Comparison of Colvard pupillometer and infrared digital photography for measurement of the dark-adapted pupil diameter

Purpose

To investigate the accuracy of pupil diameter measurement using the Colvard pupillometer and to determine the learning curve for inexperienced examiners.

Setting

Texas Tech University Health Sciences Center, Lubbock, Texas, USA.

Methods

In this population study, subjects with normal pupillary behavior were tested by 1 of 2 investigators (examiner A, examiner B). After 5 minutes of dark adaptation at 1 lux, digital infrared pupil photography of the right eye was performed, followed by measurement of the horizontal pupil diameter and vertical pupil diameter with the Colvard pupillometer. The photographs were digitally analyzed to determine the horizontal and vertical pupil diameters. During phase I of the study, examiners were masked to the results of infrared pupil photography; during phase II, they reviewed the infrared pupil photography results after each testing session. Bland-Altman plots were created to detect measurement bias; results were graphed by subject test sequence to assess learning. A test difference of less than ±0.5 mm was considered clinically acceptable.

Results

Fifty-nine subjects were tested in phase I, of whom 39 had adequate infrared pupil photography for analysis; 40 were tested in phase II, of whom 34 were included. The mean age of the analyzed subjects was 27 years (range 18 to 44 years). For all subjects, the infrared pupil photography median horizontal pupil diameter was 7.09 mm ± 0.75 (SD) (range 5.44 to 8.79 mm); the median vertical pupil diameter was 7.22 ± 0.79 mm (range 5.45 to 9.10 mm). Examiner A initially had a negative bias (Colvard pupillometer value less than infrared pupil photography value) for both horizontal and vertical pupil diameter measurements, which resolved during phase I after 23 subjects were tested; 18 of the final 19 subjects tested (11 phase I, 8 phase II) showed a test difference of less than 0.5 mm for all readings. The pupil diameter did not affect the bias. Examiner B had a strong positive bias that persisted throughout the study. Testing 26 subjects in 5 sessions during phase II did not improve the accuracy. During the final testing session, 3 of 8 subjects had a test difference of 0.5 mm or more in at least 1 dimension. The pupil diameter did not affect the bias.

Conclusion

The Colvard pupillometer is susceptible to user errors causing unidirectional bias and seems to have a steep and variable learning curve.