Vitamin D dosing by blood level, not by bottle

How far a dose moves your blood level depends mostly on where you start and what you weigh. Across 94 supplementation studies, dose per kilogram of body weight explained 34.5% of the variation in circulating 25-hydroxyvitamin D, and the relationship was logarithmic: each added unit did less than the one before.¹

What a dose does to a blood level

Vitamin D status is read from serum 25-hydroxyvitamin D, and the analyses below report it in nanomoles per liter. A lab printout in nanograms per milliliter runs on a smaller scale, so the figures here will not line up with it directly.

A meta-regression pooled 33 randomized trials of vitamin D3, covering 43 treatment arms with at least 30 adult participants each.² At a mean baseline of 50.4 nmol/L, supplementation raised 25(OH)D by 37 nmol/L, with a confidence interval of 33 to 41 and significant heterogeneity between studies.² Four things independently predicted the size of the rise: dose, trial duration, baseline concentration and age.² The baseline slope was negative, at -0.19, and that is the entire argument for dosing by blood level rather than by label.² The same capsule does more for someone who starts low.

The subgroups sketch the shape of the curve. The pooled rise was larger at doses of 800 IU a day or more, at 39.3 nmol/L; larger again after 6 to 12 months, at 41.7 nmol/L; larger in people who began below 50 nmol/L, at 39.6 nmol/L; and largest in adults over 80 years, at 40.5 nmol/L.² The age result runs opposite to the common assumption that older adults get less out of a given dose.

A separate systematic review asked why identical doses produce such different increments. It pooled 144 cohorts from 94 independent studies in people aged 10 years and over.¹ The association between dose per kilogram of body weight per day and the rise in 25(OH)D was logarithmic rather than straight.¹ Dose per kilogram explained 34.5% of the variation; the type of supplement, D2 or D3, explained 9.8%; age 3.7%; taking calcium alongside 2.4%; and baseline 25(OH)D 1.9%.¹

Two things follow. A fixed capsule is a different dose per kilogram in a small person and a large one, which is part of why the same bottle produces different blood levels in the same household. And because the curve bends, the increments shrink as the dose climbs, so the gap between a moderate dose and a large one is smaller than the ratio printed on the labels.

The level the outcome data point to

Dosing by blood level only means something if some level is worth reaching. The clearest dose-response evidence in this set comes from people with type 2 diabetes or prediabetes, pooled from 21 cohort studies and 6 randomized trials.³ Against a status of 50 nmol/L or above, the relative risk of all-cause mortality was 1.36 for insufficiency, defined as 25 to below 50 nmol/L, and 1.58 for deficiency below 25 nmol/L.³ The curves were nonlinear, with the lowest risk at about 60 nmol/L for both all-cause and cardiovascular mortality.³

Then the trials in the same population. Supplementation did not reduce all-cause mortality: a relative risk of 0.96, with an interval from 0.79 to 1.16, across 6 trials and 7,316 participants, graded low-certainty.³ The association between a low level and dying is strong. The evidence that raising the level with a capsule changes that outcome is not. A blood level is a good way to size a dose; it is not by itself proof that the dose does anything.

D3, D2 and how often you dose

The form on the label matters less than the marketing implies, and how much it matters depends on the schedule. An early meta-analysis found D3 raised 25(OH)D more than D2 overall, at P = 0.001, but once dosing frequency was separated out the advantage was clear for bolus doses, at P = 0.0002, and lost with daily supplementation.⁴

A more recent meta-analysis of 20 comparative studies pushed back. Restricted to 12 daily-dosed comparisons measured by liquid chromatography-tandem mass spectrometry, the change in total 25(OH)D was 10.39 nmol/L lower on D2 than on D3, a shortfall of 40%, with an interval from -14.62 to -6.16.⁵ Body mass index turned out to be the strongest modifier of that result, cutting heterogeneity between studies to zero, and the D2 to D3 difference lost significance mainly in people with a BMI above 25 kg/m2.⁵ The form question and the body-size question keep turning out to be the same question.

Where too much begins

A risk assessment applied the Food and Nutrition Board's own methodology to the trial literature and concluded that the board's ceiling at the time, 50 micrograms or 2,000 IU a day, was not based on current evidence and was viewed by many as too restrictive.⁶ Going through the human clinical trials, it found no toxicity in healthy adults at doses of 250 micrograms a day, which is 10,000 IU of D3, and proposed that value as the upper limit instead.⁶

That is one analysis arguing a case, published before official limits were last revised, and what it describes is an absence of harm in trial participants rather than a benefit. The useful point for dosing is that every interesting decision sits far below the ceiling, in the range where the response curve is still steep.

What the guideline actually says about testing

The Endocrine Society's clinical practice guideline suggested measuring serum 25-hydroxyvitamin D with a reliable assay as the initial diagnostic test in patients at risk of deficiency, and recommended treatment with either D2 or D3 for deficient patients.⁷ It also drew two lines that get quoted far less often: there was not sufficient evidence to recommend screening people who are not at risk of deficiency, and not sufficient evidence to prescribe vitamin D for noncalcemic benefits such as cardiovascular protection.⁷

Set against the trial data, that is a consistent position. A test is how a dose becomes a decision for someone with reason to think they are deficient. For someone without that reason, the guideline most often cited to justify testing declines to recommend it.

What remains unknown

The dose-response literature is heterogeneous in a way nobody has resolved. The meta-regression reported significant heterogeneity among its studies, and the five predictors in the systematic review leave much of the between-cohort variation unexplained.²¹ The comparison of the two forms rests on thin numbers: body mass index data existed for only 13 of 17 daily-dosed comparisons in the newer analysis, and the earlier one could not verify effects across age, sex or ethnicity at all.⁵⁴

The larger gap is at the outcome end. The mortality dose-response comes from cohort studies, with certainty ranging from very low to moderate, in a population selected for diabetes, while the randomized trials in that same population found nothing.³ Nobody has randomized people to a dose chosen from their blood level against a fixed dose and followed them for outcomes that matter. Until that trial exists, dosing by blood level is a way to reach a target efficiently rather than evidence that the target is worth reaching.