Research Methodology

What this hub is, and isn't, methodologically

NMN 101 was built through a structured literature review: a deliberate, repeated, documented search of primary and authoritative secondary sources against a consistent verification standard, applied page by page and revised on correction passes. It is not a formal systematic review in the technical sense — it does not follow a pre-registered protocol (such as PRISMA), was not conducted by multiple independent reviewers screening studies in duplicate, and its search was not exhaustively documented with a full PRISMA flow diagram. This distinction matters, and this hub does not describe its own process as a "systematic review" because that term carries specific methodological requirements this project does not claim to meet.

Search approach and databases used

Primary-source verification relied principally on PubMed/NCBI (via both standard search and the E-utilities API, which reliably returns article records even where a publisher's own page blocks automated access), publisher full-text pages (Nature, Frontiers, Scientific Reports, and others, where accessible), PubMed Central (PMC) full text where available, and DOI resolution to locate the authoritative publisher record for a given citation. Google Scholar and general web search were used only to locate candidate primary sources and their identifiers (PMID/DOI) — never as a substitute for inspecting the primary record itself once located.

Date cutoff

The evidence base was reviewed on a rolling basis across Batches 1 through 6 of this hub's construction, with a final consolidated cutoff of August 30, 2026. Individual pages carry their own "last reviewed" date reflecting when their specific content was most recently checked against the literature.

Inclusion and exclusion criteria

For the human intervention trial inventory specifically (see NMN Human Clinical Trials): included were published human intervention studies administering NMN and reporting a health-relevant outcome, identified through repeated PubMed/publisher searches rather than a single fixed query. Registered-but-unpublished trials were tracked separately and explicitly excluded from any efficacy conclusion. Preclinical (cell or animal) studies were tracked separately from the human inventory throughout, never merged into trial counts. Studies whose primary identity could not be confirmed through an authoritative bibliographic record (no PMID, no DOI, no confirmed journal indexing) were retained in the inventory for completeness but excluded from evidence synthesis and explicitly labeled Tier C.

Primary-source verification and the three-tier system

Every source cited for a material claim was classified into one of three verification tiers, established during Batch 4 and applied consistently since:

Tier A — Primary verified: the primary publication itself was directly inspected (via its own abstract, full text, or an authoritative bibliographic record) and its core study identity and outcomes were verified.

Tier B — Primary record verified, full outcomes partially verified: an authoritative primary record was confirmed to exist and match the citation, but some specific outcome detail could not be independently checked (for example, a paywalled full text where only the abstract was accessible).

Tier C — Identified but not independently verified: a study or report appears to exist, but its primary identity, full text, or indexing could not be sufficiently verified to this hub's standard.

It is important to be precise about what verification tier does and doesn't mean: verification confidence is not the same as study quality, is not the same as effect size, and is not the same as certainty of clinical benefit. A small, methodologically weak trial can still be Tier A if its record was directly and fully inspected; a large, well-conducted trial could in principle be Tier B or C if this review simply couldn't access its full record. Tier reflects how thoroughly this review could confirm a source, not how good the underlying study is.

Handling inaccessible sources

Where a source could not be accessed in full (a paywalled journal, a page blocked to automated retrieval, a non-indexed journal with no PMID/DOI), that limitation is stated explicitly in the citation's note field and reflected in its verification tier, rather than silently treated as equivalent to a fully-verified source.

Systematic reviews, meta-analyses, and duplicate populations

Systematic reviews and meta-analyses cited throughout this hub are explicitly labeled as pooled analyses of other trials, not counted as additional primary trials in any inventory total. Where multiple meta-analyses draw on overlapping sets of underlying trials, that overlap is flagged, and pooled participant counts from different reviews are never added together as if they represented distinct populations.

Endpoint classification: primary vs. secondary outcomes

For every trial cited, this hub distinguishes a study's prespecified primary endpoint from its secondary or exploratory outcomes, and states explicitly when a trial's most notable reported finding was secondary rather than primary. A secondary finding is never presented with the same evidentiary weight as a primary result.

Statistical interpretation

Where a trial or meta-analysis reports a statistically non-significant result, that is stated as null rather than reframed as inconclusive-but-suggestive. Where a meta-analysis's own authors flag risk-of-bias concerns in their included studies, or describe their own pooled finding as preliminary, that characterization is preserved rather than upgraded in this hub's own language. Small-sample statistical limitations (for example, an unusually small per-arm sample size, or an unexplained anomalous result in a control arm) are described in precise, specific terms rather than with vague language implying a formal scientific controversy where none has been identified.

Funding/COI verification

Funding and conflict-of-interest status is checked against each paper's own published disclosure statement or an equivalently authoritative primary record — never inferred from a different paper, a later disclosure, general author reputation, or an unrelated company relationship. Where a paper's own disclosure could not be located, that is labeled "Funding/COI not independently verified" rather than assumed in either direction.

Regulatory-source hierarchy

For Batch 6's regulatory content, sources are prioritized: (1) statute or regulation; (2) a regulator's own official database, decision, or determination; (3) regulator guidance documents; (4) official scientific opinions (which are explicitly distinguished from legal authorization); (5) official court decisions where materially relevant; (6) secondary legal or trade-press commentary, used only as context or to locate primary material, never as the sole basis for a stated regulatory conclusion. Where a primary regulatory document could not be directly retrieved during this review (for example, due to an access restriction on a government docket system), that limitation is stated explicitly on the relevant page rather than papered over with confident-sounding secondary-source language.

Claim-audit methodology

After a page is drafted, every material factual claim is individually reviewed against its cited source and classified DIRECT, INDIRECT, or UNSUPPORTED, per the definitions on Scientific & Editorial Policy. Claims that fail to reach DIRECT are narrowed to what the source actually supports, re-sourced to a source that does support the original wording, or removed outright. This audit is repeated on correction passes when new evidence or a sourcing gap is identified, with before/after totals reported.

Update process

Pages are revised when: new relevant primary literature is identified; a previously-cited source is found on re-inspection not to support its assigned claim; regulatory status changes; or a correction is requested and verified. Locked content from an earlier phase of this hub's construction is not edited casually — a genuine conflict between new evidence and existing locked content is reported explicitly rather than resolved by quietly rewriting the older page.

Related reading

For the policy this methodology implements, see Scientific & Editorial Policy. For the full reference index built using this methodology, see NMN Research Library. For the trial inventory that most extensively demonstrates this methodology in practice, see NMN Human Clinical Trials.