Brimonidine eye drops reveal diminished sympathetic pupillary tone in comatose patients with brain injury

Comatose patients aged \(\ge\) 18 years and admitted to the ICU for acute brain injury following cardiac arrest (n = 8) or acute cerebrovascular disorders (n = 7) were enrolled on a convenience basis when residual consciousness for all practical purposes was ruled out according to clinical exam, neuroimaging (e.g., CT post-cardiac arrest showing global edema with absence of white/gray matter distinction), and/or EEG (e.g., absence of EEG background activity). Exclusion criteria were active eye disease or a history of pupillary injury (e.g., cataract); patients treated with mono-oxidase inhibitors or tricyclic antidepressants due to possible interaction with brimonidine; “high or very high” levels of sedation (see below); and a planned brain death/organ procurement protocol (to avoid interference with brainstem reflex testing).

Healthy volunteers, matched for age and sex, were recruited from January to March 2022 via local advertisement. Exclusion criteria were active eye disease or a history of pupillary injury and mono-oxidase inhibitors and/or tricyclic antidepressants.

Before pupil measurements, a neurological examination was performed by or under supervision of a board-certified neurologist with > 15 years of experience (DK), which included (but was not limited to) a detailed evaluation of the cranial nerves and brainstem reflexes, Full Outline of UnResponsiveness (FOUR) score [32], and assessment of spontaneous movements (e.g., eye-opening, movement of extremities).

Room illuminance was measured (in lux) near the eyes using a conventional iPhone and the application “LUX Lightmeter” to ensure that illuminance remained constant throughout all measurements. Ambient light was adjusted/reduced to ensure scotopic conditions (≈25 lx). Pupil size was measured in both the eyes using automated pupillometry (NPi-200 Pupillometer, NeurOptics, USA). Measurements were taken at baseline (T0) before administering a single drop of brimonidine (2 mg/mL) into the right eye to block sympathetic tonus. Measurements were then repeated 5, 10, 20, 30, and 120 min (T5, T10, T20, T30, T120) after intervention, to demonstrate the presence or the absence of anisocoria and miosis. At each time point, the measurement of pupil size was repeated three times for each eye, and mean pupil size was calculated to account for variation.

Clinical data including cause of admission, MRI, CT, EEG, level of sedation, if applicable, and final diagnosis were collected. Levels of sedation were categorized as “none” (no sedation given during or before the examination); “low to moderate” indicating remifentanil < 15 µg/kg/h, propofol < 2 mg/kg/h, fentanyl < 5 µg/kg/h, sevoflurane < 3%, and midazolam < 0,15 mg/kg/h; and “high to very high” indicating remifentanil > 15 µg/kg/h, propofol > 2 mg/kg/h, fentanyl > 5 µg/kg/h, sevoflurane > 3%, and midazolam > 0,15 mg/kg/h [4].

Primary outcomes were the following: (1) within-eye difference, measured as mean difference change in pupil size from T0 to T30; (2) between-eye difference, measured as mean difference between the treated (right) and the non-treated (left) pupil in each subject at T30; and (3) between-group difference, measured as a comparison of mean difference change in treated (right) pupil size from T0 to T30 between comatose patients and matched controls. Secondary outcome was correlation of changes in pupil size from T0 to T30 with the FOUR score. T30 was chosen because miosis typically occurs within 30 min after brimonidine administration [12].

Data were assessed by q-q-plot and histogram to check for normal distribution. Baseline and 30-min measurements were not normally distributed, so non-parametric tests were used for further analysis. The primary outcome of within-eye difference was analyzed using Wilcoxon signed-rank test, while the primary outcomes of between-eye and between-group differences were analyzed using Wilcoxon rank-sum test. The secondary outcome (correlation between FOUR score and pupil reactivity) was assessed by Spearman rank correlation. We further performed sensitivity analyses using linear regression models to assess the effect of age, baseline pupil size, and room illuminance on the pupil size at the 30-min measurement, as well as subgroup analysis of the effect of brimonidine after 120 min and another sensitivity analysis of all outcomes, excluding comatose patients with a baseline measurement < 3 mm. Significance was set at p < 0.05. Primary outcomes were adjusted for multiple comparisons (n = 3) by the Bonferroni correction, and adjusted significance was set at 0.05/3 = 0.016. p-values are given after Bonferroni correction. All analyses were performed using R (version 4.2.0., R-packages “summarytools,” “cowplot,” “ggplot2,” “lattice,” “ggpubr,” and “tidyverse”).

The study was approved by the Ethics Committees of the Capital Region of Denmark (file number: H-19044446) and by the Knowledge Center of Data Reviews (file number: P-2021–879). Informed written consent was obtained from the patients’ next-of-kin and a trial guardian (the attending anesthesiologist), as well as from healthy volunteers.

Data is available upon reasonable request.

Fifteen comatose patients (mean (SD) age 59 (13.8) years; seven women) and fifteen healthy volunteers (mean (SD) age 55 (16.3) years; seven women) were included (Table 1). All patients had highly pathological brain injury on CT (“highly pathological” being defined as Fisher grade ≥ 3 [subarachnoid hemorrhage], hemorrhage volume ≥ 30 mL [intracerebral hemorrhage], strategic hemorrhage or infarct in brainstem [ischemic stroke or infratentorial hemorrhage], or any visible sign of anoxic brain injury including global or cortical edema [cardiac arrest]). Nine patients had EEGs done, all of which were highly pathological with suppression-burst or continuously suppressed (< 10 µV) background activity. Following a discussion with the patients’ next-of-kin, a decision to withdraw life-sustaining treatment was eventually made in all 15 patients. The mean time from pupillary examination to death was 2.6 days (range: 0–16) with all immediate causes of death being withdrawal of life-sustaining therapy. At baseline (T0), mean (range) FOUR score was 4 (0–9) and mean (range) GCS 3.6 (3–7). During the examination, one patient received high levels of sedation (a protocol violation), six patients received low to moderate levels of sedation, while the remaining eight patients did not receive any sedation. Five patients required inotropic support with norepinephrine before or during examination (Table 2).

In comatose patients, the change of pupil size from T0 to T30 (i.e., the pupillary reactivity to brimonidine) did not correlate with the FOUR scores (R =  − 0.3, p = 0.28; Fig. 5). There was a significant difference in pupil size change from T0 to T30 and T120 in the control group compared to the comatose group (Table 3). This difference was not affected by age or room illuminance. An association was observed between pupil size at T0 and the change in pupil size at T30 and T120, with larger pupil size at T0 resulting in a greater difference at T30 and T120 (Table 3).

All healthy volunteers had a baseline pupil size > 3 mm, while only nine of the comatose patients had a baseline pupil size > 3 mm. To account for this difference, we did a sensitivity analysis of all outcomes, excluding comatose patients with a baseline measurement < 3 mm (n = 6). This analysis showed that the control group still had a larger decrease in pupil size compared to the comatose group with a mean difference between groups of 1.16 mm, 95% CI: [1.08; 1.23], p < 0.001. The comatose group showed a significant change in pupil size from T0 to T30 in the treated eye, yet the effect of brimonidine was still small with a mean change of − 0.37 mm, 95% CI: [− 0.52; − 0.21], p = 0.012. Importantly, the difference between the treated and the non-treated eye remained non-significant at T30 with a mean difference of 0.13 mm, 95% CI: [− 0.21; 0.47], p = 0.93. Thus, the significant difference in the treated eye also occurred in the non-treated eye (Supplementary Table S1).

This study has limitations. It appears that the pupil did constrict in some of the comatose patients after brimonidine (Fig. 4A), but this effect if present was minor and temporary compared to that in awake volunteers (Fig. 4B). There are several possible explanations. First, tonic sympathetic tone of the dilator muscle of the iris depends on the viability of brainstem structures that activate the preganglionic sympathetics in the upper spinal cord [24, 25]. It is possible that with these cases, there was a small residual activation of the dilator from brainstem centers. For example, the locus coeruleus and other brainstem nuclei have descending projections to the intermediolateral cell column [24, 25]. These brainstem pathways likely remained partially intact in at least some of the cases we studied. We cannot rule out that a change in the intracranial pathology occurred over the 2-h study period, but we deem this unlikely. Second, it might be that the alpha-1 receptor at the dilator muscle had developed a small degree of denervation hypersensitivity during the days after the insult and before the study was conducted [28]. Even a small release of norepinephrine from the postganglionic sympathetic nerves might then produce a residual activation of the dilator that was blocked by brimonidine. Third, brimonidine also decreases aqueous humor production and thereby decreases intraocular pressure, which comes about by a direct alpha 2 agonist effect on the blood vessels. This effect is not related to the alpha 2 activity that prevents norepinephrine release from the sympathetic nerves. A decrease in intraocular pressure might therefore result in a very small decrease in pupil diameter, and in fact, the use of apraclonidine (instead of brimonidine) to diagnose Horner’s syndrome may circumvent this physiological effect [29].

The pupillary responses we report were highly variable. However, we purposely studied a diverse population of unconscious subjects to avoid limiting our study to a specific brain injury. While this did produce variable pupillary responses, our statistical analysis of this exploratory study confirmed our hypothesis that the unconscious state is associated with lack of tone in the dilator muscle of the iris. A larger cohort is needed to specifically investigate pupillary responses in coma related to strategic brainstem lesions, bi-hemispheric damage, and global metabolic/anoxic compromise, respectively.

The comatose patients had various types of brain injury because they were selected to reflect a real-life ICU setting of acute brain injuries. Since the aim was to test a consciousness biomarker that would be meaningful and reliable across the entire range of brain injuries encountered in the ICU, the comatose group was, as stated above, heterogeneous on purpose. In other words, a consciousness marker that only would work in, say, patients with a focal frontal lesion owing to traumatic brain injury, but not in patients with parietal lesions or those after cardiac arrest, would be of little practical value. We acknowledge that the cohort was small, and a larger prospective validation trial is needed to confirm that the results are generalizable, but—as stated—in this pilot study, we focused on the extremes of the consciousness levels to test if our primary hypothesis could be falsified. This approach required a comparably smaller sample size.

The challenging setting in the ICU, with critically ill patients who are subject to sudden deterioration, also resulted in missing values for the 120-min measurement for five of the 15 patients, but we do not think that the lack of this data compromises the overall findings of the study. Also, despite our exclusion criteria, one patient received high levels of sedation at the time of pupillary exam, which was a protocol violation, but we decided to keep this patient in the cohort; nevertheless, given the naturalistic ICU setting. Critically ill patients with acute brain injuries typically require use of sedatives, inotropes, and opioids, and increased intracranial pressure may affect third cranial nerve functions. Notably, coma patients with small pupils were mostly those that received remifentanil, and opioid constricted pupils have a tightly constricted sphincter muscle. However, sensitivity analysis for baseline pupil size showed that the brimonidine effect was minimal in comatose patients with pupil sizes of 3 mm or more. Still, all these factors must be considered when pupillometry is used for prognostication, as they can have a confounding effect [21], but a method to identify residual consciousness in the ICU that would be prone to major artifacts would have very limited clinical utility.

We assessed comatose patients with the FOUR score, which works well with intubated patients [32]. Scores ranged from 0 to 9 points, indicating a difference in wakefulness. However, FOUR scores were not significantly correlated to the difference in pupil size between baseline and 30 min after intervention, possibly owing to the small sample size and lack of statistical power.

Finally, the presence of sympathetic tone may be necessary for consciousness but may not be sufficient. However, all sympathetic tone is not abolished during unconsciousness. We here are referring only to sympathetic tone in the dilator muscle of the iris. Patients with a preexisting Horner’s syndrome could be conscious but would not constrict following topical administration of brimonidine. Also, patients with tonic pupils might not exhibit any change in size after brimonidine. Because no tests are perfect 100% of the time, a more complete study is required to evaluate the specificity and sensitivity of the brimonidine test for the unconscious state.

The major strengths of this study are the age- and sex-matched groups, control of room illuminance, which ensured scotopic conditions, and repeated measurements over 120 min with automated pupillometry. These factors are important as age can affect pupil rigidity, and brimonidine has a larger effect under scotopic conditions, so ensuring low room illuminance was crucial [12]. With repeated measurements, the probability of overlooking an even short-lasting miotic effect was diminished. Also, for the purpose of this proof-of-concept study, even though evoked potentials and advanced fMRI/EEG-based consciousness paradigms were not performed, we ensured that clinical exam, EEG, and neuroimaging had rendered the presence of residual consciousness close to impossible, despite no or limited sedation. Furthermore, linear regression models revealed that our results remained robust when corrected for baseline pupil size, age, and room illuminance.