We conducted a retrospective, multicenter, observational study. Centers involved in the French National Reference Center for HES (CEREO) and in the European Eosinophilic Granulomatosis with Polyangiitis study group as well as 1 US center of EGPA expertise (National Jewish Health, NJH) were queried to identify patients with: (i) at least one episode of ophthalmic vascular manifestation (CRAO, branch retinal artery occlusion (BRAO), CRVO, branch retinal vein occlusion (BRVO), Purtscher’s retinopathy, retinal vasculitis or ION); (ii) concomitant absolute eosinophilia count (AEC) ≥ 0.5 x10⁹/L when the ophthalmic vascular manifestation occurred. Exclusion criteria were the presence of any condition, comorbidity or concomitant treatment leading to thrombophilia (either constitutional or acquired), cardiac embolism, rhythmic heart disease, tight carotid stenosis (NASCET ≥70%) homolateral to the retinal involvement or other causes of Purtscher’s retinopathy (e.g. acute pancreatitis, head trauma or thrombotic microangiopathy), ION (e.g. giant cell arteritis) as well as the presence of anti-myeloperoxydase (MPO) anti-neutrophil cytoplasmic autoantibodies. A comprehensive list of exclusion criteria is provided in the Supplementary Appendix .

The PUBMED database was searched for English-language publications released up to April 2023, using the following combination of MeSH terms: (i) ‘hypereosinophilic syndrome’ (or any term referring to a condition embedded within the spectrum of clonal, reactive (including lymphocytic HES, drug-induced or paraneoplastic eosinophilia) and idiopathic HES (including single-organ and systemic eosinophil-associated diseases)); (ii) and a MeSH term referring to an ophthalmic vascular manifestation (e.g., retinal artery occlusion, retinal vein occlusion, retinal vasculitis, retinal diseases, optic neuropathy). Reference lists from selected publications were screened for additional relevant studies.

All cases were reviewed by the investigators (EC, MG) considering the entire follow-up. Using a standardized de-identified case report form, demographic (including cardiovascular and venous thromboembolism risk factors), clinical, laboratory and imaging findings at the time of the ophthalmic vascular manifestation as well as during follow-up were collected. For each patient, the underlying process underpinning blood hypereosinophilia was assessed according to the International COoperative study Group on Eosinophil disorders (ICOG-Eo) terminology (9) and thus considered as either clonal (i.e. neoplastic, including FIP1L1::PDGFRA myeloid neoplasm with eosinophilia), reactive (including all conditions that lead to the production of type 2 inflammation-related cytokines and thereby to non-clonal HE), overlapping (when embodied within the spectrum of autoimmune diseases, e.g. MPO ANCA-negative EGPA (23), IgG4-related disease (24), or eosinophilic fasciitis (25)), or idiopathic.

For patients with ≥ 3 months of follow-up, and after exclusion of patients with single-flare eosinophilia (parasitosis and drug-induced eosinophilia), studied outcomes included visual acuity at last follow-up, the recurrence of either ophthalmic or extra-ophthalmologic vascular events, the occurrence of ophthalmic complications (e.g. retinal neovascularization, intravitreal hemorrhage or neovascular glaucoma), and death. Full ophthalmic recovery was defined as the resolution of ophthalmologic symptoms (full correction of visual acuity, visual field normalization, no recurrence of transient monocular blindness). Partial recovery was defined as partial improvement in visual acuity and/or visual field.

Patient characteristics are reported as median [interquartile] ([IQR]) and frequency (percentage) for continuous and categorical variables, respectively. Visual acuity was converted into the log of the minimum angle of resolution (logMAR). Patient subsets were differentiated based on the subtype of vascular manifestation (arterial involvement vs. venous thrombosis vs. Purtscher’s retinopathy). Visual outcomes were compared using the Chi-squared test and continuous variables were compared using the Kruskal Wallis test.

All methods were carried out in accordance with relevant guidelines and regulations (i.e. the Good Clinical Practice protocol, the Declaration of Helsinki principles and the MR004 French legislation regarding observational retrospective studies) and this study was approved by the independent ethics committee of Foch Hospital (IRB00012437, approval number 23-03-04).

One hundred and twenty-three patients were screened through the CEREO, EESG and NJH databases, and 55 case reports (corresponding to 56 patients) treated for eosinophilia and concomitant ophthalmic vascular manifestations were identified through the literature review. Overall, 57 patients fulfilled inclusion criteria (20 from the observational study and 37 from the literature review, Figure 1 ). Their main characteristics are reported in Table 1 . Thirty-two (56%) were male and their median age at ophthalmic manifestation onset was 54 [44-65] years. Thirty-six (58%) had at least one cardiovascular risk factor, three (5%) had a prior history of cardiovascular disease and four (7%) had a prior history of venous thromboembolism. Among the 55 patients with available data, only one (2%) developed ophthalmic vascular involvement despite ongoing treatment with both antiplatelets and anticoagulants.

Ophthalmologic symptoms were the initial eosinophil-related organ involvement in 34 (59%) patients, while 23 (41%) had already been diagnosed with a prior eosinophil-associated disease, including 17 (30%) who were currently treated with systemic corticosteroids at the time of ophthalmic vascular involvement and four (7%) who had prior eosinophil-related venous thromboembolism. Among the seven patients with available data, the median delay between eosinophil-related first manifestation and ophthalmic symptoms was 36 [15-96] months. The median AEC at onset of ophthalmic vascular manifestation was 3.5 [1.7-6.8] x10⁹/L. Among the 40 patients (70%) with available data, the median peak AEC during follow-up was 4.6 [2-8.9] x10⁹/L, including 37 (92%) patients with hypereosinophilia > 1.5 x10⁹/L (i.e. the HES-defining threshold). Among the 48 patients with extra-ophthalmologic eosinophil-related organ involvement, the median number of organs involved was two [1-3], consisting mostly of peripheral nervous system (n=23), skin (n=28), lung (n=16), ENT (n=17) and musculoskeletal (n=12) manifestations. Of note, four (7%) patients reported features of arterial thrombosis (ischemic stroke n=2, acute coronary syndrome and gastrointestinal tract ischemia, a single patient each), nine (16%) of venous thromboembolism (pulmonary embolism n=4, lower limb deep venous thrombosis n=3, inferior vena cava thrombosis, pulmonary vein thrombosis, a single patient each), and four (7%) had eosinophilic myocarditis. Among the 13 thrombotic manifestations, seven occurred concomitantly with ophthalmic vascular manifestations. Eosinophilic granulomatosis with polyangiitis (EGPA, formerly Churg-Strauss syndrome) was the leading cause driving eosinophilia (n=32, 57%). Other associated conditions included STAT5A-associated chronic eosinophilic leukemia (n=1, 2%), lymphocytic (n=2, 4%), idiopathic (n=13, 23%), reactive (n=5, 7% including three drug adverse reactions and two parasitosis) HES as well as overlapping diseases (n=4, 7%, consisting of polyarteritis nodosa, IgG4-related disease, sclerosing cholangitis and eosinophilic fasciitis, a single patient each).

Ocular involvement was bilateral in 21 patients (37%), and in the remaining patients there was no predominant eye involved. Ophthalmic vascular manifestations included CRAO (n= 24 patients, 29 eyes), BRAO (n= 5 patients, 6 eyes), CRVO (n= 3 patients, 5 eyes), BRVO (n= 2 patients, 2 eyes), retinal vasculitis (n= 5 patients, 7 eyes), retinal vasospasm evidenced upon fluorescein angiography (one patient with bilateral involvement), Purtscher’s retinopathy (n=6 patients, 12 eyes), anterior ischemic optic neuropathy (n= 10 patients, 13 eyes) and posterior ischemic optic neuropathy (n=2 patients with unilateral involvement). Ocular symptoms (present in all patients but one) consisted of vision loss (53 patients, 71 eyes), transient monocular blindness (11 patients, 13 eyes) and visual field abnormalities (7 patients, 9 eyes). The median visual acuity at diagnosis was 1.7 [0.5-2.3] logmar. In all eyes with available data (n=64 eyes), fundus examination was always abnormal (fundus data of the 2 patients with posterior ischemic optic neuropathy were not available). Among the 39 fluorescein angiographies that were performed, imaging findings were abnormal and concordant with the clinical diagnoses suspected upon fundus examination in 37 (95%) cases. Figure 2 illustrates the retinal imaging finding of a patient with CRAO in the left eye. Among the 28 patients with imaging of supra-aortic arteries, 8 (29%) had evidence of mild (NASCET < 70%) ipsilateral carotid artery atheroma. Of note, among the 22 patients with available concomitant brain imaging (cerebral MRI or injected TDM), cerebral ischemic events were reported in two (5%) cases.

At the acute phase, forty-five (79%) patients received systemic corticosteroids (oral corticosteroids, n=23; intravenous corticosteroids n=22; starting doses ranging from 10mg to 1mg/kg/day of prednisone or equivalent before gradual tapering). Twelve (21%) patients received immunosuppressants, including cyclophosphamide (n=8), azathioprine (n=3) and methotrexate (n=1). Fibrinolytic agents and ocular hypotensive therapy were prescribed in five (9%) and four (7%) patients, respectively, while nine (16%) and 21 (38%) patients received long-term anticoagulants or antiplatelets, respectively. Other treatments consisted of antiparasitic drugs (n=1), plasma exchanges (n=2) and intravenous immunoglobulins (n=1).

During follow-up, other treatments included imatinib (n=2), hydroxyurea (n=2) and infliximab (n=1).

Full details of patients' outcomes are provided in Table 2 . Among the 50 (88%) patients with available follow-up data (including 30 with more than three months of follow-up), the median follow-up after the initial ophthalmic vascular manifestation was 10.5 [1-18] months. One (3%) patient had a recurrence of ophthalmic vascular manifestation (CRVO) and three (10%) patients had another vascular event (lower limb deep vein thrombosis n=2, pulmonary embolism n=1). In all cases, AEC was above 0.5 x10⁹/L at time of recurrence. At last follow-up, only six patients (12%) achieved full recovery. Sixteen patients (32%) achieved partial recovery, 23 patients (46%) stabilized once under treatment, while the condition of the remaining five patients (10%) worsened despite therapy. By comparison with patients with arterial involvement, the rate of ophthalmic recovery was higher in patients with either Purtscher’s retinopathy or venous involvement (recovery achieved in 8/10 eyes with Purtscher’s retinopathy and in 4/5 eyes with venous involvement vs 18/53 eyes with arterial involvement, p=0.019). Patients with venous involvement or Purtscher’s retinopathy had a better visual acuity at last follow-up than patients with arterial involvement (median logmar 0 [0-0], 1 [0-2.3] and 0.5 [0.2-0.8] for patients with venous, arterial involvement and Purtscher’s retinopathy respectively; p= 0.038). Of note, the levels of baseline absolute eosinophil counts did not correlate with long-term visual outcomes, and there were no significant differences regarding the final visual acuities of patients who received antiplatelets vs. anticoagulants as well as in patients who received only corticosteroids vs. those who received both corticosteroids and immunosuppressants (data not shown).

Three (10%) patients developed retinal neovascularization with subsequent intravitreal hemorrhage during follow-up. During follow-up, four patients died of pulmonary infection, MRSA-induced septic shock, endomyocardial fibrosis and hepatitis (a single patient each). Both patients who died of sepsis were treated by systemic corticosteroids (in addition to methotrexate or mepolizumab, a single patient each).

Thirty-two (57%) patients fulfilled MIRRA and/or ACR criteria for EGPA (all without MPO-ANCA). In the latter patients, ophthalmic vascular manifestations consisted of CRAO (17 patients, 21 eyes), BRAO (a single patient with unilateral involvement), AION (8 patients, 11 eyes), PION (a single patient with unilateral involvement), retinal vasculitis (3 patients, 4 eyes), CRVO (a single patient with bilateral involvement) and BRVO (a single patient with unilateral involvement). It was noteworthy that no patient with EGPA had Purtscher’s retinopathy. Extra-ophthalmic vascular manifestations included both eosinophilic-driven (e.g. eosinophilc myocarditis or central nervous system involvement, a single patient each) and vasculitic (e.g. mononeuritis multiplex n=18 patients and purpura, n=12 patients) manifestations. Among the 19 patients with available data, only five (26%) had mild elevation of C-reactive protein (CRP) levels (<40mg/L).

The visual prognosis of EGPA patients was poor and the median final visual acuity was 2 [0.1-2.4] logmar. Among the fourteen patients with at least three months of follow-up, only one patient recovered completely and three achieved partial recovery, whereas two patients worsened and eight had a stability of their visual acuity. Moreover, three patients (21%) had a new vascular event during follow-up (lower limb deep vein thrombosis, n=2; pulmonary embolism, n=1). None of the 32 EGPA patients died during follow-up.