Vitamin D

Chondrocytes and synovium express vitamin D receptors, vitamin D regulates subchondral bone quality and muscle function, and deficiency is common in knee-OA populations. Plausible, generic — and unusually well tested. This is one of the few supplement topics with a nearly complete randomized record, and that record is mostly null. Any copy implying vitamin D "protects cartilage" conflicts with three null structural RCTs.

The null trials

McAlindon 2013 (JAMA, n=146, 2 years): 2,000 IU/day escalated to keep serum levels high. Knee pain fell similarly in both arms, and MRI cartilage-volume loss was identical (P=0.96).

Jin 2016, the Tasmanian VIDEO trial (JAMA, n=413, deficient at baseline, 50,000 IU/month, 2 years): null on both primary endpoints — tibial cartilage-volume loss and WOMAC pain — despite correcting the deficiency.

Arden 2016, the UK VIDEO study (n=474, 800 IU/day, 3 years): no difference in joint-space narrowing (P=0.49). The 800 IU dose was probably too low — a design flaw, but a null is a null.

MacFarlane 2020, the VITAL knee-pain cohort (n=1,398, 2,000 IU/day, ~5.3 years): WOMAC knee pain did not differ from placebo at any timepoint.

Early cohort data pointed the same way: Felson 2007 found no association between baseline vitamin D and radiographic worsening in two cohorts. Publication bias, for once, runs in the honest direction — the nulls are the famous papers.

What survives

Meta-analyses squeeze out a small symptomatic signal. Gao 2017 (4 RCTs, n=1,136) found modest WOMAC pain and function improvements — better above 2,000 IU/day — but no effect on cartilage volume. Wang 2023 (8 RCTs, n=3,077) pooled symptomatic benefits and reduced effusion-synovitis, again favouring >2,000 IU/day. Both lean on secondary endpoints of individually null trials, and the pooled pain effects sit below accepted clinically-important thresholds.

The more interesting thread is observational: among Tasmanian VIDEO completers, people who stayed vitamin-D sufficient over 24 months lost less tibial cartilage (~2.1%), accumulated less effusion-synovitis, and kept better physical function than the consistently insufficient. That analysis is post-hoc and non-randomized — healthier-adherer bias applies — but it is the best human data for a sufficiency-maintenance framing.

Dose & practical notes

The defensible position: test, and correct deficiency to above 50 nmol/L (20 ng/mL) — cheap, safe, and supported by the sufficiency data — while expecting no cartilage regrowth and no slower structural loss. If any symptomatic signal exists, the meta-analytic subgroups place it at daily doses above 2,000 IU rather than monthly boluses, which have underperformed daily dosing across the vitamin D literature generally.

Safety

Wide margin at 800–2,000 IU/day. Hypercalcemia risk appears with sustained very high intakes (chronically above 4,000 IU/day unmonitored, or bolus mega-dosing) and in conditions of calcium dysregulation such as sarcoidosis and hyperparathyroidism.

What I take from it

Vitamin D is the control case for this whole section: a plausible mechanism, enthusiastic early observational data, and then real trials that said no. Fix a deficiency because deficiency is worth fixing — not because your cartilage will notice.