Browse by intervention

Magnesium threonate: what it changed, and what it did not

One intervention, every measured benefit and result that did not improve in Issue 01. See who it moved for, where it stayed neutral or went the other way, and the next hypothesis Phoenix can validate.

Research issue: Issue 01
Data through: 2026-08-09
Evidence map version: v1.1.0

Explore the evidence Get the Evidence Map

How to read this

Phoenix research is longitudinal and observational: we follow real APOE4 carriers over time and compare each member with their own earlier self, with no placebo group, no randomisation, and members choosing their own interventions.

That means these are strong signals pointing at what to test next, not proof of cause. Most nutrition, supplement and lifestyle research you read online has exactly the same design. The difference is that we say so, and we show you the confounders.

Read the full methods and limits

Most promising

Deep plus REM sleep

Verified wearable deep and REM sleep stages · Aim: higher

Sleep mask

✓Repeated positive

One curated case had 134 more deep-plus-REM minutes; the 11-member ten-tag-day cohort averaged about 17 more minutes and stayed positive at every floor.

Selected member cases

Fine Tuner · APOE3/4 · female · age 50-59
Sleep-mask nights carried 134 more minutes of deep plus REM sleep under independent recomputation.
What else changed: no nearby starts were recorded.
Next study: repeat Deep plus REM sleep after a dated Sleep mask window.

1 selected member case · not a response rate

Logged members: 19
Members with repeats: 11
Observations: 1

Sources and context

Members with a sleep-mask tag and verified wearable stages, evaluated at named minimum tagged-day floors.

report-lines-343-412, reviewed by independent verifier

1 selected case shown with this evidence.

Earlier bedtime

✓Repeated positive

Three curated earlier-bedtime cases carried part of their longer sleep in deep plus REM stages; exact group-stage counts are not published.

Selected member cases

Deep Diver · APOE4/4 · female · age not disclosed
Deep plus REM sleep carried about 43 minutes of her earlier-night advantage.
What else changed: no nearby starts were recorded.
Next study: repeat Deep plus REM sleep after a dated Earlier bedtime window.
Night Shift · APOE3/4 · female · age 50-59
Deep plus REM sleep carried 33 to 53 minutes of her earlier-night advantage.
What else changed: no nearby starts were recorded.
Next study: repeat Deep plus REM sleep after a dated Earlier bedtime window.
Omega Lifter · APOE4/4 · male · age 40-49
Deep plus REM sleep carried roughly half of his 73-to-93-minute earlier-night advantage.
What else changed: no nearby starts were recorded.
Next study: repeat Deep plus REM sleep after a dated Earlier bedtime window.

3 selected member cases · not a response rate

Observations: 3
Nights: 53

Sources and context

Curated earlier-bedtime cases with verified wearable sleep-stage data.

report-lines-343-412, reviewed by independent verifier

3 selected cases shown with this evidence.

ZenoWell taVNS

✓Repeated positive

Two curated ZenoWell cases carried 44 and 23 more deep-plus-REM minutes on use nights; five of 40 study members had valid wearable coverage.

Selected member cases

Current Convert · APOE4/4 · female · age 60-69
Use nights carried 44 more deep plus REM minutes across 60 use and 91 non-use nights.
What else changed: no nearby starts were recorded.
Next study: repeat Deep plus REM sleep after a dated ZenoWell taVNS window.
Signal Seeker · APOE3/4 · female · age 50-59
Use nights carried 23 more deep plus REM minutes across 77 total nights.
What else changed: no nearby starts were recorded.
Next study: repeat Deep plus REM sleep after a dated ZenoWell taVNS window.

2 selected member cases · not a response rate

Logged members: 40
Members with repeats: 5
Observations: 2
Nights: 228
Measurements: 228

Sources and context

Device-study members with real wearable data on logged use and non-use nights; placeholder rows are excluded.

report-lines-414-430, reviewed by independent verifier

2 selected cases shown with this evidence.

Subjective sleep quality

Phoenix daily check-in sleep-quality score · Aim: higher

Earlier bedtime

✓Repeated positive

Among APOE4/4 carriers logging early bedtime, better subjective sleep ratings outnumbered worse ratings; no public responder percentage is generated.

0 selected member cases · not a response rate

Logged members: 60
Members with repeats: 60

Sources and context

APOE4/4 carriers with early-bedtime tags and comparable subjective sleep ratings.

report-lines-343-412, reviewed by independent verifier

Mediterranean-keto pattern

✓Positive individual result

The Pattern Spotter's Mediterranean-keto days averaged 1.3 sleep-quality points higher.

Selected member cases

Pattern Spotter · APOE4/4 · female · age 60-69
Sleep quality averaged 1.3 points higher on Mediterranean-keto days.
What else changed: no nearby starts were recorded.
Next study: repeat Subjective sleep quality after a dated Mediterranean-keto pattern window.

1 selected member case · not a response rate

Logged members: 1
Members with repeats: 1
Observations: 1

Sources and context

The curated Pattern Spotter comparison of Mediterranean-keto and other days.

report-lines-433-457, reviewed by independent verifier

1 selected case shown with this evidence.

Sauna

✓Positive individual result

The Sauna Sleeper rated sleep 7.8 on sauna days versus 5.8 on comparison days, a 2.1-point advantage after rounding.

Selected member cases

Sauna Sleeper · APOE4/4 · male · age under-40
Sleep quality averaged 7.8 on 17 sauna days versus 5.8 across 97 comparison days.
What else changed: no nearby starts were recorded.
Next study: repeat Subjective sleep quality after a dated Sauna window.

1 selected member case · not a response rate

Logged members: 1
Members with repeats: 1
Observations: 114
Measurements: 114

Sources and context

The Sauna Sleeper's 17 sauna days compared with 97 non-sauna days.

report-lines-433-457, reviewed by independent verifier

1 selected case shown with this evidence.

Total sleep duration

Verified wearable total sleep duration · Aim: higher

Earlier bedtime

✓Repeated positive

Two analyses found earlier nights about 30 minutes longer: 56 of 68 eligible members under one rule and 59 of 66 eligible members under a stricter rule. No blended response rate is published.

Selected member cases

Deep Diver · APOE4/4 · female · age not disclosed
Earlier nights were 74 to 80 minutes longer across 53 nights of wearable data.
What else changed: no nearby starts were recorded.
Next study: repeat Total sleep duration after a dated Earlier bedtime window.
Night Shift · APOE3/4 · female · age 50-59
Earlier nights were 77 to 120 minutes longer across two independent recomputations.
What else changed: no nearby starts were recorded.
Next study: repeat Total sleep duration after a dated Earlier bedtime window.
Omega Lifter · APOE4/4 · male · age 40-49
Earlier nights were 73 to 93 minutes longer across two independent recomputations.
What else changed: no nearby starts were recorded.
Next study: repeat Total sleep duration after a dated Earlier bedtime window.

3 selected member cases · not a response rate

Observations: 3
Nights: 53

Sources and context

Members eligible under each named within-member median-bedtime analysis; the two cohorts remain separate.

report-lines-343-412, reviewed by independent verifier

3 selected cases shown with this evidence.

ZenoWell taVNS

✓Positive individual result

The Signal Seeker averaged 522 total sleep minutes on use nights versus 494 on comparison nights across 77 nights.

Selected member cases

Signal Seeker · APOE3/4 · female · age 50-59
Use nights averaged 522 total sleep minutes versus 494 on comparison nights across 77 nights.
What else changed: no nearby starts were recorded.
Next study: repeat Total sleep duration after a dated ZenoWell taVNS window.

1 selected member case · not a response rate

Logged members: 40
Members with repeats: 5
Observations: 1
Nights: 77
Measurements: 77

Sources and context

Device-study members with real wearable total sleep on logged use and non-use nights.

report-lines-414-430, reviewed by independent verifier

1 selected case shown with this evidence.

Mixed signals

Subjective sleep quality

Phoenix daily check-in sleep-quality score · Aim: higher

Morning caffeine

✓ / ↘Mixed results

One curated member rated morning-caffeine sleep lower while another felt no change despite a wearable-stage decline.

Selected member cases

Current Convert · APOE4/4 · female · age 60-69
Her subjective sleep rating dropped on morning-caffeine days.
What else changed: no nearby starts were recorded.
Next study: repeat Subjective sleep quality after a dated Morning caffeine window.
Fine Tuner · APOE3/4 · female · age 50-59
Her subjective sleep rating was unchanged despite lower wearable deep plus REM sleep.
What else changed: no nearby starts were recorded.
Next study: repeat Subjective sleep quality after a dated Morning caffeine window.

2 selected member cases · not a response rate

Logged members: 2
Members with repeats: 2
Observations: 2

Sources and context

Curated morning-caffeine cases with comparable subjective sleep ratings.

report-lines-343-412, reviewed by independent verifier

2 selected cases shown with this evidence.

No clear change

Deep plus REM sleep

Verified wearable deep and REM sleep stages · Aim: higher

Untimed caffeine or coffee

—Neutral result

Untimed caffeine and coffee tags showed no deep-plus-REM pattern, unlike the explicitly timed morning-caffeine tag.

0 selected member cases · not a response rate

Sources and context

Members with untimed caffeine or coffee tags and verified wearable sleep-stage data.

report-lines-343-412, reviewed by independent verifier

Total sleep duration

Verified wearable total sleep duration · Aim: higher

Sleep regularity

—Neutral result

Sleep regularity did not outperform sleep duration on any matched-night definition tested in Issue 01.

0 selected member cases · not a response rate

Sources and context

Members with matched wearable nights under the tested bedtime-regularity definitions.

report-lines-559-604, reviewed by independent verifier

What may work against

Deep plus REM sleep

Verified wearable deep and REM sleep stages · Aim: higher

Morning caffeine

↘Opposite direction

Deep plus REM sleep was 3 minutes lower at one tagged day and 8 minutes lower among 24 members with at least ten tagged days.

Selected member cases

Current Convert · APOE4/4 · female · age 60-69
Morning-caffeine days carried 111 fewer minutes of deep plus REM sleep.
What else changed: no nearby starts were recorded.
Next study: A placebo-controlled crossover found dose and clock time mattered, supporting explicit dose-and-time capture instead of one blanket caffeine label.
Fine Tuner · APOE3/4 · female · age 50-59
Morning-caffeine days carried 82 fewer minutes of deep plus REM sleep while felt sleep was unchanged.
What else changed: no nearby starts were recorded.
Next study: A placebo-controlled crossover found dose and clock time mattered, supporting explicit dose-and-time capture instead of one blanket caffeine label.

2 selected member cases · not a response rate

Logged members: 43
Members with repeats: 24
Observations: 2

Sources and context

Members with a morning-caffeine tag and wearable sleep, evaluated at separate minimum tagged-day support floors.

report-lines-343-412, reviewed by independent verifier

2 selected cases shown with this evidence.

Not yet measured

A placebo-controlled crossover found dose and clock time mattered, supporting explicit dose-and-time capture instead of one blanket caffeine label.

How to test it

Baseline: Record caffeine-free wearable nights before the timed exposure block.
Duration: 14 days
Adherence: Log dose and clock time while wearing the device every night.
Retest: Compare deep plus REM sleep after the two-week tagged block.
Falsifier: Timed caffeine tags do not reproduce lower deep plus REM sleep as dose or proximity to bedtime increases.
Eligible-member target: 40 members

Training and workouts

↘Opposite direction

Across 32 members with at least ten workout tags, workout nights averaged about 11 fewer deep-plus-REM minutes.

0 selected member cases · not a response rate

Logged members: 32
Members with repeats: 32

Sources and context

Members with at least ten workout tags and verified wearable deep-plus-REM data.

report-lines-460-507, reviewed by independent verifier
report-lines-559-604, reviewed by independent verifier

This result was withdrawn: An earlier plus-42-minute workout line reversed direction under independent recomputation.

Deep sleep

Verified wearable deep-sleep stage · Aim: higher

Training and workouts

↘Opposite direction

The Trade-Off Athlete's same-night deep sleep was 30% to 35% lower across two computations.

Selected member cases

Trade-Off Athlete · APOE4/4 · male · age 50-59
Same-night deep sleep was 30% to 35% lower across two independent computations.
What else changed: no nearby starts were recorded.
Next study: repeat Deep sleep after a dated Training and workouts window.

1 selected member case · not a response rate

Logged members: 1
Members with repeats: 1
Observations: 1

Sources and context

The curated Trade-Off Athlete comparison of workout and non-workout wearable nights.

report-lines-460-507, reviewed by independent verifier

1 selected case shown with this evidence.

Subjective sleep quality

Phoenix daily check-in sleep-quality score · Aim: higher

External stress

↘Opposite direction

External-stress weeks averaged 1.3 sleep-quality points lower; only a handful of dense loggers contributed.

0 selected member cases · not a response rate

Sources and context

Members with sufficiently dense external-stress tags and comparable other-week sleep ratings.

report-lines-433-457, reviewed by independent verifier

Travel

↘Opposite direction

Travel weeks averaged 1.7 sleep-quality points lower; only a handful of dense loggers contributed.

0 selected member cases · not a response rate

Sources and context

Members with sufficiently dense travel tags and comparable home-week sleep ratings.

report-lines-433-457, reviewed by independent verifier

Alcohol

↘Opposite direction

The Pattern Spotter's drinking nights averaged 1.72 sleep-quality points lower across 25 drinking and 35 comparison nights.

Selected member cases

Pattern Spotter · APOE4/4 · female · age 60-69
Sleep quality averaged 1.72 points lower across 25 drinking nights versus 35 comparison nights.
What else changed: no nearby starts were recorded.
Next study: repeat Subjective sleep quality after a dated Alcohol window.

1 selected member case · not a response rate

Logged members: 1
Members with repeats: 1
Observations: 60
Nights: 60
Measurements: 60

Sources and context

The Pattern Spotter's drinking nights compared with her own non-drinking nights.

report-lines-433-457, reviewed by independent verifier

1 selected case shown with this evidence.

Audit history

These historical records remain beside the current evidence so you can see why their status changed.

Deep plus REM sleep

Sauna

!Record under review

Wearable sauna analyses disagree: one wearable analysis found deep plus REM lower in most of 15 members; a corrected larger set split nearly evenly.

0 selected member cases · not a response rate

Logged members: 15
Members with repeats: 15

Sources and context

Members with sauna tags and corrected verified wearable sleep-stage coverage under each analysis version.

report-lines-559-604, reviewed by independent verifier

This result is under review: The two wearable computations remain visible and excluded from active ranking while Phoenix collects corrected nights.

Deep sleep

Alcohol

!Record withdrawn

An apparent alcohol-improves-deep-sleep result dissolved under device-coverage checks and is retained only as a retraction.

0 selected member cases · not a response rate

Sources and context

Historical alcohol-tagged wearable nights before corrected device-coverage checks.

report-lines-559-604, reviewed by independent verifier

This result was withdrawn: The positive deep-sleep result was a wearable-coverage artifact and remains excluded from active evidence.

Most promising

HbA1c

Laboratory haemoglobin A1c result · Aim: lower

Magnesium threonate

✓Repeated positive

The same two magnesium-threonate cases nudged HbA1c lower, but Issue 01 does not publish their before-and-after values.

Selected member cases

Magnesium Match · APOE4/4 · female · age 50-59
HbA1c moved lower in the same clean window; public before-and-after values were not printed.
What else changed: no nearby starts were recorded.
Next study: repeat HbA1c after a dated Magnesium threonate window.
Magnesium Mover · APOE4/4 · male · age 50-59
HbA1c moved lower in the same clean window; public before-and-after values were not printed.
What else changed: nearby starts included B-vitamin intake not captured.
Next study: repeat HbA1c after a dated Magnesium threonate window.

2 selected member cases · not a response rate

Logged members: 87
Members with repeats: 2
Observations: 2
Measurements: 4

Sources and context

Members with magnesium-threonate exposure and a repeat HbA1c result in the reviewed case set.

report-lines-159-189, reviewed by independent verifier

2 selected cases shown with this evidence.

Homocysteine

Laboratory homocysteine result · Aim: lower

Magnesium threonate

✓Repeated positive

Two clean curated cases lowered homocysteine by 28.8% and 31.9%; these are two cases, not a response rate.

Selected member cases

Magnesium Match · APOE4/4 · female · age 50-59
Homocysteine fell from 9.40 to 6.40 µmol/L, down 31.9% across a 26-month clean window.
What else changed: no nearby starts were recorded.
Next study: No human magnesium-threonate homocysteine study was found; B-vitamin exposure is the main alternative to separate.
Magnesium Mover · APOE4/4 · male · age 50-59
10.53 → 7.5 µmol/L · -28.8% · 64 days
What else changed: nearby starts included B-vitamin intake not captured.
Next study: No human magnesium-threonate homocysteine study was found; B-vitamin exposure is the main alternative to separate.

2 selected member cases · not a response rate

Logged members: 142
Members with repeats: 33
Observations: 2
Measurements: 4

Sources and context

Broad magnesium-family users with a repeat homocysteine result are support context, not clean threonate responders.

report-lines-248-343, reviewed by independent verifier
report-lines-129-156, reviewed by independent verifier

2 selected cases shown with this evidence.

Not yet measured

No human magnesium-threonate homocysteine study was found; B-vitamin exposure is the main alternative to separate.

How to test it

Baseline: Record a homocysteine result before the dated start.
Duration: 84 days
Adherence: Track magnesium and B-vitamin use throughout the window.
Retest: Repeat homocysteine after the planned window.
Falsifier: Homocysteine does not fall in clean magnesium-threonate windows once B-vitamin exposure is dated and separated.
Eligible-member target: 40 members

Keep the map

Take every result into your next experiment.

Get the Issue 01 Evidence Map by email. It groups the same evidence by outcome and by intervention, including what worked, what did not, and what Phoenix wants to test next.

Instant PDF · Issue 02 lands in your inbox the day it publishes · No spam

Turn evidence into your next test

Magnesium and homocysteine is 7 members from an answer.

Forty carriers with a dated start and a repeat blood test close it. Creatine, vitamin D3 and B12 sit further back. Your next blood test is somebody else's answer, and theirs is yours.

Take the 7-minute assessment

Full Issue 01 index

Two ways into the same evidence

Start with the result you want, or start with something you are already taking or doing. Both paths lead back to the same measured records.