Similarities and differences in systemic risk factors for retinal artery occlusion and retinal vein occlusion: A nationwide case–control study

To investigate the relationship between risk factors for retinal artery occlusion (RAO) and retinal vein occlusion (RVO) and thereby identify similarities and differences between the two types of retinal vascular occlusions. In this case–control study, 5708 patients with RAO were included and matched with three patients with RVO each. The patients with RVO were matched on sex and age at index date. All patients, personal information, diagnoses, and prescriptions were obtained from the Danish nationwide registries. Adjusted conditional logistic regression was used to investigate the association of RAO and RVO with the included risk factors. RAO was stronger associated with arterial hypertension, heart failure, ischemic heart disease, peripheral artery disease, and stroke than RVO, with effect measures ranging from 1.10 to 2.21. RVO was associated with cataract and glaucoma with effect measures of 0.80 (95% CI 0.73–0.87) and 0.65 (95% CI 0.56–0.76), respectively. Differences in the level of associations with the included risk factors suggests differences in the pathophysiologies of the two diseases. The main pathophysiology associated with RAO was atherosclerosis, whereas the main pathophysiology associated with RVO was changes in the pressure gradients of the eyes.


Introduction
Retinal vascular occlusions cause an obstruction of the retinal artery or vein, which can lead to vision loss and even blindness. A distinction is made between the arterial (RAO) and the venous (RVO) occlusion. The incidence rate of RAO is 1.8-1.9 per 100,000 person years for occlusions of the central retinal artery [1,2]. RVO is more common with an incidence rate of 1.5 per 10,000 person years, making RVO the second most common retinal cause of blindness, following diabetic retinopathy [3].
RAO and RVO share several risk factors and there are some similarities in their pathophysiologies. It is generally accepted that RAO is caused by atherosclerosis, mainly emboli originating from plaques in the carotid artery [4]. Different theories of the association between RAO and RVO have been described. One theory is compression of the vein, where the retinal artery becomes atherosclerotic. The stiff and rigid artery can then compress the lumen of the resilient vein [5][6][7]. However, the affected retinal vein has been investigated in connection with an occlusion and no changes in the lumen were determined [8,9]. Another theory is based on the pressure gradients of the eye. The eye acts as a Starling resistor, making the retinal vein susceptible to collapsing and thrombosis formation [10]. The retinal artery may contribute to this thrombosis formation by causing turbulence or changing the biochemical environment in the shared adventitia where the retinal artery comes in close proximity to the retinal vein [11].
One initial step in the understanding of the association between RAO and RVO is to elucidate clinical risk factors that the two diseases have in common and determine the level of association. The risk factors for RVO have been documented to some extent, whereas a limited number of risk factors have been investigated for RAO and mainly in smaller populations. The risk factors for RAO and RVO have mainly been investigated separately, and this study will investigate RAO and RVO combined. Our aim was to investigate the level of association for different risk factors between patients with RAO and RVO. Using Danish register-based data, risk factors identified in the existing literature for both diseases will be compared between a matched population of patients with RAO and patients with RVO.

Register data sources
Three different Danish nationwide registries were used to obtain the data for this study. The first register was the Danish Civil Registration System (CPR), where information about date of birth, date of death, vital status, migration, and sex can be extracted [12]. The second register used was the Danish National Patient Register (DNPR). The DNPR contains information on all hospital admissions along with diagnoses, procedures, and outpatient services [13]. The coding system in this register has been the Classification of Diseases version 10 (ICD-10) since 1994. The last register used was the Danish National Prescription Registry (LSR), where individual-level data on all prescription drugs sold in pharmacies in Denmark can be found. This register uses the global Anatomical Therapeutic Chemical classification code [14]. Information obtained in the three registers is crosslinked on individual level, using the personal identification number given to each resident of Denmark at birth or migration.

Study population
This was a case-control study. All inpatients and outpatients discharged with a first-time hospital diagnosis of RAO (ICD-10 codes: H340, H341, and H342, including all subcodes) constituted the case population. These diagnoses were extracted from the DNPR and linked with background information of the patients from the CPR. Data from private practicing ophthalmologists were not available. The index date was defined as the date of the RAO diagnosis. The case population was matched with a control population of patients with RVO. The matching was done on sex and age at the index date (± 1 year). The patients with RVO were identified in the DNPR (ICD-10 code H348, including all subcodes) and cross-linked with personal data from the CPR as for the cases. Afterward, we identified patients with RVO matching the RAO cases on sex and age at index date.
The flowchart in Supplementary Fig. 1 shows the selection of both the case population and the control population. The inclusion period was from January 1, 2000, through December 31, 2018. Exclusion criteria included inconsistent information from the CPR or age < 18 years at index date. Furthermore, controls were excluded if they had received a RAO diagnosis prior to their matched index date and cases were excluded if they had received a RVO diagnosis prior to their index date. Each case was matched with three random RVO patients, excluding cases with less than three possible controls. Patients with RVO could be controls for multiple cases. A total of 5708 patients with RAO were included in the time period. For each RAO case, three patients with RVO were included as comparisons, including a total of 17,124 RVO controls.

Exposure
All exposures investigated in this study were identified in the existing literature for both RAO and RVO. The literature for RAO consisted mainly of smaller studies, whereas more profound literature for RVO and associated risk factors were available. Diagnoses from up to five years prior to the RAO diagnosis or the RVO diagnosis were included for all risk factors investigated. All diagnoses were identified in the DNPR using ICD-10 codes and cross-linked with data for the respective case or control. Furthermore, drug prescriptions identified in the LSR were used for a more certain diagnosis for arterial hypertension, diabetes, and glaucoma.

Statistical analyses
For both cases and controls the baseline characteristics were described. Continuous measures were described using means and standard deviations, and categorical measures were described using proportions.
To investigate the association between risk factors for RAO and RVO, conditional logistic regression was used. The model was adjusted for sex, age at index, and potential confounders present both before the exposure and between the exposure and index date. Adjustments were made for peripheral artery disease, atrial fibrillation, ischemic heart disease, stroke, diabetes, heart failure, arterial hypertension, renal disease, cataract, and glaucoma. Odds ratios and 95% confidence intervals were summarized for comparisons of risk factors for RAO and RVO.
An analysis stratified on RAO subtype and RVO subtype was conducted, thereby comparing central retinal artery occlusion (CRAO) and central retinal vein occlusion (CRVO) in one strata and branch retinal artery occlusion (BRAO) and branch retinal vein occlusion (BRVO) in another strata. Furthermore, to ensure the introduction of anti-VEGF treatment did not influence the RVO group a sensitivity analysis stratifying the population in one strata of patients receiving their diagnosis before the year 2009 and one strata of patients receiving their diagnosis in or after the year 2009. Thereby, dividing patients according to the introduction of anti-VEGF treatment for patients with RVO.
The significance level was set at 0.05. Statistical analyses were performed using Stata Statistical Software: Release 16 (StataCorp LP, College Station, TX).

Results
This case-control study investigated risk factors shared by RAO and RVO patients. In Table 1, the baseline characteristics of both the RAO and RVO patients are summarized. The mean age of both populations was 69.0 years (sd: 11.9), and 44.2% of both populations were females. The prevalence of the investigated risk factors in each population is summarized in Table 1.
The model identified differences between risk factors for RAO and RVO. The differences and similarities between the investigated risk factors and whether each risk factor favored RAO or RVO are shown in Fig. 1. The model was adjusted for all included risk factors. Differences in the association between the exposures and the two diseases were identified for a majority of the investigated risk factors, except of atrial fibrillation, diabetes, and renal disease. Cataract and glaucoma were associated with RVO. The ORs were 0.80 (95% CI 0.73-0.87) for cataract and 0.65 (95% CI 0.56-0.76) for glaucoma. Arterial hypertension, heart failure, ischemic heart disease, peripheral artery disease, and stroke were associated with RAO with effect measures being 1. 10  Overall, the stratified analysis on retinal occlusion subtypes showed comparable results with the main analysis. The considerable differences included arterial hypertension, atrial fibrillation, and cataract in the CRAO patients and glaucoma and renal disease in the BRAO patients. The 95% CIs reached equivalence for multiple of these risk factors given smaller case and event numbers, where larger effect was identified in the main analysis. The stratified analysis on introduction of anti-VEGF treatment for RVO patients showed similar results before and after 2009. The only differences observed were arterial hypertension reached equivalence in the strata before the year 2009 and atrial fibrillation showed a stronger association in the RVO patients in the strata in and after the year 2009. The stratified analyses can be visualized in supplemental material, Table S2 and Table S3.

Discussion
The principal finding from this study suggests differences in the association between RAO and RVO and the investigated risk factors. The eye diseases were stronger associated with RVO, whereas cardiovascular diseases had stronger associations with RAO. This suggests some differences in the pathophysiology of the two diseases.
Understanding the risk factors for RAO and RVO individually and the differences between their pathophysiology is important in the management and treatment of these patients. Symptomatic patients with RVO are treated with anti-vascular endothelial growth factor to prevent neovascularization and reduce macular edema [15][16][17]. RAO is a more acute event with no treatment verified in randomized controlled trials, and some may even be harmful [18,19]. However, patients with RAO should be referred to stroke evaluation to prevent subsequent stroke events [20,21].
The strong association between RAO and cardiovascular diseases, especially heart failure, ischemic heart disease, peripheral artery disease, and stroke, supports the theory of atherosclerosis being the main cause of RAO [4,22]. The differences between RAO and RVO may suggest that cardiovascular diseases do not cause RVO as directly or strongly as they cause RAO. This theory is supported by the strong association between RVO and the investigated eye diseases since they seem to play a major role in the development of RVO as well.
The stratified analysis showed large similarities based on the mean effect estimates between the retinal vascular occlusion subtypes. However, in branch retinal vascular occlusions, renal disease was stronger associated with RAO, whereas in central retinal vascular occlusions and in the main analysis the effect measure for renal disease reached equivalence. In branch retinal vascular occlusion population, the effect measure reached equivalence. This was unexpected since glaucoma is a well-established risk factor for RVO, which was clear in the main analysis. This could be due to random variation. The mean effect estimates for arterial hypertension and cataract as well as for arterial hypertension in the strata before 2009 were like the main analysis; however, the 95% CI reached equivalence. Since the mean effect measure are similar, this may be due to the wider confidence intervals in smaller populations.
In both stratified analyses, atrial fibrillation showed slightly stronger associations with RVO in the central vascular occlusion population and the population diagnosed after or in the year 2009. This tendency was present in the main analysis as well.
Atrial fibrillation has previously been associated with venous thromboembolism [23], which supports the association with RVO. However, the 95% CI was close to equivalence, which suggests that the difference in association with atrial fibrillation may not vary largely between RAO and RVO.
Overall, possible pathophysiologies for both RAO and RVO may be elucidated from this study, in accordance with the existing literature. The association with arterial disease being the main cause of RAO was supported. The cardiovascular diseases with effect measures close to equivalence indicate similar associations between the risk factors for both RAO and RVO, which support atherosclerosis as a potential cause of RVO as well. This could be by compression as described earlier. The strong association between especially glaucoma and RVO suggests that the intraocular pressure (IOP) may contribute to the development of RVO. IOP is a well-established risk factor for both glaucoma and RVO [24,25]. The IOP will raise when the composition of fluid in the eye becomes imbalanced [26], which could result in glaucoma. Several studies, including this study, have identified glaucoma as a strongly associated risk factor for RVO [24,[27][28][29]. Different types of glaucoma have been described, depending on where in the renewal of fluid the imbalance occur [26]. The changes in the pressure inside the eye, arterial hypertension, or ocular hypertension may affect the blood supply of the eye. The retinal veins are especially vulnerable to these changes due to their thin and flexible vessel walls [30,31]. The eye acts like a Starling resistor around the retinal vein, where changes in the pressure gradients between the intraluminal pressure and the IOP could make the retinal vein susceptible to partial collapse or in some cases complete collapse [30,31]. The collapsed vein will have limited blood flow and be susceptible to thrombosis formation, which may be initiated by turbulence or changes in the biochemical environment. Turbulence could be mediated by the retinal artery or venous pulsation, which is caused by vibrations from the pulsating retinal artery, from the pulsating IOP, or from the variation between the pulse pressure between the intraocular space and the cerebrospinal fluid in the lamina cribrosa [11,[32][33][34]. The same mechanism may cause RAO and elevated IOP have been described as a risk factor for RAO as well [35]; however, the retinal artery is more resistant to changes in the pressure gradients in the eye [36]. Furthermore, the main pathophysiology supported for RAO remains atherosclerosis, where elevated IOP may increase the risk of RAO development [37].

Strengths and limitations
The nationwide Danish registries ensure full coverage of all included patients, increasing the reliability of this study. This was a large study with a total of 22,832 patients included. For each RAO case, three patients with RVO were included as controls. Another strength of this study was the matching between the included cases and the controls, which reduced the risk of confounding and completely excluding sex and age as potential confounders.
Some limitations are associated with conducting a register-based study. It was not possible to acquire information on all factors that potentially could introduce confounding. In particular, lifestyle factors, including smoking, body mass index, and alcohol consumption, were of great interest. However, these factors are not registered in the CPR, DNPR, or LSR. In addition, clinical manifestations were not available as well, which would have been of interest in the comparison of RAO and RVO. In addition, it was not possible to verify all the diagnoses obtained in the registers. The positive prediction value has been investigated for some of the investigated diagnoses. These include atrial fibrillation, heart failure, and arterial hypertension, where PPVs ranging from 76 to 95% were determined [38]. The validation of the included diagnoses was accepted and not evaluated to cause significant bias due to incorrectly registered diagnoses. Furthermore, the registers include data from hospitals, limiting some of the diagnoses to the more severe cases. The mild cases of cataract and glaucoma may only be assessed by practicing ophthalmologists, and diabetes may only be assessed by general physicians. Therefore, the results may not be applicable for the milder cases of these risk factors.
The Danish registries do not register which eye is affected. Therefore, we could not match the RAO or RVO and the risk factors by eye.
The RVO patients may not be representative for all RVO patients, since the matching limited the included RVO patients to a subpopulation of same age and gender as the RAO patients. However, the groups of patients for RAO and RVO are similar, which suggested matching on age and sex would not affect the results significantly.

Conclusion
Our study suggests that atherosclerosis is the main underlying pathophysiology of RAO, where changes in the hemodynamics of the eye can make the retinal artery more susceptible to an occlusion. Conversely, the main pathophysiology supported for RVO is changes in the pressure gradients of the eye, where atherosclerosis and cardiovascular changes may contribute to occlusion formation in the vulnerable veins. The modifiable risk factors associated with these pathogeneses should be managed aggressively to prevent the diseases and prevent subsequent cardiovascular events.