Falls · VITD-FALL-02 · v2
Several randomised trials in older people found more falls on higher vitamin D doses: a Swiss trial saw falls rise from about 48% to 67% on larger monthly doses [1], a long-term-care trial saw falls more than double [5], a dose-ranging trial found the most falls at the top doses [6], and a fall-focused trial flagged more serious falls and fall-related hospitalisations at 1000 IU per day or more [12]. But the two biggest high-dose trials, together enrolling over 26,000 people, found no overall change in falls [4][17], and a 25,000-person moderate-dose trial also found nothing [10]. One trial even found fewer falls on a modest 1000 IU daily dose [15]. The harm signal is real and shows up repeatedly, yet it tends to appear in smaller dose-finding or subgroup analyses while the largest pragmatic trials do not confirm a blanket increase, which is why the panel did not treat the claim as settled either way.
Multiple independent RCTs report increased falls at higher doses: Bischoff-Ferrari [1] found a higher fall incidence in the 60,000 IU and calcifediol arms versus 24,000 IU (P=.048), Ginde [5] reported an incidence rate ratio of 2.33 (95% CI 1.49-3.63) in long-term care, Smith [6] found a U-shaped curve with falls increased on high versus medium doses (OR 5.6, 95% CI 2.1-14.8), and STURDY [12] found more serious falls (HR 1.87) and fall-related hospitalisations (HR 2.48) at doses of 1000 IU per day or higher. Against this, the two largest high-dose monthly trials are null overall: ViDA [4] HR 0.99 (0.92-1.07) and D-Health [17] OR 1.02 (0.95-1.10), and the 25,871-participant VITAL trial at 2000 IU per day is null (OR 0.97, 0.90-1.05) [10]. D-Health [17] did show increased risk in the BMI<25 subgroup (OR 1.25, 1.09-1.43). Counter-direction evidence exists at modest doses, where Cangussu [15] found placebo fallers 1.95x more likely to fall than a 1000 IU per day group. The contested point: The division is between the reproducible harm signal in dose-finding, subgroup and secondary analyses ([1], [5], [6], [12], [17] BMI<25) and the null overall results of the two largest high-dose trials ([4], [17]) and the large moderate-dose trial [10]. It also turns on whether 'high-dose' names a single exposure at all, since the positive and null trials span daily 4000-4800 IU [6], monthly 60,000-100,000 IU [1][4][17] and 2000 IU per day [10].
All 20 papers behind this claim with their DOIs and what each does to it, the panel’s arbitration, the evidence gap map and the open research questions. Free, and sent to your inbox.
We’ll also send you other claims from Vitamin D, one at a time, and tell you when a release or a pivotal trial changes a grade. Unsubscribe by replying.
One reviewer's only job is to attack the finding. The challenges below were weighed before the grade was set.
The two largest high-dose RCTs are null for overall falls — ViDA [4] HR 0.99 (0.92-1.07) and D-Health [17] OR 1.02 (0.95-1.10) — directly contradicting a blanket 'increases fall risk' statement.
'High-dose' conflates heterogeneous regimens — daily 4000-4800 IU [6], monthly 60,000-100,000 IU [1][4][17], and 2000 IU/d [10] — whose fall effects differ, so the claim's wording is imprecise and the positive and null trials are not testing the same exposure.
Where it remains uncertain. The two largest high-dose trials are null for overall falls ([4] HR 0.99; [17] OR 1.02), and moderate-dose VITAL is also null [10], so the harm concentrates in smaller or subgroup analyses rather than the big pragmatic trials.
Aleatory
The strongest harm estimate comes from a 107-participant single-site trial [5], and subgroup splits in [12] and [17] carry wide intervals, so chance variation plausibly inflates individual effect sizes.
Epistemic
Falls appear to follow a non-monotone, possibly U-shaped relationship with dose or 25(OH)D level [6], but no trial in the corpus maps the threshold at which risk turns from benefit to harm.
Inferential
Key harm findings are prespecified-tertiary, post-hoc or subgroup analyses ([4], [12], [17]), and [19] warns such subgroup findings may be type 1 error, weakening the inference from signal to causal harm.
The load-bearing papers from a corpus of 20, each classified by what it does to the claim. Every DOI is taken from the corpus the claim was graded on.
Supports: higher fall incidence at higher monthly doses, though comparator is active 24,000 IU not placebo.
“over the 12-month follow-up, the incidence of falls differed significantly among the treatment groups, with higher incidences in the 60,000 IU group (66.9%; 95% CI, 54.4% to 77.5%) and the 24,000 IU plus calcifediol group (66.1%; 95% CI, 53.5%-76.8%) group compared with the 24,000 IU group (47.9%; 95% CI, 35.8%-60.3%) (P = .048)”
Panel auditInternal validity: medium · scope of inference: Community-dwelling adults 70+ with a prior fall, mostly vitamin D deficient, over 12 months.Key limitations: small sample · active low-dose comparator not placebo · falls a secondary endpoint · single centre
Counter: largest high-dose monthly trial shows no fall increase.
“The HR for falls-adjusted for age, sex, ethnic origin, history of recent fall, physical activity, and baseline 25(OH)D-was 0·99 (95% CI 0·92-1·07; p=0·82) for vitamin D compared with placebo”
Panel auditInternal validity: high · scope of inference: Healthy, ambulatory adults aged 50-84 (mean 66), not selected for fall risk or deficiency.Key limitations: falls post-hoc · healthy ambulatory not high-fall-risk · includes under-65s · bolus monthly dosing
Supports: >2x fall rate on 100,000 IU/mo in older LTC residents.
“Falls were more common in the high-dose group (1.47 per person-year vs 0.63 in standard-dose group; IRR = 2.33, 95% CI = 1.49-3.63, P < .001)”
Panel auditInternal validity: medium · scope of inference: Long-term care residents aged 60+ in Colorado facilities over 12 months.Key limitations: small sample · falls secondary · standard-dose not placebo comparator · institutionalised population
No high-dose arm exists, so the high-dose contrast the claim asks about is not reported.
Panel auditInternal validity: high · scope of inference: Home-dwelling women 70-80 with a prior fall, at standard 800 IU/d dosing.Key limitations: dose is standard not high · women only · vitamin D not placebo-equivalent to high-dose question
4 of 20 papers shown. The full review lists all of them, with the panel’s arbitration →
Claims are graded at each quarterly release, and re-graded between releases only when a pivotal trial lands. Every change is a new version of the claim.
A credible, dose-dependent harm signal is present in several RCTs but is contradicted by the largest high-dose trials and concentrated in subgroup or secondary analyses, warranting an Uncertain verdict.
Release 2026-q4 re-graded every claim from its corpus under protocol v0.19. An earlier release's grade is shown for the record; it came from an earlier protocol and is not a movement.
Assumptions that carry weight but are not themselves established.
The panel records the gap that would most move the claim, and the study that would close it.
Placebo-controlled daily high-dose falls trial. The clearest harm estimates come from active-comparator or secondary analyses ([1], [12]); no corpus trial tests a daily high dose against placebo with falls as the primary, adequately powered endpoint in general older adults. The panel rates this the highest-importance open gap for the claim, and names the study design that would close it: A large double-blind RCT randomising older adults to a defined daily high dose versus placebo with falls prespecified as the primary outcome.
Open question: Does a daily high dose of vitamin D increase falls versus placebo when falls are the primary, powered outcome in general older adults? Why it is still open: The null large trials used monthly dosing [4][17] or moderate daily dosing [10], and the positive daily-dose finding [6] is a secondary outcome in selected women; no placebo-controlled daily high-dose falls-primary trial exists in the corpus.
Priors is a sceptical first reader, not peer review, and not medical advice.
Release 2026-q4 · graded 26 Oct 2026 · protocol v0.19