**3. Variability of risk factor measurement and the effects on risk assessment**

It is important to note that, similar to most multivariate prediction models derived from prospective studies [14], we are making several assumptions when we apply the 5-year risk calculator to guide treatment decision: [1] we assume that the baseline variables that were measured are the most predictive of the risk of glaucoma development, [2] the model also assumes that the risk of glaucoma progression is linear, and [3] patients who are being assessed have similar clinical characteristics as the participants in the OHTS and EGPS.

However, the variability of baseline risk factor measurements is a known phenomenon. IOP is well known to vary from visit to visit [15], which can be due to regression to the mean phenomenon, diurnal variation as well as order of IOP measurement [15]. This variability is also observed in the performance of VF, hence the PSD value [16]. Results of VF examination can be affected by patients' subjectivity and the substantial test-retest variability. The PSD value is a weighted standard deviation of the differences between the measured and normal reference visual field at each test location. A higher PSD value merely suggests a more irregular 'hill of vision', which can contribute to variability in patients' responses and/or areas of focal loss. Given that, by definition, OHT subjects do not have glaucomatous VF defect, the PSD value tends to be low. Therefore, a slight variability in patients' responses can contribute to a significant change in its value. Care must be taken when interpreting PSD value as a stand-alone figure.

Therefore, the apparently more comprehensive risk stratification strategy that is based on the 5-year risk calculator can face several fundamental challenges. As discussed, the variability of IOP measurement and PSD value, even during baseline assessment, may add a considerable source of error to the risk calculation. Furthermore, due to the within-subject changes in risk factors' values during follow-up, the correlation between baseline and updated values may diminish with time [17]. One study has demonstrated that risk calculation is variable over time and that longitudinal changes in baseline variables correspond with changes in the risk estimation of glaucoma development [18]. In the study, the 5-year risk of POAG development was calculated by incorporating different measurements that assume the best-case scenario (baseline age, lowest PSD, highest CCT and lowest IOP) and the worst-case scenario (final age, highest PSD, lowest CCT and highest IOP). For the VCDR, a value of ±0.2 was applied to model interobserver and intraobserver variability (i.e. -0.2 in addition to the best-case scenario and + 0.2 in addition to the worst-case scenario). It was found that, within the same individual, the mean risk of POAG conversion could increase by almost 10-fold when comparing the worst- and best-case scenarios (5.0% vs. 45.7%, *P* < 0.01). Hence, risk stratification is dynamic, and risk estimations should be recalculated during follow-up visits as variables can fluctuate significantly within the same individual over time.

It is important to note that the VCDR data that derived the risk calculator was based on VCDR measurements on the optic disc stereophotography of the OHTS and EGPS cohorts. In the RCTs, the measurement and evaluation of the optic nerve head (ONH) were performed by highly trained, independent graders at designated optic disc centres that followed a strict protocol in a non-clinical setting [19]. In clinical practice, the assessment of ONH and the measurement of VCDR by individual ophthalmologists

#### *Visual Impairment and Blindness - What We Know and What We Have to Know*

are susceptible to intraobserver and interobserver variability [20]. The precision and quality of the ONH evaluation is unlikely to match those in the RCTs.

Nowadays, imaging technologies such as confocal scanning laser ophthalmoscopy by Heidelberg retinal tomography (HRT) and optical coherence tomography (OCT) provide an objective and reproducible measurement of optic disc parameters. However, different techniques of evaluating ONH (hence the VCDR measurement) may not agree with each other. For instance, there is a poor agreement in the optic disc measurements obtained from HRT and OCT [21]. A study that compared VCDR measurement obtained with OCT, HRT and stereophotography in untreated OHT patients showed that there were poor agreement and lack of interchangeability between different techniques [22]. This is due to the differences between

#### **Figure 1.**

*Optic disc of a normal right eye (a) and corresponding spectral-domain optical coherence tomography (OCT, (b)) delineating the anatomical measurements of disc margin. The green dots on the fundus photo represent the Bruch's membrane opening (BMO) identified by spectral-domain optical coherence tomography (OCT). The blue crosses represent the disc margin that was identified by an examiner with stereophotography. Hence, the green dots and blue crosses represent the potential disc margins that could be identified by different examiners (adapted from Chauhan BC and Burgoyne CF 2013 [25]).*

**123**

**Figure 2.**

*Risk Assessment of Ocular Hypertension and the Use of Medication*

the techniques in defining optic disc margin and optic disc cup. The assessment of ONH in HRT and stereophotography relies on the examiners to define the disc margin, which can be variable. Spectral-domain OCT demonstrated that the 'perceived disc margin' of HRT and stereophotography rarely correlate with the Bruch's membrane opening (BMO) (**Figure 1**) [23, 24], which is considered to be the true outer border of the neural tissue because axons cannot pass through an intact Bruch's membrane to exit the eye [25]. The BMO is also unaltered under larger change of IOP [26]. Hence, it is a more reliable landmark, especially for eyes with OHT. Spectral-domain OCT is arguably more accurate in defining the ONH because it defines the BMO at every clock hour. It can also reliably identify the cup margin by measuring the minimum distance between the BMO and the internal limiting membrane in all meridians; the built-in software can then define the maximal vertical diameter to be the vertical cup diameter. In comparison, the definition of vertical cup diameter by stereoscopic photography and HRT are likely to be less accurate. Both are based on subjective judgement of examiners in defining the cup and disc margin. In some cases of small cups that do not pass through the midline of the optic disc, HRT has difficulties in calculating the VCDR because it obtains the vertical cup diameter along the vertical axis at the midline of the disc. Hence, the value of VCDR becomes '0' if the cup does not pass through the

The study that compared VCDR measured by OCT, HRT and stereophotography

of the ONH in patients with untreated OHT also investigated how the degree of disagreement extended to their corresponding 5-year risk estimation when other risk factors were kept contant [22]. In the study, ONH images of 140 untreated OHT eyes (of 75 patients) were taken by fundus camera (stereoscopic images), OCT and HRT. ONH stereophotographs were evaluated with a stereo-viewer by two glaucoma specialists, and the VCDR was measured with the ImageJ software. VCDR measurements obtained with stereophotography, OCT and HRT were used to calculate the estimated 5-year risk. The study showed that there was disagreement in VCDR measurements between the three methods. This disagreement also extended to their corresponding 5-year risk estimation of POAG development [22]. When the comparison was made on the Bland–Altman plots, the range of discrepancies tended to widen with increasing mean risk, especially beyond the estimated

*Bland-Altman plot: comparison of 5-year risk estimations of POAG conversion that were calculated by vertical cup-to-disc ratio measured by different vertical cup-to-disc ratio measured by spectral-domain optical coherence tomography (OCT) and stereophotography performed by a glaucoma specialist. Notice that the range of discrepancies widen with increasing mean risk, especially beyond the estimated risk of >15%. OCT, optical coherence tomography; reader 1, stereophotography evaluation (adopted from Chan PP et al., 2019 [22]).*

*DOI: http://dx.doi.org/10.5772/intechopen.90804*

vertical midline.

#### *Risk Assessment of Ocular Hypertension and the Use of Medication DOI: http://dx.doi.org/10.5772/intechopen.90804*

*Visual Impairment and Blindness - What We Know and What We Have to Know*

quality of the ONH evaluation is unlikely to match those in the RCTs.

are susceptible to intraobserver and interobserver variability [20]. The precision and

*Optic disc of a normal right eye (a) and corresponding spectral-domain optical coherence tomography (OCT, (b)) delineating the anatomical measurements of disc margin. The green dots on the fundus photo represent the Bruch's membrane opening (BMO) identified by spectral-domain optical coherence tomography (OCT). The blue crosses represent the disc margin that was identified by an examiner with stereophotography. Hence, the green dots and blue crosses represent the potential disc margins that could be* 

*identified by different examiners (adapted from Chauhan BC and Burgoyne CF 2013 [25]).*

Nowadays, imaging technologies such as confocal scanning laser ophthalmoscopy by Heidelberg retinal tomography (HRT) and optical coherence tomography (OCT) provide an objective and reproducible measurement of optic disc parameters. However, different techniques of evaluating ONH (hence the VCDR measurement) may not agree with each other. For instance, there is a poor agreement in the optic disc measurements obtained from HRT and OCT [21]. A study that compared VCDR measurement obtained with OCT, HRT and stereophotography in untreated OHT patients showed that there were poor agreement and lack of interchangeability between different techniques [22]. This is due to the differences between

**122**

**Figure 1.**

the techniques in defining optic disc margin and optic disc cup. The assessment of ONH in HRT and stereophotography relies on the examiners to define the disc margin, which can be variable. Spectral-domain OCT demonstrated that the 'perceived disc margin' of HRT and stereophotography rarely correlate with the Bruch's membrane opening (BMO) (**Figure 1**) [23, 24], which is considered to be the true outer border of the neural tissue because axons cannot pass through an intact Bruch's membrane to exit the eye [25]. The BMO is also unaltered under larger change of IOP [26]. Hence, it is a more reliable landmark, especially for eyes with OHT. Spectral-domain OCT is arguably more accurate in defining the ONH because it defines the BMO at every clock hour. It can also reliably identify the cup margin by measuring the minimum distance between the BMO and the internal limiting membrane in all meridians; the built-in software can then define the maximal vertical diameter to be the vertical cup diameter. In comparison, the definition of vertical cup diameter by stereoscopic photography and HRT are likely to be less accurate. Both are based on subjective judgement of examiners in defining the cup and disc margin. In some cases of small cups that do not pass through the midline of the optic disc, HRT has difficulties in calculating the VCDR because it obtains the vertical cup diameter along the vertical axis at the midline of the disc. Hence, the value of VCDR becomes '0' if the cup does not pass through the vertical midline.

The study that compared VCDR measured by OCT, HRT and stereophotography of the ONH in patients with untreated OHT also investigated how the degree of disagreement extended to their corresponding 5-year risk estimation when other risk factors were kept contant [22]. In the study, ONH images of 140 untreated OHT eyes (of 75 patients) were taken by fundus camera (stereoscopic images), OCT and HRT. ONH stereophotographs were evaluated with a stereo-viewer by two glaucoma specialists, and the VCDR was measured with the ImageJ software. VCDR measurements obtained with stereophotography, OCT and HRT were used to calculate the estimated 5-year risk. The study showed that there was disagreement in VCDR measurements between the three methods. This disagreement also extended to their corresponding 5-year risk estimation of POAG development [22]. When the comparison was made on the Bland–Altman plots, the range of discrepancies tended to widen with increasing mean risk, especially beyond the estimated

#### **Figure 2.**

*Bland-Altman plot: comparison of 5-year risk estimations of POAG conversion that were calculated by vertical cup-to-disc ratio measured by different vertical cup-to-disc ratio measured by spectral-domain optical coherence tomography (OCT) and stereophotography performed by a glaucoma specialist. Notice that the range of discrepancies widen with increasing mean risk, especially beyond the estimated risk of >15%. OCT, optical coherence tomography; reader 1, stereophotography evaluation (adopted from Chan PP et al., 2019 [22]).*

risk of >15% (**Figure 2**). We should be careful when using the 5-year risk of >15% as our treatment threshold when using VCDR values that are obtained from different measurement techniques. This can dramatically alter the management approach for any OHT subjects, especially if they have a relatively high baseline estimated risk and/or larger VCDR. In the cohort of untreated OHT eyes from this study, up to 72 eyes (51.2%) would require treatment if OCT was used for assessing the VCDR, according to the ≥15% 5-year risk cut-off. On the other hand, only 54 eyes (38.6%) would require treatment if the VCDR measurements were obtained from stereophotography by one of the glaucoma specialists. Therefore, one must be cautious when applying the risk estimation obtained from the other means of measuring VCDR.
