Variations in Primary Open-Angle Glaucoma Prevalence by Age, Gender, and Race: A Bayesian Meta-Analysis

Abstract

purpose. To quantify the variation in primary open-angle glaucoma (OAG) prevalence with age, gender, race, year of publication, and survey methodology.

methods. Medline, EMBASE, and PubMed were searched for studies of OAG prevalence. Studies with defined population samplings were sought. Forty-six published observational studies of OAG prevalence (103,567 participants with 2509 cases of OAG) were identified for inclusion in the systematic review and meta-analysis. Data on the number of people and the number of cases of OAG by age, race, and gender were sought for each study. Additional information was obtained regarding whether the definition of glaucoma relied on raised intraocular pressure (IOP) and whether visual field examination was performed routinely on all individuals. Bayesian meta-analysis was used to model the associations between the log odds of OAG and age, race, gender, year of publication, method of visual field testing, and effect of reliance on IOP in the definition of OAG.

results. Black populations had the highest OAG prevalence at all ages, but the proportional increase in prevalence of OAG with age was highest in white populations. The odds ratio per decade increase in age was 2.05 in white populations (95% credible interval, 1.91 to 2.18), 1.61 (95% credible interval, 1.53 to 1.70) in black populations, and 1.57 (95% credible interval, 1.46 to 1.68) in Asian populations. The average estimated prevalence in those older than 70 years of age was 6% in white populations, 16% in black populations, and 3% in Asian populations. After adjusting for age, race, year of publication, and survey methods, men were 1.37 (95% credible interval, 1.22 to 1.53) times more likely than women to have OAG. The prevalence of OAG was one third lower in studies in which routine visual fields were not assessed and that used an IOP criterion in the definition of glaucoma; this effect was reduced to the null after adjustment for age, racial group, and year of publication.

conclusions. Although black populations had the highest prevalence of OAG at all ages, white populations showed the steepest increase in OAG prevalence with age. Men were more likely than women to have OAG.

Many large population-based studies have been conducted to determine the prevalence of glaucoma. Most have been carried out in white populations, including studies from the United states, ^{1 2 3} the United Kingdom, ^{4 5} Scandinavia, ^{6 7 8} Australia, ⁹ and The Netherlands. ¹⁰ Studies have also been conducted on Asian, ¹¹ multiracial, and black communities. ^{2 12} Glaucoma is a leading cause of vision loss worldwide and is of major public health importance: it is estimated to affect 66 million people worldwide, with at least 6.8 million people bilaterally blind from the condition. ¹³ A recent meta-analysis of six studies confirmed a higher prevalence in black Americans than in white Americans. ¹⁴ Estimates of the prevalence of any glaucoma in white populations ranged from 1% to 1.5% in those aged 40 to 65 years, rising to 2% to 7% in those older than 65; estimates in black Americans ranged from 1.5% to 3.6% and 4.6% to 9.8%, respectively, for similar age groups. ¹⁴ These racial differences are of particular interest because they may allude to mechanisms of disease. 

This article is a systematic review of the published literature on OAG prevalence. The aim was to quantify the relation of OAG prevalence to age and gender and how these relationships vary by racial group. Heterogeneity in prevalence between studies attributed to differences in survey methodology and year of publication was also explored. 

Methods

Systematic Review Process

A systematic review of all published papers, letters, abstracts, and review articles were identified using Medline, EMBASE, and PubMed electronic databases. The following text word search strategy was used: glaucoma with prevalence or incidence or survey or epidemiology. The electronic search (completed in February 2005) yielded 132 unduplicated references. Two reviewers (ARR, SM-I) completed the literature search independently. 

Inclusion Criteria and Data Extraction

Population-based surveys were included that aimed to examine an entire town/village/geographic region or that used a clearly defined random or clustered sampling procedure. Studies that were surveys or audits of hospital eye departments or clinics were excluded. Studies without a defined sampling strategy were also excluded. 

Fifty-five articles were retrieved from abstract review for more detailed evaluation. Forty-six unduplicated studies were identified after full paper review, and two reviewers (ARR, SM-I) extracted data independently from these studies. The studies excluded after full paper review were those that did not use any defined form of population sampling or did not provide sufficient data for the 95% confidence interval (or standard error) of prevalence of OAG to be determined. 

There was considerable variation in the eye examination methods and case definitions of OAG. Studies were not excluded on the basis of definitions of OAG or methods or instrumentation used to identify cases of OAG. All studies examined the optic disk, but there was wide variation in the criteria for defining glaucomatous nerve damage. A previous review of six prevalence studies ¹⁴ noted that despite variation in survey methodology and definition of OAG, results were remarkably similar across studies. To examine this directly, studies were classified on the basis of two questions. 

First, does the study design include visual field testing on all participants? Studies classified as not performing routine visual field testing included those that (i) did not perform field testing at all (very few studies), (ii) restricted field testing to persons identified at high risk of OAG (high risk was typically defined on the basis of IOP or a family history of glaucoma), and (iii) studies that selected a proportion of participants randomly or consecutively for visual field testing. 

Second, does the case definition of OAG rely on an IOP-based criterion? If a study provided data on the number of OAG cases with and without increased IOP, it was combined for the meta-analysis. Such studies were then classified as not relying on IOP for the case definition of glaucoma. 

Several studies provided additional data on suspect/probable OAG. These data have not been included in this review because case definitions for probable/suspect OAG were unclear, not stated, or subjective, and it was difficult to make comparisons across studies. 

The following additional data were sought for each study: study name and year of publication (to assist with checking for duplicate publications); total number of persons invited to participate in the survey; number who attended for examination (alternatively, the reported response rate); average or median age or age range of study participants; racial group, defined as white (European, American, Australian), black (African, American, Caribbean, European), Hispanic (Mexican or European), Asian (Chinese, Eskimo, Japanese, Indian, Thai), and mixed; number (or prevalence) of cases of OAG overall and, if available, according to age, gender, and racial group if the study included more than one racial group; and average age or midpoint of the range in each age group or mean or median age for the study or midpoint of the age range specified. If the age group was specified as younger than x, older than x, or x+, then the age band was taken to be the same width as other age groups in the same study, and the midpoint of the range was used. 

Statistical Methods

A random effects meta-analysis ¹⁵ stratified by racial group of the log odds of OAG in each study was performed. Odds were converted back to prevalence estimates and displayed as a forest plot with 95% confidence intervals for all studies to show the variation in prevalence of OAG across studies in different racial groups. Heterogeneity between studies was estimated using χ² tests (the Q statistic), and I² values, ¹⁶ which measure the proportion of variability between studies that is due to heterogeneity rather than chance, were determined. The estimate of chance variation is based on within-study variances. 

A Bayesian meta-analytical approach was used to examine simultaneously the relationship between OAG prevalence and age, racial group, gender, whether the survey included routine visual field testing on all subjects, and whether the case definition of OAG relied on IOP. Year of publication was used to examine any trends in reported OAG prevalence over time. The possible influence of the reported participation rate on OAG prevalence was also investigated. 

A Bayesian logistic meta-regression model of the log odds of OAG with age was constructed based on the age-specific data extracted. Some studies contributed data for just one age group, whereas others contributed up to several age groups. The model took into account this hierarchical structure of the data by estimating the prevalence of OAG at 40 years of age within each study separately, but the change with age was estimated across all studies. Racial group was included as a study level covariate, and an interaction between age and racial group was investigated. Other variables were included as study level covariates in the meta-regression (see Statistical Appendix). A separate meta-analysis was performed to examine the effect of gender because only a subset of studies published data stratified this way. 

The usefulness of the Bayesian approach is that variation at all levels is taken into account so that final estimates reflect all sources of variability. It also takes into account that the within- and between-study estimates of variance are themselves estimated with error. Bayesian analyses yield posterior distributions, in this case for odds and odds ratios (OR). From the posterior distribution we report the median value as our best estimate and the 95% credible interval (CrI) as a measure of uncertainty. The CrI specifies the range of values within which the true value is expected to lie with 95% probability (it is the range between the 2.5% and 97.5% percentiles of the posterior distribution). Noninformative normal priors were used for log odds or log ORs, and noninformative gamma priors for the corresponding variances. 

Results

From 46 studies we obtained data on 103,567 participants, including 2509 with OAG. Table 1 ^{1 2 3 4 5 6 7 8 9 10 11 12 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50} summarizes the studies by racial group. Two studies reported on more than one racial group, and results were reported separately for each racial group. Figure 1shows the prevalence estimate for each study along with 95% confidence intervals, with pooled estimates and χ² tests for heterogeneity by racial group. There is considerable heterogeneity between studies within each racial group (in all cases, P < 0.0001). The I² statistics ¹⁶ were greater than 90% for all racial groups, showing that most heterogeneity was the result of variability between the studies rather than of chance variability. Because no significant differences in prevalence were observed between studies in white and Hispanic populations or between white populations and a single mixed-race population (the Mamre population in South Africa), these were all included as white studies. The effect of excluding Hispanic studies and the mixed-race population had little impact on the results (data not presented). The pooled random effects prevalence estimate of OAG is 1.4% (1.0%–2.0%) in Asian populations, 4.2% in black populations (95% CI, 3.1%–5.8%), and 2.1% (95% CI, 1.6%–2.7%) in white populations. 

Forty-one of the 46 studies published data stratified by age. Inspection of the data showed a single outlying study in an Asian population with an exceptionally low prevalence of POAG (0.03%) ¹⁹ that was excluded from the analyses. Figure 2shows the raw data from each study with OAG prevalence plotted on a logarithmic scale against age. Figure 2also shows that prevalence increases proportionately with age for each racial group (i.e., prevalence increases exponentially with age) and that the prevalence of OAG is higher in black populations at all ages but the increase in prevalence (or odds) of OAG is steeper for white populations; increases with age in black and Asian populations are similar. A more formal analysis is presented in Table 2 , in which interactions between age and race are included. A multiple-variable meta-regression analysis showed that in white populations the OR of OAG per decade is 2.05 (95% CrI, 1.91–2.18), indicating a doubling in OAG prevalence per decade, whereas in black populations the OR is 1.61 per decade (95% CrI, 1.53–1.70) and in Asian populations it is 1.57 per decade (95% CrI, 1.46–1.68). Inspection of the 95% CrI for the interaction terms between age and race excluded an OR of 1.0 and is evidence for different slopes with age between white and black populations and between white and Asian populations. Further inspection of the data showed two potential outlying studies in black populations (Fig. 2)with higher-than-average prevalence and relatively shallower gradient with age. ^{31 37} Exclusion of these two studies increased the OR with age to 1.86 per decade (95% CrI, 1.73–2.01). However, the 95% CrI for the posterior distribution for the difference in the age effect between white and black populations still excluded an OR of 1.0 and therefore suggested that the increase in prevalence with age was greater in white populations. After controlling for age, there was no evidence of any difference in the prevalence of OAG among black populations from America, Europe, West Indies, or Africa. The log-linear association with age was formally assessed by including a quadratic term in age in the model. There was no evidence of curvature, suggesting that the association between OAG prevalence and age is exponential. 

Prevalence estimates in studies published in the 1960s–70s were on average lower than in studies published in the 2000s. Even after allowing for differences in survey methods, age, and racial group, the OR was 0.46. For studies from the 1980s and the 1990s, the corresponding ORs were 0.48 and 0.93 (Table 2) . 

It was possible to determine whether routine visual fields were assessed and whether criteria for OAG case definition used IOP for all studies. Studies that relied on IOP to define OAG and that did not examine visual fields routinely (17 studies) had a 35% lower odds of OAG (crude OR, 0.65; 95% CrI, 0.33–1.27) compared with studies that did examine visual fields and did not rely on IOP to define OAG (20 studies). Typically, the former studies tested visual fields on various subgroups, which are biased by design and would be expected to miss cases of OAG. After adjusting for age, racial group, and year of publication, the difference was reversed (Table 2)but the 95% CrI are wide, indicating no material difference. In studies in which visual field tests were performed routinely (25 studies), the odds of OAG were similar in those that relied on IOP for OAG definition and those that did not (OR, 1.00; 95% CrI, 0.52–1.86). The subgroup of three studies in which visual field tests were not performed routinely and did not rely on IOP for OAG case definition appeared to have higher odds of OAG than the baseline group (Table 2) , but the 95% CrI was very wide and should be interpreted cautiously. 

A survey response rate was available for 39 studies, and rates were classified as three groups: 50% to 75%, more than 75% to 85%, and more than 85%. Survey response rate did not seem to influence the prevalence of OAG; ORs comparing the three groups were approximately 1.00, and 95% CrIs were wide. In addition, there was no evidence of difference in OAG prevalence between studies that reported a response rate compared with those that did not. 

Published estimates by gender were available for 25 studies with 61,267 participants and 1355 cases of OAG. Overall the prevalence of OAG in men was 1.37 times that in women (95% CrI, 1.22–1.53) after adjusting for age, racial group, publication year, and survey methods. On stratification by racial group, the OR of OAG in men compared with women was 1.46 in white populations (95% CrI, 1.24–1.73; 31,647 persons with 576 cases of OAG), 1.27 in black populations (95% CrI, 1.05–1.55; 6177 persons with 455 cases of OAG), and 1.36 in Asian populations (95% CrI, 1.05–1.76; 23,443 persons with 324 cases of OAG). In studies that reported by gender, the associations with age, racial group, survey methods, and year of publication were similar to those reported in all studies. 

Discussion

This systematic review quantifies the rate of change in the prevalence of OAG with age for different racial groups. It is based on a large number of participants (103,567 persons, including 2509 with OAG) from a diversity of geographic populations. Although the average prevalence at all ages was higher in black populations than in white or Asian populations, the rate of change of OAG prevalence with age appears highest in white populations, in whom the odds of OAG double per decade. The equivalent proportionate increase in black or Asian populations was lower at approximately 1.6, equivalent to a 60% increase per decade. The reasons for these racial differences in disease trends with age remain unclear. 

Given that the methods of examination have changed markedly since the 1960s, it is not surprising to find an increase in OAG prevalence even with adjustment for age. This may reflect a true increase in prevalence or improved diagnostic criteria and examination techniques. The prevalence of glaucoma was lower in studies in which routine visual field tests were not performed and in which IOP was relied on as a defining criterion for glaucoma compared with studies in which visual field tests were performed routinely and IOP did not determine the diagnosis for OAG. There was no clear effect after allowing for age and racial group, which is consistent with findings predominantly in white populations, published in a previous review. ¹⁴  

Whether a gender difference exists in the prevalence of POAG has been controversial. Numerous studies have reported that the prevalence of OAG is higher in men, ^{1 10 12 48} others have reported a higher prevalence in women, ^{6 31} and yet others have reported no gender difference in prevalence. ^{2 3 4 5 9 11 49 51} Heterogeneity in the reporting of these gender effects may be explained by lack of power, with insufficient numbers of OAG cases within individual studies to detect consistent gender differences together with appropriate adjustment for age. If, for example, the true prevalence of OAG is 1.5% in women and 2.2% in men within a given age group (based on the findings from the present study), the required sample size for an individual study would have to be approximately 16,500 to detect a statistically significant difference at the 5% level (i.e., P = 0.05) and 90% power. There is no single study of this size. Pooling data from different studies allowed the gender effect to be determined with greater statistical certainty. Data from 61,267 subjects (including 1355 OAG cases) suggests that OAG prevalence in men is approximately 1.4 times that in women (adjusted for age, racial group, publication year, and survey methods); this effect was remarkably consistent across the three racial groups. This finding disagrees with a previous meta-analysis ¹⁴ that did not find a difference in the prevalence of OAG between men and women after controlling for age in white, black, or Hispanic populations. ¹⁴ However, this pooled analysis was based on one study in a black population (n = 2395) and one study in an Hispanic population (n = 4773) and was probably underpowered to detect gender difference in these two racial groups. Although the analysis was probably sufficiently powered to answer this question for the six studies in white populations (n = 22,556), investigators had access to additional unpublished data for these studies by gender that were not in the original study publications and were therefore able to perform an individual patient data meta-analysis. Only two of the studies included in that review had published data by gender; therefore, the other 4 studies were not included in our gender analyses. Differences can arise between meta-analyses that are performed using published estimates and those based on individual patient data because of differences in the accuracy or structure of the data, analytical approach, and studies included. An individual level meta-analysis allows a more flexible approach to the analyses and is preferable if raw data can be obtained for all relevant studies. If, however, raw data can only be obtained for a limited number of all relevant studies identified, the review may not be representative and the results may be biased. In the current international review, it is likely that it would not have been possible to obtain the raw data for a considerable proportion of the studies, especially as some were conducted many years ago. We therefore chose to undertake a meta-analysis of published data and estimates. The estimated OR in white populations for males compared with females from the previous review was 0.97 (95% CI, 0.79–1.20). We can use this result from the six studies in white populations as a “Bayesian prior” and combine it with our review data (excluding the two studies from our data to avoid double counting). The OR for men compared with women then becomes 1.23 (95% CrI, 1.06–1.42), indicating that OAG prevalence is estimated to be 23% higher in men than in women. This is the best available estimate, and it includes their data ¹⁴ and our review data appropriately combined, and the estimate is consistent with a recent very large study ⁵² looking at the prevalence of glaucoma treatment in the United Kingdom, where the OR of treatment for glaucoma in men compared with women was 1.24 (95% CI, 1.19–1.28). ⁵² A biological hypothesis for a gender difference in prevalence remains unclear and may reflect differences in environmental exposure between men and women; perhaps similar parallels can be drawn with well-known gender differences in other noncommunicable diseases such as cardiovascular disease. 

The use of Bayesian methods in this review is novel and allowed for the full hierarchical structure of the data to be accounted for using a multilevel approach. In addition, the precision of the variability within and across studies is taken into account and is reflected in the final estimates and 95% CrI. 

The drawback of any observational study examining the prevalence of disease is nonresponse. The participation rate ranged from 50% to 94%, but this did not appear to influence the prevalence of OAG. It is uncertain whether nonparticipants were more likely to represent those with OAG or those without OAG, who do not perceive a need for examination. However, a common finding in these glaucoma surveys was that approximately half of glaucoma cases were previously undiagnosed. The possibility of response rate bias is important because meta-analysis of surveys affected by a consistent bias will always be subject to that bias. 

Table 3summarizes the estimated prevalence of OAG by age and racial group. These estimates were similar to gender-specific estimates from a smaller number of studies (25 studies), though the prevalence of OAG was higher in men than in women at all age groups (data not presented). Table 3shows that the difference in prevalence of OAG between racial groups varied with age. In 40- to 49-year-olds, the prevalence of OAG in black populations was approximately 7 times higher than the prevalence in white populations, whereas by age 80 to 89 years the prevalence was approximately 2.5 times higher. The prevalence in Asian populations was similar to the prevalence in white populations at ages 40 to 69; thereafter, it was relatively higher in white populations. These estimates are shown graphically by the straight line in Figure 2 . 

Despite the caveats associated with the derivation of these estimates, a weighted average in those older than 70 years gives a pooled prevalence of OAG of approximately 6% in white populations, 16% in black populations, and 3% in Asian populations. These figures represent a considerable public health burden of OAG, which is likely to increase with time because of an aging population. OAG leads to irreversible visual field loss that negatively affects a person’s mobility and independence. ⁵³  

Conclusions

This review summarizes quantitatively the international literature on the effect of age, racial group, survey methods, year of publication, and gender on the prevalence of OAG. OAG prevalence increases more rapidly in white than in black and Asian populations, but at all ages black populations have the highest prevalence estimates. OAG is more common in men than in women across all racial groups. Understanding these differences in disease prevalence may give further clues to mechanisms and types of glaucoma and a sound basis for future health service provision. 

Appendix 1

Statistical Appendix

Bayesian Meta-Regression Model

All Bayesian models were fitted using WinBugs version 1.4 (MRC Biostatistics Unit, Cambridge, UK). 

Let r _kj be the number of cases of OAG in study k, age category j.  

\[r_{\mathrm{kj}}{\sim}\mathrm{Bin}\ (p_{\mathrm{kj}},\ n_{\mathrm{kj}})\]

 

\[\mathrm{logit}\ (p_{\mathrm{kj}}){=}{\theta}_{\mathrm{kj}}\]

Where p _kj is the prevalence of OAG in study k, age category j.  

\[{\theta}_{\mathrm{kj}}{=}{\alpha}_{\mathrm{k}}{+}{\beta}X_{\mathrm{kj}}{+}{\Omega}Z_{\mathrm{k}}{+}{\psi}V_{\mathrm{k}}{+}{\delta}_{1}X_{\mathrm{kj}}Z_{\mathrm{k}}{+}{\delta}_{2}X_{\mathrm{kj}}V_{\mathrm{k}}{+}{\Sigma}{\gamma}_{\mathrm{k}}{\theta}_{\mathrm{k}}\]

where X _kj is the mean age in the jth age category of the kth study (centered to age 40 years); Z _k is coded 1 if racial group is black for kth study 0, otherwise; V _k is coded 1 if racial group is Asian for kth study 0, otherwise; β is the increase in log odds of OAG with age in white populations; Ω is the difference in log odds of OAG at 40 years of age between studies in black and white populations; ψ is the difference in log odds of OAG at 40 years of age between studies in Asian and white populations; δ₁ is the difference in the slope with age between studies in black and white populations; δ₂ is the difference in the slope with age between studies in Asian and white populations; γ_k are the associated effects with three other study level covariates, θ_k: whether routine visual fields were scheduled for all participants, whether IOP was used in the case definition of glaucoma, and year of publication. 

Prior Distribution at Level 1: Within Study

 

\[{\alpha}_{\mathrm{k}}{\sim}N(\mathrm{A},{\tau})\]

where α_k is the log odds of OAG in the kth study (thereby retaining the within-study structure in the model) at 40 years of age. 

Prior Distributions at Level 2: Across Studies

 

\[A{\sim}N\ (0,\ 1{\times}10^{{-}5})\]

where A is the mean log(odds) of OAG at 40 years.  

\[{\tau}{\sim}{\Gamma}\ (0.01,\ 0.01)\]

where 1/τ is the between-study variance in log odds of OAG at age 40 modeled in terms of precision in WinBugs.  

\[{\beta}{\sim}N\ (0,\ 1{\times}10^{{-}5})\]

 

\[{\Omega}{\sim}N\ (0,\ 1{\times}10^{{-}5})\]

 

\[{\psi}{\sim}N\ (0,\ 1{\times}10^{{-}5})\]

 

\[{\delta}_{1}{\sim}N\ (0,\ 1{\times}10^{{-}5})\]

 

\[{\delta}_{2}{\sim}N\ (0,\ 1{\times}10^{{-}5})\]

 

\[{\gamma}_{\mathrm{k}}{\sim}N\ (0,\ 1{\times}10^{{-}5})\]

 

 Supported by the Department of Health, National Health Service Executive London Regional Office Grant RDC01931.

 Submitted for publication March 20, 2006; revised May 18, 2006; accepted August 18, 2006.

 Disclosure: A.R. Rudnicka, None; S. Mt-Isa, None; C.G. Owen, None; D.G. Cook, None; D. Ashby, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Prevalence Surveys Included in the Review

Race or Ethnicity	Author	Year	Name/Location	Age (y)	Sample Size	OAG Cases
Asian	Alsbirk 17	1973	Greenland	40+	396	5
	Arkell 18	1987	Kotzebue, Alaska, USA	15–70+	1686	1
	Hu 19	1989	Shunyi, Beijing	40+	3000	1
	Shiose 11	1991	Japan	30–70+	8924	224
	Rauf 20	1994	Southall, London, UK	30–80+	184	5
	Foster 21	1996	Hovsgol, Mongolia	40–89	942	5
	Jacob 22	1998	Vellore, India	30–60	972	4
	Foster 23	2000	Singapore	40–81	1232	22
	Dandona 24	2000	Andhra Pradesh, India	30–102	1399	27
	Metheetrairut 25	2002	Bangkok, Thailand	60–104	2092	73
	Bourne 26	2003	Rom Klao, Bangkok, Thailand	50–70+	790	16
	Ramakrishnan 27	2003	Aravind, India	40–90	5150	64
	Iwase 28	2004	Tajimi, Japan	40–80+	3021	119
	Rahman 29	2004	Dhaka, Bangladesh	35–85	2347	29
Black	Wallace 30	1969	Jamaica, West Indies	35–74	574	6
	Mason 31	1989	St. Lucia, West Indies	30–70+	1679	147
	Tielsch 2	1991	Baltimore, MD, USA	40–80+	2395	100
	Wormald 32	1994	London, UK	35–60+	873	32
	Leske 12	1994	Barbados, West Indies	40–86	4498	308
	Buhrmann 33	2000	Kongwa, East Africa	40–80+	3247	100
	Rotchford 34	2002	Kwazulu-Natal, South Africa	40–80+	1005	28
	Ekwerekwu 35	2002	Alum-Inyi, Nigeria	30–80+	664	14
	Rotchford 36	2003	Temba, South Africa	40–97	839	31
	Ntim-Amponsah 37	2004	Ghana, West Africa	30–100	1785	149
Hispanic	Quigley 38	2001	Proyecto VER, Nogales and Tucson, AZ	41–90+	4774	94
	Anton 39	2004	Segovia, Spain	40–79	510	10
Mixed	Salmon 40	1993	Mamre, South Africa	40–70+	987	15
White	Hollows 4	1966	Ferndale, Wales	40–74	4231	20
	Bankes 41	1968	Bedford, UK	20–80+	5941	45
	Leibowitz 1	1980	Framingham, MA, USA	<65–75+	2631	50
	Bengtsson 6	1981	Dalby, Sweden	58.5–68.5	1511	13
	Martinez 42	1982	Gisborne, New Zealand	65–90+	481	20
	Gibson 43	1985	Melton Mowbray, UK	76–85+	484	32
	Tielsch 2	1991	Baltimore, MD, USA	40–80+	2913	32
	Ringvold 7	1991	Norway	65–89+	1871	63
	Klein 3	1992	Beaver Dam, WI, USA	43–75+	4926	104
	Coffey 5	1993	Roscommon, Ireland	50–80+	2186	41
	Dielemans 10	1994	Rotterdam, The Netherlands	55–75+	3062	34
	Leske 12	1994	Barbados, West Indies	40–86	133	1
	Giuffre 44	1995	Casteldaccia, Sicily	40–99	1062	13
	Hirvela 45	1995	Oulu, Finland	70–95	500	52
	Ekstrom 8	1996	Tierp, Sweden	65–74	760	29
	Mitchell 9	1996	Blue Mountains, Australia	49–80+	3654	87
	Cedrone 46	1997	Ponza, Italy	40–80+	1034	26
	Bonomi 47	1998	Egna-Neumarkt, Italy	40–80+	4297	84
	Reidy 48	1998	North London, UK	65–100	1547	47
	Wensor 49	1998	Melbourne, Australia	40–90+	3271	56
	Kozobolis 50	2000	Crete, Greece	40–80+	1107	31
Total					103,567	2,509

Figure 1.

 

Meta-analysis of prevalence of OAG stratified by racial group.

Figure 2.

 

Prevalence of glaucoma (on a logarithmic scale) against age for each study separately. Symbol size is inversely proportional to prevalence standard error. Straight line: predicted prevalence from Bayesian meta-regression adjusted for racial group, publication year, and survey methods.

Table 2.

 

Odds Ratios from Bayesian Meta-Regression Model Adjusted for All Factors Except Gender

Factor	No. Studies in Meta-Analysis	Adjusted OR (95% CrI)
Effect per decade increase in age by racial group
White	24	2.05 (1.91–2.18)
Black	10	1.61 (1.53–1.70)
Asian	14	1.57 (1.46–1.68)
Year of publication
2000-2005	15	1.00
1990-1999	20	0.93 (0.60–1.40)
1980-1989	5	0.48 (0.24–0.91)
1960-1979	4	0.46 (0.20–0.98)
Study design factors VF*/IOP, †
VF only	20	1.00
VF and IOP	5	1.00 (0.52–1.86)
IOP only	17	1.25 (0.75–2.13)
Neither	3	2.07 (0.90–4.67)
Men vs. women
Overall	25	1.37 (1.22–1.53)
White	16	1.46 (1.24–1.73)
Black	2	1.27 (1.05–1.55)
Asian	8	1.36 (1.05–1.75)

 Gender comparison was performed in a subset of 25 studies, and effects were adjusted for all factors listed in the table.

*  Were visual field tests performed routinely on all participants?

†  Was IOP used as a defining criterion for the case definition of OAG?

Table 3.

 

Estimated Prevalence According to Age and Race

Age Range (y)	Predicted Prevalence of OAG (95% CrI)
		White	Black	Asian
30–39	—	1.8 (1.2–2.7)	0.4 (0.3–0.6)
40–49	0.4 (0.3–0.6)	2.9 (1.9–4.4)	0.6 (0.4–1.0)
50–59	0.8 (0.5–1.2)	4.6 (3.1–6.8)	1.0 (0.6–1.6)
60–69	1.6 (1.1–2.5)	7.2 (4.9–10.6)	1.6 (1.0–2.4)
70–79	3.3 (2.2–4.9)	11.2 (7.6–16.1)	2.5 (1.6–3.8)
80–89	6.6 (4.4–9.7)	16.9 (11.7–23.8)	3.8 (2.3–5.9)
90–95	10.8 (7.2–15.8)	22.5 (15.7–31.2)	—