Schematic showing participationin the Beaver Dam Eye Study.
The 5-year incidence of dry eyeby age and sex in the Beaver Dam Eye Study. Numbers constitute the samplesizes.
Moss SE, Klein R, Klein BEK. Incidence of Dry Eye in an Older Population. Arch Ophthalmol. 2004;122(3):369-373. doi:10.1001/archopht.122.3.369
To estimate the 5-year incidence of dry eye and to examine its associationwith risk factors.
The population of Beaver Dam, Wis, that was 43 to 84 years of age (n= 5924) was examined in the 1988-1990 (n = 4926), 1993-1995 (n = 3722), and1998-2000 study phases (n = 2962). At the 1993-1995 examination, when dryeye data were first collected, and the 1998-2000 examination, 2783 subjectsparticipated, and 44 were interviewed. Of these, 2802 provided dry eye history.The incidence cohort consisted of 2414 subjects not reporting dry eye in the1993-1995 examination. Risk factor information was ascertained at the 1993-1995examination and included demographics, medical history, cardiovascular diseaserisk factors, medications, and lifestyle factors.
During the 5-year interval between examinations, a history of dry eyedeveloped in 322 of 2414 subjects, for an incidence of 13.3% (95% confidenceinterval [CI], 12.0%-14.7%). Incidence was significantly associated with age(P<.001). After adjusting for age, incidence wasgreater in subjects with a history of allergy or diabetes, who used antihistaminesor diuretics, and with poorer self-rated health (P<.05).Age-adjusted incidence was less in subjects using angiotensin-converting enzymeinhibitors or consuming alcohol (P<.05). It wasnot significantly associated with sex, blood pressure, hypertension, serumtotal or high-density lipoprotein cholesterol level, body mass index, historyof arthritis, gout, osteoporosis, cardiovascular disease, thyroid disease,or smoking, and use of caffeine, vitamins, antianxiety medications, antidepressants,calcium channel blockers, or anticholesterolemics.
Incidence of dry eye is substantial. However, there are few associatedrisk factors. Some drugs (eg, diuretics and antihistamines) are associatedwith a greater risk, whereas others (angiotensin-converting enzyme inhibitors)are associated with lower risk.
Dry eye syndrome is a common source of great discomfort in the elderlypopulation, and it can seriously affect quality of life.1- 3 Itsmanagement can be a frustrating experience for patients and their eye careproviders. It has been shown to be associated with rheumatoid arthritis andother autoimmune diseases.4 Several studieshave reported on the prevalence of dry eye and have examined these and otherrisk factors.5- 10 However,because these studies are cross-sectional, they can demonstrate only thata factor is associated with dry eye and not whether the factor precedes thedry eye. This is important because people with dry eye may modify certainbehaviors, such as diet and drug use, in response to the condition.
Therefore, longitudinal studies of dry eye are needed. We previouslyreported on the prevalence of dry eye in the Beaver Dam Eye Study.9 We are now able to examine the incidence of dry eyein the same cohort 5 years later. Thus, the purpose of this report is to estimatethe 5-year incidence of dry eye in an elderly population and to examine itsassociation with various risk factors.
The methods and procedures used to identify and examine the Beaver DamEye Study population have been previously published.11,12 Briefly,a private census of Beaver Dam, Wis, was conducted to identify residents aged43 to 84 years (n = 5924). Figure 1 presentsthe subsequent examination and follow-up history of this cohort through the1988-1990, 1993-1995, and 1998-2000 study examination phases. At each phase,all attempts were made to examine the participants in the same order. Reasonsfor nonparticipation and comparisons between participants and nonparticipantsat the 1988-1990 and 1993-1995 examinations have been reported elsewhere.11,12 Information on the presence of dryeye was first collected at the 1993-1995 examination, which is the startingpoint for incidence of dry eye. Thus, analysis is based on the 2827 participantsin the 1993-1995 and 1998-2000 examinations. Comparisons of these subjectswith the 513 who had died and the 382 who did not participate are presentedin Table 1.
All examinations have followed similar protocols, which were approvedby our institutional human subjects committee. Informed consent was obtainedfrom each participant at each examination. The examination included a medicalhistory questionnaire; measurement of height, weight, and blood pressure;determination of refractive error and visual acuity; dilation of the pupils;stereographic color fundus photography for evaluation of age-related maculopathy;slitlamp and retroillumination photography of the lenses for evaluation ofcataract; and collection of urine and blood samples for a series of standardlaboratory tests.
Systolic and diastolic blood pressures were recorded as the averageof 2 measurements. Hypertension was defined as a systolic blood pressure of160 mm Hg or greater, a diastolic blood pressure of 95 mm Hg or greater, ora history of hypertension with use of antihypertensives. Body mass index wasdefined as weight in kilograms divided by the square of height in meters.Subjects were considered to have diabetes if they reported a history of diabetesmellitus; treatment with insulin, oral hypoglycemic agents, or diet; or receivinga new diagnosis during participation in the study. The criterion for diagnosiswas a glycosylated hemoglobin value of greater than 2 SDs above the mean fora given age-sex group and a random blood glucose level of greater than 200mg/dL (>11.1 mmol/L). Arthritis, fractures, osteoporosis, gout, thyroid disorder,and stroke were determined by history. A history of cardiovascular diseasewas defined as a history of angina, heart attack, or stroke. Participantswere asked to bring to the examination all prescription and over-the-countermedications that they were regularly taking. The examiner asked whether therewere other medications being taken that were not brought. If there were, theparticipant was asked to contact the examiner later with the medication name.Participants, their physicians, and their pharmacies were contacted when necessaryto verify medication use. The name of each drug was entered into the databaseand assigned a code reflecting that of the American Hospital Formulary Servicefor each active ingredient.13 Heavy drinkingwas defined as current or past consumption of 4 or more servings of alcoholicbeverages daily. The average weekly consumption of alcohol in grams was computedas the sum of alcohol from each 0.355-L (12-oz) serving of beer, 0.118-L (4-oz)serving of wine, and 0.044-L (1.5-oz) serving of liquor or distilled spirits.Each serving of beer, wine, and liquor was considered to contain 12.96, 11.48,and 14.00 g of alcohol, respectively. A current or ex-smoker was an individualwho had smoked at least 100 cigarettes in his life. Pack-years smoked wascomputed as the number of packs (20 cigarettes) smoked each day times thenumber of years smoked. The average daily consumption of caffeine in milligramswas computed as the sum of caffeine from each 0.237-L (8-oz) serving of brewedcoffee (103 mg), instant coffee (57 mg), hot or iced tea (36 mg), hot chocolate(6 mg), and caffeine-containing soda (46 mg).
History of dry eye was not determined at baseline. At the 5- and 10-yearfollow-up examinations, the presence of dry eye was determined by means ofsubject self-report. Dry eye was defined as a positive response to the question,"For the past 3 months or longer, have you had dry eyes?" For subjects needingfurther prompting, this was described as a "foreign-body sensation with itchingand burning, sandy feeling, not related to allergy." Because history of dryeye was not obtained until the 1993-1995 examination, that juncture is regardedas the baseline for these analyses. All covariate information pertains tothat point. Incidence of dry eye was defined as a positive report of dry eyeat the 10-year examination or interview in persons who did not report havingdry eye at the 5-year examination. Differences in rates of incidence weretested for statistical significance by means of Mantel-Haenszel procedures.14 Age-adjusted incidence rates were computed by thedirect method.15 Proportions of 0.426, 0.306,0.208, and 0.060 for the groups aged 48 to 59, 60 to 69, 70 to 79, and 80to 91 years, respectively, were used for the standard population. These proportionsreflect the age distribution of the set of participants at risk for incidenceof dry eye. Logistic regression was used to examine the association of severalvariables with the incidence of dry eye.
Among the 2827 subjects who participated in the 1993-1995 and 1998-2000examinations, 388 had reported having dry eye in the 1993-1995 examination,and 25 were missing dry eye information at 1 or both examinations (Figure 1). This left 2414 subjects at riskfor incidence of dry eye. Among the 3722 subjects participating in the 1993-1995examination, the presence of dry eye (n = 534) at that time had no effecton the age- and sex-adjusted odds of dying before the 1998-2000 examination(odds ratio [OR], 1.01; 95% confidence interval [CI], 0.77-1.32).
Among the 2414 subjects at risk for incidence of dry eye, age variedfrom 48 to 91 years with a mean ± SD of 63 ± 10 years. Men constituted44% of the cohort, and 99% were white. Dry eye was reported in 322 of 2414subjects at risk, for a 5-year incidence of 13.3% (95% CI, 12.0%-14.7%). Incidenceof dry eye increased significantly with age (P<.001),from 10.7% in subjects aged 48 to 59 years to 17.9% in those 80 years or older(Figure 2). The incidence of dryeye was also higher overall in women (14.7%) compared with men (11.7%) (P = .04; Figure 2).However, after adjusting for age, the difference was no longer statisticallysignificant (P = .06).
Because many of the risk factors examined for their associations withincidence of dry eye are correlated with age, further analysis was age adjusted. Table 2 presents the results of this analysis.Persons with diabetes or a history of allergy reported a higher 5-year incidenceof dry eye than those without. The use of certain medications was also associatedwith incidence of dry eye. Persons using antihistamines, diuretics, or steroidshad a higher incidence, whereas those using angiotensin-converting enzyme(ACE) inhibitors had a lower incidence. The differences in incidence betweensubjects not taking or taking antianxiety drugs, antidepressants, calciumchannel blockers, anticholesterolemics, multivitamins, and hormone replacement(women only) were not statistically significant (data not shown). There wereno significant differences between subjects with (n = 388) and without (n= 2414) dry eye in the 1993-1995 examination with respect to discontinuinguse of a drug if they were using it or beginning to use it if they were not,with 1 exception. People with dry eye in the 1993-1995 examination were morelikely to begin using antidepressants by the 1998-2000 examination than thosewithout dry eye (11.8% vs 5.3%; P<.001).
Additional significant factors in the age-adjusted analysis (Table 2) included alcohol consumption.Persons consuming alcohol in any amount were less likely to report dry eyeincidence than nonusers. Also, subjects reporting themselves to be in poorerhealth compared with others their age had a higher incidence. Persons witha higher total cholesterol level had a higher age-adjusted incidence of dryeye, but this association did not reach statistical significance (P = .06). Hypertension and smoking history were also not associatedwith incidence of dry eye (Table 2).
Several additional characteristics were examined for associations withincidence of dry eye that were not statistically significant. These includedbody mass index; systolic and diastolic blood pressure; high-density lipoproteincholesterol level; ratio of total to high-density lipoprotein cholesterol;leukocyte count; hematocrit level; a history of gout, fractures, or osteoporosis;pack-years smoked; caffeine consumption; and a history of heavy drinking (datanot shown).
Some factors known to be associated with prevalence of dry eye werenot found to be associated with incidence. These included arthritis and thyroiddisease (Table 2). When we examinedthe effect of dry eye on the incidence of a history of arthritis and thyroiddisease, we found that those with dry eye in the 1993-1995 examination (n= 388) did not have a significantly different incidence of arthritis comparedwith people without (n = 2414) (27.5% vs 21.3%; P =.08). The corresponding incidences of thyroid disease were also not significantlydifferent (3.9% vs 4.9%; P = .47). However, in peoplewith (n = 322) compared with people without (n = 2092) incidence of dry eye,arthritis was more likely to develop (28.3% vs 20.3%; P = .01), but not thyroid disease (5.2% vs 4.8%; P = .76).
To determine which factors were independently associated with incidenceof dry eye, logistic regression models were developed. Age was included inevery model. Other variables were selected in stepwise fashion. Subjects whowere missing information for any of the variables are excluded from the analysis.The results of this procedure are presented in Table 3. In addition to older age, subjects with poorer self-ratedhealth, with a history of allergy, or who used diuretics were at higher riskfor incident dry eye. Those using ACE inhibitors were at lower risk. Whenthe use of antihistamines and corticosteroids is added to the model, the associationof history of allergy became weaker (OR, 1.30; 95% CI, 0.98-1.72), suggestingthat the medications taken for allergy may be the material factor.
Large epidemiological studies of dry eye are few. In addition, all previousobservational studies of dry eye have been cross-sectional or prevalence studies.5- 10 Longitudinalor incidence studies have the advantage compared with cross-sectional studiesin that the risk factors were observed before the occurrence of dry eye. Therefore,any risk factors discovered are more (although not necessarily) likely tobe in the causal pathway. This report is, to our knowledge, the first to examinethe incidence of dry eye. We found the overall 5-year incidence of dry eyeto be 13.3%.
However, as a consequence, few studies are available for direct comparisonwith ours. Some report an increase in prevalence of dry eye with age,6,10 but this is not consistent.5 Our incidence results agree with those of the formerstudies. However, as seen in the multivariable analysis, the effect of agemay be attenuated when other confounding factors such as general health ormedication use are considered. Also, some prevalence studies report more dryeye in women,6,9 but, again, thisfinding is not universal.5,10 Wenoted a somewhat greater incidence in women, but after controlling for age,the difference between women and men was not statistically significant. Thehigher prevalence of dry eye in women may be a result of men with dry eyenot being observed because they have died.
There are a number of inconsistencies between results for incidenceof dry eye and prevalence results we presented earlier9 andother prevalence studies. For instance, prevalence studies, including theprevalence phase of this study, have noted an association of arthritis andthyroid disease with dry eye.6,9 However,we failed to find a relationship between a history of arthritis or thyroiddisease and the incidence of dry eye. We can only speculate as to why thisis so. First, it may be that dry eye precedes arthritis and thyroid diseaseor that they occur more or less together. We observe this in the case of arthritis,but not thyroid disease. Also, we are not able to distinguish the type ofarthritis present. We might observe a stronger relationship if we could focuson rheumatoid arthritis.
It is generally believed that certain drugs are associated with prevalenceof dry eye.16- 18 Theseinclude diuretics, antihistamines, and psychotropics. Except for diureticsand to a lesser extent antihistamines, we found no associations of these drugsor other commonly used drugs such as calcium channel blockers and anticholesterolemicswith the incidence of dry eye. Also, except for the antidepressants, we foundno predilections to change medications as a function of earlier dry eye status.In the case of antidepressants, people with dry eye at the 1993-1995 examinationwere more likely to be taking an antidepressant 5 years later than peoplewithout dry eye. In a cross-sectional study, this would result in the appearanceof a relationship with prevalence, even if there was no relationship withincidence. However, this explanation would not apply to antianxiety agents.
We found a protective effect for the use of ACE inhibitors. To our knowledge,this is the first report of any relationship between these drugs and dry eye.Other studies have suggested that ACE inhibitors have anti-inflammatory effects.19,20 We can only speculate as to whetherthis is the basis for the connection. Further study may be worth pursuing.
There are some limitations to this study. First, we relied on self-reportsof the study subjects to determine the presence of dry eye with no objectivetesting. However, the tests commonly performed are notoriously lacking insensitivity and specificity.6,21 Thus,we believe allowing the subjects to be the arbiters of their own signs andsymptoms to be valid for this somewhat nebulous condition. In addition, inthe prevalence results, several well-known risk factor associations with dryeye were found, supporting the validity of this approach.9 However,we consequently are not able to differentiate between tear-deficient and evaporativedry eye. Risk factors for these 2 entities may differ. This might weaken anyassociations that are present. Differences between this study and others inhow dry eye is defined also make it difficult to contrast findings among them.However, this would affect quantitative comparisons more than comparisonsof general associations between risk factors and dry eye. The 5-year intervalbetween examinations may also be a limitation. The 5-year incidence reportedherein may not be a true cumulative incidence because occurrences of dry eyebetween examinations that had resolved would not be detected. In addition,the 5-year interval may be too long a period to attempt to associate somerisk factors with the outcome. The result of these circumstances would mostlikely be a bias toward the null hypothesis and underestimation of incidence.Furthermore, as an older population, the cohort has experienced attritiondue to death before and after the determination of dry eye. This has the potentialto introduce bias if the risk factors and dry eye are both associated withmortality. However, this is not likely, as we do not find dry eye to be associatedwith mortality. Again, the effect would at most be a diminution of relationships.In addition, as the study cohort is a middle-class, white population, theresults of this investigation may not be applicable to other groups of differingethnicity or socioeconomic class. However, because it is population based,our study is more likely to be representative than a clinic-based cohort.Finally, because of the number of associations examined, some statisticallysignificant results may be due to chance. Thus, it is important to considerthe reported results not in isolation, but in a context that includes otheravailable evidence and biological plausibility or mechanisms.
Incidence of dry eye is common in the older population, occurring ata rate of 13.3% during 5 years in persons aged 48 to 91 years. There are fewindependently associated risk factors, especially those subject to intervention.However, some drugs, such as diuretics and antihistamines, are associatedwith increased risk for dry eye, whereas ACE inhibitors are associated withdecreased risk. This latter association deserves further research. Perhapsof equal significance are the relationships that were not found. Factors suchas arthritis and thyroid disease, which are generally regarded as being associatedwith dry eye and have been confirmed by prevalence studies, were not foundto predict incidence. Nevertheless, this in no way invalidates the cross-sectionalresults, for these may simply be contemporaneous events.
Corresponding author: Scot E. Moss, MA, Department of Ophthalmologyand Visual Sciences, University of Wisconsin–Madison, 610 N Walnut St,454 WARF, Madison, WI 53726-2336 (e-mail: firstname.lastname@example.org).
Submitted for publication June 12, 2003; final revision received October6, 2003; accepted November 17, 2003.
This study was supported by grant EY06594 from the National Institutesof Health, Bethesda, Md (Drs R. Klein and B. E. K. Klein).