The Lens Opacities Case-Control StudyRisk Factors for CataractM. Cristina Leske, MD, MPH; Leo T. Chylack, Jr, MD; Suh-Yuh Wu, MA; The Lens Opacities Case-Control Study Group

\s=b\The Lens Opacities Case-ControlStudy evaluated risk factors for age-re-lated nuclear, cortical, posterior subcap-sular, and mixed cataracts. The 1380participants were ophthalmology outpa-tients, aged 40 to 79 years, classified intothe following groups: posterior subcap-sular only, 72 patients; nuclear only, 137patients; cortical only, 290 patients;mixed cataract, 446 patients; and con-

trols, 435 patients. In polychotomous lo-gistic regression analyses, low edu-cation increased risk (odds ratio[OR]= 1.46) and regular use of mul-tivitamin supplements decreased risk(OR =0.63) for all cataract types. Dietaryintake of riboflavin, vitamins C, E, andcarotene, which have antioxidant poten-tial, was protective for cortical, nuclear,and mixed cataract; intake of niacin,thiamine, and iron also decreased risk.Similar results were found in analysesthat combined the antioxidant vitamins(OR =0.40) or considered the individualnutrients (OR =0.48 to 0.56). Diabetesincreased risk of posterior subcapsular,cortical, and mixed cataracts (OR =1.56).Oral steroid therapy increased posteriorsubcapsular cataract risk (OR = 5.83).Females (OR =1.51) and nonwhites(OR = 2.03) were at increased risk onlyfor cortical cataract. Risk factors for nu-clear cataract were a nonprofessionaloccupation (OR =1.96), current smoking(OR = 1.68), body mass index (OR = 0.76),and occupational exposure to sun-

light (OR =0.61). Gout medications(OR =2.48), family history (OR =1.52),and use of eyeglasses by age 20 years,which is an indicator of myopia(OR = 1.44), increased risk of mixed cata-ract. The results support a role for thenutritional, medical, personal, and otherfactors in cataractogenesis. The poten-tially modifiable factors suggested bythis study merit further evaluation.

(Arch Ophthalmol. 1991;109:244-251)

ge-related cataract is the maincause of blindness and visual im¬

pairment throughout the world.1"4 Withthe aging of the population, the overallprevalence of visual loss due to lensopacities continues to increase everyyear. More than 15 million personsworldwide are estimated to be blindfrom cataract, and this number is ex¬

pected to reach 40 million by the year2025' in the absence of new or more

efficiently delivered interventions. Be¬cause surgery is the only treatment,the growing need for surgical re¬sources further adds to the consider¬able socioeconomic impact of the condi¬tion. The problem is especially criticalin developing countries but affects themore developed countries as well; theeconomic impact of cataract surgeryalone in the United States is estimatedto be more than $2.5 billion per year.5

See also pp 196, 252, and 256.

Oxidative damage plays a major rolein cataractogenesis.'' Animal, laborato¬ry, and some clinical and epidemiologiedata support the relationship of cata¬ract to nutritional factors such as lowlevels of proteins and vitamins withantioxidant potential, ie, riboflavin, vi¬tamins C and E, and the carotenoids.7"3Other studies have identified severalpersonal, medical, and environmentalcataract risk factors,21""2" lending fur¬ther support to the hypothesis thatopacities result from the cumulativeeffect of various insults to the lens.Furthermore, some evidence suggeststhat different mechanisms, such as

protein aggregation and osmotic dam¬age, may cause cataracts in differentregions of the lens.6'7

Given the magnitude and serious¬ness of the cataract problem, the iden¬tification of protective factors has ahigh priority.1"' The Lens OpacitiesCase-Control Study (LOCS) was de¬signed to meet this need by a multifac-torial, comprehensive investigation.Its primary aim was to evaluate riskfactors for nuclear, cortical, and poste¬rior subcapsular opacities, and to de¬termine whether these types of cata¬ract differ when regarding their riskfactors. Another aim was to compareLOCS findings with those of parallelcase-control studies conducted in In¬dia18 and Italy; the latter study began

after LOCS and followed its design andmethods. This article presents resultsfor all LOCS participants, as well as

comparisons with the Indian study, forthe following groups of hypothesizedrisk factors2'6""'1""1": (1) nutritional fac¬tors (including intake of proteins, vita¬mins, and calcium; body mass index);(2) medical and medication history (in¬cluding diabetes and cardiovasculardiseases; use of oral steroids, antihy-peruricemics, aspirin, smoking, and al¬cohol); (3) personal characteristics(including education and occupation,race, iris color); and (4) environmentaland other factors (including sunlightexposure, myopia and use of eyeglass¬es, family history).

METHODSIdentification of Study Population

The LOCS was designed to have fourcase groups (nuclear, cortical, posteriorsubcapsular [PSC|, or mixed cataract) andone control group (no cataract). Cases andcontrols were general ophthalmology outpa¬tients at the Massachusetts Eye and EarInfirmary and the Brigham and Women'sHospital, Boston, Mass, from December1985 to December 1988. Cases had age-related nuclear, cortical, or PSC opacities inat least one eye that explained a decrease invisual acuity, if any; controls had no lensopacities and visual acuities of 20/20 orbetter in both eyes. All patients were aged40 to 79 years, were Massachusetts resi¬dents, and were able to provide informedconsent and complete an interview in En¬glish. Persons with diagnoses that pre¬cluded full pupillary dilatation (to at least 6mm), lens evaluation at the slit lamp, orlens photography were ineligible, as were

patients with other types of cataract or eyepathology causing a loss of visual acuity.

Medical records of incoming patientswere screened for eligibility by a patientrecruiter following preset criteria for inclu¬sion and exclusion, as well as for achievingbalance in the number and age of cases andcontrols (eg, limiting the weekly number ofcontrols to achieve a 3:1 case-control ratioand selecting the oldest available controls).Records of all patients screened were re¬ferred to a LOCS ophthalmologist and thento an epidemiologist for further review ofeligibility criteria and possible diagnoses ofexclusion. After this review, the ophthal¬mologists of potentially eligible patientswere contacted to determine contraindica¬tions for participation; ifnone were present,the patient was invited by letter and tele¬phone to complete a special study visit fordata collection and definitive classificationinto the appropriate study group.

Accepted for publication July 27, 1990.From the Division of Epidemiology, the De-

partment of Preventive Medicine, School of Medi-cine, State University of New York at StonyBrook (Dr Leske and Ms Wu); and the Center forClinical Cataract Research, The Brigham andWomen's Hospital, The Massachusetts Eye andEar Infirmary, and Harvard Medical School, Bos-ton, Mass (Dr Chylack).

Reprint requests to the Division of Epidemiolo-gy, Department of Preventive Medicine, SUNYat Stony Brook, HSC, L3-099, Stony Brook, NY11794-8036 (Dr Leske).

Downloaded From: by a Western University User on 09/12/2018

Classification Into Study GroupsThe type and degree of lens opacification

were graded at the slit lamp and from lensphotographs following the protocol of theLens Opacities Classification System I(LOCS I) that was developed for use in thisstudy.21'22 This system, which is based onstandard photographs, provides ordinalscores of the degree of opacification sepa¬rately for the nuclear (NO, Nl, or N2),cortical (CO, Cla, Clb, or C2), and PSC(PO, PI, or P2) regions of the lens. Evalua¬tion of the system has shown high interob-server and intraobserver reproducibility,22as measured by weighted ( ) statistics2'1;similar results were found in quality controlassessments conducted throughout thestudy ( = 0.74 to 0.93), using the samemethods.22 One ophthalmologist performed63.7% of the LOCS I slit-lamp classifica¬tions and the rest were performed by threeophthalmologists (13.0%, 12.4%, and10.9%). Slit-lamp and photographic grad-ings yielded similar results; assignment intostudy groups was based on the slit-lampgradings.

Patients were classified as having a singletype of cataract (ie, nuclear, cortical, orPSC) when there was only one type ofopacity present, as defined by the LOCS Iscores of both eyes. The nuclear group,therefore, consisted of patients with onlynuclear opacities (LOCS I score = Nl or N2)in both eyes or with nuclear opacities in one

eye and no opacities in the other eye.Similar criteria were used to define thecortical group (LOCS I score = Clb or C2)and the PSC group (LOCS I score = PI orP2). The mixed cataract group consisted ofpatients with various combinations of nucle¬ar, cortical, and PSC opacities in one orboth eyes.

Data Collection

Data were collected by trained studypersonnel who were unaware of the specificrisk factors of major interest; the interview¬er was also unaware of the case or controlstatus of the participant. A standardizedprotocol was followed to collect the dataoutlined below.

(1) Measurements of height, weight,hand-grip strength, and blood pressure(average of two measurements with a ran¬dom zero sphygmomanometer); iris color(graded as light, medium, or dark usingphotographic standards).

(2) Nutritional intake assessment withthe food frequency questionnaire devel¬oped, described, and evaluated by Block etal24 and Cummings et al, which was de¬rived from dietary intake data from theSecond Health and Nutrition ExaminationSurvey.21''2' This instrument provides esti¬mates of intake of calories, fats, carbohy¬drates, vitamins (A, C, E, thiamine, ribofla-vin, niacin, carotene), minerals (calcium,phosphorus, iron, sodium, potassium), anddietary fiber, which are based on usualfrequency of consumption and portion sizeof specific foods.24 The food-frequencymethod was selected because it providesinformation on long-term dietary patterns*;it provides rankings of intake in the studygroups and thus permits meaningful com-

parisons among these groups.(3) Standardized interview to assess med¬

ical care practices (frequency of visits, typeof physician, and main reason for visits inthe last 5 years); history of diabetes, cardio¬vascular diseases, arthritis, and major med¬ical problems; use (at least once a week forat least 1 month) of medications for diabe¬tes, cardiovascular diseases, and arthritis,of oral steroids, antihyperuricemics, aspi¬rin, prescription eyedrops, vitamin andmineral supplements; use of cigarettes,pipe, or cigars; use of beer, wine, or liquor;race (self-reported), marital status, years ofeducation, major lifetime occupation; occu¬

pational exposure to sunlight, bright indoorlights, or arc welding; leisure time in thesun; residence and travel to areas of highsun exposure; use of hats, sunglasses, andregular eyeglasses; hair color (at age 20years), freckles, and skin sensitivity to thesun; ancestry, sibship size, and family histo¬ry of medical and ocular conditions in par¬ents and siblings.

(4) Complete ophthalmologic examina¬tion, including refractive error and visualacuity (Humphrey Automated Refractor);applanation tonometry; LOCS I classifica¬tion; and ocular diagnoses.

(5) Color slit-lamp (Zeiss), retroillumina-tion (Neitz CTR), and black and whitephotography (Topcon SL-45 Scheimpflug)of the lens.

AnalysisTo evaluate risk factors in a single model

for PSC, nuclear, cortical, and mixed cata¬racts, each case group was compared withthe controls and with the other case groupsusing polychotomous logistic regression.1 '

This multivariate method is an extension ofstandard logistic regression "; while the lat¬ter method is dichotomous (eg, it comparesone disease category with a referentgroup), polychotomous logistic regressionallows evaluations of risk factors for sever¬al mutually exclusive disease categories,while simultaneously adjusting for con¬

founding and other exposures. In case-con¬trol studies with more than one case group,polychotomous logistic regression has beenused to compare risk factors between eachcase group and the controls in a singlemodel; it also allows the comparison of riskfactors among the various case groups.

The analysis comprised three major stepsas outlined below. First, associations witheach risk factor were evaluated by age- andsex-adjusted analyses; Mantel-Haenszeltests for odds ratios (ORs) and logisticregression analyses were used. Second, thevariables found to be significantly (P<,05)associated with any cataract type in thefirst phase of analyses were grouped into(1) nutritional, (2) medical, (3) personal, and(4) environmental and other risk factors.Polychotomous logistic regression, adjust¬ing for age and sex, was conducted withineach group. Third, variables significantly(P<.05) associated with at least one of thefour cataract types in the second analyticstep were retained to form a full model formultivariable polychotomous regression.The similarity in risk factors across somecataract types in the full model led to areduced model, in which (1) coefficients for

a given variable were constrained to beequal across selected cataract types whentheir estimates in the full model were al¬most equal, and (2) coefficients for certainvariables for selected cataract types wereconstrained to be zero when their estimatesin the full model were not significantlydifferent from those of controls.1"23 Thecomparison between the full model and thefinal reduced model was made by a 2 test ofthe improvement of log likelihood.

For these analyses, nutritional intakevariables were grouped as low, medium,and high using the lowest quintile, thethree middle quintiles, and the highestquintile, respectively; ORs compared highvs low, using the latter as the referencegroup. Because estimates of nutrient intakefrom food frequency questionnaires may notnecessarily adjust for total caloric intake,dietary intake data were evaluated usingthe absolute values and also adjusting suchvalues to account for caloric intake.2" ' Simi¬lar results were found with and withoutsuch adjustments, and the final analysesused the absolute nutrient estimate values.The high correlations among dietary nutri¬ents" lead to multicolinearity problems inmultivariate modeling and cause difficultiesin interpreting results. An alternative is toselect one nutrient as a proxy for the analy¬sis, but this does not account for possiblecumulative or synergistic effects among nu¬trients. Furthermore, the report of a signif¬icant association is likely to be misinter¬preted as the effect of the selected nutrientalone. These two problems are decreased ifdietary nutrients are grouped into indexes.Because of the interest in evaluating hy¬pothesized associations with all the vita¬mins with antioxidant potential, an index ofantioxidant vitamin intake was defined bygrouping riboflavin, vitamin C, vitamin E,and carotene (the sum of -carotene,ß-carotene, lutein, cryptoxanthin, otherxanthins, and lycopene). The index wasdefined as follows: (1) low, if at least threeof the vitamins were in the lowest quintileand the rest were not in the highest quin¬tile; (2) high, if at least three of the vitaminswere in the highest quintile and the restwere not in the lowest quintile; and(3) medium, otherwise. All steps of theanalyses evaluated associations with indi¬vidual nutrients as well as with the antioxi¬dant index.

RESULTS

Of the 2491 charts initially screened,1772 (71.1%) were of potential casesand 719 (28.9%) were of potential con¬

trols; 529 were found to be ineligibleafter further review and were ex¬cluded. Of those eligible, 1428 (72.8%)completed study visits with identicalpercent participation for cases and forcontrols; lack of time or interest wasthe main reason given for nonpartici-pation. Nonparticipating cases andcontrols were similar to participants inage, residence, race, and diagnoses,and had a higher proportion of women.We were unable to determine possiblebiases due to selective participation; a

Downloaded From: by a Western University User on 09/12/2018

Table 1.—Age, Sex, Race, and Education in the Four Cataract Case Groups and Controls

Controls(n = 435)

Type of Cataract

PosteriorSubcapsu¬

lar Only(n = 72)

Cortical Only(n = 290)

Nuclear Only(n = 137)

Mixed(n = 446)

Age, yMean ± SD 59.7 ± 8.5 61.8 ± 8.7 65.3 ± 7.2 69.0 ± 6.5 67.5 ± 7.5Median 68.040-49, % 13.6 9.7 2.4 0.7

12.160-69, %

£70, %

Sex, %M

Race,W

46.4

53.6

85.7 94.4

56.2

63.1

76.9

44.5

62.0

90.4

43.3

Other

Education, y<12, %

12, %

2:13, %

3.5

17.638.144.3

25.038.936.1

4.5

29.936.633.4

34.630.934.6

36.2

mail survey of all nonparticipants toascertain major risk factors in thatgroup had only a 26.4% response.

After the study visit was completed,48 persons were found to be ineligible,leaving a total of 1380 participants: 945cases (68.5%) and 435 controls (31.5%).This article is based on interview andexamination data from these partici¬pants, who were classified into thefollowing study groups: 72 PSC only,137 nuclear only, 290 cortical only,446 mixed cataract, and 435 controls.The mixed group consisted of the fol¬lowing subgroups: PSC and cortical,112; PSC and nuclear, 91; nuclear andcortical, 105; and PSC, nuclear, andcortical, 138. Of the cases, 59.3% had aLOCS I score of 2 and 44.0% had avisual acuity of 20/40 or worse in one orboth eyes. Table 1 presents the distri¬bution of study groups by age, sex,race, and educational status; Table 2lists the diagnoses among the controls.

Cases were similar to controls infrequency of physician visits and otherhealth care use variables. Verificationof the medical and medication historydata obtained by interview was at¬tempted for the 1134 participants(82.6%) who had a primary care physi¬cian and authorized the release of suchinformation. Physician response afterthree mailings was 85.6% for cases and85.8% for controls. Results, whichwere similar across all case groups andcontrols, showed excellent agreementbetween self-report and physician re¬

port of diabetes ( = 0.81) and hyper¬tension ( = 0.79), fair to good agree¬ment for coronary heart disease, use of

oral steroids, and gout medications( = 0.50 to 0.64), and poor agreementfor arthritis ( = 0.37) and aspirinusage (k = 0.28). In the first step ofstatistical analysis, several variableswere significantly associated with oneor more types of cataract as outlinedbelow.

Nutritional Factors

Most of the ORs for nutrient intakewere under unity; the number of sta¬tistically significant associations wasrelated to the size of each case group.A decreased risk of some cataracttypes was associated with the follow¬ing: high dietary intake of vitamin A,vitamin C, vitamin E, riboflavin, nia-cin, thiamine, and iron; high values ofthe antioxidant index; ever use of mul-tivitamin supplements on a regular ba¬sis (at least once a week for at least 1year); and high body mass index (kilo¬gram per square meter) (Table 3).

Personal Characteristics

Increased risk for certain case

groups was associated with nonwhiterace, low education (<12 years), non-

professional occupation, and iris color(dark and medium vs light) (Table 4).

Medical History Information

Risk of some cataract types in¬creased with the following: diabeteshistory (for at least 1 year); ever use oforal steroids and gout medications on a

regular basis (at least once a week forat least 1 year); and current smoking(at least one cigarette a day for at least1 year and still smoking) (Table 4).

Table 2.—

Distribution of DiagnosesAmong the Controls

Diagnostic Categories No. (%)Retractive error 206 (47.3)

Presbyopia 85Myopia 22Hyperopia 28Unspecified 71

Normal results ofeye examination 93(21.4)

Eye pathology 136(31.3)Eyelid 54Lacrimal 16Vitreous 14Retina 9Ocular hypertension 7Conjunctiva 4Cornea 3Orbit 3Other 26

Total 435 (100.0)

Environmental and Other Factors

Occupational exposure to bright sun¬

light (at least 2 h/d for at least 2months) decreased nuclear cataractrisk; family history of cataract in par¬ents and/or siblings; and use of eye¬glasses before the age of 20 yearsincreased mixed cataract risk (Table4).

After the second analytic step, allthe above factors remained statistical¬ly significant. In the third analyticstep, therefore, these variables formedthe full model for polychotomous logis¬tic regression, controlling for age andsex, to evaluate risk factors by cata¬ract type. The statistically significantdietary factors were highly intercorre-lated. As outlined in the "Methods"section, polychotomous logistic regres-

Downloaded From: by a Western University User on 09/12/2018

Table 3.—Mantel-Haenszel Age and Sex-Adjusted Odds Ratios* (OR) and 95% Confidence Intervals (CI)for Significant Nutritional Variables by Case Group

Risk Factors

Type of Cataracts

Posterior SubcapsularOnly (n = 72)OR (95% CI)

Cortical Only(n = 290)

OR (95% CI)

Nuclear Only(n - 137)

OR (95% CI)

Mixed(n - 446)

OR (95% CI)Nutritional intake

Vitamin A, IU 1.35 (0.59, 3.08) 0.86 (0.52, 1.42) 0.45 (0.23, 0.88)t 0.60 (0.37, 0.96)tVitamin C, mg 0.73 (0.31, 1.73) 0.80 (0.50, 1.29) 0.48 (0.24, 0.99)t 0.72 (0.46, 1.12)Vitamin , 1.18 (0.51, 2.75) 0.59 (0.35, 0.99)t 0.66 (0.33, 1.32) 0.58 (0.37, 0.93)tRiboflavin, mg 1.85 (0.76, 4.49) 0.59 (0.36, 0.97)t 0.72 (0.35, 1.45) 0.65 (0.40, 1.05)Niacin, mg 1.22 (0.52, 2.87) 0.59 (0.34, 1.00)t 0.57 (0.29, 1.13) 0.47 (0.30, 0.76)tThiamine, mg 1.30 (0.50, 3.38) 0.51 (0.30, 0.86)t 0.60 (0.30, 1.20) 0.55 (0.34, 0.91)tIron, mg 1.09 (0.49, 2.41) 0.62 (0.37, 1.04) 0.59 (0.30, 1.13) 0.49 (0.30, 0.80)tAntioxidant vitamins index 0.89 (0.22, 3.55) 0.42 (0.18, 0.97)t 0.52 (0.15, 1.80) 0.39 (0.19, 0.80)t

Multivitamin supplements 0.40 (0.21, 0.77)t 0.52 (0.36, 0.75)t 0.55 (0.33, 0.92)t 0.70(0.51, 0.97)tBody mass index§ 0.88 (0.67, 1.15) 1.03 (0.88, 1.20) 0.78 (0.62, 0.97)t 1.09 (0.94, 1.25)

Changes in risk factors are as follows: nutritional intake, high vs low; multivitamin supplements, yes vs no; body mass index, 1 SD increase.tP< .05. indicates tocopherol equivalents.§Odds ratios are based on coefficients from the logistic regression controlling for age and sex.

Table 4.—Percent Frequencies and Mantel-Haenszel Age-and Sex-Adjusted Odds Ratios (OR)for Significant Personal, Medical History, and Other Variables by Study Group*

Type of Cataract

Risk Factors

Control, %(n = 435)

OR (95% CI)

Posterior SubcapsularOnly, %(n = 72)

OR (95% CI)

Cortical Only, %(n = 290)

OR (95% CI)

Nuclear Only, %(n = 137)

OR (95% CI)

Mixed, %(n = 446)

OR (95% CI)Personal characteristics

Nonwhite 14.3 5.60.38 (0.14, 1.08)

23.12.32 (1.53, 3.53)t

9.61.05 (0.52, 2.15)

15.21.37 (0.88, 2.12)

Low education 25.01.46 (0.82, 2.62)

29.91.85(1.27, 2.68)t

34.61.99(1.21, 3.27)t

33.01.90 (1.34, 2.70)t

Nonprofessional 71.81.02 (0.58, 1.80)

77.51.11 (0.77, 1.60)

86.11.81 (1.02, 3.22)t

79.91.29 (0.91, 1.81)

Darker iris color 58.00.70 (0.42, 1.18)

71.01.47 (1.05, 2.08)t

72.21.92 (1.18, 3.12)t

66.61.19 (0.88, 1.61)

Medical historyDiabetes 15.3

1.47 (0.70, 3.08)17.9

1.98(1.25, 3.13)t5.1

0.47 (0.19, 1.19)17.5

1.96 (1.28, 3.00)tOral steroids 9.7

5.73(2.14, 15.3)t1.0

0.48 (0.13, 1.80)3.6

2.18(0.65, 7.28)3.6

1.56 (0.59, 4.14)Gout medications 4.2

1.12 (0.31, 4.02)3.4

1.26 (0.50, 3.14)4.4

1.26 (0.38, 4.24)8.1

2.71 (1.35, 5.42)tCurrent smoking 23.9

1.64 (0.87, 3.08)16.5

1.10 (0.78, 1.84)20.4

2.30 (1.30, 4.07)117.8

1.47 (0.99, 2.18)Environmental and other exposures

Work in sunlight 40.1 49.31.28 (0.72, 2.26)

34.10.91 (0.64, 1.30)

26.30.53 (0.30, 0.94)t

35.10.80 (0.57, 1.12)

Use of eyeglasses 26.81.58 (0.88, 2.83)

22.81.14 (0.78, 1.67)

19.71.21 (0.73, 2.00)

25.01.50 (1.06. 2.11)t

Family history 24.20.87 (0.47, 1.59)

29.21.10 (0.77, 1.57)

30.21.32 (0.82, 2.13)

37.41.72 (1.24, 2.37)t

*CI indicates confidence interval.tP< .05.

sion analyses were conducted sepa¬rately for each of the individual nutri¬ents and the antioxidant index; theseanalyses yielded similar results. Toevaluate associations with all vitaminswith antioxidant potential and accountfor their possible joint effects, the finalanalyses used the index defined a prio¬ri to represent and summarize thesenutrients; results for the individual

nutrients are presented in the text.The full model considered four case

groups (PSC, cortical, nuclear, andmixed cataracts) and one referencegroup—the controls. Before reachingthe final reduced model, all case

groups were combined and comparedwith the controls by dichotomous lo¬gistic regression to evaluate whetherrisk factors were similar across all

cataract types. test of devianceshowed that combining all the fourcase groups into one was not accept¬able simplification (deviance = 438.4;ri/=54;P<.001).

Table 5 presents the ORs with 95%confidence intervals based on the coef¬ficients estimated from the final re¬duced model. 2 test of devianceindicated that this reduced model was

Downloaded From: by a Western University User on 09/12/2018

Table 5.—The Final Polychotomous Logistic Regression Model; Odds Ratios* (OR) Are Based on Coefficients From the Model

Risk Factors

Type of Cataract

PosteriorSubcapsular

Only OR (95% CI)Cortical OnlyOR (95% CI)

Nuclear OnlyOR (95% CI)

MixedOR (95% CI)

Personal characteristicsAge 1.03 (0.99, 1.06) 1.09 (1.06, 1.11)+ 1.17 (1.13, 1.21)t 1.13 (1.11, 1.16)1

0.67 (0.39, 1.16) 1.51 (1.07, 2.12)1: 1.00 (0.61, 1.64) 1.10 (0.80, 1.50)Nonwhite 2.03 (1.40, 2.93)1Low education 1.46 (1.06, 2.02) 1.46 (1.06, 2.02)1 1.46 (1.06, 2.02)1 1.46 (1.06, 2.02)1Nonprofessional 1.96 (1.10, 3.50)1Darker iris color 1.52 (0.99, 2.35)

Nutritional statusMultivitamin supplements 0.63 (0.47, 0.84)t 0.63 (0.47, 0.84)t 0.63 (0.47, 0.84)t 0.63 (0.47, 0.84)tAntioxidant vitamins

index§ 0.75 (0.18. 3.09) 0.40 (0.21, 0.78)t 0.40 (0.21, 0.78)1 0.40 (0.21, 0.78)tBody mass index # 0.76 (0.62, 0.95)1:

Medical historyDiabetes§ 1.56 (1.03, 2.35)* 1.56 (1.03, 2.35)1 0.45 (0.19, 1.06) 1.56 (1.03, 2.35)1Oral steroids 5.83 (2.39, 14.2)tGout medications # 2.48 (1.42, 4.25)tCurrent smoking 1.68 (1.03, 2.75)1 #

Environmental and other exposuresUse of eyeglasses 1.44 (1.06, 1.94)1Work in sunlight 0.61 (0.37, 0.99)1Family history # 1.52 (1.17, 1.99)t

'Changes in risk factors are as follows: age, 1-year increase; body mass index, 1 SD increase; antioxidant index, high vs low; the remaining risk factors, yes vs no.The symbol "#" designates ORs of 1 for those coefficients constrained to be equal to the controls. CI indicates confidence interval.

tComparisons of cases vs controls: P< .01.IComparisons of cases vs controls: .01 < < .05.¡¡Comparisons among groups: antioxidant vitamins index—(cortical = nuclear = mixed) vs PSC, = .20; diabetes

—

(PSC = cortical = mixed) vs nuclear, < .01.

compatible with the observed data (de¬viance =40.4; df=42; =.58). In thistable, similar coefficients were con¬strained to be equal across case groupsfor low education (all groups), regularuse of multivitamins (all groups), anti¬oxidant vitamins index (all groups ex¬

cept PSC), and diabetes (all groupsexcept nuclear). The symbol "#" desig¬nates ORs of 1.00 for those groupsconstrained to be equal to the controls.The results found in this final modelare similar in direction and magnitudeto those obtained from the Mantel-Haenszel analyses in Tables 3 and 4.Given the size of the study groups, thepower to detect ORs of at least 2 forfactors with a frequency of at least15% among the controls was 57% forPS, 94% for cortical, 79% for nuclear,and 98% for mixed cataracts (a = .05;two-sided tests). It is possible, there¬fore, that only large ORs could bedetected for the PSC group.

As expected, all case groups wereassociated with age (except for PSC);the ORs provide a measure of theincrease in cataract risk for an increaseof 1 year in age. Other variables re¬lated to all cataract types were educa¬tion and use of multivitamins. Havingless than 12 years of education in¬creased the risk (OR = 1.46) for everycataract type; regular use of multivita¬min supplements was protective(OR = 0.63) for each of the four cata-

ract types. The lower use of multivita¬min supplements in cases than controlswas observed in all age, sex, race, andeducation subgroups, regardless of thecriteria to define such use (Table 6).

High intake of several dietary nutri¬ents decreased the risk of cortical,nuclear, and mixed cataracts; the PSCgroup was not significantly differentfrom the other case groups (P= .20) orfrom the controls (P = .35), but thepower to detect associations in thisgroup was low because of sample size.Results for the index summarizing in¬take of vitamins with antioxidant po¬tential are given in Table 5 (OR = 0.40);similar results were found in separateanalyses for the individual nutrients.When each nutrient was entered in thefinal model, ORs were very similar tothose presented in Table 3 and re¬mained statistically significant in thelarger case groups. Odds ratios rangedfrom 0.48 to 0.56 with upper 95% confi¬dence limits under unity for riboflavin,niacin, and thiamine in cortical cata¬ract and for vitamin E, niacin, thia¬mine, and iron in mixed cataract.

Posterior subcapsular, cortical, andmixed cataracts were associated with a

history of diabetes for at least 1 year(OR = 1.56). Other risk factors were

specific for only one cataract type.Thus, the use of oral steroids forat least 1 year was associated onlywith PSC (OR = 5.83). Females

(OR = 1.51) and nonwhites (OR =2.03)were at increased risk only for corticalcataract. Risk factors specific for nu¬clear cataract were a nonprofessionaloccupation (OR = 1.96), current smok¬ing (OR = 1.68), occupational exposureto sunlight (OR = 0.61), and body mass

index, where an increase of 1 SD re¬duced risk (OR = 0.76). A dark/mediumiris color, which was associated withnuclear cataract in the initial analyses,had 95% confidence intervals includingunity in the final model. Regular use ofgout medications (OR = 2.48), familyhistory of cataract (OR = 1.52), and useof eyeglasses before age 20 years(OR = 1.44) were risk factors only formixed cataract.

COMMENT

The study found general risk factorsfor all cataract types as well as specificrisk factors for some case groups.These findings support the multifac-torial hypothesis of cataractogenesisand reinforce the need to evaluatecataract types separately. Interpreta¬tion of the results, however, must firstconsider alternative explanations, suchas selection bias, misclassification, biasin the ascertainment of risk factors,and confounding.

Identification of Cases and Controls

We did not select cases from a cata¬ract surgery setting since the reasons

Downloaded From: by a Western University User on 09/12/2018

Table 6.—

Percent of Cases and Controls Ever Using Multiple Vitamin Supplementsby Demographic Characteristics According to Two Definitions*

DemographicCharacteristics

A, %

Cases(n = 945)

Controls(n = 435)

B, %

Cases(n = 945)

Controls(n - 435)

Age, y40-59 30.860-69 31.6 38.6 23.8 33.7

18.3 31.1Sex

M 28.632.6 46.8 23.1

RaceW 22.6 32.8

Other 24.2 33.3

Education, y<12 16.212 32.0 36.0

:13 34.3 48.2 25.7 37.7

A indicates at least once a week for at least 1 month; B, at least once a week for at least 1 year.

that lead to patient selection and refer¬ral for surgery may cause biases32'33that are difficult to assess; in addition,the selection of an appropriate controlgroup for surgical cases is problematic.Our cases and controls were outpa¬tients visiting the same general eyeservices, an approach that increasedtheir comparability on socioeconomic,demographic, health care utilization,and other factors. Biases would occurif risk factor status had a differentinfluence on the outpatient visits ofcases and controls, eg, because of self-referral or physician referral. Theprobability of these biases seems

small, since the possible association ofcataract with most of the factors understudy is not well known. Ascertain¬ment bias could exist for diabetes be¬cause diabetics consult ophthalmolo¬gists more frequently than nondiabet-ics; such bias appears unlikely, sincediabetes frequency was not increasedin all the case groups (Table 4). Noevidence suggests that controls withdifferent diagnoses had different riskfactors, since the distribution of thesefactors was similar in controls withrefractive error (n = 206), normal re¬sults of eye examinations (n = 93), or

eye abnormalities (n = 136). Knowl¬edge of risk factor status is also unlike¬ly to have affected patient selection forthe study, because (1) a detailed proto¬col for case and control selection was

followed, (2) the recruiting staff wasunaware of the specific factors understudy, and (3) most risk factors (eg,nutrient intake, use of multivitamins,etc) were determined after LOCSparticipation.

Misclassification of persons into thecase and control groups was minimizedby using a simple and reproducible

system of lens classification; in addi¬tion, more than half of the classifica¬tions were performed by one examin¬er. If random misclassification didoccur, its effect would be to decreasethe study's ability to detect significantdifferences among the groups.

Nonparticipation could cause bias ifthe frequency of the risk factors understudy was related to participation sta¬tus. Cases and controls had identicalparticipation rates; no differences be¬tween participants and nonparticipantscould be determined. Other case-con¬trol studies of elderly patients reportlower participation rates of 43% to65%.34

Identification of Risk Factors

Interviewer bias was minimized byfollowing a standardized protocol andformat; the interviewer was also un¬aware of the case or control status ofthe participant and of the specificstudy hypotheses. Misclassifications ofrisk factor status due to inaccuratemeasurements (eg, of body mass, nu¬trient intake) are unlikely to differbetween cases and controls. Recallbias could have occurred if cases weremore (or less) likely than the controlsto remember risk factors such as nutri¬ent intake, diabetes, smoking, or useof vitamins, steroids, gout medica¬tions, or eyeglasses. This possibilitycannot be eliminated but seems unlike¬ly, since the case-control differenceswere seen only for those self-reportedfactors and were specific (except formultivitamins) for some cataract typesonly. Furthermore, the results of thephysician verification of medical infor¬mation showed no difference in theresults for cases and controls, withacceptable to excellent agreement for

use of medications and diabetes. Cata¬ract family history, however, is espe¬cially susceptible to recall bias andsuch information was not verified.

Other Issues

To address the issue of confounding,multivariate analysis methods wereused. Because many variables wereevaluated and multiple comparisonswere made, the possibility of chancefindings must be considered; there wasa specific rationale, however, for in¬cluding each of the risk factors evalu¬ated in the study. In sum, although thepossibility of biases and other limita¬tions can never be totally eliminated,the potential of such problems to affectthe associations observed seems low.

General Cataract Risk Factors

Low education has been identifiedconsistently as a cataract risk factor inother studies, including the parallelIndia-US study.1"'1718 Our results alsosuggest that use of multivitamin sup¬plements is protective for all types ofcataract, with users having less thantwo thirds the risk of nonusers. Thefollowing was the first question used toascertain nutritional supplement in¬take: "Have you ever in your lifetimetaken any vitamins or any minerals(like calcium) on a regular basis (atleast once a week for at least 1month)?" Further questions obtaineddata on the type of preparation, fre¬quency, and length of use. The mostfrequent supplements used were mul¬tivitamin preparations. To obtain a rig¬orous definition, only those personswho took multivitamin supplementsregularly at least once a week for atleast 1 ,year were considered users.Although comparisons are limited bythe differences in study populationsand definitions of "user," the frequen¬cy and patterns of supplement use inour controls by age, sex, race, educa¬tion, and type of preparation are con¬sistent with national data.35,36 The dif¬ferences in multivitamin use betweencases and controls persisted after con¬

sidering other risk factors (Table 5)and were consistent across age, sex,race, and education subgroups (Table6), giving further support to the con¬clusion that the association is real.Similar results were found by anothercase-control study, where 175 cataractsurgery patients consumed significant¬ly less supplementary vitamins C andE than their age- and sex-matchedcontrols.37

Specific Cataract Risk Factors

Two or More Cataract Types.—High values of the index of antioxidant

Downloaded From: by a Western University User on 09/12/2018

vitamin intake were protective, de¬creasing the risk of cortical, nuclear,and mixed cataract by more than half(Table 5). Of the individual nutrients,vitamin E, riboflavin, niacin, thiamine,and iron remained significantly associ¬ated with cortical and mixed cataractin the final analyses, a result that maybe related to the larger size of thesecase groups. These findings are consis¬tent with the India-US study and otherepidemiologie studies that reported as¬sociations of cataract with low nutrientintake1" " and decreased blood levels ofantioxidants such as vitamin C, vita¬min E, and the carotenoids.18,38,39 Adecreased intake of vitamins C and E,riboflavin, and carotene is also consis¬tent with evidence from laboratory andexperimental animal studies that sug¬gest that antioxidant nutrients protectagainst oxidative damage to the lens.7,8Dietary intakes of thiamine, niacin,and iron were similarly decreased incases, which could be explained by thesimilar food sources for such nutri¬ents.26 The associations with nutrientintake seemed to be independent oftotal caloric intake; these findings sug¬gest that specific nutrients in the diet,rather than total intake, are related tocataract risk.

Diabetes was associated with PSC,cortical, and mixed cataracts but nu¬clear cataract was unrelated to diabe¬tes in this study. In fact, diabetesfrequency tended to be low in thenuclear group (Tables 3 and 5). Inthe final model, results for the nu¬clear group were significantly dif¬ferent (P<.01) from those of theother groups. While the diabetes-cata¬ract association has been well estab¬lished,2'"1317 it was not found in otherstudies where nuclear opacities werethe most frequent cataract type.19'4"'41The lack of a relationship betweendiabetes and nuclear cataract suggeststhat the pathogenic mechanism for dia¬betic cataract predominantly affectsother lenticular zones. Previous stud¬ies have found the diabetes-cataractassociation only among persons youn¬ger than 65 years.21113 While severalexplanations have been proposed forthe lack of association at older ages,another possibility is the preponder¬ance of nuclear cataract with advanc¬ing age.

One Cataract Type.-PSC Onhj.-The PSC patients were younger andthe group tended to have more whitesand men than the other groups (Table1). Only the use of oral steroids wasidentified as a PSC risk factor, con¬

firming the relationship found in otherstudies.21' However, the low power ofthe study to detect risk factors for

PSC may account for the lack of othersignificant associations, such as withdietary intake variables.

Cortical Only. —A higher prevalenceof cortical cataract in women than menhas been reported17'42; this is consistentwith our finding that women were atincreased risk only for cortical cata¬ract. The risk of cortical cataract fornonwhites was double that for whites,an association also reported previous¬ly. 17 A high frequency of cataract existsin tropical areas of the world such as

India,1"4 where skin pigmentation isdarker than in areas of lower cataractprevalence. The association with non-white race found in our study was

independent of educational, occupa¬tional, and other factors and may indi¬cate a genetically determined influenceon pathogenesis.

Nuclear Only. —A nonprofessionallifetime occupation increased nuclearcataract risk, an association foundafter controlling for educational sta¬tus. Age- and sex-adjusted analy¬ses by specific occupational categor¬ies showed significantly more factoryworkers among the nuclear group thanthe controls (OR =2.2 [1.08 to 4.31]).

Current smoking increased nuclearcataract risk, an association previouslyreported by others.2" This associationwas present only for current smokers,a group that had the longest durationof exposure to cigarettes.

Persons with nuclear cataract had adecreased occupational exposure tosunlight, which is consistent with theirhigh frequency of factory workers, ie,an indoors type of occupation. Somestudies have reported that sunlightincreases risk of cortical and mixedcataracts, but they found no relation¬ship with nuclear cataract1719'40; othercase-control studies did not find defini¬tive associations with sunlight expo¬sure.4tul Our inability to confirm thatrisk of any cataract increases withsunlight exposure could be due to thelimited range of such exposure in oururban study population; it could also beexplained by the limitations of the self-reported data on sun exposure.

Persons in the nuclear group had lowbody mass; this finding is consistentwith their low frequency of diabetes,which is related to body mass. Similarresults were found in the parallel studyconducted in India18 where nuclear cat¬aract was associated with low bodymass and low intake of some nutrients.

Mixed Only.—Use of gout medica¬tions was associated only with mixedcataract. The cataractogenic potentialof allopurinol, especially for PSC, hasbeen suggested in some reports4"' butnot confirmed in others.4,i'47 Another

significant factor was the use of eye¬glasses before the age of 20 years, avariable used as an indicator of myo¬pia. Because cataract often inducesmyopia, analyses based on current re¬fractive error would be difficult tointerpret. Since the onset of myopiais predominantly during adolescence,most persons with clinically significantmyopia would be expected to wearcorrective lenses for distance vision bythe age of 20 years. While there was a

high agreement between this variableand a refractive error of one diopter or

less, the reason for using eyeglasses atan early age was not verified and theassociation with myopia is suggestiveonly. Family history of cataract wasalso related to mixed cataract risk. Asmentioned previously, recall bias couldhave influenced this finding.

CONCLUSIONS

The LOCS results confirm earlierreports that cataract is related to lowsocioeconomic and nutritional status.This conclusion is supported by theassociation with low education, a non-

professional occupation, intake of vita¬mins and iron, and body mass. Thesignificant protective effects of dietarynutrients parallel those for use of mul¬tivitamin supplements. These resultspersisted after controlling for otherfactors and are consistent with reportsof a protective role for antioxidantvitamins.7"9'37"39 While interpretation ofsuch evidence have been hampered byméthodologie limitations of some stud¬ies, LOCS was carefully designed toevaluate these factors while exploringother possible cataractogenic agents aswell. Although these results show thatnutritional factors were independentlyrelated to cataract in multivariate ana¬

lyses, further confirmation and care¬ful interpretation of these results isneeded.

In addition to the general risk fac¬tors above, diabetes was confirmed asa risk factor for all cataract types,except nuclear. A new finding sug¬gested by the study is that diabeteshas a different association with nuclearthan with other types of cataract. Nu¬clear cataract was also significantlydifferent from the other groups withregard to current smoking, occupation,work in sunlight, and body mass. In¬terpretation of these findings mustconsider that our lens classificationsystem defines nuclear cataract basedon the opalescence of the nucleus; nu¬clear color is graded separately in thissystem.21 As such, our criteria for nu¬clear cataract are different from thoseused in other studies.

Other personal characteristics, such

Downloaded From: by a Western University User on 09/12/2018

as female sex and nonwhite race, in¬creased the risk of cortical cataract.The cataractogenic potential of oralsteroids (PSC) and gout medications(mixed cataract) is supported by thestudy. Supportive evidence is also pro¬vided for the association of mixed cata¬ract with myopia and family history.

The LOCS results suggest a multi-factorial etiology in cataractogenesis,where personal, nutritional, medical,and other exposures accelerate lensopacification, perhaps through thecommon pathway of oxidative damage.Although some general risk factorswere identified, the different types ofcataract had different risk factors,thus supporting the premise that cata¬ract types need to be evaluatedseparately.

Given its prevalence, even a modestdecrease in cataract risk has majorpublic health implications. From thisviewpoint, risk factors amenable tomodification or intervention are themost interesting findings of this study,because they offer a possibility forcataract prevention. A potential formodifying cataract risk is suggested bythe associations with nutritional intakeand use of multivitamin supplements.The study also suggests a role forother potentially modifiable factors,such as use of some medications andsmoking. The results highlight a needfor periodic lens evaluations amongpersons with diabetes and those usingoral steroids or gout medications. Suchperiodic monitoring could lead to earlyidentification of lens changes and ap¬propriate patient treatment. Becausereducing cataract risk would have ma¬

jor implications, future studies areneeded to confirm and evaluate thepotentially modifiable factors sug¬gested by this study.

This investigation was supported by grantEY05733 from the National Eye Institute, Be¬thesda, Md.

The Lens Opacities Case-Control Study GroupPrincipal Investigator.—M. Cristina Leske,

MD.Co-investigators.—Leo T. Chylack, Jr, MD;

Robert D. Sperduto, MD; Leslie Hyman, PhD;Roger Grimson, PhD; and Barbara A. Under¬wood, PhD.

Coordinating Center.—Dis Leske, Grimson,and Hyman; Robert Kehoe, MSc; Phyllis Neusch-wender; Linda Warheit-Roberts, MPH: Suh-YuhWu, MA (Stony Brook, NY); and Dr Sperduto(National Eye Institute, Bethesda, Md).

Clinical Center.—Dr Chylack; Deborah Deck¬er, BA; Patricia Khu, MD; Daniel McCarthy, BS,COA; Paul Oppedisano, BS; and CarolineWehner, MS (Boston, Mass).

Biochemical Centers.—Dr Underwood andJohn Wallingford, PhD (National Eye Institute,Bethesda, Md); Satish Srivastava, PhD; NaseemAnsari, PhD; and Anja Schlüter (University ofTexas Medical Branch, Galveston).

Advisors.-Roy C. Milton, PhD (National EyeInstitute, Bethesda, Md); Gladys Block, PhD (Na¬tional Cancer Institute, Rockville, Md); and Ju-

dith Wylie-Rosett. EdD, RD (Albert EinsteinCollege of Medicine, Bronx, NY).

References1. Kupfer C. The conquest of cataract: a global

challenge. Trans Ophthalmol Soc UK. 1984;104:1\x=req-\10.

2. Leske MC, Sperduto R. The epidemiology ofsenile cataracts: a review. Am J Epidemiol.1983;118:152-165.

3. To Restore Sight: The Global Conquest ofCat-aract Blindness. New York, NY: Helen Keller In-ternational Inc; 1986.

4. International Agency for the Prevention ofBlindness. World Blindness and Its Prevention.New York, NY: Oxford University Press; 1984;2.

5. National Advisory Eye Council. 1987 Evalu-ation and Update, A Report of the National Advi-sory Eye Council. Washington, DC: US Dept ofHealth and Human Services. In: Vision Research \p=m-\

A National Plan. 1987. NIH publication 87-2755.6. Report of the Cataract Panel. A Report of the

National Advisory Eye Council. Washington, DC:US Dept of Health and Human Services. In: VisionResearch\p=m-\ANational Plan: 1983-1987. 1984;2(pt3). NIH publication 84-2437.

7. Bunce G. Nutrition and cataract. Nutr Rev.1979;37:337-343.

8. McLaren DS, Halasa A. Nutritional and met-abolic cataract. In: Bellows JB, ed. Cataract andAbnormalities of the Lens. New York, NY: Grune& Stratton; 1975:255-264.

9. Senile cataract and vitamin nutrition. NutrRev. 1989;47:326-328. Editorial.

10. Kahn HA, Leibowitz HM, Ganley JP, et al.The Framingham Eye Study, II: association of oph-thalmic pathology with variables previously mea-sured in the Framingham Heart Study. Am J Epi-demiol. 1977;106:33-41.

11. Hiller R, Kahn H. Senile cataract extractionand diabetes. Br JOphthalmol. 1976;60:283-286.

12. Hiller R, Giacometti L, Yuen K. Sunlightand cataract: an epidemiologic investigation. Am JEpidemiol. 1977;105:450-459.

13. Ederer F, Hiller R, Taylor H. Senile lenschanges and diabetes in two population studies. AmJ Ophthalmol. 1981;91:381-395.

14. Philips CI, Bartholomew RS, Clayton R, etal. Cataracts: a search for associations or causativefactors. In: Regnault F, ed. Symposium on theLens. Princeton, NJ: Excerpta Medica; 1980:19-25.

15. Clayton RM, Cuthbert J, Philips CI, et al.Analysis of individual cataract patients and theirlenses: a progress report. Exp Eye Res. 1980;31:553-566.

16. Chatterjee A, Milton RC, Thyle S. Preva-lence and aetiology of cataract in Punjab. Br JOphthalmol. 1982;66:35-42.

17. Hiller R, Sperduto RD, Ederer F. Epidemi-ologic associations with nuclear, cortical and poste-rior subcapsular cataracts. Am J Epidemiol. 1986;124:916-925.

18. Mohan M, Sperduto RD, Angra SK, et al.India-US case-control study of age-related cata-racts. Arch Ophthalmol. 1989;107:670-676.

19. Taylor HR, West SK, Rosenthal FS, et al.Effects of ultraviolet radiation on cataract forma-tion. N Engl J Med. 1988;319:1429-1433.

20. West S, Munoz B, Emmett EA, Taylor HR.Cigarette smoking and risk of nuclear cataracts.Arch Ophthalmol. 1989;107:1166-1169.

21. Chylack LT Jr, Leske MC, Sperduto R, KhuP, McCarthy D, LOCS Research Group. Lens opac-ities classification system. Arch Ophthalmol.1988;106:330-334.

22. Leske MC, Chylack LT Jr, Sperduto R, KhuP, Wu S-Y, McCarthy D, et al. Evaluation of a lensopacities classification system. Arch Ophthalmol.1988;106:327-329.

23. Fleiss JL. Statistical Methods for Rates andProportions. 2nd ed. New York, NY: John Wiley &Sons Inc; 1981:217-225.

24. Block G, Hartman AM, Dresser CM, CarrollMD, Gannon J, Gardner L. A data-based approachto diet questionnaire design and testing. Am J

Epidemiol. 1986;124:453-469.25. Cummings SR, Block G, McHenry K, Baron

RB. Evaluation of two food frequency methods ofmeasuring dietary calcium intake. Am J Epide-miol. 1987;126:796-802.

26. Block GB, Dresser CM, Hartman A, CarrollMD. Nutrient sources in the American diet: quanti-tative data from the NHANES II survey, I: vita-mins and minerals. Am J Epidemiol. 1985;122:13-26.

27. Block GB, Dresser CM, Hartman A, CarrollMD. Nutrient sources in the American diet: quanti-tative data from the NHANES II survey, II: mac-ronutrients and fats. Am J Epidemiol. 1985;122:27-40.

28. Willett W. Nutritional Epidemiology. NewYork, NY: Oxford University Press; 1990.

29. Marshall RJ, Chisholm EM. Hypothesistesting in the polychotomous logistic model with an

application to detecting gastrointestinal cancer.Stat Med. 1985;4:337-344.

30. Breslow NE, Day NE. The Analysis ofCase-control Studies: Statistical Methods in Can-cer Research. Geneva, Switzerland: IARC Scientif-ic Publications. 1980;1:192-246.

31. Willett W, Stampfer MJ. Total energy in\x=req-\take: implications for epidemiologic analyses. Am JEpidemiol. 1986;124:17-27.

32. Kennedy RH, O'Fallon WM, Melton LJ. Theimpact of referral bias on evaluation of clinical andepidemiologic features of cataract surgery. InvestOphthalmol Vis Sci Suppl. 1984;25:1.

33. Morgan L, Schwab IR. Age and visual acuityin men vs women undergoing cataract surgery.Invest Ophthalmol Vis Sci Suppl. 1984;25:2.

34. Kelsey JL, O'Brien LA, Grisso JA, HoffmanS. Issues in carrying out epidemiologic research inthe elderly. Am J Epidemiol. 1989;130:857-866.

35. Block G, Cox C, Madans J, Schreiber GB,Licitra L, Melia N. Vitamin supplement use bydemographic characteristics. Am J Epidemiol.1988;127:297-309.

36. Moss AJ, Levy AS, Kim I, Park YK. Use ofVitamin and Mineral Supplements in the UnitedStates: Current Users, Types ofProducts, and Nu-trients. Hyattsville, Md; National Center forHealth Statistics; 1989. Advance data from vitaland health statistics; No. 174.

37. Robertson JMcD, Donner AP, TrevithickJR. Vitamin E intake and risk of cataracts in hu-mans. Ann N Y AcadSci. 1989;570:372-382.

38. Jacques PF, Hartz SC, Chylack LT Jr,McGandy RB, Sadowski JA. Nutritional status inpersons with and without senile cataract: bloodvitamin and mineral levels. Am J Clin Nutr.1988;48:152-158.

39. Jacques PF, Chylack LT Jr, McGandy RB,Hartz SC. Antioxidant status in persons with andwithout senile cataract. Arch Ophthalmol. 1988;106:337-340.

40. Collman GW, Shore DL, Shy CM, Chec-koway H, Luria AS. Sunlight and other risk factorsfor cataracts: an epidemiologic study. Am J PublicHealth. 1988; 78:1459-1462.

41. Dolezal JM, Perkins ES, Wallace RB. Sun-light, skin sensitivity, and cataract. Am J Epide-miol. 1989;129:559-568.

42. Sperduto RD, Seigel D. Senile lens and se-nile macular changes in a population based sample.Am J Ophthalmol. 1980;90:86-91.

43. Hanna C, Dreis MW, Cosgrove KW. Cata-racts associated with allopurinol therapy. Am JOphthalmol. 1982;94:137-140.

44. Lerman S, Megaw JM, Gardner MS. Allopu-rinol therapy and cataractogenesis in humans. AmJ Ophthalmol. 1982;94:141-146.

45. Lerman S, Megaw JM, Fraunfelder FT.Further studies on allopurinol therapy and humancataractogenesis. Am J Ophthalmol. 1984;97:205\x=req-\209.

46. Jick H, Brandt DE. Allopurinol and cata-racts. Am J Ophthalmol. 1984;98:355-358.

47. Clair WK, Chylack LT Jr, Cook EF, Gold-man L. Allopurinol use and the risk of cataractformation. Br J Ophthalmol. 1989:173-176.

Downloaded From: by a Western University User on 09/12/2018