured an association of serum vitamin D and glaucomatous progression, given the larger number of female patients. The fact that the patients in our cohort had some reason to have levels of serum vitamin D measured is a potential source of bias, but this also increased the likelihood of having subjects with abnormal levels of vitamin D, which likely would have increased the chance of detecting an association with rates of glaucomatous change if such associations existed. To fully avoid any confounding effects of this potential bias, a prospective design should be utilized in future studies in which serum vitamin D measurements are taken routinely of all patients seen in glaucoma clinic. It should also be noted that our analyses did not directly address whether supplementations of vitamin D can be a protective factor for glaucoma progression. While supplementations would likely be reflected in the measured serum vitamin D levels, prospective randomized controlled trials would be better suited to directly answer such question.

In our study, we included all eyes with at least 2 SAP and OCT tests during a minimum follow-up of 6 months. It may be argued that eyes with only a few tests would not yield accurate rates of change. However, exclusion of such eyes leads to removal of useful data that can still be used to evaluate the associations at the population level. Importantly, removal of such data can potentially bias the analysis of association between covariates and change over time. For example, it could lead to inappropriate exclusion of some eyes with fast progression that occurred over a short period of time. In such a situation, it is important to use appropriate statistical methodology that takes into account the fact that slopes from eyes with few tests are relatively imprecise. This can be handled by linear mixed models and BLUPs, where estimates of rates of change “borrow strength from the population”, so that eyes with few tests provide less contribution to the overall estimated effect of the covariate, but no data are wasted.^25,26 Regardless, we also assessed the association of serum vitamin D and rates of glaucomatous change in subsets of eyes with 5 or more SAP or OCT tests and found similar results to the overall cohort. Of note, it is important to acknowledge that given the retrospective nature of our study, it was not possible to control for type, frequency, and adherence to glaucoma treatment. However, as the effects of treatment (or lack of it) should ultimately be reflected on the measured rates of change, this would be unlikely to have influenced our results, unless there exist yet uncovered relationships between patterns of glaucoma care and vitamin D levels.

In our study, we used all available vitamin D serum measurements that were acquired within the follow-up period, as defined based on the first and last OCT test, and took their average to represent a patient’s overall vitamin D serum level. This design has inherent limitations as it is not possible to guarantee a temporal association between vitamin D levels and subsequent changes in structural and functional tests. This could only be assessed by a prospective design including periodic measurements of vitamin D levels and structural and functional glaucoma testing. As another limitation, the use of average levels of vitamin D ignores potentially wide fluctuations that may be seen seasonally in such levels. However, unless there would be reasons to believe that patients with progressive glaucoma would have measurements taken at different seasons compared to those without progression, there is little reason to believe this should have affected the association as assessed in our study.

While we classified eyes in our cohort using objective criteria previously published,²² the initial detection of cases included all subjects in our cohort with ICD codes for glaucoma or glaucoma suspects. This could have resulted in some eyes with glaucoma or glaucoma suspects not having been included in our cohort due to miscoding or misdiagnosis. Another relevant question is whether glaucoma should be diagnosed by eye or by patient and whether test measurements should be averaged for each patient, when looking at the association with a systemic measurement such as vitamin D level. While serum vitamin D measurements and demographics represent data at the subject level, rates of MD and RNFL thickness progression as well as other eye-specific covariates (e.g., CCT, IOP) are data at the eye level. In addition, because a patient can have one glaucomatous and one suspect eye, we classified glaucoma at the eye level and not patient level. Doing so avoided having to average rates of progression and other measurements at the subject level. Instead, we appropriately controlled for the inter-eye correlation in our linear model. This also meant that a patient’s serum vitamin D values were included twice in the model if both eyes of that patient were included in our cohort. However, when we picked one eye (with more severe disease) per patient and re-ran the analyses, the conclusions were unchanged. In univariable analyses, mean serum vitamin D was not associated with RNFL progression (β = 0.000; p = 0.99) or MD progression (β = 0.012, p = 0.65). In multivariable analyses, mean serum vitamin D was also not associated with RNFL progression (β = 0.004, p = 0.93) or with MD progression (β = 0.035, p = 0.36).

It is interesting to note that while we found that mean IOP was significantly associated with rates of OCT change, we were not able to find a statistically significant association between mean IOP and rates of SAP MD loss. We believe that the explanation is likely related to the fact that rates of change in SAP may be confounded by the subjective nature of perimetry, as well as by nonlinearities in translating retinal ganglion cell (RGC) loss to visual sensitivity thresholds.³¹ Despite this, the lack of association between mean IOP and rates of SAP MD loss is a limitation of our paper and may affect the generalizability of our findings.

In conclusion, we were not able to find a statistically significant association between serum vitamin D levels and rates of functional or structural loss in glaucoma. Our findings suggest that serum vitamin D level may not be a useful modifiable factor influencing clinical outcomes in the disease.