In plain words

One small trial found that NMN made muscle respond better to insulin in women with prediabetes, and critics say its two groups started out unequal. A second trial in adults found no difference from placebo. Summaries that pooled several trials, which may include these two, disagree with each other.

The short answer

Two randomized trials point different ways. In 25 overweight or obese postmenopausal women with prediabetes, Yoshino and colleagues (2021) found muscle insulin sensitivity rose over 10 weeks on NMN and not on placebo; the ledger has its dose only from a later review. In 78 adults on 300, 600 or 900 mg a day for 60 days, Yi and colleagues (2022) found no difference from placebo in HOMA-IR. Three pooled analyses, which may include these trials, disagree and add no separate count. Certainty is low because both trials are small and measured insulin differently, and critics challenged the first over unequal starting liver fat.

Where the evidence sits

Insulin sensitivity improved

■ Randomized trial, Yoshino et al., 2021 · ledger has no n · 10 weeks · also reported by Yang et al., 2026
◪ Pooled analysis, Wang et al., 2025 · may overlap studies counted here

No difference from placebo

■ Randomized trial, Yi et al., 2022 · 78 people · 60 days
◪ Pooled analysis, Yang et al., 2026 · may overlap studies counted here
◪ Pooled analysis, Chen et al., 2024 · may overlap studies counted here

■ measured by the source's own study · ◪ pooled analysis, may overlap studies counted here · □ reported, reviewed or argued, no new data

Evidence weight6161 / 100?

Who says what

SourceFindingDesignPeopleDose and lengthFunding and independence
Yoshino et al., 2021Insulin sensitivity roseRandomized, placebo-controlled25 women, prediabetes250 mg/day, 10 weeks (per Yang)Not in ledger
Yi et al., 2022No HOMA-IR differenceRandomized, placebo-controlled78 adults300 to 900 mg/day, 60 daysNot in ledger
Wang et al., 2025Insulin resistance reducedMeta-analysis412Not in ledgerNot in ledger
Yang et al., 2026HOMA-IR trend, not significantMeta-analysis160Not in ledgerNot in ledger
Chen et al., 2024Lost significance without one trialMeta-analysisNot in ledgerNot in ledgerNot in ledger
Brenner et al., 2021Unequal starting liver fatComment on the trialSame 25Not in ledgerNo new data
Yoshino et al., 2021

Insulin sensitivity rose

Design
Randomized, placebo-controlled
People
25 women, prediabetes
Dose and length
250 mg/day, 10 weeks (per Yang)
Funding
Not in ledger
Yi et al., 2022

No HOMA-IR difference

Design
Randomized, placebo-controlled
People
78 adults
Dose and length
300 to 900 mg/day, 60 days
Funding
Not in ledger
Wang et al., 2025

Insulin resistance reduced

Design
Meta-analysis
People
412
Dose and length
Not in ledger
Funding
Not in ledger
Yang et al., 2026

HOMA-IR trend, not significant

Design
Meta-analysis
People
160
Dose and length
Not in ledger
Funding
Not in ledger
Chen et al., 2024

Lost significance without one trial

Design
Meta-analysis
People
Not in ledger
Dose and length
Not in ledger
Funding
Not in ledger
Brenner et al., 2021

Unequal starting liver fat

Design
Comment on the trial
People
Same 25
Dose and length
Not in ledger
Funding
No new data

Did NMN raise insulin sensitivity in prediabetes?

Yoshino and colleagues (2021) gave postmenopausal women with prediabetes and overweight or obesity NMN or placebo for 10 weeks; clamp-measured muscle glucose disposal rose on NMN only. Yang and colleagues (2026) appear to describe it: 25 such women on 250 mg a day. The authors concluded:

These results demonstrate that NMN increases muscle insulin sensitivity, insulin signaling, and remodeling in women with prediabetes who are overweight or obese

Did the groups start out alike?

Brenner and colleagues (2021) noted that the 13 women on NMN started with 6.3% liver fat and the 12 on placebo with 14.8% (P = 0.003), with no dose in the ledger:

Given that a target of NMN is liver fat clearance, this was not an effectively randomized trial.

Did higher doses change HOMA-IR?

Yi and colleagues (2022) gave 78 adults 300, 600 or 900 mg a day or placebo for 60 days:

The HOMA-IR showed no statistically significant differences for all NMN-treated groups as compared to placebo at day 60.

HOMA-IR rose from baseline at 600 and 900 mg (39 people, 60 days) and on placebo.

What do the pooled analyses add?

Three meta-analyses, each a pooled estimate that may include the trials above, disagree. Wang and colleagues (2025) found NMN reduced insulin resistance in 412 people, with no dose or duration in the ledger. Chen and colleagues (2024) found a marginal HOMA-IR improvement lost significance without one trial; the ledger has no n, dose or duration. Yang and colleagues (2026) pooled 160 people in five comparisons, with no dose or duration in the ledger:

HOMA-IR showed a mild improvement trend: MD = −0.22, 95% CI −0.55 to 0.11, p = 0.19, I2 = 0.0%.

Why the sources disagree

The first trial, as Yang and colleagues describe it, gave 250 mg a day to 25 women with prediabetes for 10 weeks and measured muscle glucose disposal with a clamp. The second gave 300 to 900 mg a day to 78 adults for 60 days and measured HOMA-IR against placebo. Brenner and colleagues add a further axis: unequal liver fat at the start of the first trial. The meta-analyses, which may include both trials, differ on which trials they pooled; Chen and colleagues lost significance after dropping one.

What would settle it

An independent randomized trial of NMN at 250 mg a day against placebo for at least 10 weeks in overweight or obese postmenopausal women with prediabetes, with groups balanced on liver fat at the start, the clamp as the pre-registered primary measure and HOMA-IR as a secondary one. A matching result from a second research group would raise certainty above low.

How this page was made

The 6 sources are two randomized trials, three meta-analyses that may include those trials, and a published comment on one trial, found through a literature check of PubMed records on NMN in people; studies of other compounds, animals or disease groups were left out. Every claim was copied word for word from its source into the Longevity Circle ledger and checked against the original text, and a person approved each one. "Contested" means the independent trials point in different directions; certainty follows the GRADE levels used in medical reviews. Drafting was assisted by AI.

Changelog: 23 Sep 2026, first version, 6 sources.