Correcting vitamin D deficiency fast — and where I overshot

Educational only. This is a record of my own N=1 experiment — it is not medical advice, and the doses described are what I took, not a recommendation. High-dose vitamin D belongs with a lab-confirmed deficiency and a doctor who is monitoring your calcium.

In February 2026 a blood test put my 25-OH vitamin D under 30 nmol/L — roughly 12 ng/mL, which is severe deficiency by any definition anyone uses. I wanted it corrected, and I wanted it corrected quickly rather than drifting upward over a year. So I read everything I could find, built a protocol around the best evidence I could get to, and ran it.

It worked. Twelve weeks later I was at 118 ng/mL — far past where I meant to stop. This article covers both halves: what actually makes correcting vitamin D deficiency fast possible, and how I sailed past the target while following careful advice. The second part turns out to be the more useful lesson, because the mistakes were not where I expected them.

How I got deficient in the first place

Worth stating, because the causes explain why an aggressive correction was reasonable rather than reckless.

I live at northern latitude, where winter UV-B is effectively zero for months — your skin can be in full sun and still produce nothing. On top of that I had spent three and a half weeks abroad taking no supplements at all, including three days of gastrointestinal illness that stripped magnesium, zinc and potassium along with everything else.

But the real lesson sits further back. I had supplemented before and it had not worked. Two reasons, both of which I only understood later: 4,000 IU is a maintenance dose, not a repletion dose, and I was taking it alongside psyllium husk. Soluble fibre binds fat-soluble vitamins and carries them straight through you. I had been diligently taking a supplement and diligently cancelling it out.

That combination — a deep hole, no sun, and a year of supplementation that had achieved nothing — is why I stopped tinkering and used a proper loading dose.

One thing I should be plain about, because it matters for how you read the rest of this. The 20,000 IU per day I used sits well above what any mainstream guideline recommends for repletion, which typically runs in the range of 2,000 to 6,000 IU daily for correcting a deficiency, sometimes with higher weekly dosing under supervision. I arrived at 20,000 through protocols used at the aggressive end of the functional medicine world. It is not a standard loading dose, and the rest of this article is partly an account of what that decision cost me.

Why dose is not the bottleneck — absorption is

Here is the thing almost every guide gets wrong. They argue about IU. The number on the bottle is the least interesting variable, because vitamin D is fat-soluble and a fat-soluble vitamin taken without fat is largely wasted.

The evidence points to roughly 20–25 grams of fat in the same meal for reliable absorption. I thought I had this covered: I was taking my dose with 10 ml of olive oil. That is about 9 grams. Less than half the requirement, and I had built my entire protocol on it.

My week 2.5 finger-prick test showed almost no movement off the floor. That result was the useful one of the whole protocol, because it forced me to look for the leak rather than reach for a higher dose. Once I added a real fat source to that meal — oily fish, cheese, nuts, avocado, anything that pushes the meal past 20 grams — the curve started moving.

The three rules that made the difference, in order of impact:

• Take D3 with a genuinely fatty meal. A teaspoon of oil does not count. Aim for 20–25 g of fat in that meal.
• Keep fibre supplements far away from the dose. I take psyllium at 07:00 and D3 at lunch — a three-hour gap minimum.
• Use D3, not D2. Cholecalciferol raises serum 25(OH)D substantially more effectively than ergocalciferol per unit.

None of this costs anything. All of it changes whether the dose does anything at all.

The cofactor that nearly derailed it

When I started loading, my tracked sleep quality dropped noticeably — from around 86% to 72% on my wearable’s scoring. Nothing else in my life had changed.

The likely explanation is magnesium. It is a required cofactor in the hydroxylation steps that convert vitamin D into its active form, so starting a high dose sharply increases demand. My intake at the time was 300 mg. Raising it to 450 mg, split between lunch and evening, brought the sleep score back — and it then settled above where it had started.

Some caution is warranted here. This is a single subject, unblinded, with other variables in motion. It is a strong hypothesis, not a finding. But the mechanism is well described, and if you start loading D3 and something feels off, magnesium is the first place I would look.

I also took vitamin K2 at 400 µg throughout, on the standard reasoning that it directs calcium toward bone rather than soft tissue. That is mechanistic reasoning rather than proven fact, though — no trial demonstrates that K2 prevents hypercalcaemia at high D3 doses. Treating it as a safety net was, in hindsight, more comfort than evidence.

What the twelve weeks actually looked like

Point	What happened	What it meant
Start (February)	Under 30 nmol/L, ~12 ng/mL	Severe deficiency, loading dose justified
Week 2.5	Finger-prick barely off the floor	Absorption problem, not a dosing problem
Week 3	Fat intake corrected to 20–25 g	The actual turning point
Week 3–4	Sleep quality dipped, magnesium raised to 450 mg	Cofactor demand rose with the dose
Week 11	Finger-prick read clearly “sufficient”	I concluded I was on track. I was not.
Week 13 (12 May)	Venous lab: 118.0 ng/mL	Far past target. Stopped D3 the next day.

Roughly ten weeks of real movement, from severe deficiency to well above the sufficient range. So the fast correction is achievable. The problem is what happens when it keeps going and you cannot see it.

Where it went past the target

Two things failed at once, and neither was the protocol itself.

The instrument went blind

On 27 April my cassette test showed a clear result in the sufficient band. I read it as somewhere around 35–50 ng/mL. Two weeks later the lab said 118.

This is assay saturation, and it is systematic rather than bad luck. These finger-prick cassettes lose resolution above roughly 60–70 ng/mL — past that point everything produces a similar line, so a good result and an excessive one look the same. They are excellent at confirming deficiency and useless for managing the upper range. If you are running a loading protocol, they will tell you when you are low and then quietly stop telling you anything at all.

The setpoint was too high

I was aiming at 60–80 ng/mL. Even overshooting a target of 50 would have left me somewhere reasonable. Overshooting 80 put me at 118.

That is the part worth sitting with. Every individual piece of advice I followed was defensible; the number I pointed all of it at was not. Careful execution aimed at the wrong setpoint gets you further from where you want to be, not closer.

Why my target was too high to begin with

I did research the target before starting. I asked an AI to lay out the landscape, and what came back was coherent: 60–80 ng/mL as the optimal range for systemic benefits, correct unit conversions, appropriate warnings about cofactors and testing, and four authorities behind it — Peter Attia, the Vitamin D Council, Michael Holick, and the Life Extension Foundation. Four independent sources converging on one range is normally the strongest signal that a question is settled. That coherence is exactly why I stopped looking.

Tracing the four sources

When I finally traced them, each one came apart:

Source	Actual position	Worth knowing
Peter Attia	40–60, aims ~50	Calls above 80 excessive; explicitly rejects 10–20,000 IU dosing
Michael Holick	40–60	Documented industry payments; his 2011 guideline was replaced in 2024
Vitamin D Council	50–80	Defunct since 2019; founder has since revised his own reasoning
Life Extension	50–80	Sells vitamin D supplements
Bryan Johnson	~40 steady state	Paused his 5,000 IU dose after reaching 134
Endocrine Society 2024	No threshold established	Found no trial evidence supporting a specific target level

Filter out the misattribution and the commercial interest and everything left converges on 40–60. Only the two weakest sources still support 60–80.

What the trials actually show

Look past the personalities and the same conclusion appears. Benefit in the literature comes from leaving the deficient range, not from climbing above the sufficient one — no trial has randomised people to 60–80 versus 40–50 and found an advantage. Meanwhile the costs keep climbing: calcium absorption tracks upward with vitamin D status, and one analysis found the calcium-phosphate product rising sharply once 25(OH)D passes about 50 ng/mL.

What I take from this is not that vitamin D is overhyped — I was genuinely deficient and correcting it was right. It is that an AI summary hands you the centre of gravity of everything written on a subject, not the weight of the evidence behind it. On vitamin D, supplement sellers write far more than trial researchers do.

The guardrails I would insist on now

Correcting a deficiency quickly is a legitimate goal. But speed only works with instruments that keep working at the top of the range, and that is where I came up short.

1. Set the target at 50 ng/mL, not 80. Overshoot is normal; build the margin into the setpoint rather than hoping it does not happen.
2. Book the venous lab at week 8, not week 12. My finger-prick check at week 11 was worthless precisely when it mattered.
3. Get calcium and PTH on the same form. I asked, was told these fell outside the standard panel, and let it drop. That was wrong — serum calcium is one of the most routine tests there is. What was missing was not the capability but a request reason. Bring the indication, not just the request: “I have been taking 20,000 IU daily since February — isn’t serum calcium indicated?” is a question the system has an answer for.
4. If you cannot get the monitoring, lower the dose. This is the one I would put above the others. Aggressive dosing without the ability to check calcium is not a calculated risk. It is an uncalculated one.

I stopped D3 entirely on 13 May and have taken none since. Vitamin A came down from 10,000 to 5,000 IU, K2 stayed on to handle calcium released as levels fall, and my target from here is 50 — not because a study proved 50 beats 60, but because it is the number that survives once you strip out everything that was being sold to me.

Frequently asked questions

How fast can you correct a vitamin D deficiency?

It depends almost entirely on how low you start, how well you absorb, and what dose a doctor considers appropriate for you. Repletion dosing is a clinical decision, not something to copy from a stranger’s blog — and it needs a lab test booked before you begin, not after you start wondering. For context on my own case: I moved from severe deficiency to well above the sufficient range in roughly ten weeks, but I used a dose far above standard repletion guidance and paid for it by overshooting.

Was 118 ng/mL dangerous?

Almost certainly not. Clinical toxicity is generally described above 150 ng/mL, and hypercalcaemia typically appears higher still. I had no symptoms. But without calcium and PTH data I cannot say so with confidence, and that gap in the record is permanent.

How long does it take to come back down?

Slowly. 25-OH vitamin D has a half-life measured in weeks, and fat tissue releases stored vitamin D gradually on top of that. Stopping is not a reset button — expect months, not days.

Are finger-prick vitamin D tests worth using?

For confirming deficiency, yes. For managing a loading protocol, no. Mine read “sufficient” while I was at 118, and it was not defective — that is simply what these assays do at the top of their range.

Where this leaves me

One wrinkle worth noting: I spent July in Peru at around 3,400 metres, close to the equator, outdoors every day. UV-B intensity rises roughly 10–12% per 1,000 metres of altitude, and near the equator the sun sits close to overhead. My body has almost certainly been making meaningful amounts of vitamin D on its own, which means my next test will not show a clean decay curve from 118. It shows something more useful anyway: where I land with sunlight and without pills. That is the number a maintenance dose should be built on.

The correction worked. That part I would do again, with the fat and the cofactors and the fibre timing exactly as they ended up. What I would change is the number I aimed at and the instrument I trusted to tell me when to stop.

If there is one thing this whole exercise taught me, it is that the figures we treat as settled are often just figures that got repeated a lot. Mine included.

Disclaimer. This article is published for educational purposes only and does not constitute medical advice, diagnosis, or treatment. I am not a clinician. Every dose, timing and target described here reflects what I personally did under conditions specific to me, following a deficiency confirmed by a laboratory — none of it is a recommendation for anyone else. Vitamin D is fat-soluble and accumulates in the body, and high doses carry real risks including hypercalcaemia. Consult a qualified healthcare professional before starting or changing any supplementation, and make sure calcium is on the request form.