Dietary intake of vitamin D during adolescence and risk of adult-onset systemic lupus erythematosus and rheumatoid arthritis
Vitamin D has immunomodulatory properties with potential etiologic implications for autoimmune diseases. The relevant exposure time during which vitamin D may influence disease risk is unknown. Our objective was to examine the relationship between reported vitamin D intake during adolescence and adult-onset rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE) incidence in prospective cohort studies of women, the Nurses' Health Study (NHS) and the Nurses' Health Study II (NHSII).
Food frequency questionnaires concerning high school diet completed by 73,629 NHS (1986) and 45,544 NHSII (1998) participants were used to calculate nutrient intakes during adolescence. Incident RA and SLE cases prior to 2006 (NHS) and 2007 (NHSII) were confirmed by medical record review. Cox proportional hazards models calculated relative risks and 95% confidence intervals of incident RA and SLE according to quintile cutoffs of vitamin D intake. Age- and calorie-adjusted and multivariable-adjusted (including sun exposure factors) analyses were completed. Random-effects models were used to meta-analyze estimates of association from the 2 cohorts.
Incident RA was confirmed in 652 NHS and 148 NHSII participants and SLE was confirmed in 122 NHS and 54 NHSII participants over a mean followup time of 351 months (NHS) and 209 months (NHSII). Age- and calorie-adjusted and multivariable-adjusted models did not show significant associations between adolescent vitamin D intake and risk of adult-onset RA or SLE.
We did not find associations between adolescent dietary vitamin D intake and adult RA or SLE risk among NHS and NHSII women, suggesting that other time periods during the life course should be studied.
Vitamin D has been implicated as a potential etiologic factor in several related autoimmune diseases, including multiple sclerosis (MS), rheumatoid arthritis (RA), and systemic lupus erythematosus (SLE) (1–6). The large, prospective Iowa Women's Health Study of women ages 55–69 years observed an inverse association between both dietary and supplemental vitamin D intake and RA risk (7), yet a prospective study of vitamin D intake and RA risk among women in the Nurses' Health Study (NHS) found no association (8).
There is mounting evidence for the immunomodulatory effects of vitamin D (1, 4). The hormonally active form of vitamin D mediates immunologic effects by binding to nuclear vitamin D receptors (VDRs), which are present in most immune cell types involved in innate and adaptive immunity. VDRs are expressed constitutively in monocytes, activated macrophages, dendritic cells, natural killer cells, and T and B cells. Activation of VDRs has potent antiproliferative, prodifferentiative, and immunomodulatory functions that are both immune enhancing and immunosuppressive (9).
Early life exposures have been implicated in the etiology of RA and SLE (10–12). Costenbader et al demonstrated that a Northeastern and Mid-Atlantic region of residence in the US at age 15 years was associated with an increased risk of adult-onset RA (from age 30–55 years at inception), suggesting a Northern gradient in RA risk that was confirmed with a detailed geospatial analysis (13, 14). This suggests a potential vitamin D association with RA risk. In a prior study of these cohorts, we did not observe a statistically significant association between vitamin D dietary intake during adult years, from ages 25–55 years, and future risks of RA or SLE (8). However, the relevant time window during the lifespan when vitamin D exposure acts to augment the future risk of RA or SLE is not known.
Adolescence is a time of great physiologic and hormonal change, as well as the start of many health-related behaviors, including smoking and alcohol use, obesity, and physical inactivity. These adolescent health-related behaviors contribute to the risk of noncommunicable diseases in adults, such as hypertension and cancer, independent of adult exposures (15, 16). We hypothesized that vitamin D intake during these years could be relevant many years later in influencing the development of RA and SLE. There is evidence that adolescent diet may affect the risk of adult-onset MS (17, 18). Adolescence is a critical time for immunologic development, and exposures during that time period may have consequences for future disease risk (19).
To date, there have been no studies examining the association between vitamin D intake during adolescence and the risk of RA and SLE in adulthood. We hypothesized that low vitamin D intake from dietary and supplement sources during high school years would increase the risk of developing RA and SLE in adulthood, and studied this association in 2 large, prospective cohort studies of women.
Significance & Innovations
Few studies to date have explored the causal nature of the association between vitamin D status and autoimmune diseases, including rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE), due to limitations by studies' cross-sectional designs.
In a prospective cohort of women, we did not observe an association between reported adolescent dietary vitamin D intake and adult-onset RA or SLE.
Our findings provide insights into when vitamin D intake may be most important and invite future studies into the role of vitamin D intake, a modifiable exposure, at other time periods during the life course.
PATIENTS AND METHODS
The study population included women from 2 longitudinal cohort studies, the NHS and the Nurses' Health Study II (NHSII). The NHS began in 1976 when 121,700 female US registered nurses ages 30–55 years were enrolled (birthdates from 1921–1946). In 1989, the NHSII began with 116,670 female registered nurses ages 25–42 years at enrollment (birthdates from 1947–1964). Therefore, these cohorts represent 2 distinct birth cohorts that began high school in the years 1935–1960 and 1961–1978. Both cohorts were contacted every 2 years by questionnaire to update diet, medications, anthropometrics, and incident physician-diagnosed illnesses. Total response rates to the followup questionnaires were >90% in each cycle. Deaths in the cohorts were usually reported by next of kin and confirmed by the National Death Index.
In 1986, the NHS questionnaire included a 116-item food frequency questionnaire (FFQ) about diet during the past year and a 24-item FFQ that asked the participants to recall intake during high school. This FFQ was completed and returned by 75,458 NHS participants. In 1988, participants reported their alcohol consumption between ages 18 and 22 years. In 1997, NHSII participants were asked if they would be willing to complete a supplemental FFQ about diet during high school (124-item HS-FFQ); 49% agreed and a total of 45,948 NHSII participants completed and returned the HS-FFQ in 1998.
Incident cases of RA and SLE were identified until the end of followup (June 2006 in NHS and June 2007 in NHSII) and confirmed using a 2-stage case validation process previously described elsewhere in detail (11, 20). Participants who self-reported any connective tissue disease (CTD), including RA and SLE, on the biennial followup surveys were asked to complete the previously validated Connective Tissue Disease Screening Questionnaire (21). If positive, medical records were independently reviewed by 2 board-certified rheumatologists trained in chart abstraction to confirm the self-reported diagnosis against standardized American College of Rheumatology (ACR) classification criteria for RA and SLE (22, 23).
Participants who were missing dates of birth or death, those who had confirmed CTD (including RA and SLE) before the start of followup, women who reported CTD during followup but for whom the diagnosis was not confirmed by medical record review, and participants with an implausible caloric intake (<500 or >5,000 kcal) were excluded.
The 1986 NHS HS-FFQ included 24 items regarding dietary recalled intake between ages 13 and 18 years. It has been shown to be highly reproducible when women were queried regarding their high school diet and alcohol intake again in 1994 (r = 0.57, range 0.38–0.74), and was not significantly correlated with current diet as an adult (r = 0.25) (24). For the oldest participants in the cohort (age 65 years in 1986), diet at age 18 years occurred 47 years in the past. For the youngest women (age 40 years in 1986), it was 22 years since high school (24). There was no effect of age on the reproducibility of recall in this cohort.
The 1998 NHSII 124-item HS-FFQ was administered to nurses at a mean age of 43.8 years (range 33.6–53.3 years), yielding a maximum recall time of >25 years (25). Prior reports on reproducibility found that Pearson's correlation for nutrient intake was 0.65, with validity correlations at 0.40 for nutrients, when comparing nurses' recall with the responses of mothers of NHSII participants. Current adult diet was only weakly correlated with recalled adolescent diet. Therefore, the NHSII HS-FFQ is a reasonable record of adolescent diet. A subsequent validation study of the HS-FFQ in young adults who had completed other detailed FFQs during high school showed that the HS-FFQ reasonably captures adolescent diet as recalled by adults (26).
Both the NHS and NHSII HS-FFQs were designed to include foods common during high school (e.g., milkshakes, snack foods). For each food item, a unit or portion size was specified and subjects were asked how often, on average, they had consumed the specified amount of each item. Nine possible response categories were provided, ranging from “never or less than once per month” to “6 or more times per day” (27). Nutrient intakes on the NHS FFQ and NHSII HS-FFQ were computed for each participant by multiplying the frequency of consumption of each unit of food by the nutrient content of the specified portions, and then summing the contributions from all foods (27–29). Nutrient values in foods were obtained from the US Department of Agriculture, food manufacturers, and independent academic sources. Since the composition of some foods has changed over time, in order to provide the best approximation of intake during adolescence, food composition data from the relevant time period (1940s, 1950s, 1960s, and 1970s) were used, whenever available (29, 30). Vitamin D and retinol nutrients were energy adjusted by using the residuals from the regression of nutrient intake on total caloric intake (31, 32).
Data were available on additional adolescent covariates related to vitamin D exposure and disease risk from the NHS and NHSII questionnaires, including teenage smoking history, alcohol use, reproductive factors (e.g., onset of menses), physical activity during adolescence, ethnicity, sunscreen use, and susceptibility to burns during childhood. Body mass index (BMI) at age 18 years was calculated based on reported weight at age 18 years and height at cohort entry. Information on the US state of residence at birth and state of residence at age 15 years was collected from NHS participants in 1992 and from NHSII participants in 1993. The residences were subsequently categorized into 3 regions: North, Middle, and South. This information was used to derive the geographic latitude for the nurse from birth and in childhood and adolescence.
Energy-adjusted intakes of retinol were also considered potential confounders, since there is evidence that vitamin D and retinol both require retinoid X receptor proteins for their actions. A high level of retinol intake could potentially antagonize the actions of vitamin D (33, 34).
In each cohort, we ran separate models for RA and SLE with followup time in person-months extended from time of NHS and NHSII enrollment to the earliest of date of RA or SLE diagnosis, date of death, or end of followup (June 2006 in NHS and June 2007 in NHSII). Total adolescent dietary vitamin D intake exposure was divided into quintiles using the distribution of intake in the cohort. Cox proportional hazards regression models, stratified on age and cohort time, were used to estimate relative risks and 95% confidence intervals for each category of total vitamin D intake, using the lowest quintile as the referent. Multivariable-adjusted models were adjusted further for potential confounding factors, including BMI at age 18 years, age at menarche, adolescent smoking pattern and alcohol intake, adolescent sunscreen use and sun sensitivity, latitude of residence at birth and at age 15 years, physical activity during adolescence, and birth weight. Total energy and energy-adjusted intakes of retinol were also included in the multivariable-adjusted models. Random-effects meta-analytical models were used to obtain a single combined estimate of association across the 2 cohorts for RA and for SLE (35).
The cohort consisted of 73,629 nurses in the NHS followed through June 2006 and 45,544 nurses in the NHSII followed through June 2007, after excluding 1,382 NHS participants for reporting caloric intakes <500 kcal per day and 53 NHS participants for reporting >5,000 kcal consumption per day. No NHSII participants were excluded for this reason. Incident RA was confirmed in 652 NHS and 148 NHSII participants (total n = 800, n = 772 [97%] met at least 4 ACR criteria for RA and n = 28 met at least 3 ACR criteria and reviewers' consensus opinion of RA). Incident SLE was confirmed in 122 NHS and 54 NHSII participants (total n = 176) using a minimum of 3 ACR criteria and reviewers' consensus opinion of SLE (96% met at least 4 ACR criteria for SLE). This was over a mean ± SD followup time of 351 ± 38.6 months for NHS participants and 209 ± 14.0 months for NHSII participants.
Means across the 5 categories of calorie-adjusted vitamin D intake ranged from 135.4 IU to 764.6 IU per day for NHS participants and from 144.0 IU to 643.2 IU for NHSII participants. The majority of women reported vitamin D supplement intakes during high school years that were <400 IU per day (88% of NHS women and 94% of NHSII women), and only a small percentage received the currently recommended 600 IU per day (36) during their adolescence (14% of NHS women and 10% of NHSII women). The mean ± SD age of the women at the time of the FFQ completion was 43 ± 7.2 years for the NHS and 35 ± 4.7 years for the NHSII (Tables 1 and 2). Approximately 10% of women were overweight at age 18 years (BMI >25 kg/m2), and ∼25% of women had menarche at age <12 years. Both the NHS and NHSII participants are predominantly white. Approximately 35% of women resided in Northern regions of the US at birth and at age 15 years, and a minority of all women reported no sun sensitivity during adolescence.
Table 1. Characteristics of participants in the Nurses' Health Study by calorie-adjusted vitamin D quintiles*
|Age, mean ± SD years||43 ± 7.2||43 ± 7.2||43 ± 7.2||43 ± 7.1||43 ± 7.2|
|BMI >25 kg/m2 at age 18 years||13||11||11||10||9|
|Age at menarche <12 years||22||23||23||24||25|
|Northern birth residence||28||34||39||41||39|
|Northern residence at age 15 years||28||34||39||41||39|
|Sun sensitivity (burn) during adolescence||35||36||35||35||34|
|Any sunscreen use during adolescence||82||84||84||85||86|
|Smoking in adolescence (age ≤18 years)||14||14||13||13||12|
|Adolescent alcohol intake||57||58||59||55||56|
|Physically active 1–3 months/year during adolescence||28||29||28||29||29|
|Retinol, mean ± SD IU||1,079 ± 364||1,331 ± 360||1,504 ± 368||1,706 ± 500||5,272 ± 4,102|
Table 2. Characteristics of participants in the Nurses' Health Study II by calorie-adjusted vitamin D quintiles*
|Age, mean ± SD years||35 ± 4.5||35 ± 4.7||35 ± 4.6||35 ± 4.6||35 ± 4.8|
|BMI >25 kg/m2 at age 18 years||12||11||9||8||7|
|Age at menarche <12 years||25||24||25||25||23|
|Northern birth residence||29||32||34||35||36|
|Northern residence at age 15 years||28||31||34||35||35|
|Sun sensitivity (burn) during adolescence||47||47||48||47||47|
|Any sunscreen use during adolescence||32||34||34||35||30|
|Smoking in adolescence (age <19 years)||27||25||23||22||20|
|Adolescent alcohol intake||74||76||76||75||72|
|Physically active 1–3 months/year during adolescence||32||31||30||31||30|
|Retinol, mean ± SD IU||2,014 ± 1,186||2,317 ± 1,106||2,523 ± 1,172||2,919 ± 1,380||5,082 ± 2,691|
Among the NHS participants, ∼15% reported no sunscreen use during adolescence, with the majority of women reporting “little sunscreen use.” In contrast, ∼60% of NHSII women reported no sunscreen use during adolescence. The majority of all women were nonsmokers in high school, with 43% of NHS women and 25% of NHSII women reporting no alcohol intake. Approximately 43% of NHS women and 22% of NHSII women reported no physical activity during adolescence and ∼25% of both NHS and NHSII women reported high socioeconomic status, as defined by their father's occupation.
With the exception of mean retinol level, none of these characteristics appeared to vary across calorie-adjusted quintiles of vitamin D intake in either the NHS or NHSII. As expected, retinol intake increased with increasing vitamin D intake, likely representing multivitamin and food sources.
Age-adjusted models did not show significant associations between adolescent vitamin D intake and risk of adult-onset RA and SLE when comparing the fifth quintile of vitamin D intake to the first quintile (Table 3). In age-adjusted analyses of additional covariates, the only factors associated with an increased risk of adult-onset RA in the NHS were adolescent smoking initiation, alcohol use during adolescence, Northern region of residence at age 15 years, and higher rates of physical activity. Also in the NHS, younger age at menarche was associated with an increased risk of SLE (see Supplementary Table 1, available in the online version of this article at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)2151-4658).
Table 3. Estimated RRs and 95% CIs of the association between adolescent vitamin D intake and incident RA and SLE*
|RA, vitamin D intake (IU/day)|| || || || || || || || || || || |
| Quintile 1||142.2||123||1.00||1.00||147.6||30||1.00||1.00||1.00||1.00|| |
| Quintile 2||228.8||125||0.99 (0.77–1.27)||0.99 (0.76–1.28)||227.4||32||1.10 (0.66–1.81)||1.11 (0.67–1.84)||1.01 (0.81–1.26)||1.01 (0.80–1.27)||0.69|
| Quintile 3||303.3||135||1.04 (0.81–1.33)||1.02 (0.78–1.34)||315.7||31||1.07 (0.64–1.79)||1.07 (0.63–1.80)||1.04 (0.84–1.30)||1.03 (0.81–1.31)||0.88|
| Quintile 4||385.3||141||1.10 (0.87–1.41)||1.02 (0.77–1.36)||404.0||25||0.87 (0.51–1.49)||0.87 (0.50–1.53)||1.06 (0.85–1.32)||0.99 (0.77–1.28)||0.62|
| Quintile 5||692.0||128||1.03 (0.80–1.32)||0.80 (0.57–1.14)||586.7||30||1.02 (0.61–1.70)||1.06 (0.59–1.91)||1.03 (0.82–1.28)||0.86 (0.64–1.17)||0.42|
| P for trend¶|| || ||0.75||0.19|| || ||0.81||0.91||0.35||0.21|| |
|SLE, vitamin D intake (IU/day)|| || || || || || || || || || || |
| Quintile 1||142.2||22||1.00||1.00||147.6||11||1.00||1.00||1.00||1.00|| |
| Quintile 2||228.8||24||1.03 (0.58–1.85)||1.07 (0.59–1.95)||227.4||11||0.98 (0.42–2.28)||0.94 (0.40–2.21)||1.02 (0.63–1.64)||1.02 (0.62–1.67)||0.80|
| Quintile 3||303.3||29||1.24 (0.71–2.17)||1.27 (0.68–2.35)||315.7||8||0.59 (0.23–1.53)||0.55 (0.20–1.47)||0.95 (0.47–1.92)||0.91 (0.41–2.04)||0.16|
| Quintile 4||385.3||22||0.97 (0.54–1.75)||0.92 (0.47–1.82)||404.0||12||1.01 (0.44–2.31)||0.82 (0.34–2.01)||0.98 (0.61–1.59)||0.88 (0.52–1.52)||0.84|
| Quintile 5||692.0||25||1.09 (0.62–1.95)||1.14 (0.52–2.53)||586.7||12||1.08 (0.48–2.46)||0.65 (0.24–1.76)||1.09 (0.68–1.75)||0.92 (0.49–1.71)||0.38|
| P for trend¶|| || ||0.88||0.85|| || ||0.75||0.41||0.74||0.75|| |
When we adjusted for BMI at age 18 years, ancestry, age at menarche, adolescent smoking pattern and alcohol intake, adolescent sunscreen use and sun sensitivity, latitude of residence at birth and at age 15 years, physical activity during adolescence, and birth weight, there appeared to be some suggestion of a protective effect for SLE with increasing quintiles of vitamin D intake among the NHSII cohort; however, it did not reach statistical significance (Table 3). The test for trend across the quintile cutoffs was nonsignificant for the NHS, the NHSII, and the pooled analysis for RA and SLE. Other factors that were significant in the age-adjusted analysis for RA and SLE were no longer significant in the fully-adjusted models.
In this study, we did not observe an association between reported dietary intake of vitamin D during adolescence and risk of RA or SLE in adulthood. In the multivariable models for RA or SLE, we also did not observe any association for sun exposure, sunscreen use, or physical activity during adolescence, which could be proxies for sun exposure. This suggests that adolescence may not be a key time period for vitamin D deficiency leading to adult-onset RA or SLE after age 25 years. Whether vitamin D deficiency might lead to early-onset RA or SLE in adolescence or early adult life cannot be addressed by our study.
With regard to timing of exposure, past large, prospective cohort studies have identified exposure during the perinatal period as potentially related to risk of developing RA and SLE in adulthood (10, 37). However, due to hormonal changes and continued immune development, there is good reason to suspect that adolescence is a potentially important time to examine exposures in the etiology of SLE (19).
There is epidemiologic evidence that vitamin D intake is inversely associated with a risk of other autoimmune diseases, including MS and type 1 diabetes mellitus (38, 39). However, it is strongest for MS, where a clear latitude effect with the highest risk at Northern latitudes and vitamin D deficiency in adult life are associated with a higher risk of MS, as well as some studies suggesting MS risk is related to vitamin D status at different ages, possibly starting in utero and extending through early childhood and adolescence (2, 38). Living in Northern latitudes is also associated with an increased risk of MS, and this is thought to be mediated by reduced vitamin D from decreased solar exposure (40).
The studies of vitamin D intake during adulthood and risk of incident adult-onset RA come to contradictory conclusions (7, 8). One study found a strong protective effect of baseline high vitamin D intake in diminishing the risk for RA through 11 years of followup (7), whereas another revealed no association between repeated measures of intake and risk of RA (8).
Many cross-sectional and case–control studies have reported lower levels of 25-hydroxyvitamin D (25[OH]D) in patients with existing SLE compared to controls, and lower mean levels in African American patients when compared with white patients (41, 42). Renal disease also may lower levels of 1,25-dihydroxyvitamin D, and symptoms of fatigue and photosensitivity may contribute to vitamin D insufficiency (43, 44). Whether vitamin D levels are low prior to SLE onset or as a result of disease has not been established. However, a prospective study of incident adult RA and SLE in the same cohort found no association between adult dietary vitamin D intake prior to onset of RA or SLE (8). Our past null findings may have been in part due to the fact that dietary sources account for only a small proportion of circulating vitamin D, whereas other sources that account for a larger proportion, such as ultraviolet light exposure, were not considered in that study.
Some limitations of our study include the use of the FFQ for calculated vitamin D intake as a proxy for vitamin D status. Although this is a fair proxy for relative ranking of individuals regarding vitamin D intake, it is not possible to extrapolate directly to circulating 25(OH)D levels. Results of the high school vitamin D intake questionnaire in these cohorts have been highly reproducible, but given the time windows concerned, it is not possible to compare to the relevant circulating vitamin D levels (24). Differential misclassification of the reported vitamin D intake is highly unlikely between the cases and controls, since the HS-FFQs were administered years prior to RA or SLE diagnosis. Small numbers of incident cases also reduced our power to detect an association if one existed. We examined incident cases only and we did not include RA and SLE cases diagnosed before 1976 for the NHS and before 1989 for the NHSII, thereby omitting cases diagnosed in young adulthood; therefore, our study only addresses the association for RA onset after age 25 years. Although vitamin D deficiency is more prevalent in patients with darker skin pigmentation due to the decreased cutaneous conversion of vitamin D to its active form via ultraviolet B exposure, the NHS cohorts are predominantly white, limiting the ability to study the vitamin D associations within African Americans.
Our study had a number of strengths. It was a large, prospective cohort study with longitudinal data spanning up to almost 30 years. It is also the first prospective study to date on adolescent dietary vitamin D intake in relation to adult-onset RA and SLE. In our study, we were able to adjust for factors from adolescence that could have been associated with circulating vitamin D in our multivariable models, including sunscreen use, sun sensitivity, and latitude of residence.
In 2011, the Institute of Medicine increased the recommended dietary allowance from 400 IU to 600 IU of vitamin D per day in those ages 1–70 years and to 800 IU among those ages >70 years (36). Although we observed that the majority of nurses reported insufficient vitamin D intakes in their high school years, we did not find associations between reported dietary intake of vitamin D during adolescence and risk of RA or SLE in adulthood. In contrast to prior studies of adult RA and SLE, in the current study, we adjusted for epidemiologic factors that influence circulating vitamin D levels, such as indices of sun exposure. Our results rule out a strong association between high school dietary intake of vitamin D containing foods and supplements and the future development of adult-onset RA and SLE. However, inaccurate reporting of adolescent diet would reduce our power to detect an association. Future studies are required to examine the relationship between circulating vitamin D levels throughout the life course and the risk of RA and SLE.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Hiraki had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Hiraki, Costenbader, Karlson.
Acquisition of data. Hiraki, Costenbader, Karlson.
Analysis and interpretation of data. Hiraki, Munger, Costenbader, Karlson.
The authors would like to thank Channing Laboratory, Department of Medicine, Brigham and Women's Hospital, and Harvard Medical School.