Task-Evoked Pupillary Dynamics Are Altered in Post-COVID Syndrome, 2026, Smit, Hohberger+

[SNT Gatchaman]

Task-Evoked Pupillary Dynamics Are Altered in Post-COVID Syndrome

Spoiler: Authors

Smit, Alexander; Fleischmann, Philipp; Knauer, Thomas S; Mardin, Christian Y; Michelson, Georg; Zott, Julia; Güttes, Moritz; Sarmiento, Helena; Ilgner, Miriam; Jakobi, Marie; Rech, Jürgen; Hohberger, Bettina

BACKGROUND/OBJECTIVES
Post-COVID syndrome (PCS) is frequently associated with persistent cognitive complaints such as fatigue and impaired concentration, yet objective markers related to cognitive dysfunction are lacking. Pupillary oscillation metrics have emerged as non-invasive indicators of task-related cognitive load and autonomic regulation. This study investigated the Index of Pupillary Activity (IPA) and the Low/High Index of Pupillary Activity (LHIPA) in a large cohort of patients with PCS compared with healthy controls.

METHODS
In this cross-sectional study, 526 participants (397 PCS patients, 129 controls) performed a standardized virtual reality-based stereoscopic task at three disparity levels: 275 arcsec (high difficulty), 550 arcsec (medium difficulty), and 1100 arcsec (low difficulty), using a head-mounted display with integrated eye tracking. Continuous pupillometry data were recorded, and IPA and LHIPA were calculated. Linear mixed-effects models with random intercepts for participants were applied, adjusting for age, sex, and task difficulty.

RESULTS
Both IPA and LHIPA were significantly lower in PCS patients than in controls at all three task difficulty levels in post hoc model-based contrasts. In adjusted mixed-effects models, PCS was also associated with lower overall IPA (β=−0.111, 95% CI −0.160 to −0.062, p<0.001) and lower overall LHIPA (β=−0.164, 95% CI −0.253 to −0.074, p<0.001). Lower task difficulty was associated with higher values of both metrics: for IPA, β=0.164 at 550 arcsec and β=0.287 at 1100 arcsec (both p<0.001); for LHIPA, β=0.161 at 550 arcsec and β=0.254 at 1100 arcsec (both p<0.001), relative to 275 arcsec. Thus, both indices showed an inverse association with task difficulty. Age was negatively associated with both metrics, whereas male sex was positively associated with both. No significant interaction between cohort and task difficulty was observed.

CONCLUSIONS
PCS was associated with reduced IPA and LHIPA during a standardized stereoscopic task. These findings indicate altered task-related pupillary dynamics in PCS and may reflect altered cognitive-load processing and autonomic regulation. LHIPA, and with caution also IPA, may contribute to the objective assessment of task-related pupillary alterations in PCS.

Web | DOI | PDF | Medical Sciences | Open Access

 

[Hutan]

Introduction

Pupillometry is an ophthalmic method based on a non-invasive and quick measurement of pupillary response. Previously, pupillometry has been successfully used in patients with familial dysautonomia and autonomic neuropathies [17]. Recent data suggest that patients with PCS also show an altered pupillary response: a reduction of duration of pupil constriction (n = 25 patients with PCS) [18], an altered pupillary response after different lighting conditions (n = 44 patients with PCS) [19], or a reduction in pupil dilation and constriction (n = 65 patients with PCS, requiring hospitalization during acute COVID-19) [20].

Those studies of the pupillary light reflex primarily assess reflexive autonomic and brainstem-mediated responses to luminance changes, whereas task-evoked pupil fluctuations reflect cognitive effort, arousal regulation, and mental workload during active task performance [21,22]. In PCS, previous pupillometric studies have mainly described abnormalities in light-evoked or basic pupillary responses, including altered contraction dynamics and reduced dilation/constriction responses [18–20]. Task-evoked oscillatory pupil metric seems to be an interesting approach as an objective marker for cognitive dysfunction in PCS.

To assess task-evoked pupil dynamics related to cognitive load, the IPA was proposed as a frequency-based measure of rapid pupil diameter oscillations during task performance [21]. The LHIPA further relates low- and high-frequency components of the pupil signal. Because luminance-related changes and baseline drift typically occur on slower timescales than task-evoked microdilations, these indices aim to emphasize cognitively driven pupil dynamics while reducing contamination by ambient light variation [23].

Recent data suggest that IPA may reflect cognitive impairment in PCS during a virtual 3D task [24]. In addition, subsequent machine learning analyses identified IPA and LHIPA among the most informative features to distinguish patients with PCS from healthy controls [25]. Therefore, the present study examined IPA and LHIPA in a large cohort of patients with PCS and healthy controls during a standardized virtual-reality stereoscopic task with graded difficulty.

By combining two frequency-based pupillary metrics with a controlled task paradigm and mixed-effects modeling, this study aimed to examine whether task-evoked pupil oscillations differ between patients with PCS and healthy controls and whether these measures may serve as objective indicators of altered cognitive load processing in PCS.

 

[Hutan]

Participants


Method

In each stereoscopic stimulus, one of the four balls was rendered closer to the partic-ipant than the other three balls. Task difficulty was manipulated by binocular disparity, using three disparity levels: 275, 550, and 1100 arcsec. The 275-arcsec condition represented the smallest disparity difference that could be implemented with the display geometry of the headset. Smaller disparity values correspond to higher task difficulty because the depth difference between the balls becomes harder to resolve. Participants were instructed to identify the ball perceived as closest and to indicate their response with the corresponding arrow key on a wired keyboard as quickly and accurately as possible.

Each participant completed three runs. At the beginning of each run, all four balls were presented at the same depth (idle phase). After the participant initiated the run by pressing an arrow key, 81 stereoscopic stimuli were presented in randomized order. These stimuli comprised all combinations of 3 disparity levels, 9 stimulus positions, and 3 repetitions. The first run served as familiarization, the second as training, and data from the third run were used for statistical analysis.

The duration of the analyzed third run was defined as the total time required to complete the 81-stimulus stereoscopic task. For each stimulus, the system recorded response time (time from stimulus onset to gtkey press), response correctness, disparity level, and stimulus position. Simultaneously, binocular pupil diameter, gaze direction, gaze origin, and eye openness were recorded continuously during task performance