Background

Central retinal artery occlusion (CRAO) is a devastating form of acute ischemic stroke [1] that usually causes permanent visual loss in one eye. While a majority of acute ischemic strokes affect the brain (cerebral ischemic stroke), a small minority (<1%) affect the retina and – although disabling – are typically recognized later than cerebral ischemic strokes. When the retina’s main source of metabolic substrates – the central retinal artery – is occluded, rapid loss of visual function ensues. Only a minority of affected patients recover a best-corrected visual acuity (BCVA) of 20/100 or better [2], resulting in long-term quality of life impairments with respect to overall social functioning, employment, mental health, and level of dependency, and a propensity to falls in the elderly [3,4].

Clinical framework

Intravenous (IV) tissue plasminogen activator (tPA; which almost exclusively refers to alteplase [5] or tenecteplase [6]) is effective in improving functional status after cerebral ischemic stroke [7]. When a clot blocks an artery, IV tPA can break up the clot, restore blood flow (i.e., achieve reperfusion) and increase the chance of salvaging blood-starved tissue [8]. The odds of reperfusion are associated with a the size of the occluded blood vessel and the length of the occluding particle [9]. The central retinal artery has a diameter of only 160um at the optic nerve head [10,11] and is thus ideally poised to benefit from IV tPA. The risk of developing atrial fibrillation (AF; the most common arrythmia in the US and a cause of blood clots) after CRAO is equivalent to that seen in matched patients with cerebral ischemic stroke and higher than in controls [12]. Furthermore, patients with CRAO also have simultaneous cerebral ischemic stroke approximately 20–30% of the time when magnetic resonance imaging (MRI) of the brain is performed concurrently) [13–15], although overt neurological symptoms occurring in tandem with CRAO are less common. These observations argue that similar processes may underpin both disorders.

A recent systematic review and meta-analysis using a standardized definition of visual recovery (attainment of a final BCVA of ≥20/100) found that use of IV tPA improves the proportion of acute CRAO patients with a favorable outcome [16]. In this study, 37.3% of patients treated with IV tPA recovered functional vision in the affected eye, compared with 17.7% of untreated patients. The risk of symptomatic intracranial hemorrhage (the most feared side-effect of IV tPA) was low, with no published cases when treatment was given within 4.5 hours of time last known well (LKW) at the time of this analysis. In the TenCRAOS clinical trial, there was one sICH in a participant that was treated with tenecteplase (presented at ESOC in May 2025).

Biological framework

The retina requires a constant supply of substrates to maintain tissue viability [17]. As the chief conduit for these metabolic substrates, if the central retinal artery is blocked retinal viability is temporarily maintained by i) passive diffusion of oxygen from the outer retina and choroid and ii) collateral flow from the choroidal circulation. By analogy to cerebral ischemic stroke, there is a retinal “ischemic penumbra” with observed anoxic, hypoxic and normoxic compartments [18]. Pre-clinical data support 4.5 hours as a rational time cutoff for IV tPA administration in CRAO. In non-human primates, occlusion of the central retinal artery proximal to its entry into optic nerve for more than 105 minutes causes severe retinal tissue injury [19–21]. However, in older animals with experimentally-induced chronic hypertension and an atherogenic diet, irreversible tissue loss does not occur until 240 minutes after occlusion [22]. This suggests that in elderly people at high vascular risk, the retina may be more robust to transient impairments in blood flow [23].

The potential benefit of IV thrombolysis for CRAO remains an area of active discussion. Concerns about the use of IV thrombolysis in CRAO include the possibility that IV thrombolysis does not reach the occluded retinal vessel in meaningful concentrations, that collateralization may be less robust that in the cerebral circulation, and that other mechanisms of CRAO – including calcific emboli, intimal dissection, or arteritic CRAO – would not benefit from IV thrombolysis [24].

Rationale for individual participant data meta-analysis (IPMA)

Three randomized, controlled, trials (RCTs) enrolling patients with acute non-arteritic CRAO are either complete or at an advanced stage of conduct: THEIA, TenCRAOS, and REVISION. The THEIA trial (NCT03197194) randomized 70 patients with acute CRAO within 4.5 hours of time LKW to either IV alteplase or 300 mg acetylsalicylic acid (ASA) delivered orally. A higher proportion of patients randomized to IV thrombolysis achieved the primary end point but results were not statistically significant [25]. The TenCRAOS trial (NCT04526951) is at an advanced stage of conduct and preliminary results were presented at the European Stroke Organization Conference (ESOC) in May 2025. The REVISION trial (NCT04965038) is actively enrolling with a potential sample size up to 422 participants, and approaching the first interim analysis (120 participants) as of the time of this submission. In the early phase of REVISION, alteplase was the thrombolytic employed, the protocol was amended to use tenecteplase instead. The primary comparator in REVISION is IV placebo.

An IPMA confers several benefits that are relevant to the present clinical issue. First, it would permit derivation of a pooled estimate of treatment effect from the 3 trials. Such an estimate is expected to be more precise than an estimate derived from any individual RCT. Secondly, each trial may individually be considered underpowered to detect a small but clinically meaningful effect size. Pooling data from the three trials will enhance statistical power to assess the effect of IV tPA on clinical outcomes. Third, an IPMA will allow exploration of individual patient-level sources of heterogeneity in a manner that would not be possible in a study-level meta-analysis (owing to the phenomenon of ecological fallacy). Fourth, signals of safety concerns are more likely to manifest in a larger cohort of study participants [26] and may provide important context to pooled estimates of efficacy. Finally, IPMA is considered the highest echelon of evidence with respect to evidence grading and this study may be of use to professional bodies in updating guideline recommendations for CRAO [27].

Study objectives

Constituent study objectives are outlined in Table 1 (overleaf). The primary study objective is to address the following question, reflecting a common dilemma in clinical practice:

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Study Objectives.

https://doi.org/10.1371/journal.pone.0342739.t001

“In adults ≥18 years old with acute, non-arteritic, CRAO and disabling visual symptoms (a BCVA of worse than 20/100 at onset; population), does treatment with IV thrombolysis (either alteplase or tenecteplase) within 4.5 hours of time LKW (intervention) increase the odds of achieving a final BCVA of ≥20/63 (outcome) when compared with no IV thrombolysis, placebo, and/or oral anti-thrombotic therapy (comparator)?”

Study design

This systematic review and IPMA is prospectively registered through PROSPERO (#1154900). We will report it in accordance with the PRISMA-IPD reporting guideline [28,29]. Case record forms from individual trials, analytic code, and summary results of individual clinical trials will be disseminated as Supplemental Material to the main manuscript. The extracted data from studies may be made available upon request to the corresponding authors of each individual clinical trial, in accordance with any applicable regulations and data sharing policies. No automation tools will be used in the screening or data extraction processes. This study has been deemed exempt by the Duke Institutional Review Board (Pro00117992) as this is deemed non-human subjects research given the de-identified nature of the underlying data.

Eligibility criteria

We will include only RCTs that enroll patients with acute, non-arteritic CRAO who are randomized and treated within 4.5 hours of time LKW. We will permit inclusion of trials where the comparator is no IV thrombolysis, placebo, and/or oral anti-thrombotic therapy. RCTs that also permit the deployment of conservative therapies [30] will only be included if use of such therapies is permitted in both arms (treatment and control). We will exclude trials that enrolled fewer than 20 patients (because such small trials tend to entail bias and provide unstable estimates of treatment effect) as well as any trials where IV thrombolysis was given alongside another trial-mandated intervention (medical or surgical therapy).

Information sources and search strategy

We will search MEDLINE (via Ovid), Embase (via Elsevier), the Cochrane Library (via Wiley), Web of Science – Science Citation Index and Social Science Citation Index (via Wiley), and the ClinicalTrials.gov registry from inception of each registry through June 30th, 2026. We will use database-specific subject headings and keywords related to IV thrombolysis, IV tPA, alteplase, tenecteplase, and CRAO (and its permutations). No restrictions will be placed on the publication year. We will place no limitations on language. An experienced medical librarian will devise and conduct the searches, with input on keywords from other authors. The search strategy will be independently peer-reviewed by another librarian using a modified Peer Review of Electronic Search Strategies (PRESS) checklist [31]. We will hand-search the bibliographies of included studies. We will not search literature produced outside of traditional publishing avenues such as unpublished conference materials, theses, government or commercial white papers, or technical reports (sometimes referred to as the “grey literature”). Completed trials identified through ClinicalTrials.gov will be included). The full, reproducible search strategy for each included database will be disseminated as part of the Supplemental Material to our primary manuscript.

Screening process

All unique citations will be imported into Covidence [32], a web-based systematic review platform for screening. Two investigators will independently screen all titles and abstracts. Next, potentially eligible records will be independently screened by two investigators. At both stages, conflicts will be resolved through discussion and arbitration by a third study member where necessary.

Data collection process

Once all trials that meet our study criteria re identified, a minimum dataset will be designed by the collaborating authors and reviewed by the present study’s statistician. Subsequently, this minimum dataset will be provided to each trial statistician. Trial statisticians will extract participant-level data by direct access to study databases. At a minimum, fully de-identified data from the THEIA, TenCRAOS, and REVISION trials will be provided to Duke University under a data transfer and utilization agreement (DTUA) between Duke University and the primary institution associated with each RCT. If further RCTs that meet our selection criteria are identified, we will contact the corresponding author of the trial manuscript to request individual participant-level data.

Two investigators will independently inspect, clean, and collate the datasets. Indicator variables will be created where necessary. All BCVA measurements will be harmonized and converted to logMAR per Table 2 (overleaf). Discrepancies identified at this point will be resolved through discussion, and we will maintain communication with each trial’s primary statistician in case queries arise during data cleaning and collation. A single data frame will be created from the pooled source trial datasets containing indicator variables identifying the source trial for each participant. The resulting data frame will represent the master dataset for the present study, and imputation of missing data will only be performed after this step. As a form of quality control, descriptive and inferential statistics from the parent RCT will be re-calculated and cross-checked against conference presentation results and any corresponding publications.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Best-corrected visual acuity conversion algorithm (Adapted from Schulze-Bonsel et al., 2006 [33]).

https://doi.org/10.1371/journal.pone.0342739.t002

Data items

At a minimum, the following data elements will be included in our study:

1. 1) Study-Level Parameters

Study design, blinding method, comparator, randomization procedure, country/countries of study conduct, total number of patients included.

1. 2) Participant-level parameters

• Baseline Attributes:

Age, biological sex, National Institutes of Health Stroke Scale (NIHSS) score on presentation, presenting BCVA, past medical history (including hypertension, hyperlipidemia, diabetes mellitus, atrial fibrillation, tobacco exposure, and chronic kidney disease). We will collect data on race and ethnicity, where possible, for descriptive purposes. However, we anticipate that because of legal restriction in certain countries where constituent trials are conducted, this will likely not be possible for all participants. Therefore, we do not anticipate being able to include these parameters within any modeling or subgroup analyses.

• Treatment Parameters:

Time of symptom onset, time to emergency department (ED) presentation, time of randomization, treatment assignment (intervention versus control), time of intervention delivery (if delivered successfully), IV thrombolytic agent used (alteplase or tenecteplase), dose of IV thrombolytic agent received, concurrent therapies employed (including ocular massage, hemodilution, anterior chamber paracentesis, topical or systemic intraocular pressure-lowering drugs, or hyperbaric oxygen therapy), and control/active control used (placebo, anti-thrombotic therapy, or other).

• End Points:

Follow-up BCVA, final BCVA, modified Rankin Scale score (mRS), symptomatic intracranial hemorrhage (sICH; per the modified Safe Implementation of Thrombolysis in Stroke-Monitoring Study [SITS-MOST] definition which includes any new intracranial hemorrhage associated with a decline of ≥4 points on the National Institutes of Health Stroke Scale (NIHSS) score [36]) occurring within 36 hours of thrombolysis administration, any ICH, orolingual angioedema, gastrointestinal hemorrhage, other extracranial hemorrhage, and mortality. We will ascertain the precise definition of each complication employed as we anticipate that this will vary by trial.

Risk of bias in individual studies

Two investigators will independently assessed risk of bias in individual studies using the Cochrane Risk of Bias 2.0 tool for interventional studies [37]. Discrepancies in risk of bias assessments will be resolved through discussion. Risk of bias will be presented in graphical and narrative form using established methodology.

Synthesis methods

Key participant, and study-level attributes (“Data Items”, above) will be disaggregated by intervention status (treatment versus control), presented in tabular form, and synthesized narratively. Counts and percentages will be used to summarize categorical variables and medians with first (Q1) and third (Q3) quartiles will be used to summarize the distribution of continuous variables. We will calculate absolute standardized differences to illustrate differences across groups with respect to baseline participant- and study-level attributes.

Publication bias assessment

To investigate the presence of reporting biases/small study effects, we will create funnel plots for each outcome and visually inspect the plot for evidence of asymmetry. This is likely to be highly exploratory given the low numbers of anticipated studies in this analysis.

Grades of recommendation assessment development and evaluation (GRADE) assessment

We will prepare an evidence profile, assess the overall quality of the evidence, and offer a recommendation for future clinical practice using the GRADE framework [47].

Dissemination plan

We will submit a manuscript arising from this analysis to an international, peer-reviewed journal and present the results at one or more scientific conferences (such as the International Stroke Conference, European Stroke Organization Conference, or World Stroke Congress).