In our US population-based study using the NHANES database, we found that AS, cigarette smoking, and higher age were positively associated to self-reported uveitis. Its main strengths include the wide diversity of the database in terms of geographic characteristics and socioeconomic status and its representation of the national noninstitutionalized population, making it the first nationwide study in the country comparing the presence of self-reported uveitis with different demographic and clinical factors.
For our analysis, we looked at known risk factors for the development of uveitis, such as, UC, and CD. It has been previously reported that HLA-B27-positive spondyloarthritides have the highest association with non-infectious uveitis in the adult population in North America and Europe [5, 13,14,15]. In our study, we found a statistically significant predominance of all three diagnoses in the uveitis group compared with the non-uveitis group, which supports the results from the previous literature.
Given that all analyses on our population-based study accounted for the multi-stage nature of NHANES, these results are nationally representative, and thus, we can extrapolate them in order to obtain an estimate of the actual prevalence of the disease in the country, which is of 5.4 per 1000 residents (95% CI 3.4–8.5/1000). This number differs from previous epidemiologic studies carried out by Thorne et al. [12] and Gritz and Wong [18], who reported an overall prevalence of 121 cases per 100,000 adults (95% CI 117.5–124.3) in the USA and 115.3 per 100,000 persons in Northern California, respectively. If we would only consider the patients from our database who reported having received treatment for uveitis (n = 19) as a way of confirming their diagnoses, the prevalence estimate in our population would still be higher than those previously mentioned (3.4/1000 persons; 95% CI 2.3–5.1/1000). A possible explanation could be that our study is not limited to a specific type of uveitis, as it is the one from Thorne et al., which mainly looked at the prevalence of noninfectious uveitis, and that it is based on the national population, as opposed to Gritz and Wong’s study, which may give more regional results based on Northern California population, and Thorne et al., who only studied privately insured patients, excluding an important part of society. Also, those two studies calculated the prevalence of the disease during a 1-year period, while we measured the lifetime prevalence of the participants, so age may be playing a role in this prevalence estimate as a risk factor for the disease.
Prior studies have shown that ocular inflammatory disorders have a higher incidence in women compared with men, particularly in women of childbearing age, and that this gender difference increases with increasing age [17]. Sex hormones and the presence of an extra X chromosome in women are thought to play important roles in the development of these immune-mediated diseases [17, 35, 36]. While this gender difference has been found in studies performed in several developed countries [5, 6], a reverse gender predilection has been noted in developing countries, such as India or Turkey, where a male predominance was found [14, 15, 19,20,21]. In our study which examined residents in the USA, a slight female predominance was found, although it was not statistically significant.
As previously mentioned, a positive association between age and the development of uveitis was found in our analysis, where the greater the age, the higher the odds of suffering from uveitis. This finding supports results obtained in the Pacific Ocular Inflammation Study carried out in a Hawaiian population and Gritz and Wong’s Northern California Epidemiology of Uveitis Study [1, 18]. Several other studies have reported the highest incidence of uveitis in persons between the ages of 20 and 40 years [10, 14, 19,20,21,22,23]. This differs from our results, which showed a higher mean age of 53 ± 13 years old, with most of the patients concentrated between the ages of 40 to 60 years. However, our population participated in a survey that asked “Have you ever been diagnosed with uveitis?”, which may be the reason for this discordance, since the older the participants, the higher the chance for them to have had uveitis in the past. The age of diagnosis was not an available variable on the dataset. Although age has been a risk factor that has not been unanimously related with uveitis, our study adds evidence of age being a related risk factor in a large-scale population-based study.
Our analysis of ethnicity did not reveal any specific ethnic group predominance, which may be partly due to the small sample size or a true lack of correlation. Our results contrast with other studies that have shown slight group differences, including the Pacific Ocular Inflammation Study which found a higher incidence of uveitis in the white population when compared with Asians and Pacific Islanders [1]. Furthermore, Nguyen et al. reported a higher incidence of uveitis in African American patients with inflammatory bowel disease in comparison with Caucasian patients [37].
A link between vitamin D deficiency status and uveitis has also been previously reported. A study done in China in 2010 looking at patients with Vogt–Koyanagi–Harada (VKH) disease described a possible association between VKH and vitamin D deficiency [28]. A more recent study carried out in the Massachusetts Eye and Ear Infirmary determined that there is a correlation between vitamin D deficiency and non-infectious uveitis [27]. All patients in our study population had normal levels of vitamin D in serum samples, which differs from the results of these two previous studies.
Our present study also demonstrates that a positive history of cigarette smoking is associated with higher odds of a uveitis diagnosis. This higher rate of uveitis was obtained through the former smokers’ group, leading us to theorize that patients stopped smoking when they were diagnosed with the disease due to their doctor’s recommendation. This association between uveitis and tobacco smoking supports the results obtained by Lin et al., who concluded that cigarette smoking is a risk factor for all anatomic types of uveitis as well as infectious uveitis, and Yuen et al., who found that cigarette smoking had a strong association mainly with noninfectious uveitis [24, 25]. It has also been found that tobacco smokers have a higher risk for active disease, the need for higher doses of steroids in order to control their uveitis, an earlier development of the disease, higher incidence of bilateral ocular inflammation, higher odds of disease recurrence, and higher rates of cystoid macular edema as a complication [26, 38, 39]. This positive association between tobacco smoking and uveitis, however, may not apply to all forms of uveitis because tobacco smoking does not have any negative ocular effects in patients with Behçet’s disease [40].
Cigarette smoking is the leading preventable cause of death in the USA [41]. It causes more than 480,000 deaths each year in the USA and accounts for nearly one in five deaths [41,42,43]. Chronic exposure to tobacco smoke or nicotine has an effect on the immune system, increasing the release of pro-inflammatory cytokines such as TNF-α, IL-1, IL-6, IL-8, and GM-CSF, as well as decreasing the release of anti-inflammatory ones, such as IL-10. These findings support the theories that cigarette smoking has a positive correlation with the development of many chronic inflammatory and autoimmune diseases, although the causal and pathophysiological relationship still remains unclear [44,45,46,47,48,49,50,51].
Despite the advantages of using such a large database, our study has several limitations intrinsic of the NHANES design. Some of these weaknesses are the small sample size of those with uveitis, which may have underpowered the study and thus prevent us from identifying well-known disease predictors, such as gender; the fact that it is generalizable only to the noninstitutionalized US residents, excluding the institutionalized population in the country; and the cross-sectional nature of the study, which precludes the analysis of the pathology over time. This can also lead to an overestimation of the prevalence and risk factors.
Moreover, self-report of uveitis by the participants may refer to either an acute isolated attack or to chronic disease, which is not specified in the questionnaires. The questionnaire also does not ask the participants to specify the location of the uveitis (anterior, intermediate, posterior, or panuveitis). It is also possible that, since the diagnosis of uveitis was not verified by an ophthalmologist, patients may have confused their diagnosis with other types of eye inflammation. The fact that only 19 out of 27 patients with uveitis reported having received treatment raises suspicion on the accuracy of the diagnosis, since it is not the standard of care to leave uveitis untreated. On the other hand, uveitis may not have been diagnosed in all participants. Some participants may not have been aware of a uveitis episode and therefore not reported it. Therefore, recall bias may influence the results and add more uncertainty to the prevalence estimate. A study published in 1991 suggested that self-reported ocular disease should not be the only source of information for prevalence estimates and that clinical determinations are necessary [52], and a more recent paper trying to determine the validity of self-reported eye disease in a Latino population concluded that the sensitivity is generally very low, so there is a high chance that patients who did not report the disease actually had it, while the specificity was high for all the diagnoses [53]. However, none of these two studies asked for a diagnosis of uveitis, and only self-reported cataract, glaucoma, diabetic retinopathy or macular degeneration were studied. There is no evidence in the literature of the validity of self-reported uveitis for epidemiologic studies.
Other limitations of the study may be that when looking at the smoking history, we just focused on the general categories of whether patients were never, former, or current smokers; we did not evaluate specific quantities such as number of cigarettes smoked per day, years of smoking, or the timing of smoking in relation to the diagnosis of uveitis; and the fact that correlation between two variables does not prove causation [54, 55]. Moreover, the 95% confidence intervals of multivariable models were wide, indicating that there may be some imprecision in the effect estimates.
