|Home | About | Journals | Submit | Contact Us | Français|
To identify geographic and climatic risk factors associated with exfoliation syndrome (ES).
This is a retrospective, observational study of 626901 beneficiaries in a managed-care network throughout the continental United States (excluding California) under ophthalmic surveillance from 2001–2007. International Classification of Disease (ICD-9CM) codes (365.52, 366.11) were used to identify 3367 incident ES cases. We assessed the risk for ES by geographic latitude tier in the continental United States and then assigned state-level climatic data (e.g., ambient temperatures, elevation, sun exposure) to individuals based on their place of residence. Multivariable-adjusted Cox regression models were used to assess associations with the hazard of developing ES.
Compared with living in the middle geographic tier of the United States, residing in the northern tier (above 42°N) was associated with an increased hazard of developing ES (adjusted Hazard Ratio (HR)=2.14 [Confidence Interval (CI)=1.94–2.35]). Conversely, living in the southern geographic tier (below 37°N) was associated with a reduced hazard of ES (HR=0.83 [CI=0.75–0.93]). These associations did not materially change after excluding whites. After adjusting for joint environmental effects, for every one-degree increase in July high temperature, the hazard of ES decreased by 9% (HR=0.91 [CI=0.89–0.93]) and for every one-degree increase in January low temperature, the hazard decreased 3% (HR=0.97 [CI=0.96–0.98]. For each additional day of sunshine exposure annually, the hazard of ES increased by 1.5% (HR=1.02 [CI=1.01–1.02]) for persons living with average levels of other climatic factors.
Ambient temperature and sun exposure may be important environmental triggers of ES.
Exfoliation syndrome (ES) is an extracellular deposit disorder that causes significant ocular morbidity. It is the most common cause of secondary open-angle glaucoma.1 Elevated intraocular pressure commonly results because exfoliation material and uveal pigment lodge in the ocular outflow pathway, although glaucoma associated with ES is clearly multifactorial.2 In addition, ES is cataractogenic3 and cataract surgery in affected eyes can be fraught with complications,4 due largely to inherent zonular instability.5,6 In ES, grayish-white deposits are readily visible in the ocular anterior segment using minimal magnification. These aggregates consist of macromolecules mostly involved in basement membrane biosynthesis,7–11 although markers of complement activation, oxidative stress, ischemia, and inflammation are also found.12 In contrast to the ocular deposits in ES, electron microscopy is typically needed to demonstrate exfoliation material in visceral organs,13,14 which are more insulated from environmental influences such as ambient temperature, compared with the eye.15 Furthermore, while ES has non-ophthalmic manifestations such as hyperhomocystemia16–21 and sensorineural hearing loss,22–27 the evidence for other prominent systemic features is not entirely clear.28
In a landmark genome-wide association study, common variants in the gene LOXL1 (coding for lysyl oxidase-like 1) were associated with ES among participants in Iceland and Sweden, where ES is hyperendemic.29 Subsequent gene association studies confirmed these findings in persons with ES throughout the world. In the aggregate, these studies indicate that although LOXL1 risk genotypes are present in ≥ 92% of ES patients, they are also seen in ≥ 74% of controls,30 suggesting that other genetic or environmental influences contribute to ES.
ES prevalence varies worldwide from 0% to greater than 20%.31–33 Although some notable exceptions exist,34,35 point estimates of ES prevalence tend to increase with latitude in the Northern hemisphere, from 1% in Sri Lanka (7°N) to greater than 20% in Sweden (64°N).33 We analyzed a fully annotated health-care claims database with beneficiaries residing throughout the continental United States (U.S.), a region spanning 15° of latitude, to investigate latitude as a risk factor for ES. Furthermore, we assessed whether selected climatic variables are associated with ES.
The University of Michigan Institutional Review Board found this study, with its de-identified database, to be exempt from IRB review.
The i3 InVision Data Mart database (Ingenix, Eden Prairie, MN) contains detailed, fully de-identified records of all beneficiaries in a large, U.S. managed-care network. We used data on all beneficiaries in this database receiving any eye care from January 1, 2001, through December 31, 2007. This subset comprised patients with at least one International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9CM),36 code for an eye-related diagnosis (360–379.9) or Current Procedural Terminology (CPT)37 code for an eyerelated visit, diagnostic or therapeutic procedure (65091–68899 or 92002–92499), or other ICD-9CM or CPT code assigned by an ophthalmologist or optometrist during their time in the plan. For this patient subset, we had all medical claims (inpatient, outpatient, skilled nursing facility) for any medical conditions and detailed sociodemographic information. This data source has been used previously to identify risk factors associated with other medical conditions.38, 39
Because ES is strongly age-associated, we identified all individuals aged 60 years and older who were in the database for one or more consecutive years and had at least one visit to an eye-care provider. Individuals in the plan for < 365 days or with discontinuous plan enrollment were excluded (Figure 1). As continental U.S. geographical region (i.e., northern, middle, southern tiers) was the exposure of interest, California residents were excluded, because that state spans the southern and middle tiers (approximately 9° of latitude) and we did not know patients’ specific location within their state of residence. Because the numbers of Hawaiians and Alaskans were insufficient for subgroup analysis, persons from these states were excluded. Beneficiaries with an ICD-9CM billing code of 365.52 (pseudoexfoliation glaucoma) or 366.11 (pseudoexfoliation of the lens capsule) were classified as having ES.
Analyses were performed using SAS 9.2 software. Patient characteristics were summarized using means and standard deviations for continuous variables and frequencies and percentages for categorical variables. For all analyses, a p-value <0.05 was considered statistically significant.
Cox regression models were developed to compare the hazards of ES by the variables of interest, with adjustment for sociodemographic variables, ocular conditions, and systemic diseases.40 For the analyses, we used the first year of beneficiary enrollment in the plan as a look-back period. To capture incident ES cases, case identification commenced after 1 year of medical-plan enrollment. Patients receiving an ES diagnosis in their first year of enrollment were considered prevalent cases and were excluded. Patients were followed until development of the event (ES) or censoring (i.e., on leaving the plan or December 31, 2007, the study-period end date).
Patients’ age at event or censoring was determined. Using age as the time axis and region of residence as the key predictor of interest, the Cox model was left truncated at the age of index (1 year after medical-plan entry). Models were adjusted for sex, race, education, diabetes mellitus, hypertension, hyperlipidemia, obesity, myocardial infarction, peripheral vascular disease, systemic hypotension, skin cancer (surrogate for long-term sun exposure), migraine headaches, sleep apnea,cataract, pseudophakia/aphakia, diabetic retinopathy, age-related macular degeneration, and Charlson index score (an overall health measure)41 (eTable 1).
Using information in the database on beneficiaries’ state of residence at enrollment, we categorized patients according to residence in a northern tier above 42° North (CT, ID, MA, ME, MI, MN, MT, ND, NE, NH, NY, OR, RI, SD, VT, WA, WI, WY), middle-tier (CO, DE, IA, IL, IN, KS, KY, MD, MO, NJ, NV, OH, PA, UT, VA, WV), or southern-tier below 37° North (AL, AR, AZ, FL, GA, LA, MS, NC, NM, OK, SC, TN, TX) (Figure 2).42
Multivariable-adjusted Cox regression models were used to estimate the hazard of ES for each U.S. state of residence. Missouri was designated as the reference state because this middle-tier state has relatively uniform altitude and is close to the geographic center of the continental U.S. For each other state, patients’ hazard of developing ES relative to patients in Missouri was estimated.
We calculated the state-specific incidence of ES as the number of new cases in the state divided by the number of person-years at risk in which enrollees were under ophthalmic care. Poisson regression was used to model the ES incidence rate for all states using the number of incident cases per state as the dependent variable and the cumulative person-years at risk as the offset. For each state, we assessed the possible association between ES incidence and the following variables, using National Climatic Data Center data43: mean annual rainfall and snowfall (in inches), mean annual numbers of sunny days and days with precipitation, mean high temperature in July, mean low temperature in January, UV index, mean elevation (in feet) above sea level, and mean latitude/longitude.
Tests to check for multicollinearity were performed, and factors highly correlated with others were removed from the models. Only factors significant at p<0.05 by backward-selection procedures were retained. Next, based on the coefficients estimated by the Cox model, including the significant factors, we calculated the multivariable hazards of developing ES for residents of each U.S. state, relative to Missourians.
Of 626,901 patients who met the inclusion criteria, 3367 (0.54%) had at least one ES diagnosis. Of the 3367 ES patients, 2194 (65.2%) had ICD-9CM code 366.11 (exfoliation syndrome), 894 (26.7%) had ICD-9CM code 365.52 (exfoliation glaucoma), and 279 (8.3%) had both codes documented. The mean age (±SD) was higher among patients with ES than among other patients (71.7±6.8 vs. 68.1±6.8; p<0.001). A greater proportion of patients with ES were female (66%) than male, and were non-Hispanic white (92%) rather than another race (p<0.0001). The age-adjusted proportion of patients with cataract, pseudophakia/aphakia, macular degeneration, branch or central retinal venous occlusion, and hearing loss was higher among ES patients than among the others (p<0.0001 for each comparison) (Table 1). These associations between ES and age,32,33,44–47 lens status,3 retinal venous occlusive disease,48–50 and sensorineural hearing loss22,23,25–27 are consistent with previous studies in which patients were identified using standardized ophthalmic or histopathologic examination.
Residence in the northern geographic tier was associated with a 114%-increased hazard of ES (adjusted HR=2.14 [95% CI=1.94–2.35]), and in the southern tier a 17%-decreased hazard (adjusted HR=0.83 [CI=0.75–0.93]), compared with residence in the middle tier. To determine whether the observed regional differences are attributable to a disproportionate number of whites’ living in the northern tier, we performed a secondary analysis excluding white patients. The increased hazard among northern-tier residents remained significant (adjusted HR=3.27 [CI=2.17–4.92]); however, the hazard of ES among nonwhites did not differ between residents of southern-tier versus middle-tier states (adjusted HR=1.31 [CI=0.87–1.97]).
Residents of 25 of the 46 states had elevated hazards of ES relative to Missourians (Table 2). North Dakotans (adjusted HR=7.77 [CI=3.86–15.62]) and Minnesotans (adjusted HR=5.62 [CI=4.34–7.29]) had the highest hazards of ES. The majority of states in which patients’ likelihood of ES was considerably elevated were in the upper Midwest (IA, NE, MN, MT, ND, WI) and, to a lesser extent, New England (CT, MA, ME). Patients’ hazard of ES was elevated in 13 of the 18 northern-tier states (72%), compared with 5 of the 13 southern-tier states (38%) (AZ, FL, NM, OK, NC).
Poisson regression analysis revealed that four variables were associated with increased ES incidence: an increase in the annual number of sunny days (p=0.0001), decreased mean high July temperature (p<0.0001), decreased mean January low temperature (p<0.0001), and lower elevation above sea level (p=0.004) (Table 3).
Additional regression models assessed whether joint environmental effects were present, and two interactions were found: a positive interaction between sunny days and January low temperature (p<0.0001), and a negative interaction between sunny days and elevation above sea level (p<0.0001). After adjustment for the two interactions and other potential confounders, every 1-degree increase in July average temperature decreased the hazard of ES by 9% (HR=0.91 [CI=0.89–0.93]). Every 1-day increase in sunny days increased the likelihood of ES (HR=1.02 [CI=1.01–1.02]) at medium altitude (1053 feet) and medium average January temperature (23.5°F) (Table 4). This association was attenuated by higher altitude (p interaction<0.0001) and lower mean January temperature (p interaction<0.0001) (Table 4). While every 1-degree increase in mean January low temperature decreased the hazard of ES by 2.3% (HR=0.98 [CI=0.97–0.98]), the association was enhanced in locales with fewer sunny days (Table 4) (p interaction<0.0001). A 100-foot rise in altitude increased the hazard of ES by 2% (HR=1.02 [CI=1.01–1.03]) in locales with the fewest sunny days (164 days), but not in those with the most sunny days (Table 4) (p interaction<0.0001). Our findings did not materially change whether we adjusted for all skin cancers or all skin cancers besides melanoma, which unlike other skin cancers, is not necessarily related to chronic sun exposure (results not shown).
Little is known about whether geographic factors are independently associated with the development of ES. In our sample, individuals in northern states had considerably higher hazards of ES relative to those in middle and southern states, and persons living in upper-Midwestern states had substantially higher hazards of ES relative to Missourians. Our models demonstrate that greater sunshine exposure and lower ambient temperatures, in summer and winter, increase the likelihood of ES. The increased association with altitude was found only in states with relatively few sunny days.
The prevalence of ES is known to vary worldwide, with a gradient by latitude. Reported ES prevalences in Sri Lanka (latitude=7°N),51 South India (12°N),52 Pakistan (30°N),53 Greece (39°N),54 and Sweden (62°N)33 are 1.1%, 3.8%, 6.5%, 11.9%, and 23%, respectively. In two clinic-based U.S. studies, which might overestimate the burden of ES because of referral bias, the prevalences were low: 3% among adults aged 60 or older in North Carolina (latitude=35°N),46 and 1.4% in Louisiana (latitude=30°N).55 In our study the crude prevalences of ES in various U.S. states were considerably lower than previously reported estimates for Europe and Asia. Mississippi (32°N) had the lowest crude prevalence (0.23%), whereas Minnesota (45°N) had the highest (2.84%). The lower prevalences in our study have many possible explanations. Nevertheless, in our study, where patients’ residences spanned a latitude range of 15°, living in the northern continental U.S. tier was associated with an increased hazard for ES; residing in North Dakota was associated with the highest risk for ES relative to living in Missouri. Considering the effect of latitude was the same in whites and nonwhites suggests that a trend toward genetically predisposed Northern Europeans’ populating northern-tier states does not explain our findings. Colder temperatures in summer and winter months increased the hazard of ES. Many of the highest reported ES prevalences are from countries with cold mean temperatures. For example, ES prevalences among Icelanders, Finns, and Lapps are greater than 20%.32,34,56,57 One explanation may be that the extracellular deposits of ES represent a nucleation reaction58 that is prone to develop at lower temperatures. While the temperature in the vascular iris may be close to the core body temperature, the temperatures in the avascular ocular segments, such the anterior chamber and lens, may be susceptible to ambient temperatures.15 One notable exception to the cold-precipitation hypothesis was a study in which ES prevalence among Eskimos in Alaska, Canada, and Greenland was 0%.59 Perhaps in Eskimos a thicker iris and more abundant periorbital fat help to keep ocular temperatures high enough to prevent extracellular deposit formation.60 Alternatively, the finding could be related to cultural practices—e.g., style of dress or housing design—that may modulate climate impact.
The hazard of ES increased in states with more sunny days, and sunshine exposure appears to modify the associations with winter temperature and elevation. Studies have reported strong associations between ES and climatic droplet keratopathy,61 a condition associated with ultraviolet-light exposure.62 Furthermore, high ES prevalences have been found in populations with considerable sun exposure, including Australian Aborigines (latitude=27°S)63 and Navajo Indians (latitude=37°N).64 The cornea transmits ultraviolet rays,65 and ultraviolet radiation may add to the impaired elastogenesis caused by abnormal LOXL1 function, although further research is needed to confirm this.
The association between elevation and ES was modified by sunshine exposure. In locales with relatively few sunny days, higher altitude was associated with increased risk of ES, whereas the opposite was true for locales with more sunshine. The reason for this finding is not entirely clear and may be confounded by other variables. To date, little is known about the potential effect of altitude on ES prevalence. At 6%, the prevalence of ES was relatively high among 50 Navajo Indians, aged 60 years or older, living on an Arizona reservation at 36°N and an altitude of 1500 meters.64 These data suggest that high altitude may contribute to ES; more studies with detailed estimation of exposure to altitude are needed to better understand its impact.
In a study of 350 Aboriginies living in different regions of Australia63, ES was associated with lower latitudes (north of the 29th parallel south) and greater levels of total global radiation exposure. Temperature, relative humidity, evaporation rate, rainfall, sunlight, and ultraviolet radiation exposure were not significant. Direct comparisons between that study and ours are difficult because of differences in climatic factors between Australia and the U.S., the sociodemographic characteristics of the samples, and the covariates included for adjustment in the analyses. We did not have information on total global radiation levels for each US state to incorporate this factor into our analyses.
Using a large administrative database to investigate the environmental variables associated with ES has several benefits. The large number of ES cases provided ample power to study the relationship between geographic factors and ES. In addition, our sample is geographically and sociodemographically diverse. Clinic or hospital-based studies are affected by selective referral of severe cases to ophthalmic centers, but our sample was not limited to referral cases, or patients seen by specialists or at academic medical centers.
Our study also has several limitations. First, misclassification of ES likely existed in our database; however, our cases had known comorbid conditions, demographic features, and ocular characteristics consistent with ES. Second, we cannot rule out the possibility that differential detection of ES in northern tier versus southern tier states explains our results, although this seems unlikely. For example, residence in North Dakota—a state with no academic ophthalmology centers—was associated with the highest ES risk relative to Missouri, which has several such centers. Yet we still cannot rule out detection bias, whereby northern eye care providers are more prone to detect ES than practitioners in the other tiers. Third, the claims database contained no zip-code-level information on beneficiaries. Therefore, the average levels of each environmental factor for a given state used for these analyses may not precisely reflect individuals’ actual exposure. We minimized such misclassification by excluding residents of California, a state that spans 9° of latitude. Fourth, we knew neither how long beneficiaries lived in their state nor whether they generally spend most or all months of the year living there. For example, the increased hazard of ES associated with residing in Florida may reflect recent migration from a northern state, but such migratory trends would have driven the tier-related results to the null. Furthermore, considerable ancillary evidence indicates that sunshine is implicated in ES.61,63,66,67 In addition, we cannot capture the extent to which individuals residing in a given U.S. state are exposed to the environmental conditions characteristic of that state. Some enrollees spend significant time outdoors because of their occupation or hobbies, whereas others may have limited exposure to environmental conditions. This ecologic bias may affect our findings. Finally, all participants were U.S. residents with health insurance; our findings may be nongeneralizable to other populations and regions.
These data suggest that climatic factors may contribute to ES. More work exploring the relation between individual-level environmental exposures (adjusting for lifestyle choices that might modify the ocular impact of climatic factors) and ES is needed. Discovery of environmental factors linked to ES could lead to primary prevention measures for this condition.
Grant support: National Eye Institute K23 Mentored Clinician Scientist Award (1K23EY019511-01), R01 EY011671, RO1 EY015473, American Glaucoma Society Clinician Scientist Award, Blue Cross Blue Shield of Michigan Foundation, and from unrestricted grants from Research to Prevent Blindness (Pasquale, University of Michigan), Core Center for Vision Research NIH EY007003 (Reed).
None of the authors have any financial disclosures or conflicts of interest to disclose.
All the authors had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis
eTable 1 (online only): International Classification of Diseases 9th Revision, Clinical Modification Codes Used in the Analysis