Total intake of different minerals and the risk of multiple sclerosis

Study design

The NHS and NHSII are prospective cohort studies that started in 1976 (NHS, 121,700 participants) and 1989 (NHSII, 116,671 participants), including female registered nurses who lived in 11 (NHS) or 14 (NHSII) US states, and were 30–55 (NHS) and 25–42 (NHSII) years old at inclusion. They are followed over time with biennial self-administered questionnaires collecting information on a variety of lifestyle exposures and health-related outcomes, with a high response rate (about 90%) in both cohorts in each questionnaire cycle.

Standard protocol approvals, registrations, and patient consents

The institutional review board of Brigham and Women's Hospital approved these cohort studies. The return of a completed questionnaire implied informed consent by the participants.

Assessment of total intake of different minerals

Diet was assessed every 4 years in 80,920 women in NHS (1984–2002) and 94,511 women in NHSII (1991–2007), using a validated semiquantitative food frequency questionnaire (FFQ), described in detail elsewhere.^16,17 The FFQs collected information on average portion sizes (e.g., 1 slice of bread, 2 slices of bacon, 1 8-oz glass of milk) and frequency of usual consumption (“never or less than once per month,” “1–3 per month,” “1 per week,” “2–4 per week,” “5–6 per week,” “1 per day,” “2–3 per day,” “4–5 per day,” or “6+ per day”) over the previous year of about 130 different food items covering dairy products, fruits, vegetables, breads/cereals/starches, eggs/meats, beverages, and sweets/baked goods/miscellaneous. In addition, participants were asked about their supplement use. Follow-up questions inquired about brands and frequency (multivitamins) and daily supplementation doses (e.g., zinc: <25, 25–74, 75–100, >100 mg).

Minerals are essential nutrients, some of which are needed in small amounts (trace elements). The FFQ covers the major contributing food and supplemental sources.¹⁸ In this study, the main potassium sources (adequate intake [AI]¹⁹: 4,700 mg/d) were coffee (contributing 4.0%–8.8% of total intake depending on the year and the cohort), milk (3.6%–9.3%), and potatoes (4.2%–7.3%); the main magnesium sources (Recommended Daily Allowance [RDA]¹⁹: 320 mg/d) were multivitamins/magnesium supplements (3.8%–13.6%), coffee (4.0%–10.0%), milk (4.5%–10.1%), and cold cereals (3.5%–5.6%); the main calcium sources (RDA: 1,000 mg/d) were dairy products (17.3%–80.5%) and multivitamins/calcium supplements (9.6%–36.5%); phosphorus (RDA: 700 mg/d) came primarily from dairy products (16.7%–22.5%) and meat (3.3%–7.7%). Top iron sources (RDA: 18 mg/d in women) were multivitamins/iron supplements (22.3%–42.0%), cold cereals (10.2%–17.8%), and beef (2.5%–7.0%); top zinc sources (RDA: 8 mg/d) were multivitamins (10.5%–36.9%), zinc supplements (6.2%–16.2%), beef (3.8%–14.9%), and cold cereals (4.9%–6.9%); top manganese sources (AI: 1.8 mg/d) were multivitamins (9.6%–34.9%), cold cereals (7.5%–12.3%), tea (4.1%–9.8%), and oats (3.1%–7.3%); copper (RDA: 0.9 mg/d) came primarily from multivitamins (13.5%–39.6%), liver (2.4%–8.1%), and potatoes (3.1%–7.2%). The FFQ neither captures copper in drinking water derived from water pipes nor selenium as the amount in plants depends primarily on soil enrichment and varies thus by location.²⁰

The mineral content in different food items and supplements was determined using the Harvard²¹ FFQ Nutrient Database, which is regularly updated and expanded for average nutrient content. Intakes of each mineral were estimated by multiplying the content of each source by the woman's reported frequency of intake, with frequency weighted such that “1 per day” equaled 1. Total intake of a specific mineral was calculated by summing intake across the different sources. Subsequently, we adjusted the mineral intake for energy intake using the residual method to isolate each mineral's contribution to risk independent of total caloric intake or food quantity consumed.²²

The measurement of these minerals by FFQ has been validated in the context of the Women's Lifestyle Validation Study (WLVS), conducted among participants in the NHS and NHSII with data collection over an approximately 1-year period in 2010–2012 to assess the measurement error that can occur when dietary exposures are self-reported.²³ Using a slightly expanded version of the FFQ used in the original nurses cohorts with in total 152 food items to include newer foods, the energy-adjusted de-attenuated (random error–corrected) Spearman correlations between the assessment by FFQ with that from 7-day dietary records including foods and supplements were overall strong (potassium: r = 0.65, magnesium: r = 0.75, calcium: r = 0.74, phosphorus: r = 0.67, iron: r = 0.58, zinc: r = 0.68, copper: r = 0.63), as were the correlations with that from 24-hour dietary recalls (potassium: r = 0.64, magnesium: r = 0.73, calcium: r = 0.68, phosphorus: r = 0.68, iron: r = 0.55, zinc: r = 0.53, copper: r = 0.52). Manganese was not assessed in the WLVS, but within a previous study using NHSII data collected by FFQs, manganese intake highly correlated with that of magnesium (r = 0.65) and copper (r = 0.62),²⁴ arguing that FFQs are likely to well capture manganese.

Case ascertainment

Women reported their diagnosis of MS on the biennial questionnaires. We subsequently asked them for permission to contact their treating neurologist and review their medical record. Since 2003, our study neurologist (T.C.) has been reviewing all medical records for confirmation. In this study, we defined the outcome as neurologist-confirmed probable or definite MS diagnosis according to date of diagnosis. Cases of possible MS were not included. There were 479 new cases of probable or definite MS diagnosed during follow-up (130 in NHS, 349 in NHSII) among women who also completed the FFQs. In secondary analyses, we defined the outcome according to date of first symptom. There were overall 289 new cases (59 in NHS, 230 in NHSII) diagnosed with probable or definite MS during follow-up with information on date of onset and diet.

Covariates

We categorized ancestry, reported in 1992 (NHS) and 1989 (NHSII), into Southern European, Scandinavian, Other Caucasian, or Other in this study. Women reported their state of residence at age 15 in 1992 (NHS) and 1993 (NHSII). The states were categorized into northern, middle, and southern tier, as previously described.²⁵ Smoking status was updated biennially including information on number of cigarettes smoked per day, categorized subsequently into pack-years: never, <10, 10–24, and ≥25. Women reported their height at baseline and their weight at age 18 in 1980 (NHS) and 1989 (NHSII). We calculated their body mass index (BMI) in kg/m² using these measures, and categorized them as underweight, normal, overweight, or obese (<18.5, 18.5 to <25.0, 25 to <30, and ≥30 kg/m²) according to the cutoffs suggested by the WHO. Supplemental vitamin D intake was categorized into none, <400, and ≥400 international units per day. Total energy intake was calculated from the FFQ.

Statistical analyses

We estimated the association between total intakes of different minerals and MS risk during follow-up using Cox proportional hazards regression. To optimize power, we used the date of diagnosis to define the outcome in the main analyses, as fewer cases had a certain date of onset. Study participants contributed, therefore, time at risk (person-years) from date of return of the baseline diet questionnaire (NHS: 1984, NHSII: 1991) to date of MS diagnosis, date of death, or loss to or end of follow-up (NHS: June 2004, NHSII: June 2009), whichever occurred first. We used 1986 as baseline for copper as the assessment in 1984 was incomplete in the NHS. We included the energy-adjusted minerals as exposure in all the analyses and replaced missing exposures with last available values if the participant was not lost to follow-up. We assessed age- and total energy–adjusted as well as multivariable-adjusted hazard ratios (HR) and 95% confidence interval (CI) at a significance level of 0.05. Women with daily caloric intake of below 500 or above 3,500 kcal and those who returned only the baseline questionnaire were excluded from the analyses.

We assessed the total intake at baseline continuously by incremental steps of 100 mg/d for potassium, magnesium, calcium, and phosphorus, and 1 mg/d for iron, zinc, manganese, and copper. For the categorical analyses, we ranked participants into quintiles according to their (1) baseline and (2) cumulative average intake. We also examined whether more extreme exposures were associated with MS risk by ranking participants in deciles according to their baseline total mineral intake. In all categorical analyses, we also assessed whether there was a statistically significant linear trend across the medians within each quantile. The multivariable models were adjusted for ancestry, latitude of residence at age 15, BMI at age 18, pack-years of smoking, supplemental vitamin D, and total energy intake. The latter 3 were updated biennially and introduced into the models as time-varying covariates. We conducted analyses separately for each cohort and then pooled the results using fixed-effects models with the inverse of the variance of the risk estimates as the weight, and a Q statistic to assess heterogeneity.²⁶ Mantel extension tests were performed to examine the trend across the pooled categories.

In secondary analyses, we used date of onset instead of diagnosis to define the outcome and repeated the analyses of continuous and categorical exposures in quintiles (baseline and cumulative intake). Differences in the findings could indicate reverse causation when using diagnosis date, i.e., that the disease or related behavioral changes affected mineral intake after disease onset. As we found an inverse association between zinc intake and MS risk in these secondary analyses, we explored whether changes in zinc intake from diet or supplements after clinical disease onset could explain the discrepancy between results based on date of diagnosis and those based on date of first symptom. To examine this hypothesis, we predated the outcome by 6 years from diagnosis date among MS cases to approximate first symptom. We additionally compared the results for total zinc intake to intake from food only. Further, we assessed whether pre-onset intake, derived from information on the last FFQ prior to first symptom, was important, considering the suggested antimicrobial effects of zinc^13,27 and the possible role of viral infections in triggering MS symptoms.²⁸ For this we used total and supplemental intake (including multivitamins) in quintiles and zinc-only supplement use (yes–no). We adjusted these pre-onset analyses also for α-linolenic acid (ALA) intake (in quintiles) based on a previously reported inverse association between ALA and MS in these cohorts.²⁹ For purposes of comparison, last FFQ analyses were also conducted for the other minerals.

Data availability

The datasets analyzed in the current study are not publicly available because of restricted access, but further information about the datasets is available from the corresponding author on reasonable request.