Article Citation: Du Y, Pohlmann JE, Chatzidakis S, et al. Quantitative pupillometry predicts neurologic deterioration in patients with large middle cerebral artery stroke. Annals of Neurology. 2025;97:930–941. https://doi.org/10.1002/ana.27178.

Background:

Quantitative pupillometry is a relatively new and increasingly utilized clinical tool that harnesses several metrics of pupillary response to generate the Neurological Pupil Index (NPi). A score of 0 indicates an unreactive pupil, while a score of 5 signifies a normal, reactive pupil. A threshold of 3 or lower is often used to define decreased pupillary reactivity. This study examines whether changes in the NPi predict neurologic deterioration – a question of clinical importance, as early detection could allow timely, targeted medical or surgical intervention.

Methods:

The authors conducted a single-center prospective study of patients admitted to the neurosciences ICU at Boston Medical Center between 2019 and 2024. Eligibility criteria included stroke size more than half of the MCA territory, presentation within 24 hours of stroke onset, and collection of at least three quantitative pupillometry measurements. Pupillometer assessments were performed every 1, 2, or 4 hours by nursing staff, corresponding to the frequency of ordered neurological checks. The primary outcome was the interval from last seen well to neurological deterioration, defined as a decrease in GCS score, increase in NIHSS, or new absence of pupillary reflex. Secondary outcomes included neurological worsening due to cerebral edema. A Cox proportional hazards model was used to investigate the association between NPi values and time to neurological deterioration.

Results:

777 total patients were screened, with 71 patients included in the final analysis; the primary reason for screen failure was stroke size < ½ MCA territory. Neurologic deterioration occurred in 45% of patients, and the majority were female (59%). Among those who experienced neurologic deterioration, 65% had right-sided strokes. Overall, 76% underwent mechanical thrombectomy, decompressive hemicraniectomy was performed in 27%, and mortality was 31%.

Patients with lower NPi had a significantly higher risk of neurologic deterioration (HR 2.46, 95% CI 1.68 to 3.61) and higher rates of cerebral edema. Incorporating additional pupillometry measurements such as dilation velocity into the Cox models improved prediction of time to deterioration, as these patients demonstrated lower average dilation velocity preceding clinical decline. Declines in NPi and dilation velocity occurred within 0–4 hours before deterioration and were not observed in controls (mean NPi change −0.63 vs −0.02 in controls).

The optimal NPi threshold for predicting deterioration was 4.01, and the optimal 12-hour decline in NPI was -0.15. These thresholds demonstrated high negative predictive values (>90%) but low positive predictive values (<30%).

Commentary:

This study is the first to demonstrate that reproducible changes in quantitative pupillometry precede neurological deterioration—beginning up to 12 hours prior, and becoming more pronounced within 4 hours of decline. Serial pupillometry trends appear more clinically informative than any single NPi value or threshold, and incorporating additional parameters such as dilation velocity enhances predictive accuracy. Given the low positive but high negative predictive value of proposed thresholds, these metrics may be most useful to exclude imminent deterioration. The inclusion of a diverse patient population is commendable, though the single-center design and small sample size limit generalizability.

Impact on clinical practice:

By the time poor pupillary reactivity is detectable to the naked eye, interventions for malignant stroke—such as hyperosmolar therapy or surgical decompression—are often too late to prevent irreversible injury. Reliable physiologic biomarkers are therefore needed to predict patient decline. Traditional predictors of decline, such as changes in mental status or focal exam findings, may be delayed, subtle, or confounded by other factors, leading to a reactive rather than proactive approach to management of malignant stroke. Although pupillometry is increasingly utilized in neurocritical care, optimal implementation protocols remain undefined. This study provides important preliminary data to support standardized pupillometry monitoring and highlights its potential role as an early, objective tool to guide timely intervention.