Regional Variation of and Association of US Birthplace With Vitiligo Extent

Importance  Little is known about population-based risk factors and regional differences for vitiligo.

Objective  To determine the impact of place of birth and residence on vitiligo extent.

Design, Setting, and Participants  A prospective questionnaire-based study using an online questionnaire with 2786 adults (72.2% of whom resided in the United States) with a history of physician-diagnosed vitiligo.

Exposures  Regions of birth and residence.

Main Outcomes and Measures  Body surface area (BSA) of vitiligo lesions.

Results  Patients with vitiligo who were born outside the United States had lower odds of vitiligo-affected BSA greater than 25%, even after controlling for race/ethnicity, sex, and current age (logistic regression; adjusted odds ratio [aOR], 0.57 [95% CI, 0.46-0.60]). Birthplace in all continents was associated with lower odds of affected BSA greater than 25% than was birthplace in North America. Adults born outside the United States had less affected BSA whether they resided inside (aOR, 0.58 [96% CI, 0.41-0.81]) or outside the United States (aOR, 0.60 [95% CI, 0.48-0.76]). Birthplace and residence at latitudes closer to the equator were associated with lower rates of affected BSA greater than 25% (P ≤ .002). The prevalence of affected BSA greater than 25% varied greatly by state of residence (range, 27.3% in Maryland to 100% in North Dakota, South Dakota, and Wyoming) (global Moran index = 0.37; P < .001; G statistic = 0.62; P < .001). Spatial regression models that controlled for the regional variation were constructed and confirmed that birthplace outside the United States was associated with lower odds of affected BSA greater than 25% (aOR, 0.61 [95% CI, 0.45-0.83]) but not race/ethnicity.

Conclusions and Relevance  There was significant statewide and intercontinental variation for rates of extensive vitiligo. These results suggest that previously unrecognized regional environmental risk factors, especially early in life, play an important role in vitiligo. Additional studies are needed to confirm these early findings and identify such risk factors.

Trial Registration  clinicaltrials.gov Identifier: NCT01401374

Vitiligo is a condition typified by loss of pigment cells and is believed to be primarily autoimmune in nature.¹ The worldwide prevalence of vitiligo is estimated to be 0.5% to 2% of the population; still, little is known about the epidemiology of vitiligo in the United States and internationally.² Furthermore, population-based risk factors for vitiligo have rarely been explored. In particular, environmental risk factors for vitiligo and vitiligo severity are not well established. We hypothesized that different regional environmental exposures have an impact on the distribution of vitiligo extent in the United States and internationally.

As the populations of developed countries increase in age, so does their risk for chronic diseases.³ This suggests that 1 or more exposures in non-Westernized or developing countries may protect against chronic disease. Indeed, a previous study⁴ found that foreign-born American children had dramatically lower prevalence of eczema and allergic disease than US-born American children. Furthermore, low prevalence of vitiligo (0.18%) has been reported in children in the Sinai Desert, suggesting a possible role for latitude in the development of the disease.⁵ We hypothesized that foreign birthplace and/or residence may similarly be associated with lower frequency and/or severity of vitiligo. The goal of this study was to determine whether there is regional variation of vitiligo extent and whether vitiligo extent is lower in adults living outside the United States and foreign-born American adults.

This prospective questionnaire-based study was approved by the institutional review board at St Luke’s-Roosevelt Hospital Center for adults (≥18 years old, male or female) with vitiligo. Data were deidentified, confidential, and posed no risk to participants; informed consent waiver was granted by the institutional review board at St. Luke’s-Roosevelt Hospital Center. Participants were not compensated. The survey was validated as previously described.⁶ We also administered the survey to an additional 47 foreign-born patients from multiple countries of origin and found there were no significant differences across different groups. The questionnaire was uploaded to https://www.surveymonkey.com, and members of nonprofit support groups for vitiligo vulgaris were invited to participate by responding to the survey. The support groups included a large audience from the United States and more than 100 countries from Europe, Australia, Africa, Asia, and South America. Responses from initiation of the study in June 2010 through July 2012 were reviewed. The questionnaire took an average of 16 minutes (range, 10-28 minutes) to complete. Exclusion criteria were not being diagnosed as having vitiligo by a physician, vitiligo onset reported at birth (to exclude possible cases of nevus depigmentosus or pigmentary mosaicism), and incomplete survey responses.

Questionnaire items were developed to determine both novel risk and prognostic factors for vitiligo. The a priori aims of the questionnaire included identification of regional variation of vitiligo extent. The questionnaire consisted of items pertaining to demographics, phenotype of vitiligo, places of birth, and current residence (city, state, and country). There were 77 questions overall (55 closed questions and 22 open questions). The questions related to this study and response rates are listed in eTable 1 in the Supplement.

All data processing and statistics were performed with SAS software (version 9.3; SAS Institute Inc). Numerical responses were verified for noninteger or implausible values (eg, <0 or >100 years for age). States and countries were recoded into Federal Information Processing Standards’ 2-letter codes and were also manually reviewed for accuracy. Cities, states, and countries were converted into longitude and latitude via 3 approaches: (1) geocoding using SAS software (proc geocode) and manual lookup in (2) Google Maps and (3) Yahoo Maps. Distributions of absolute latitudes of birthplace and current place of residence were compared between adults with vitiligo-affected body surface area (BSA) of 1% to 25% vs those with more than 25% affected. χ² Tests were used to compare absolute latitude of ±0° to 24° vs ±25° to 90°.

Logistic regression models were constructed to examine the effects of country of current residence and birth on vitiligo-affected BSA (binary outcome). For many countries, there were few respondents, which precluded robust estimates of effect size and 95% CIs. Therefore, composite variables were created for residence and/or birthplace in the United States (binary) and for the respective continents (categorical). Adjusted odds ratios (aORs) were calculated for multivariate models that included race/ethnicity (white, African American/black, Hispanic, Asian/Southeast Asian/Indian, multiracial/other), sex (male or female), age (continuous), and duration of vitiligo (continuous).

Statewide prevalences of affected BSA greater than 25% were plotted using a color-coded heat map for lowest, intermediate, and highest tertiles, respectively (proc gmap procedure). Spatial autocorrelation cluster analysis was performed to characterize significant variability of affected BSA greater than 25% between states. The spatial autocorrelation statistics global Moran index and Geary C statistic and local Moran index G(i)d were used to evaluate for significant spatial variation. Significant spatial clusters encompass states with high prevalence of affected BSA greater than 25% that are surrounded by other states with similarly high prevalences. In contrast, spatial outliers encompass states with high prevalence that are surrounded by states with low prevalence. State maps of the United States were plotted with prevalences of affected BSA greater than 25%, as well as local Moran index G(i)d, that indicate regions with spatial clustering of similar prevalences.

Multiple logistic regression with random effects for spatial correlation and covariates was performed to determine the impact of residence and/or birthplace outside the United States (binary predictors) on affected BSA greater than 25% (proc glimmix procedure). For models of birthplace outside the United States, spatial random effects were modeled for country of birth, using latitude and longitude combinations. In contrast, models of residence outside the United States modeled spatial random effects for country of residence. An exponential spatial covariance structure was used, where the covariance between 2 observations depends on a Euclidean distance metric based on longitude and latitude at the county level. Covariates included race/ethnicity, sex, and age based on potential confounding and significant spatial autocorrelation for these variables. The aORs were calculated by including them in the full models. Sensitivity analyses were performed for patients who were diagnosed as having vitiligo when they were younger than 18 years or at 18 years or older.

Missing values were encountered from nonresponse of patients to various questions (see eTable 1 in the Supplement for response rates). Complete case analysis was performed, that is, missing values were ignored throughout the study. P < .05 was taken to indicate statistical significance for all estimates. However, the multiple dependent tests performed in this study increase the risk of falsely rejecting the null hypothesis. Therefore, P values near .05 should be interpreted with caution.

Data were collected on a total of 2960 respondents; 174 were excluded owing to age younger than 18 years or not having their vitiligo diagnosed by a physician, resulting in data analyzed for a total 2786 patients. Sex, age, race/ethnicity, duration of disease (years), and vitiligo-affected BSA for respondents are listed in Table 1.

In bivariate models, affected BSA greater than 25% was significantly associated with race/ethnicity, where African Americans had the highest (43.1%) and Asians had the lowest (26.8%) prevalences, respectively (P < .001) (Table 1). An affected BSA greater than 25% (P < .001 for all comparisons) were also associated with older age, female sex, and being diagnosed as having vitiligo before age 18 years. All of these were therefore considered as potential confounding variables and included in multivariate analyses.

A total of 2011 patients (72.2%) resided in the United States, while 775 (27.8%) resided outside the United States. Similarly, 1750 (62.8%) were born in the United States and 1036 (37.2%) patients were born outside the United States. Of those who were born outside the United States, 738 (71.2%) also resided outside the United States. Individual countries of residence and their relationship to affected BSA are presented in eTable 2 in the Supplement.

Overall, individuals with vitiligo who were born outside the United States had significantly lower odds of affected BSA greater than 25% (logistic regression; OR, 0.52 [95% CI, 0.44-0.61]; P < .001). This association remained significant in multivariate models that included race/ethnicity, sex, current age, and duration of vitiligo (P < .001). In bivariate models, patients who resided in any continent outside of North America had significantly lower odds of affected BSA greater than 25% (Europe: OR, 0.51 [95% CI, 0.38-0.68], P < .001; Australia: OR, 0.52 [95% CI, 0.30-0.88], P = .01; Africa: OR, 0.45 [95% CI, 0.26-0.79], P = .006; Asia: OR, 0.33 [95% CI, 0.23-0.47], P < .001; and South America: OR, 0.45 [95% CI, 0.25-0.81], P = .008) (Table 2 and Table 3). In multivariate models, all continents remained significant (P ≤ .04).

Similarly, patients with vitiligo who currently resided outside the United States had significantly lower odds of affected BSA greater than 25% (logistic regression; OR, 0.59 [95% CI, 0.49-0.70]; P < .001). This association remained significant in multivariate models that included, race/ethnicity, sex, current age, and age at vitiligo diagnosis (P < .001). In bivariate models, patients who resided in Europe (OR, 0.58 [95% CI, 0.46-0.7l]; P = .01), Asia (OR, 0.39 [95% CI, 0.27-0.57]; P < .001), and South America (OR, 0.14 [95% CI, 0.03-0.60]; P = .03) had significantly lower odds of affected BSA greater than 25% (Table 2 and Table 3). In multivariate models, however, only residence in Europe (P = .002) remained significant.

Adults with vitiligo who were born outside the United States had significantly less disease-affected BSA whether they resided inside (OR, 0.52 [95% CI, 0.40-0.67]; P < .001) or outside (OR, 0.52 [95% CI, 0.43-0.62]; P < .001) of the United States (Table 2 and Table 3). These associations remained significant in multivariate models that controlled for race/ethnicity, sex, current age, and duration of vitiligo (P ≤ .001). However, US-born adults with vitiligo who resided outside the United States did not have significantly different odds of affected BSA greater than 25% (P ≤ .60).

Given the intercontinental variation of BSA greater than 25% described in the previous subsection, we sought to determine whether affected BSA greater than 25% varied by latitude. Adults with affected BSA greater 25% were less likely to be born or reside in latitudes between −24° and 24° than those with BSA of 1% to 25%; that is, they were born or lived farther from the equator (birthplace: 7.1% vs 13.0%, P < .001; residence: 3.0% vs 5.5%, P = .002) (Figure 1).

Given the intercontinental variation of affected BSA greater than 25%, we sought to determine whether there is also statistically significant US statewide variation. The prevalence of affected BSA greater than 25% varied greatly depending on what state adults lived in within the United States (range, 27.3% in Maryland to 100% in North Dakota, South Dakota, and Wyoming) (Figure 2A). The global Moran index was 0.37 (P < .001; z score = 4.79), indicating clustering of vitiligo severity in nonrandom patterns. Similarly, the G statistic was 0.62 (P < .001; z score = −4.44), indicating clustering of higher rates of affected BSA greater than 25% in nonrandom patterns. Local G(i)d were maximal at several foci, including the northern Great Plains and southern states (Figure 2B). All patients with affected BSA greater than 25% had neighbors within a 1° radius of longitude and/or latitude with affected BSA greater 25% and all patients with affected BSA of 25% or less had neighbors with affected BSA of 25% or less. A total of 44 of 50 (88.0%) and 36 of 50 (72.0%) continental US states and districts, respectively, were bordered by at least 1 other state and/or multiple states with similar BSA ranking. Similar results were found when using longitude and latitude estimates from Google and Yahoo Maps (data not shown).

The distribution of race/ethnicity of participants did not significantly vary across states (P ≥ .13 for all races/ethnicities). However, the distribution of individuals born outside the United States varied greatly across states (range, 5.0% in Kentucky to 37.5% in Hawaii and Nevada). The global Moran index was 0.26 (P < .001; z score = 4.01), and the G statistic was 0.76 (P = .009; z score = −2.61), indicating clustering of foreign-born US adults in nonrandom patterns.

Given the significant statewide variation of affected BSA greater than 25%, we explored spatial regression models of vitiligo BSA greater 25%. Three different multivariate logistic regression models were created: (1) fixed-effects models of race/ethnicity, age, and sex; (2) spatial regression models of race/ethnicity, age, and sex; and (3) spatial regression models of US birthplace, race/ethnicity, age, and sex. In model 1, all 3 variables were significant predictors of affected BSA greater than 25%, including significantly lower odds of BSA greater 25% in Hispanic Americans (aOR, 0.61 [95% CI, 0.43-0.86]; P = .005) and Asian Americans (aOR, 0.42 [95% CI, 0.24-0.73]; P = .002) (Table 4). In model 2, however, race/ethnicity was no longer a significant predictor of affected BSA greater than 25% (P ≤ .25). In model 3, birthplace outside the United States was associated with significantly lower odds of affected BSA greater than 25% (aOR, 0.61 [95% CI, 0.45-0.83]; P = .001), whereas race/ethnicity was not (P ≤ .90). Furthermore, there were no significant 2- or 3-way interactions between US birthplace, race/ethnicity, and other demographic factors (P ≥ .12). Sensitivity analyses were performed for model 3, in patients who were diagnosed as having vitiligo when they were younger than 18 years or 18 years or older. There were no significant differences for fixed effects in any of the models (data not shown). However, models of patients diagnosed as having vitiligo when they were younger than 18 years were better fitting with more variance attributed to spatial effects.

The present study demonstrates that (1) adults with vitiligo born outside the United States have significantly lower self-reported BSA of vitiligo. (2) This was true whether the foreign-born adults lived inside or outside the United States. (3) The associations remained significant even after controlling for race/ethnicity, sex, current age, and age at vitiligo onset. (5) There were significant intercontinental differences of vitiligo BSA with significantly less extensive disease in adults born in South America, Europe, Australia, Africa, and Asia compared with those born in North America. (6) Vitiligo extent was inversely associated with latitudes of birth and residence closer to the equator. (7) We further found significant US statewide variation of vitiligo BSA with apparent “hot spots” of more extensive disease occurring in the northern Great Plains and southern states. (8) The associations of US birthplace and vitiligo extent remained significant even after controlling for the effects of spatial clustering of disease. These data indicate that birthplace in the United States is a previously unrecognized factor in the epidemiology of vitiligo. Previous studies of American and international cohorts with vitiligo demonstrated associations with vitamin D levels and comorbid autoimmune disease.^7-10 Vitamin D levels are inherently related to UV exposure and regional factors, including cloud cover, haze, altitude, and longitude.¹¹ Thus, the association between vitamin D levels and vitiligo suggest there are regional differences of disease prevalence and/or severity. To our knowledge, however, previous studies have not examined the role of birthplace in the epidemiology of vitiligo. The present study suggests that adults with vitiligo born in any continent outside of North American, particularly regions closer to the equator, have significantly less extensive vitiligo.

The protective effects of foreign birthplace against more extensive vitiligo are reminiscent of those found in eczema. Silverberg et al⁴ recently found that children born outside the United States have significantly lower prevalence of allergic disorders, including eczema, asthma, hay fever, and food allergies. The risk of allergic disease significantly increased after residing in the United States for 10 or more years. Together with the results of this study, birthplace in the United States seems to be a risk factor in the epidemiology of multiple inflammatory skin diseases. It is possible that the diagnosis and report of the vitiligo are more precise in the United States, which may account for the observed findings. However, this is less likely because the inverse association between foreign birthplace and vitiligo extent was significant, even in those who were currently residing in the United States. Furthermore, there were no significant differences of effect size between foreign-born adults residing inside or outside the United States. An epidemiologic study in the southern Sinai Desert, a region close to the equator, showed extremely low prevalence of vitiligo in children (0.18%), suggesting that UV exposure and other factors related to latitude may be a protective factor in vitiligo.⁵ The protective effects of foreign birthplace against eczema and allergic disease support the “hygiene hypothesis,” which suggests that that either infections or certain microbial exposures in early childhood may confer protection against atopic disorders.¹² It may be that such early-life exposures are similarly protective against other inflammatory and/or autoimmune disease. This notion is supported by our finding that residence in or outside the United States did not modify the protective effects of foreign birthplace against extensive vitiligo. Future studies are warranted to determine the role of birthplace and early-life environmental exposures on subsequent risk of vitiligo and other autoimmune disease.

In the present study, Hispanic American and Asian American adults were found to have less extensive vitiligo than white Americans in fixed-effect models. On the one hand, these associations were no longer significant after controlling for spatial autocorrelation of BSA across the United States. On the other hand, place of birth remained significant even after controlling for spatial autocorrelation, race/ethnicity, and other demographic factors. Moreover, there were no significant interactions between foreign birthplace and race/ethnicity, suggesting that the protective effects of foreign birthplace occur in all racial and ethnic groups. Previous studies suggested that there may be racial/ethnic differences of vitiligo, particularly in the emotional sequelae owing to the notable appearance in darker patients¹³ and the effect of vitiligo on racial identity,¹⁴ and superior response to tacrolimus¹⁵ and narrowband UV-B phototherapy with or without afamelanotide.¹⁶ The present study suggests that environmental factors related to regions of residence and birth may play an equally or even more important role than genetic and/or inherent factors related to race/ethnicity. To our knowledge, prior studies have neither identified nor explored the role of immigration patterns on vitiligo disease expression.

This study has several strengths, including being prospective, with a large diverse sample size, validated questionnaire for vitiligo extent, and the use of spatial autocorrelation and regression analysis to determine regional differences of vitiligo extent. The patient population includes a large sample of men and women and representation of various adult ages and vitiligo extent from the United States and internationally. The use of spatial autocorrelation and regression analyses allowed for identification of areas within the United States with higher rates of extensive vitiligo. Furthermore, incorporation of spatial analysis into regression models allowed for correction of biased estimates and spurious associations between race/ethnicity and vitiligo-affected BSA.

This study has several limitations. The invitation to participate in the survey was distributed by vitiligo support groups, which may have resulted in a selection bias toward patients with more severe vitiligo. The distribution of vitiligo-affected BSA in some countries may not be representative of their respective populations because of relatively small number of respondents from some countries and potential selection bias. To address these issues, comparisons between the United States and North America and abroad were limited to composite variables for regions and continents. History of physician-diagnosed vitiligo and vitiligo extent were self-reported and not assessed clinically. Nevertheless, the instrument used in this study for self-reported BSA was previously validated and found to be highly correlated with clinical assessment by a dermatologist.⁶ Moreover, self-report of diagnoses is ubiquitous in epidemiology and large-scale, population-based studies. Despite the large sample size, there were smaller sample sizes for patients from some international countries and regions. Therefore, we were able to assess only the effect sizes of foreign birthplace and residence at the continental level and not at the country or city level. Socioeconomic status was not assessed and may be a confounding factor. Finally, this study surveyed adults who were already diagnosed as having vitiligo. We were thus not able to determine the prevalence of vitiligo in the United States and abroad. Future US population-based studies are needed to determine the overall and statewide prevalences of vitiligo in the United States.

Adults born outside the United States seem to have less extensive vitiligo than those born in America. There was also significant statewide variation of vitiligo extent, which suggests that there are previously unrecognized environmental factors that drive the extent and severity of vitiligo in the United States and international populations. Additional large-scale clinical studies are needed to confirm these findings and to better understand how birthplace and early life exposures have an impact on the risk of vitiligo.

Accepted for Publication: April 11, 2014.

Corresponding Author: Jonathan I. Silverberg, MD, PhD, MPH, Department of Dermatology, Northwestern University, 680 Lakeshore Dr, Ste 1400, Chicago, IL 60611 (jonathanisilverberg@gmail.com).

Published Online: July 9, 2014. doi:10.1001/jamadermatol.2014.899.

Author Contributions: Drs J. I. Silverberg and N. B. Silverberg had full access to all the data in the study and take responsibility for the integrity of the data and accuracy of the data analysis.

Study concept and design: J. I. Silverberg, N. B. Silverberg.

Acquisition, analysis, or interpretation of data: J. I. Silverberg, Reja.

Drafting of the manuscript: All authors.

Critical revision of the manuscript for important intellectual content: J. I. Silverberg, N. B. Silverberg.

Statistical analysis: J. I. Silverberg, Reja.

Administrative, technical, or material support: N. B. Silverberg.

Study supervision: N. B. Silverberg

Conflict of Interest Disclosures: None reported.