The Association Between Intraocular Pressure and Visual Field Worsening in Treated Glaucoma Patients

Abstract

Purpose:

To describe the relationship between mean treated intra-ocular pressure (IOP) and visual field (VF) worsening and understand how this relationship is affected by glaucoma severity and IOP range.

Methods:

1,446 eyes of 869 treated glaucoma patients with at least five longitudinal reliable VF tests and IOP measures were included. Mixed-effects linear models were employed to understand the effect of eye-specific mean treated IOP on mean deviation (MD) slope. Models included interaction terms to assess the differing relationships between MD slope and mean IOP by glaucoma severity (suspect/mild vs. moderate/advanced) and splines to account for the differing effects of mean IOP on MD slope at different IOP ranges (above or below 21mmHg).

Results:

In suspect/mild glaucoma, when treated IOP values were <21 mmHg, a 1mmHg increment in IOP was not associated with an increase in the rate of VF worsening (p>0.05) but when treated IOP values rose above 21mmHg, a 1mmHg increment in IOP was associated with faster VF worsening (−0.09 dB/yr per 1mmHg increment, p<0.05). In moderate/advanced disease, a 1mmHg increment in treated IOP was associated with faster VF worsening both below and above 21 mmHg, but the effect was much more pronounced in the higher range (−0.02 dB/yr per mmHg increment below 21 mmHg vs −0.74 dB/yr per mmHg increment above 21mmHg, p<0.05 for both).

Conclusions:

In treated eyes, IOP remained associated with worsening in eyes with more advanced glaucoma throughout the IOP range but was only associated with worsening in eyes with less severe glaucoma at higher IOP.

Precis:

In treated eyes with mild/suspect glaucoma, IOP increments are associated with worsening mean deviation once IOP reaches the 20s. In moderate/advanced eyes, IOP increments are associated with worse VF performance across the entire IOP range.

Introduction:

Elevated intraocular pressure (IOP) is a well-known risk factor for the development of glaucoma^1,2 and the worsening of visual function measured with automated visual field (VF) testing³. IOP is also the only glaucoma risk factor that can be modified, and treatment options including medical¹, laser⁴, or surgical intervention.⁵ As such, IOP has served as a frequent outcome in clinical trials and is routinely tracked to assess the risk of glaucoma development and worsening in patients under clinical care. Previous longitudinal studies have demonstrated that higher mean IOP^1,5–9, maximum (peak) IOP^3,10, and standard deviation (SD)^3,10,11 of IOP are associated with higher rates of glaucoma worsening.

However, there are unresolved issues regarding the effects of IOP on VF worsening. In particular, the Early Manifest Glaucoma Trial (EMGT) has suggested that for every 1mmHg decrease in IOP from baseline there is a 10% reduction in risk of VF worsening and that on further visits the risk of VF worsening decreases by 11 to 13%¹² for every additional mmHg of mean IOP lowering. On the other hand, the Advanced Glaucoma Intervention Study (AGIS) suggests that as long as the maximum IOP is consistently kept below 18mmHg, net VF worsening is unlikely to occur⁵. Thus, while EMGT suggests the lower the IOP the better, AGIS suggests that lowering IOP to prevent maximum IOP from rising above 18mmHg may not have a significant additional effect, as long as lowering is consistent across time. From a patient management standpoint, the resolution of this dilemma has significant implications. Less invasive medical², laser¹³ and angle-based procedures^14–16 can often achieve mean IOPs in the mid to high teens which, if found consistently across appointments, would be suggested to be safe by AGIS, whereas subconjunctival filtering surgery^17,18, with its higher rates of complications¹⁹, is often required to achieve lower IOPs suggested to be meaningful by EMGT.

Therefore, the primary objective of this study is to clarify, in a treated clinical population, the effect of mean treated IOP on VF worsening. In this study, we specifically choose to focus on a real-world treated clinical population as much of the work studying the impact of IOP and glaucoma worsening has been done in the setting of major glaucoma clinical trials^2,3,5,12 which may not be indicative of the effects of IOP control in a real-world clinical setting where the clinicians main goal is to achieve IOP control with treatement²⁰ rather than adhere to pre-specified study guidelines. While there has been some prior^21–23 retrospective work that has addressed this question, sample sizes were limited. Therefore, we performed our analysis using a large database of clinical and VF data from patients treated for glaucoma at a tertiary care glaucoma practice. To achieve our primary objective of understanding the impact of mean treated IOP on VF worsening, we first analyze the effect of a 1mmHg change in treated IOP on VF worsening across various treated IOP ranges (i.e. asking if the effect of going from a treated IOP of 24 to 23 mmHg on VF worsening the same as going from an IOP of 14 to 13 mmHg?). Second, we assess whether various stages of glaucoma progress at different rates at the same levels of treated IOP (i.e. is an eye with advanced glaucoma more likely to progress at a treated IOP of 17 compared to an eye with mild glaucoma?). As disease progression during treatment is undesirable, our findings will help clarify the frequency and extent of treatment failures at various disease stages and IOP levels.

Methods

The study protocol was approved by the Johns Hopkins University School of Medicine Institutional Review Board (IRB) and adhered to the tenets of the Declaration of Helsinki. A waiver of consent was obtained to review VF data and obtain patient information via chart review. The study was HIPAA compliant.

Study Participants

Patients included in this study were 18 years of age or older and were evaluated at the Wilmer Eye Institute between 1998 and 2012 as previously described.²⁴ All study participants had a glaucoma-related diagnosis (glaucoma suspect or any other form of glaucoma). We analyzed all eyes that that had five or more VFs which were obtained with the Humphrey Field Analyzer (HFA II, Carl Zeiss Meditec Inc., Dublin, CA) using the Swedish Interactive Threshold Algorithm (SITA) Standard test protocol and the 24-2 pattern. Patients could have either one or both eyes included in the analyses. We excluded VFs if the percentage of false positives was >15% for all stages of disease and false negatives were greater than 25% for mild/suspect disease and 50% for moderate/advanced disease.²⁴

VF and Clinical Data Collection

VF data were retrieved for eyes meeting the inclusion criteria listed above. Mean deviation (MD) at baseline was used as the measure of disease severity: baseline MD>−6 was defined as mild/suspect disease, baseline MD >−12 and ≤ −6 as moderate disease and baseline MD ≤ −12 as advanced disease. Eyes with MD<−28 at baseline were not included. Due to insufficient numbers to model advanced disease in isolation (n for advanced: 162, n for moderate: 224, n for mild/suspect: 1062) we combined moderate and advanced disease into one subgroup. MD slope (decibel [dB] change per year) was calculated for each eye using standard linear regression modeling and was used as the primary dependent variable for the study.

We performed a chart review to determine baseline demographic characteristics (age at the time of the first VF, sex, glaucoma type, glaucoma or cataract surgery before the baseline VF, interim glaucoma or cataract surgery, and baseline number of glaucoma medications). Additionally, we collected treated IOP values for every visit where a VF was obtained. From these longitudinally collected treated IOP values, we calculated the mean treated IOP over follow-up for each eye. Additionally, we calculated the maximum treated IOP over follow-up and the SD of treated IOP values over follow-up. To eliminate the impact of outlier treated IOP values, on IOP metrics eyes with IOP values > 50mmHg were not included. Mean, maximum, and SD of treated IOP were used as the primary independent variables in the models described below.

Modeling of Influence of IOP characteristics and VF MD

We created scatter plots of the effects of mean, maximum, and SD of IOP over follow-up on MD slope and fit LOWESS curves to visualize the impact of differences in treated IOP parameters on MD slope, stratified by baseline disease severity. Based on these plots, we created mixed-effects linear regression models in which MD slope was the dependent variable and mean treated IOP over follow-up was the primary independent variable. Based on the appearance of the LOWESS curves demonstrating differing rates of MD slope change across different ranges of treated IOP metrics, we introduced spline terms into our models. The position of the spline knots was chosen based on the technique described by Muggeo²⁵. Additionally, we added interaction terms to the models to capture the different relationships between MD slope and treated IOP metrics for various disease stages.

Models also included terms to account for confounders that may impact the relationship between MD slope and treated IOP including, age, sex, race, glaucoma diagnosis, presence of baseline or interim glaucoma surgery, presence of baseline or interim cataract surgery, and the number of glaucoma medications at baseline. Additionally, to account for the clustering of two eyes from the same patient we employed a linear mixed-effects modeling approach with random intercepts. To allow for interpretability of the intercept value from the models, we centered covariates at the mean values for continuous variables and the largest group for categorical variables (female, Caucasian, phakic patient with no baseline or interim surgery). Based on the intercept and coefficient values from the models described above, we calculated the number of years it would take for the average eye in our dataset (female, white, phakic with no baseline or interim surgery) to experience a 1dB loss VF MD at various levels of mean treated IOP over follow-up.

We also wanted to understand the impact of treated IOP measures on the likelihood of VF worsening. Therefore, we developed logistic regression models to understand the relationship between treated IOP metrics and the likelihood of a −0.5dB per year MD change which is significantly higher than the mean rate of change in a clinical population (−0.05dB/yr).²⁶ Models used similar parameters to linear regression models described above except spline terms were not included due to insufficient numbers beyond the spline knots. All statistical analyses were performed in R version 3.5.1 (R Foundation for Statistical Computing, Vienna, Austria).

Results:

Demographic and ocular characteristics

A total of 10,051 VFs and clinical exams from 1,446 eyes of 869 patients were included in the analysis. The mean age was 64 (SD: 12.3) years, 66.2% were Caucasian, and 54.5% were female (Table 1). The majority of eyes had primary open-angle glaucoma (54.3%) and mild/suspect disease (73.2%). A minority of eyes had prior incisional glaucoma surgery or prior cataract surgery (25.7% and 21.1% respectively). Two hundred seventy-six eyes (19.9%) underwent glaucoma surgery, and 161 eyes (11.1%) underwent cataract surgery over the course of follow-up. Eyes were on a mean of 1.3 (SD 1.3) glaucoma medications at baseline. Mean IOP at baseline was 16.7 mmHg (SD: 4.6 mmHg) and decreased to 15.8 (SD 3.6) over the study period. Each eye had on average 6.7 VFs done over the course of follow-up (SD: 1.3) with a mean follow-up time of 5.14 yrs (SD: 1.20, median: 5.39 yrs, range: 0.48 yrs to 11.65 yrs)

Table 1.

Demographic and Ocular Characteristics of Subjects Studied

Patients, n	869
Mean Age, years (SD)	64.3 (12.3)
Sex
Female, n (%)	474 (54.5)
Race
White, n (%)	575 (66.2)
Black, n (%)	217 (25.0)
Other, n (%)	77 (9.0)
Eyes, n	1,446
Glaucoma Type
POAG, n (%)	759 (54.3)
POAS, n (%)	402 (28.7)
PACG, n (%)	92 (6.6)
PACS, n (%)	25 (1.8)
Other, n (%)	121(8.4)
Prior Glaucoma Surgery
Yes, n (%)	371 (25.7)
Interim Glaucoma Surgery
Yes, n (%)	276 (19.0)
Baseline Lens Status
Phakic, n (%)	1141 (79.2)
Pseudo-Phakic, n (%)	296 (20.5)
Aphakic, n (%)	9 (0.6)
Interim Cataract Surgery
Yes, n (%)	161 (11.1)
Mean Number of Glaucoma Meds (SD)	1.3 (1.3)
Treated IOP Data
Mean Baseline IOP, mmHg (SD)	16.7 (4.6)
Mean treated IOP over Follow-up, mmHg (SD)	15.8 (3.6)
Eyes with Mean Treated IOP >21, n (%)	111 (8.1)
Suspect/Mild, n (%)	102 (7.5)
Moderate/Advanced, n (%)	9 (0.6)
Mean Max Treated IOP, mm Hg (SD)	19.8 (5.0)
Mean Treated IOP SD (SD)	2.7 (1.6)
Visual Fields, n	10,051
Mean No. of VFs per Eye, (SD)	6.7 (1.3)
MD at Baseline
>−6 dB, n eyes (%)	1,059 (73.2)
≤−6 dB, n eyes (%)	387 (26.7)
Mean MD Slope, dB/yr (SD)	−0.31 (0.71)

SD: Standard Deviation, POAG: Primary Open Angle Glaucoma, POAS: Primary Open Angle Suspect, PACG: Primary Angle Closure Glaucoma, PACS: Primary Angle Closure Suspect, IOP: Intraocular Pressure, mmHg: Millimeters of Mercury, VF: Visual Field, MD: Mean Deviation, dB/yr: Decibels per Year

Description of dependent and independent variables

The mean rate of change of VF MD over time for all eyes was −0.31 dB/yr (SD: 0.71) (Table 1 and Figure 1). Eyes with moderate or advanced glaucoma at baseline had a faster rate of VF MD loss over time compared to eyes with mild/suspect glaucoma (Table 2). On average, eyes with mild/suspect disease lost −0.28 dB/yr (95% CI: −0.42 to −0.13) whereas eyes with moderate/advanced disease lost approximately −0.42 dB/yr (95% CI −0.60 to −0.24, p<0.01 comparing moderate/advanced disease to mild/suspect disease).

Figure 1.

Density Plot of Mean Deviation Change Over Time by Disease Stage

Table 2.

Table 2a. Effect of Mean Treated IOP on VF Progression
Disease Severity at Baseline	Mean MD Slope, dB/yr (95% CI)	Change in MD Slope for 1 mmHg Increase in Mean Treated IOP when Mean IOP ≤21 (95% CI)	Change in MD Slope for 1 mmHg Increase in Mean Treated IOP when Mean IOP >21 (95% CI)
Mild/Suspect	−0.28 (−0.42 to −0.13)*	−0.01 (−0.02 to 0.01)	−0.09 (−0.14 to −0.03)*
Moderate/Advanced	−0.42 (−0.60 to −0.24)*	−0.02 (−0.04 to 0.00)*	−0.74 (−0.88 to −0.60)*
Table 2b. Effect of Maximum Treated IOP on VF Progression
Disease Severity at Baseline	Mean MD Slope, dB/yr (95% CI)	Change in MD Slope for 1 mmHg Increase in Max Treated IOP when Max IOP ≤31 (95% CI)	Change in MD Slope for 1 mmHg Increase in Max Treated IOP when Max IOP>31 (95% CI)
Mild/Suspect	−0.29 (−0.44 to −0.14)*	−0.01 (−0.02 to 0.00)	−0.14 (−0.20 to −0.09)*
Moderate/Advanced	−0.41 (−0.59 to −0.23)*	−0.01 (−0.03 to 0.00)	−0.26 (−0.33 to −0.20)*
Table 2c. Effect of SD of Treated IOP on VF Progression
Disease Severity at Baseline	Mean MD Slope, dB/yr (95% CI)	Change in MD Slope for 1 mmHg Increase in SD of Treated IOP when SD IOP ≤5 (95% CI)	Change in MD Slope for 1 mmHg Increase in SD of Treated IOP when SD IOP >5 (95% CI)
Mild/Suspect	−0.29 (−0.44 to −0.14)*	−0.03 (−0.07 to 0.01)	−0.17 (−0.24 to −0.10)*
Moderate/Advanced	−0.39 (−0.56 to −0.21)*	−0.01 (−0.07 to 0.04)	−0.22 (−0.30 to −0.15)*

VF: Visual Field, MD: Mean Deviation, dB/yr: Decibels per Year, CI: Confidence Interval IOP: Intraocular Pressure, mmHg: Millimeters of Mercury, SD: Standard Deviation

^*=p<0.05

Mean (14.3, SD: 3.5) and maximum treated IOP (18.7, SD:5.1) of eyes with moderate/advanced disease tended to be lower than the mean (16.4, SD 3.5) and maximum treated IOP (20.2, SD 4.9) of eyes with mild/suspect disease (p<0.01 for both comparisons) (Figure 2). The SD of treated IOP was slightly higher in eyes with moderate/advanced eyes compared to eyes with mild disease (2.9 vs. 2.7 respectively, p=0.01).

Figure 2.

Density Plots of Mean, Maximum and Standard Deviation of Intraocular Pressure by Disease Stage

Influence of IOP on VF Change

The influence of mean IOP on change in VF MD varied by disease severity and overall level of treated IOP (Figure 3 and Table 2). After conducting a formal analysis for the ideal location spline knots²⁵, a single knot at 21mmHg was found to be the ideal location for a spline knot, and therefore we added a spline knot at this location. Additionally, because the impact of treated IOP parameters on MD slope appeared to vary by disease severity at baseline (Figure 3), we added interaction terms to the models to account for the varying effect of treated IOP on MD slope by disease severity.

Figure 3.

Effect of Mean, Maximum and Standard Deviation of Intraocular Pressure on Visual Field Mean Deviation Slope

Table 2a and Figure 3 (top) demonstrate that for mean IOPs below 21 mmHg, treated IOP appeared to not affect the rate of VF worsening in eyes with mild/suspect disease (−0.01 decrease in MD slope per 1 mmHg increment in IOP p=0.26). For eyes with moderate/advanced disease, at mean treated IOPs below 21, a 1mmHg increment in mean treated IOP did result in a statistically significant higher rate of VF worsening (−0.02 dB/yr per 1mmHg increment in IOP, p<0.05). Above a threshold of 21mmHg, both eyes with mild/suspect and moderate/advanced disease show higher rates of VF worsening with higher levels of mean treated IOP. However, the influence of a 1mmHg treated IOP increment was much more profound in eyes with moderate/advanced glaucoma compared to eyes with mild/suspect glaucoma (−0.74 dB/yr for 1mmHg increment in mean treated IOP above 21 mm Hg vs. −0.09 dB/yr per 1mmHg increment in mean treated IOP, p<0.05 for both). 68 of 377 eyes with moderate/advanced disease (18.0%) with mean IOP values <21 underwent rapid VF worsening (−1.0 dB/yr or worse), compared to 54 of 952 (5.7%) eyes with mild/suspect disease (p<0.01 when comparing moderate/advanced eyes to mild/suspect eyes).

Based on the results of our models, we found that various levels of mean treated IOP over follow-up affected eyes with mild/suspect disease and moderate/advanced disease differently (Table 3). Our models estimate that, on average, it will take approximately 3.2 years (95% CI: 2.9 to 4.0 years) for an eye with mild/suspect glaucoma and a mean IOP of 20 mmHg over follow-up to lose one dB of MD. Whereas for a treated eye with moderate/advanced glaucoma it would only take 1.9 years (95% CI 1.5 to 2.4) to achieve the same MD loss. At a mean treated IOP of 30, on average, an eye with mild/suspect disease would take 0.8 years (95% CI 0.7 to 1.9) to experience 1dB of VF loss, whereas a treated eye with advanced/moderate disease would achieve this same level of loss in just 0.1 years (95% CI 0.1 to 0.2).

Table 3.

Average Number of Years Required to Lose 1dB of MD at Various Mean Treated IOPs (95% CI)

Disease Severity at Baseline	10 mmHg	15 mmHg	20 mmHg	25 mmHg	30mmHG
Mild/Suspect	4.7 (2.8 to 6.9)	3.8 (3.4 to 4.1)	3.2 (2.9 to 4.0)	1.5 (1.1 to 2.7)	1.0 (0.7 to 1.9)
Moderate/Advanced	3.1 (2.4 to 4.1)	2.3 (2.3 to 2.4)	1.9 (1.5 to 2.4)	0.3 (0.2 to 0.3)	0.1 (0.1 to 0.2)

dB: Decibels, MD: Mean Deviation , IOP: Intraocular Pressure , CI: Confidence Interval, mmHg: Millimeters of Mercury

The likelihood of experiencing a rate of VF MD worsening of at least −0.5 dB/yr varied across the level of mean (Figure 4) treated IOP for eyes with mild/suspect vs. moderate/advanced disease at baseline. Higher levels of mean treated IOP were associated with a higher likelihood of VF worsening in both mild/suspect and moderate/advanced disease (Table 4). Additionally, we found that lower mean IOP was not associated with VF improvement (defined as MD slope >0) in adjusted regression models in both the mild/suspect group (OR 0.99 [95% CI 0.92 to 1.05]) or the moderate/advanced group (OR 0.95 [95% CI 0.86 to 1.05])

Figure 4.

Effect of Mean, Maximum and Standard Deviation of Intraocular Pressure on Proportion of Eyes with Visual Field Worsening (Worsening Defined as Mean Deviation Slope <−0.5 dB/year).

Table 4.

Table 4a. Effect of Mean Treated IOP on Odds of VF Loss of 0.5 dB per Year
Disease Severity at Baseline	Odds Ratio for 1 mmHg Increase in Mean Treated IOP (95% CI)
Mild	1.09 (1.02 to 1.16)*
Moderate/Advanced	1.13 (1.04 to 1.22)*
Table 4b. Effect of Maximum Treated IOP on Odds of VF Loss of 0.5 dB per Year
Disease Severity at Baseline	Odds Ratio for 1 mmHg Increase in Max Treated IOP when Max IOP ≤33 (95% CI)
Mild	1.09 (1.05 to 1.15)*
Moderate/Advanced	1.09 (1.03 to 1.15)*
Table 4c. Effect of SD of Treated IOP on Odds of VF Loss of 0.5 dB per Year
Disease Severity at Baseline	Odds Ratio for 1 mmHg Increase in SD of Treated IOP when SD IOP ≤5 (95% CI)
Mild	1.39 (1.20 to 1.60)*
Moderate/Advanced	1.28 (1.09, 1.51)*

IOP: Intraocular Pressure, VF: Visual Field, MD: Mean Deviation, dB: Decibels per Year, CI: Confidence Interval, SD: Standard Deviation

^*=p<0.05

Sensitivity Analyses:

We attempted to adjust for the impact of outliers in our dataset. There were four eyes with an MD slope <−5.0 dB/yr which may have unduly influenced the modeling results. Additionally, there were 17 eyes with mean IOP values >25 mmHg over follow-up. We separately removed these two sets of eyes from the analysis and found that there were no notable differences in the primary results stated above which reinforces the notion that higher mean treated IOP result in faster rates of VF MD loss and that the effects of IOP on VF loss varies by the absolute level of IOP and disease severity.

To account for the fact that some eyes included in the analysis were glaucoma suspects or ocular hypertensives and may not have the same risk of VF worsening as eyes with manifest glaucoma, we ran the models without these eyes. Again, our results showed no notable difference from the primary results stated above and in the tables and figures. Additionally, to account for the fact that we merged eyes with moderate and advanced glaucoma into one subgroup for greater statistical power, we performed a sensitivity analysis which categorized eyes into three separate subgroups for mild, moderate, and advanced disease. We found that the main results of our work did not change except in the subgroup of eyes with advanced glaucoma with a mean IOP >21. There were very few eyes in this group (n=3) and unsurprisingly there was no significant relationship between mean IOP and VF worsening in this subset of eyes.

We attempted to account for eyes that may have had many clinical visits over a short period of time and therefore have a disproportionate influence on IOP metrics. For any eye with VF and clinical measurements less than 6 months apart, we used mean IOP values of the visits within the 6-month time frame as input into our models. Again, the primary results of the models in this sensitivity analysis were not notably different from the primary results described in the tables and figures.

Discussion:

Our study demonstrates that, even amongst patients undergoing management at an academic center, higher mean treated IOP is associated with higher rates of VF loss and that the effects of mean treated IOP on VF loss vary by disease stage. In particular, eyes with moderate/advanced disease tend to progress at a faster rate at the same treated IOP levels compared to treated eyes with mild disease, and tend to progress at mean treated IOP levels in the mid to high teens, unlike eyes with mild/suspect disease which, on average, do not progress with treated IOPs in this range. Additionally, our work suggests that incremental differences in treated IOP have vastly different effects when occurring at different absolute levels of IOP. For instance, the impact of a 1 mmHg mean treated IOP increment in an eye with a mean treated IOP in the twenties results in a much larger rate of VF loss than in an eye with a mean treated IOP in the teens, suggesting that the impact of IOP is non-linear across the range of IOP.

Our results suggest that glaucoma stage at baseline is an important modifier of the effect of IOP measures on VF worsening, even in treated eyes. In particular eyes with advanced disease tend to progress more at the same levels of mean, maximum and SD treated IOP (Table 2, Figures 3 and 4). These findings are in agreement with prior work which has demonstrated that the diagnosis of advanced glaucoma at baseline is associated with a higher risk of VF loss over follow-up despite treatment.^6,27–29 Furthermore, our data suggest that the baseline rate of VF loss is higher in treated eyes with moderate/advanced disease compared to mild disease at baseline (Table 2), despite the fact that these eyes were treated more aggressively (with lower treated IOPs). Our study also demonstrates that increments in IOPs tend to have more significant impacts on the rate of VF loss in treated eyes with moderate/advanced glaucoma at baseline compared to eyes with mild/suspect disease. Specifically, treated eyes with moderate/advanced disease tended to deteriorate faster with higher IOP even over the lower IOP range, unlike treated eyes with mild disease, which showed no significant difference in the rate of deterioration over these lower treated IOP levels. These results suggest that when physicians follow recommendations that IOP should be kept in the mid-teens or lower in eyes with moderate/advanced disease^5,30,31 and low 20s or lower in eyes with mild/suspect disease^2,32–34, VF progression is minimal.

As IOPs were treated, our data do not indicate ideal treatment guidelines, but rather indicate when treatment failure is most likely. Given that worsening occurred even in the lower IOP range in advanced glaucoma patients, even though these individuals are likely to be more visually impaired and have less functional reserve, one could argue that treatment of this group could have been more aggressive. However, additional IOP lowering in this group needs to be balanced against the risk of treatment (particularly for surgical therapy) or burden of therapy. Moreover, worsening may be harder to detect in this group given the greater variability in VF testing in advanced disease^24,35.

Our work also demonstrates that the absolute level of treated IOP modifies the effect of changes in IOP on VF worsening, even in clinical populations where physicians are setting IOP goals to minimize disease worsening. In particular, in eyes with moderate/advanced disease, higher IOP over the lower range of IOP tended to have a small, albeit statistically significant impact on the speed of VF worsening, whereas at higher IOP ranges (above 20 mm Hg) the effect of equivalent increments in IOP on VF worsening is particularly worse. While prior work has demonstrated that the risk of VF worsening decreases by 10–20% per mmHg reduction in IOP^1,12,36,37, prior studies have not generally focused on whether the effect of IOP on glaucoma worsening depends on the absolute level of IOP. This is an important distinction for clinicians who treat glaucoma patients, as interventions that reduce the IOP from the mid-20s to the mid to high teens are likely to have a more substantial impact at slowing the rate of VF loss compared to interventions that lower a patients IOP from the mid-teens to the low teens or single digits. Often, the later interventions including traditional filtering surgeries such as trabeculectomy which are often associated with high rates of complications.¹⁹ Additionally, though some work has found that lower IOP levels are associated with VF improvement³⁸, logistic regression modeling with our dataset did not demonstrate that lower mean IOP was associated with VF improvement (MD slope greater than zero). We hypothesize that much of the VF improvement in this study was due to VF learning effect³⁹ which is likely independent of IOP.

We also found that higher maximum IOP and SD of treated IOP measurements are associated with greater rates of VF loss in both eyes with mild and moderate/advanced damage. In particular, when maximum treated IOP values were in the 30s, they were associated with significant VF worsening. These data confirm the findings of AGIS⁵ as well as other groups,⁴⁰ which demonstrate that even peak IOP levels (>18mmHG in AGIS) are associated with higher rates of worsening. Our results differ from these studies as we used splines terms to determine if maximum treatment IOP had different effects over the range of observed treatment IOPs. As such, our results cannot be directly compared to these prior studies. Based on these results, it is important for clinicians managing glaucoma patients to note that even the occasional spike of IOP into the 30s suggests rapid future worsening, and likely the need for more aggressive IOP lowering. Our work also agrees with prior work suggesting that higher inter-visit fluctuation in IOP (as measured by higher SD IOP) is associated with faster rates of VF loss^3,10. However, unlike prior work, our study demonstrates this effect is significant only when the SD of treated IOP is >5mmHG both for mild disease and moderate/advanced disease. These results indicate that while IOP fluctuations may play a role in glaucoma worsening, the rates of occurrence of levels of fluctuation that influence worsening are relatively low in a treated clinical population. Moreover, such measures are of lower practical significance, given the difficulty in calculating SD IOP in clinical practice. Finally, in the setting of treatment, variability in treated IOP may reflect changes in management due to progression (i.e a surgical procedure such as trabeculectomy that dramatically lowers IOP but increases SD IOP), as opposed to a cause of progression.

Our data provide valuable insight regarding how long it takes certain levels of IOP to produce damage to visual function in various subsets of treated glaucoma patients. Based on the results presented in table 3, at a treated IOP of 25, on average eyes with mild disease will take 1.5 years to lose 1dB of VF whereas for an eye with advanced disease the time it takes to lose a similar amount of field is 0.3 years. However, it must be noted that the course of each patients VF worsening is unique as demonstrated by the variation in the rates of VF loss and the likelihood of worsening at various levels of IOP in figure 3 and 4 respectively, therefore it is of paramount importance that the clinician follows each patients course individually with serial glaucoma testing as recommended by current guidelines⁴¹ before making concrete treatment decisions. However, the results of this work suggest when treatment is likely to be successful for patients treated in glaucoma clinics:

1. For patients with mild or suspect disease, when IOP below 20 is achieved, progressive VF loss was minimal and uncommon. This suggests that physicians (at least in this group), are sufficiently aggressive when achieving IOP lowering over this IOP range.

2. For moderate/advanced disease, lower treatment IOPs even in the less than 20 mm Hg range is associated with slower VF loss, although the relative impact of IOP decrements in this range likely is less than the impact of IOP decrements over the higher range of IOP (the 20s and above).

3. A treated IOPs in the 30s at any stage of disease indicates a poor clinical course. Patients with IOPs in this range despite being followed may have had unique reasons for not undergoing further therapy (aversion to surgery, poor compliance or medication tolerance, high surgical risk), and should be counseled about the high likelihood and rate of vision loss.

There are several limitations to the work presented here. First, we have used VF worsening as our primary outcome measure as our database did not include any structural data from OCT and may miss worsening that is occurring structurally but not functionally. Additionally, due to the lower prevalence of moderate and advanced disease in our population, we were unable to separate these two groups due to a lack of statistical power. Future work with a larger dataset including both structural parameters and more eyes may allow us to address these issues. Finally, these results should only be applied to a treated clinical population, they are likely not generalizable to an untreated group of patients (i.e. studies of new glaucoma patients, community-based glaucoma screenings, etc.). In untreated patients with glaucoma and baseline IOPs in the high 20s, some prior work has shown that a IOP lowering of 20–30% from pre-treatment levels results in no visual field loss at 5 years of followup.⁴² However, when compared to other studies that assessed newly diagnosed glaucoma patients, the results of our logistic modeling, which suggest that there is a 9 to 13% higher likelihood of VF worsening for every mmHg increment in IOP are consistent with the previously demonstrated results from EMGT³² . However, it should be noted that EMGT defined worsening with pointwise criteria and optic nerve photos unlike our definition of worsening which depended on MD slope less than 0.5db/yr, so these similar numbers may have been coincidental.

In summary, our data demonstrate that, in treated eyes, glaucoma worsening related to IOP is dependent on the stage of glaucoma as well as the treated IOP range. On average, eyes with advanced disease and/or a higher treated IOP range tend to have more VF loss with further treated IOP increments compared to eyes with less advanced disease and/or lower treated IOP range.