Introduction

Retinal artery/arteriole occlusion (RAO) is a major cause of sudden painless vision loss that often results in devastating impairment.1 The disease entity encompasses central RAO (CRAO), branch RAO (BRAO) and RAO. Typical RAO is most commonly caused by an impacted embolus in the narrowest part of the central retinal artery, where it enters the sheath of the optic nerve,2 or branch retinal arterioles. The main sources of emboli are the carotid artery and heart.2 Various cardiac conditions or invasive cardiovascular procedures cause RAO.3 Because the central retinal artery arises from the ophthalmic artery (OphA), the first major branch of the internal carotid artery,4 significant stenosis of either the OphA or internal carotid artery can result in the development of CRAO. Although much rarer, emboli from other intracranial vascular branches can also result in CRAO due to variations in the origin and course of the OphA.3 Thus, the presence of intracranial vascular abnormalities, such as unruptured intracranial aneurysm (UIA) and invasive vascular treatment procedures can cause thromboembolic complications in retinal vessels.

A UIA>7 mm has a higher risk of rupture and is warranted for treatment by endovascular treatment (EVT) and surgical clipping.5 Endovascular coil embolisation is widely practised in cases of UIA given its better safety, anatomical accessibility and higher success rate,5 whereas the surgical approach outweighs embolisation in complete occlusion and a lower recurrence rate. In cases of paraclinoid aneurysms, which arise from the segment of the internal carotid artery between the distal dural ring and origin of the posterior communicating artery, endovascular access is preferred over direct surgery because of technical difficulties caused by the anatomical complexity.6 7 Because paraclinoid aneurysms are located near or contain branches of the OphA, RAO might occur more frequently in paraclinoid aneurysms with or without treatment. In previous studies, most RAO cases were found post-treatment for paraclinoid aneurysms,8–15 although some have been reported in other types of UIAs.16–18 Currently, the frequency or risk factors of RAO occurring after EVT remain elusive.19 In this study, we first analysed a nationwide cohort to investigate the incidence rate of RAO in UIA according to the treatment protocol and to evaluate the risk factors associated with the incidence of RAO in patients with UIA. We further analysed the hospital cohort using detailed clinical information, including the location and size of UIAs.

MethodsData source and study population

We used data recorded from January 2016 to July 2022 from the National Health Claims Database of the HIRA service in Korea. The data source has been described in detail elsewhere.20 Briefly, the HIRA database provides extensive information regarding all health claims in Korea, either submitted through the Korean National Health Insurance scheme or Medical Assistance Programme.21 This claims database includes direct medical costs, personal information, demographics, diagnoses, procedures and medical prescriptions of all Korean residents. As the database is based on unique identification numbers (Korean Resident Registration Number), that are given to all Korean residents at birth, it can be used to obtain the healthcare records and demographics of patients without duplication or omission.20

We identified 423 559 eligible patients diagnosed with UIA between January 2016 and July 2022 (figure 1). All diagnoses were based on the International Classification of Diseases, 10th edition (ICD-10). To clarify the primary risk of UIA on the incidence of RAO, we set a ‘washout period’ and excluded patients with any previous hospital claims for UIA, CRAO or BRAO between 2016 and 2017 (n=168 692). Patients diagnosed with RAO before the diagnosis of UIA (n=332) or before treatment for UIA (n=65) were excluded. Lastly, patients who underwent both microsurgical clipping and EVT for UIA were excluded (n=1230). Finally, 253 240 participants were included in this analysis.

Figure 1

Flow chart of patients excluded and included in the analysis in the current study. HIRA, Health Insurance Review and Assessment.

Exposure

We categorised patients with UIA into three groups based on subsequent treatment: observation (n=208 993), microsurgical clipping (n=14 168) and EVT groups (n=30 079). Patients were considered to have received specific treatment if relevant ICD-10 codes for microsurgical clipping (S4641 or S4642) and EVT (M1661 or M1662) were given after the diagnosis of UIA in the claims data.

Outcome and covariates

The primary outcome was the incidence rate of early (within 60 days) or delayed (after 60 days) RAO and the period from the index date of diagnosis or treatment of UIA to the incidence of RAO. The earliest date of the claim for RAO was considered the incident time of the disease. The secondary outcomes included factors associated with RAO development after EVT.

Baseline patient characteristics, including age, sex, type of medical insurance, residence and medical conditions, such as hypertension, diabetes, dyslipidaemia, transient ischaemic attack, ischaemic stroke, haemorrhagic stroke, myocardial infarction, chronic kidney disease (CKD), malignancy, hyperthyroidism, hypothyroidism, chronic liver disease, chronic obstructive pulmonary disease and Charlson Comorbidity Index (CCI), were collected. To ensure accuracy, comorbidities were established based on one inpatient or two outpatient records with ICD-10 codes in the database. Prescription histories of antithrombotic agents or anticoagulants within 6 months before baseline were obtained from the claims data. Details of the ICD-10 codes used to define the medical conditions, comorbidities, drug treatments and procedures are presented in online supplemental table 1.

Hospital cohort study with detailed clinical information

As the claims data did not include detailed information on UIAs, such as location or size, we conducted additional analyses using a hospital cohort comprising consecutive patients who were treated with microsurgical clipping or EVT for UIAs at a tertiary high-volume hospital between January 2018 and June 2022. In total, 2569 consecutive patients with UIAs (2896 UIAs) treated with microsurgical clipping or EVT were eligible for the study. Patients were categorised based on treatment: 1176 patients with 1369 aneurysms treated with microsurgical clipping, and 1393 patients with 1527 aneurysms treated with EVT. Therapeutic decisions were discussed by a multidisciplinary team of neurosurgeons and neurointerventionists, and a final decision was made by consensus.

The analysed aneurysmal factors included the number of aneurysms, maximum aneurysm diameter, aneurysmal location, accompanying vascular problems and whether re-embolisation was performed. The locations of the aneurysms were classified as the internal carotid artery, which encompasses the distal paraclinoid artery, paraclinoid artery and proximal paraclinoid artery; anterior cerebral artery, including the anterior communicating artery; middle cerebral artery; or posterior circulation, including the vertebral artery, basilar artery, posterior cerebral artery and cerebellar arteries. Considering the proximity of the OphA branch and paraclinoid artery, the risk factors associated with development of RAO in paraclinoid aneurysms were assessed separately. Procedure-related factors included the emboliszation method (single or multiple microcatheter technique) and use of devices (balloon, stent, flow diverter or flow disruptor). For details of the collected data from the retrospective chart review and routine treatment protocols for EVT, seee online supplemental methods.

All patients who had acute-onset vision loss after treatment for UIA underwent complete ophthalmic evaluations, including best-corrected visual acuity, intraocular pressure, slit lamp biomicroscopy, dilated fundoscopic examination, fundus photography and spectral-domain optical coherence tomography. A visual field test and fluorescein angiography were additionally performed, if necessary. The diagnosis of RAO was made clinically by ophthalmologists at the initial assessment and additionally confirmed by two retinal specialists (HRK and YJK) during a retrospective chart review.3

Statistical analyses

Baseline characteristics are reported as mean±SD or frequency (per cent). Using the HIRA database, the cumulative incidence of RAO according to the subgroups was calculated using the Kaplan-Meier method. Age-adjusted and sex-adjusted standardised incidence ratios (SIR) of RAO were calculated using the RAO incidence rate in the entire Korean population between 2018 and 2019 as a reference. HRs and 95% CIs based on the Poisson distribution for RAO associated with the subgroups were calculated using Cox proportional hazards models, with untreated UIA as the reference. HRs were adjusted for age, sex and the presence of hypertension, diabetes mellitus or dyslipidaemia. The adequacy of the proportional hazard assumption was confirmed using log-minus-log survival plots and Schoenfeld residuals. Univariable and multivariable analyses of risk factors for RAO in UIA were performed.

For hospital cohort data analysis, Pearson’s χ2 test or Fisher’s exact test was used to determine the differences between categorical variables. The means for differences between continuous variables were analysed using the Student’s t-test. A logistic regression model was used to evaluate factors associated with RAO. The results are reported as ORs and 95% CIs.

Statistical analyses were performed using SAS (V.9.4, SAS Institute), SPSS (V.25.0; IBM) and R (V.3.5.3, R Foundation for Statistical Computing, Vienna, Austria). P values <0.050 were considered statistically significant.

ResultsBaseline characteristics of the nationwide cohort

Of the 2 53 240 patients with UIA, 208 993 (82.5%) were observed without further treatment, 14 168 (5.6%) underwent microsurgical clipping and 30 079 (11.9%) underwent EVT. The mean ages were 62.46±13.82, 60.14±9.92 and 60.32±11.76 in the observation, microsurgical clipping and EVT groups, respectively. The three groups differed in age, sex, residence, comorbidities of systemic diseases, CCI and prescription history of antithrombotic agents or anticoagulants. Patients with UIA who required treatment with either microsurgical clipping or EVT were significantly younger, predominantly female and had urban residency than those who were observed without further treatment. The EVT group was more likely to have received aspirin or antiplatelet therapy than the other groups. The general characteristics of the nationwide cohort are presented in table 1.

Table 1

Baseline characteristics of the nationwide cohort by subsequent treatment

Risks of RAO

The cumulative incidence of RAO was significantly higher in the EVT group (62.41 per 100 000 person-years) than in the other groups (30.49 and 18.39 per 100 000 person-years in the observation and microsurgical clipping groups, respectively). In the EVT group, early RAO (25 patients (0.08%)) was markedly observed compared with the other groups (observation group: 45 patients (0.02%), microsurgical clipping group: 4 patients (0.03%)) (figure 2 and online supplemental figure 1).

Figure 2

Cumulative incidence of retinal artery occlusion in patients with unruptured intracranial aneurysm according to the Kaplan-Meier method. EVT, endovascular treatment.

When age, sex and the presence of hypertension, diabetes or dyslipidaemia were adjusted, microsurgical clipping was not associated with higher RAO risks in comparison to conservative treatment (early RAO: HR=1.33, 95% CI: 0.48 to 3.72; delayed RAO: HR=0.45, 95% CI: 0.14 to 1.40). In the EVT group, early RAO developed in 25 (0.1%) patients and delayed RAO developed in another 25 (0.1%), with HRs of 4.00 (95% CI: 2.44 to 6.56) and 1.74 (95% CI: 1.13 to 2.68), respectively (figure 3). The characteristics of the patients who developed early and delayed RAO are presented in online supplemental table 2.

Figure 3

Cox regression analysis of retinal artery occlusion (RAO) risks by onset date. (A) Early RAO, which occurred within 60 days after the diagnosis or treatment of unruptured intracranial aneurysm (UIA), (B) delayed RAO, which occurred after 60 days or longer from the diagnosis or treatment of UIA. Group 1, patients with untreated UIA; Group 2, UIA patients treated with microsurgical clipping and Group 3, UIA patients treated with endovascular treatment. The Cox model was adjusted for age, sex and the presence of hypertension, diabetes mellitus or dyslipidaemia. The incidence rate is per 100 000 person-years.

Next, we investigated whether the RAO risk increased when each group was compared with the general population. The incidences of RAO in the general population were 13.96 and 9.40 per 100 000 person-years among men and women, respectively (online supplemental table 3). Age-adjusted and sex-adjusted SIR of RAO was also significantly higher in the EVT group (SIR=2.87, 95% CI: 1.24 to 4.02, p<0.001) than in the entire Korean population, whereas the values in observation (SIR=1.13, 95% CI: 0.93 to 1.19, p=0.120) and microsurgical clipping groups (SIR=0.87, 95% CI: 0.51 to 1.75, p=0.713) were similar to that of the general population (online supplemental table 4).

Risk factors of RAO after EVT for UIA

We assessed the risk factors for RAO after EVT for UIAs. In univariable analysis, female sex (HR=0.58, 95% CI: 0.33 to 1.02, p=0.058), CKD (HR=4.02, 95% CI: 1.45 to 11.17, p=0.022) and the use of a balloon microcatheter (HR=2.93, 95% CI: 1.16 to 7.39, p=0.008) were associated with RAO after EVT. Age, socioeconomic status, residence, hypertension, diabetes, dyslipidaemia, stroke, CCI and the use of anticoagulant/antithrombotic agents were not significantly associated with RAO after EVT. Multivariable analysis revealed that CKD (HR=3.98, 95% CI: 1.42 to 11.19, p=0.009) and the use of a balloon microcatheter (HR=3.33, 95% CI: 1.29 to 8.65, p=0.013) were significantly associated with RAO after EVT (table 2).

Table 2

Results of univariable and multivariable analyses for the risk factors of retinal artery occlusion after endovascular treatment for unruptured intracranial aneurysm in the nationwide cohort

Hospital cohort analysis with detailed clinical information

In total, 1393 patients (female=1,075 (77.2%); mean age, 59.5 years) underwent EVT and 1176 patients (female=859 (73.0%); mean age, 60.6 years) underwent microsurgical clipping. Although RAO did not occur after microsurgical clipping, 11 cases (0.8%) occurred after EVT (p<0.001). All RAO cases occurred in the same direction as the right/left positions of the UIAs. Anatomically, RAO appeared to be prevalent after treatment of aneurysms located in the internal carotid artery (paraclinoid aneurysms, 10 (90.9%); and distal paraclinoid aneurysm located at posterior communicating artery, 1 (9.1%)), with borderline statistical significance (p=0.063). Previous studies reported that RAO cases after UIA treatment also occurred after EVT, and most UIAs were paraclinoid aneurysms. The detailed characteristics of the patients who developed RAO after treatment with UIAs are presented in online supplemental table 5. Subjective vision decreases or visual field defects of varying degrees occurred within 1–2 days of treatment. At the last follow-up, three patients (patients 3, 9, 11) had partial vision loss and one patient with CRAO (patient 2) had permanent vision loss. Representative images of RAO after EVT are shown in online supplemental figure 2. Notably, patients with RAO were younger and used balloon microcatheters more frequently than patients without RAO in the EVT group (table 3).

Table 3

Characteristics of patients with hospital-based unruptured intracranial aneurysm (UIA) who underwent microsurgical clipping or endovascular treatment

We conducted a subgroup analysis of paraclinoid aneurysms which is close to the OphA. Among the 612 patients with 638 aneurysms, 10 (1.6%) patients developed RAO after EVT. Multivariable logistic regression analysis showed that the independent risk factors for RAO were the use of a balloon microcatheter (OR=7.56, 95% CI: 1.35 to 42.44, p=0.022). The details of the risk factor analysis are presented in table 4.

Table 4

Risk factors of retinal artery occlusion (RAO) in paraclinoid aneurysm treated with endovascular treatment

Discussion

In this nationwide, population-based study, we elucidated the risk of RAO in patients with UIA according to subsequent treatment and the associated risk factors for RAO. Few case reports have been published regarding the development of RAO after endovascular coil embolisation for UIA, and there is no evidence regarding the association between RAO and UIA. Our results provide detailed information regarding the increased risk of RAO and the time of incident RAO, especially in patients treated with EVT for UIA. The rapid increase in the incidence of RAO during the first 60 days after treatment of UIA implicates a direct relationship between EVT and RAO. As patent artery occlusion (2%) and thromboembolism (2.4%) are the most common complications of this procedure,5 16 it is possible that coil embolisation itself may cause RAO. Moreover, the use of certain devices, direct contact with the wall of the intracranial artery or plaque break-off during catheter passage may play a role in the development of thrombus-related RAO.16 17 In our study, the use of a balloon microcatheter was significantly associated with higher risk of RAO in the HIRA dataset and hospital cohort. In cases of balloon-assisted coil embolisation, intermittent OphA occlusion caused by balloon inflation precedes coil insertion. Previous studies revealed that collateral OphA flow from the external carotid artery was present in 60%–84.6% of cases.22 23 Usually, distal flow is maintained because of collateral flow, and the possibility of RAO is low. It is assumed that, however, because of this collateral, a thrombus may develop at the proximal segment of the OphA during balloon inflation. Specifically, the usual blood flow from the internal carotid artery to the OphA changes to a flow directed from the external carotid artery to the OphA during balloon inflation, increasing the risk of thrombus formation at the site of flow retention. After balloon deflation, the thrombus may wash out into the OphA, causing RAO.

The microsurgical treatment of paraclinoid aneurysms remains challenging because of their anatomical features. Meticulous surgical techniques, including anterior clinoidectomy, optic strut removal and dissection of the distal dural ring, are required to fully expose the aneurysm. Simultaneously, surrounding anatomical structures (eg, optic nerves, oculomotor nerves, perforating arteries and cavernous sinuses) should be preserved. In previous studies, the rate of visual complications after microsurgical clipping was 4.3%–30%.24–26 Visual dysfunctions included optic neuropathy,24 visual field defects,25 26 vision loss of unmentioned cause and oculomotor nerve palsy25 other than retinal vessel occlusion. The recent introduction of new devices (eg, balloon microcatheter, stent, flow diverter and flow disruptor) to treat wide-neck aneurysms that are unsuitable for simple coil embolisation has broadened the range of endovascular approaches. Therefore, EVT is widely used for the treatment of paraclinoid aneurysms. In our hospital cohort, RAO did not occur after microsurgical clipping, whereas a few cases occurred after EVT (total EVT group, 0.8%; paraclinoid aneurysms treated with EVT, 1.6%). This implies a higher risk of RAO in cases of paraclinoid aneurysms than in other UIAs. Although the location of the UIA showed borderline significance owing to the small number of RAO cases in the hospital cohort, all RAO cases were reported as internal carotid artery aneurysms, mostly paraclinoid aneurysms. Perhaps the position of the balloon during balloon-assisted coil embolisation might be similar, resulting in a similar risk of consecutive RAO. Interpretations of the results were limited because the number of RAO after EVT was very small and none occurred after microsurgical clipping. Further studies are warranted to recruit a larger cohort encompassing untreated UIAs and the location of the UIAs.

We speculate that the presence of UIA itself might cause thromboembolic complications in the OphA and retinal artery. A comparison of patients with UIA with the general population showed that untreated UIA and microsurgical clipping for UIA did not increase the risk of consecutive RAO. Blood flow from the heart and carotid artery is directed towards the OphA, and emboli from those vessels mainly cause RAO.2 However, intracranial vessels are located more distal to the OphA branch and have different courses of circulation. Therefore, UIAs other than paraclinoid aneurysms are less likely to cause RAO. Different aetiologies may play a role in the occurrence of RAO in the UIAs located elsewhere. Bae et al reported a case of cilioretinal artery occlusion after EVT of the anterior communication artery.16 Considering the distal location of the aneurysm from the OphA and the direction of blood flow, it is unlikely that the thrombus caused RAO. Air embolism during heparin injection into the catheter may cause microembolism-related retinal infarction.27 Structural changes in the arterial wall, breakdown of blood products or inflammation due to catheter-induced stimulation of the endothelium may result in vasospasm and then hypoxia.16 28

Interestingly, the presence of CKD was significantly associated with an increased risk of RAO in EVT group. The correlation between end-stage renal disease and RAO risk has been well described.29 30 As RAO and CKD share common risk factors (eg, hypertension, diabetes mellitus, hyperlipidaemia and atherosclerosis), the incidence of RAO can be higher in patients with CKD.30 Moreover, chronic inflammation, which is prevalent in CKD, may contribute to the development of RAO. However, these studies were limited to patients on dialysis. The relationship between the presence of CKD and RAO needs to be evaluated further.

The strength of this study is that it was based on a nationwide population-based dataset, representative of the entire Korean population. Furthermore, we introduced an inpatient dataset to supplement clinical data that were not provided in the nationwide cohort. However, this study has the following limitations. First, the data were based on diagnostic codes and insurance claims; therefore, it is possible that misclassification or miscategorisation could have occurred. Second, information related to the clinical characteristics of UIA could not be assessed in the nationwide cohort, given that factors other than treatment method might interfere with the risk of RAO. To compensate for this, a comparison with the general population using age-adjusted and sex-adjusted SIR and additional analysis using the hospital cohort were conducted. Third, because not all patients with UIA underwent ophthalmic examinations, asymptomatic patients with RAO were not evaluated. Lastly, confounding factors, such as smoking history and alcohol consumption, which were not available in the HIRA dataset, could not be adjusted.

Because the most dreaded risk of intracranial aneurysms is rupture, which leads to a mortality rate as high as 80%, timely treatment is essential for UIA. However, caution should be exercised when deciding on UIA treatment, as indiscriminate endovascular or surgical treatment may increase the risk of RAO and irreversible vision loss. With the development of EVT, the treatment of UIA, especially paraclinoid aneurysms, has become easier. However, clinicians should be aware of the increased risk of RAO after EVT for paraclinoid aneurysms and the use of balloon-assisted coil embolisation. Also, further studies are warranted to recruit a larger cohort to elucidate associated factors of RAO in patients with UIAs.