Comparison of the Humphrey Field Analyzer and MP-3 Microperimeter in Patients With Glaucoma, Classified by Severity and Misalignment of Test Points

Abstract

Purpose: To compare Nidek MP-3 microperimetry and ZEISS Humphrey Field Analyzer (HFA) visual field (VF) results in patients with primary open-angle glaucoma (POAG), classified by VF defect severity and to describe a novel method for simulating sensitivity distribution changes assumed to be due to fixation errors.

Methods: This retrospective study used the MP-3 and HFA 10-2 tests to examine the VF in eyes with mild, moderate, or severe POAG, using 68 identical test points. Reliability indices and measurement durations were compared. The relationship between pointwise sensitivities of the devices was analyzed.

Results: Thirty-five eyes from 23 patients (10 mild, 12 moderate, and 13 severe POAG) were evaluated. In the severe POAG group, the MP-3 had a significantly lower false-positive (FP) rate and higher false-negative rate than those of the HFA 10-2 (all, P = 0.03). A significant negative correlation existed between the root mean square error, derived from regression analysis comparing the VF sensitivity between devices at each test point, and VF sensitivity of the MP-3 (P < 0.01). In 18 eyes, a shift in the sensitivity distribution occurred, with a significant correlation between the HFA 10-2 mean deviation and the presence of this shift (P < 0.01).

Conclusions: Reliability indices differed significantly between the two devices in severe POAG cases. A shift in the sensitivity of the test points of the two VF tests was detected in approximately one-half of the eyes.

Translational Relevance: The MP-3 provides lower FP rates in advanced POAG and may assist in the management of this patient cohort.

Introduction

The Humphrey Field Analyzer (HFA; Carl Zeiss Meditec, Dublin, CA, USA) is a standard automated perimetry device used to assess glaucomatous visual field (VF) damage.¹ However, it is widely acknowledged that VF testing devices are subject to significant variability, which can potentially hinder timely detection of disease progression.^2,3 One reason for this variability is eye movements during the test.^4,5 To overcome this issue, the MP-3 microperimeter (Nidek, Aichi, Japan) is equipped with an autotracking system whereby the position of the retina is accurately followed throughout the VF test, thereby enabling stable and precise projection of the target stimulus onto the retina. 

Several studies have compared or evaluated retinal sensitivity of the 10-2 field HFA (HFA 10-2) and MP-3 tests in patients with glaucoma^6–8 and assessed the structure–function relationship and test–retest reproducibility of the two devices. Previous studies have found a significant correlation between mean retinal sensitivities determined by each device, as well as similar structure–function relationships between mean retinal sensitivities (as measured with each device)^6,8 and ganglion cell and inner plexiform layer thickness.⁷ Further study has reported the structure–function relationship between pointwise VF sensitivity and corresponding ganglion cell complex (GCC) thickness to be improved when using the MP-3 compared to the HFA 10-2.⁸ However, no study has compared the reliability indices, including false-positive (FP) and false-negative (FN) response rates, of the devices across different severities of glaucoma. Furthermore, no study has compared the sensitivity distribution of each VF test point between the two devices, as has been done with the HFA and the “imo” head-mounted perimeter (CREWT Medical Systems, Tokyo, Japan).⁹ Understanding the difference between the reliability indices of the HFA 10-2 and MP-3 tests is crucial in contextualizing new research on MP-3 microperimetry within the conventional literature utilizing HFA. Therefore, the aim of this study was to compare the reliability indices and sensitivity distribution of the HFA 10-2 and MP-3 tests in patients with primary open-angle glaucoma (POAG), classified by disease severity. 

Methods

Study Design and Participants

This was a retrospective, single-center study that involved patients with POAG who visited the Nagoya University Hospital (Nagoya, Japan) between April 2023 and August 2024. This study was approved by the Institutional Review Board of Nagoya University Graduate School of Medicine (approval no. 2017-0283-11456). The study protocol adhered to the tenets of the Declaration of Helsinki and the Japanese Guidelines for Life Science and Medical Research. The Institutional Review Board waived the requirement for informed consent, due to the retrospective nature of the study and its minimal risk to participants, as the data were collected as part of the routine care of the patients. Patients were included in the study if they met the following criteria: (1) a clinical diagnosis of open-angle glaucoma, based on the presence of glaucomatous VF defects, including three or more contiguous total deviation points at P < 0.05, two or more contiguous points at P < 0.01, a 10-dB difference across the nasal horizontal midline at two or more adjacent points, or a mean deviation (MD) worse than −5 dB,¹⁰ on HFA 24-2 or HFA 30-2 and clinically determined glaucomatous changes (based on the optic disc appearance on fundoscopy and the presence of nerve fiber layer defect or GCC thinning in optical coherence tomography [OCT] with mydriasis); (2) the identification of glaucoma as the only pathology causing VF damage; (3) the absence of other systemic or ocular disorders, including cataracts (except clinically insignificant senile cataract (Lens Opacities Classification System III <grade 2 nuclear opalescence/color and <grade 2 cortical and posterior subcapsular cataracts); and (4) age of ≥20 years. To compare the results of MP-3 and HFA 10-2 tests across different glaucoma severities, we adopted the modified glaucoma staging system¹¹ to stratify the included eyes, based on the MD parameter of HFA 10-2: mild (>−6 dB), moderate (−12 to −6 dB), or severe (<−12 dB) glaucoma. 

The best-corrected visual acuity (BCVA) was measured using a standard decimal visual acuity chart. The axial length was the average of five or more measurements using partial coherence interferometry (ZEISS IOLMaster with Advanced Technology 5 software). The anterior segment was examined using slit-lamp biomicroscopy after mydriasis, performed by an ophthalmologist (TI or KY). All examinations were performed by skilled certified orthoptists or ophthalmologists with at least 3 years of training. All records were separately reviewed by two ophthalmologists (TI and MM) and verified for agreement. In cases of discrepancy, two ophthalmologists (TI and KY) considered the accuracy of the measurements and jointly reached a decision about inclusion. 

Optical Coherence Tomography

The thickness of the GCC was measured using OCT (Nidek RS-3000 Advance). The OCT system was equipped with a tracking feature that reduces the effects of eye movements, thereby allowing for a more precise measurement of the thickness at each retinal point. In all cases, the synchronous serial interface quality was confirmed to be 7/10 or better. The GCC thicknesses of the superior and inferior semicircular retinal areas were averaged to obtain the mean thickness. 

Visual Field Testing

The HFA 10-2 and MP-3 VF tests were conducted in no particular order within a month apart. No patient had any additional treatments or surgeries between the two VF tests. In cases where the two VF tests were not conducted on the same day, we confirmed that the patients had stable glaucoma, based on well-controlled intraocular pressure and progression-free GCC thinning or progression-free VF defect, per OCT or retest HFA, respectively. The HFA 10-2 test was performed in a dim room with a correcting lens holder, based on the refractive error measured by the KR-8900 Autorefractor/Keratometer (Topcon, Tokyo, Japan), whereas the MP-3 test was carried out without a correcting lens holder or spectacles because of the equipped automatic refraction-correcting function. A white-on-white HFA 10-2 measurement was carried out using the Swedish interactive threshold algorithm (SITA) standard test and standard Goldmann III stimulus size. The MP-3 test was conducted using the 4-2 full-threshold staircase strategy with a standard Goldmann III stimulus size and 68 test points that were positioned identically to those in the HFA 10-2 test grid. The background luminance of the MP-3 was set to 31.4 apostilb (asb), which was identical to that of the HFA 10-2. The maximum luminance and stimulus dynamic range of the MP-3 were set to 10,000 asb and 0 to 34 dB, respectively. 

Only patients with reliable VFs were included in the analyses. Reliable VFs of the HFA 10-2 test were defined as those having a fixation loss (FL) rate < 20%, FP rate < 15%, and FN rate < 33%, as determined by the SITA standard.¹⁰ Reliable VFs of the MP-3 test were similarly defined as those having an FP rate < 15% and FN rate < 33%; FL was not applicable because the MP-3 has an auto-tracking system, thereby enabling precise projection of the stimulus only at the predefined retinal positions. 

Shifting of the Test Points

The HFA 10-2 field was inverted superoinferiorly for comparison with the MP-3 results. In cases in which the sensitivity of either system was reported as <0 dB, it was censored at –2 dB, as in previous reports.^12,13 HFA 10-2 measurements, unlike MP-3 measurements, can be prone to fixation instability because of the absence of an eye-tracking system. Therefore, a novel method was considered for simulating changes in the sensitivity distribution of the test points, assuming these occur as a result of fixation errors, and shifts in the sensitivity distributions between the HFA 10-2 and MP-3 were assessed as follows. 

Thirteen superimposed patterns were generated by shifting all HFA 10-2 test points from their corresponding MP-3 counterparts by up to two points (4°) horizontally and vertically, within a shifted distance of 4°. For each superimposed pattern, a univariable linear regression analysis between all corresponding pointwise sensitivities of HFA 10-2 and MP-3 was performed to identify the pattern with the lowest root mean square error (RMSE) of the actual sensitivity of the test points for the regression line. The position of this pattern was then defined as the “alignment position” (Fig. 1). The allowable range of the horizontal or vertical shifts (two points, 4°) was determined, based on the finding that eye movements within 2° occur in approximately 70% of the overall fixation time in patients with glaucoma.^4,14 The RMSE and P value were calculated for the superimposed pattern, free of horizontal/vertical movement, and for the alignment position. Furthermore, to study the variability of the sensitivities of HFA 10-2 and MP-3 points without horizontal or vertical shifts, the RMSE was calculated by comparing their actual sensitivities on the MP-3. To analyze the relationship between the sensitivities of HFA 10-2 and MP-3 points after correcting for the effect of the shift, we identified the common test points between the two perimetries at the alignment position for all eyes, and we calculated the slope of the univariable linear regression analysis between their sensitivities. 

Figure 1.

 

Representative case for comparing the alignment points of the HFA 10-2 and MP-3. (A, B) Representative case of POAG of the left eye in a 76-year-old male. Sensitivity results for the MP-3 (A) and HFA 10-2 (B) are shown. Thirteen alignment patterns were generated by shifting the vertically inverted HFA 10-2 results (C) against the MP-3 results (A) by one point at a time, horizontally or vertically, within a total shift distance of 4°. The pattern with the lowest RMSE from the regression line of the univariable linear regression analysis between sensitivities of each corresponding point of HFA 10-2 and MP-3 is defined as the alignment position. (D) For the superimposed position without any horizontal or vertical shifts, the RMSE is 5.69. (E) In the alignment position, shifted zero points horizontally and one point vertically, the RMSE is 4.36.

Statistical Analysis

Before the analysis, all left-eye data were converted to a right-eye format. For the BCVA, decimal values were converted to the logarithm of the minimum angle of resolution (logMAR) units. Generalized linear regression analyses were performed between the three different POAG severity groups. Univariable linear regression analysis was performed to compare the sensitivities of the HFA 10-2 and MP-3 at all common test points, after alignment. In addition, we conducted logistic regression analyses with backward stepwise model selection, using the Akaike information criterion to determine whether there was a shift in the VF sensitivity distribution between the two devices. Explanatory variables included age, BCVA (in logMAR units), axial length, mean GCC thickness, and HFA 10-2 indices; these included MD, pattern standard deviation (PSD), FL rate, FP rate, FN rate, foveal sensitivity, and measurement duration. 

For the mixed-effects linear/logistic model regression analyses, the right or left eye and the patients were treated as random effects. Statistical significance of the results was set at P < 0.05. Analyses were performed using scikit-learn 0.24.0, based on Python 3.6.7 and R 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria). 

Results

Fifty-eight eyes from 31 patients with POAG initially underwent VF measurements. We excluded eight eyes of seven patients, based on the HFA 10-2 reliability indices alone; nine eyes of nine patients, based on those of the MP-3 alone; and six eyes of five patients, based on those of both HFA 10-2 and MP-3. Thirty-five eyes from 23 patients were consequently analyzed. Demographic information of the included cases is presented in Table 1. For 26 eyes, HFA 10-2 and MP-3 tests were carried out on the same day; the HFA 10-2 was performed before the MP-3 test in all 26 eyes. For patients who had the tests completed on separate days, the HFA 10-2 test preceded the MP-3 test in seven eyes, whereas the MP-3 test preceded the HFA 10-2 test in two eyes. The mean ± SD (range) of the MD of the excluded cases was –14.2 ± 9.7 dB (–31.3 to –0.5), with a statistically significant difference between the MD of the included and excluded eyes (P < 0.01). We found significant differences in measurement durations between the different severities of POAG for the HFA 10-2 but not for the MP-3 (P = 0.03 and P = 0.76, respectively) (Table 2). There were no significant differences in the reliability indices of either device across all severities. 

Table 1.

 

Demographic Information

 

Table 2.

 

Comparison of Demographic Information Among the Mild, Moderate, and Severe POAG Groups

 

Table 3 presents comparisons of the reliability indices and measurement durations between the HFA 10-2 and MP-3 tests in each POAG severity group. The duration of the VF testing was significantly longer for the MP-3 than for the HFA 10-2 across all severities (mild, P < 0.01; moderate, P = 0.03; severe, P < 0.01). With regard to the reliability indices, a significant difference in the FP and FN rates was found between the HFA 10-2 and MP-3 in the severe group (both P = 0.03); however, no significant difference was observed in either index between the two devices in the mild group (P = 0.79 and P = 0.57, respectively) or moderate group (P = 0.88 and P = 0.30, respectively). In 18 of the 35 eyes, the alignment position differed from the superimposed position without a horizontal or vertical shift (Table 4). 

Table 3.

 

Comparison of Reliability Indices and Measurement Durations Between the HFA 10-2 and MP-3 in Each POAG Severity Group

 

Table 4.

 

RMSE and P Values Without and With Horizontal or Vertical Shifts

 

A map of the slope of the univariable linear regression line for the sensitivities of the HFA 10-2 and MP-3 at all common test points after position alignment is displayed in Figure 2A. In the univariable linear regression analysis of all VF sensitivities between the HFA 10-2 and MP-3 after alignment, the P value was <0.01 (y = 0.75x + 2.0). The P value in the univariable linear regression analysis without alignment was also <0.01 (y = 0.80x + 2.7). Figure 2B shows the RMSE for each of the 68 VF test points, calculated from the regression analysis of the HFA 10-2 and MP-3 sensitivities in the superimposed position without horizontal or vertical shift. Figure 2C shows the average MP-3 sensitivity for each of the 68 test points. A significant but moderately weak negative correlation was found between the average MP-3 sensitivity and the RMSE in the regression analysis of the sensitivities of both devices (Fig. 2D), indicating greater discrepancy between the VF sensitivities of the HFA 10-2 and MP-3 at lower MP-3 sensitivity levels (R² = 0.29, P < 0.01). 

Figure 2.

 

Heat maps of the comparison of the visual field sensitivities of the HFA 10-2 and MP-3, RMSE, and mean MP-3 sensitivity for each test point; the graph shows the relationship between the RMSE and the MP-3 VF sensitivity. (A–C) Each tile in the field is colored in grayscale, with darker and brighter shades for higher and lower values, respectively (A, B), and with darker and brighter shades for lower and higher values, respectively (C). The heat maps are shown in accordance with the VF test points in MP-3 perimetry of the right eye, considering that the position of the retina was accurately followed throughout the VF test. The slope of the univariable linear regression line between HFA 10-2 and MP-3 sensitivities was calculated, after aligning the position at each of the 68 test points (A). The RMSE calculated from the regression analysis of the VF sensitivity of the HFA 10-2 and MP-3 (B), as well as the mean value for the VF sensitivity of MP-3 (C), were calculated without any horizontal or vertical shift. (D) Finally, the RMSE values derived from the regression analysis of each of the 68 VF stimulus points are plotted against the MP-3 VF sensitivity. In the univariable linear regression analysis, P < 0.01 (black approximate line, y = −0.17x + 8.1).

Furthermore, we analyzed the predictability of the shift in alignment position. There were significant correlations between shifting and explanatory variables, including the MD and PSD of the HFA 10-2 test (Table 5), indicating that higher MD or lower PSD values of the HFA 10-2 measurements were correlated with more cases in which a shift in the sensitivity distribution of one point or more was observed in the alignment position (all, P < 0.01). Multivariable logistic regression analysis, which adapted a backward stepwise model selection, also revealed that higher MD values of the HFA 10-2 measurements were correlated with the presence of a shift of one point or more (P < 0.01). 

Table 5.

 

Univariable Logistic Regression Analyses Between Shifting and Explanatory Variables

 

Discussion

In this study, we compared HFA 10-2 and MP-3 perimetry results in patients with mild, moderate, or severe POAG. We found significant differences in reliability indices between the two devices in the severe group and a shift in the distribution of pointwise sensitivities between the two VF measurements in approximately one-half of the eyes. Of note, our findings revealed that the autotracking system of the MP-3 and the SITA strategy of the HFA 10-2 have distinct impacts on the reliability and sensitivity of VF measurements in severe POAG cases. Clinicians should consequently consider these device-specific characteristics when interpreting VF results, particularly in advanced disease stages. Similar to the findings of previous reports indicating that microperimetry, which directly projects stimuli onto retinal regions of interest, is suitable for investigating patients with macular diseases who have lost central vision,^15–17 it is considered that the MP-3 would be effective in evaluating patients with POAG whose central visual field is severely impaired. Our study provides new insights into optimizing perimetry strategies for the better management of POAG, potentially leading to patient-specific and effective care. 

A previous report⁸ compared mean recorded sensitivities using both tests and demonstrated that the MP-3 had a mean sensitivity approximately 3 dB lower than that of the HFA 10-2. By taking into account the shifting of the sensitivity distributions of both devices, we examined the retinal sensitivity at each VF test point, after the hypothetical alignment of the two measurements. We found that the MP-3 retinal sensitivity was assumed to be approximately 70% of that of the HFA 10-2. This difference may be attributable to the varying severity of cases between the previous report⁸ and our study. However, this finding also indicates that, as the mean sensitivity increases, the difference in retinal sensitivity results from the HFA 10-2 and MP-3 will exceed 3 dB. This factor may be one reason for the larger difference in mean sensitivity values between the HFA 10-2 and MP-3 in a previous report¹⁸ on patients with a mean sensitivity of approximately 22 dB with the HFA 10-2 (in which the difference was approximately 9 dB), compared with a report⁸ on patients with a mean sensitivity of approximately 18 dB with the HFA 10-2 (in which the difference was approximately 3 dB). It may be suggested that there would be substantial discrepancy in retinal sensitivity measurements between the MP-3 and HFA 10-2 in normal or subnormal eyes. In this study, very few 34-dB sensitivity points were identified by using the MP-3 (two points in N013, one point in N040, and one point in N68); hence, the narrower stimulus dynamic range (0–34 dB) of the MP-3 would be expected to have little ceiling effect on the comparability of MP-3 and HFA 10-2 results. It would not be appropriate to assume that the retinal sensitivity measures of the two devices are identical in a population that is expected to have high mean retinal sensitivity (e.g., in glaucoma screening). There are several possible reasons why the VF sensitivity measured with the MP-3 was lower than that measured with the HFA 10-2, including the prolonged measurement duration of the MP-3 and the resultant fatigue,^19,20 the difference between SITA and full-threshold strategies,²¹ and the possibility that the MP-3 test stimulates exactly the same point on the retina with narrower light, thereby causing light adaptation at the test point and reducing retinal sensitivity because of repeated target presentation.⁸ 

In severe POAG, the FP rate was lower and the FN rate was higher in MP-3 VF measurements than in the HFA 10-2 measurements. These differences may be attributed to the different testing strategies of the two devices. A previous report²² demonstrated that FP rate estimates were slightly higher with the SITA strategy than with the full-threshold strategy, and it was noted that FPs are calculated differently in SITA than they are in the full-threshold test in which classic catch trials are employed.¹⁴ Of note, our study revealed a significant difference in the reliability indices of the HFA 10-2 and MP-3 only in severe POAG. In patients with advanced POAG, alternative estimation of FP response rates in the SITA strategy may not work as intended, because in SITA the minimum response time (∼180 ms) is adjusted, based on the patient's individual mean response time, which may be improperly adjusted and result in higher FPs than actually calculated, as previously suggested.¹⁴ In instances in which the mean response time in a patient with advanced POAG caused their minimum response time to be adjusted so that it was longer than usual, potentially some true responses (although occurring at a quicker speed than the mean response time for that patient) may have been incorrectly counted as FPs. On the other hand, it has been emphasized that FPs are the least variable in test–retest studies and occur less frequently than FLs or FNs.²³ An increase in the FP rate can lead to inaccuracies in the results of the MD (e.g., introducing a 33% FP error during the examination improves the MD by more than 6 dB from the baseline).²⁴ Therefore, in VF measurements of severe glaucomatous cases in which the MD is assumed to be low, a low FP rate would be preferable. 

The differences in the FN rates between the two devices may also be attributed to the testing strategy. The FN response parameter was originally intended to reflect patient reliability in computerized perimetry.^10,25 Based on this perspective, it is possible that the prolonged duration of MP-3 measurements and the resultant fatigue may reduce patient reliability. However, previous studies have emphasized that FN rates in patients with glaucoma represent an ocular condition rather than a patient condition,^26,27 and full-threshold test procedures have been shown to yield approximately twice as many FNs as SITA procedures.²⁸ It may be reasonable to see a marked difference in FN rates across tests with the MP-3 (equipped with full-threshold procedures) and the HFA 10-2 (equipped with SITA procedures) in patients with advanced POAG, who are assumed to have a poor ocular condition. 

Of note, the duration of MP-3 VF measurements is significantly longer than that of the HFA 10-2. In our study, the duration of HFA 10-2 measurements was significantly shorter in the mild POAG group than in the moderate or severe groups but remained constant with the MP-3, regardless of the severity. The faster measurement algorithm (SITA) in the HFA 10-2 likely accounted for the difference in duration, whereas the lack of such a feature in the MP-3 may explain its consistently longer measurement duration, as previously reported.⁸ Considering the consistency in reliability indices for mild or moderate POAG, the HFA 10-2 and MP-3 devices can similarly be used to assess visual function in these disease stages, whereas the HFA 10-2 offers the advantage of faster testing times. 

Analysis of the retinal sensitivity distribution in MP-3 perimetry revealed that a lower VF sensitivity in the MP-3 corresponded to the larger discrepancy in the VF sensitivity between the HFA 10-2 and MP-3. The difference in retinal sensitivity caused by increased test–retest and intra-test variability with disease severity, as previous reports suggest,^29,30 is a possible reason. By contrast, the misalignment of the retinal sensitivity distribution tended to be more frequent in patients with milder POAG. Additionally, there was no statistical correlation between the misaligned VF sensitivity distribution and age or HFA 10-2 reliability indices such as FL, FP, or FN rates. It is possible that factors other than reliability indices or age were involved. We propose the influence of eye movements as a contributing factor to the difference in distribution. Previous reports have shown that eye movements within 2° occurred in approximately 70% of the overall fixation time in patients with glaucoma^4,14; however, it has also been reported that small eye movements (i.e., less than 3°) during VF measurement are unavoidable, even in well-trained healthy individuals.^31,32 It was reported that patients with glaucoma can have multiple defects in different areas of the visual field and make rather fewer saccades than healthy controls.³³ We hypothesized that, in patients with advanced POAG, in whom the extent or depth of visual field defects is greater, the areas of the retina capable of fixation are limited; thus, the shift in the distribution of retinal sensitivity tends to be less pronounced. FLs are considered to be elevated consistently after the third decade of life,³⁴ and our results showed significant difference in the age of patients with severe POAG (P = 0.04). However, there was no significant difference in FL rates across POAG severities (P = 0.26), which would support this hypothesis. It is difficult to define an appropriate compensatory eye movement strategy for glaucoma solely using the parameters equipped in the HFA 10-2 and MP-3. Several tracking recording methods have been proposed to measure eye movements that may affect the reliability of VF measurements.^4,35 Validations with the addition of such parameters by implementing gaze-tracking or eye- and head-tracking technologies in the MP-3 can be a topic for further study. 

Although our results showed that there was no significant correlation between shifting and axial length (P = 0.15), it should be noted that we did not exclude patients with myopia greater than –6 DS in the present criteria. Furthermore, depending on the method of refractive correction, the retinal location of peripheral test points may be different among eyes with highly different refractive statuses. Previous reports have suggested that a different pattern of VF defect exists in patients with myopia,^36–39 and the type of refractive correction used can also impact the VF results in these patients.^40,41 Factors such as back vertex distance, spectacle magnification, and spectacle minification differences can influence retinal eccentricity and subsequently affect VF test outcomes. Future work investigating the relationship between myopic glaucoma and fixation on VF tests is recommended. 

This study had several limitations. The evaluation of the anterior segment and grading of cataract for each patient was carried out by an ophthalmologist (TI or KY); however, the progression of a cataract between tests may reduce sensitivity or inflate reliability indices. Therefore, the 1-month gap (mean ± SD [range], 20.0 ± 9.9 days [5–28]) between the HFA 10-2 and MP-3 tests and the order in which the VF testing were performed may have affected the sensitivities and/or the indices. Excluding nonreliable VF cases on HFA 10-2 and MP-3 testing likely contributed to the small sample size and affected the results; for example, patients with complete loss of central visual function, in whom reliability indices are expected to show inappropriate values, were excluded from the study and therefore could not be evaluated. This factor may have inflated the likelihood of a shift occurring in milder POAG cases. Performing the HFA 24-2 or HFA 30-2 tests before the study to confirm glaucomatous VF defect and to manage these patients may have affected the results because patients are familiar with the HFA but not MP-3, and it has been suggested that reliability parameters (including FL and FP rates) and global indices (including MD and mean sensitivity) may improve with repeated testing.⁴² This factor may make comparisons less appropriate. Our novel method for simulating the alignment of test points in both tests models only for x and y position shifts (i.e., horizontal and vertical, respectively); shifts along the z position were not considered. The fact that the MP-3 projects the target light directly onto the retina, whereas the HFA projects it onto a screen, may not guarantee an exact correspondence of eccentricity on the retina, as previous studies suggest.^43,44 Finally, the purpose of this study was to examine the characteristics of MP-3 and HFA 10-2 in routine clinical practice; therefore, testing with the HFA 10-2 was performed using the SITA standard thresholding algorithm, whereas the MP-3, which is not equipped with the SITA strategy, was performed using the 4-2 full-threshold staircase strategy. Thus, to confirm that changes were due to the influence of eye movement/retinal position, as hypothesized, it would have been valuable to perform the HFA 10-2 using the 4-2 full-threshold strategy. 

In conclusion, this study revealed significant differences in reliability indices between the HFA 10-2 and the MP-3 perimeter in patients with severe POAG. A shift between the sensitivities of the test points of the two VFs was found in approximately one-half of the eyes. Despite the disadvantage of longer VF measurement durations, the MP-3 may be advantageous in the evaluation of severe POAG, as the target stimulus is projected to a precise location on the retina and the FP rate is lower than that of the HFA 10-2. Accurate reporting of FP is essential in determining a patient's true underlying visual function and whether a retest is needed. Therefore, the MP-3 providing lower FP rates in advanced POAG may assist in the management of this patient cohort. 

Acknowledgments

Supported by a grant from Research Activity Start-Up (24K23528 to TI); a Grant-in-Aid for Young Scientists from JSPS KAKENHI (21K16870 to RT); a Suda Memorial Glaucoma research grant (to RT); and the Japan Glaucoma Society Research Project Support Programme (RT). 

Disclosure: T. Inooka, None; R. Tomita, None; T. Kominami, None; M. Mochizuki, None; K.M. Nishiguchi, Bayer (F, R), Novartis (C, F, R), Santen (F, R), Senju (C, F, R), Chugai (C, R); K. Yuki, None