Predicting the Extent of Damage in the Humphrey Field Analyzer 24-2 Visual Fields Using 10-2 Test Results in Patients With Advanced Glaucoma

Abstract

Purpose: To predict Humphrey Field Analyzer 24-2 test (HFA 24-2) results using 10-2 results.

Methods: A total of 175 advanced glaucoma eyes (175 patients) with HFA 24-2 mean deviation (MD_24-2) of < −20 dB were prospectively followed up for five years using HFA 10-2 and 24-2 (twice and once in a year, respectively). Using all the HFA 24-2 and 10-2 test result pairs measured within three months (350 pairs from 85 eyes, training dataset), a formula to predict HFA 24-2 result using HFA 10-2 results was constructed using least absolute shrinkage and selection operator regression (LASSO). Using 90 different eyes (testing dataset), the absolute differences between the actual and LASSO-predicted MD_24-2 and that between the slopes calculated using five actual and LASSO-predicted MD_24-2 values, were adopted as the prediction error. Similar analyses were performed for the mean total deviation values (mTD) of the superior (or inferior) hemifield [hemi-mTD_sup.24-2(-hemi-mTD_inf.24-2)].

Results: The prediction error for the LASSO-predicted MD_24-2 and its slope were 2.98 (standard deviation [SD] = 1.90) dB and 0.32 (0.33) dB/yr, respectively. The LASSO-predicted hemi-mTD_sup.24-2 (hemi-mTD_inf.24-2), and its slope were 3.02 (2.89) and 3.76 (2.72) dB, and 0.37 (0.41) and 0.44 (0.38) dB/year, respectively. These prediction errors were within two times SD of repeatability of the simulated stable HFA 24-2 VF parameter series.

Conclusions: HFA 24-2 results could be predicted using the paired HFA 10-2 results with reasonable accuracy using LASSO in patients with advanced glaucoma.

Translational Relevance: It is useful to predict HFA24-2 test from HFA10-2 test, when the former is not available, in advanced glaucoma.

Introduction

Glaucoma is the leading cause of blindness worldwide.¹ Glaucomatous visual field (VF) changes usually manifest in the mid-peripheral VF; visual function is retained in the central region until the late stages of the disease. In severely advanced glaucoma, VF damage is often characterized by large arcuate scotomata in the upper and lower hemifields, which form a ring that threaten the central visual function.^2,3 The Humphrey Field Analyzer Central 10-2 test (HFA 10-2 test; Carl Zeiss Meditec, Dublin, CA, USA) is used for the detailed measurement of the central 10° when monitoring such patients.^4,5 Nonetheless, to estimate the glaucoma-induced overall VF damage in such patients, visual sensitivity remaining outside the central 10° must also be measured using a program, such as the HFA 24-2 test. One of the challenges faced in the clinical setting is that performing a single VF test with sufficient frequency is a significant burden^6–8; undergoing an HFA 24-2 test in addition to the HFA 10-2 test adds to this burden. Consequently, the number of HFA 10-2 tests, which are important to monitor the central visual function of patients with advanced glaucoma, might be further limited by the need to perform additional HFA 24-2 tests on certain clinic visits. Indeed, in our prior study, non-negligible proportion (17.3%) of eyes showed further progression of the VF deterioration with the HFA 24-2 test, compared to 26.9% with the HFA 10-2 test.⁹ 

In advanced glaucoma, it is possible to predict HFA 10-2 test results using HFA 24-2 test data¹⁰ and various machine-learning methods.^11–15 Among these, the least absolute shrinkage and selection operator (LASSO) regression has the best prediction performance. An additional advantage of this method over others is that it allows us to know the final equation and visualize the mechanism of prediction (i.e., coefficients and standard error of each parameter in the final regression equation), unlike support vector machine and deep learning.¹⁶ If the HFA 24-2 test results can be predicted using the HFA 10-2 test results of the same eye via LASSO regression with reasonable accuracy, it would allow physicians to assess the status of the HFA 24-2 VF without actually having to performing it. Thus we investigated the validity and usefulness of LASSO-predicted HFA 24-2 test results using the HFA 10-2 test results of the same eye. 

Methods

This study was approved by the ethical review committees of each participating institute (No. 544-2:) and registered as No: UMIN000001004. The study was conducted in accordance to the tenets of the Declaration of Helsinki. Written informed consent was obtained from all participants. Patients with severely advanced glaucomatous VF damage were consecutively recruited from among those visiting the outpatient clinics of seven institutes (the Advanced Glaucoma Study).⁹ The inclusion criteria were as follows: (1) glaucoma was the only cause of VF damage or impairment of visual acuity; (2) experience with VF examination using an HFA with the Swedish interactive threshold algorithm standard (SITA-S) central 24-2 program (HFA 24-2) and ≥ 2 of such reproducible VF test results per eye before enrollment; (3) a reproducible mean deviation (MD) ≤ −20 dB in either eye using this program; (4) best-corrected visual acuity of ≥ 20/40 (0.5) in both eyes; (5) no clinically significant cataract and inadequate intraocular pressure (IOP) control that would warrant reliable VF test results and no need for aggressive therapy during the subsequent five years of follow-up. 

VF examinations using the SITA-S HFA 24-2 and HFA 10-2 programs in the Advanced Glaucoma Study⁹ were used in the analysis in the current study. All VFs were required to be reliable VF measurements, defined as a fixation loss rate ≤ 20% and false-positive rate ≤ 15%. HFA 10-2 testing was performed every six months, and HFA 24-2 testing was conducted once a year. 

The current study initially included 175 eyes of 175 patients; of these patients, 149 patients did not undergo surgical treatment over the study course.⁹ Among the 149 patients without surgery, 90 were randomly chosen to be included in the testing dataset, and the remaining 85 eyes were used as the training dataset. The training dataset was used to construct a model to predict the HFA 24-2 test results from those of the HFA 10-2. The testing dataset was used to evaluate the validity of the predicted HFA 24-2 test results. 

Prediction of HFA 24-2 Test Results Using the HFA 10-2 Test Results and a LASSO Regression Model

It is widely acknowledged that ordinal statistical models, such as ordinary least squares linear regression, may be overfitted to the original sample, especially when the number of predictor variables is large, such as in the current training dataset. Tibshirani et al.¹⁶ previously applied a shrinkage method to regularize the sum of the absolute values of the regression coefficients to mitigate these issues in linear/logistic modeling.¹⁷ This LASSO regression method has been used in studies of human perception, genetic analyses,^18,19 and glaucoma.^20,21 The penalized log-likelihood function to be maximized was estimated using the following formula^16,17:  

\begin{eqnarray*} \mathop \sum \limits_{i = 1}^n \big [(yixi\beta - \log \left( {1 + e_{}^{xi\beta }} \right) \big ] - \ \lambda \mathop \sum \limits_{j = 1}^p \beta _j^{} \end{eqnarray*}

where xi is the i-th row of a matrix of n observations with p predictors, β is the column vector of the regression coefficients, and λ represents the penalty applied. 

Using the training dataset, all the HFA 24-2 and 10-2 test result pairs measured within three months were identified (350 pairs from 85 eyes). Using the 68 total deviation (TD) and MD values of the paired HFA 10-2 test, the MD (MD_24-2) of the paired HFA 24-2 test were predicted using LASSO regression. Similarly, mean TD (mTD_24-2) value of the superior hemifield (hemi-mTD_sup.24-2) and the inferior hemifield (hemi-mTD_inf.24-2) were predicted using the 68 TD values and MD value of the paired HFA 10-2 test with the LASSO regression. Additionally, the formulas to predict the MD_24-2 and hemi-mTD_sup.24-2(_inf.24-2) from paired HFA 10-2 results were also obtained. 

Analyses of the Validity of the LASSO-Predicted HFA 24-2 Results Using the Testing Dataset

Comparison Between the LASSO-Predicted MD_24-2 and Hemi-mTD_sup.24-2(_inf.24-2) Values and the Actual HFA 24-2 VF Parameter Values

The prediction error for the LASSO-predicted MD_24-2 was calculated as the absolute difference between the MD_24-2 of the first actual HFA 24-2 VF and the LASSO-predicted MD_24-2 derived from the first HFA 10-2 result, as below:  

\begin{eqnarray*} && {\rm{Absolute\ difference}}\\ && = |{\rm actually\ measured\ MD _{24 - 2}} - {\rm LASSO} - {\rm MD_{24 - 2}}| \end{eqnarray*}

 

The prediction error for the LASSO-predicted hemi-mTD_sup.24-2(_inf.24-2) was also calculated similarly. 

Comparison of the MD_24-2 and mTD_sup.24-2(_inf.24-2) Slopes Obtained Using Five LASSO-Predicted Data to Those Obtained Using Five Actual HFA 24-2 Results

Using five actual and LASSO-predicted MD_24-2 values obtained over five years in the testing dataset, the actual and LASSO-predicted MD_24-2 slopes were calculated, respectively, using OLSR. The slope-prediction error was calculated as the absolute difference between the actual MD_24-2 slope and the LASSO-predicted MD_24-2 slope in the same eye. Similar calculations were done for the actual and LASSO-predicted hemi-mTD _sup.24-2(_inf.24-2) slopes. 

Clinical Validity of the LASSO-Predicted HFA 24-2 VF Parameters

A stable HFA 10-2 result series has previously been simulated to confirm the specificity of different criteria for progression of the HFA 10-2 test results.⁹ Among the 175 patients, we identified 51 stable eyes defined as no significant (P > 0.05) change in the MD and hemi-mTD during the follow-up period in both of the actual HFA 24-2 and 10 HFA 10-2 test results. Using the first and second actual HFA 24-2 test results of these 51 stable eyes, 10,200 simulated stable series of five HFA 24-2 test results were generated (200 simulated series of five VF values for each of the 51 eyes). Ninety series were randomly selected from the 10,200 simulated series, and the differences in the first and second MD_24-2 and hemi-mTD_sup.24-2(_inf.24-2) were calculated. These were subsequently compared to the prediction errors for the LASSO-predicted HFA 24-2 VF parameters. Thereafter, using the randomly chosen 90 pairs of the simulated stable series were randomly selected, the MD_24-2 and hemi-mTD_sup.24-2(_inf.24-2) slopes (dB/year) were calculated. These were subsequently compared to the prediction error for the LASSO-predicted slopes of MD_24-2 and hemi-mTD_sup.24-2(_inf.24-2). 

In addition, progression of the actual MD_24-2 and hemi-mTD_sup.24-2(_inf.24-2) was defined as a slope < −0.5 dB/year with P < 0.05. Rapid progression, which needs a quick response, including surgery for advanced glaucoma, was defined as a slope < −1.0 dB/year with P < 0.05. The usefulness of the LASSO-predicted MD_24-2 or LASSO-hemi-mTD _sup.24-2(_inf.24-2) slope in detecting progression of the disease was examined using the area under the receiver operating characteristic curve (AUC). 

All statistical analyses were performed using R (version 3.6.1; The R Foundation for Statistical Computing, Vienna, Austria). The R package “glmnet” was used to perform the LASSO regression. 

Results

From our prior study in advanced glaucoma,⁹ HFA 24-2 and HFA 10-2 test results of 85 and 90 patients were randomly assigned to the training and testing dataset, respectively. The average MD_24-2 was −25.8 ± 2.8 (standard deviation) and −25.2 ± 3.5 dB and the average age was 65.5 ± 12.5 and 62.7 ± 12.5 years in the training and testing dataset, respectively. The patient demographics, including the VF characteristics of both datasets, are summarized in Table 1. The formulas to predict the MD_24-2 (Supplement 1) and hemi-mTD_{sup.24-2(inf.24-2)} (Supplements 2 and 3) using the paired HFA 10-2 test results and LASSO regression were linear equations containing TD values of 31 and 32 (35) test points of the paired HFA 10-2 results, respectively. 

Table 1.

 

Characteristics of the Study Participants in the Training and Testing Datasets

 

Prediction Errors of the LASSO-Predicted MD_24-2 and Hemi-mTD_sup.24-2(_inf.24-2) and Their Respective Slopes Calculated Using HFA 10-2 Results

The average prediction errors for the LASSO-predicted MD_24-2 and hemi-mTD_sup.24-2(_inf.24-2) were 2.98 ± 1.90 (SD) dB (N = 90) and 3.02 ± 2.89 (3.76 ± 2.72) dB, respectively (Table 2). The average prediction error for the LASSO-predicted MD_24-2 and hemi-mTD _sup.24-2(_inf.24-2) slopes were 0.32 ± 0.33 dB/year (N = 90) and 0.37 ± 0.41 (0.44 ± 0.38) dB/year, respectively (Table 3). 

Table 2.

 

LASSO-Predicted HFA 24-2 VF Parameters Using the HFA 10-2 Test Results and Their Prediction Errors

 

Table 3.

 

Slopes of the LASSO-Predicted HFA 24-2 VF Parameters Using the HFA 10-2 Test Results and Their Prediction Errors

 

Clinical Validity of the LASSO-Predicted HFA 24-2 VF Parameters

The difference between the first and second MD_24-2 hemi-mTD _sup.24-2(_inf.24-2) of the 90 randomly selected stable HFA 24-2 series was −0.011 ± 1.58 dB and −0.0021 ± 2.06 (−0.020 ± 1.72) dB, respectively (Table 2). The 2 × SD of the difference between the first and second MD_24-2 or hemi-mTD _sup.24-2(_inf.24-2) of the stable HFA 24-2 series, which should be primarily very close to zero, indicates the statistical fluctuation of the difference. The average prediction errors of the above LASSO-predicted MD_24-2 or hemi-mTD _sup.24-2(_inf.24-2) were within or almost within the range of 2 × SD (Table 2). The average MD_24-2 and hemi-mTD _sup.24-2(_inf.24-2) slope calculated using the simulated stable HFA 24-2 series was −0.0040 ± 0.25 dB/year and −0.0061 ± 0.32 (−0.0020 ± 0.27) dB/yr, respectively (Table 3). The 2 × SD indicates the statistical fluctuation range of the MD_24-2 or hemi-mTD _sup.24-2(_inf.24-2) slope calculated using the stable HFA 24-2 VF series, which should be primarily very close to zero. The prediction error for the LASSO-predicted MD_24-2 or hemi-mTD _sup.24-2(_inf.24-2) was also within the range of 2 × SD. (Table 3). 

The AUCs for detecting the actual MD_24-2 and hemi-mTD_sup.24-2(_inf.24-2) slopes < −0.5 dB/year (P < 0.05) using the LASSO-predicted MD_24-2 and hemi-mTD_sup.24-2(_inf.24-2) slopes were 0.90 (95% confidence interval [CI], 0.82–0.95), 0.59 (95% CI, 0.48–0.70), and 0.89 (95% CI, 0.81–0.95), respectively. The AUC for detecting the progression of the actual MD_24-2, hemi-mTD_sup.24-2 or hemi-mTD_inf.24-2 slope < −1.0 dB/year (P < 0.05) using LASSO-predicted MD_24-2 slope was 0.93 (95% CI, 0.86–0.97), with an optimum sensitivity and specificity of 100.0% and 84.0%, respectively (Figure). 

Figure.

 

The AUC for detecting a progression of <−1.0 dB/year (P < 0.05) in the actual MD_24-2, hemi-mTD_sup.24-2, or hemi-mTD_inf.24-2 slope using the LASSO-predicted MD_24-2 slope. The AUC was 0.93 (95% CI = 0.86 to 0.97) with an optimum sensitivity and specificity of 100.0% and 84.0%, respectively.

Discussion

In this study, 175 eyes with advanced glaucoma, a baseline mean MD_24-2 = −25 dB, and an MD_10-2 = −23 dB were prospectively followed up for five years using the HFA 10-2 (twice a year) and HFA 24-2 (once a year) tests. A training dataset consisting of 350 pairs of HFA 10-2 and 24-2 results from 85 eyes, obtained within three months, was used to construct an equation to predict HFA 24-2 results using HFA 10-2 results and LASSO regression. We estimated the clinical validity and usefulness of the LASSO-predicted MD_24-2 and hemi-mTD _sup.24-2(_inf.24-2) calculated from the actual HFA 10-2 results. We found that the LASSO-predicted MD_24-2 and hemi-mTD _sup.24-2(_inf.24-2) were not significantly different from the actual HFA 24-2 VF values. Furthermore, the prediction errors of the LASSO-predicted MD_24-2 and hemi-mTD _sup.24-2(_inf.24-2) averaged approximately 10% to 15% of the absolute values of the actual MD_24-2 or hemi-mTD_sup.24-2 (_inf.24-2) values. In eyes with advanced VF damage, fluctuation in the measured VF sensitivity outside the central 10° is usually large.²² Considering the average MD_24-2 and hemi-mTD_sup.24-2 (_inf.24-2) values of −25.2 and −26.4 (−23.9) dB, respectively, in the testing dataset eyes, the prediction errors of the LASSO-predicted MD_24-2 and hemi-mTD_sup.24-2 (_inf.24-2) of 2.98 to 3.76 dB were within the clinically acceptable error range. These values were also within the 2 x SD range of fluctuation of the simulated stable MD_24-2 or hemi-mTD_sup.24-2 (_inf.24-2) series values (Table 2). In other words, the prediction errors were within the range of statistical fluctuations of the stable HFA 24-2 VF parameters in eyes with an MD_24-2 of −25.0 dB. For the comprehensive assessment of VF damage outside the central VF of 10° in eyes with advanced glaucoma, assessment of the whole VF or hemifields would be more clinically relevant than that of VF sensitivities at each test point. In such eyes, performing HFA 10-2 is unavoidable because it is assessment is important to monitor the remaining functional central vision.^4,5 However, it may be difficult to measure both HFA 10-2 and 24-2 VFs a sufficient number of times in general clinical practice.^6–8 Supporting this, in our prior study, a non-negligible proportion (17.3%) of eyes showed further progression of the VF deterioration with the HFA 24-2 test, compared to 26.9% with the HFA 10-2 test.⁹ Moreover, significant risk factors for the progression of HFA 24-2 test were identified in our prior study.⁹ The currently proposed method to predict the HFA 24-2 VF status in such cases may be clinically useful. This would help avoid HFA 24-2 VF measurements to allow HFA 10-2 tests, which are more important, to be performed. 

In eyes with advanced glaucoma with several low-sensitivity test points and limited room for progression, a progression analysis relying on pointwise progression, such as the Humphrey's Glaucoma Progression Analysis (Carl Zeiss Meditec), may be of limited value.²² Glaucoma causes diffuse VF loss in addition to localized deterioration²³; the former becomes predominant when the number of test points reaches very low sensitivity. This implies that the progression analysis of MD or hemi-mTD in the superior or inferior HFA 24-2 hemifields would be of more clinical concern, assuming that the HFA 10-2 tests were closely monitored. The LASSO-predicted MD_24-2 and hemi-mTD_{sup.24-2(inf.24-2)} slopes over the five-year follow-up period yielded a prediction error range of 0.32 to 0.44 dB/yr. This was favorably compared to the 2 × SD statistical fluctuation of the MD_24-2 and hemi-mTD_{sup.24-2(inf.24-2)} slope values (0.50–0.64 dB/yr) of the simulated stable series. However, the LASSO-predicted slope values themselves were significantly more positive than the actual slopes. This may be at least partly because the actual HFA 10-2 results had relatively more saved test points than the actual HFA 24-2 results in the eyes severely damaged by glaucoma. Because the differences between the LASSO-predicted and actual MD_24-2 and hemi-mTD_{sup.24-2(inf.24-2)} slopes were all approximately −0.18 dB/yr, we might use −0.18 dB/yr as a correction factor when predicting MD_24-2 and hemi-mTD_{sup.24-2(inf.24-2)} slopes. Furthermore, the actual MD_24-2, or hemi-mTD_{sup.24-2(inf.24-2)} slope value < −1.0 dB/yr (P < 0.05), which is a potentially dangerous and rapid progression of the HFA 24-2 results, could be detected with an AUC of 0.93 and an optimum sensitivity and specificity of 100.0% and 84.0%, respectively. This suggests the clinical relevance of the LASSO-predicted MD_24-2 slope. 

This study has limitations. First, the local spatial information (i.e., each HFA 24-2 TD progression rate) could not be predicted. A model to predict the 52 HFA 24-2 TD values using the corresponding HFA 10-2 results yielded an average prediction error of approximately double or triple of that of the MD_24-2 or hemi-mTD_{sup.24-2(inf.24-2)} (data not shown). Second, the current results were obtained in a group of patients with advanced glaucoma whose MD_24-2 progression rate was relatively slow, averaging −0.22 dB/yr. The current results may not be applicable in patients with a faster progression of MD_24-2 and hemi-mTD_{sup.24-2(inf.24-2)}.^24–26 However, in a study of 417 Japanese patients with glaucoma at various stages of VF damage, whose IOP was controlled at around 13.5 mm Hg, we found a mean MD_24-2 slope of −0.26 dB/yr.²⁷ This suggests that patients, such as those in the current study, might not be a small number in the real world data. Third, it has been suggested that including spectral-domain optical coherence tomography (SD-OCT) parameters could further improve the prediction accuracy of VF status.^27–29 However, SD-OCT was not widely used when the study started. Therefore further studies are required to determine whether the inclusion of SD-OCT parameters may improve the prediction accuracy of HFA 24-2 VF status using HFA 10-2 results in patients with advanced glaucoma. Last, there may be a “floor effect” in VF in eyes with advanced glaucoma. However, MD is an average value of multiple test points, which means there may still be test points much higher than 20 dB even in VF with MD = −20 dB,⁹ although a considerable number (17.3%) of eyes progressed despite the even worse MD value (−25.85 dB) in our prior study. Thus this aspect would have only marginal effect on the current results. 

In conclusion, our method enabled the prediction of MD_24-2 and mTD_24-2 in each hemifield of HFA 24-2 using 68 total deviation (TD) and MD values of the HFA 10-2 results and the LASSO regression with reasonable accuracy in eyes with severe VF damage. Similarly, the progression rates of MD_24-2 and mTD_24-2 in each hemifield of HFA 24-2 with 5 VF tests were predicted with such an accuracy using the five predicted MD_24-2 and mTD_24-2 values. The method proposed in this study may be used to approximately determine the HFA 24-2 VF status using HFA 10-2 results, thus saving time by avoiding HFA 24-2 VF testing and filling in the gaps where HFA 24-2 test measurements could not be obtained. This may be useful in routine clinical practice in managing patients with severe VF damage where closer monitoring of central VF using HFA10-2 is needed. 

Acknowledgments

Supported by the Japan Glaucoma Society, Tokyo, Japan. The funding organization had no role in the design or conduct of this research. 

Author Contributors: The authors were involved in the design and conduct of the study (AR, SK, IT, YK, KA, YY, IS, UK, IA, and MA); collection, management, analysis and interpretation of data (AR, MA); and preparation, review and approval of the manuscript (AR, SK, IT, YK, KA, YY, IS, UK, IA, and MA). AR is the guarantor. 

Collaborators: Department of Ophthalmology, University of Tokyo Graduate School of Medicine (Tokyo, Japan): Makoto Araie, MD, PhD (Current affiliation: Kanagawa Dental University, Yokohama Clinic), Atsuo Tomidokoro, MD, PhD (Current affiliation: Higashinakano Tomidokoro Eye Clinic, Tokyo, Japan), Kenji Sugisaki, MD (Current affiliation: Department of Ophthalmology, International University of Health, and Welfare, Mita Hospital, Tokyo, Japan), Asaoka Ryo, MD, PhD (Current affiliation: Department of Ophthalmology, Seirei Hamamatsu General Hospital, Shizuoka Japan), Hiroshi Murata, MD. Department of Ophthalmology, Faculty of Life Sciences, Kumamoto University (Kumamoto, Japan): Hidenobu Tanihara, MD, PhD, Masaru Inatani, MD, PhD (Current affiliation: Department of Ophthalmology, Faculty of Medical Sciences, University of Fukui, Fukui, Japan), Toshihiro Inoue, MD, PhD. Yoshikawa Eye Clinic (Tokyo, Japan): Keiji Yoshikawa, MD. Division of Ophthalmology, Department of Surgery, Kobe University Graduate School of Medicine (Kobe, Japan): Akira Negi, MD, PhD, Hidetaka Maeda, MD, PhD (Current affiliation; Maeda Eye Clinic, Osaka, Japan), Akiuasu Kanamori, MD, PhD, Yuko Nakanish, MD, PhD. Department of Ophthalmology, Nihon University School of Medicine (Tokyo, Japan): Yoshio Yamazaki, MD, PhD (Current affiliation: Yamazaki Eye Clinic, Tokyo, Japan.), Kenji Mizuki, MD, PhD (Current affiliation: Hakata Mizuki Eye Clinic, Fukuoka, Japan). Department of Ophthalmology, Saga University Faculty of Medicine (Saga, Japan): Satoshi Okinami, MD, PhD, Ryo Iwakiri, MD, PhD (Current affiliation: Department of Ophthalmology, National Hospital Organization Ureshino Medical Center, Saga, Japan), Shinichirou Ishikawa, MD, PhD. Department of Ophthalmology, Tokyo Post and Telecommunication Hospital (Tokyo, Japan): Shun Matsumoto, MD, PhD, Kenichi Uchida, MD (Current affiliation: Tokyo Kyosai Hospital, Tokyo, Japan), Koichi Mishima, MD, PhD (Current affiliation: Department of Ophthalmology, Kanto Central Hospital of the Mutual Aid Association of Public School Teachers, Tokyo, Japan), Hodaka Nemoto, MD (Current affiliation: Department of Ophthalmology, University of Tokyo Graduate School of Medicine, Tokyo, Japan). Tajimi Iwase Eye Clinic (Gifu, Japan): Aiko Iwase, MD, PhD. 

Disclosure: R. Asaoka, Oculus (I), Reichert (I), Nidek (I), Kowa (I); K. Sugisaki, Japan Glaucoma Society (R), Senju Pharmaceutical (I), Kowa (I), Alcon Japan (I); T. Inoue, Kowa (R, I), Novartis Pharma (R, I), Senju Pharmaceutical (R, I), Pfizer Japan (R, I), Otsuka Pharmaceutical (R, I), Tomey (I), Santen Pharmaceutical (I), Japan Focus Company (I), Beyer (I), Glaukos (I), Wakamoto Pharmaceutical (I), Mitsubishi Tanabe Pharma (I), Allergan (I), Alcon Japan (R); K. Yoshikawa, Santen Pharmaceutical (I), Senju Pharmaceutical (I), Otsuka Pharmaceutical (I), CREWT Medical Systems (I), JFC Sales Plan (I), R E Medical (I); A. Kanamori, Santen Pharmaceutical Co. (I), Otsuka Pharmaceutical Co. (I), Senjyu Pharmaceutical Co. (I), Kowa Pharmaceutical Co. (I), Nitten Pharmaceutical Co. (I); Y. Yamazaki, Kowa (I), Santen Pharmaceutical (I), Senju Pharmaceutical (I), Novaritis Pharma (I); S. Ishikawa, Santen (I), Senjyu (I), Kowa (I), Otuka (I), Alcon (I), Seed (I), Nitten (I); K. Uchida, None; A. Iwase, Carl Zeiss Meditec (I), Kowa (I), Otsuka (I), Pfizer (I), Santen (I), Senju (I), Novartis (I), JGSTK-DiscAnalysis (P); M. Araie, Pfizer (I), Santen Pharmacy (I), Topcon Medical System (I), Otsuka (I), Senju (I), Aerie (I), Kowa (I), JGSTK-DiscAnalysis (S)