Association of Demographic, Familial, Medical, and Ocular Factors With Intraocular Pressure

Objective  To describe the distribution and associations of demographic, familial, medical, and ocular factors with intraocular pressure (IOP).

Methods  A cluster stratified random sample of urban and rural residents of Victoria, Australia, aged 40 years and older. Participants completed an interview and underwent a standardized dilated ophthalmic examination including measurement of IOP with an electronic applanation tonometer (Tono-Pen). Glaucoma status(possible, probable, definite) was determined by a consensus panel. The main outcome measure was IOP.

Results  The mean age of the 4576 participants was 59 years, 53% were women, 32% were born overseas, and 132 had open-angle glaucoma. Geometric mean (SD) IOP was 14.3 (±1.5) mm Hg. The relationship between IOP and nuclear sclerosis, iris color, and family history of glaucoma depended on glaucoma status. In those with glaucoma, family history of glaucoma and country of birth were significantly associated with IOP in multivariate models (model: r² = 0.08, P = .01). In the group without glaucoma, place of residence, use of alcohol, iris color, vitamin E intake, and spherical equivalent were associated with IOP (model: r² = 0.01, P = .006).

Conclusion  In participants with glaucoma, genetic factors seem to be stronger predictors of IOP, whereas in those without glaucoma, lifestyle and physiological factors seem to play a greater role.

ALTHOUGH OCULAR hypertension has traditionally been associated with increased risk of open-angle glaucoma, not all who have elevated intraocular pressure (IOP) have or will develop the disease.^1,2 Furthermore, the role of IOP in inducing glaucomatous optic disc damage is controversial and not well understood.^3-5

Several studies have reported characteristics associated with IOP.^6-19 Many, but not all, studies report an increase in IOP with age. Other factors reported to be associated with IOP include ethnicity, sex, diabetes, hypertension, myopia, iris color, nuclear sclerosis, and family history of glaucoma. Few studies have evaluated the relationship of lifestyle factors such as smoking, alcohol use, and diet with IOP or the differences in determinants of IOP between those with and without glaucoma. Nor have many studies been conducted in ethnically diverse populations.

The Melbourne Visual Impairment Project is a large, ethnically diverse population-based study of the prevalence of and risk factors for age-related eye diseases in Victoria, Australia. We report the distribution of IOP in the population and evaluate its association with demographic, familial, medical, and ocular factors among people who have not been treated for glaucoma.

Details of the methods are reported elsewhere.²⁰ Briefly, the sample was drawn from 9 randomly selected clusters composed of adjacent pairs of census collector districts in Melbourne and from 4 in rural Victoria. A door-to-door private census identified all residents eligible for participation in the study. Eligibility requirements included age 40 years and older and residence of 6 months or longer at the current address. All identified eligible residents were invited to complete an interview and examination at a local examination center. Interpreters were used when participants did not speak English and home visits were conducted when participants were unable to attend the local examination center.

Participants completed a standardized questionnaire in the door-to-door census and at the local examination center. The interviews elicited data regarding the participant's sociodemographic characteristics, diet, medical history, medication use, history of eye disease, and current visual symptoms.

Visual field assessment was conducted using a Humphrey Field Analyzer with the 24-2 Fastpac statistical package (Humphrey Instruments Inc, San Leandro, Calif). A Bjerrum tangent screen visual field was performed when Humphrey visual fields were unobtainable. When this was unobtainable, a confrontation field was done.

Intraocular pressure was measured after the instillation of benoxinate hydrochloride (0.4%) in each eye. The measurement was taken with an electronic applanation tonometer (Tono-pen; Oculab, Gendale, Calif) and repeated if the result was 21 mm Hg or greater. Intraocular pressure measurements confirmed greater than 21 mm Hg were checked with a Goldmann applanation tonometer. All IOP measurements were made after 10 AM.

Dilated fundus examination included measurement of vertical cup-disc ratio and photography with a retinal camera (Topcon TRC 50EX; Topcon Corporation, Tokyo, Japan). Paired stereoscopic photographs of the optic disc and macula were taken and subsequently assessed through stereo viewers. Vertical cup-disc ratios were measured and recorded in addition to any abnormalities of the choroid and retina.

Diagnosis of glaucoma was by a consensus panel of 6 ophthalmologists, 2 of whom were glaucoma subspecialists.²¹ Records of participants considered as having glaucoma were submitted to the panel for evaluation. Suspected glaucoma cases were classified as definite, probable, or possible open-angle glaucoma, ocular hypertensive, presumed ocular hypertensive, primary angle-closure glaucoma, secondary glaucoma, or no glaucoma present.

Two different definitions of myopia were evaluated: less than −1.0 diopter spherical equivalent and less than −0.5 diopter spherical equivalent. Iris color was categorized using standard photographs and nuclear opacity was classified by the Wilmer system.^20,22

SAS statistical software version 6.10 (SAS Institute, Cary, NC) was used in the data analyses. Variables considered in the analysis are shown in Table 1. The natural log of data for vitamin intake, nuclear opacity, and IOP (right eye) were used in the analyses as this transformation yielded an an approximatley normal distribution for these variables. Means reported for IOP are geometric. Data for right eyes were used in analyses of eye-specific factors. Correlation analysis and analysis of variance with Scheffé multiple comparison were used to evaluate univariate statistical relationships with IOP. Interaction terms are the product of the factors under investigation. General linear modeling was used for multivariate analysis with a stepwise approach to model construction. All variables with a univariate P value less than.15 were included in the full models. Variables with a P value less than .05 were considered statistically significant and remained in the final models. Participants with a history of glaucoma treatment(either surgery or medication use), angle-closure glaucoma, or secondary glaucoma were excluded from the analysis. Participants with possible, probable, or definite glaucoma were included in the group considered to have open-angle glaucoma.

A total of 4744 (88% of eligible) people participated in the study, 3271 from the urban cohort and 1473 from the rural cohort. A total of 4576 people were included in the analyses. Seventy-six people were excluded from the analyses because they had been treated for glaucoma with medication, surgery, or both. An additional 91 people were excluded from the analyses because data for IOP were missing and 1 was excluded because of presence of primary angle-closure glaucoma. The mean age was 59 years, 53% were women, and 32% were born overseas(13.4% from Italy or Greece, 8.7% from the British Isles, and 9.9% were born elsewhere). A total of 132 participants were classified as having open-angle glaucoma: 58 possible, 24 probable, and 50 definite.

The distribution of IOP was somewhat skewed to the right (Figure 1). Geometric mean (SD) IOP was 14.3 (1.5) mm Hg for the entire study population and differed significantly between those with and without glaucoma. Among those with glaucoma, geometric mean IOP was 17.9 mm Hg (1.3) mm Hg and among those without glaucoma was 14.2 mm Hg (1.2) (F = 70.93, P<.001). Geometric mean IOP did not vary significantly between the 3 levels of glaucoma classification. A total of 2.8% of the participants had IOP greater than 21 mm Hg, 29.9% of whom had glaucoma. In those with glaucoma, 28.8% had an IOP greater than 21 mm Hg compared with 1.5% in those without glaucoma (χ²₁ = 434, P<.001).

All factors considered in the analysis were first analyzed for statistical interaction between the glaucoma status of the participant and the factor. The relationship between IOP and nuclear sclerosis, smoking status, iris color, and family history of glaucoma depended on the participant's glaucoma status. In the separate univariate analyses for those with and without glaucoma, neither smoking nor nuclear sclerosis were independently associated with IOP at the P<.05 level in either group. The relationship between IOP and family history of glaucoma was statistically significant only in the group with glaucoma. Iris color was independently related to IOP in both groups, but the magnitude of difference in geometric mean IOP between light- to dark-colored eyes was greater in the group without glaucoma (Table 2 and Table 3).

In multivariate analyses for the group without glaucoma, residential location, previous but not current alcohol use, iris color, spherical equivalent of the eye, and vitamin E intake remained significantly associated with IOP(Table 4). Compared with urban residents, geometric mean IOP was higher among rural residents. Intraocular pressure increased with increasing pigmentation of the iris. Both spherical equivalent of the eye and vitamin E intake were inversely associated with IOP. Current alcohol users had somewhat higher geometric mean IOP compared with those who had stopped using alcohol. The final model explained only 1.1% of the variation in IOP for the group without glaucoma.

The final model for the group with glaucoma explained 7.8% of the variation in IOP for this group. Factors remaining significant in this model included family history of glaucoma and country of birth. Family history of glaucoma and Mediterranean country of birth were both associated with increased IOP(Table 4).

Body mass index data were available for a subset of 3696 participants without glaucoma and were independently associated with IOP (r = 0.11, P<.001). Once data for body mass index were entered into the model, neither alcohol use nor vitamin E intake retained statistical significance. Together with body mass index, residential location, iris color, and spherical equivalent of the eye explained 1.9% of the variation in IOP in the group without glaucoma. Data for body mass index were available for 99 participants with glaucoma. Body mass index did not reach statistical significance when added to the final model for the group with glaucoma (F = 1.05, P = .31).

The results of this study indicate that the relationship between IOP and other factors is dependent on the glaucoma status of the individual. Family history of glaucoma and country of birth were significantly associated with IOP in multivariate models for the group with untreated glaucoma. In contrast, among the group without glaucoma, use of alcohol, iris color, vitamin E intake, and spherical equivalent are associated with IOP. The final model for the group with glaucoma explains a much greater proportion (7.8%) of the variation in IOP than the final model for the group without glaucoma (1.1%). A slightly greater proportion of the variation in IOP is explained when body mass index is added to the model for the group without glaucoma (1.9%) for the subset in whom body mass index data were available. However, the the power of the factors investigated to explain variation in IOP is low for both those with and without glaucoma. Thus, other factors must be evaluated to better understand the variation in IOP.

Of concern in this study is the potential bias introduced into the results due to the criteria we used to select participants. Specifically, we excluded all participants who had a history of glaucoma surgery or who used medications to lower IOP. This assumes that participants had equal access to health care and hence equal likelihood of having diagnosed glaucoma and consequent treatment to lower IOP. Our finding that Mediterranean country of birth was related to increased IOP among participants with glaucoma raises the question of whether this association is due to a difference in glaucoma detection and treatment, which led to selection into the analysis a group with more severe glaucoma and higher IOP. We find no evidence that ethnicity was related to undiagnosed glaucoma and thus a greater likelihood of more advanced glaucoma with higher IOP. Among Mediterranean participants with glaucoma, 58% were undiagnosed cases compared with 68% of those who were Australian born. Furthermore, 90% of Mediterranean-born participants with diagnosed glaucoma reported either glaucoma surgery or treatment to lower IOP compared with 73% of those who were Australian born. No statistical differences in glaucoma diagnosis or glaucoma treatment were noted between participants from different countries of birth. We also found no statistical difference in rates of glaucoma diagnosis or treatment for other demographic variables used in these analyses. Therefore, it seems unlikely that this study is affected by selection bias, given that diagnosis and treatment rates are comparable.

The distribution of IOP in this study is similar to that reported for white populations in other studies using similar methods, although geometric mean IOP is somewhat lower when compared with the total population or with those with and without glaucoma (Table 5).^{5,9,14,19,23,24} This may be an artifact of the method used for calculating means. We report geometric means, which tends to lower the mean by reducing the weight of values in the extreme high range of the distribution. Among those with glaucoma, the proportion with IOP greater than 21 mm Hg is similar to that reported in the Egna-Neumarkt²³ and the Beaver Dam Eye studies.¹⁴ Similar to results from the Egna-Neumarkt Study, the distribution of IOP tends to be skewed to the right and this distribution is not entirely accounted for by glaucoma cases.

Results of previous studies evaluating difference in mean IOP between men and women are inconsistent.^{6,9,11,13,14,19,23,24} We find no significant difference in geometric mean IOP between men and women, controlling for age. In contrast to many,^{7,11,13,23,24} but not all,^9,14,16,17 we find that IOP decreases with increasing age, although the difference is less than 1 mm Hg between the geometric mean for the 40- to 49-year-old group compared with those older than 90 years, for the group without glaucoma. Furthermore, the association of age with IOP for either the group with or without glaucoma is not significant in multivariate models.

Numerous studies have reported a positive association between IOP and systolic blood pressure.^{6,7,9,13,14,19} We were unable to directly assess this relationship, but did find that among those without glaucoma, those with current treatment of self-reported hypertension had lower IOP compared with those who are not currently treated. Also, those with a history of cardiovascular disease (defined by current medication use) had significantly lower IOP in univariate analyses than those without a history of cardiovascular disease or hypertension. However, neither of these differences attained significance in multivariate models for either the group without or with glaucoma. It is interesting to note that increasing body mass index, which is associated with hypertension, is correlated with increasing IOP for the group without glaucoma but not for those with glaucoma. Body mass index was associated with IOP in both the Barbados Eye Study¹⁹ and Beaver Dam Eye Study.¹⁴

Although not unprecedented,¹¹ the relationship between IOP and iris color is difficult to interpret. Among those without glaucoma, those with brown irides had, on average, significantly higher IOP than those with blue or gray irides, albeit the difference was less than 1 mm Hg. However, the trend was not strictly linear, with increasing IOP as pigmentation increased. Those with green or light brown irides had, on average, lower IOP than did those with blue, grey, or green irides. This pattern was also noted in the group with glaucoma, although it was not statistically significant in either the univariate or multivariate models. More participants of Mediterranean(59%) and other ethnicity (53%) had brown irides than those of Australian or British descent (12%). Subsequent analysis, which included interaction of ethnicity with iris color, did not demonstrate dependence of the relationship between iris color and IOP on ethnicity for either the group with or without glaucoma. Hence, it seems unlikely that the relationship of iris color with IOP is confounded by ethnicity.

The decrease in IOP with increasing vitamin E intake among those without glaucoma may indicate an antioxidant effect. We were unable to locate other studies that included analyses of the relationship between IOP and dietary factors. However, a recent review of biological mechanisms of glaucoma suggests that oxygen free radicals may play a toxic role in induction of retinal ischemia and neural degradation.²⁵ The relationship we note between IOP and vitamin E intake may reflect this to the extent that IOP and retinal health are related.

In those without glaucoma, previous but not current use of alcohol was related to lower IOP compared with those who currently use or have never used alcohol. The lack of a dose-response effect suggests that this association acts as a surrogate for some other lifestyle factor that we are unable to identify. However, neither vitamin E intake nor alcohol consumption retain statistical significance once body mass index is entered into the models. Body mass index had a weak but statistically significant negative correlation with vitamin E intake (r = -0.05), but the relationship between IOP and vitamin E was not found to depend on body mass index in subanalyses for the group without glaucoma. Body mass index was, on average, 26 kg/m² for those who currently use alcohol, have used alcohol in the past, or have never used alcohol, and alcohol use was not related to vitamin E intake. These analyses tend to support that alcohol consumption, vitamin E intake, or both act as a surrogate for dietary or lifestyle factors. We are unable to evaluate other lifestyle factors, such as exercise, which may shed further light on the relationships noted in this study.

The higher average IOP among rural residents without glaucoma may be related to other medical factors. A significantly higher proportion of rural participants reported untreated hypertension than in urban areas and body mass index was marginally higher, on average.

It is postulated that refractive error is related to IOP by influencing the shape of the eye and subjecting it to greater stress as the spherical equivalent decreases.⁸ However, we were unable to demonstrate that geometric mean IOP was significantly different by either definition of myopia used in this study. We find some evidence that in those without glaucoma there is a tendency for IOP to increase as the spherical equivalent decreases. This finding is comparable to results from the Beaver Dam Eye Study in which refractive error explained 0.6% of the variation in IOP in multivariate models.¹⁴

Both a history of diabetes and increasing glycolated hemoglobin level have been reported to be associated with increased IOP.^14,18,19 We find little support for the association of diabetes with IOP in this study. Although not statistically significant, IOP was higher among diabetic persons but glycosylated hemoglobin level was not correlated with IOP.

Among those with glaucoma, both family history of glaucoma and Mediterranean country of birth suggest genetic predisposition to higher IOP. Both findings are supported by other studies.^5,14,19 In the Barbados Eye Study, those with a family history of glaucoma had a geometric mean IOP of 18.4 mm Hg compared with 17.8 mm Hg in those without a positive family history.¹⁹ Studies in the United States report racial differences in IOP.^5,11

This study differs from many others in the way that we stratified our analyses. We chose to test statistically whether or not the relationship of IOP to factors hypothesized to influence IOP was dependent on the glaucoma status of the participants. This in part may explain the difference in results between our study and those of the Barbados Eye Study and the Beaver Dam Study—the 2 studies using statistical approaches most similar to ours. In the Barbados Eye Study, only participants with no history of glaucoma or glaucoma treatment were included in analyses, whereas in the Beaver Dam Study, all participants were included in analyses. Both studies are able to explain approximately 10% of the variation in IOP with their analyses. In contrast, we are able to explain less than 2% of the variation of IOP in the group without glaucoma. This may in part be explained by the nature of the variables available in our study to explain IOP. In the Beaver Dam Eye Study, 3.9% of the total variation was explained by systolic blood pressure alone. We were unable to include many of the physiological factors included in both of these studies.

When we model the associations with IOP without stratifying by glaucoma status, we are able to explain 4.4% of the variation in IOP. In these models, family history of glaucoma and glaucoma status are significant and explain 1.2% and 3.2% of the variation in IOP, respectively. This suggests that when we ignore the dependence of IOP on glaucoma status, the model is weighted to the IOP of the glaucoma cases.

In conclusion, in individuals without glaucoma, IOP may be influenced by physiological and lifestyle factors. In comparison, genetic factors seem to be better predictors of IOP in those with glaucoma.

Accepted for publication October 18, 2000.

This study was supported in part by the National Health and Medical Research Council, Canberra, Australia; and the Victorian Health Promotion Foundation, the Estate of Dorothy Edols, the Ansell Ophthalmology Foundation, and the Jack Brockhoff Foundation, Melbourne, Australia.

Corresponding author and reprints: LeAnn M. Weih, PhD, 32 Gisborne St, East Melbourne, Victoria 3002, Australia (e-mail: leann@unimelb.edu.au).