Risk Factors and Frequency of Examinations for Detecting Visual Field Deterioration in Patients with Newly Diagnosed Exfoliation Glaucoma in Sweden

Abstract

Précis:

The study identified risk factors for exfoliation glaucoma and recommended re-evaluating target intraocular pressure (IOP) after 5 visual fields to slow disease progression.

Purpose:

This study aimed to establish risk factors for exfoliation glaucoma and determine the earliest time points for estimating disease progression.

Patients:

A total of 96 patients with newly diagnosed exfoliation glaucoma were included. Included patients were required to perform at least 7 visual field tests within a 3-year period (±3 months). All patients were treated at inclusion.

Methods:

This was a nonrandomized, prospective cohort study. The predictors measured included IOP, mean deviation (MD), and visual field index (VFI). Progression was assessed using the rate of progression based on MD, VFI, and “Guided Progression Analysis.” Linear or logistic regression models were developed based on the variables studied. An analysis of variance was used to establish the earliest time point. At the earliest time point, the models were retested. The area under the receiver operating characteristic curve was calculated.

Results:

The general rate of progression of the cohort was −3.84 (±2.61) dB for the MD values and 9.66 (±6.25) % for the VFI values over 3 years. The IOP, MD, and VFI values at diagnosis were predictors of progression for both linear and logistic regression. Analysis of variance and post hoc Tukey test showed significant values at 24 months for MD and VFI. The area under the curve at 24 months showed significant values for MD and VFI.

Conclusions:

The predictors studied (IOP, MD, and VFI) showed moderate accuracy at baseline but excellent predictive capacity at 24 months postdiagnosis. Re-evaluating the target IOP at 24 months can effectively slow down disease progression.

Glaucoma is a group of optic nerve diseases that primarily alters a patient’s visual field. Several types of glaucoma have been described, with primary open angle glaucoma and exfoliation glaucoma (EXFG) being the most common in Sweden.¹ EXFG is caused by increased production of protein material that accumulates in the trabecular meshwork, increasing the intraocular pressure (IOP). Although the causes and mechanisms of exfoliation have not yet been elucidated, several genes have been linked with the disease.^2,3

Previous studies have described EXFG as a rapidly progressive form of glaucoma,^4–6 and visual field testing is the gold standard for assessing glaucoma progression, as recommended by the European Glaucoma Society.⁷ According to the Swedish guidelines for glaucoma care, at least 7 (1 baseline and 6 follow-up) visual fields are needed over a 3-year period to detect progression.⁸ As EXFG seems to be a much faster progressing glaucoma, detecting progression as early as possible is crucial to stop the disease and preserve the visual field. A target IOP is usually established at diagnosis, with revisions made after 3 years based on visual field progression according to the Swedish guidelines for glaucoma.⁸

The first step in glaucoma treatment is establishing a diagnosis, followed by monitoring disease progression in each patient.⁹ Models have been used to help with this, with several glaucoma models described in previous studies.^6,10 However, most of these models included glaucoma patients with both old and new diagnoses, and evidence for models based solely on newly diagnosed glaucoma patients is limited.

The present study aimed to study risk factors for visual field deterioration in patients with newly diagnosed EXFG and to determine how early these factors could accurately predict progression.

MATERIALS AND METHODS

This study was a nonrandomized prospective cohort study that included all patients newly diagnosed with EXFG. Patients were recruited from January 1, 2012, to December 31, 2017 (6 years). All patients attending the Ophthalmology Department at the Skaraborg Hospital with a newly diagnosed EXFG were asked to participate in the study. Signed informed consent was obtained from all participants. The Ethical Committee granted the study ethical approval (DN:119-12).

Inclusion Criteria

1. Patients newly diagnosed with EXFG. The definition used for EXFG was according to the European Glaucoma Society guidelines.⁷

2. Age at diagnosis: 85 years or below.

Exclusion Criteria

1. Patients unable to perform reliable visual field tests, defined as visual fields with ≤15% false positives and/or ≤20% fixation losses.

2. Patients with advanced visual field damage, defined as mean deviation (MD) ≤−18 dB and/or visual field index (VFI) ≤40%. These patients were excluded to avoid “floor effects.”^11,12

3. Patients suffering from other significant eye diseases during the 3-year follow-up period, such as central retinal venous occlusion and branch retinal venous occlusion.

4. Patients who could not be followed up for 3 years due to dementia, general illness, or moving to another part of the country.

5. Patients who underwent glaucoma surgery, uncomplicated cataract surgery, or selective laser trabeculoplasty (SLT) were not excluded from the study.

At the recruitment visit, all patients underwent ophthalmological examinations. The age and sex of the participants were also recorded. Patients with unilateral or bilateral glaucoma were observed, and in cases of bilateral glaucoma, 1 eye was randomly selected for the study.

Visual acuity was evaluated using a Snellen chart, and the IOP was recorded using a Goldmann applanation tonometer. The average value was derived from 3 separate IOP measurements. Central corneal thickness was measured using an ultrasound device (Tomey Pachymetry; Tomey Corp., Nagoya 451–0051). Gonioscopy was performed to inspect the trabecular meshwork by using a goniolens with undilated pupils. On the basis of Shaffer classification, gonioscopy was categorized (0–4), and the pigmentation of the trabecular meshwork was measured from 0 to 3.

The patient was then examined by an ophthalmic nurse using a Humphrey Field Analyzer (Carl-Zeiss, Straße 22, 73447) to perform a visual field test with the 24-2 technique of the Swedish Interactive Threshold Algorithm (SITA fast).

The patient’s pupils were subsequently dilated, and exfoliation was observed in the anterior region of the eye using a slit lamp. A 90-D lens was used to examine the optic nerve, and the cup-to-disc (C/D) ratio was recorded. After determining a target IOP, the patient was administered IOP-lowering eye drops. The target IOP was fixed at 20 mm Hg; if the patient had an IOP < 20 mm Hg, the target IOP was determined as a 20% IOP reduction. To evaluate the effect of the medication, the IOP was measured 1 month after the recruitment appointment. Over the following 3 years (±3 months), patients underwent examinations every 6 months (±1 month), and visual acuity was assessed during follow-up visits. In addition, IOP values were recorded before the visual field test. The patients completed at least 7 visual fields in total, including 1 at the beginning (baseline) and 6 over the course of the 3-year follow-up. All patients were followed up according to the Swedish Guidelines for glaucoma care.⁸

If the IOP needed to be lowered, new drugs were administered and/or SLT was performed. For analysis purposes, the SLT was scored as Yes/No. The presence or absence of cataract surgery was recorded as Yes/No. The amount of drugs taken at the end of the 3-year period was measured using the amount of medicines (compounds) taken and not the number of bottles.

End Points

The primary endpoint of the study was to analyze the visual field progression in patients with newly diagnosed EXFG. Deterioration in the visual field was measured in 3 different ways: 2 as continuous variables and the third as a binomial (Yes/No) variable.

The first approach was based on MD values, which is the oldest method for measuring visual field progression but is still in use.¹³ However, cataract development has been shown to modify MD values, making it an imprecise parameter for establishing glaucoma progression. The rate of progression (ROP) was measured in decibels per year (dB/y).

The second method was the VFI values. The device calculated the VFI values as percentages (%). For example, a normal visual field shows 100% VFI. The ROP was measured as a percentage per year.

The third technique was guided progression analysis (GPA). The device included a GPA that was automatically executed. The GPA is an “event analysis,” differentiating it from the MD-ROP and the VFI-ROP, which are regression analysis. The machine compares every single point to similar points found in earlier assessments. The GPA classified progression as no, possible, or likely. For analytical purposes, results were divided into progression and no progression (binary variables). Therefore, progression included both “possible” and “likely” progression.

Statistics

Statistical analyses were performed using SPSS software (IBM). First, the baseline clinical characteristics of the cohort were tested using a t test or χ² test to determine the differences according to GPA analysis. Correlations among IOP/MD, IOP/VFI, and MD/VFI were tested using Pearson coefficients. Further, the same correlations were retested at the time point indicated by the analysis of variance (ANOVA)/Tukey test comparisons as the earliest time for differences. Again, at 36 months (end of the study), correlations between MD-ROP and VFI-ROP were tested using the Pearson coefficients, and correlations between MD-ROP/GPA and VFI-ROP/GPA were tested using “point biserial correlation.” To detect “multicollinearity” among the predictors at inclusion, the variance inflation factor (VIF) was calculated.

The relationship between the IOP at diagnosis and MD-ROP or VFI-ROP at 36 months was tested using linear regression analysis. In addition, linear regression analysis was performed to test the association between MD at diagnosis and MD-ROP. A similar approach was used to diagnose the VFI and VFI-ROP. Linear regression was used to adjust for the significant covariates. Logistic regression analysis was used to test the relationship between the IOP at diagnosis and GPA. The logistic regression was adjusted for significant covariates. On the basis of the GPA analysis, a receiver operating characteristic area under the curve (AUC) was calculated. MD and VFI values at different time points were first tested for homoscedasticity using the Levene test. If homoscedasticity was present, the groups were tested using one-way ANOVA. In the case of significant results, a post hoc Tukey test was performed to test the significance between the groups and establish the time point for the earliest significant values. At the earliest point, linear regression for the association between IOP, MD, and VFI values with MD and VFI-ROP was performed as described above. In addition, a new logistic regression analysis of the association between IOP, MD, and VFI with GPA was performed. On the basis of the GPA, the AUC was calculated to determine the accuracy, sensitivity, and specificity of the model. The significance level was set at P < 0.05.

RESULTS

A total of 96 patients were included in the study. Sixteen patients were excluded from this study. Reasons for exclusion included glaucoma surgery in 2 cases, low adherence to check-up visits in 4 cases, and low-quality visual fields in the remaining cases.

The average age of patients at inclusion was 70.33 (±6.04) years, with a sex distribution of 49 females and 47 males (51/49%). The mean visual acuity was 0.8 (±0.23) in Snellen units, while the mean IOP at inclusion was 32.52 (±5.54) mm Hg. EXFG was unilateral in 66 patients and bilateral in 30 patients (69/31%). The mean central corneal thickness was 546.06 (±34.08) µm. At gonioscopy, the average values for anterior chamber depth and pigmentation were 3.16 (±0.68) and 2.5 (±0.52), respectively. The average cup/disc ratio was 0.79 (±0.09). Only 15 out of 96 patients (15%) underwent cataract surgery (pseudophakia) at the time of inclusion.

All patients were followed for 3 years (±3 months). The overall ROP of the whole cohort was −3.84 (±2.61) dB for the MD values over 3 years, which is approximately −1.28 dB/y or a visual field deterioration of 4.2%/y. With regard to the VFI values, the ROP was 9.66 (±6.25) % over a 3-year period, which is ~3.22%/y. In relative terms, this indicates a relative visual field deterioration of 3.2%/y (as VFI is already a percentage value). At the end of the study period, patients were evaluated using the GPA. The cohort was divided into 2 groups according to the results of the GPA analysis: progression and no progression. On the basis of the GPA, 37 patients did not show any progression in their visual fields, while disease progression was observed in 59 patients. At inclusion, there was a significant difference in the IOP values between patients who progressed and those who did not. Patients who progressed had an average IOP of 3 mm Hg higher than those who did not progress (t test; P=0.005). Age was also significantly different between patients who progressed and those who did not. Progressing patients were, on average, 3 years older than nonprogressing patients (t test; P=0.03). The MD and VFI values at inclusion were also different between the groups, with MD values being double among progressors compared with nonprogressors (t test; P<0.001). On the other hand, the VFI values at inclusion were ~10% lower among progressors compared with nonprogressors (t test; P <0.001). (Table 1).

TABLE 1.

Baseline Characteristics at the Inclusion of the Patients Based on the GPA Analysis

	No progress (N=37)	Progress (N=59)	Test	P
Age (y) (SD)	71.29 (±6.61)	74.15 (5.59)	t test	0.03*
Sex (F/M) (%)	21/16 (57/43)	29/30 (49/51)	χ2	0.40
VA (Snellen) (SD)	0.82 (0.23)	0.84 (0.22)	t test	0.73
CCT (µm) (SD)	541.67 (35.83)	542.86 (33.26)	t test	0.87
IOP at diagnosis (mmHg) (SD)	30.54 (5.88)	33.75 (5.01)	t test	0.005*
MD at diagnosis (dB) (SD)	−3.7 (3.44)	−7.58 (4.96)	t test	<0.001*
VFI at diagnosis (%) (SD)	92.05 (9.80)	82.61 (14.72)	t test	<0.001*
Phakic/pseudophakic at diagnosis (%)	29/8 (78/22)	45/14 (76/24)	χ2	0.88

CCT indicates central corneal thickness; F, female; IOP, intraocular pressure; M, male; MD, mean deviation; VA, visual acuity; VFI, visual field index.

^*Significant values at P≤0.05.

Upon inclusion, a significant correlation was found between the IOP and MD values (Pearson coefficient; P= 0.001, r=0.51). Similar results were obtained for the correlation between IOP and VFI values at inclusion (Pearson coefficient: P=0.001, r=0.55). The correlation between the MD and VFI values at inclusion was high (Pearson coefficient; P<0.001, r=0.92). The VIF test for multicollinearity at inclusion was 5.32 for IOP, 14.26 for MD, and 15.13 for VFI.

During the 3-year follow-up period, 6 patients underwent cataract surgery in the no-progress group, and 8 patients underwent surgery in the progress group. The difference was not significant (χ² test; P=0.54). During the same period, 2 patients were treated with SLT in the no-progress group, while 20 patients were treated with SLT in the progress group. Differences were considered statistically significant (χ² test; P<0.001). The number of medicines used was significantly higher in the progress group (around 3 medicines) than in the no-progress group (around 2 medicines) (t test; P<0.001). The ROP was much higher in the progress group than in the no-progress group. Considering the MD values, the ROP was double in progressors than in nonprogressors (−1.38 dB/y vs. −0.6 dB/y) (t test; P<0.001). Similar results were obtained when the VFI values were calculated as the ROP. The ROP was around 3 folds higher in progressors than in no progressors (2.95%/y vs. 0.93%/y) (t test; P<0.001) (Table 2). The results for the whole cohort are shown in Table 1, Supplemental Digital Content 1, http://links.lww.com/IJG/A846.

TABLE 2.

Clinical Characteristics of the Cohort During the Follow-up Period (3 years) According to the GPA.

	No progress N=37	Progress N=59	Test	P
Cataract surgery under follow-up (%)	6/31 (16/84)	8/51 (14/86)	χ2	0.54
SLT treatment under follow-up (%)	2/35 (5/95)	20/39 (34/66)	χ2	<0.001*
IOP reduction at 12 months (mmHg) (SD)	10.59 (5.40)	13.46 (4.62)	t test	0.006*
Percentage reduction of IOP at 12 months (%) (SD)	32.74 (11.79)	38.92 (8.57)	t test	0.004*
IOP at 36 months (mmHg) (SD)	17.01 (1.85)	17.98 (1.78)	t test	0.81
MD at 36 months (dB) (SD)	−5.41 (3.54)	−12.42 (5.81)	t test	<0.001*
VFI at 36 months (%) (SD)	88.54 (10.09)	72.76 (17.25)	t test	<0.001*
No. medicines at 36 months (SD)	2.10 (0.81)	3.06 (0.73)	t test	<0.001*
MD rate of progression (dB/year) (SD)	−0.6 (0.5)	−1.38 (0.9)	t test	<0.001*
VFI rate of progression (%/year) (SD)	0.93 (0.46)	3.55 (1.73)	t test	<0.001*

SLT indicates selective laser trabeculoplasty; IOP, intraocular pressure; MD, mean deviation; VFI, visual field index.

^*Significant values at P≤0.05.

The average IOP at inclusion for the entire cohort was 32.52 mm Hg (±5.54), whereas at the six-month visit, the average IOP was 21.19 mm Hg (±1.91). Although there was a difference in IOP between progressors and nonprogressors at inclusion, this difference was not observed at 6 months or during the remaining follow-up period. The evolution of the IOP values is shown in Figure 1.

FIGURE 1.

Evolution of the IOP values in the 3 years’ follow-up period. The bars represent the 95% CI for the mean. IOP indicates intraocular pressure.

The evolution of the MD and VFI values during the 3-year follow-up period is shown in Figures 2 and 3. It can be observed that there was a difference in the MD and VFI values between progressing and nonprogressing patients at inclusion, and this difference increased with time. The most significant deterioration in VFI occurred between diagnosis and 18–24 months, after which the VFI stabilized. A similar trend was observed for the MD values.

FIGURE 2.

Evolution of the MD values in the 3 years’ follow-up period. The bars represent the 95% CI for the mean. MD indicates mean deviation.

FIGURE 3.

Evolution of the VFI values in the 3 years’ follow-up period. The bars represent the 95% CI for the mean. VFI indicates visual field index.

The IOP, MD, and VFI values at diagnosis were found to be predictors of glaucoma development, and their associations were tested using different models. First, a linear regression model using IOP at diagnosis was used as the predictor and MD-ROP as the dependent variable. Another model used IOP at diagnosis as a predictor and VFI-ROP as the dependent variable. A significant association was found in both the models (P=0.007 and 0.006, respectively). Finally, a third logistic regression model used IOP at diagnosis as a predictor and GPA (binary) as the dependent variable. This model also showed a significant association (P=0.007).

Further models were constructed using the MD and VFI at diagnosis as predictors. Significant associations were found between MD at diagnosis and MD-ROP (P=0.005) and between MD at diagnosis and GPA (P=0.001). In addition, VFI at diagnosis was included in another model to test the association between VFI at diagnosis and VFI-ROP, which revealed a significant association (P=0.003). Finally, a logistic regression model showed that VFI at diagnosis was associated with GPA (P=0.005) (Table 3).

TABLE 3.

Regression Analysis Using IOP, MD, and VFI at Diagnosis as a Predictor in the GPA, MD-ROP, and VFI-ROP Models

	GPA				MD-ROP			VFI-ROP
Effect predictor	OR (95% CI )	PA	R2 †	P	β coeff. (95% CI)	R2 ‡	P	β coeff. (95% CI)	R2 ‡	P
IOP	1.12 (1.03–1.22)	69.1	0.15	0.01*	0.05 (0.02–0.08)	0.19	0.007*	0.15 (0.05–0.24)	0.18	0.006*
MD	1.06 (0.98–1.14)	71	0.23	0.001*	0.34 (0.01–0.60)	0.14	0.005*	NA	NA	NA
VFI	1.07 (0.99–1.15)	68	0.2	0.005*	NA	NA	NA	0.12 (0.04–0.19)	0.17	0.003*

The models were adjusted for age, SLT treatments, and the number of medications.

^†Adjusted Nagelkerke R².

^‡Adjusted R².

GPA indicates glaucoma progression analysis; MD, mean deviation; NA, nonapplicable; OR, odds ratio; PA, predictive accuracy (from the classification table); ROP, rate of progression; VFI, visual field index.

^*Significant values at P≤0.05.

Receiver operating characteristic of AUC analysis was conducted to test the accuracy of the model at diagnosis, and sensitivity and specificity values were obtained. At diagnosis, 3 predictors (IOP, MD, and VFI) were tested. For the IOP at diagnosis, the AUC was 0.74 (0.64–0.84), P<0.001; for the MD, the AUC was 0.76 (0.66–0.76), P<0.001, and for the VFI, the AUC was 0.77 (0.66–0.77), P<0.001. The AUC values showed acceptable discrimination of the model (Fig. 4). Four values were chosen from the list of cutoff values to be included in the table (Table 4), indicating the highest sensitivity and specificity. For IOP, the optimal values were ~30–31 mm Hg. For MD, the optimal values were around −4 dB. For VFI, the most accurate predictive values were between 91.50 and 93.50% (Table 4).

FIGURE 4.

The figure shows the area under curve (AUC) for the 3 predictors (IOP, MD, and VFI) at diagnosis. For the IOP at diagnosis, the AUC was 0.74 (0.64–0.84), P<0.001; for the MD, the AUC was 0.76 (0.66–0.76), P<0.001, and for the VFI, the AUC was 0.77 (0.66–0.77), P<0.001. IOP indicates intraocular pressure; MD, mean deviation; ROC, receiver operating characteristic; VFI, visual field index.

TABLE 4.

Receiver Operating Area Under Curve (AUC) Analysis for the Predictors IOP, MD, and VFI at Diagnosis

	AUC (95% CI)	P	Cutoff values(1)	Sensitivity	Specificity
IOP	0.74 (0.64–0.84)	<0.001*	29.50 mm Hg	82	51
	—	—	30.50 mm Hg	77	57
	—	—	31.50 mm Hg	72	60
	—	—	32.50 mm Hg	58	76
MD	0.76 (0.66–0.76)	<0.001*	−4.54 dB	78	70
	—	—	−4.31 dB	76	73
	—	—	−4 dB	73	77
	—	—	−3.65 dB	70	78
VFI	0.77 (0.67–0.77)	<0.001*	89	78	45
	—	—	91.50	73	52
	—	—	93.50	70	75
	—	—	94.50	60	82

The cutoff values included in the table were the ones showing highest sensitivity and specificity.

AUC indicates area under curve; IOP, intraocular pressure; MD, mean deviation; VFI, visual field index.

^*Significant values at P≤0.05.

The MD and VFI values of the entire cohort at 7 time points were studied. The time points were 0 (inclusion), 6, 12, 18, 24, 30, and 36 months. All the MD and VFI values were tested separately for the equality of variance (homoscedasticity) using Levene test. The Levene test for MD and VFI values showed homodestacity (P=0.07, 0.10, respectively). ANOVA was performed to detect the differences among the values, and afterward, a post hoc Tukey test was performed. The ANOVA results were significant (P<0.001 and <0.001) for the MD and VFI values (Table 5). Tukey test of the MD values showed a significant difference between the baseline and 24, 30, and 36 months (P=0.03, 0.001, and <0.001, respectively). Significant differences were also found between 6, 24, 30, and 36 months (P=0.04, 0.002, P<0.001, respectively). In addition, significant differences were found at 12, 30, and 36 months (P=0.03, 0.003, respectively). Similar results were obtained when VFI values were considered (see Tables 2, Supplemental Digital Content, http://links.lww.com/IJG/A846 and 3, http://links.lww.com/IJG/A846).

TABLE 5.

ANOVA Analysis of MD and VFI Values in the Whole Group (N=96) in 36 Months

	Time (mo)							P †	P ‡
	0	6	12	18	24	30	36
MD (dB) (SD)	−6.10 (4.81)	−6.16 (5.81)	−6.79 (4.93)	−7.51 (5.06)	−8.42 (5.41)	−9.22 (5.90)	−9.75 (6.09)	0.07	<0.001*
VFI (%) (SD)	86.22 (13.44)	85.22 (13.87)	84.14 (14.41)	79.97 (16.42)	78.37 (17.09)	77.31 (17.85)	76.07 (18.28)	0.10	<0.001*

dB indicates decibel; MD, mean deviation; VFI, Visual field index.

^†Levene test.

^‡Analysis of variance test (ANOVA).

^*Significant values at P≤0.05.

At the earliest time point (24 months), the predictor models were recalculated in the same manner as at diagnosis, and 4 linear regression models were performed. The first 2 models tested the association between IOP and MD-ROP and between IOP and VFI-ROP. The third test assessed the association between the MD and MD-ROP. The fourth model tested the association between the VFI and VFI-ROP. Three logistic regression models were used to test the associations between IOP/GPA, MD/GPA, and VFI/GPA. The models showed no association between IOP and the progression of MD-ROP and VFI-ROP (P=0.09, 0.06). The IOP values at 24 months were not predictors of progression. In contrast, an association was found between MD and MD-ROP (P<0.001) and between MD and GPA (P<0.001). Similar results were found for the associations between VFI and VFI-ROP (P<0.001) and VFI and GPA (P<0.001) (Table 6).

TABLE 6.

Regression Analysis Using IOP, MD, and VFI at 24 Months as Predictor in the GPA, MD-ROP, and VFI-ROP Models

	GPA				MD-ROP			VFI-ROP
Effect predictor	OR (95% CI )	PA	R2 †	P	β coefficient (95% CI)	R2 ‡	P	β coefficient (95% CI)	R2 ‡	P
IOP	NA	NA	NA	0.47	NA	NA	0.09	NA	NA	0.06
MD	0.7 (0.59–0.82)	81	0.44	<0.001*	0.08(0.06–0.11)	0.36	<0.001*	NA	NA	NA
VFI	0.94 (0.89–0.98)	82	0.46	<0.001*	NA	NA	NA	0.6 (0.06–0.12)	0.40	<0.001*

The models were adjusted for age, SLT treatments, and number of medications.

^†Adjusted Nagelkerke R².

^‡Adjusted R².

GPA indicates glaucoma progression analysis; MD, mean deviation; NA, nonapplicable; OR, odds ratio; PA, predictive accuracy (from the classification table); ROP, rate of progression; VFI, visual field index.

^*Significant values at P≤0.05.

The correlations between the predictors of IOP/MD and VFI were studied at 24 months. No correlations were found between the IOP and MD (Pearson coefficient; P=0.06) or between IOP and VFI (Pearson coefficient; P=0.07). However, a strong correlation was found between the MD and VFI values (Pearson coefficient; P<0.001, r=0.91).

An AUC analysis was performed again at 24 months to test the accuracy of the model. The predictors included IOP, MD, and VFI values at 24 months. The IOP values were not significantly different (P=0.42). The MD values showed an AUC of 0.86 (0.79–0.94) and were significant (P<0.001). Similar results were obtained for the VFI values, with an AUC of 0.88 (0.82–0.96) (P<0.001). In both cases (MD and VFI), the AUC showed excellent model discrimination at 24 months (Fig. 5). With sensitivity and specificity around 80%, the cutoff levels for MD were around −6.65 to −6.9 dB. In the case of VFI values at 24 months, with a sensitivity and specificity of ~80%, the cutoff level was ~86% (Table 7).

FIGURE 5.

The figure shows the area under curve (AUC) for the 3 predictors (IOP, MD, and VFI) at 24 months. The IOP values were not significantly different (P=0.42). The MD values showed an AUC of 0.86 (0.79–0.94) (P<0.001). Similar results were obtained for the VFI values, with an AUC of 0.88 (0.82–0.96) (P<0.001). IOP indicates intraocular pressure; MD, mean deviation; ROC, receiver operating characteristic; VFI, visual field index.

TABLE 7.

Receiver Operating Characteristics (ROC) Area under Curve (AUC) analysis at 24 months for the predictors IOP, MD and VFI at diagnosis

	AUC (95% CI)	P	Cutoff values(1)	Sensitivity	Specificity
IOP	0.54 (0.43–0.66)	0.42	NA	NA	NA
MD	0.86 (0.79–0.94)	<0.001*	−7.75 dB	84	70
	—	—	−6.9 dB	81	77
	—	—	−6.65 dB	78	78
	—	—	−5.7 dB	76	85
VFI	0.88 (0.82–0.96)	<0.001*	84.50	84	75
	—	—	85.50	81	78
	—	—	86.50	78	82
	—	—	87.50	77	87

The cutoff values included in the tables were the ones showing highest sensitivity and specificity.

AUC indicates area under curve; IOP, intraocular pressure; MD, mean deviation; VFI, visual field index.

^*Significant values at P≤0.05.

The correlation at 36 months between the MD-ROP and VFI-ROP measurements for the entire cohort was high (Pearson test; P= <0.001; coef= 0.95 (0.94–0.96). The correlation between MD-ROP and GPA was also good [point biserial calculation; P=<0.001; coef: 0.58 (0.43–0.70)]. The correlation between VFI-ROP and GPA was also good [point biserial calculation, P<0.001; coef: 0.73 (0.62–0.81)].

DISCUSSION

The present study showed a relatively high rate of visual field progression in patients newly diagnosed with EXFG. The GPA analysis revealed approximately two thirds of the glaucoma patients developed progression within 3 years. These findings support the results of previous studies that identified exfoliated glaucoma as an aggressive form of glaucoma.^4,6,14–16 However, it must be considered that the patients included in this study were newly diagnosed, and the progression of the disease may have attenuated later.

The present study recruited consecutive patients who visited the Eye Department. All patients were referred to the department by an optometrist because of elevated IOP readings. According to our recommendations, an optometrist should refer patients for consultations if their IOP is >25 mm Hg, as measured by a noncontact tonometer, at least twice. Therefore, patients with EXFG and lower IOP values were not included in this study. However, EXFG is a recognized type of “high IOP glaucoma”; hence, the number of excluded individuals should be minimal and most likely had no impact on the study’s findings.

Patients with advanced visual field progression had higher IOP values compared with nonprogressors. These findings were similar to previous studies, as IOP is the most well-known risk factor of glaucoma progression.¹⁷ Patients who showed an advanced progression in the 3-year follow-up period showed more deteriorated visual fields at inclusion than nonprogressors. At inclusion, most patients had a unilateral presentation of EXFG (69%). These results were consistent with those of previous studies.¹⁸ The MD and VFI values at inclusion differed significantly between progressors and nonprogressors. These results were consistent with those of previous studies that showed that patients with more damaged visual fields showed greater progression.¹⁹ The other factor that also differed between the progressors and nonprogressors was age; older patients showed more advanced progression than younger patients. These results were also consistent with those of previous studies.⁵

In this study, 3 variables (IOP, MD, and VFI) were chosen as predictors for the models. These 3 predictors could be combined using the same model. However, correlation analysis showed interactions between the predictors. Furthermore, the collinearity test VIF showed high values, indicating an interaction between the predictors. Logically, a higher IOP before diagnosis induced more damage to the MD and VFI at the time of inclusion. MD and VFI were not independent of the IOP values, and MD and VFI were not independent of each other. Combining all 3 predictors into the same model would have created confusion, and it would not have been possible to know which of the predictors was the cause of the effects as they interacted with each other.

Models based on the values at inclusion showed a significant association between IOP/MD and VFI values, and MD-ROP, VFI-ROP, and GPA. This indicated that the IOP, MD, and VFI values at inclusion were predictors of glaucoma deterioration in patients with newly diagnosed EXFG. The AUC showed values around 74%, 76%, and 77% for all predictors at inclusion. The sensitivity and specificity of the models at inclusion were 72/60%, considering an IOP cutoff value of 31.50 mm Hg. Considering the increased IOP (ie, 32.50 mm Hg), the sensitivity decreased to 58%. Meanwhile, the specificity increased to 76% (Table 4). This meant that when considering a higher IOP, the true positive rate decreased, including the negative cases among the positive ones. For the MD values, sensitivity and specificity were 76/73% at −4.31 dB. For the VFI values, sensitivity and specificity were 70/75% at 93.50% VFI. Better cutoff values of MD and/or VFI at inclusion would have decreased the sensitivity, thereby decreasing the true positive rate.

The present study tested the predictors at inclusion and how early the course of disease progression could be established. The problem in determining whether progression has occurred is the interindividual and intraindividual variation in the visual fields. The question is whether the variation in field parameters is a deterioration or only a normal biological variation. For this reason, ANOVA with a post hoc Tukey comparison test was performed. The models were retested 24 months after diagnosis and at the earliest time point detected using Tukey comparisons. The AUC increased from 76% at inclusion to 86% at 24 months for the MD values. Meanwhile, for the VFI values, the AUC increased from 77% at inclusion to 88% at 24 months. These results showed that at 24 months, the models had excellent discrimination between progressors and nonprogressors.²⁰ Even considering the specificity and sensitivity of the model, the values increased from 76/73% to 81/77% when MD values were considered. In the case of VFI, the specificity and sensitivity increased from 70/75% at inclusion to 81/78% at 24 months. In this fast-progressing glaucoma patient cohort, waiting for 7 visual fields (1 baseline and 6 follow-ups) to determine progression was probably unnecessary. Five visual fields were shown to be sufficiently accurate for assessing the progression and revaluation of the target IOP.

A “target” IOP was set at diagnosis at 20 mm Hg. This target IOP was achieved at 12 months (20.15 ± 1.84 mm Hg). Usually, the target IOP is re-evaluated after 7 visual fields (3 years) according to the Swedish guidelines for glaucoma care.⁸ In the present study, the average IOP at 24 months was 18.92 (±1.56) mm Hg, and the average IOP at 36 months was 17.94 (±1.8) mm Hg. However, according to the GPA analysis, several patients with glaucoma still showed progression. Therefore, the target IOP should be re-evaluated earlier to identify fast progressors and further reduce the IOP among them.

The prediction accuracy (PA) of logistic regression was estimated from the classification table shown in the regression analysis.^21,22 PA is a general measure of a model’s ability to discriminate between positive and negative values. The PA based on IOP/MD and VFI at inclusion was 69/71/68%, respectively. The model was retested 24 months after inclusion, showing no significant values for IOP but significant values for MD and VFI. The PA of MD was 81%, meanwhile for VFI was 82%.

The present study had certain limitations. All included patients were diagnosed with EXFG; the results would probably not apply to other clinical variants of glaucoma, thus creating diminished external validation and generalizability of the models.

One study limitation is treatment escalation, where patients with advanced glaucoma were treated aggressively. It would be unethical not to treat the patients according to the disease’s progression. A target IOP of 20 mm Hg was established at inclusion to mitigate this limitation. Furthermore, all included patients were born in Sweden; genetic mechanisms may have been involved; therefore, the results from this study cannot be applied to other populations. Glaucoma progression was measured using visual fields, which remains the gold standard for evaluating disease progression.⁷ No anatomic measurements of the optic nerve were performed (Optical Coherence Tomography). Furthermore, the study did not include patients with advanced glaucoma damage because of difficulties in follow-up. Therefore, our results only apply to patients with early or moderate glaucoma. Furthermore, selection bias must be considered, as patients who did not collaborate to perform the visual examination were excluded. However, the number of patients who were excluded for this reason was low. Finally, the models did not account for certain risk factors contributing to glaucoma progression, such as disc hemorrhages and blood pressure.

This study provides important clinical findings for the daily care of patients with EXFG. The cohort demonstrated a rapid average progression of −1.28 dB/y, which is considered fast progression.^4,23 However, there was a significant variation in progression rates among individuals, ranging from −0.1 to 4 dB/y. Therefore, each patient must be assessed individually. Despite setting an IOP target of 20 mm Hg, around 2/3 of glaucoma patients still experienced progression of the disease. The study found that performing 5 reliable visual fields is sufficient to differentiate between progressing and nonprogressing patients. In addition, high IOP levels at diagnosis (>31 mm Hg) predicted faster visual field deterioration. Finally, MD values at > −4.31 dB and/or VFI at 93.50% at diagnosis were also associated with faster progression and should be considered during diagnosis.

In conclusion, this study presented several risk factors for newly diagnosed EXFG. The models were based on both continuous and binary outcomes. The predictive capacities increased when the risk factors were retested 24 months after diagnosis. Upon inclusion, the risk factors showed sensitivity and specificity of around 70% and AUC of 77%. However, at 24 months, risk factors showed sensitivity and specificity ~80% and an AUC of 88%. The risk factors seemed to have a very good discriminative capacity at 24 months (5 visual fields) after diagnosis. Therefore, it may not be necessary to wait for 36 months to reevaluate the target IOP in patients newly diagnosed with EXFG. Instead, the target IOP can be reevaluated after 5 visual fields to slow down the progression of this aggressive type of glaucoma.

Supplementary Material

ijg-33-168-s001.pdf ^{(144.3KB, pdf)}