Assessment of Binocular Vision Function With Neos™ (ROCKET) (ROCKET)

The investigation uses a prospective, multicentre, cross-sectional, paired within-participant design. Each participant receives both assessments on the same day and serves as their own comparator, reducing the influence of differences between participants. There is no random allocation. Although the investigation is not formally blinded, clinicians performing the reference examination do not have access to automated neos™ results when determining the clinical classifications. Investigational results do not influence treatment decisions.

The investigation comprises two sequential phases. An initial tuning phase includes 10 selectively recruited participants: five with preserved fusion and stereopsis without suppression, and five with absent fusion and stereopsis and suppression present. These data are used to refine algorithm parameters and thresholds. The algorithm configuration is then locked and remains unchanged throughout the validation phase. Tuning-phase participants are excluded from all analysis sets.

During validation, eligible patients are recruited consecutively when study personnel and procedures are available. Recruitment includes patients with and without suspected or confirmed binocular dysfunction, reflecting the population undergoing clinical binocularity assessment. Approximately 174 participants are expected to enter validation to obtain around 157 evaluable participants. The final enrolment depends on the observed distribution of clinical reference classifications and the proportion of evaluable examinations.

The investigational configuration uses neos™ software version 4.0.0-rocket, incorporating additional binocular visual stimuli and algorithms. The examination typically takes less than 15 minutes. It includes gaze holding and ocular alignment, stimuli for assessing suppression, fusion and stereopsis, pupillary function, saccades, smooth pursuit and visual-field assessment.

The clinical reference examination is performed by an appropriately trained healthcare professional using established methods selected according to clinical findings and participant cooperation. These may include prism bar testing, Bagolini striated lenses, Worth 4-Dot testing and standardized stereotests. Not every test is required for every participant; suitable reference information must be obtained for each primary classification. When specified test results are unavailable, unusable or omitted for clinical reasons, the protocol allows the orthoptist's clinical diagnosis to determine the reference classification.

Prespecified rules define the reference outcomes. Sensory fusion is present when binocular single perception is demonstrated using Bagolini striated lenses or Worth 4-Dot testing. Stereopsis is present when stereoscopic disparity of 600 arcseconds or finer is detected. Central suppression is classified using a hierarchical rule: it is absent when stereopsis or sensory fusion is present. When both are absent, suppression is present if clinical testing demonstrates failure to perceive input from one eye; otherwise, it is classified as absent.

Diagnostic performance is assessed separately for each primary endpoint using a two-by-two comparison of neos™ and clinical reference classifications. Sensitivity and specificity are reported with two-sided 95% exact Clopper-Pearson confidence intervals. An endpoint meets its prespecified success criterion when the lower confidence bounds for both sensitivity and specificity exceed 65%. There is no composite success criterion across the three endpoints and no multiplicity adjustment. Agreement on the affected eye is summarized descriptively among participants classified as suppression-present by both methods.

Validation recruitment ends when all prespecified reference-class requirements are met: at least 41 evaluable participants without sensory fusion, 73 with stereopsis, 73 without stereopsis and 22 with suppression. An operational review at approximately 80 evaluable participants, or earlier if indicated, examines reference-class prevalence to update enrolment projections. It does not examine agreement between neos™ and the reference assessment. No formal interim efficacy or futility analysis is planned.

Secondary analyses compare neos™ fusion break-off angles with clinical prism-bar measurements using Bland-Altman analysis. Agreement in stereopsis severity categories is assessed using weighted Cohen's kappa. Suppression-related outputs and other eye-movement, pupil and visual-field measurements are characterized descriptively, with exploratory comparisons across clinically defined groups. No formal performance goals apply to these secondary analyses.

Primary analyses include participants with conclusive results from both methods for the endpoint concerned. Missing classifications are not imputed. Participants may contribute to one endpoint analysis even if another endpoint is not evaluable. Evaluability rates and reasons for missing or inconclusive results are reported. Safety assessments include validation-phase participants who begin the neos™ examination. No routine follow-up visit is planned; ongoing adverse events are followed as clinically appropriate until resolution or stabilization.