Low vitamin D at birth is not associated with autoimmune problems 2 decades later (no surprise)
Neonatal vitamin D levels and autoimmune disorders: a Danish population-based cohort study
European Journal of Epidemiology June 2025 https://doi.org/10.1007/s10654-026-01436-9
Henriette Thisted Horsdal, Berit Heitmann, Sanne Grundvad Boelt, Esben Agerbo, Thomas Werge, Nis Borbye-Lorenzen, Carsten Bøcker Pedersen, John J. McGrath & Xiaoqin Liu
To examine the associations of neonatal 25-hydroxyvitamin D (25(OH)D) and vitamin D-binding protein (DBP), and their corresponding genetic predictors, with the risk of nine autoimmune disorders. We conducted a population-based cohort study of a random sample of individuals born in Denmark between 1981 and 2005, sourced from the iPSYCH2012 study. We measured 25(OH)D and DBP concentrations in neonatal dried blood spots. We identified individuals diagnosed with selected autoimmune disorders (multiple sclerosis, rheumatoid arthritis, psoriatic arthritis, type 1 diabetes mellitus, autoimmune thyroiditis, Graves’ disease, celiac disease, Crohn’s disease, and ulcerative colitis). Cox regression was employed to estimate hazard ratios (HRs) for any autoimmune disorder and for nine specific diagnoses, in relation to 25(OH)D, DBP, and their polygenic scores (PGSs). Among 20,404 eligible individuals, 757 (3.7%) developed an autoimmune disorder. Neither neonatal 25(OH)D nor DBP was associated with the risk of any autoimmune disorder, with HRs of 1.07 (95% CI, 0.99–1.15) and 0.99 (95% CI, 0.92–1.07) per standard deviation (SD) increase, respectively. Similarly, PGSs for 25(OH)D and DBP did not show an association with any autoimmune disorder, with HRs of 1.04 (95% CI, 0.97–1.12) and 0.98 (95% CI, 0.91–1.05) per SD increase, respectively. Findings were similar when exposures were analyzed in tertiles and across nine individual autoimmune disorders. Neonatal vitamin D is not associated with the risk of autoimmune disorders. Similarly, genetic predictors of vitamin D status, as reflected in PGSs for 25(OH)D and DBP, are not associated with the risk of autoimmune disorders.
A study found that only 8% of the future vitamin D variances were associated with levels 2 decades before
There's essentially one study that directly measures individual-level predictability out to two decades, and its answer is that a single baseline reading is a poor predictor of where someone lands 20 years later.
The definitive one is Kubiak, Kamycheva & Jorde (2021, Eur J Clin Nutr), using three waves of the Tromsø study — 1994/95, 2007/08, and 2015/16 — with 1702 subjects who had valid 25(OH)D measurements in all three surveys, using z-scores to adjust for seasonal variation. The tracking decayed exactly as you'd expect a noisy biomarker to:
- The correlation between 25(OH)D z-scores from Tromsø 4 and Tromsø 6 (~13 years) was 0.40, declining to 0.29 for the correlation between Tromsø 4 and Tromsø 7 (~21 years).
An r of 0.29 means baseline explains roughly 8% of the variance in an individual's level 21 years on. The category-shuffling makes it concrete:
*26% of subjects in the lowest baseline sextile had moved into the three highest sextiles 21 years later, and
*35% of those in the highest baseline sextile had dropped into the lowest three sextiles. Their own conclusion is the gap-flag you'd write: tracking declines over time, and use of a single 25(OH)D measurement as an indicator of vitamin D status is questionable in long-lasting observational studies.
The shorter-horizon studies show the decay curve leading into that:
- Same group's earlier 14-year Tromsø analysis (Jorde 2010): correlation between 1994 and 2008 measurements ranged from 0.42 to 0.52 depending on seasonal adjustment, versus 0.80 between baseline and 12-month values in the intervention arm.
- A longitudinal-variability analysis found the correlation decreasing with time between tests, from 0.83 for tests in the same year to 0.55 after 3 years.
- OsteoPerio, 5 years apart in postmenopausal women: Spearman r of 0.61 overall (0.68 for same-season draws), with an intraindividual CV of 24.6% and a modest ICC of 0.59.
- Adolescence (Poopedi 2015), spanning ages 11–20: no significant correlation between 25(OH)D in the earlier versus later years of adolescence, though shorter 2-year intervals correlated well (r ≥ 0.65).
What none of this shows — worth being explicit about, given how the question is usually mis-stated:
It does not say vitamin D status is random. Short-interval reproducibility is decent (r ≈ 0.6–0.8 at 1–5 years), which is why single measurements remain defensible for short-follow-up epidemiology. The 20-year decay isn't measurement noise — assay stability in frozen samples is fine over decades. It's real biography: significant predictors of change in 25(OH)D include baseline level, change in BMI, and vitamin D supplementation at follow-up. People start supplementing, gain or lose weight, move, age. In Tromsø the whole cohort's mean drifted upward (57 → 59 → 72 nmol/L) as supplementation spread — a secular trend layered on top of individual drift.
And it does not mean nobody has modeled future/average status — but that's a different claim. The Nurses' Health/HPFS prediction models regress 25(OH)D on determinants (age, UV-B, intake, BMI, season) and land at model r² of 0.25 to 0.33 — predicting a contemporaneous level from covariates, not forecasting an individual 20 years out.
So the honest one-liner for a wiki page: the only direct evidence (a single 21-year cohort) puts the baseline-to-future correlation at ~0.29, and essentially no other study has even attempted the 20-year individual horizon. That's a genuine "profitable ignorance" gap — the tracking question is cheap to ask of any biobank with repeat draws, and almost nobody has.