Attributable Risk Estimates for Cataract to Prioritize Medical and Public Health Action

Abstract

purpose. Cataract is the most common cause of blindness in the world. The purpose of this study was to estimate the population attributable risk associated with identified risk factors for cortical, nuclear, and posterior subcapsular (PSC) cataract in a representative sample of the Victorian population aged 40 years and older.

methods. Cluster, stratified sampling was used and participants were recruited through a household census. At locally established test sites, standardized clinical examinations were performed to assess cataract and personal interviews were conducted to quantify potential risk factors. Multivariate logistic regression was used to determine the independent risk factors associated with the three types of cataract, and the population attributable risk was calculated.

results. A total of 3271 (83% of eligible) of the urban residents and 1473 (92%) rural residents participated. The urban residents ranged in age from 40 to 98 years (mean, 59 years), and 1511 (46%) were men. The rural residents ranged in age from 40 to 103 years (mean, 60 years), and 701 (48%) were men. The overall prevalence of cortical cataract was 12.1% (95% CL 10.5, 13.8), nuclear cataract 12.6% (95% CL 9.61, 15.7), and PSC cataract 4.93% (95% CL 3.68, 6.17). Significant risk factors for cortical cataract included age, female gender, diabetes for greater than 5 years, gout for greater than 20 years, arthritis, myopia, average annual ocular UV-B exposure, and family history of cataract (parents or siblings). Significant risk factors for nuclear cataract included age, female gender, rural residence, age-related maculopathy, diabetes for greater than 5 years, smoker for greater than 30 years, and myopia. The significant risk factors for PSC cataract were age, rural residence, thiazide diuretic use, and myopia. Of the modifiable risk factors, ocular UV-B exposure explains 10% of the cortical cataract in the community, and cigarette smoking accounts for 17% of the nuclear cataract.

conclusions. Because of the near universal exposure to UV-B in the environment, ocular protection has one of the highest modifiable attributable risks for cortical cataract and would therefore be an ideal target for public health intervention. Quit smoking campaigns can be expanded to incorporate information about the excess cataract in the community associated with long-term smoking. Nonmodifiable risk factors such as age, gender, and long-term medication use have implications for the timely referral and treatment for those at higher risk of cataract.

Cataract is the most common cause of blindness in the world. ¹ Cataract surgery is the most commonly performed ophthalmic procedure in Australia, and the number of cataract extractions has increased at more than double the rate that would be expected simply by changing demographics. ² Although safe, effective surgery is available as a means of secondary prevention, strategies for primary prevention have the potential for saving health care dollars and improving the visual function and independence of the growing number of elderly people in the population. 

Reviews of the epidemiology of age-related cataract reveal both modifiable and nonmodifiable risk factors, such as age and gender. ^{3 4} Potential modifiable risk factors include UV light exposure, smoking, certain medications, and low antioxidant intake. Although a number of studies about the risk factors for cataract have been completed, we know of no studies that have reported the attributable risk of these various risk factors in the population and the implications for community health planning. There has been considerable discussion in the literature about the appropriate use of attributable risk estimates in public health. ^{5 6 7 8} One of the most frequent errors noted in the interpretation of the attributable risk estimate is to equate this fraction with the proportion of cases having a given risk factor. ⁹ Another feature of the attributable risk that is often misinterpreted is the inability to sum the estimates for the individual attributable risks; in fact, they may sum to greater than 1.0. ⁹ Attributable risk estimates are best used to prioritize medical and public health interventions on the basis of the magnitude of potential effect on the disease outcome in the community. ⁷  

The purpose of this study was to estimate the population attributable risk associated with previously identified risk factors for cortical, nuclear, and PSC cataract in a representative sample of the Victorian population aged 40 years and older. 

Methods

Details about the methodology for the Visual Impairment Project have been published previously. ¹⁰ Briefly, cluster-stratified sampling was used to randomly select nine urban areas and four rural areas to recruit residents aged 40 year and older. A household census was used to recruit eligible participants and to gather information about demographics. Eligible residents were then invited to attend a local examination site for a more detailed interview and examination. The study protocol was approved by the Royal Victorian Eye and Ear Hospital Human Research Ethics Committee and adheres to the Declaration of Helsinki for research involving human subjects. 

Lens opacities were graded clinically at the time of the examination and subsequently from photographs using the Wilmer cataract photograph-grading system. ¹¹ Cortical and posterior subcapsular (PSC) opacities were assessed on retroillumination. Cortical opacities were measured as the proportion (in 1/16) of pupil circumference occupied by opacity, whereas the height and width of the PSC opacities were recorded. Nuclear opacities were observed by slit lamp. For this analysis, cortical cataract was defined as ≥4/16 opacity, nuclear cataract was defined as opacity equal to or greater than Wilmer standard 2.0, and PSC cataract was defined as PSC opacity≥ 1 mm², independent of visual acuity. All people not defined as cases of cataract were used as controls for the analyses. Previous analyses have shown that modeling the definition of cataract as prevalent cataract, including as cases or excluding entirely previous cataract surgery, and considering different cutoff values for level of lens opacity does not alter the risk factors substantially. ¹²  

In the case of previous cataract surgery (n = 249, 3.8%), the treating ophthalmologist was contacted whenever possible to obtain details about the type of opacity present before cataract surgery. Prior cataract surgeries were classified as cataract cases. Bilateral congenital cataracts or cataracts secondary to intraocular inflammation or trauma were excluded from the analysis. Three cases of bilateral secondary cataract and eight cases of bilateral congenital cataract were excluded from the analyses. 

Photographs were graded separately by two research assistants and discrepancies were adjudicated by an independent reviewer. Except for missing photographs, the photogrades were used in the analyses. Photogrades were available for 4301 (91%) cortical assessments, 4147 (88%) nuclear assessments, and 4304 (91%) PSC assessments. Intraclass correlation coefficients to assess the agreement between the clinical and photogrades were 0.86 for cortical opacity, 0.89 for nuclear opacity, and 0.73 for PSC opacity. People were categorized according to the severity of the opacity in their worse eye. 

A standardized questionnaire was used to obtain information about education, employment, and ethnic background. Specific information was elicited on the occurrence, duration, and treatment of a number of medical conditions, including ocular trauma, arthritis, diabetes, gout, hypertension, and mental illness. Information about the use, dose, and duration of smoking, alcohol, analgesics, and steroids were collected. A food frequency questionnaire was used to determine current consumption of dietary sources of antioxidants and use of vitamin supplements. ¹³ Lifetime ocular UV-B exposure was determined for each individual from information about place of residence, time spent outdoors, and the use of ocular protective devices, including hats, spectacles, and sunglasses. ¹⁴  

Data were collected either by direct computer entry with a questionnaire programmed in Paradox (Borland International Inc., Scotts Valley, CA) with internal consistency checks or on self-coding forms. Open-ended responses were coded at a later time. Data that were entered on the self-coded forms were entered into a computer with double data entry and reconciliation of any inconsistencies. Data range and consistency checks were performed on the entire data set. SAS version 6.1 (SAS Institute, Cary, NC) was used for statistical analyses. 

Ninety-five percent confidence limits (CL) around the age-specific rates were calculated according to Cochran ¹⁵ to account for the effect of the cluster sampling. Ninety-five percent CLs around age-standardized rates were calculated according to Breslow and Day. ¹⁶ The strata-specific data were weighted according to the 1996 Australian Bureau of Statistics census data ¹⁷ to reflect the cataract prevalence in the entire Victorian population. 

Bivariate analyses with t-tests andχ ² were first used to evaluate risk factors for cataract. Any factors with P < 0.10 were then fitted in a backwards, stepwise, logistic regression model. For the final multivariate models, P < 0.05 was considered statistically significant. Design effect was assessed in two manners. First, cluster-specific models where all variables from the overall model were forced into models stratified by cluster were calculated. Second, the final model was run with cluster as a fixed effect. The design effect was assumed to be additive, and an adjustment made in the variance by adding the variance associated with the design effect before constructing the 95% CLs. The adjusted method of attributable risk estimation based on regression models was used to quantify the attributable risks for the three cataract types. ¹⁸  

Results

Study Population

A total of 3271 (83% of eligible) of the urban residents and 1473 (92%) rural residents participated. Nonparticipants differed from participants only in language spoken at home; they were more likely to speak a language other than English at home. ¹⁹ The participation rate for people who spoke Greek at home was 76% compared with 85% participation for people who spoke English at home. The study population is representative of both the Victorian population and the Australian population as a whole. The urban residents ranged in age from 40 to 98 years (mean, 59 years), and 1511 (46%) were men. The rural residents ranged in age from 40 to 103 years (mean, 60 years), and 701 (48%) were men. 

Cataract Prevalence and Risk Factors

The prevalence of all three types of cataract increases with age; however, the relative distribution of the three types of cataract changes with age (Fig. 1) . Cortical cataract is more common than nuclear cataract until 70 years of age. PSC is the least common type of cataract at ages greater than 50 years but is the most prevalent cataract in the age group 40 to 49 years. PSC cataract has been shown to have the greatest impact on vision. The overall prevalence of cortical cataract was 12.1% (95% CL 10.5, 13.8), nuclear cataract 12.6% (95% CL 9.61, 15.7), and PSC cataract 4.93% (95% CL 3.68, 6.17). A total of 1057 (23% of the study population) cases of cataract were identified. The distribution and overlap of cataract type was as follows: 342 (32%), pure cortical cataract; 344 (33%), pure nuclear cataract; 91 (8.6%), pure PSC cataract; 143 (14%), cortical and nuclear cataract combined; 57 (5.4%), nuclear and PSC cataract combined; 36 (3.4%), cortical and PSC cataract combined; and 44 (4.2%), all three types of cataract. 

The following potential risk factors were evaluated in bivariate analyses: age, gender, location, education, country of birth, parents’ country of birth, occupation, iris color, hypertension, diabetes, gout, diarrhea, aspirin use, paracetomol use, topical steroids, systemic steroids, inhaled steroids, allopurinol, hormone replacement therapy, arthritis, smoking history, alcohol use, mental illness, diazepams, lithium carbonate, calcium channel blocker, ACE inhibitors, α- andβ -blockers, thiazide diuretics, potassium-sparking diuretics, loop diuretics, family history of cataract (parents or siblings), myopia ≥ −1 diopters (D), age-related maculopathy, eye injury, body mass index, vitamin A intake and supplements, vitamin C intake and supplements, vitamin E intake and supplements, β-carotene intake and supplements, selenium intake and supplements, zinc intake and supplements, glaucoma, use of insulin, and ocular UV-B exposure. For the nutrient supplements, duration of use was also investigated. Continuous variables were evaluated as both continuous variables and categorical (ordinal) variables. 

Significant bivariate risk factors for cortical cataract included the following: lifetime ocular UV-B exposure, female gender, education level, occupation, hypertension, use of calcium channel blockers, use of ACE inhibitors, use of α-blockers, duration of β-blocker use, duration of thiazide diuretic use, use of loop diuretics, duration of diabetes, use of insulin, duration of gout, duration of allopurinol use, arthritis, duration of aspirin use, duration of acetaminophen use, duration of hormone replacement therapy, duration of diazepam use, cigarette smoker, alcohol use, family history of cataract, age-related maculopathy, higher body mass index, previous eye injury, and lower intakes of β-carotene or vitamin E. ¹² Significant bivariate risk factors for nuclear cataract included the following: lifetime ocular UV-B exposure, female gender, rural residence, education, country of birth, occupation, iris color, duration hypertension, duration of calcium channel blocker use, duration of ACE inhibitor use, use of α-blockers, years of β-blocker use, duration of thiazide diuretic use, use of loop diuretics, duration of diabetes, duration of gout, use of allopurinol, duration of arthritis, duration of aspirin use, duration of acetaminophen use, use of systemic or inhaled corticosteroids, duration of hormone replacement therapy, any treatment for mental condition, duration of diazepam use, cigarette smoking, alcohol drinking, myopia, glaucoma, age-related maculopathy, higher body mass index, prior eye injury, and lower vitamin C, vitamin E, or β-carotene consumption. ¹² Bivariate risk factors for PSC cataract included the following: lifetime ocular UV-B exposure, rural residence, country of birth, hypertension duration, use of calcium channel blocker, use of ACE inhibitor, use of β-blocker, use of thiazide diuretic, diabetes, gout, arthritis duration, use of aspirin, use of acetaminophen, family history of cataract, myopia, glaucoma, age-related maculopathy, alcohol use, and lower intakes of vitamin C, vitamin E, or β-carotene. ¹²  

These significant risk factors from the bivariate analyses were then evaluated in backwards, stepwise, logistic regression models. The significant risk factors from the multivariate models varied by cataract type (Tables 1 2 3) . Significant risk factors for cortical cataract included age, female gender, diabetes for greater than 5 years, gout for greater than 20 years, arthritis, myopia, average annual ocular UV-B exposure, and family history of cataract (parents or siblings). Significant risk factors for nuclear cataract included age, female gender, rural residence, age-related maculopathy, diabetes for greater than five years, smoker for greater than 30 years, and myopia. The significant risk factors for PSC cataract were age, rural residence, thiazide diuretic use, and myopia. The potential affect of study design was assessed by modeling the cluster and by conducting cluster-specific models but did not affect the estimates of the odds ratios or SEs. 

Population Attributable Risk

Population attributable risk for the individual risk factors identified in the multivariate, logistic regression models was calculated according to Bruzzi. ¹⁸ They are not additive because the odds ratios are derived from logistic regression. For all three types of cataract, the factor with the highest attributable risk was age (Tables 1 2 3) . For cortical and nuclear cataract, the next highest attributable risk was associated with gender. 

Conclusions

To our knowledge, this is the first study to report not only the risk factors for the three major types of age-related cataract but also the population attributable risk associated with these risk factors. Because of the sampling scheme, these results are representative of the Australian population as a whole and would likely be generalizable to other developed countries. Attributable risk estimates calculated from multiple logistic regression analyses have the advantage that they control for potential confounders. However, they are not additive, so it is not possible to get a reliable estimate of the total amount of cataract potentially preventable if the identified risk factors could be eliminated in the population. Rather, they are useful for priortizing interventions. 

The risk factors for the three types of cataract are similar to what has been reported previously, ^{3 4} with the exception that we did not find steroid use to be associated with PSC cataract, even at the bivariate level. The primary difference with the present study is that sampling strategy resulted in a study sample that is representative of the entire population, and therefore a meaningful estimate of population attributable risk can be made. If the prevalence of a risk factor in the population is very low (as we found for the use of steroids), the population attributable risk will also be low. 

The risk factors for cataract identified can be divided into those that are modifiable and those that are not modifiable. The modifiable risk factors are potentially amenable to change through public health interventions, whereas some of the nonmodifiable risk factors have implications for timely detection and treatment by health care practitioners. The majority of risk factors for all three types of cataract are nonmodifiable. 

For all three types of cataract, age had the greatest attributable risk. For cortical and nuclear cataract, female gender had the second highest attributable risk. The third highest attributable risk for cortical cataract was family history of cataract. Although all none of these three risk factors are amenable to primary intervention, they do have implications for secondary prevention. People at higher risk of developing cataract should be advised to be alert for changes in their vision associated with the development of cataract. Furthermore, health care providers should be informed about the high-risk groups so that they can refer these patients for timely treatment of cataract before the onset of severe functional impairment. Systemic diseases (such as diabetes, gout, and arthritis) or medication for systemic diseases (thiazide diuretic use) were associated with cataract. Therefore, primary prevention of the systemic diseases would also serve as primary prevention for cataract, although the attributable risks were quite low, with the exception of arthritis and cortical cataract (PAR = 11%). 

Although we found that myopia was associated with all three types of cataract, we do not think that this association was causal. We have previously reported in this population that a myopic shift occurs in the presence of increasing nuclear opacity. ²⁰ The myopic shift with increasing opacity, especially nuclear sclerosis, has long been recognized. 

The largest attributable risks for cataract were documented for ocular UV-B exposure and cortical cataract (PAR = 10%), and long-term smoking and nuclear cataract (PAR = 17%). The large percentage of the population with these exposures contributes to their high attributable risks. These data provide further support for public health campaigns to decrease the prevalence of smoking in the community and to increase ocular protection from UV-B by avoidance of the midday sun and use of shade, sunglasses, and brimmed hats. These factors are especially important because they are the most obvious targets for primary public health prevention strategies of the identified risk factors for the three types of cataract. 

In conclusion, knowledge of attributable risk as well as odds ratios associated with risk factors for cataract will allow informed decision making to prioritize primary and secondary prevention strategies for cataract in the community. 

 Supported in part by grants from the National Health and Medical Research Council, the Victorian Health Promotion Foundation, the Jack Brockhoff Foundation, the Ansell Ophthalmology Foundation, and the Estate of the late Dorothy Edols. Rural field work was supported in part by the APEX and Rotary Clubs of Bright, Victoria. CAM is the recipient of the Wagstaff Research Fellowship in Ophthalmology from the Royal Victorian Eye and Ear Hospital.

 Submitted for publication September 17, 2000; revised July 10, 2000; accepted July 25, 2000.

 Commercial relationships policy: N.

 

Figure 1.

 

Age-specific rates of cortical, nuclear, and PSC cataract (including prior surgery).

Table 1.

 

Population Attributable Risk for Risk Factors Associated with Cortical Cataract in People Aged ≥40 Years

Risk Factor	Cases*	Controls*	OR, †	PAR, † , ‡
Age (years)
40–59	53 (12)	2230 (63)	1.00
60–69	144 (33)	828 (24)	6.74 (4.84, 9.38)
70+	239 (55)	464 (13)	20.0 (14.5, 27.7)	0.80 (0.77, 0.83)
Gender
Male	174 (40)	1636 (46)	1.00
Female	262 (60)	1886 (54)	1.53 (1.20, 27.7)	0.21 (0.18, 0.24)
Diabetes status
No diabetes or duration ≤ 5 years	406 (93)	3455 (98)	1.00
Diabetes for >5 years	30 (6.9)	67 (1.9)	2.65 (1.61, 4.38)	0.04 (0.03, 0.05)
Gout status
No gout or duration ≤ 10 years	412 (95)	3436 (98)	1.00
Gout for >10 years	24 (5.5)	86 (2.4)	1.72 (1.02, 2.92)	0.02 (0.015, 0.03)
Arthritis status
No arthritis	254 (58)	2755 (78)	1.00
Arthritis	182 (42)	767 (22)	1.38 (1.10, 1.74)	0.11 (0.10, 0.13)
Refractive status
Not myopic or myopia ≤ 1 D	376 (86)	3071 (87)	1.00
Myopia > 1 D	60 (14)	451 (13)	1.55 (1.11, 2.15)	0.05 (0.035, 0.06)
Average annual ocular UV-B exposure
≤0.02 Melbourne sun-years	311 (71)	2782 (79)	1.00
>0.02 Melbourne sun-years	125 (29)	740 (21)	1.55 (1.19, 2.01)	0.10 (0.085, 0.12)
Family history
No cataract	300 (69)	2790 (79)	1.00
Cataract	136 (31)	732 (21)	1.82 (1.43, 2.33)	0.14 (0.12, 0.16)

*  Values in parentheses are percentages.

†  Values in parentheses are 95% confidence limits.

‡  PAR was derived from multiple logistic regression, and therefore is not additive.

Table 2.

 

Population Attributable Risk for Risk Factors Associated with Nuclear Cataract in People Aged ≥40 Years

Risk Factor	Cases*	Controls*	OR, †	PAR, † , ‡
Age (years)
40–59	21 (4.1)	2489 (62)	1.00
60–69	127 (25)	1004 (25)	14.8 (9.21, 23.8)
70+	369 (71)	492 (12)	81.1 (51.1, 128.6)	0.93 (0.90, 0.96)
Gender
Male	218 (42)	1886 (47)	1.00
Female	299 (58)	2099 (53)	1.63 (1.30, 2.05)	0.25 (0.19, 0.25)
Residence
Urban	291 (56)	2849 (71)	1.00
Rural	226 (44)	1136 (29)	2.01 (1.60, 2.52)	0.22 (0.20, 0.24)
Age-related maculopathy status
No age-related maculopathy	364 (70)	3605 (90)	1.00
Age-related maculopathy	153 (30)	153 (10)	1.72 (1.33, 2.23)	0.12 (0.11, 0.13)
Diabetes status
No diabetes or duration ≤ 5 years	485 (94)	3896 (98)	1.00
Diabetes for >5 years	32 (6.2)	89 (2.2)	1.67 (1.02, 2.75)	0.02 (0.018, 0.03)
Smoking status
Non-smoker or smoke ≤ 30 years	330 (64)	3250 (82)	1.00
Smoke for >30 years	187 (36)	735 (18)	1.89 (1.48, 2.41)	0.17 (0.16, 0.18)
Refractive status
Not myopic or myopia ≤ 1 D	432 (84)	3495 (88)	1.00
Myopia > 1 D	85 (16)	490 (12)	2.59 (1.87, 3.58)	0.10 (0.09, 0.11)

*  Values in parentheses are percentages.

†  Values in parentheses are 95% confidence limits.

‡  PAR was derived from multiple logistic regression, and therefore is not additive.

Table 3.

 

Population Attributable Risk for Risk Factors Associated with PSC Cataract in People Aged ≥40 Years

Risk Factor	Cases*	Controls*	OR, †	PAR, † , ‡
Age (years)
40–59	57 (28)	2467 (57)	1.00
60–69	55 (27)	1088 (25)	2.32 (1.58, 3.40)
70+	92 (45)	790 (18)	5.11 (3.60, 7.25)	0.52 (0.48, 0.57)
Residence
Urban	123 (60)	3045 (70)	1.00
Rural	81 (40)	1300 (30)	1.53 (1.14, 2.06)	0.14 (0.12, 0.16)
Thiazide diuretic use
Nonuser	178 (87)	4114 (95)	1.00
User	26 (13)	231 (5)	2.07 (1.33, 3.24)	0.07 (0.05, 0.08)
Refractive status
Not myopic or myopia ≤ 1 D	154 (75)	3821 (88)	1.00
Myopia > 1 D	50 (25)	524 (12)	3.05 (2.16, 4.30)	0.16 (0.15, 0.18)

*  Values in parentheses are percentages.

†  Values in parentheses are 95% confidence limits.

‡  PAR was derived from multiple logistic regression, and therefore is not additive.

The authors thank the following field team members for their contributions: Simon Barty, Marie Bissinella, Caroline De Paola, MEd Psych, Cara Fu, Charles Guest, PhD, Sharon Lee, Patricia Livingston, PhD, Claire McKean, Yury Stanislavsky, PhD, Cathy Walker, Matthew Wensor, MSc, and Carolyn Wood.