Relative afferent pupillary defect in glaucoma: a pupillometric study


Lada Kalaboukhova
Institute of Neuroscience and Physiology/Ophthalmology
Sahlgrenska University Hospital
SE 431 80 Mölndal
Tel: +46 31 3433256
Fax: +46 31 412904


Purpose:  To study the presence of relative afferent pupillary defect (RAPD) in patients with glaucoma with the help of a custom-built pupillometer.

Methods:  Sixty-five participants were recruited (32 with open-angle glaucoma and 33 healthy subjects). All underwent standard clinical examination including perimetry and optic disc photography. Pupillary light reflexes were examined with a custom-built pupillometer. Three video sequences were recorded for each subject. Alternating light stimulation with a duration of 0.5 seconds was used, followed by a 1 second pause. Mean values of pupil area ratio (PAR), pupil contraction velocity ratio (PCVR), and pupil dilation velocity ratio (PDVR) were calculated. Receiver operating characteristic (ROC) curves were constructed for each of the three parameters. Intra-individual variability was estimated.

Results:  PAR and PDVR differed significantly between the group with glaucoma and the control group (P < 0.0001). PAR was more sensitive for glaucoma detection than the other pupillometric parameters (PCVR and PDVR). The area under the receiver operating characteristic curve was largest for PAR. At a fixed specificity of 90%, sensitivity for PAR was 86.7%.

Conclusion:  Measuring RAPD with infrared computerized pupillometry can detect optic neuropathy in glaucoma with high sensitivity and specificity. The method is fast and objective. Pupil area amplitude measurements were superior to pupil velocity measurements for the detection of RAPD in glaucoma


A relative afferent pupillary defect (RAPD) is an important sign of asymmetric or unilateral lesions of the anterior visual pathway. RAPD can be observed in cases of asymmetrical retinal or optic nerve damage, including glaucomatous optic neuropathy. Although primary open-angle glaucoma usually affects both eyes, it is often an asymmetric disease. Thus, measurement of RAPD may have potential as an objective diagnostic test in glaucoma.

RAPD can be detected in daily clinical practice with the swinging flashlight test. This procedure is described elsewhere (Levatin 1959). It is a commonly used test in neurology and neuro-ophthalmology (Wilhelm 1998), but requires some skills and experience from the examiner.

Pupillometry can be performed with the help of a number of commercially available instruments, for example the Procyon P2000 SA Pupillometer (Procyon Instruments Ltd., Haag-Streit, UK) (Wickremasinghe et al. 2005). The use of pupillometers is not widespread. Some clinical applications are: examination of the scotopic pupillary size before refractive surgery (Schmitz et al. 2003); examination of the light reflex in different drug studies (Marx-Gross et al. 2005); and measurement of daytime sleepiness (Wilhelm et al. 2001). For research purposes, custom-built pupillographs and pupillometers are commonly used (Schmid et al. 2000; Wilhelm et al. 2001).

In this study, we measured RAPD in patients with bilateral or unilateral glaucomatous optic neuropathy and compared these measurements with those of healthy subjects. Measurements were made with a custom-built infrared pupillometer using previously determined stimulus parameters (Kalaboukhova et al. 2006). Sensitivity and specificity for detecting glaucoma and variability of the method were estimated.

Materials and Methods


Patients with open-angle glaucoma were recruited from the out-patient glaucoma clinic at Sahlgrenska University Hospital, Mölndal, Sweden. Diagnosis of glaucoma was based on optic disc and retinal fibre layer evaluation (ophthalmoscopy and stereoscopic colour fundus photographs). Glaucomatous optic neuropathy was defined as presence of generalized thinning of the retinal nerve fibre layer together with unusually small rim in relation to optic disc size and/or localized defects of the retinal nerve fibre layer together with localized thinning of the neuroretinal rim (‘notch’). Corresponding visual field changes were observed in the majority of glaucoma cases. Patients with pseudoexfoliative glaucoma, pigment glaucoma and normal-pressure glaucoma were included in the study. In cases of unilateral glaucoma, the contralateral eye was classified either as normal or as ocular hypertensive. Ocular hypertension was defined as intraocular pressure of more than 21 mmHg on at least two occasions and absence of glaucomatous optic neuropathy or visual field defects.

Healthy participants were recruited from patients who had visited the eye department for reasons other than glaucoma, most commonly external eye infections. They had no signs of glaucomatous optic neuropathy and had intraocular pressure of 21 mmHg or lower.

Exclusion criteria were: angle-closure glaucoma; retinal disease; optic nerve disease other than glaucoma; amblyopia; strabismus; any intraocular surgery except uncomplicated cataract extraction; use of any miotics or other medicine affecting pupil motility; and history of any intraocular inflammation.

The study was approved by the local institutional ethics committee and followed the tenets of the Declaration of Helsinki. All participants received written information about the pupillometry and gave their written consent.


All subjects underwent standard ophthalmological examination, including perimetry (Humphrey Field Analyser II, SITA 24-2, Carl Zeiss Meditec Inc., Jena, Germany), ophthalmoscopy, Goldmann applanation tonometry and fundus photography (Nikon NFC-50, Nikon, Tokyo, Japan; film: Kodak Ektachrome 100 Pro, Eastman Kodak Company, New York, NY, USA) after pupil dilation.

Visual field test was classified as abnormal if Glaucoma Hemifield Test was abnormal or borderline and pattern standard deviation was out of the normal range (P < 10%).

Reproducibility of the visual field test was estimated based on the number of fixation errors and false negative and positive errors. The quality of the examination was judged as poor if one of these indices exceeded 20%. In cases of poor test reliability, the examination was repeated.

For RAPD measurements, a custom-built pupillometer was used (Fig. 1). It has been described previously (Kalaboukhova et al. 2006). In short, it consists of two digital high-resolution video cameras [JAI M40 (60 Hz); JAI A-S, Copenhagen, Denmark), one for each eye; two white diode lamps (each is a set of 15 small diodes, ø 1.25 mm) (Nerlite S-40 Spot Illuminator, RVSI/NER, Weare, New Hampshire, USA) for alternating light stimulation; and a background infrared illumination device (Sony HVL-IRH2, Sony, Tokyo, Japan). Two pasteboard screens (one on each side of the nose) separate the two eyes, preventing stray light entering from the contralateral side.

Figure 1.

 Pupillometer. Two separate video cameras (1); infrared diode lamp for background illumination (2); two white diode lamps generating light stimuli (3); two pasteboard screens (4).

The RAPD examination was performed in a dark room. The distance between the light sources and the examined eyes was about 18 cm, and they were placed at an angle of 10° temporal to the visual axis. Before the start of the recording, the subject was allowed about 15 seconds of adaptation to the alternating light stimuli. The intensity of the alternating stimulus light was 1000 cd/m2, which was found in a preliminary study to be the optimum light intensity (unpublished data). Three video sequences of pupillary reaction to alternating light stimulation (0.5 seconds for stimuli and 1 second for pause) were recorded in each case. Duration of each video recording was 16 seconds (960 frames). The pause between recordings was about 1 min. During the pause a dim light was used for room illumination.

During the test the subject was asked to look at a small dark red distant fixation lamp, to avoid accommodative miosis.

Measurements of pupil area were performed with the help of digital image processing using two morphological functions: gray-scale opening and closing. Description of the effect of the functions can be found elsewhere (Baxes 1994). Thresholding and binary contrast enhancement were used to increase the contrast of the image. The pupil area in each video frame was measured as the number of white pixels. The area of each pupil was presented graphically (Fig. 2).

Figure 2.

 Graphical presentation of pupillary reaction to light showing relative afferent pupillary effect (RAPD) in the left eye. Broken line, right eye; unbroken line, left eye; Amax and Amin, maximal and minimal pupil size; downward arrow (no filled circle), right eye stimulated; downward arrow (with filled circle), left eye stimulated.

A special program written in Visual Basic was used for subsequent calculations. Maximum (Amax) and minimum (Amin) pupil area and pupil area amplitude (Amax − Amin) were calculated for the right and left pupil in each stimulus cycle (Fig. 2). The mean amplitude between all stimulus cycles was computed for right and left eye during right and left eye stimulation, respectively. The mean amplitude of right and left eye was then averaged. Pupil area ratio (PAR) was defined as the ratio between the averaged mean amplitudes during right and left eye stimulation, respectively.

A ratio of the slopes of the pupil area during contraction and dilation was also computed. Definitions of pupil constriction velocity (PCV) and pupil dilation velocity (PDV) are presented in Fig. 3. Pupil constriction velocity ratio (PCVR) and pupil dilation velocity ratio (PDVR) were defined as the ratio between the mean slope during right and left eye stimulation, in the same way as for PAR.

Figure 3.

 Curves demonstrating pupillary reaction to light. Broken line, left eye; unbroken line, right eye; Max, maximum pupil area; Min, minimum pupil area; PCV, pupil contraction velocity (calculated from the middle portion of the curve between Max and Min); PDV, pupil dilation velocity (calculated from the middle portion of the curve between Max and Min).

A PAR, PCVR or PDVR value below 1 denoted a right RAPD; a value above 1 denoted a left RAPD. To facilitate comparisons, a ratio below unity was inverted and assigned as negative value. For example, a ratio of −1.5 corresponded to a right RAPD, +1.5 to a left RAPD.

Mean values of PAR, PCVR, PDVR (PARm, PCVRm and PDVRm) were calculated for three video sequences in each case.

Variability of PAR, PCVR and PDVR between sequences was estimated with the help of the within-subjects effects test. The mean deviations for three measurements from mean value were also calculated.

Receiver operating characteristic (ROC) curves were constructed for mean values of pupillometric parameters –PARm, PCVRm and PDVRm.

Correlations between PARm and visual field indexes (mean deviation (MD), pattern standard deviation (PSD)), were estimated by calculations of Pearson's correlation coefficient.

SPSS statistical package version 11.5 (SPSS, Chicago, Illinois, USA) was used for statistical calculations.


Sixty-five participants (32 with glaucoma and 33 healthy subjects) were recruited. Because of light reflexes produced by an intraocular lens, which disturbed image analysis, the results of one healthy participant were excluded from subsequent calculations. Four subjects with poor reliability of the repeated visual field examination were also excluded. Therefore, results of 30 healthy participants and 30 subjects with glaucoma were used in the following analysis. Demography of the population is presented in Table 1. Age and gender differences between the two groups were not significant. Intraocular pressure (IOP) was significantly higher (P < 0.0001) in the glaucoma group in spite of local pressure-lowering medication in eyes with glaucoma and ocular hypertension. Six patients were treated with a prostaglandin analogue, 15 with combination of prostaglandin and β-blocker and/or carboanhydrase inhibitor, and six with β-blocker and/or carboanhydrase inhibitor. Three glaucoma patients were untreated at the time of examination. Visual field indices (MD and PSD) differed significantly between eyes in the glaucoma group and control eyes (P < 0.0001).

Table 1.   Demography of the study population (n = 60).
 Glaucoma groupHealthy subjects
  1. SD, standard deviation; IOP, intraocular pressure; RE, right eye; LE, left eye; HFA, Humphrey Visual Field Analyser II; ΔMD, difference in mean deviation between right and left eyes; ΔPSD, difference in pattern standard deviation between right and left eyes.

Age (years ± SD)65 ± 1063 ± 8
Female / male14 / 1616 / 14
IOP (mmHg):
 RE [mean ± SD (range)]18.5 ± 4 (13–28)15 ± 3 (9–21)
 LE [mean ± SD (range)]19 ± 4 (12–30)15.7 ± 3 (10–21)
 Mean MD (RE / LE)1.4 / −3.610.28 / 0.25
 Mean PSD (RE / LE)5.58 / 4.631.7 / 1.6
ΔMD [dB (range)]6.3 (0.31–18.80)0.03 (−1.5 to −2.7)
ΔPSD [dB (range)]5.3 (0.15–13.40) −0.09 (−0.84 to 0.80)

Among the 30 patients with glaucoma, 17 had unilateral and 13 had bilateral glaucoma. Ten patients with unilateral glaucoma had ocular hypertension in the other eye. Two eyes with obvious glaucomatous optic disc damage and high IOP were classified as preperimetric glaucoma. Difference in IOP between the right and left eye was significantly larger in unilateral glaucoma than in bilateral glaucoma (P = 0.04). A detailed description of the glaucoma group is presented in Table 2.

Table 2.   Characteristics of the glaucoma group (n = 30).
(n = 7)
(n = 10)
(n = 13)
  1. Δ IOP, difference in intraocular pressure between right and left eyes; Δ MD, difference in mean deviation between right and left eyes; Δ PSD, difference in pattern standard deviation between right and left eyes; PARm, mean value of pupillary area ratio for three measurements; PCVRm, mean value of pupil contraction velocity ratio for three measurements; PDVRm, mean value of pupil dilation velocity ratio for three measurements.

ΔIOP [mean (range), mmHg]      4 (1–7)    4 (0–11)      2 (0–6)
ΔMD [mean (range), dB]8.3 (0.31–17.34)6.1 (0.42–18.77)5.3 (1.57–12.49)
ΔPSD [mean (range), dB]6.8 (0.28–12.17)5.3 (0.15–3.42)4.4 (0.41–9.01)
PARm [mean (range)]1.6 (1.15–3.24)1.4 (1.02–1.80)1.37 (1.10–2.01)
PCVRm [mean (range)]1.2 (1.02–1.61)1.1 (1.03–1.27)1.07 (1.01–1.13)
PDVRm [mean (range)]1.25 (1.09–1.81)1.2 (1.00–1.69)1.17 (1.04–1.40)

Comparisons of mean values of the three pupillometric parameters are presented in Table 3. The difference of PARm and PDVRm between the two groups was significant (P < 0.0001). In Fig. 4, point diagrams of PARm distribution are presented for the glaucoma and control groups (Fig. 4A), and for glaucoma subgroups (Fig. 4B).

Table 3.   Comparisons of the pupillometric parameters in glaucoma and control groups (paired t-test).
Pupillometric parametersGlaucoma
(n = 30)
(n = 30)
  1. PARm, mean value of pupillary area ratio for three measurements; PCVRm, mean value of pupil contraction velocity ratio for three measurements; PDVRm, mean value of pupil dilation velocity ratio for three measurements.

PARm [mean (range)]1.4 (1.02–3.24)1.1 (1.00–1.26)< 0.0001
PCVRm [mean (range)]1.1 (1.01–1.61)1.07 (1.00–1.17)0.242
PDVRm [mean (range)]1.2 (1.00–1.81)1.08 (1.00–1.28)< 0.0001
Figure 4.

  (A)  Point diagram of mean pupil area ratio (PARm) distribution in glaucoma and control groups. Open squares, healthy subjects; filled squares, subjects with glaucoma. (B)  Point diagram of mean pupil area ratio (PARm) distribution in different glaucoma groups. Filled squares, subjects with bilateral glaucoma; open squares, subjects with glaucoma in one eye and ocular hypertension in the other; asterisks, subjects with unilateral glaucoma.

By comparing the three recordings from each subject, the intra-individual variability of RAPD measurements was determined. Mean deviation from the mean value was 5.8%, 5.7% and 6.1% for PAR, PCVR and PDVR, respectively. The variation between the recordings was not significant (P > 0.9 for PAR and PCVR; P > 0.2 for PDVR) (Table 4).

Table 4.   Intra-individual variability.
SourceTest of within-subjects effectsMean deviation from mean value
[% (range)]
Type III sum
of squares
  1. F, correlation coefficient; PAR, pupillary area ratio; PCVR, pupil contraction velocity ratio; PDVR, pupil dilation velocity ratio; df, degree of freedom; SE, standard error.

PAR0.03120.0160.0370.9635.8 (0–22)4.2
PCVR0.11820.0500.7000.9325.7 (0.3–14)3.5
PDVR1.17020.8851.5170.2246.1 (0.3–30)5.6

The area under the ROC curves (AUC) (Fig. 5) for PARm, PCVRm and PDVRm were 0.923, 0.613 and 0.755, respectively (Table 5). The ROC curve for PCVRm was reversed because values of PCVRm had a tendency to be larger in the control group than in the glaucoma group. The AUC differed significantly from the reference line for PARm and PDVRm (P < 0.0001–0.001) (Table 5). For PARm at a cut-off point of 1.16, sensitivity was 86.7% and specificity was 90%. Sensitivity for PCVRm and PDVRm at a fixed specificity of 90% was significantly lower: 36.7% (at a cut-off point of −1.02) for PCVRm and 53.3% for PDVRm (at a cut-off point of 1.15).

Figure 5.

 ROC curves for: mean values of pupil dilation velocity ratios (PDVRm) (0.755); mean values of pupil contraction velocity ratios (PCVRm) (0.613); mean values of pupil area ratios (PARm) (0.923).

Table 5.   Statistics of the area under ROC curve for the pupillometric parameters.
 AreaSEASAsymptotic 95% confidence interval
Lower boundUpper bound
  1. PARm, mean value of pupillary area ratio for three measurements; PCVRm, mean value of pupil contraction velocity ratio for three measurements; PDVRm, mean value of pupil dilation velocity ratio for three measurements; SE, standard error; AS, asymptotic significance – level of significance between the presented parameters and the reference line.


Figure 6 presents an example of visual field maps, optic disc photographs and RAPD curves of a patient with early glaucoma of the right eye.

Figure 6.

 Patient with early glaucomatous damage of the right eye. (A) Optic disc photo of the right eye. (B) Optic disc photo of the left eye. (C) Visual field of the right eye (HFA SITA 24-2). (D) Visual field of the left eye (HFA SITA 24-2). (E) Curves demonstrating pupillary reaction to light (relative afferent pupillary defect of the right eye); broken line, right eye; unbroken line, left eye; PARm = −1.59; downward arrow (with filled circle), right eye stimulated; downward arrow (no filled circle), left eye stimulated.

Pearson's correlation coefficients were R = −0.64 (P < 0.0001) for ΔMD and PARm, and R = 0.55 (P < 0.0001) for ΔPSD and PARm(Fig. 7).

Figure 7.

 (A) Correlation between mean value of pupil area ratio (PARm) and the difference of mean deviation between right and left eyes (ΔMD) (HFA SITA 24-2) for glaucoma and control groups (R = −0.64). (B) Correlation between PARm and the difference of pattern standard deviation between right and left eyes (ΔPSD) (HFA SITA 24-2) for glaucoma and control groups (R = 0.55).

We measured RAPD in three normal subjects with neutral filters before one eye (0.3 log, 0.6 log and 0.9 log). In all three subjects there was a linear correlation between filter density and PAR. However, the slope of the regression line varied between the normal subjects (Fig. 8). Therefore, 0.6 log filter resulted in PARm values from 1.3 to 2.3. The reason for this variation is unclear and merits further study (outside the scope of this paper). In all three normal subjects, the 0.6 log neutral filter resulted in PARm far outside normal limits.

Figure 8.

 Relative afferent pupillary defect (measured as mean values of pupil area ratio − PARm) in three healthy subjects using a neural density filter (0.3 log, 0.6 log or 0.9 log) before one eye. Asterisks, subject 1; filled triangles, subject 2; open squares, subject 3.


The aim of this study was to measure RAPD in patients with glaucoma using a custom-built infrared pupillometer.

In daily ophthalmological practice the swinging flashlight test is used to detect RAPD. In spite of the subjectivity of the test, it has been reported that this simple clinical procedure in the hands of an experienced observer may be more sensitive than pupillometric methods (Wilhelm 1998).

With the equipment we used, artifacts caused by blinking and by light reflexes in pseudophakic eyes with small pupils were major problems. We reduced blink artifacts by removing the whole stimulus cycle containing the blink from analysis. Light reflexes were reduced or removed by digital image processing. Still, in one case with pseudophakia with small pupils, RAPD analysis could not be performed.

The custom-built pupillometer based measurements on pupil area instead of commonly used pupil diameter (Loewenfeld 1999). We did not allow eyelids and eyelashes to interfere with the measurement. Each film was scrutinized for such artifacts.

The results showed that amplitude comparison was superior to velocity comparison for the detection of RAPD in glaucoma. At a cut-off point of 1.16 for PARm, specificity was 90% and sensitivity 86.7%.

Another interesting result was that of the two pupil velocity parameters: only dilation velocity showed any asymmetry between eyes with asymmetrical glaucoma damage. Dilation velocity was slower in the most affected eye, while contraction velocity did not seem to be affected by neuronal loss. The explanation for this result is obscure.

In a few cases we registered a small RAPD in healthy participants. It has been reported that normal subjects may have a small asymmetry of pupillary reflexes (Wilhelm 1998). It has been suggested that asymmetrical innervation of the pupillary muscles or asymmetrical neural input of the right and left eyes may cause such asymmetry (Kawasaki et al. 1995). In our study we averaged pupillary movements between right and left eyes, making the latter explanation less likely. Asymmetric cataract has also been claimed to cause RAPD – interestingly, in the less affected eye (Lam & Thompson 1990). However, in the present study we excluded subjects with dense cataract.

Two patients in this study had an obvious difference in glaucoma damage between the right and left eye, but no significant RAPD. The reason for this finding was not clear. It is well known that it can take several years before neural damage becomes visible as visual field defects (Quigley et al. 1989). It could be that the neuronal damage in these patients was more symmetrical than was obvious from the visual field results. It was interesting that both patients had central corneal thickness exceeding 600 µm.

In conclusion, measuring RAPD with infrared pupillometry is an objective method to detect glaucoma with good sensitivity and specificity. Although sensitivity and specificity were high, they fell short of being ideal for screening purposes. At present, pupillometry on its own cannot serve as a screening device for glaucoma. This also applies to other diagnostic tests in glaucoma. Further refinement of the measurement techniques and the algorithms for automated curve analysis are needed before pupillometry can become a part of the routine diagnostic armamentarium in glaucoma.


This study was supported financially by Göteborgs Läkaresällskap, Herman Svensson Foundation, För Blinda Och Synsvaga and Föreningen De Blindas Vänner.