Multiple Systemic Vascular Risk Factors Are Associated With Low-Tension Glaucoma

Précis: Multiple systemic vascular-associated conditions including systemic hypertension and hypotension, diabetes mellitus, migraine headache, peripheral vascular disease, Raynaud syndrome, and anemia were associated with low-tension glaucoma. Purpose: The purpose of this study was to identify systemic risk factors associated with low-tension glaucoma. Patients and Methods: A retrospective case-control study design was employed to identify patients seen at the Mayo Clinic Department of Ophthalmology between 2005 and 2015 with low-tension glaucoma and an age-matched and sex-matched control group, each containing 277 patients. Results: The low-tension glaucoma group had more myopic refractive errors (−1.6 vs. −1.0 D, P<0.001), lower intraocular pressure (14.2 vs. 15.2 mm Hg, P<0.001), and a higher cup-to-disc ratio (0.7 vs. 0.3, P<0.001). The low-tension glaucoma group was significantly less likely to be obese (body mass index >30, P=0.03). This group had a significantly higher prevalence of systemic hypertension [odds ratio (OR): 1.64, P=0.004], diabetes mellitus (OR: 3.01, P<0.001), peripheral vascular disease (OR: 2.61, P=0.009), migraine headache (OR: 2.12, P=0.02), anemia (OR: 2.18, P=0.003), systemic hypotension (OR: 4.43, P<0.001), Raynaud syndrome (OR: 3.09, P=0.05), and angiotensin-converting enzyme inhibitor (OR: 1.64, P=0.01) or calcium channel blocker use (OR: 1.98, P=0.004). After adjusting for systemic hypertension, calcium channel blocker use remained significant (OR: 1.70, P=0.03). No significant difference was found between groups with respect to hyperlipidemia, obstructive sleep apnea, coronary artery disease, carotid stenosis, stroke, or statin, angiotensin-converting enzyme inhibitor, angiotensin receptor blocker, beta-blocker, or metformin use. Conclusions: Multiple vascular-associated conditions were associated with low-tension glaucoma including systemic hypertension, diabetes mellitus, peripheral vascular disease, migraine headache, Raynaud syndrome, anemia, systemic hypotension, and calcium channel blocker use. This study strengthens the evidence for the vascular hypothesis of low-tension glaucoma.

G laucomatous optic neuropathy (GON) is a progressive, characteristic optic neuropathy that leads to retinal ganglion cell loss which may progress to visual field loss when not adequately treated. GON is currently the second leading cause of overall blindness, second only to cataract, and the leading cause of irreversible blindness worldwide. 1 The most significant risk factor and only reliable therapeutic target for GON is intraocular pressure (IOP). Other known risk factors include, age, family history, race, central corneal thickness, and myopia. 2,3 Though IOP is the principle driving force for the progression of GON, and available treatments aim to reduce IOP by pharmacologic, laser, or surgical intervention, nearly one third of patients have evidence of GON despite never having elevated IOP (defined as > 21 mm Hg), and are termed normal or low-tension glaucoma (LTG). Furthermore, in some populations, LTG is the prevailing type of open-angle glaucoma. 4 This suggests that there may be other variables, risk factors, or mechanisms of cellular injury in the initiation and/or progression of GON other than IOP alone. Genetic predisposition, inflammatory and immune responses, structural and mechanical stress, mitochondrial dysfunction and oxidative stress, and vascular dysfunction may modulate the pressure-associated perceived stress of the retinal ganglion cell or may even act in an IOP-independent mechanism. 5,6 Of these, one of the leading proposed mechanisms of retinal ganglion cell injury, particularly in LTG, is based on the vascular hypothesis.
The vascular hypothesis of GON is fairly wellestablished. It suggests that GON occurs in part due to diminished perfusion of the optic nerve by the peripapillary microcirculation, leading to retinal ganglion cell stress and ultimately cell death and atrophy. 7,8 Perfusion of the optic nerve can be estimated as ocular perfusion pressure (OPP), which is typically estimated as a function of IOP and mean arterial pressure such that either higher IOP or lower mean arterial pressure leads to diminished perfusion of the nerve. 9 Decreased OPP has been shown to be associated with GON, and chronic optic nerve ischemia leads to axonal loss and optic disc excavation consistent with GON. 10 Multiple vascular factors can influence OPP including systemic blood pressure, blood pressure lability, vascular resistance, cardiovascular health, serum lipid levels, atherosclerosis, and vascular inflammation. 7,8,11,12 Furthermore, low blood pressure, particularly nocturnal hypotension, has been shown to have associations with GON and visual field loss. [13][14][15][16] Finally, peripheral vascular regulation, measured by nailfold capillaroscopy, has also been associated with low blood pressures in glaucoma patients suggesting that blood pressure and vascular dysregulation affect OPP. 11 However, some of the previously described risk factors are controversial with respect to their effect on GON, and other risk factors have not been thoroughly studied. For example, systemic hypertension and diabetes mellitus have been shown to be associated with primary open-angle glaucoma (POAG) [17][18][19][20][21][22][23][24][25][26][27][28] and LTG, 29,30 while other studies did not find a positive association. [31][32][33] Other diseases such as obstructive sleep apnea (OSA), 34,35 migraines, 36,37 Raynaud syndrome, 16 dyslipidemia, 30 stroke, 30 and coronary artery disease 30 have been studied, but the evidence is less conclusive. Finally, other conditions like anemia, which would be expected to affect oxygen and nutrient delivery to the optic nerve, has not been definitively associated with GON previously. Therefore, conditions that affect systemic and peripheral blood flow and alter optic nerve perfusion are likely additional risk factors for the initiation and/or progression of GON but knowledge gaps exist.
Systemic medications used to treat conditions that may affect tissue perfusion are potential modifiers of OPP and, therefore, and may be confounders when studying conditions such as systemic hypertension or diabetes mellitus and have led to confusion in the medical literature. For example, calcium channel blocker use has been shown to have a protective effect in LTG progression, 38,39 while other studies have shown a negative effect with POAG. 40,41 Systemic beta-blockers have been associated with a higher frequency of disc hemorrhages 42 as well as progression in LTG patients. 43,44 However, systemic beta-blocker use has also been shown to be protective in POAG. 40,41 Much of the data regarding systemic hypertension therapy is confounded by multiple factors at play that could affect OPP including systemic hypertension, systemic hypotension, and IOP. For example, certain vascular risk factors may occur in concert with one another and be grouped into different phenotypes that predispose patients to LTG. Flammer syndrome has been described as a constellation of symptoms resulting from primary vasculature dysregulation that may predispose patients to glaucoma. Characteristics of Flammer syndrome more commonly occur in women and include nocturnal hypotension, low body mass index (BMI), Raynaud syndrome, migraine headache, and decreased blood flow with cold provocation. 45 This may represent a distinct phenotype in LTG patients in contrast to patients with other vascular risk factors such as systemic hypertension, diabetes mellitus, atherosclerotic vascular disease, and OSA who are more commonly older, male, and have a higher BMI. To better define vascular risk factors, including systemic medication use, we selected patients with LTG that show GON without a history of high IOP for further study.

Data Source and Participants
This is a retrospective case-control review of patients seen at Mayo Clinic in Rochester, MN, diagnosed with LTG between 2005 and 2015. The study was approved by the Mayo Clinic Institutional Review Board. Two groups of patients were enrolled: a study group of LTG patients, defined as those with an office visit with a LTG diagnosis code (ICD-9 code 365.12), and an age-matched and sex-matched control group without glaucoma or other optic neuropathy. All patients were seen at Mayo Clinic Department of Ophthalmology in Rochester between January 1, 2005, and December 31, 2015. Inclusion criteria for the study group included a visit billingcode diagnosis of LTG and age at the visit of 40 years or older. Patients were excluded if not meeting inclusion criteria, not having a comprehensive medical visit on record since 2005, having a history of nonglaucomatous type optic neuropathy, or having any form of glaucoma other than LTG. Inclusion criteria for the control group was an age at visit of at least 40 years old with the visit billing-code diagnosis for refractive error and refraction. Exclusion criteria were the same as the LTG group plus no prior diagnosis of LTG. The 2 populations were identified and enrolled using Advanced Cohort Explorer software to retrospectively review all visits in Ophthalmology at Mayo Clinic in Rochester, MN, between 2005 and 2015.
All data was collected by reviewing the complete medical record. Baseline patient characteristics included age, sex, race or ethnicity, visit date, and BMI. Medical history was obtained by searching the medical record for diagnoses of systemic hypertension, dyslipidemia or hyperlipidemia, diabetes mellitus, OSA, cerebrovascular event or stroke, carotid stenosis, peripheral vascular disease, migraine headaches, lupus, Raynaud syndrome, anemia, and systemic hypotension or syncope. Medication history including the use of statins, angiotensin-converting enzyme (ACE) inhibitors, angiotensin receptor blockers (ARBs), beta-blockers, calcium channel blockers, thiazide diuretics, and metformin was recorded. In addition, a history of alcohol use and tobacco smoking was recorded. For the purposes of this study, we focused on LTG as a systemic disease, and ophthalmic findings recorded at the visit date were recorded in each eye separately and averaged between the 2 eyes into a single value for study. Since we incorporated a control population, we selected metrics that could be obtained from a general eye examination that included IOP (in mm Hg) measured by applanation or rebound tonometry (Icare), cup-to-disc ratio, refractive error (in diopters of spherical equivalence), central corneal thickness (in microns), and documentation of the presence of a disc hemorrhage measured by the clinical examiner at the single visit date.
Patients with LTG were further classified into 2 separate phenotypes that were based on metabolic phenotypes that tend to be clustered in a single patient. Phenotype 1 was defined as LTG patients with risk factors that comprise or are associated with metabolic syndrome: systemic hypertension, diabetes mellitus, peripheral vascular disease, coronary artery disease, or OSA, 6,46 without any characteristics of phenotype 2. Phenotype 2 was defined as LTG patients with Raynaud syndrome, migraine headache, anemia, or systemic hypotension without any of the criteria from phenotype 1. If a patient had any condition from both phenotypes, they were excluded from this subgroup analysis.

Data Collection
A total of 370 unique patients were identified with a diagnosis of LTG and meeting inclusion criteria for the study group. Of those, 93 patients were excluded. A total of 277 patients were enrolled into the study group. A total of 19,523 potential control patients meeting inclusion criteria were identified. Of these, 277 control patients were randomly selected in a 1:1 fashion by matching to the study group by sex, age ( ± 5 y), and within 1 year of visit date (Fig. 1). Initial matches were selected using a SAS function to select potential controls based on the GREEDY algorithm. For the algorithm, a distance formula is calculated based on how close the potential control is to the case, and the potential control with the closest distance to the case is selected for the study. Further evaluation of the controls was used to determine if they should be excluded for other reasons. A control patient was replaced by a new, matched control from the eligible population if exclusion criteria were met to keep a total of 277 patients in each group.

Statistical Analyses
Categorical variables were presented as percentages, and continuous variables were described a mean and standard deviation (SD). Overall comparisons of continuous variables were initially compared with 2-sample t tests. In addition, odds ratios (ORs) and results from the logistic regression models were presented. These logistic regression models were used to further investigate the factors of interest while adjusting for potential confounding factors based on their statistical significance and relation to one another. After identifying important univariate relationships between the study and control groups, medication use was further studied in the logistic regression model after adjusting for systemic hypertension to remove the confounding relationship between systemic hypertension and antihypertensive medications. A P-value <0.05 was considered significant. Analysis was completed using SAS, version 9.4 (SAS Institute Inc., Cary, NC).

Demographics and Inclusion
Of the initial 370 patients identified as candidates for inclusion in the study, a total of 93 patients were initially excluded from the study leaving 277 patients remaining in each group. The most common reason for exclusion was the lack of a general medical examination since 2005 (n = 62). Other reasons for exclusion were the lack of a matched control (n = 17), another form of glaucoma (n = 7), or a coexisting non-GON (n = 7) (Fig. 1). There was an equal number of males and females in each group with females representing 57.4% of the population. There was no significant difference in age between the 2 groups with the average age of visit within the study and control groups being 69.5 and 68.9 years old, respectively (P = 0.6). Most of the patients self-identified as white in the LTG and control group (90.3% and 93.5%, respectively), and there was no significant difference between white and nonwhite (P = 0.8) or other ethnicities between the 2 groups (Table 1).

Ocular Metrics
In patients with LTG, there was no significant difference between the 2 eyes in IOP (P = 0.7), refractive error (P = 0.9), central corneal thickness (P = 0.9), or cup-to-disc ratio (P = 0.4) suggesting a bilateral phenotype in most patients. When comparing LTG patients to controls, the mean cup-to-disc ratio was significantly higher (0.7 vs. 0.3, P < 0.001), confirming the presence of GON. Mean IOP was lower in the LTG group than the control group (14.2 vs. 15.2 mm Hg, P < 0.001). Mean refractive error revealed patients with LTG were more myopic than the control group (−1.60 vs. −0.10 D, P < 0.001). In addition, patients with LTG trended toward thinner corneas compared with controls (545 vs. 585 µm, P = 0.08). Finally, a total of 21 patients in the LTG group had disc hemorrhages, while none of the controls had disc hemorrhages (P < 0.001) ( Table 2).

Systemic Metrics
Patients with LTG did not have a significantly different mean BMI compared with controls (28.1 vs. 28.6, P = 0.3). However, when subdividing BMI into categories of healthy weight (BMI = 18.5 to 25), overweight (BMI = 25 to 30), or obese (BMI > 30), patients with LTG were more likely to have a healthy weight BMI, while control patients were more likely to be obese (OR: 1.69, P = 0.02) ( Table 3) Table 3).

Medication Use
Patients with LTG were more likely to be taking ACE inhibitors (OR: 1.64) or calcium channel blockers (OR: 1.98) compared with controls. There was no difference in the use of statins (P = 0.8), ARBs (P = 0.4), thiazides (P = 0.7), betablockers (P = 0.5), or metformin (P = 0.5) between the 2 groups (Tables 2, 4). Since many of the patients on these medications may also have systemic hypertension, we controlled for systemic hypertension. After controlling for systemic hypertension, calcium channel blocker use remained significantly associated with LTG (OR: 1.70, P = 0.03), while ACE inhibitor use no longer showed an association (P = 0.2). There remained no significant difference for ARB (P = 0.1), beta-blocker (P = 0.7), or thiazide use (P = 0.09) ( Table 5).

DISCUSSION
Since GON is likely a multifactorial disease with a number of systemic risk factors including those that may affect a patient's hemodynamics and ultimately their OPP, we identified patients with LTG and performed a case-control study with age-matched and sex-matched nonglaucomatous    The higher cup-to-disc ratio in patients with LTG compared with controls provided validation of the study population. Furthermore, disc hemorrhages were observed in 8% of LTG patients, while no disc hemorrhages were documented in the control population. LTG patients had a lower average IOP than the general population, though many of these patients had received IOP-lowering treatment. This finding provides support to the hypothesis of IOP-independent or indirectly IOP-dependent mechanisms of retinal ganglion cell injury in LTG. Consistent with previous literature, our LTG patients had a more myopic refractive error and trended toward a thinner central corneal thickness. 47,48 LTG has been studied to have a higher prevalence in Asian populations. 4 Though our specific patient populations did not show a significant difference in race between LTG and control patients, the majority of our study population self-identified as white.
The vascular hypothesis of GON suggests that systemic vascular dysregulation leads to retinal ganglion cell loss and may play a more prominent role in LTG compared with other types of open-angle GON. Systemic vascular dysregulation and lower end-organ perfusion are more likely in patients with systemic vascular comorbidities such as high or low blood pressures, anemia, cardiac disease, or stroke. Prior studies have investigated risk factors in ocular hypertension and GON with varying results. An association between LTG and systemic hypotension and nocturnal blood pressure fluctuations have been previously described. 13,16,29,49 Other studies have described an association between systemic hypertension, LTG, and optic disc hemorrhages. 18,22,23,30 From a blood pressure standpoint, our study supports the parabolic nature of risk with both systemic hypertension and hypotension being risk factors for LTG. Coronary artery disease and stroke have been reported to have an association with glaucoma, 12,18,50 however, our study did not find an association with either condition. Some studies have shown an association with dyslipidemia and LTG, 30 while others have not. 22,23 Our study adds to the list of LTG studies that did not show an association with dyslipidemia.
Other risk factors in our study included Raynaud syndrome, migraine headache, peripheral vascular disease, and anemia. These conditions may reduce perfusion of the optic nerve by directly decreasing OPP, secondary to vasospasm or diminished oxygen-carrying capacity of the blood, that may ultimately lead to retinal ganglion cell stress, atrophy, and death. 7,8 Raynaud syndrome is caused by peripheral vasospasm which can lead to dysregulation of peripheral perfusion. It has been suggested to be an independent risk factor for the development of LTG. 16,51 Our study also found a significant association between LTG and Raynaud syndrome. Migraines are also thought to be precipitated by a vasospastic mechanism, which could alter OPP, and previous reports suggest that migraine headaches could be a risk factor for the development and progression of glaucoma. 36,37 Our study supports this claim by finding a significant association between migraine headaches and LTG. A previous study described an association between peripheral vascular disease and GON, 52 but prior evidence supporting an association with LTG and peripheral vascular disease is limited. 53 Our study supports this association between peripheral vascular disease and LTG. Anemia, by decreasing the oxygen-carrying capacity of the blood, mechanistically would presumably be a vascular risk factor for LTG by reducing optic nerve oxygen delivery. Our study found a significant association between anemia and LTG, and to our knowledge, this is the first study to show an association between GON and anemia.
Systemic medication use and the risk of GON is a topic of recent interest 54 and the literature has conflicting evidence of GON risk with multiple medications. For example, calcium channel blockers have been associated with a higher  risk of developing POAG 40,41 but have also been shown to have a protective role, slowing visual field progression in LTG patients. 38,39 After adjusting for systemic hypertension, our study found calcium channel blockers use to be significantly associated with LTG. The use of ACE inhibitors and ARBs in the setting of LTG is less well-studied. Previous literature has suggested a protective role of ACE inhibitor and ARBs use with glaucoma progression, 54 while others found no association with ACE inhibitors or ARBs. 40,41 After adjusting for systemic hypertension, no significant associations were found for ACE inhibitors, ARBs, or thiazide use. Previous reports have shown nonsignificant reductions of glaucoma development with systemic beta-blocker use. 40,41 Others have reported that LTG patients treated with topical beta-blockers are more likely to progress and have lower OPP. 44,55 Systemic beta-blockers alter systemic vascular regulation but also play a role in aqueous production and IOP reduction, which may make finding associations between beta-blockers and LTG difficult. Our study found no significant difference in the prevalence of systemic beta-blocker use and LTG. In addition to vascular dysregulation, systemic inflammation and metabolic stress may lead to disrupted axonal transport and retinal ganglion cell death. 5,8 Diseases such as diabetes mellitus, obesity, and OSA can lead to glucose intolerance, metabolic dysfunction, and a heightened inflammatory response. 56,57 Though we studied diabetes mellitus in the context of vascular dysregulation, it should be noted that diabetes mellitus and other forms of metabolic stress may have multiple mechanisms that could contribute to retinal ganglion cell loss. Diabetic patients have shown higher rates of LTG prevalence and glaucoma progression in some studies. 15,22 In contrast, the Advanced Glaucoma Intervention Study did not find diabetes to be associated with progressive visual field loss. 58 It has even been suggested that diabetes may confer a decreased risk for glaucoma. 59 Our data showed that LTG patients were significantly more likely to have diabetes mellitus, supporting the hypothesis for diabetes being a risk factor for LTG.
A correlation between OSA and LTG has been previously studied. 34,35 Our data did not show a significant association between OSA and LTG. In our study, LTG patients were less likely to be obese than the control population. Since a higher BMI is the leading risk factor for OSA, we hypothesize that this is the major difference in our study compared with prior studies in the literature showing an association between LTG and OSA. Some studies showed that a lower BMI has also been associated with a higher risk of glaucoma. 59-61 LTG patients have been shown to have a lower BMI compared with the general population and their POAG counterparts. 62,63 In our population, LTG patients were significantly more likely to have a normal BMI (< 25) rather than be obese ( > 30). This would support the hypothesis that a lower BMI could be a risk factor for LTG, whereas a higher BMI could serve a protective role in LTG. Tobacco smoking and alcohol use has been described as a risk factor for glaucoma in some studies but not others. 64,65 Our data did not find an association with tobacco smoking or alcohol use. Flammer syndrome has been described as a constellation of symptoms that include female sex, low BMI, nocturnal hypotension, Raynaud syndrome, migraine headache, and decreased blood flow with cold provocation. Traditionally, systemic hypertension, diabetes mellitus, and peripheral vascular disease are thought to cause vascular dysregulation through atherosclerosis and increased resistance to blood flow and may be related to a metabolic syndrome phenotype. In contrast, our phenotype 2 patients are more similar to patients with Flammer syndrome in terms of female sex, lower BMI, systemic hypotension, migraine headaches, and Raynaud syndrome. This would also help explain a parabolic relationship between BMI and LTG. This suggests that patients who are female with a normal BMI, systemic hypotension, migraine headache, Raynaud syndrome or anemia may develop LTG at a younger age.
This study is limited by its nature of the retrospective review. The diagnosis of LTG was based on a billing diagnosis and was made by a variety of providers including optometrists, comprehensive ophthalmologists, and glaucoma specialists. Though the LTG group was validated by disc hemorrhages, low IOP, and optic nerve cupping compared with controls, it is possible some LTG patients were missed in this analysis due to a lack of appropriate billing code. Furthermore, though we found no significant difference in ocular metrics between the 2 eyes in LTG patients, it is possible that a small number of patients with LTG may have the highly asymmetric disease, which we would not have captured using the mean values between the 2 eyes. The outcome measures were obtained from chart review and confined by the accuracy and detail of available data in the medical record. Determination of presence or absence of disease was based on diagnosis in the provider's assessment, report, and plan portion of the clinical note. Note that many of the systemic diseases studied may be correlated with one another, which could confound associations identified in our study.