Article Text

Download PDFPDF

Original article
Pupillometry: a non-invasive technique for pain assessment in paediatric patients
  1. Mark A Connelly1,2,
  2. Jacob T Brown3,
  3. Gregory L Kearns1,3,4,
  4. Rawni A Anderson3,
  5. Shawn D St Peter1,5,6,
  6. Kathleen A Neville1,3,4,7
  1. 1Departments of Pediatrics, University of Missouri-Kansas City, Kansas City, Missouri, USA
  2. 2Department of Developmental and Behavioral Sciences, The Children's Mercy Hospital, Kansas, Missouri, USA
  3. 3Divisions of Clinical Pharmacology, The Children's Mercy Hospital, Kansas, Missouri, USA
  4. 4Department of Pharmacology, The University of Missouri, Kansas, Missouri, USA
  5. 5Department of Surgery, The University of Missouri, Kansas, Missouri, USA
  6. 6Center for Prospective Surgery Research, the Children's Mercy Hospital, Kansas, Missouri, USA
  7. 7Department of Hematology and Oncology, The Children's Mercy Hospital, Kansas, Missouri, USA
  1. Correspondence to Dr Kathleen A Neville, Division of Clinical Pharmacology and Therapeutic Innovation, Children's Mercy Hospitals and Clinics, 2401 Gillham Road, Kansas City, MO 64108, USA; kaneville{at}


Objective Pupillometry has been used to assess pain intensity and response to analgesic medications in adults. The aim of this observational study was to explore proof of concept for the use of this technique in paediatric patients. Changes in pupil parameters before and after opioid exposure also were evaluated.

Design and setting This was a single-centre, prospective study conducted at an academic paediatric medical centre.

Patients Children 9–17 years of age undergoing elective surgical correction of pectus excavatum were enrolled into a protocol approved by the human ethical committee (institutional review board).

Interventions Pupil size and reactivity were measured using a handheld pupillometer. Pain was assessed using age-appropriate, validated pain self-report scales.

Results Thirty patients were enrolled. Each point change on a 10 cm visual analogue pain intensity scale was associated with a statistically significant mean change of 0.11 mm/s in maximum pupil constriction velocity, and of approximately 0.4% in pupil diameter. As expected, there was an association between total opioid dose (expressed as morphine equivalents) and pupil diameter. Age, sex and baseline anxiety scores did not correlate significantly with pupillary response.

Conclusions The association of maximum pupillary constriction velocity and diameter with pain scores illustrates the potential for using pupillometry as a non-invasive method to objectively quantitate pain response/intensity in children. The technique holds promise as a pharmacodynamic ‘tool’ to assess opioid response in paediatric patients.

  • Pain
  • Pharmacology
  • Analgesia
  • Technology

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What is already known

  • Current pain assessment in children can be challenging and current strategies are not always accurate.

  • A single study in adults found that the pupil dilation reflex in response to a painful stimulus correlated with subjective pain scores.

What this study adds

  • In postoperative children certain quantitative pupillometry readings significantly correlated with self-reported pain scores.

  • Pupillometry may be a viable tool for objective pain assessment in children.


Research over the past several decades has led to significant improvement in the recognition and treatment of pain in children.1 ,2 However, despite significant advances in the understanding of the mechanisms of pain,3 considerable gaps remain in our knowledge of how pain is interpreted and expressed by children. As a consequence of the challenges in quantifying pain in children, management often is suboptimal.4–6

Self-reported pain ratings are the primary metric for quantifying pain because of their ease of use and compatibility with the commonly held belief that pain is best defined by the individual's own perception.7 However, self- reports of pain by children reflect an array of developmental, psychological and social influences that may obscure the information required to make a reliable assessment of pain and analgesic response.8–10 Furthermore, self-reporting tools are not useful in non-verbal infants or children.

In adults, pupillometry has been studied as a method to quantitate pain11 and to assess opioid pharmacodynamics.12 ,13 The extent of pupil dilatation can provide an index of nociceptive input via autonomic innervation of the iris muscles,14 ,15 while the extent of attenuation in this pupillary response during exposure to opioid analgesics can provide an index of pharmacological effect by reflecting the extent of occupancy of μ and κ opioid receptors in the central nervous system.16 Currently available handheld pupillometers are safe, well-tolerated by children, easy to use and modestly priced.17 They have the potential for providing an objective measurement of pain intensity and treatment response in children. To determine their utility in paediatric patients, we assessed pupillary response in a cohort of patients following a painful surgical procedure by comparing pupillometry measurements with simultaneous self-reported pain intensity.


This was a single-centre, prospective, within-subjects design study approved by the local institutional review board and conducted at an academic paediatric medical centre. Written permission/assent was obtained from all participants. Participants were enrolled from a population of children undergoing pectus excavatum repair with bar placement. All patients were given the opportunity to participate in the study. Postoperatively, all patients received either patient controlled analgesia (PCA) or analgesia via epidural catheter.18 Participants were excluded if they had inadequate baseline cognitive functioning or understanding of the English language to adequately understand the pain self-report questionnaires, or if they had any known allergies to the pain medications used in the postoperative pain management protocol. The final sample contained 30 patients between the ages of 9 years and 17 years (63% male; 97% Caucasian; 3% African-American). Of the study participants, 13 received epidural analgesia while 16 were on intravenous PCA pumps. During the study, five participants that initially had an epidural switched to PCA.

Preoperative medication and anaesthesia was standardised in all patients. At the conclusion of surgery, participants receiving PCA were started on intravenous hydromorphone by PCA (approximately 5–6 µg/kg continuous infusion, 5–6 µg/kg demand dose, with a 6 min lockout) with a loading dose if necessary. An additional 8 µg/kg dose of hydromorphone was available every 2 h for self-reported pain scores greater than 4/10. For participants who received epidural analgesia, a thoracic epidural catheter was placed at the level of T6 to T8. An initial bolus of approximately 0.3 mL/kg of 0.1% ropivacaine, fentanyl (1–1.2 µg/kg) and clonidine (1.8–2 µg/kg) was administered, followed by an infusion of 0.1% ropivacaine, fentanyl (2 µg/mL) or hydromorphone (4 µg/mL), and clonidine (1µg/mL or 1.5 µg/mL) solution at approximately 0.3 mL/kg/h. Postoperatively, bolus doses of 0.05 mL/kg of the ropivacaine/fentanyl/clonidine solution were available every 20 min for pain scores greater than 4/10 (maximum of 2 mL/dose or a cumulative dose of 14 mL/h). Patients still in pain received ropivacaine 0.2%, lidocaine 1% or a bolus of clonidine (1 µg/kg; maximum every 8 h). If pain relief while on the epidural protocol was judged to be inadequate, participants were then switched to intravenous analgesia via morphine or hydromorphone PCA. No pupillometry assessments were collected on postoperative day 0 in order to avoid confounding pupillometry readings by medications administered perioperatively. Starting on postoperative day 1, the cumulative dose of opioid analgesics administered 3 h prior to each pain assessment was recorded and converted to intravenous parenteral morphine-equivalent doses expressed in milligrams using published conversion factors (0.15 for hydromorphone, 0.01 for fentanyl and a factor of 3 for epidural to systemic conversion).19 The total equianalgesic dose was divided by patient weight to derive a standardised dose equivalent expressed as mg/kg of morphine equivalent.

Self-report pain measures

Pain intensity was assessed at rest at baseline (preoperatively) and every 3 h during postoperative days 1–3. At each pain assessment, self-reported pain scores were recorded using a 10 cm horizontal visual analogue scale (VAS) with anchors of no pain (0) and worst possible pain (10).20 Pain scores related to movement (ie, coughing) were not assessed in the study participants. Information on emotional status was assessed at each pupillometry reading using an abbreviated form of the Positive and Negative Affect Schedule for Children, a previously validated technique.21

Pupillometry technique

Pupillometry measurements were obtained by a research coordinator at baseline (preoperatively) and then simultaneously with self-reported pain intensity every 3 h beginning at 0800 and ending at 2000 during postoperative days 1–3 (5 readings per day). Each pupillometry measurement was performed in triplicate, and the average of the three measurements served as the final data point for each time period. Potential bias associated with operator error was minimised by having only three individuals, all fully trained in the use of the pupillometer, record all observations.

A NeurOptics PLR-100 infrared pupillometer (NeurOptics, Irvine, California, USA) was used to measure dark-adapted pupil diameter and pupillary response dynamics under controlled dim light conditions using the standardised procedure recommended by the manufacturer. The procedure involves positioning the soft rubber cup of the instrument over the eye to be used for measurement. The resting (maximum) pupil diameter (mm) is then recorded, followed by measurement of the pupil constriction velocity (mm/s), minimum diameter (mm), time to minimum diameter (s), constriction amplitude (resting minus minimum diameter), relaxation velocity (mm/s) and final diameter (mm) in response to a standard light stimulus. The procedure is brief, non-invasive, has no associated risk, and has been validated as a measure of analgesic drug effect in adults.16 ,22

Data analysis

The primary hypothesis tested was that pupillometry measurements would correlate with postoperative self-report pain scores. A secondary hypothesis was that pupillometry measurements would correlate with opioid exposure expressed as morphine equivalents. Hierarchical linear modelling (HLM) was used to test the primary and secondary study hypotheses. HLM is a recommended analytical approach used to establish associations for variables measured serially over time within each individual.23 In HLM models, the dependent variable (a given pupil parameter) is regressed onto predictor variables (pain and opioid exposure) for each individual at every time point while controlling for unknown sources of measurement error. Individual models are then averaged together to determine the overall unique relationship between each predictor and the outcome variable (ie, the expected change in the outcome variable for each unit change in the given predictor variable) while controlling for error from unspecified individual differences. To test the primary study hypothesis, model b-coefficients representing the average relationship between pain intensity and pupil parameters (while controlling for opioid exposure) were evaluated for statistical significance against a t-distribution. If significant, this was interpreted as support for the primary study hypothesis. Similarly, model b-coefficients representing the average relationship between opioid exposure and pupil parameters (while controlling for pain) were evaluated for statistical significance against a t-distribution. If significant, this was interpreted as support for the secondary study hypothesis. Child age, sex and baseline anxiety were evaluated in all models as covariates and removed if they did not significantly moderate findings. Estimates of central tendency (mean, median) and variability (SD, quartiles) were computed on the primary study variables for descriptive purposes and for evaluating statistical assumptions. HLM was conducted using HLM6 for Windows (Scientific Software International, Skokie, Illinois, USA, 2004). All other statistical analyses were conducted using subroutines available in SPSS V.18 (SPSS for Windows, SPSS, Chicago, Illinois, USA, 2004). The significance limit set for all statistical analyses was α=0.05.


With one exception of a subject who refused to provide assent, all of the remaining patients who were initially approached agreed to participate, thus yielding a 97% accrual. No discomfort, injury or other apparent study-related adverse effects occurred with any study participant and no pupillometry measurements were refused at any point during the study.

Age, sex and anxiety scores were found to not significantly correlate with pupillary response and were excluded from subsequent analyses. Maximum pupil size, per cent change in pupil size, and maximum constriction velocity were directly correlated with the VAS (while controlling for opioid dose and baseline pupillary measurements) (table 1). The two most robust of these were maximum constriction velocity and per cent change in pupil size. As illustrated in figure 1, each point change on the 10-point VAS was associated with a mean change of 0.11 mm/s in maximum constriction velocity (p=0.004). Likewise, as illustrated in figure 2, each point change on the 10-point VAS was associated with a mean change of 0.39% in pupil size (p=0.018). Relationships between maximum constriction velocity and pain at the level of an individual are displayed in figures 3 and 4 for the patients with the highest and lowest correlations between these variables, respectively. As would be expected on a physiological basis regarding pupil function, the regression of maximum constriction velocity onto resting pupil diameter demonstrated a statistically significant moderate association (p<0.01), with 56% of the variability in pupil diameter being unique from maximal constriction velocity (r2=0.441, SE of the estimate=0.597; illustration not shown).

Table 1

Results of hierarchical linear modelling evaluating the association of pain intensity with pupillometry metrics

Figure 1

The association of maximum pupil constriction velocity with pain intensity determined by patient self-report using a validated visual analogue scale (VAS). The association was statistically significant (p<0.01). The error bars reflect the ±SE of the mean for each point.

Figure 2

The association of the per cent of pupil constriction with pain intensity determined by patient self-report using a validated visual analogue scale. The association was statistically significant (p<0.01). The error bars reflect the ±SE of the mean for each point.

Figure 3

The relationship between pain intensity and maximum constriction velocity over time for the individual with the highest and lowest relationships between these variables. VAS, visual analogue scale.

Figure 4

The relationship between pain intensity and maximum constriction velocity over time for the individual with the lowest relationship between these variables. VAS, visual analogue scale.

The exploratory (ie, secondary) hypothesis tested was to explore the potential associations between pupillary response and opioid exposure. As expected given the effects of opioids on pupillary function, all measurements of pupillary response were inversely associated with opioid dose (table 2). As illustrated in figure 5, every 1 µg/kg increase in opioid dose was associated with an average 2.02 mm reduction in pupil size. For every µg/kg increase in opioid dose, minimum pupil size decreased 0.82 mm, overall change in pupil size (before and after light stimulus) decreased 17.88%, maximum constriction velocity decreased 2.64 mm/s, average constriction velocity decreased 1.26 mm/s, and dilation velocity decreased 0.89 mm/s as compared with baseline (table 2).

Table 2

Pupillary parameter values before and during analgesic medication administration

Figure 5

The association between baseline pupil size and daily drug dose (expressed as µg/kg morphine equivalents). The association was statistically significant (p<0.01). The error bars reflect the ±SE of the mean for each point.


We found statistically significant associations between self-reported pain intensity and several measures of pupillary response in children who were evaluated after a uniform, painful surgical procedure. Of the associations tested, the relationship between maximum pupillary constriction velocity and VAS pain intensity (figure 1) provided the best surrogate for quantitating pain intensity in children.

Expressing pupillary function as a percentage (increase) in pupil constriction functionally provides a measure of maximum effect (Emax) for a given observation (figure 2). This ‘ceiling’ could potentially be impacted by differences in the concentration versus effect profile for agents used to manage pain (eg, opioids, clonidine, ropivacaine, lidocaine) in our cohort. This is especially pertinent to the use of clonidine, a centrally acting α2 adrenergic agonist which has been shown to have pupillary effects.24 Maximum constriction velocity on the other hand is a dynamic parameter, which reflects how efficiently the pupil responds to a given stimulus. Given that an increase of 2 or more on the 0–10 VAS scale is associated with a clinically significant change in pain intensity,25 an observed change in maximal pupil constriction velocity of ≥0.22 mm/sec could be used to guide intervention to optimise pain management. It could also provide an objective surrogate to aid in the differentiation of pain perception versus actual pain intensity. Finally, it is important to recognise that either maximal constriction velocity or the per cent of pupil constriction is dependent upon the baseline status of the pupil, as illustrated by our finding of a significant correlation between maximum constriction velocity and resting pupil diameter. This caveat does not limit the utility of pupillometry, especially when maximum constriction velocity is used as the surrogate end point.

The association between daily drug dose (expressed as morphine equivalents) and pupil size (figure 3) was expected. Nonetheless, controlling for opioid effect in HLM analyses, we were able to demonstrate the independent effect of pain on pupillary response. However, further research is needed to support the use of pupil size as a predictor of analgesic dose requirement given potential confounders in our uncontrolled study (eg, different drug combinations including non-opioid compounds).

Our findings are consistent with previous studies that have examined pupillometry with traditional physiological surrogates of pain assessment in adults and children. Barvais et al16 compared pupil response with arterial pressure, heart rate and bispectral index (BIS) variations following a painful stimulus in healthy adult patients receiving propofol and remifentanil. They found that in response to a noxious stimulus, pupil dilatation was more sensitive and better correlated with opioid concentrations than heart rate, blood pressure and BIS monitoring. In a similar study performed in anaesthetised children, Constant et al14 compared changes in pupil diameter with heart rate, blood pressure and BIS monitor variation in response to a skin incision on the lower limb. Similar to the adult data,16 these investigators were able to show that the change in pupil diameter had a greater effect size than BIS variation or haemodynamic markers.

While precise pupil measurements and subjective pain assessment have not been previously compared in children, Aissou et al2,6 performed a study in 100 adults comparing changes in the pupil dilatation reflex and self-reported pain scores via a 5-point verbal rating scale. Shortly after extubation, patients receiving intravenous morphine in a postoperative setting had their pupil size monitored and recorded for 10 s during a standardised stimulus of constant pressure (200 kPa) within 2–3 cm from their skin incision. A statistically significant relationship was observed between the VAS and maximum pupil diameter, as well as a threshold of per cent change in pupil diameter in patients who required additional doses of morphine.

As with all physiological surrogates, pupillometry does have limitations. These include the lack of specificity of the pupil dilatation reflex in response to pain and clinical situations in which a patient is in pain but has constricted pupils. Also, as our study enrolled verbal, cognitively appropriate participants capable of self-reporting pain, additional research is needed to generalise findings to non-verbal patients or those with significant cognitive impairment.

Summary and conclusions

Our data preliminarily illustrate the utility of pupillometry as a physiological surrogate capable of assessing pain intensity in paediatric postsurgical patients ranging in age from 9 years to 17 years. The device was well-accepted, easy to use and produced reliable measurements of several parameters reflective of pupillary function. Of all the parameters evaluated, maximum constriction velocity appeared to have the greatest association with VAS pain intensity ratings and provided the most dynamic measurement. Our findings suggest that this technique may have value for assessing pain intensity and/or response to treatment in non-verbal patients and in patient populations experiencing different types of pain (eg, nociceptive pain vs complex pain; acute pain vs chronic pain; surgical pain vs traumatic pain). Pupillometry may also prove to demonstrate its application as a pharmacodynamic surrogate capable of exploring the impact of age on the exposure-response relationship for opioids in paediatric patients.


This study was supported in part by a grant from the Katherine B Richardson Endowment fund. The guidance and technical assistance provided by Dr Ralph Kauffman in the preparation of this manuscript is gratefully acknowledged.



  • Contributors MAC was significantly involved in study design, data collection and data analysis/interpretation, as well as drafting and revising the manuscript. JTB assisted with interpretation of the data and drafting/revising the manuscript. GLK assisted with interpretation of the data and drafting/revising the manuscript. RAA had significant involvement in data collection. SDSP had significant involvement in the study conception/design. KAN was significantly involved in study design, data collection and data analysis/interpretation, in addition to revising the manuscript, approving the final version for publication, and the accuracy and integrity of all aspects of the research.

  • Competing interests None.

  • Ethics approval Children’s Mercy Hospital Institutional Review Board.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Linked Articles

  • Drug therapy
    Richard F Howard Christina Liossi