Discussion

Of those over 40 years old in European derived populations, 0.4% are estimated to have PACG. Three-quarters of cases occur in female subjects. Our analysis suggests that PACG is between almost two and four times more common than previous estimates1 ,6 and there are 1.60 million people in Europe, 581 000 people in the USA and 130 000 people in the UK with PACG today. Accounting for the ageing population structure in these areas, cases are predicted to increase by 19% in the UK, 9% in Europe and 18% in the USA within the next decade. Our estimate of PACG prevalence is higher than previous estimates of 0.1%6 and 0.25%.1 The reasons for this are multifactorial. First, the accepted definition of PACG has changed considerably, data from new population studies are now available (Friedman et al, unpublished work)14 ,16 and also the analysis methods used to derive overall prevalence values differ.

Interpretation of PACG prevalence values from population studies is complicated by the inconsistent case definitions used (see table 1 and supplementary table 1) with some studies diagnosing PACG based only on a narrow anterior chamber angle with raised IOP19–21 (ie, PAC).2 In 2002, Foster et al proposed a standardised definition for PACG requiring structural and/or functional evidence of glaucomatous optic neuropathy in the presence of an ‘occludable’ angle with evidence of previous trabecular meshwork obstruction by peripheral iris (eg, raised IOP or peripheral anterior synechiae (PAS)).2 They defined an occludable angle as ≥270° of posterior trabecular meshwork not seen on gonioscopy in keeping with previous population studies.22 Thomas et al used a definition requiring ≥180° of irido-trabecular contact (ITC) for their studies in India.23 Later analysis by Foster et al showed the original 270° ITC definition to be too stringent as it excluded half of all individuals who had angle closure (seen by the presence of PAS) and glaucomatous optic neuropathy from being defined as PACG.24 The original description for occludable angle by Becker and Shaffer is believed to be the most inclusive definition7 with angle closure ‘probable’, and ‘possible’ with anterior chamber widths of 10° and 20° (grades 1 and 2), respectively.25 This variability in the gonioscopic definitions of an occludable angle is highlighted by comparison of the Egna-Neumarkt study11 where a Shaffer grade 0 was required for PACG diagnosis, whereas in the Ponza12 study, a Shaffer grade 2 or less was required. Clearly this will influence disease prevalence. Unpublished data from the Salisbury Eye Study show that using the ≥270° ITC definition, cases numbers are more than halved compared with when the ≥180° ITC definition is used (Friedman et al, unpublished work).14 While a revised consensus definition for an occludable angle has not been published, 180° or more of ITC has been proposed7 and is now the accepted clinical and research definition.26 Additionally, the advent of anterior segment OCT imaging suggests that gonioscopy does not detect all cases with ITC.27 We also found marked variability in the exact detail given for PACG definitions with PAS being described as part of the definition only in the Egna-Neumarkt11 ,18 and Salisbury Eye Evaluation Glaucoma Studies,14 and is assumed to have been part of the Wroclaw16 study based on their POAG definition. Furthermore, a raised IOP is only listed as a part requirement of a PACG definition in the Egna-Neumarkt,11 ,18 Ponza12 and Wroclaw16 studies. Ultimately, we used a pragmatic approach to identify studies with PACG definitions similar to the current ISGEO definition and included studies where the authors believed trabecular meshwork obstruction by peripheral iris to be the cause of the glaucoma. There were no studies with PACG definitions identical to the current ISGEO definition (although the Wroclaw16 study was closest), presumably as there are no prevalence studies that have been published following the ISGEO definition. The criteria for performing gonioscopy differed between studies, with gonioscopy typically only being performed once glaucoma was suspected,11 ,13–15 ,18 while others additionally performed gonioscopy if there was a shallow anterior chamber found on slit lamp or Van-Herrick grading.12 ,16 ,28 This may influence the identification of PACG cases.

Although there are no data comparing POAG and PACG disease severity in European populations, data from East Asia suggest that the risk of serious vision loss is considerably higher in PACG.5 As treatments including peripheral iridotomy and lens extraction can effectively arrest disease progression,29 ,30 there should be greater emphasis on accurate gonioscopy to discriminate between PACG and POAG when suspects are referred for glaucoma phenotyping. Following from this, we suggest that all primary glaucoma cases should be considered to be PACG until the anterior chamber angle is shown to be open on gonioscopy.

Our analysis has several limitations. The primary limitation being a consequence of the heterogeneity in PACG definitions and methods of anterior chamber angle assessment used as discussed. It is possible in our literature search that we missed prevalence studies that were not PubMed listed and we attempted to correct for this by reviewing the reference lists of all the prevalence studies we identified. In our model, we estimated the age-specific prevalence for 60–69 years old to be lower than that for those 50–59 or 70+ years old. This is surprising as PACG prevalence, like POAG, is believed to increase with age.1 This dip in the age-specific prevalence is also seen in the Ponza study,12 implicating this as a possible cause. However, a sensitivity analysis shows that even if there were no dip in observed prevalence in the Ponza study, it would still be present in our estimated age-specific prevalence values. We agree that it is likely that our 60–69-year age-specific prevalence is an underestimation, presumably due to the small number of studies contributing to the analysis. However, it is also possible that this finding is real and could be explained by different pathophysiological mechanisms that predominate at different ages, or that there may be increased cataract surgery rates in this age group leading to a cohort effect. Another limitation is that none of the PACG population studies from which we drew data enrolled all eligible participants and thus there is potential for response bias. Angle closure disease is more common in female subjects, who have longer life expectancy than male subjects, and our model did not account for this potentially underestimating case numbers. We also did not account for different PACG prevalence in ethnic minority populations. Again, this is likely to underestimate case numbers as PACG prevalence is higher in people of Asian and East Asian descent compared with European descent.1 In the USA especially, ethnic groups other than Caucasians form a large part of the whole population. In the Salisbury14 and Baltimore17 eye studies, 77% and 55% subjects respectively were Caucasian with the remainder being of African-American descent; the other studies did not report on the ethnicity of their populations. Finally, we assumed that the true PACG prevalence will not vary for any reason except population dynamics. This is unlikely to be the case as an analysis of UK hospital episode statistics showed the incidence of PACG appeared to be decreasing after reaching a peak in the late 1990s. The decline was attributed to increasing rates of phacoemulsification.31 More recent hospital episode statistics data show the decrease in PACG cases previously reported has reversed, with higher case numbers now than ever before and also that the incidence of acute angle closure in the UK appears to have halved in the past decade.32 These findings may in part be explained by increased physician awareness and increasing rates of phacoemulsification and laser peripheral iridotomy.

This analysis provides a current evidence-based estimate of PACG prevalence in European derived populations. PACG is more common than previously thought, and all primary glaucoma cases should be considered to be PACG until the anterior chamber angle is shown to be open on gonioscopy.