Without normative data, it is hard to know what is abnormal. Research suggests that an abnormal pupillary light reflex (PLR) is reflective of early changes in intracranial physiology.1–3 The pupillometer is a handheld device that was developed in response to a need for standardized and reliable pupil assessments. Because of the limitations of subjective observations of the PLR,4,5 there is growing adoption of pupillometer technology.6 Seeing these changes as trends is important. However, a lack of normative data specific to the population of neurocritically ill patients limits the interpretation of PLR data. The purpose of this analysis is to explore distributions for PLR data from automated pupillometry in a diverse cohort of neurocritically ill patients.
The full assessment of cranial nerve function, which includes the PLR, is a fundamental element of the neurological examination. A major portion of the functional status of the optic (cranial nerve II) and oculomotor (cranial nerve III) cranial nerves can be assessed by evaluation of the PLR.7,8 Abnormalities in this examination reflect not only cranial nerve deficits but also pathologic changes in the nearby intracerebral environment, leading to cranial nerve dysfunction.9,10 Recent literature supports that simply reporting a pupil as reactive, brisk, or sluggish fails to capture the entirety of the PLR.8 A small pupil will not reach high constriction velocity (CV) because the percent change in size is small. A large pupil may have a brisk response, but failure to fully constrict would be reflected as a low Neurological Pupil Index (NPi) value (eg, the pupil that briskly constricts from 6 mm to 5 mm).
There are limitations to the traditional method of assessing the PLR with a handheld flashlight. The pupil size, shape, and reactivity to light are influenced by ambient lighting, skin tone, eye color, quality of the light source, and subject cooperation.4,5 Using the terms brisk and sluggish in clinical practice to describe the PLR is vague and subjective. Variables in the pupillary assessment such as CV are reportedly sensitive to changes in intracranial pressure.11,12 Consistent and reliable pupil examinations with trending data may assist the team of providers in clinical decision making and lead to therapeutic interventions that may influence patient outcomes. Timely intervention (eg, hyperosmolar therapy, hyperventilation as a bridge to surgical intervention) after observation of a new pupillary abnormality has been associated with improved survival or recovery.13
Having objective PLR data is important to provide high-quality clinical care. Providing clinicians with a deeper understanding of the reference ranges for those data is expected to drive clinical interventions.14 There is currently only 1 handheld pupillometer sold in the United States; the NPi-200 (developed by Neuroptics, Inc) is used in acute and critical care settings.10,15–18 In addition to providing a digital reading for the initial size of the pupil, the NPi-200 provides several unique values that are unavailable from the manual examination (Figure 1). These include the latency, CV, dilation velocity (DV), and NPi, and each of these elements can be evaluated in detail. Pupil size is measured is millimeters and documented to the hundredth place. Latency is the time from when the pupil is first exposed to light to the onset of pupil constriction (measured in seconds and documented to the hundredth). Both CV and DV are measured in millimeters per second (mm/s) and documented to the hundredth. Constriction velocity is the speed at which the pupil changes from baseline to its smallest size over time. Dilation velocity is the change from the smallest pupil size back to baseline over time.
The NPi is derived from a proprietary algorithm that incorporates constriction latency, CV, and DV. The NPi is indexed to a baseline set of samples obtained from healthy volunteers, with scores ranging from 0 to 5.19,20 NPi scores greater than 3 are considered to be reflective of normal PLR, whereas a score of 0 indicates the absence of a PLR. Previous research in healthy volunteers found that larger pupil size was correlated with faster CV and higher NPi scores.21 The NPi is a derived variable with no specific units and documented to the tenth. In contrast to the limited interrater reliability of human observations,5 the pupillometer has a very high interdevice reliability.22
A major hurdle in interpreting pupillometry values is the diversity of pathophysiologic conditions and medical diagnoses hypothesized to impact PLR. For example, a distal middle cerebral artery infarct is expected to have less impact on PLR than would a large intracranial hemorrhage that results in central brainstem herniation.7 The gap in literature is a lack of descriptive data and measures of central tendency for PLR values with respect to patients admitted to the neurocritical care unit (NCU). This work, therefore, is not intended to serve as the pinnacle of automated PLR research, nor is it intended to provide inferential analyses; rather, this study will serve as a foundation for future exploration.
This is a planned descriptive analysis of data from the Establishing Normative Data for Pupillometer Assessments in Neuro-Intensive Care (END-PANIC) registry, which is approved by the local institutional review board and registered with ClinicalTrials.gov (NCT02804438). All data were collected as standard of care, and subject consent was not required. Data collection methods have been previously described in detail.23 In brief, END-PANIC is a prospective multicenter registry of pupillometer assessments and associated clinical data abstracted from patients admitted to the NCU at participating hospitals and prescribed to have PLR assessments as part of their standard care. Time and date values are stamped independently for left eye (OS) and right eye (OD). Pupillometer values for OS and OD (ie, NPi, minimum and maximum pupil size, DV, CV, and percentage of constriction) are obtained during routine care, and research staff downloads the values to an electronic spreadsheet. Similarly, data from the electronic medical record are abstracted such that pupillometer readings can be merged with electronic medical record data into a single time-stamped database. Shared data are maintained on a secured server, and statistical analysis is performed using SAS v9.4 for Windows (SAS Institute).
There were 72 472 PLR readings from 2140 patients included in this analysis. The END-PANIC registry currently includes 2655 patients admitted to 3 hospitals. Of these, there were 2371 patients with Glasgow Coma Scale (GCS) scores. Subjects in whom pupil examinations were not performed and observations without pupil reactivity were excluded (n = 231). The highest daily GCS (best neurologic function) was matched to automated PLR data without regard to medical diagnosis. Most observations provided OS and OD readings. Averaging OS/OD may fail to capture significant variation because of intracranial pathology (eg, mass effect with midline shift). Therefore, OS and OD values are evaluated separately (eg, a single assessment of both eyes from 1 patient would yield 2 PLR readings). Age was categorized as youngest (<25 years), middle (25–50 years), older (51–75 years), and elderly (>75 years). All values are reported as mean (standard deviation), unless otherwise noted.
To provide normative data in a large population of patients with variable severity of illness that might significantly impact the generalizability of those values, we chose to stratify patients using the GCS. Despite recognized limitations,24–26 the GCS provides a common denominator for discussing patients with neurologic injury. One of the flaws of the GCS is that it is limited in measuring level of consciousness in specific conditions, such as sedated or intubated patients.24 However, the GCS is a widely used organ-specific measure of severity of illness in neurocritical care wherein GCS 13–15 is mild, GCS 9–12 is moderate, and GCS 3–8 is severe. With a full understanding of the limitations, this analysis intentionally evaluates PLR data in relationship to GCS values.
This study provides only descriptive analytics. Nominal data are summarized using frequency, ordinal data were summarized using mean and interquartile range, and continuous data were summarized using mean and standard deviation. Central tendencies were developed for each PLR variable at each GCS level (mild, moderate, and severe). A random-effects mixed model for longitudinal data was constructed to account for clustering of subjects. This model was run using the Kenward-Roger degree of freedom.
This analysis includes 2140 subjects from 3 US hospital NCUs, including 2 university hospitals and 1 community hospital. Subjects were predominantly white (n = 1345, 62.9%), non-Hispanic (n = 1582, 73.9%), and female (n = 1,134, 53.0%), and the mean age was 52.3 years (IQR, 40–69; Supplemental Digital Content 1, available at http://links.lww.com/JNN/A199). The PLR values and central tendencies for the entire sample taken as a single cohort, and stratified by GCS score severity, are provided in Table 1.
Neurologic Pupil Index
The mean (SD) NPi for OS (4.14 [0.83]) and OD (4.15 [0.80]) were similar (P = .2199). A paired t test also showed a statistically significant difference in NPi for OS and OD within subjects (P < .001). A random-effects mixed model indicates significant within-subject variation in OS and OD NPi values (P < .001). The NPi values for mild (4.3 [0.64]), moderate (4.2 [0.83]), and severe (4.1 [0.89]) injury were statistically significantly different (P < .001). The difference in NPi values for men (4.3 [0.73]) versus women (4.1 [0.86]) was significant (P < .001). There was a statistically significant difference in NPi values by age comparing youngest (4.0 [0.88]), middle (4.0 [0.84]), older (4.2 [0.73]), and elderly (4.4 [0.56]; P < .001) patients.
For size, the OS (3.47 [1.16] mm) and OD (3.49 [1.17] mm) were clinically similar but statistically different (P < .0027). A paired t test also showed a statistically significant difference in size for OS and OD within subjects (P < .001). A random-effects mixed model indicates significant within-subject variation in OS and OD pupil sizes (P < .001). The size values for mild (3.2 [1.14] mm), moderate (3.6 [1.14] mm), and severe (3.8 [1.07] mm) injury were statistically significantly different (P < .001). The difference in size for men (3.3 [1.12] mm) versus women (3.6 [1.18] mm) was significant (P < .001). The difference in size values for youngest (3.6 [1.35] mm), middle (3.8 [1.33] mm), older (3.4 [1.05] mm), and elderly (3.2 [0.97] mm) patients was significant (P < .001).
Pupil latency is measured in seconds (s). Latencies for the OS (0.26 [0.06] s) and OD (0.26 [0.06] s) were similar (P = .0607). A paired t test showed a statistically significant difference in latency for OS and OD within subjects (P < .001). A random-effects mixed model indicates significant within-subject variation in OS and OD pupil latencies (P < .001). The latency values for mild (0.26 [0.06] s), moderate (0.26 [0.06] s), and severe (0.24 [0.05] s) injury were clinically similar but statistically significantly different (P < .001). The latency values for men (0.25 [0.05] s) versus women (0.26 [0.06] s) were also statistically significantly different (P < .001). The difference in latency values for youngest (0.25 [0.06] s), middle (0.25 [0.06] s), older (0.26 [0.05] s), and elderly (0.26 [0.05] s) patients was found to be significant (P < .001).
The mean (SD) CVs for OS (1.62 [0.86] mm/s) and OD (1.62 [0.87] mm/s) were similar (P = .8609), and a paired t test found no statistically significant difference in CV within subjects (P < .59). A random-effects mixed model indicates significant within-subject variation in OS and OD CVs (P < .001). The CV values for mild (2.0 [0.82] mm/s), moderate (1.7 [0.82] mm/s), and severe (1.3 [0.80] mm/s) injury were statistically significantly different (P < .001). The difference in CV values for men (1.6 [0.82] mm/s) versus women (1.7 [0.9] mm/s) was statistically significant (P < .001). The difference in CV values for youngest (1.6 [0.95] mm/s), middle (1.8 [0.99] mm/s), older (1.6 [0.80] mm/s), and elderly (1.5 [0.74] mm/s) patients was significant (P < .001).
For size, the OS (0.71 [0.39] mm/s) and OD (0.72 [0.39] mm/s) were similar (P = .2623). A paired t test also showed a statistically significant difference in DV for OS and OD within subjects (P = .002). A random-effects mixed model indicates significant within-subject variation in OS and OD DVs (P < .001). The DV values for mild (0.58 [0.37] mm/s), moderate (0.76 [0.37] mm/s), and severe (0.91 [0.37] mm/s) injury were statistically significantly different (P < .001). The DV values for men (0.67 [0.37] mm/s) versus women (0.74 [0.40] mm/s) were different (P < .001). The difference in DV values for youngest (0.68 [0.44] mm/s), middle (0.80 [0.44] mm/s), older (0.71 [0.36] mm/s), and elderly (0.67 [0.33] mm/s) patients was significant (P < .001).
Findings were explored based on admission diagnosis. The mean (SD) values for NPi ranged from 4.06 (0.81) for nonsurgical neurologic injuries other than stroke (eg, seizure) to 4.31 (0.69) for acute ischemic stroke and were statistically significantly different by diagnosis (P < .001), although it is noted that these values are clinically similar. The mean (SD) pupil size ranged from 3.2 (1.10) mm for intracerebral hemorrhage (ICH) to 3.7 (1.13) mm for neoplasm (P < .001). Latency ranged from 0.24 (0.04) s for traumatic brain injury to 0.28 (0.05) s for acute ischemic stroke (P < .001). Constriction velocity ranged from 1.48 (0.91) mm/s for ICH to 1.86 (0.93) mm/s for neoplasm (P < .001). Dilation velocity values ranged from 0.63 (0.39) mm/s for ICH to 0.83 (0.40) mm/s for neoplasm (P < .001). Although these differences are statistically significant, we are unable to find literature discussing the clinical relevance of this finding.
The results provide insight for clinicians who have already adopted or those transitioning their practice to incorporate automated pupillometry. These findings not only serve as a baseline for future comparisons and hypothesis testing that includes PLR data from automated pupillometry but also provide point estimates and variance for PLR measures when used in clinical practice. The ability of quantitative pupillometry to deliver replicable and reliable results, and its relationship to associated outcomes, has been previously established for pupillary measurements in previous studies.2,9,12,21,22,27-30 However, until now, only a small cohort of healthy volunteers has been assessed for normative data.12
Paired testing of OS and OD values, including mixed methods, provides support that differences in measured values are expected. This finding confirms and extends clinical and pathophysiologic findings. The vast majority of intracranial lesions has some focality and would be expected to impact 1 eye PLR differently than the other eye. For example, a left frontal ICH would be expected to impact the left PLR differently than it would impact the right PLR up until the lesion expands to a terminal point (eg, downward herniation). The results provide insight into expected findings and emphasize the need for monitoring trends. The left and right PLR should be examined separately with OS values compared and trended with previous OS values, and OD values compared and trended with previous OD values.
This analysis was not designed to provide inferential statistics or to describe PLR effects within discreet diagnoses. This analysis examines the distribution of pupil reactivity values. A recognized limitation is the use of nonrandomized sampling. All observations were obtained from patients admitted to NCUs and therefore were more likely to have abnormal PLR. However, this does represent the largest known sample of automated PLR readings, and the heterogeneity of the sample enhances external validity. It is a recognized limitation of the study that the registry does not collect specifically data on the presence or absence of second or third cranial nerve pathology. However, these are rare findings in the NCU and unlikely to impact the results.31
This report is the first to provide measures of central tendency for PLR values across a large cohort of subjects. Separate observations of OS and OD, including trends, should be monitored throughout the NCU stay. These data provide baseline normative values for future research into automated pupillometry in clinical practice.
1. McNett M, Moran C, Janki C, Gianakis A. Correlations between hourly pupillometer
readings and intracranial pressure values. J Neurosci Nurs
2. McNett M, Moran C, Grimm D, Gianakis A. Pupillometry trends in the setting of increased intracranial pressure. J Neurosci Nurs
3. Natzeder S, Mack DJ, Maissen G, Strässle C, Keller E, Muroi C. Portable infrared pupillometer
in patients with subarachnoid hemorrhage: prognostic value and circadian rhythm of the Neurological Pupil Index (NPi). J Neurosurg Anesthesiol
4. Omburo L, Stutzman S, Supnet C, Choate M, Olson DM. High variance in pupillary examination findings among postanesthesia care unit nurses. J Perianesth Nurs
5. Olson DM, Stutzman S, Saju C, Wilson M, Zhao W, Aiyagari V. Interrater reliability of pupillary assessments. Neurocrit Care
6. Lee MH, Mitra B, Pui JK, Fitzgerald M. The use and uptake of pupillometers in the intensive care unit. Aust Crit Care
7. Bader MK, Littlejohns LR, Olson DM. AANN Core Curriculum for Neuroscience Nursing
. 6th ed. Glenview, IL: American Association of Neuroscience Nurses; 2016.
8. Shoyombo I, Aiyagari V, Stutzman SE, et al. Understanding the relationship between the Neurologic Pupil Index and constriction velocity values. Sci Rep
9. Yokobori S, Wang KKK, Yang Z, et al. Quantitative pupillometry and neuron-specific enolase independently predict return of spontaneous circulation following cardiogenic out-of-hospital cardiac arrest: a prospective pilot study. Sci Rep
10. Zafar SF, Suarez JI. Automated pupillometer
for monitoring the critically ill patient: a critical appraisal. J Crit Care
11. Pickworth WB, Welch P, Henningfield JE, Cone EJ. Opiate-induced pupillary effects in humans. Methods Find Exp Clin Pharmacol
12. Taylor WR, Chen JW, Meltzer H, et al. Quantitative pupillometry, a new technology: normative data and preliminary observations in patients with acute head injury. Technical note. J Neurosurg
13. Clusmann H, Schaller C, Schramm J. Fixed and dilated pupils after trauma, stroke, and previous intracranial surgery: management and outcome. J Neurol Neurosurg Psychiatry
14. Aoun SG, Stutzman SE, Vo PN, et al. Detection of delayed cerebral ischemia using objective pupillometry in patients with aneurysmal subarachnoid hemorrhage. J Neurosurg
15. Schallenberg M, Bangre V, Steuhl KP, Kremmer S, Selbach JM. Comparison of the Colvard, Procyon, and Neuroptics pupillometers for measuring pupil diameter under low ambient illumination. J Refract Surg
16. Hsieh YT, Hu FR. The Correlation of pupil size measured by Colvard pupillometer
and Orbscan II. J Refract Surg
17. Bootsma S, Tahzib N, Eggink F, de Brabander J, Nuijts R. Comparison of two pupillometers in determining pupil size for refractive surgery. Acta Ophthalmol Scand
18. Du R, Meeker M, Bacchetti P, Larson MD, Holland MC, Manley GT. Evaluation of the portable infrared pupillometer
19. Larson MD, Behrends M. Portable infrared pupillometry: a review. Anesth Analg
20. Chen JW, Gombart ZJ, Rogers S, Gardiner SK, Cecil S, Bullock RM. Pupillary reactivity as an early indicator of increased intracranial pressure: the introduction of the Neurological Pupil Index. Surg Neurol Int
21. Boev AN, Fountas KN, Karampelas I, et al. Quantitative pupillometry: normative data in healthy pediatric volunteers. J Neurosurg
. 2005;103(suppl 6):496–500.
22. Zhao W, Stutzman S, DaiWai O, Saju C, Wilson M, Aiyagari V. Inter-device reliability of the NPi-100 pupillometer
. J Clin Neurosci
23. Olson DM, Stutzman SE, Atem F, et al. Establishing normative data for pupillometer assessment
in neuroscience intensive care: the "END-PANIC" Registry. J Neurosci Nurs
24. Nik A, Sheikh Andalibi MS, Ehsaei MR, Zarifian A, Ghayoor Karimiani E, Bahadoorkhan G. The efficacy of Glasgow Coma Scale (GCS) score and Acute Physiology and Chronic Health Evaluation (APACHE) II for predicting hospital mortality of ICU patients with acute traumatic brain injury. Bull Emerg Trauma
25. Reith FC, Brennan PM, Maas AI, Teasdale GM. Lack of standardization in the use of the Glasgow Coma Scale: results of international surveys. J Neurotrauma
26. McNett MM, Amato S, Philippbar SA. A comparative study of Glasgow Coma Scale and Full Outline of Unresponsiveness scores for predicting long-term outcome after brain injury. J Neurosci Nurs
27. Olson DM, Fishel M. The use of automated pupillometry in critical care. Crit Care Nurs Clin North Am
28. Truong JQ, Ciuffreda KJ. Comparison of pupillary dynamics to light in the mild traumatic brain injury (MTBI) and normal populations. Brain Inj
29. Halpern JI, Gordon WH Jr. Trochlear nerve palsy as a false localizing sign. Ann Ophthalmol
30. Solari D, Rossetti AO, Carteron L, et al. Early prediction of coma recovery after cardiac arrest with blinded pupillometry. Ann Neurol
31. Fang C, Leavitt JA, Hodge DO, Holmes JM, Mohney BG, Chen JJ. Incidence and etiologies of acquired third nerve palsy using a population-based method. JAMA Ophthalmol