Study Population
We carried out a frequency-matched case-control study of 343 cases and 334 controls to investigate potential risk factors for cataract which included years of outdoor exposure and intake of antioxidant micro- nutrients. Detailed information on the methodology and the nutritional results has been described previously [18].
The study was conducted from July 1994 to September 1995 in the city of Burjassot, on the Mediterranean east-coast of Spain, within the province of Valencia. All cases and controls were recruited among patients ages 55 to 74 attending the ophthalmology outpatient day-care clinic at the town's primary health-care center (Centro de Especialidades de Burjassot, hereafter referred to as CEB). The CEB belongs to the National Health Care System and is a primary referral health care centre for all individuals living in the same geographical area. The majority of the population in the Valencia province uses the National Health Care System as the main medical service for cataract problems.
A case consisted of any patient from 55 to 74 years of age who was diagnosed with nuclear, cortical, posterior subcapsular, or mixed cataract (i.e. any combination of the types listed) in at least one eye, and of grade ≥ 1 for cortical and posterior subcapsular cataract and ≥ 2 for nuclear opacities, according to the LOCS system Version II (Lens Opacification Classification System II) [19]. The lenses were assessed by slit lamp examination after pupil dilatation and graded using the grading scale for opalescence from 0 to III for nuclear cataract, from 0 to IV for cortical cataract, and from 0 to III for posterior sub capsular cataract. Grade 0 refers to no opacities for any of the subtypes.
Each participant's lens status was independently classified by two observers, and agreement reached for every diagnosis.
If the patient was aphakic or pseudophakic in one eye, he/she could still be a case if the other eye was diagnosed as having any type of cataract. Controls consisted of other patients not affected by cataracts and frequency-matched by age (within 5 years) and gender with the cases. Among these patients, both eyes had no lens opacities, according to the LOCS II classification system. Participants with unilateral congenital or traumatic cataract were classified according to their other eye. We excluded patients if they (i) were on special diets; (ii) had an intra-ocular pressure of ≥ 21 mm Hg; or (iii) had conditions or medical treatments known to be associated with an increased risk of cataract (including people with diabetes). Diabetics were excluded for two main reasons (i) Diabetics undergo regular ophthalmologic examinations for risk of diabetic retinopathy. Surveillance bias therefore may make diabetic patients more likely to be diagnosed with cataract at an earlier stage than the general population from which the cases were drawn (ii) diabetics follow special diets. Cases were also excluded if they had a congenital or traumatic cataract or had a cataract associated with exposure to radiation, or toxic agents.
The study design adhered to the tenets of the Declaration of Helsinki and was approved by the Ethical Committees of the London School of Hygiene and Tropical Medicine, England, and the CEB, Spain. Written informed consent was obtained from each participant.
Interview data
Data were collected by trained study interviewers from the Valencia Institute of Public Health (IVESP) who were unaware of cataract status and the hypotheses under study. The data collected included measurements of height and weight, as well as a structured questionnaire for information on socio-demographic variables, history of severe episodes of diarrhoeal illness; all medications currently being used, "usual" diet using a validated Food Frequency questionnaire; current and past tobacco and alcohol consumption; use of vitamin supplements; and information on outdoor exposure (see below). Blood antioxidant vitamin levels were also analyzed [18].
Measurement of Sunlight Exposure
Years of outdoor exposure were estimated according to a modified version of the empirical model developed by Rosenthal et al. [20]. The original Rosenthal exposure model uses three data sources (i) a questionnaire to collect personal lifetime outdoor exposure; (ii) UVR meteorological data (average annual ultraviolet radiation flux, reflectiveness of terrains); and (iii) laboratory data on ocular protection provided by hats, glasses and sunglasses.
We used a questionnaire designed for use among the general population and provided by Dr S. K. West in 1993 (Dana Center for Preventive Ophthalmology, Johns Hopkins University, USA). This was slightly modified to include some additional information (e.g. season of exposure). The questionnaire's main component was the chronological history of time spent outdoors between 10 AM to 4 PM (when UVR is at its maximum in the Northern Hemisphere), during working periods from age 25 to the date of interview (or diagnosis). Eighty percent of all cases were newly diagnosed at the time of the interview, but 20% had a clinic record of a diagnosis of cataract within the recruitment period (1994 to 1995). For these, information on sunlight exposure was collected up to the date of diagnosis.
Periods during retirement or spent working as a housewife were considered as employment periods. We also collected information on time spent outdoors on weekends during the summer but did not use it because it was found to be inconsistently reported by people in different occupations. The questionnaire assumed that everybody had 2 days per week-end. However, for the most frequent occupations among the participants such as farmers, bricklayers, and shop assistants, the usual work week was 6 days; and for others, such as housewives and retired people, 7 days. For people who worked more than 5 days and reported leisure time activities it was not possible to establish how much of the daily exposure over the week-end was due to work or leisure activities and therefore there was a possibility of double counting the hours of sunlight exposure. The outdoor exposure information was collected for each occupation, with winter (October to March) and summer (April to September) periods separated. For those who were working in shifts, each shift was considered a separate job, since exposure to sunlight might vary considerably depending on the job schedule. We did not have information on local meteorological data regarding radiant UV-B for the period under investigation. Instead, we used the average monthly number of sunlight hours in Valencia from 1938 to 1990, covering most of the exposure period for the study participants [17], to account for differences in summer and winter exposures. This led to weighing winter years of exposure by a factor of 0.39, and the summer ones by a factor of 0.61. We separately calculated the total number of years of winter and summer exposure for every job period and then added the contributions from all of them. The weighted average of the winter and summer years of exposure thus represents the cumulative number of years of outdoor exposure between 10 AM and 4 PM from age 25 to the date of interview or diagnosis.
Shown below is the model for determining years of outdoor exposure (OE).
OE = {0.39 Σ
pw years
pw [hrs
pw x days
pw] + 0.61 Σ
qs
years
ps
[hrs
qs x days
qs]}/(365.25*24)
Where
pw = Indicator for summer employment periods p
q s = Indicator for winter employment periods q
years
pw
= Total number of years in employment period pw
years
qs
= Total number of years in employment period qs
dayspw= Number of weekdays per year in employment period pw
days
qs
= Number of weekdays per year in employment period qs
hrs
pw
= Number of hours per day between 10 AM and 4 PM spent outdoors during employment period pw
hrs
qs
= Number of hours per day between 10 AM and 4 PM spent outdoors during employment period qs
Finally, data on the use of protective devices such as hats, spectacles, and sunglasses while outdoors was collected and was included in an alternative definition of years of outdoor exposure, hereafter referred to as "corrected years of outdoor exposure".
The responses for the use of protective devices ranged from "never," "less than half the time," "half the time," and "more than half the time," to "all the time." This measure weighed the number of hours spent outdoors during a given employment period by 1, provided no protective factor was used, or by a term dependent on which type of protection was used. We used the same weights as those arrived at by Rosenthal et al. in studies of the ocular dose of UV radiation from exposure to sunlight [21–23]. According to their results, the level of ocular protection provided by spectacles is 79%, i.e. the percentage of time that a person wears spectacles while outdoors is multiplied by a factor of 0.21. The protection provided by the use of sunglasses is 93% (hence the factor is 0.07); and that provided by hats an average of 47% (ranging from 75% for a wide-brimmed hat and 20% for a short-brimmed-hat).
Statistical Analysis
The analyses were first carried out taking cataract as a general outcome (all types of cataracts combined), and then, repeated by the specific type of cataract to explore whether the exposures had any different effect by type of cataract. The two measures of exposure, outdoor years and corrected outdoor years, were categorized into fifths according to the quintiles of the controls distribution, with the lowest category treated as the reference group. The associations between the categories of these two exposure variables and risk of cataract were evaluated in terms of odds ratios (ORs) with 95% confidence intervals (95% CI), using two separate age-and sex-adjusted logistic regression models.
We examined the effects of potential confounders, namely dietary intake; plasma antioxidant levels; episodes of severe diarrheal illness; supplementary use of antioxidant vitamins; body mass index; regular use of aspirin, anti-hypertensive and anti-gout medication by including these factors sequentially into the two separate logistic regression models and examining whether they changed or modified the original ORs. Education, "pack-years" of cigarette consumption and alcohol history were forced into all models. Final models used multiple logistic regressions to estimate odds ratios adjusted by smoking history, alcohol consumption, and education and serum levels of ascorbic acid, retinol and lycopene. Overall significance and linear trends were assessed using Likelihood Ratio Tests (LRT) [24].
To explore whether either measure of years of outdoor exposure were relevant for a particular type of cataract, the two models were also fitted using all the controls and either case subset with pure nuclear, pure posterior, or pure cortical. Participants were classified as having a "pure" type of cataract when only one type of lens opacity was present in one or in both eyes. All statistical analyses were performed using Stata 6 software [25].
