Pupillometric Evaluation of the Dynamics of the Pupillary Response to a Brief Light Stimulus in Healthy Subjects

Abstract

Purpose.: To investigate the correlation between measurements of amplitude (A) and peak velocity (V) of constriction in the pupil light reflex of normal subjects, and to determine the effects of stimulus intensity, pupil size, and age on this relationship.

Methods.: The pupil response to a variable intensity 1.0 second light stimulus presented under open-loop conditions (Maxwellian optics) was measured using infrared video techniques in 43 healthy subjects aged 20 to 75 years old.

Results.: In response to the “standard-intensity” light stimulus, mean measurements of A and V were 1.92 mm (SD 0.39) and 5.65 mm/s (SD 1.17), respectively. Over a 4.0 log unit range of stimulus intensities measurements of A and V were seen to co-vary with the data being best fit by the equation $V=0.86+2.65A$ (linear regression coefficient, R = 0.919, P < 0.001). In each subject the regression plot was used to normalize the velocity estimates for A = 1.0 mm; these normalized velocity estimates showed no significant correlation with the starting size of the pupil or the age of the subject.

Conclusions.: There is a strong linear correlation between amplitude and peak velocity of constriction for the pupil light reflex in normal subjects. This relationship is unaffected by the stimulus intensity, size of the pupil, or age of the subject. Clinicians and researchers must keep this interdependence in mind when drawing inferences from the observed (or measured) speed of the pupillary constriction to light.

Introduction

In clinical practice, assessment of the pupillary response to light is an integral part of the evaluation of function in the anterior visual pathways, midbrain and autonomic nervous system. It usually is achieved by shining a torch into the eye and examining the reflex constriction of both pupils. The problem that arises is how to interpret what one sees—it is clear that even among healthy people there is considerable variation in this pupillary response. Should we be more interested in the extent or the speed of the constriction? Neuro-ophthalmologists often are referred patients whose pupils are said to be “sluggish,” an ill-defined term that seems to imply a slow or feeble constriction that is less impressive than expected. Often these pupils show a brisker response when examined at the slit-lamp, possibly because of the brighter light used to elicit the reflex or because the pupillary response is being viewed under greater magnification. 

There is a corresponding lack of consensus and clarity about what to measure among research studies. Since the introduction of infrared video techniques to record the pupil, ¹ the amplitude of the light reflex response (either absolute or relative to the starting diameter) has been used as the principle outcome measure in most studies, but other variables often are considered, including latency, ² velocity ³ (either peak or average), or acceleration. ⁴ In one study, waveform partitioning was used to evaluate the discriminatory power of different time segments of the pupillary response, with the conclusion that pupil size changes during “window IV” (which they defined as between the time of maximum constriction velocity and the time of peak constriction) offered the optimum sensitivity and specificity for detection of disease. ⁵ In most of these pupillometric studies multiple variables have been measured simultaneously, but the interdependence of these variables rarely is considered. 

The present study aimed to investigate the relationship between the two commonly measured pupillometric variables (amplitude and peak constriction velocity) which correspond most closely to what is looked for in routine clinical examinations, namely the “extent” and the “speed” or “briskness” of the response. In addition consideration has been given as to how this relationship might be affected by the brightness of the light used to elicit the reflex, the starting size of the pupil and the age of the subject (or patient). Preliminary results have been presented at the 10th European Neuro-ophthalmology Society (EUNOS) Meeting in Barcelona, Spain, June 2011. 

Methods

This study used standard video pupillometry techniques to measure the extent and speed of the pupillary constriction to a brief light stimulus. Participants were healthy males or females ≥18 years old with clinically normal pupils. Exclusion criteria included any ophthalmic, neurologic, or systemic disease; previous ocular surgery or laser treatment; or topical or systemic medications likely to affect the pupillary response to light. All subjects gave informed consent, and the study was approved by the local Research and Ethics Committee, and adhered to the tenets of the Declaration of Helsinki. 

Pupil recordings were made in the dark, with the head supported on a chinrest. The subject was asked to fixate on a small red light-emitting-diode (LED) located 4 meters away. A custom-built binocular pupillometer provided simultaneous video recordings of both pupils, infra-red LEDs illuminated the iris, and the contrasting dark pupil was identified and measured (with a spatial resolution of 0.03 mm and temporal resolution of 40 ms) using automated edge-detection software written in LabVIEW (National Instruments Corporation Ltd., Austin, TX). 

The light stimulus consisted of a square-wave pulse of white light (duration 1.0 second) projected through both pupils. In each eye the stimulus beam was converged to a width no more than 1.0 mm in the plane of the pupil, ensuring identical light flux in all subjects irrespective of starting pupil size (“open-loop” or Maxwellian optics). In an emmetropic eye this stimulus illuminates the central 7⁰ of retina, centered on the fovea. Two stimulus protocols were used. In the first, after 10 seconds of darkness the subject was presented with a single pulse of “standard intensity” light (3.96 log trolands) projected into both eyes: this light intensity was enough to produce approximately 30% constriction of the pupil (i.e., not bright enough to saturate the pupillary reflex or produce an afterimage). In the second protocol, performed a few minutes later and also after 10 seconds of darkness, each subject was presented with a pseudo-random sequence of 10 light stimuli of varying intensity (from 0.0 to −4.0 log units attenuation of this “standard” brightness) projected into both eyes with an inter-stimulus interval of 8 seconds (test duration 90 seconds). 

All pupil recordings were analyzed off-line using custom software written in LabVIEW (National Instruments Corporation, Ltd.) An example of a typical pupillogram is shown in Figure 1. From the raw data, cursors were applied manually to the onset and peak of the pupillary reflex response to measure the resting diameter and response amplitude (A), respectively. In addition, a first derivative of the raw data was extracted and used to estimate the peak velocity of constriction (V, in millimeters per second; note that for simplicity this outcome measure is shown unsigned throughout this text). 

Figure 1

 

Pupillogram (continuous line) and first derivative (dashed line) showing the reflex response of the pupil to a 1.0-second light stimulus (S). Cursors were manually fitted to the onset and peak of the reflex to allow measurement of the initial pupil size (diameter at onset) and the amplitude of the response (A). Peak velocity of the reflex constriction was estimated directly from the first derivative plot as shown (V).

All statistical analysis was performed using SigmaStat (Systat Software, Inc., Chicago, IL). Normality in the distribution of any variable was assessed using the Kolmogorow-Smirnov test. The correlation between dependent and independent variables was tested using linear regression. 

Results

Forty-three normal subjects were recruited to this study (24 men and 19 women) with a mean age of 39.0 years (SD 16.0, range 20–75). 

Pupillary responses to the standard intensity light stimulus were measured in a randomly selected eye of each subject. Mean response amplitude was 1.92 mm (SD 0.39) and peak constriction velocity was 5.65 mm/s (SD 1.17). The amplitude measurements show a Gaussian distribution (Fig. 2A) across a wide range (from 1.19–3.15 mm). Velocity measurements also are dispersed over a wide range (3.83–9.27 mm/s); their distribution is positively skewed (Fig. 2B) but does not depart significantly from normality (P > 0.20). A scatter plot of amplitude versus velocity measurements shows positive correlation between these variables; pupils showing larger amplitude responses to the light stimulus also achieved higher peak velocities of constriction, and vice versa (Fig. 2C). 

Figure 2

 

Frequency histograms showing the distribution of amplitude (A) and velocity (B) measurements in response to presentation of a standard intensity light stimulus in a randomly selected eye from each of 43 normal subjects. Amplitude measurements are distributed normally, but velocity measurements show a small positive skew. (C) Scatter plot of amplitude and velocity measurements, showing a positive correlation between these variables.

In the second part of this experiment each subject was presented with a pseudo-random sequence of 10 stimuli in which the intensity of the light was varied over a 4 log unit range (in 0.5 log unit steps). An example of the resulting pupillogram and its first derivative is given in Figure 3A, which shows the pupillary responses to the first three stimuli in the sequence, at intensities of 0.0, −4.0, and −1.5 log units attenuation, respectively. Measurements of the amplitude and velocity of constriction of the pupillary responses to all 10 light stimuli are plotted in Figure 3B. These data confirmed that in this subject amplitude and velocity measurements remain positively correlated across the whole range of tested stimulus intensities (linear regression coefficient R = 0.983, P < 0.001). Similar results were obtained in all 43 subjects. 

Figure 3

 

(A) Pupillogram and first derivative showing the pupillary responses to three consecutive stimuli of different light intensities (labeled #1, #2, and #3 corresponding to 0.0, −4.0, and −1.5 log units attenuation of the standard brightness, respectively). Amplitude and velocity of constriction are seen to vary according to the intensity of the light stimulus. (B) Scatter plot of amplitude and velocity measurements from the pupillary responses to all 10 stimuli in the sequence; the 4.0 log unit range of stimulus intensities used has led to a wide dispersion of amplitude and velocity measurements in this subject, but there is a positive correlation between these two outcome measures (best fit linear regression shown as solid line; regression coefficient R = 0.983, P < 0.001). Dashed lines illustrate use of this linear regression to predict the constriction velocity expected for a response amplitude of 1.0 mm (“normalized velocity” or NV).

Because of this interdependence between amplitude and velocity of constriction, it was necessary first to normalize the velocity measurements with respect to response amplitude before investigating the possible influence of pupil size and age on peak constriction velocity. This was achieved in each subject by using the regression plot to extrapolate the constriction velocity that would be expected if the stimulus intensity had been adjusted to produce exactly a 1.0 mm response amplitude (Fig. 3B, dashed lines). Only one randomly chosen eye was selected from each subject to prevent duplication of the independent variable in the subsequent analysis. The mean value for these estimates of normalized velocity was 3.56 mm/s (SD 0.40, range 2.79–4.41). A scatter plot of these normalized velocity estimates (“NV”) against corresponding measurements of the resting pupil diameter is shown in Figure 4A. No clear relationship emerges across a wide range of pupil size (3.87–7.84 mm), and linear regression analysis gives a coefficient R = 0.242 (P = 0.118). A similar approach also reveals no apparent influence from age. Figure 4B shows a scatter plot of normalized velocity estimates against a wide range of ages (20–75 years), but the linear regression coefficient is not significant (R = 0.193, P = 0.215). If the pupil measurements in the youngest 10 subjects (mean age 24, range 20-26) are compared to those in the oldest 10 subjects (mean age 65, range 53–75), no significant difference is found in their normalized velocity estimates (mean 3.51 mm/s in the younger subjects, 3.42 m/s in the older subjects, t = 0.488, P = 0.632). 

Figure 4

 

Scatter plots of normalized velocity estimates from one (randomly selected) eye of 43 normal subjects as a function of their pupil size (A) or age (B). Solid lines denote the best fit from linear regression analyses, but in neither case was a significant correlation found between these variables.

The interdependence of velocity and amplitude measurements for all recordings from both eyes in this cohort of normal subjects is illustrated in Figure 5 (N = 658 recordings, average of 8.1 measurements per tested eye). When all these data from across the entire 4-log unit range of tested stimulus intensities are considered, the previously noted positive correlation between these two variables is seen to be curvilinear. There is a slight reduction in measured velocity for response amplitudes <0.50 mm, although the correlation for all amplitudes ≥0.50 mm appears to be linear. Despite this minor nonlinearity at one end of the graph, the data can be fit well by the equation 

 

where V is velocity and A is amplitude (R = 0.919, P < 0.001). The upper and lower 95% prediction intervals either side of the regression line also are shown in Figure 5. These show that the lower limit to constriction velocity expected in normal subjects is given by the equation 

 

 

Figure 5

 

Scatter plot of amplitude and velocity estimates from both eyes of all 43 normal subjects in response to light stimuli across the whole 4.0 log unit range of tested intensities (N = 658 reflex responses). Linear regression provides a good description of the covariation of these outcome measures for all reflex responses with an amplitude >0.50 mm. Below this level there is a degree of nonlinearity with velocity estimates lower than expected. The best fit to the data from linear regression analysis (shown as solid line; equation V = 0.862 + 2.653A) has a correlation coefficient R = 0.919, P < 0.001. Dashed lines show the upper and lower 95% prediction intervals.

Discussion

In the first part of this experiment a cohort of healthy normal subjects was presented with an identical light stimulus (the open-loop optics ensuring that the light flux was similar in every subject regardless of pupil size), and yet the pupillary responses varied widely with velocity and amplitude measurements dispersed over a 3- to 4-fold range (respectively). This variability in the pupillometric recordings mirrors what is seen clinically and presumably reflects differences in the supranuclear control of reflex gain. Expressed in terms of coefficient of variation (SD/mean as %), both pupillary outcome measures in this study displayed an inter-person variability of approximately 20%, which is similar to that reported with some other dynamic ocular biometrics, such as intraocular pressure (23%), ⁶ but significantly greater than structural biometrics, such as the thickness of the central cornea (7%) ⁷ or the retinal nerve fiber layer (10%). ⁸ The pupillometric amplitude measures were distributed normally; the positive skew to the velocity estimates may have been an artefact associated with imperfect pupil capture by the relatively slow video cameras used in this study (25 Hz). The results found in this study reinforced the clinical impression that pupils in some healthy subjects may show unimpressive responses to light, in terms of extent and speed. 

These measurements of amplitude and velocity co-varied with a highly significant regression coefficient in all subjects and across a wide range of stimulus intensities. This relationship has been alluded to briefly by others, ⁹ but as far as I am aware has not been explored previously in any detail. For the weakest light stimuli eliciting the smallest pupillary responses there may be some nonlinearity in this relationship, although caution must be applied when interpreting data at this end of the graph because the signal-to-noise ratio is much worse, inevitably giving rise to proportionately greater measurement uncertainty. In practice, clinically and in research studies, much brighter light stimuli always are used for evaluating the reflex, and within this normal “working range” of stimulus intensities the relationship between amplitude and velocity measurements is strikingly linear in normal eyes. A similarly striking and tight linear relationship between amplitude and peak velocity is well established for saccadic eye movements (at any rate for amplitudes <20 degrees; thereafter the graph asymptotes indicating system saturation); the rigidity of the interdependence between amplitude, velocity, and duration of saccades in the normal population has been called the “main sequence relationship” ¹⁰ and is thought to represent the optimal speed–accuracy trade-off in the ocular motor system. ¹¹ The advantages of this relationship in the pupillary system are less clear, and it seems more likely that it simply reflects the “transfer function” of the effector (i.e., the neuromuscular apparatus and the mechanical properties of the iris). 

Before we can apply knowledge of this amplitude–velocity interdependence to patients with disease, it is important first to identify any confounding influences on this relationship. The main uncontrolled variables likely to be relevant are pupil size and age. Pupil size in the normal population varies across a 3-fold range even in complete darkness, and might be expected to affect this amplitude–velocity relationship because of differences in the mechanical properties of the iris, and in the balance of autonomic tone in the sphincter and dilator muscles. However, the results from this study suggested that pupil size does not affect constriction velocity independently of response amplitude. It may be that the sample size of 43 subjects was insufficient to detect a small effect, and certainly none of the study participants had pupils at either extreme of what can be encountered in the normal population (<2 or >8 mm) where any effect of pupil size is more likely to be seen. Nevertheless, the data presented here indicated that large and small pupils reach similar peak constriction velocities for any given amplitude of response. 

It also might be expected that older people have shown slower pupillary responses, perhaps reflecting the consequences of senescence in the ageing iris smooth muscles. A small number of previous studies have evaluated pupil kinetics and age, and have reported reduced peak velocities of constriction in older subjects, ¹² but these studies used closed-loop stimulus conditions where the smaller pupils of the elderly will have allowed less light flux to elicit the reflex. Accordingly, the same studies also found reduced reflex amplitudes in the older subjects, and it is not clear whether the velocity had been affected independently of amplitude. In this study Maxwellian optics were used to deliver an identical light stimulus to all subjects regardless of pupil size. A linear regression analysis between age and amplitude measurements under these open-loop conditions shows a small but significant negative correlation (R = 0.357, P = 0.019) indicating that the older subjects did tend to show smaller pupillary responses to the same light stimulus. However, a similar effect of age was found on velocity measurements, so that there was no evidence of a specific slowing of the reflex in the elderly over and above that explained by changes in reflex amplitude. 

In conclusion, amplitude and peak velocity measurements from the pupil light reflex are interdependent, showing a striking and tight linear correlation over the working range of stimulus intensities used in clinical practice. This relationship is unaffected by the size of the pupil or the age of the person, and probably reflects intrinsic properties of the neuromuscular apparatus and iris mechanics. The practical implication from this study for the clinician is that observation of the speed of pupillary constriction in response to a light stimulus cannot be used in isolation to make inferences about pathology – instead, it must be interpreted in the context of the response amplitude. In other words, the pupil is expected to constrict slowly whenever the response is small, and pathology can be inferred only if these slow pupillary movements are present in the face of a normal amplitude of response. For the researcher, the 95% prediction intervals for this regression plot can be used to define the lower limit of constriction velocity in the normal population (given by: $V=2.65A-0.58$). It is proposed that this lower limit is used to provide a more rigorous definition of the term “sluggish pupil,” which now can be understood to mean pupillary light reflexes that show an abnormally reduced constriction velocity relative to amplitude. An investigation of what circumstances give rise to sluggish pupils is under way. 

Footnotes

 The author alone is responsible for the content and writing of this paper.

Footnotes

 Disclosure: F.D. Bremner, None

Footnotes

 Presented at the 10th European Neuro-ophthalmology Society (EUNOS) Meeting in Barcelona, Spain, June 2011.