Quantitative pupillometry and the Neurological Pupil index

1. Bedside vignettes: why this matters in the PICU#

Vignette A. The 9-year-old TBI on Q1 hour pupil checks#

A 9-year-old is admitted with severe TBI after a fall, ICP monitor in place, GCS 6T. The bedside flow sheet shows hourly pen-torch pupillary exams (4 mm bilateral, brisk) for 8 hours. The pupillometer is added to the routine at hour 9. The first NPi readings are 4.2 on the right, 4.1 on the left, normal. Six hours later: right NPi 3.1, left NPi 4.0 (asymmetry > 0.7 is abnormal even when both are above 3). The clinical pen-torch exam still reads "4 mm bilateral, brisk." An urgent CT shows new right uncal swelling and early midline shift. The team hyperosmolises and accelerates the decompressive craniectomy decision. The pupillometer caught it before the eye did. 1 2

Vignette B. The 5-year-old after cardiac arrest, day 3#

A 5-year-old is on day 3 after an out-of-hospital arrest, normothermic, sedation off. Bilateral NPi = 0 with absent constriction velocity, latency unrecordable. Bedside FOUR score = 1 (motor only, extensor). SSEP shows bilateral absent N20. MRI shows diffuse cortical and basal ganglia injury. The multimodal prognostication framework (NPi + FOUR + SSEP + MRI + 72 h post-arrest timing) supports the conversation with the family about prognosis. No single number, including NPi 0, is enough by itself; the value of the pupillometer here is one quantitatively reliable leg of the multimodal prognostication stool. 3 4

Vignette C. The 2-month-old with normal NPi and dilated pupils on hyoscine#

A 2-month-old with bronchiolitis is on a high-dose hyoscine patch for secretions and on a one-time dose of glycopyrrolate before fundoscopy. Hours later her pupils measure 6 mm bilateral and "non-reactive" on pen-torch. The team panics, calling neurology. The pupillometer reads: max pupil 6.2 mm, min 6.0 mm, CV 0.0 mm/s, NPi 0. But the brainstem reflexes (corneals, cough, gag) are intact, the GCS is at her pre-illness baseline, and there is no other concerning sign. NPi 0 from pharmacological mydriasis looks identical to NPi 0 from CN III failure. Always interrogate the drug list before interpreting a flat pupillometer trace. 5

2. What pupillometry is, and what it is not#

The pupillometer is a handheld infrared video camera that delivers a calibrated light pulse (typically 1000 lux, 0.8 s) to the pupil and records the diameter at high frame rate (30 to 60 fps) for the constriction and dilation phases. The device extracts:

• Maximum pupil diameter (MAX): pre-stimulus baseline, in mm.
• Minimum pupil diameter (MIN): peak constriction, in mm.
• Percent change (% CH): (MAX − MIN) / MAX × 100.
• Constriction velocity (CV): average rate of constriction in mm/s.
• Latency (LAT): time from stimulus onset to start of constriction, in ms.
• Dilation velocity (DV): average rate of return to baseline, in mm/s.
• Neurological Pupil index (NPi): a proprietary algorithm that combines the above into a single 0 to 5 score with reference to a healthy-population model.

The NPi is not the same as constriction velocity, and the two move independently. A blown CN III pupil shows NPi 0 with CV 0; a healthy pupil shows NPi 4 to 5 with CV ~1 to 2 mm/s; an early-warning pupil with subtle CN III compromise may show NPi 2.5 to 3.0 with CV slightly reduced and LAT slightly prolonged before the pen-torch exam picks up anything. The 0.7 unit asymmetry threshold is the most sensitive single signal of unilateral early uncal herniation.

Two things follow.

Pupillometry quantifies what your pen-torch already does. The pen-torch returns "brisk / sluggish / fixed" with moderate inter-rater agreement (kappa ~0.4 to 0.6). The pupillometer returns continuous numbers with near-zero inter-rater variability (kappa > 0.9). The qualitative-to-quantitative move is the single biggest reason to add a pupillometer to your unit. 6

The NPi is calibrated against an adult healthy population. Pediatric reference data (Freeman 2020, Kerscher 2023, Jiang 2023, Kirschen 2025) suggest the adult thresholds (NPi < 3 abnormal) hold reasonably in children > 1 month, but normative ranges in neonates and preterm infants are different and a "low" NPi in the first weeks of life may not have the same prognostic weight. 5 7

3. Anatomy and pathway: from light to pupil#

Fig. 1

The pupillary light reflex traverses a four-neuron arc.

1. Afferent (CN II): retinal ganglion cells → optic nerve → optic chiasm → optic tract → pretectal nuclei of the midbrain.
2. Inter-neuron: pretectal nucleus → bilateral Edinger-Westphal nuclei (which is why the consensual response exists).
3. Efferent parasympathetic (CN III): Edinger-Westphal nucleus → CN III preganglionic fibres → ciliary ganglion → short ciliary nerves → iris sphincter (constriction).
4. Efferent sympathetic: descending hypothalamic → ciliospinal centre (C8 to T2) → superior cervical ganglion → long ciliary nerves → iris dilator (dilation, also opposed when constriction is forced by light).

Where the lesion sits determines the pupillometer pattern.

This anatomy is why a unilateral falling NPi in a TBI patient is the bedside signature of evolving uncal herniation hours before the pen-torch sees it. The pupillometer detects subtle CV reduction and LAT prolongation while the visual estimate of "pupil size and reactivity" still reads normal. 10

4. The signal: what a typical NPi waveform looks like#

A normal trace has six readable features.

1. A stable MAX pre-stimulus baseline, recorded over ~1 s before the light pulse.
2. A short latency (200 to 300 ms) between stimulus onset and the start of constriction. LAT > 350 ms is abnormal.
3. A brisk constriction phase at CV ~1 to 2 mm/s. CV < 0.6 mm/s is abnormal.
4. A deep nadir (MIN) giving > 10% change from MAX in healthy adults, > 13% in healthy children.
5. A smooth re-dilation phase at DV ~0.5 to 1.0 mm/s. Abnormally fast dilation can suggest sympathetic overactivity (sepsis, pain).
6. A return to baseline within 3 to 5 s.

In a patient with early uncal herniation, the latency and CV are usually the first to drift before the MAX, MIN, or % CH show change. This is the basis of the NPi's early-warning utility: it composites these subtle changes into a single trending number.

5. The numbers to record#

Record both sides, every time, in the same order. Most units document the trend graphically on the bedside chart alongside ICP, CPP, and GCS.

6. What is normal? Age-banded reference values#

Sources: 5 7 8 9 6. Pediatric data are still emerging; the adult thresholds (NPi < 3 abnormal, < 2 strongly abnormal) appear to hold for children > 6 months.

7. What is abnormal? Pattern library#

Decision tree: "what is the NPi telling me?"#

flowchart TD
  Reading[Pupillometer reading] --> Abn{NPi < 3 or asymmetry > 0.7?}
  Abn -->|No| Normal[Continue routine]
  Abn -->|Yes| Drug{New anticholinergic or opioid?}
  Drug -->|Yes| Pharm[Pharmacological; reassess]
  Drug -->|No| Trend{Trend over last 4 h?}
  Trend -->|Stable abnormal| Care[Document, increase frequency]
  Trend -->|Falling| Esc[Escalate: imaging, ICP, NSG]
  Esc --> Action[CT, hypertonic, head up, NSG]

8. Try it: interactive widgets#

9. Pupillometry-driven management#

The pupillometer does not titrate a single drug or pressure on its own. It triggers decisions.

9.1 ICP and CPP#

A falling NPi in a TBI patient with an ICP monitor in place forces a recheck of ICP, CPP, sedation, head position, PaCO2, temperature, and serum sodium. A falling NPi in a patient without a monitor strongly justifies placing one (or, at minimum, urgent imaging).

9.2 Hyperosmolar therapy#

A falling NPi with an ICP > 20 mmHg (all pediatric ages, Kochanek 2019) supports an immediate hypertonic saline or mannitol bolus. The NPi response after osmotherapy can be re-checked at 30 to 60 min: a recovering NPi is reassurance; a continuing fall mandates escalation. 11

9.3 Sedation titration#

A patient on opioid infusions may have small pupils with preserved NPi; a patient on benzodiazepines may show slightly low NPi (2.5 to 3) without structural cause. Sedation-induced low NPi tends to recover with a daily sedation hold; structural low NPi does not. The daily sedation hold has dual purpose: it lets the clinical exam and the pupillometer both report on the brain rather than the drug.

9.4 Prognostication after cardiac arrest#

In the multimodal prognostication framework, persistent NPi 0 bilaterally at 72 h post-arrest, combined with bilateral absent SSEP N20, a markedly suppressed EEG, and diffuse cortical injury on MRI, is strongly associated with poor neurological outcome. No single modality, including NPi, is sufficient for prognostication; the framework requires concordance across at least 3 modalities. 3 4

10. Clinical contexts: NPi across acute brain injuries#

10.1 Severe TBI#

The primary published indication for pupillometry. Two registry datasets:

• ORANGE adult multicentre cohort (Oddo 2023): NPi < 3 sustained > 1 h was associated with poor outcome (mRS 4 to 6) independently of GCS and CT findings. 2
• Petrosino 2025 ORANGE pediatric pilot: smaller cohort but consistent direction; falling NPi over 6 to 12 h preceded clinical herniation by a median 9 h. 10

In pediatric severe TBI, the pupillometer adds quantitative early warning to the BTF / pBTF management bundle. It does not replace ICP monitoring but it is a useful adjunct for the patient whose ICP monitor is delayed. 11 7

10.2 Aneurysmal SAH#

NPi is added value in SAH for two reasons. First, early signs of evolving hydrocephalus or rebleed are caught earlier than by pen-torch. Second, in the DCI window (days 3 to 14), a unilateral fall in NPi often precedes the focal neurological deficit that classically defines clinical DCI. Pair with daily TCD and qEEG. 12 13

10.3 Pediatric AIS#

In a child receiving thrombectomy or thrombolysis for arterial ischaemic stroke, the pupillometer monitors for post-recanalisation oedema and the rare but devastating hyperperfusion haemorrhage. A new asymmetry > 0.7 or a falling NPi within 24 h of recanalisation triggers BP review and urgent imaging. 14 15

10.4 HIE and post-cardiac arrest#

The pupillometer's role here is prognostic, not real-time titration. Bilateral NPi 0 at 72 h post-arrest in a normothermic, sedation-free patient is one leg of the multimodal prognostication framework. Do not prognosticate during hypothermia or in the first 72 hours; sedation, hypothermia, and metabolic confounders depress NPi reversibly. 3 4

10.5 Pediatric ECMO#

VA-ECMO carries 5 to 15% risk of acute neurological injury (stroke, ICH, anoxic brain injury). Daily pupillometry is part of the ELSO neurological surveillance bundle. A new asymmetry > 0.7 or a falling NPi triggers head CT, fontanelle ultrasound, or MRI. The technical challenge: the constant ECMO circuit traffic around the bed and frequent transfers make routine pupillometry harder than in a stable ICU patient. 16 17

10.6 Meningitis and encephalitis#

In bacterial meningitis with evolving cerebral oedema or vasculitic vasospasm, the pupillometer can catch early herniation signs in a patient where the clinical exam is depressed by the underlying illness. A falling NPi in this context warrants repeat imaging and consideration of EVD or ICP monitoring. 18 19

10.7 Brain-death determination#

Pupillometry is not a brain-death determination tool by itself. It is, however, a useful adjunct that documents the absence of any constriction response with a number rather than a qualitative descriptor. NPi 0 with CV 0 bilaterally adds quantitative weight to the clinical brain-death exam. Pediatric brain-death criteria still require the standard clinical exam + apnoea test repeated at the age-appropriate interval. 20 21

10.8 DKA cerebral oedema#

A child being rehydrated for DKA whose NPi drops 0.7 or more units between hours 4 and 24 of treatment should prompt cerebral oedema treatment (mannitol or hypertonic saline) regardless of the absolute value. The pupillometer here gives a quantitative early-warning signal that supplements clinical vigilance for headache, vomiting, bradycardia, and hypertension. 22 23

10.9 Refractory status epilepticus#

In a child on continuous infusion (midazolam, pentobarbital, ketamine) for refractory status, pupillometry can detect oversedation (very low NPi, very low CV, slow DV) and titrate downward, especially when EEG burst suppression is the chosen target and the bedside exam is unreliable. 24 25

11. Multimodal integration: pupillometry in the MMM/MNM stack#

12. Setup and technique: a step-by-step#

12.1 Equipment#

• NeurOptics NPi-200 or NPi-300 handheld pupillometer (the dominant device; proprietary NPi algorithm).
• Single-patient-use eyecups (the soft rubber cushion that holds the device against the orbital rim).
• A docking station to upload measurements to the patient record (or paper logging into the bedside flow sheet).
• A quiet area with controllable ambient light (ambient should be standardised across measurements; the device is reasonably robust to ambient changes but consistency helps).

12.2 The measurement (one side)#

1. Identify the patient. Scan the patient's wristband or enter the patient ID into the device.
2. Position the patient. Supine, head neutral, head-of-bed 30°. Eyes open if possible (gently retract the lid if not, but do not press on the globe).
3. Apply the eyecup. Single-use; cushion the device against the orbital rim, not pressing on the eyeball. The device automatically aligns the camera to the pupil.
4. Trigger. Press the measurement button. The device delivers a 1000 lux, 0.8 s pulse and records the response for 5 s.
5. Wait for the result. NPi, CV, LAT, MAX, MIN, DV, % CH appear on screen. The device flags asymmetry > 0.7 automatically if a paired-side reading is in the buffer.
6. Repeat contralaterally within 30 s ideally so that ambient and patient conditions are matched.
7. Document both sides on the bedside flow sheet alongside other neuro observations.

12.3 Frequency#

• Severe TBI: hourly during the acute window (first 24 to 48 h), then Q2 to Q4 hour as stable.
• SAH (DCI window): Q2 to Q4 hour during days 3 to 14.
• Post-arrest: Q4 hour during the prognostication window (24 to 72 h post-rewarming).
• ECMO: Q4 to Q6 hour as part of the ELSO neurological surveillance bundle.
• Stable post-op neuro: Q4 to Q8 hour or as per local protocol.

12.4 What to do with the data#

• Plot the trend. Most pupillometers can export to the EHR; if not, transcribe NPi and CV onto the bedside chart alongside ICP, CPP, and GCS.
• Set asymmetry alerts. Ask nursing to escalate if Δ NPi > 0.7 or single-side NPi < 3 emerge.
• Pair with clinical exam. Every pupillometer reading should be paired with a brief documentation of patient state (awake / drowsy / sedated; pupil size estimate; light response by pen-torch).

12.5 Pediatric-specific tips#

• Sleeping infants: try to capture at a similar arousal level each time; document state.
• Toddler resistance: enlist the parent; use a brief, single-pass measurement; offer a quick reward if appropriate.
• Small pupils: at MAX < 2.5 mm the algorithm may struggle; document and re-measure after pupil natural dilation in dim light.
• Eyelid swelling / orbital trauma: may preclude measurement; document the reason rather than recording a spurious low NPi.

12.6 Calibration and quality control#

• The device self-calibrates on power-up; no daily user calibration required.
• Replace the eyecup between patients and clean per manufacturer guidance.
• Annual servicing is recommended; check the device clock and software version periodically.

13. Pitfalls#

• Pharmacological mydriasis (atropine, scopolamine, nebulised ipratropium drift, post-fundoscopic dilating drops) creates NPi 0 in an otherwise neurologically intact patient. Always interrogate the drug list.
• Opioids constrict the pupil but preserve the reflex; pinpoint pupils with NPi 4 to 5 are normal on opioid.
• Sedation depresses NPi by ~0.5 to 1 unit, more with deep benzodiazepine or propofol. Document sedation status with every measurement.
• Hypothermia depresses NPi; do not prognosticate during therapeutic hypothermia.
• Ambient light variation: NPi is reasonably robust but extreme bright vs dark conditions shift readings; aim for consistent room conditions.
• Pre-existing anisocoria (Adie's, post-surgical iris, congenital) gives baseline asymmetry; document on admission.
• NPi 0 in a patient with intact corneals and gag is more likely pharmacological or local (Adie's, eye trauma) than catastrophic CN III dysfunction; clinical correlation always.
• Neonates and small preterm infants: data are sparse, normative ranges different; treat as investigational.
• Operator-eyecup contact pressure: pressing too hard on the globe can transiently raise IOP and alter the reflex; rest the eyecup, do not press.
• Spurious asymmetry from probe misalignment: a single asymmetric reading should be confirmed within 5 min before triggering escalation.

14. Combine with…#

• Clinical exam, GCS, FOUR: the pupillometer is the quantitative complement to the pen-torch.
• ICP: structural raised pressure as the underlying driver of NPi drop.
• TCD: pairs PI and NPi as twin early-warning indices for evolving ICP rise.
• EEG: structural NPi + electrical EEG in multimodal prognostication.
• Evoked potentials: SSEP N20 + NPi in post-arrest prognostication.
• Foundations: brainstem anatomy: the parasympathetic CN III pathway as the substrate.

15. Evidence summary#

16. Recent literature (2022 to 2025)#

• Oddo 2023 (ORANGE): 514-patient adult multicentre cohort showing NPi < 3 sustained > 1 h independently predicts poor outcome (mRS 4 to 6) after acute brain injury. 2
• Kerscher 2023: pediatric TBI cohort confirming the NPi < 3 threshold as abnormal in children. 7
• Jiang 2023: pediatric NPi normative dataset, comparable to adult ranges from 6 months upward. 8
• Petrosino 2025 (ORANGE-Peds pilot): pediatric TBI pilot data showing falling NPi precedes clinical herniation by a median 9 h. 10
• Oddo 2025 (pupillometry post-arrest): confirms persistent NPi 0 at 72 h post-arrest as one leg of multimodal prognostication; not sufficient alone. 3
• Kirschen 2025: pediatric NPi in post-arrest prognostication, expanding the evidence base in children. 9

17. Self-check#