In critically ill patients, pupillary light reflex (PLR) is commonly used to explore brainstem dysfunction. After retinal stimulation from light, part of the fibres transmitting the neural impulse along the optic nerve stimulates the parasympathetic Edinger–Westphal nucleus cells in the midbrain’s pretectal area. The preganglionic fibres from the Edinger–Westphal nucleus travel along the oculomotor nerve and activate the ciliary ganglion, whose postganglionic fibres stimulate the iris sphincter muscle causing pupillary constriction (Fig. 1a).

Fig. 1
figure 1

Anatomy of the pupillary light reflex and graphic chart of the pupillary light response as displayed on a pupillometer screen. a Once the retina is stimulated by light, a neural impulse is transmitted along the optic nerve to the pretectal area of the midbrain and reaches the pretectal olivary nucleus (OPN), whose fibres radiate to the Edinger-Westphal nucleus. Here, the impulse stimulates parasympathetic preganglionic cells, whose fibres travel along the oculomotor nerve and stimulate the ciliary ganglion. The postganglionic fibres of the ciliary ganglion innervate the iris sphincter muscle and cause pupillary constriction. The fibres from the OPN on each side activate the Edinger-Westphal nuclei on both sides, ensuring a bilateral (consensual) pupillary reflex. b The pupillary response curve (pupillogram) represents the variation of the pupillary diameter immediately before, during, and after the light stimulation. The constriction latency is the time interval between the start of the light stimulation and the onset of pupillary constriction. The amount of constriction is the difference in mm between the maximum pupil size during the latency and the minimum pupil size reached at the peak of constriction. During the recovery phase, the pupil dilates again. This phase is slower than the constriction phase and depends both on suppression of parasympathetic stimulation of the pupil sphincter and on sympathetic stimulation of the iris dilator muscle. Pupillometric measures are reported in detail on ESM Table 1

Full size image

Standard PLR (s-PLR) assessment using a penlight is simple and inexpensive, but qualitative and prone to subjectivity [1]. Unlike s-PLR, automated pupillometry (AP) standardises the intensity, the distance from the eye and duration of the light stimulus and displays a quantitative and highly reproducible measurement of the pupillary response.

Automated pupillometry technique

Pupil reactivity should be measured in the dark to avoid interference from ambient light. AP is performed using pupillometers, portable devices including an infrared and a visible light source coupled with a camera. The device measures the baseline pupil size using infrared light, which does not stimulate the retina, then delivers a 3 s flash of visible light and records the pupillary response with the camera (Fig. 1b). Measured parameters generally include maximum and minimum pupil size, constriction latency (Lat), velocity (CV), and percentage (CH), and dilation velocity (DV) (Fig. 1b and ESM Table 1). The NeuroOptics pupillometer summarises these variables in a proprietary parameter, the neurological pupil index (NPi), whose values range from 0 to 5, with normal values being ≥ 3. Modern pupillometers can display a graphic representation of the change in pupillary size during the test, generally known as the pupillogram (Fig. 1b). In addition, they show a video playback of the elicited pupillary reflex, store patient data, and provide a trend of the AP values over time to track the patient’s clinical trajectory, making it a portable, non-invasive monitoring tool at the bedside.

Clinical use of automated pupillometry

Neurological assessment and detection of intracranial hypertension

New-onset unreflective anisocoria frequently occurs in neurologic emergencies and signals neurological deterioration. In severe brain injury from multiple aetiologies, intracranial hypertension (ICHT) may cause transtentorial herniation (TTH) of the medial temporal lobe and ipsilateral pupillary dilation due to either compression of the oculomotor nerve or distortion of the midbrain. The rationale of assessing or monitoring AP in unconscious brain-injured patients is to detect a functional compromise of the midbrain early, to prompt additional investigations and treatment. In a small case series [2], NPi changes preceded clinical deterioration by a median of 7.4 h in 73% of the events and returned to normal at a median of 43 min after treatment. In a single-centre cohort [3] of 54 patients with severe traumatic brain injury (TBI) and predominantly focal injuries who underwent intracranial pressure (ICP) monitoring and repeated AP assessment, episodes of sustained elevated ICP were associated with a concomitant NPi fall. In 17 patients who developed refractory ICHT requiring decompression, NPi was < 3 for 38% of the monitored time vs 1% of the time in 15 patients with non-refractory ICHT. In another study [4] on 23 ICP-monitored patients with nontraumatic supratentorial intracerebral haemorrhage who underwent AP every 30 min (total 1973 measurements), low CV had the highest sensitivity (89[79–95]%) and high Lat had the highest specificity (83[81–85]%) for detecting ICHT. Due to a low incidence of ICHT (total 74/1934 ICP readings in seven patients), the positive predictive values for both CV and Lat were also low (7[6–9]% and 8[5–11]%, respectively). Interestingly, the median values of CV, CH and DV decreased linearly with increasing ICP values from 0 to ≤ 20 mmHg.

In summary, in comatose patients with acute brain injury, deterioration of AP parameters such as NPi and CV on serial AP measurements may herald the occurrence of ICHT and prompt additional investigations, such as brain imaging or ICHP treatment. In that regard, pupillometry can be comparable to other tools such as transcranial Doppler or optical nerve sheath diameter. Because of its quantitative and reproducible results, AP is preferable to the conventional assessment of pupil size and reactivity using visual inspection.

Prognostication in acute brain injury

In patients with cardiac arrest, hypoxic-ischaemic brain injury (HIBI) occurring during and after resuscitation is the leading cause of death and disability [5]. Although there are no clinically available methods to measure cerebral blood flow during resuscitation directly, an indirect estimate of brain perfusion may be provided by pupillary reactivity. In a case series of 30 patients [6], 25 (83%) had detectable PLR using AP for at least part of the resuscitation, and continuous presence—or absence for < 5 min—of PLR during resuscitation was associated with survival and a good neurological outcome. Conversely, no patients with absent or gradually deteriorating PLR during resuscitation survived. An NPI ≤ 2 predicts a poor outcome with 98% specificity on hospital admission in patients who are comatose after return of spontaneous circulation [7].

Midbrain is relatively resistant to anoxia. In comatose cardiac arrest survivors with HIBI, a bilaterally absent PLR indicates a very likely poor outcome [8]. However, a visual pupillary assessment may occasionally miss a present PLR in patients with a good neurological outcome and cause a falsely pessimistic prediction. This is more likely to occur when the pupil size is small. In an observational European multicentre study on 456 adult patients with post-cardiac arrest HIBI [9], PLR on visual assessment was incorrectly rated as being bilaterally absent in five over 78 patients who eventually recovered (false positive rate [FPR] for poor outcome 6[2–14]%) while AP was 100% accurate (FPR 0[0–2]%). In those patients, the mean pupillary size was 1.9 ± 0.22 mm. The current Guidelines for Post-resuscitation Care [10] recommend pupillometry over s-PLR to prognosticate patients with HIBI after cardiac arrest (ESM Fig. 1).

Besides HIBI, AP changes may predict poor outcome in patients with acute brain injury from multiple aetiologies [11]. A multicentre prospective study [12] is investigating the ability of NPi to predict 6-month poor neurological outcome in adults admitted to ICU with acute brain injury from trauma or haemorrhagic stroke.

Limitations of automated pupillomentry

Like s-PLR, AP requires that the mechanism of vision and the afferent and efferent pathways of PLR are intact. Another potential confounder is the interindividual variability in pupil size and reactivity [13] whose impact on AP reliability remains to be investigated. Both anaesthetic agents [14] and opioids at clinical doses [15] that affect s-PLR may also alter some AP parameters. However, NPi remains generally unaffected. AP is more accurate than s-PLR when the pupil size is small, as it may occur due to sedation. However, a consistent threshold of AP values for predicting the correlated clinical event remains to be established. AP requires equipment and consumables, whose costs can make this technology unsuitable for low-resource settings. However, recent smartphone apps for performing AP in these contexts have recently been developed [16].

Take-home message

AP ensures a standard, objective, and reproducible assessment of PLR. In patients with HIBI, AP can detect minimal pupillary reactions that are not detected by visual evaluation and avoid falsely pessimistic predictions. In patients with brain injury at risk of ICHT, AP can be used as a non-invasive track-and-trigger system to detect subtle and progressive changes in PLR that indicate an impending neurological deterioration or increased ICP. Prospective clinical trials are needed to demonstrate the potential clinical benefits of early assessing and monitoring patients with acute brain injury using automated pupillometry.