We have audited clinical data from two settings of the same tertiary glaucoma referral centre in the UK, to explore the outcomes of newly referred patients with suspected angle closure. Based on our data, one quarter of all glaucoma referrals within a year were for suspected angle closure in the absence of other glaucoma-related factors (PACS referrals4). There are very limited data on the burden of PACS referrals on hospital-based services. With the ZAP data having shifted the management of PACS away from the widespread use of LPI,3 such knowledge is important for service development and resource allocation. In a study conducted in Scotland, 4.6% (33/715) of all glaucoma referrals within a six-month period were PACS referrals, which is a lot lower than what was found it our study.14 However, comparisons with the Scottish data may not be appropriate. Unlike the rest of the UK, the Scottish general ophthalmic services (GOS) contract reimburses optometrists to conduct supplementary eye examinations.15 Therefore, an optometrist is likely to combine information on other risk factors before deciding on a referral.14
The accuracy of PACS referrals was low, with only 11% of patients referred for ‘narrow angles’ found to have occludable angles on gonioscopy. Previous studies have reported low or moderate accuracy of optometrist-initiated glaucoma referrals in the UK16–24, Sweden,25,26 and Australia.27 However, these studies focused on the detection of glaucomatous damage. To our knowledge, there are no previous studies on the accuracy of PACS referrals initiated by community optometrists. PACS referrals are based on the recommendation that patients with LACD of ≤ 25% (van Herick Grade 1 or 2)9 should be referred to secondary eyecare services.28,29 In a recent Cochrane Review, the pooled sensitivity and specificity estimates for LACD ≤ 25% were 83% and 88%, respectively.30 However, these estimates are likely to be overestimated due to the high risk of bias in 75% of the studies included in the meta-analysis.31 In addition, at a population level the PPV of LACD (the probability of a patient having occludable angles, given an LACD of ≤ 25%) is driven low by the low prevalence of angle closure disease. In a study by Foster et, LACD ≤ 25% had sensitivity of 99% and specificity of 66% in detecting occludable angles.32 Nonetheless, with the prevalence of angle closure at 6.8%, the PPV of the 25% cut-off was 17%. This is consistent with the 11% PPV of PACS referrals found in our study. The study by Foster et al was conducted in an Asian population and LACD was assessed by two senior Ophthalmologists. In the UK population, known to have lower prevalence of angle closure33 and with LACD assessed by less experienced observers, the PPV of the test is expected to be lower than 17%, which is exactly what we found. The PPV of the Van Herick test can be improved by increasing the pre-test probability of angle closure disease. Therefore, the new RCOphth guidance advises that patients with suspected angle closure should only be referred on the basis of elevated IOP, glaucoma or certain ocular, systemic and social risk factors constituting “PACS plus”.4 Our data fully support the new guidance and also suggest that such a strategy would be unlikely to put patients at risk of losing vision. Out of the 393 PACS referrals who were seen in clinic, there were no patients with glaucoma; out of the 42 patients with occludable angles on gonioscopy, only one had PAC and all other patients were PACS.
The timeframes of the two audits presented in this report should be considered in data interpretation (figure). Until the first half of 2019, all glaucoma referrals at Moorfields at Bedford were being assessed in the face-to-face clinic, whereas the virtual clinic was mainly for monitoring patients who were stable or at low risk of vision loss. With AS-OCT having enabled non-invasive high-resolution imaging of the anterior chamber angle,8,13,34 by the end of the same year the virtual clinic was expanded to receive all newly referred patients, including those with suspected angle closure. This model of care had been previously implemented at Cayton Street, which represents the main glaucoma virtual clinic at Moorfields Eye Hospital. This also explains the difference in the methods between the two audits. The majority (78%) of PACS in the first audit and half (52%) of PACS in the second audit underwent prophylactic LPI. The first audit covered a one-year period just before the publication of the ZAP trial,3 whereas the second audit covered a 5-month period just after the publication of this landmark study. However, it takes time to bridge the gap between a randomized controlled trial and clinical practice. Therefore, the management outcomes in these audits reflect former everyday clinical practice, with preventive LPI routinely performed in PACS.35 Clinicians would now be less likely to perform prophylactic LPI in PACS who do not have additional risk factors. Similarly, clinicians would be more likely to discharge PACS to their optometrists for annual review, rather than to keep them under monitoring in the hospital setting.
Although glaucoma virtual clinics are growing in numbers,5,36 there are no standardised criteria for further assessment in a face-to-face setting. Based on angle width criteria12 on AS-OCT (an iridocorneal angle of < 20o, later tightened to < 10o), 66% of patients with suspected angle closure in the virtual clinic required further assessment with gonioscopy. Of these, 1/3 had an intervention, 1/3 were kept under monitoring and 1/3 were discharged. These management outcomes raised the question of whether the triaging process could be more efficient. Determining angle width on AS-OCT requires familiarity with the built-in software and involves the identification of the scleral spur as a reference point, which may not always be easy to discern.37 For the same reason it often relies on the clinician’s impression, rather than an actual measurement. Conversely, determining whether there is ITC on an AS-OCT image is a binary question and, therefore, expected to be less variable. Using gonioscopy as the reference standard, AS-OCT had sensitivity of 74% and specificity of 82% in detecting ITC ≥ 1 quadrants. Clinic-based studies have previously reported high sensitivity (98%13 and 84%38) and moderate specificity (55%13 and 58%38) of AS-OCT in detecting ITC ≥ 1 quadrants, compared with gonioscopy. Community-based studies have also confirmed that AS-OCT identifies more ITC, compared with gonioscopy.39,40 The reasons behind the diagnostic disagreement between gonioscopy and AS-OCT have been previously discussed.13,34 However, direct comparisons with the literature may not be appropriate. The above studies were conducted in Asians, with all assessments performed by senior clinicians or research-trained observers. Our data reflect every day clinical practice, with AS-OCT images assessed by various clinicians and with gonioscopy performed in the context of busy glaucoma clinics.
Given its high sensitivity in detecting ITC,13,34,38−40 AS-OCT is a suitable modality for the assessment of the anterior chamber angle in a glaucoma virtual clinic, because patients with occludable angles are unlikely to be missed. This is also supported by our findings: 72% of those with ITC in ≥ 1 quadrants versus 5% of those without ITC on AS-OCT had an intervention. In addition, ITC ≥ 1 quadrants on AS-OCT as a triaging criterion seems to be more effective in reducing false positives, compared with angle width < 10o. In the re-audit, by changing the triaging criterion to ITC in ≥ 1 quadrants on AS-OCT, the proportion of patients who needed further assessment with gonioscopy was reduced from 66–46% and the proportion of patients who were discharged from the face-to-face clinic was reduced from 32–24%, compared to the initial phase of the audit. These data do not suggest that AS-OCT should replace gonioscopy for the detection of ITC. Gonioscopy remains the reference standard for confirming the presence of ITC and for diagnosing angle closure.4,41−43 However, it is a clinical skill that requires considerable training and experience, it involves contact with the patient’s eye and it is subject to variability.13,34 Therefore, gonioscopy is not suitable for a technician-led glaucoma virtual clinic.
This report presents real world data, reflecting common clinical practice. Therefore, our findings are likely to be applicable to other settings within similar healthcare systems. Real-world data and real-world evidence have become increasingly important in health care decisions.44,45 However, there are also shortcomings related to their use,44 and this is denoted in the limitations of this report. Gonioscopy was performed by various clinicians with different levels of experience. To minimise misclassification bias in the detection of occludable angles, medical records were reviewed manually to ensure that all details of the gonioscopic assessment are captured. Also, those who performed gonioscopy were not masked to the optometrist’s referral (first audit) or to the initial assessment in the virtual clinic (second audit). On the other hand, the low accuracy of PACS referrals found in our study does not suggest bias towards a higher detection of occludable angles. In addition, we have no information on the training level of each referring optometrist, which may have impacted the PPV of the Van Herick assessment. However, given that Moorfields at Bedford is the only NHS glaucoma service in the region, we are likely to have captured most, if not all, glaucoma referrals within the given period. Therefore, our data are likely to reflect all levels of experience in optometric practice. Conversely, we may have missed some patients with suspected angle closure in the virtual clinic because the identification of patients for the audit partly relied on the consultants who routinely performed the reviews. The lack of standardisation in the triaging process from the virtual to the face-to-face clinic is another limitation. While the clinicians were guided by the AS-OCT criteria, the overall decision to refer the patient for further assessment may have been affected by other factors. This is an inherent limitation of the virtual setting because, to date, there are no previously published criteria to guide this process. In addition, we do not have gonioscopic data for the patients who were discharged directly from the virtual clinic.
In conclusion, using real world data from a tertiary glaucoma specialist practice in the UK, we found that PACS referrals represent a substantial burden to hospital eye services and their accuracy is low. We also found that PACS referrals mostly represent individuals at low risk of vision loss, even if occludable angles are confirmed on gonioscopy. Our data also suggest that AS-OCT is a useful modality in a glaucoma virtual clinic for the triaging of patients who need further assessment with gonioscopy. Compared to angle width < 20o or < 10o, the presence of ITC ≥ 1 quadrants on AS-OCT appears to be a more effective triaging criterion to identify those who need further assessment with gonioscopy.