Diagnostic Efficacy of Normalization of Corneal Deformation Variables by the Intraocular Pressure in Glaucomatous Eyes

Abstract

Purpose: To analyze the diagnostic efficacy of normalization of corneal deformation variables by the IOP in healthy, primary-angle closure (PACG), and primary open angle glaucoma (POAG) patients.

Methods: Fifty-nine healthy, 83 POAG, and 57 PACG eyes, matched for age and pachymetry, were included in a prospective, observational, cross-sectional study. Goldmann applanation tonometry (GAT-IOP), Corvis-ST IOP, IOPg (Goldmann correlated), and IOPcc (cornea compensated) from the ocular response analyzer were acquired. Corneal hysteresis (CH) and corneal resistance factor (CRF) from the ORA, and deformation amplitude (DA) from Corvis-ST were analyzed. Further, ratios of CH, CRF, and DA to IOP were assessed among the groups and defined as new variables (e.g., ratio [CH, IOPcc] was the ratio of CH to IOPcc).

Results: Goldmann applanation tonometry–IOP, IOPcc, and IOPg of PACG and POAG eyes were significantly higher than normal (P < 0.05). Corvis-ST IOP of healthy eyes was similar to POAG (P > 0.05) but lower than PACG (P = 0.02). Corneal hysteresis and CRF of PACG and POAG were significantly lower than normal (P < 0.0001). The ratio (CH, IOPcc), ratio (CRF, IOPcc), and ratio (DA, IOPcc) of healthy eyes were significantly higher than those of PACG and POAG eyes (P < 0.0001). The sensitivity and specificity of ratio (CRF, IOPcc) and ratio (DA, IOPcc) was significantly better than CRF and DA alone in PACG eyes (P < 0.001). However in POAG, only the sensitivity and specificity of ratio (DA, IOPcc) was significantly better than DA alone (P < 0.001).

Conclusions: Presence of glaucoma may be indicated better by ratio variables than by CH, CRF, or DA alone.

Intraocular pressure continues to remain the only modifiable risk factor among eyes diagnosed with glacuoma.^1–3 Thinner corneas were found to be associated with increased risk of glaucoma.^1,4–6 Central corneal thickness (CCT) also affected the measured value of IOP in addition to the type of measurement method and corneal biomechanical properties.⁷ While there are several variables that can confound the measurement of IOP in vivo, there exists a direct correlation among IOP, corneal biomechanical properties and CCT, when evaluated in ex vivo and in mathematical models.^7,8 Thus, there is a continuing need to assess the correlation among corneal biomechanical properties, CCT, and IOP in glaucoma patients.⁹ 

The ocular response analyzer (ORA; Reichert Inc., Depew, NY, USA) was the first applanation device that was capable of measuring the IOP and quantifying the corneal deformation in response to an applied air-puff in glaucomatous eyes.⁹ The ORA reports two variables, corneal hysteresis (CH) and corneal resistance factor (CRF), which are used to quantify corneal deformation in glaucomatous eyes.⁹ In primary-angle closure glaucoma (PACG) and POAG, CH and CRF were reduced and correlated with elevated IOP before treatment.⁹ After treatment, eyes with lower CH and CRF underwent greater reduction in baseline IOP.⁹ After normalization of IOP, CH and CRF may still be lower than the normal, which may indicate some residual corneal remodeling induced by elevated IOP.² 

Corvis-ST (OCULUS Optikgerate Gmbh, Wezlar, Germany) is another bidirectional applanation device that uses a high-speed Scheimpflug camera to capture a cross-section of deforming cornea during applanation.^10–12 Unlike the ORA, Corvis-ST actually measures corneal deformation, which may have an inverse correlation with corneal stiffness (e.g., lower deformation may imply higher corneal stiffness and vice versa [Sinha, et al. IOVS 2014;55:ARVO E-Abstract 3701]). In glaucomatous eyes, deformation variables reported by Corvis-ST were lower than the corresponding magnitudes in healthy eyes mainly due to elevated IOP.^10–12 Therefore, the aim of this study was to compare corneal deformation variables reported by the ORA and Corvis-ST in a group of age and CCT matched healthy, PACG, and POAG subjects. Further, ad-hoc variables were analyzed among the groups of patients by normalizing variables reported by the devices with the measured IOP. These variables were calculated as the estimated change in the corneal deformation variable (e.g., change in CH per unit change [change by 1 mm Hg] in IOP). The hypothesis was that an aggregate variable combining corneal deformation and IOP would have a higher sensitivity and specificity in detection of glaucoma among eyes with matched CCT. 

Methods

This prospective, observational, cross-sectional study was approved by the institutional research and ethics committee of Narayana Nethralaya Multi-specialty hospital, Bangalore, India and conducted in accordance with the tenets of Declaration of Helsinki. Written informed consent was obtained from the patients. 

Study Population

Patients were selected from the glaucoma and general outpatient department of a tertiary eye care center located in southern region of India. Patients were classified as healthy, POAG, and PACG. One randomly selected eye of each patient was included in the study. The healthy group included patients who presented for glasses, allergy, or cataract surgery with normal angles and optic nerve head. Clinical evaluation of patients included IOP measurement with Goldmann applanation tonometry (GAT), Corvis-ST (OCULUS Optikgerate Gmbh), and Ocular response analyzer (ORA; Reichert Inc.) followed by gonioscopy. Diagnosis of glaucoma was performed as described by International Society of Geographical and Epidemiological Glaucoma (ISGEO) glaucoma classification working group.¹³ Patients with open angles on gonioscopy in all quadrants and with glaucoma suspicion due to high IOP or disc changes were subjected to dilated disc evaluation and visual fields. Patients, who had significant visual field changes, such as asymmetry along horizontal meridian, cluster in arcuate area on a reliable field test, and were correlating with disc findings in either eye, were classified as POAG. Patients with narrow angles on gonioscopy with signs of occlusion, such as patchy pigmentation of trabecular meshwork, pigments on anterior lenticular surface, sphincter changes, or patients with synicheal closure (which indicate primary angle closure) along with an abnormal optic nerve head disc, and correlating with visual fields change in either eye, were considered as PACG. In all the three groups, patients with refractive error between −4 diopters (D) and +3 D with cylinder limited to −2 D were recruited. All patients were on medication for management of their IOP at the time of the study and most (91.5% in PACG and 80% in POAG group) were on a combination therapy of prostaglandins and beta-blockers for a period of at least 6 months. Further, patients with clinical conditions that can affect corneal biomechanical properties were excluded, such as keratoconus, pellucid marginal corneal degeneration, ectatic corneal disorders, prior refractive surgery, prior cornea surgery, prior retina surgery, prior collagen crosslinking, cataract surgery done within last year, and aphakia. Also, all patients with secondary glaucoma were excluded. 

Study Design

Patients underwent a complete eye examination, which included manifest refraction, IOP measurement, slit-lamp examination, gonioscopy, disc evaluation, and dilated fundoscopy. Intraocular pressure was measured with GAT-IOP, ORA, and Corvis ST. A waveform score greater than 6.0 was used to select the best measurement from the ORA. Central corneal thickness was quantified with the Corvis ST. All measurements were done between 9 AM and 5 PM, with the interval between successive measurements exceeding 5 minutes or more. The sequence of IOP measurement does not appear to affect the measured IOP value as shown in an earlier study.¹⁴ From each device, two measurements were taken and average of the two was used for further analyses. 

Statistical Analysis

All continuous variables were reported as median ±95% confidence interval (CI) of the median or mean ± SEM after confirming normality of distribution with the Kolmogorov-Smirnov test. Because some of the variables weren't normally distributed, the nonparametric Kruskal-Wallis test was used for group comparisons. Four measures of IOP were compared among the groups: GAT-IOP, Corvis-ST IOP, IOPg (Goldmann correlated), and IOPcc (cornea compensated). Corneal hysteresis (in mm Hg) and CRF (in mm Hg) from ORA, and deformation amplitude (DA) from Corvis-ST were compared among the groups. Deformation amplitude was the measured displacement in millimeters of the corneal apex during air-puff applanation. Pattern standard deviation (PSD) and visual fields index (VFI) from visual fields measurement, and thickness of the retinal nerve fiber layer (RNFL; Optovue Inc., Fremont, CA, USA) were used for further comparisons with the corneal deformation indices using correlation coefficients. 

Corneal hysteresis is the difference between the pressures at which the cornea becomes flat during applanation (P₁, P₂; i.e., CH = P₁ − P₂). Similarly, CRF = P₁ − k × P₂, where k is a correction factor that accounts for varying CCT among eyes. If IOP was elevated, CH and CRF were lower, indicating that P₁ and P₂ were closer to each other in magnitude compared with healthy eyes.⁹ Thus, ad-hoc ratio variables were defined as the ratio (CH, IOP) = CH/IOP and ratio (CRF, IOP) = CRF/IOP. Therefore, it was evident that the ratio (CH, IOP) and ratio (CRF, IOP) would be lower in magnitude than CH and CRF alone in glaucoma eyes having elevated IOP compared with healthy eyes. Similarly, ratio (DA, IOP) would be lower than DA alone in case of eyes with elevated IOP. Area under the curve, sensitivity and specificity of CH, CRF, DA, and deformation coefficients to detect PACG and POAG were analyzed with receiver operating characteristics curve (ROC). It was hypothesized that the deformation coefficients may have better sensitivity and specificity to diagnose PACG and POAG from healthy eyes than CH, CRF, or DA alone. A P value less than 0.05 was considered statistically significant. Sample size was validated using GAT-IOP as the primary variable (the current gold standard in the measurement of IOP), and a type I/II error of 0.05 (P)/0.2 (80% power). All statistical analyses and sample size calculations were performed in MedCalc v15.8 (MedCalc Inc., Ostend, Belgium) and G*Power v3.1.9.2 (Universität Düsseldorf, Düsseldorf, Germany). 

Results

Table 1 lists the median with 95% CI of the variables reported by ORA and Corvis-ST. The number of eyes was 59, 57, and 83 in healthy, PACG, and POAG groups, respectively. Based on GAT-IOP, the minimum total sample size needed to achieve a type I/II error of 0.05/0.2 was 21. Thus, the study sample size was adequate. From Table 1, age (P = 0.96) and CCT (P = 0.37) were similar among the groups (P = 0.96). From Table 1, median GAT-IOP of PACG (16 mm Hg, P = 0.002) and POAG (15 mm Hg, P = 0.002) groups was significantly higher than the healthy group. IOPg of PACG was significantly higher than IOPg of POAG (17.6 mmHg, P = 0.003) and healthy (P = 0.003) group. Further, median IOPcc of PACG (P < 0.0001) and POAG (17.2 mm Hg, P < 0.0001) was significantly higher than the healthy eyes group. Median Corvis-ST IOP of healthy was similar to Corvis-ST IOP of POAG (16.5 mm Hg, P > 0.05) but lower than Corvis-ST IOP of PACG (16.75 mm Hg, P = 0.02). Among all the IOPs, IOPcc achieved the greatest statistically significant difference among the groups (P < 0.0001) and was used for further analysis. Figure 1 shows a comparison of all the IOP's of the different groups. From Table 1, the CH of PACG (8.2 mm Hg, P < 0.0001) and POAG (7.7 mm Hg, P < 0.0001) groups was significantly lower than the CH of healthy group. Similarly, the median CRF of PACG (9.0 mm Hg, P < 0.0001) and POAG (8.4 mm Hg, P < 0.0001) groups was significantly lower than the CRF of healthy eyes group. These trends are shown in Figure 2. However from Table 1, median DA of PACG (1.05 mm) was significantly lower than POAG (1.1 mm, P = 0.04) and was similar to healthy eyes (P > 0.05). Figure 3 shows the trend in DA among the groups. The mean spherical equivalent refractive error was −0.12 ± 0.19 D and +0.57 ± 0.29 D in the POAG and PACG group, respectively (P = 0.04). The healthy eyes had no refractive error. 

Table 1

 

Median (95% Confidence Interval) of Age, IOP, Thickness, and Corneal Deformation Variables in Healthy, PACG and POAG Subjects

 

Figure 1

 

A comparison of the IOP's by GAT-IOP, IOPcc, IOPG, and Corvis-ST of healthy, PACG, and POAG groups. Symbols on top of box-whisker plot indicate statistical significance (e.g., there was statistically significant difference between GAT of normal [#] and of PACG [#], and between GAT of normal [# #] and of POAG [# #]). P < 0.05 was considered statistically significant.

Figure 2

 

Cornea resistance factor (CRF) and corneal hysteresis (CH) in healthy, PACG, and POAG groups. Symbols on top of box-whisker plot indicates statistical significance (e.g., there was statistically significant difference between CRF of healthy [+] and of PACG [+], and between GAT of healthy [++] and of POAG [++]). P < 0.05 was considered statistically significant.

Figure 3

 

Deformation amplitude (DA) in healthy, PACG, and POAG groups. Symbols on top of box-whisker indicate statistical significance. P < 0.05 was considered statistically significant.

Since IOPcc achieved the highest statistical significance, it was used to calculate the ratio variables using CH, CRF, and DA. From Table 1, the median ratio (CH, IOPcc) of healthy (0.57) was significantly higher than the ratio (CH, IOPcc) of PACG (0.40, P < 0.0001) and POAG (0.37, P < 0.0001) groups. A similar result was obtained with the ratio (CRF, IOPcc) and ratio (DA, IOPcc). Table 2 lists the results of the ROC analyses. From Table 2 in the PACG group, CH had the highest area under the ROC curve of 0.78 with a sensitivity and specificity 64.9% and 84.75%, respectively. Among the ratio variables, the ratio (CRF, IOPcc) had the highest area under the ROC curve of 0.84 with a sensitivity and specificity 66.7% and 92.9%, respectively. From Table 2, the sensitivity and specificity of the ratio (CRF, IOPcc) and ratio (DA, IOPcc) to detect PACG was significantly better than the same of CRF (P < 0.0001 in column 5) and DA (P < 0.0001 in column 5) alone, respectively. 

Table 2

 

Area Under the Receiver Operating Characteristic Curve (AUROC), Sensitivity, Specificity, and Cut-Off in Healthy, PACG and POAG Subjects.

 

From Table 2 in the POAG group, CH had the highest area under the ROC curve of 0.83 with a sensitivity and specificity of 66.3% and 93.2%, respectively. Among the ratio variables, the ratio (CRF, IOPcc) had the highest area under the ROC curve of 0.80 with a sensitivity and specificity of 66.7% and 92.9%, respectively. From Table 2, the sensitivity and specificity of the ratio (CRF, IOPcc) and ratio (CH, IOPcc) to detect POAG from healthy eyes were similar to the same of CRF (P = 0.05) and CH (P = 0.1) when analyzed alone, respectively. However, the ratio (DA, IOPcc) was significantly better than DA alone in detection of POAG from healthy eyes (P = 0.025 in Table 2). Overall, the deformation coefficients had better diagnostic ability to differentiate glaucomatous eyes from healthy eyes. However, none of the variables were able to differentiate between PACG and POAG (P > 0.05). 

The correlations between the ratio variables and measures of severity of the disease were also assessed. Median PSD of the PACG and POAG group was 8.03 dB (95% CI: 6.55–9.01) and 7.64 dB (95% CI: 5.52–9.36), respectively (P = 0.66). Mean RNFL thickness of the PACG and POAG group was 63.82 ± 9.34 μm and 71.26 ± 7.99 μm, respectively (P = 0.57). Similarly, VFI of the PACG and POAG group was 64.58 ± 7.78% and 73.75 ± 4.87%, respectively (P = 0.31). Thus, the PACG and POAG were similar in terms of the severity of the disease. However, PSD, RNFL thickness, and VFI did not correlate with any of the ratio parameters (P > 0.05). 

Discussion

Age and CCT are important determinants of corneal deformation and IOP.⁷ In this study, most measures of IOP were higher in glaucoma than in healthy subjects. There was some discordance among devices within the glaucoma group (e.g., IOPcc was the different between healthy and POAG but Corvis-ST reported difference between healthy and PACG; Fig. 1). Unlike recent studies,⁹ CCT was similar among the healthy and glaucoma groups in this study and this allowed us to investigate the correlation between corneal deformation and IOP among healthy, PACG, and POAG eyes exclusively. Using multivariate analyses, there is ample evidence to support that the CH and CRF are lower in PACG and POAG subjects compared with the healthy, even when matched for age and CCT.⁹ Using multivariate analyses, DA was found to be lower in glaucomatous eyes with elevated IOP.^10–12 In this study, DA was lower in PACG but similar between healthy and POAG eyes. Deformation amplitude is direct measure of corneal deformation unlike CH and CRF, which are qualitative measures. Intuitively, more compliant corneas should have higher DA. 

Currently, there is no conclusive evidence to support that the cornea is biomechanically altered in glaucomatous eyes and the normalization of CH and CRF after reduction of IOP indicated that corneal deformation in glaucomatous eyes was similar to healthy eyes at normal IOP's.^2,14 From Table 2, CH, CRF, and DA were similar among the healthy and glaucomatous eyes as area under the ROC curve was quite low and had low sensitivity and specificity. Because CCT and age were similar among the healthy and glaucomatous eyes (Table 1), they were unlikely to influence the statistics. However, IOP did differ among the groups and IOPcc achieved the highest statistical significance (P < 0.0001 in Table 1). Therefore, ratio variables were defined and yielded interesting observations. These ratio variables essentially normalized CH, CRF, and DA to unit IOP change as explained previously. Firstly, they yielded higher area under the ROC curve in PACG (Table 2) indicating the influence of IOP on CH, CRF, and DA. Secondly, DA normalized by IOPcc had a significant improvement in area under the ROC curve in both PACG (P < 0.0001 in Table 2) and POAG eyes (P = 0.025 in Table 2). 

From a recent study, the ratio (CRF, IOPcc) was estimated from the mean values before and after treatment.² The ratio (CRF, IOPcc) was 0.31, 0.34, 0.27, and 0.51 before trabeculectomy, phacotrabeculectomy, Ahmed glaucoma valve implantation, and phacoemulsification, respectively.² The same increased to 0.78, 0.78, 0.73, and 0.61 after trabeculectomy, phacotrabeculectomy, Ahmed glaucoma valve implantation, and phacoemulsification, respectively.² The IOPcc after treatment was in the range of 11.4 to 14.9 mm Hg.² In our study, the median IOPcc in the healthy eyes was 15.7 mm Hg. Thus, in the eyes that underwent phacoemulsification with a mean IOPcc of 14.9 mm Hg after treatment,² the ratio (CRF, IOPcc) was similar to the ratio values of the healthy eyes in this study. This clearly shows that lower IOP after treatment results in higher values of ratio (CRF, IOPcc)² and the ratio will have the same magnitude as in an IOP-matched healthy eye. In another study, mean IOPg reduced from 31.55 to 11.47 mm Hg after treatment.¹⁵ Corneal hysteresis increased from 6.83 to 9.22 mm Hg.¹⁵ Therefore, the estimated corresponding ratio (CH, IOPg) increased from 0.22 to 0.80. In the previous study,² estimated ratio (CH, IOPcc) increased from 0.25 to 0.75 after trabeculectomy. Thus, both studies^2,15 achieved the same outcome in their patients after trabeculectomy. This shows that the ratio variables (ratio of variable to IOP) allowed repeatable evaluation of outcomes after incorporating the reduction in IOP following treatment and showed that corneal deformation after IOP reduction in glaucoma eyes was similar to the deformation of normal corneas at the same IOP level. 

Both CH and CRF are actually pressure based variables unlike DA. Thus, the ratio of CH and CRF to IOP provided a measure of relative span between the pressures at which the corneal becomes flat. With elevated IOP, this difference was lower indicating increased resistance to deformation because DA was also lower, when the CH and CRF were lower. A recent review of published work concluded that CCT along with the CH and CRF may be important predictors of glaucoma progression.⁹ However, another study concluded that adjusting the CH for IOP did not improve its sensitivity and specificity.⁵ Adjustment for CCT and IOP generally was done through statistical models (e.g., multivariate models).⁹ Combining corneal deformation variables and IOP as a ratio was the novel aspect of this study and these ratios showed significantly improved sensitivity and specificity to detect glaucoma particularly PACG (Table 2) unlike the previous study.⁵ This indicated that the measurement of corneal deformation normalized by IOP may be able to diagnose glaucoma better than unadjusted corneal deformation variables. Further, better metrics of corneal deformation other than CH, CRF, and DA were needed to improve the sensitivity and specificity, which may be useful for clinical diagnoses eventually. 

While the spherical equivalent was significantly different between the PACG and POAG groups, this difference wasn't clinically significant to cause significant differences in corneal deformation. A recent study found no statistically significant difference in CH and CRF between low myopia (mean ± SD: −1.64 ± 0.77 D) and hyperopia (mean ± SD: +2.2 ± 0.5 D) groups.¹⁶ Thus, refractive error wasn't a confounder in the outcomes of the study. It is also well known that prostaglandins can cause decrease in CCT. However, medication wasn't a confounder in this study because most patients were on prostaglandins and normal eyes were age as well as CCT matched. Both CCT and IOP vary in a general population. Therefore, combining corneal deformation with IOP and CCT may result in improved detection of glaucoma. This needs to be investigated further through newer indices. A limitation of this study is the requirement to understand these correlations at different IOP levels, and also at different severity levels of disease. There is also a need to assess correlations between deformation coefficients and structural changes at and around the optic disc further. If in vivo estimation of corneal biomechanical properties in glaucoma subjects improves,¹⁷ an altogether new method for early detection or prediction of glaucoma may be developed. Further, the implications of the ratio variables on the rate of progression of PACG and POAG as well in other variants of glaucoma need to be assessed in future studies. 

Acknowledgments

Disclosure: S. Tejwani, None; S. Devi, None; S. Dinakaran, None; R. Shetty, None; P. Meshram, None; M. Francis, None; A. Sinha Roy, Carl Zeiss, Inc. (F, R), Avedro, Inc. (F), Topcon Medical Systems, Inc. (F), Bioptigen, Inc. (F), P