Vitamin D {>10,000 IU daily) is safe to use for Multiple Sclerosis; some RCTs found benefits (need much more)

Effects of high-dose vitamin D supplementation in patients with multiple sclerosis: A systematic review of efficacy and safety - June 2026

Mult Scler Relat Disord 2026 Jun 13:112 10.1016/j.msard.2026.107324

PDF

Summary by Claude AI - June 2026

A 2026 systematic review (PRISMA, PROSPERO-registered, RoB-2, dual reviewers, narrative synthesis only) of "high-dose" vitamin D (≥10,000 IU/day) in adult MS. Eleven RCTs, n ranging ~32–232, follow-up 12–78 weeks, almost all on a background of IFN-β. Three headline findings, correctly tiered by their own confidence:

  • Most robust: 25(OH)D rose reliably and severalfold — confirms exposure/adherence, nothing more.
  • Surprisingly clean safety: only rare, transient, asymptomatic hypercalcemia/hypercalciuria; no renal signal. Their strongest contribution.
  • Weak/inconsistent efficacy: no significant effect on relapse rate or EDSS progression; MRI mixed (only Hupperts/SOLAR showed a CUA reduction, ~32%); immune modulation biologically plausible (Th17/IL-17↓, anti-EBNA-1↓) but not translating to clinical endpoints.

Conclusion is appropriately hedged: safe under monitoring, promising adjunct, needs larger/longer trials.

Critique

1. The "11 RCTs" headline is inflated roughly 2×. This is the central problem. Several entries are the same trial reported for different outcomes or at different stages:

  • SOLAR cluster: Hupperts 2019 is the parent SOLAR trial; Rolf 2017/18 (anti-EBV), Muri/Muris 2016 (SOLARIUM immune substudy), and Smolders 2019/20 (NfL) are all substudies of it. That's one trial counted four times — and counted four times in the risk-of-bias plot as if independent.
  • Johns Hopkins cluster: Bhargava 2015 and Sotirchos 2015/16 are the same Calabresi/Mowry trial (10,400 vs 800 IU, 6 mo). Worse, Bhargava 2015 is a Neurology conference abstract (S38.001) — which their own eligibility criteria explicitly exclude.
  • Isfahan cluster (inferred): Ashtari 2015, Ashtari 2016, and Toghianifar 2015 are all n=94, same author group, same IFN-β comparator, ~3 months. Almost certainly one cohort reported for IL-10, QoL, and IL-17/EDSS respectively — not three trials.

Collapse these and you have roughly five independent trials (SOLAR, Hopkins, Isfahan, Burton 2010, Dörr/EVIDIMS), not eleven. The review never flags this non-independence, so the apparent evidence base is overstated and the RoB "traffic light" double-counts shared randomization quality.

2. No meta-analysis despite poolable outcomes. They cite heterogeneity, but 25(OH)D, ARR, and EDSS were reported across multiple arms with comparators. At minimum a pooled ARR/EDSS estimate (even with wide CIs) would be more informative than vote-counting. Narrative synthesis here hides effect sizes.

3. A reported "positive" is likely a baseline-imbalance artifact. They cite Toghianifar 2015 as showing significantly lower EDSS in the vitamin D arm as a clinical signal. But the sibling paper (Ashtari 2015) states baseline EDSS already differed favoring the vitamin D group. A cross-sectional end-EDSS difference in a confounded, possibly non-independent cohort is not efficacy. This should have been caught.

4. Achieved 25(OH)D and duration — the real reason for null efficacy — is underweighted. Several "high-dose" arms achieved modest levels (Dörr reached only ~62 ng/mL; Hupperts only fourfold from a deficient baseline) over ≤48 weeks. Null clinical results under short duration and unimpressive achieved levels don't refute the hypothesis; they fail to test it. The review says this in passing but doesn't foreground it — the dose-on-the-label ≠ level-achieved distinction is the key interpretive lever and gets one sentence.

5. IFN-β confounding is acknowledged but fatal to the efficacy question. Nearly all participants were on IFN-β, which suppresses baseline inflammatory activity (floor effect on MRI/relapse). The review correctly notes this but still treats the null clinical findings as somewhat informative. They're largely uninformative for the adjunctive-on-top-of-low-activity question.

6. Dose threshold applied loosely + citation sloppiness. Dörr is 20,400 IU every other day (~10,200/day averaged); Hupperts' first 4 weeks (6670 IU/day) sits below threshold. Defensible but inconsistent. Citation errors throughout: Dörr "2020" in text vs "2022" in Fig 2; Rolf "2017" vs "2018" in refs; Sintzel "2017" vs "2018"; Bhargava (n=32) and Sotirchos (n=40) listed as separate sample sizes for what is one enrolled cohort. Cassard/VIDAMS is cited in the discussion as supporting evidence though correctly excluded (5000 IU < threshold) — fine, but it underscores that the threshold cuts out the largest dedicated high-dose RRMS RCT, leaving the review's evidence base small.

Strengths worth keeping: the dose-stratified question (separating ≥10,000 IU from moderate-dose reviews) is genuinely useful and underdone; methods are transparent; and the safety synthesis is the most defensible part — reassuring on short-to-medium-term tolerability, with the honest caveat that these trials are too short/small to detect rare harms.

Bottom line for citing it: usable as a safety reference and as a tidy roster of the high-dose MS RCTs, but the "11 RCTs" framing and the absence of a pooled estimate mean it overstates breadth and understates why efficacy looks null (modest achieved levels, short duration, IFN-β floor). The honest one-line takeaway it almost reaches but doesn't quite state: high-dose D3 in MS hasn't failed — it's been tested at insufficient achieved 25(OH)D for insufficient time on top of an active DMT.

This is a sensitive clinical topic; the above is critique of the evidence, not treatment guidance.

Related in VitaminDWiki

High Dose Vitamin D

Coimbra Protocol for MS (~100,000 IU daily)