Diagnostic capability of Pulsar perimetry in pre-perimetric and early glaucoma

This study aimed to compare the diagnostic capability of Pulsar perimetry (Pulsar) in pre-perimetric glaucoma (PPG) and early glaucoma (EG) with that of Flicker perimetry (Flicker) and spectral-domain optical conference tomography (SD-OCT). This prospective cross-sectional study included 25 eyes of 25 PPG patients, 35 eyes of 35 EG patients, and 42 eyes of 42 healthy participants. The diagnostic capability using the area under the curve (AUC) of the best parameter and agreement of detectability between structural and functional measurements were compared. For PPG patients, the AUC of Pulsar, Flicker, OCT-disc, and OCT-macular was 0.733, 0.663, 0.842, and 0.780, respectively. The AUC of Flicker was significantly lower than that of OCT-disc (p = 0.016). For EG patients, the AUC of Pulsar, Flicker, OCT-disc, and OCT-macular were 0.851, 0.869, 0.907, and 0.861, respectively. There was no significant difference in AUC among these methods. The agreement between structural and functional measurements expressed by kappa value ranged from −0.16 to 0.07 for PPG and from 0.01 to 0.25 for EG. Although the diagnostic capability of Pulsar in the PPG and EG groups was equal to that of Flicker and SD-OCT, the agreements between structural and functional measurements for both PPG and EG were poor.

. Demographics of normal participants and glaucoma patients. PPG = pre-perimetric glaucoma; EG = early glaucoma; IOP = intra-ocular pressure; SAP = standard automated perimetry; MD = mean deviation; PSD = pattern standard deviation. Data are expressed as mean ± standard deviation. Statistical comparison was performed with Bonferroni test. Table 6 shows the statistical comparison of the test duration and reliability index for both the false positive (FP) and false negative (FN) response rates between Pulsar and Flicker perimetry. The reliability index of Flicker was significantly worse than that of Pulsar in the control group, the PPG group, and the EG group (p < 0.001). Test duration of Pulsar was significantly shorter than that of Flicker in the control group (p < 0.001), the PPG group (p < 0.001), and the EG group (p < 0.001).

Discussion
In the current study, we found that the diagnostic capability of Pulsar was equal to Flicker and SD-OCT for both PPG and EG patients. However, the agreement of detectability of structural and functional measurements was poor, and structural measurements appeared to be more sensitive than functional measurements in PPG patients. In contrast, functional measurements using Pulsar were equal to structural measurement in EG patients.
To date, many studies have reported on the diagnostic capability for PPG using specific functional measurements with methods such as FDT (AUC = 0.666 to 0.802) 20, 35-37 , Flicker (AUC = 0.800) 20 , SWAP (AUC = 0.660   Table 2. Comparison of each parameter measured with Pulsar, Flicker, and SD-OCT. PPG = pre-perimetric glaucoma; EG = early glaucoma; CP = corrected probability; OCT = optical coherence tomography; CDR = cup to disc ratio; DD = disc diameter; mRNFL = macular retinal nerve fiber layer; mGCL = macular ganglion cell layer and inner plexiform layer; mGCL + + = mRNFL and mGGCL. Data are expressed as mean ± standard deviation. Statistical comparison was performed with a Bonferroni test. to 0.704) 20,37 , and Pulsar (AUC = 0.733) 31 . Multiple studies have also investigated the diagnostic capability for PPG using structural measurements with methods such as OCT (AUC = 0.527 to 0.938) [38][39][40] , HRT (AUC = 0.740 to 0.914) 39,40 , and GDx (AUC = 0.688 to 0.894) 39 . Although it is difficult to make a direct statistical comparison between the AUC results of the current study and those of previous studies because of differences in sample size and characteristics, we report an AUC for PPG of 0.733 using Pulsar, similar to the results of previous studies on specific functional measurements 20,31,[35][36][37] . Although the AUC from Pulsar did not differ from other devices in the current study, the AUCs from specific functional measurements were slightly lower than those that have been reported by previous studies. When sensitivity values at the best cut-off or at fixed specificity were compared for each device, the structural measurements from OCT appeared more sensitive than the specific functional measurements reported by the current study and by a previous study 31 . This might be because PPG was determined by clinical structural assessment of the optic disc shape based on the general mechanism of pathogenesis 1, 41, 42 . The best parameter for discriminating between the control group from the PPG patients was cup shape, not cpRNFL thickness. Indeed, the diagnostic capability of structural measurements of the optic nerve head is better than cpRNFL thickness for classifying PPG 43,44 . For EG, the diagnostic capabilities using AUC of both Pulsar and Flicker were equal to those determined from SD-OCT. This is in agreement with a previous study reporting that the diagnostic capability of Pulsar using AUC was equal to FDT, HRT II, and GDx VCC 31 . Functional measurements appear to be more sensitive for EG diagnosis than structural measurements in both a previous study 31 and the current study; however, we report the opposite for PPG diagnosis. This might be due to the fact that all EG patients had abnormal SAP results corresponding with Anderson-Patella criteria 45 in addition to a glaucomatous optic nerve head. Additionally, as glaucoma progresses from PPG, superior or inferior RNFL thinning occurs along with changes to the optic nerve head. Thus, it could be reasonably expected that functional measurements are more sensitive than structural measurements for EG diagnosis.
Although the sensitivity for PPG at best cut-off of Pulsar was lower than the sensitivity of OCT, specificity at best cut-off was higher than OCT. However, both sensitivity and specificity for EG at best cut-off with Pulsar was equal to OCT. The agreement between structural and specific functional measurements in both PPG and EG was poor. However, PPG and EG were detected with either structural or specific functional measurements with 100% sensitivity (Fig. 2). There was no method able to accurately detect glaucoma using only one parameter. Based on the current study, the combination of OCT for structural measurement and Pulsar for selective functional measurement should be recommended to reliably diagnose PPG and EG.
Reliability indices of FP and FN for Flicker were worse than Pulsar, and an especially high FP rate was demonstrated for Flicker in both PPG and EG. Flicker fusion frequency was measured at each test point with a fixed high contrast stimulus of 0 dB, while contrast sensitivity was measured with a fixed temporal frequency of 10 Hz for Pulsar. Eyes with PPG or EG, such as those investigated in this study, can still respond to Flicker stimuli, although with decreased sensitivity. Flicker is difficult for PPG or EG patients to accurately respond to, as the flickering target is close to threshold at slightly decreased sensitivity regions. It was reported that threshold variability increases even the SAP measurements at slightly decreased sensitivity region 46 .
The test duration of Pulsar was shorter than that of Flicker despite the use of the same tendency oriented perimetry (TOP) algorithm. This may be due to the difference in number of test points between Pulsar and Flicker. The 32 P test point of Pulsar is similar to the original Octopus 32 test point program with a 6-degree interval, but the 4 points at the superior and inferior were each removed. In contrast, Flicker was measured with the original Octopus 32 test point program. Another reason for the difference in stimulus presentation time might be that a presentation time of 500 msec was applied for Pulsar, but only a 1 sec presentation time was applied for Flicker.
Pulsar and Flicker each have several advantages and disadvantages. Pulsar may have the advantage of ease of use and less fatigue compared with Flicker because it demonstrated good reliability indices and a shorter test duration in the current study, and a previous study reported that Pulsar is not associated with a learning effect 47 , but that Flicker is 48 . However, it was reported that Pulsar was affected with intraocular straylight as well as SAP 49 . In contrast, Flicker didn't affect ocular media opacity 50 . Thus, Pulsar may have the disadvantage of robustness to media opacities compared with Flicker.
The current study's main limitation was that the rate of glaucoma type was different between PPG and EG, and PPG patients in particular had almost normal tension glaucoma. Further studies will therefore be required to confirm our results.
In conclusion, the diagnostic capability of Pulsar for PPG and EG was equal to that of Flicker and OCT. However, the agreement between structural and functional measurements for PPG and EG was poor. The  Table 3. Results of receiver operating characteristic analysis between control and pre-perimetric glaucoma eyes. AUC = area under the curve; SE = standard erro; Se = sensitivity; Sp = specificity; src; spatial resolution contrast; PPG = pre-perimetric glaucoma; EG = early glaucoma; CP = corrected probability; OCT = optical coherence tomography; CDR = cup to disc ratio; DD = disc diameter; mRNFL = macular retinal nerve fiber layer; mGCL = macular ganglion cell layer and inner plexiform layer; mGCL + + = mRNFL and mGCL + . Data were expressed as mean ± standard deviation. The highest AUC were expressed by italic bold numbers.
structural measurements from OCT were more sensitive than the specific functional measurements from Pulsar for PPG, while specific functional measurements by Pulsar were more sensitive than the structural measurement by OCT for EG. Therefore, a combination of structural and functional measurements is recommended to reliably diagnose early glaucoma.

Methods
This prospective cross-sectional study was reviewed and approved by the Kitasato University Hospital Ethics Committee (no. B14-129  Table 4. Results of receiver operating characteristic analysis between control and early glaucoma eyes. PPG = pre-perimetric glaucoma; EG = early glaucoma; CP = corrected probability; OCT = optical coherence tomography; CDR = cup to disc ratio; DD = disc diameter; mRNFL = macular retinal nerve fiber layer; mGCL = macular ganglion cell layer and inner plexiform layer; mGCL + + = mRNFL and mGGCL + . Data are expressed as mean ± standard deviation. The highest AUC are expressed by italic bold numbers in each device. The diagnosis of glaucoma was determined via a fundus examination using slit-lamp indirect ophthalmoscopy and 90-diopter lens by one of three glaucoma specialists (MK, KM, or NS) and based on previous SAP results. All glaucoma patients and normal participants underwent a comprehensive ophthalmic examination, including noncycloplegic refraction testing, visual acuity testing at 5 meters using a Landolt ring chart, intraocular pressure measurement, ocular axial length measurement, and slit-lamp and fundus examination by a glaucoma specialist (MK, KM, or NS). Glaucoma patients were included in this study if they had a corrected visual acuity of 20/20 or better, cylindrical power of −1.50 diopter or less, and spherical equivalent of −8.00 to +5.00 diopter. These criteria were also applied to normal participants, with the added criteria of an intraocular pressure of 21 mmHg or less, a normal optic disc appearance, and no ophthalmic diseases in the absence of refractive error.  After comprehensive ophthalmic examination, all glaucoma patients and normal participants underwent an initial SAP measurement. This SAP measurement was performed with an HFA (Carl Zeiss Meditec, Dublin, CA) 24-2 or 30-2 SITA Standard. SAP results were considered reliable if the fixation loss was <20%, the false positive rate was <15%, and the false negative rate was <33%.
Early glaucoma patients. After SAP measurement, glaucoma patients were classified as EG if they showed structural glaucoma changes such as rim thinning, notching, and nerve fibre layer thinning or defects, and if they showed abnormal SAP results corresponding with Anderson-Patella criteria 45 .   Pulsar perimetry. Pulsar perimetry was performed using the Octopus 600 perimeter 32 P TOP algorithm.
The 32 P test point is similar to the original Octopus 32 test point program with a 6 degree interval, but the 4 points at the superior and inferior were each removed because of limitations of the angle of field of the monitor. The stimulus consisted of a circular, sinusoidal, 5-degree diameter grating pattern that was presented for 500 msec. The stimulus underwent a counter phase pulse motion at 10 Hz, in which both spatial resolution (from 0.5 to 6.3 cycle/degree on a 12-step log scale) and contrast (from 3 to 100% on a 32-step log scale) were simultaneously modified. Threshold sensitivity is expressed in spatial resolution contrast units (src). Refractive error was corrected to distance by inserting trial lenses with the spherical equivalent correction into the eye piece. The presentation ratios of FP and FN catch trials were configured to 10% of the total number of stimuli presented for Octopus 600 testing reliability, which corresponds with those of the HFA SAP performed with the SITA protocol.
Flicker perimetry. Flicker perimetry was performed using the Octopus 311 perimeter (Haag-Streit, Koeniz, Switzerland) 32 TOP algorithm. The stimulus consisted of a Goldmann size III (0.43 degree) target with a luminance of 1527 cd/m 2 (4800 apostilbs) that was presented for 1 sec. The flicker targets were presented under a supra-threshold condition with a background luminance of 10 cd/m 2 (31.5 apostilbs), and critical flicker frequency values were evaluated at each test point. Threshold sensitivity is expressed in critical flicker frequency (Hz). The presentation ratios of FP and FN catch trials were configured to 10% of the total number of stimuli presented for Octopus 311 testing reliability, which corresponds with those of HFA SAP performed with the SITA protocol.
SD-OCT. SD-OCT imaging was performed using 3D OCT-2000 version 8.1.1 (Topcon, Tokyo, Japan) in the 3D optic disc horizontal raster scan mode (OCT-disc) with a 512 × 128 scan resolution and 6 × 6 mm scan area and in the 3D macular vertical raster scan mode (OCT-macular) with a 512 × 128 scan resolution and 7 × 7 mm scan area. This device operates at a speed of 50,000 A-scans per second and has a depth and lateral resolution of 6 and 20 μm or less, respectively. It requires a pupil size of 2.5 mm or larger for imaging. Ocular magnification was corrected based on Littmann's formula 51 .
Outcome measures and exclusion criteria. The main outcome measures were the diagnostic capability of each device using the best cut-off parameter for discriminating between healthy and glaucomatous eyes and the agreement of detectability between structural and functional measurements. The secondary outcome measures were the comparison of reliability indices, including FP and FN, and the test duration between Pulsar and Flicker. All examinations were performed within a three-month period. The results of the first examination were excluded to avoid learning effects. Right eye results were converted to left eye format for analysis. The exclusion criteria were as follows: fixation loss >20% and false positive rate >15% in HFA measurement; reliability factor >15%, which is the average of the FP and FN rates in Flicker and Pulsar; and image quality index <30 in SD-OCT.
Statistical analysis. Normality of the data distribution was evaluated using the Shapiro-Wilk test. Test results were compared using either paired t-tests or Wilcoxon signed-rank tests. A Bonferroni test was used to correct for multiple testing. The best cut-off parameter for each device for discrimination between healthy and glaucomatous eyes was decided by the highest AUC based on receiver operating characteristic analysis. The detectability of each device was assessed using the AUC of the best cut-off parameter by the DeLong method. Kappa statistics were calculated to evaluate agreement of detectability between structural and functional measurements. All data were analysed using commercially available SPSS version 22.0 (IBM Japan Ltd, Tokyo, Japan) and MedCalc version 16.1 (MedCalc Software, Ostend, Belgium).