Population
Adults aged 18 or older; general and high-risk populations kept separate when baseline risk differs.
ITProtocol v0.8
How this evidence audit will be completed.
Rules published before the full search, so future decisions can be checked against the plan rather than reconstructed after the result.
Version0.8 · 13 August 2026
RegistrationPending
Current corpus21 curated source cards
Definitive target50–100 study families
01 · Review question
One question, four outcomes, no collapsed comparators.
The primary question is whether processed or unprocessed red-meat intake is associated with colorectal cancer, type 2 diabetes, cardiovascular events or mortality in adults. Short-term cardiometabolic markers form a separate evidence stream.
Exposure
Processed and unprocessed red meat recorded separately, including amount, frequency, preparation and product definition.
Comparator
Lower or zero intake, another intake category, or a named replacement food. An implicit comparator is never relabelled as a substitution.
Outcomes
Colorectal cancer; incident type 2 diabetes; cardiovascular events and all-cause mortality; short-term markers shown separately.
02 · Eligibility
Different evidence streams answer different questions.
Included
- Prospective and other longitudinal human studies for incident disease or mortality.
- Randomized or controlled feeding trials for cardiometabolic markers—not as direct proof of multi-year events.
- Case-control cancer evidence kept distinct from prospective cohorts.
- Mechanistic evidence as support, never a human absolute-risk estimate.
- Systematic reviews and official assessments for discovery and triangulation, without double-counting overlapping primary studies.
- No planned date, language or publication-status limit.
Excluded from the primary causal synthesis
- Cross-sectional and ecological associations.
- Animal-protein aggregates that cannot isolate red or processed meat.
- Acute meal responses without a controlled comparator.
- Editorials, narrative reviews and unsupported claims.
- Duplicate reports counted as if they were separate studies.
03 · Search plan
The search must be reproducible down to the syntax and date.
The final strategies will be peer reviewed by an information specialist and published verbatim with platform, execution date and result count. That standard has not yet been reached.
Planned information sources
- MEDLINE/PubMed, Embase and CENTRAL
- Trial registries and backward/forward citation searching
- Relevant reference lists and official WHO/IARC and WCRF material
- Publisher correction and discussion pages
- Crossref/Retraction Watch and exact-DOI PubPeer monitoring
Broad PubMed seed query · v0.1
("red meat"[Title/Abstract] OR "processed meat"[Title/Abstract]
OR beef[Title/Abstract] OR pork[Title/Abstract] OR lamb[Title/Abstract]
OR mutton[Title/Abstract])This exposure-led seed remains the starting point for the definitive strategy. The operational v0.3 layer below combines it with outcome concepts, Create Date and a one-day overlap; controlled vocabulary and validated design filters still require information-specialist review.
Operational retrieval · v0.4 · query v0.3
The screening receipt now preserves what was actually read.
Create Date targets citations newly created in PubMed, including old publications whose Entry Date can be backdated. The rolling window repeats the previous calendar day so records added after the prior 08:00 run are not skipped. ESummary and EFetch preserve the citation and complete screening material. This remains a discovery layer, not the definitive multi-database search.
Exact executed query
(("red meat"[Title/Abstract] OR "processed meat"[Title/Abstract]
OR beef[Title/Abstract] OR pork[Title/Abstract] OR lamb[Title/Abstract]
OR mutton[Title/Abstract])
AND ("colorectal cancer"[Title/Abstract]
OR "colorectal neoplasms"[Title/Abstract]
OR "type 2 diabetes"[Title/Abstract]
OR cardiovascular[Title/Abstract] OR "heart disease"[Title/Abstract]
OR stroke[Title/Abstract] OR mortality[Title/Abstract] OR death[Title/Abstract]
OR LDL[Title/Abstract] OR cholesterol[Title/Abstract]
OR "blood pressure"[Title/Abstract] OR glycemic[Title/Abstract]
OR glucose[Title/Abstract] OR insulin[Title/Abstract]
OR inflammation[Title/Abstract] OR TMAO[Title/Abstract]
OR homocysteine[Title/Abstract] OR cardiometabolic[Title/Abstract]))
AND ("2026/08/12"[Create Date] : "2026/08/13"[Create Date])Title/abstract dispositions
PMIDDecisionPrimary reason
42589697 ↗Nutrition as Modulator of Oxidative Stress in Cancer Prevention and TreatmentJournal Article · Review · Full abstract retrieved
ExcludeNarrative review with a broad diet-and-cancer discussion; no eligible primary red-meat contrast.
42583264 ↗Global, regional, and national tuberculosis burden associated with six suboptimal dietary exposuresJournal Article · Full abstract retrieved
ExcludeTuberculosis burden is outside the prespecified outcomes.
42588654 ↗Real-World Evidence for Breast Cancer Risk Stratification: Insights from the P.I.N.K. StudyJournal Article · Full abstract retrieved
ExcludeBreast cancer is outside the prespecified outcomes.
42592398 ↗The Urgent Need to Address Modifiable Risk Factors to Prevent Young-Onset Colorectal Cancer in MexicoJournal Article · Review · Full abstract retrieved
ExcludeNarrative review without an original eligible red-meat contrast or a reproducible systematic-review selection flow.
No record entered the decision corpus; the scientific conclusion is unchanged.
These are search-flow decisions, not claims that the papers are invalid. Plausibly relevant full-text exclusions remain in the separate screening register.
04 · Selection and appraisal
The unit is the study family—not the PDF.
Merge
Combine sources and remove duplicate records.
Screen
Two reviewers independently screen titles/abstracts and full texts.
Explain
Record one primary reason for each plausibly relevant full-text exclusion.
Link
Join registrations, reports, supplements, corrections and secondary analyses into one study family.
Resolve
Settle disagreements by consensus or a third reviewer.
Verify
Two reviewers check every decision-relevant extraction and appraisal.
Fields extracted
Population, exposure measurement, comparator, outcome definition, follow-up, relative and absolute estimates, uncertainty, adjustment set, missing data, sensitivity analyses, funding, conflicts, sponsor role, registration, data/code access and post-publication notices or discussion.
Design-specific tools
- ROBINS-E
- Non-randomized exposure studies
- RoB 2
- Randomized trials
- AMSTAR 2 / ROBIS
- Systematic reviews
- GRADE domains
- Outcome-level certainty with explicit reasons
05 · Synthesis
Pool only what is defensibly comparable.
- 01
Keep processed and unprocessed meat, outcomes, designs and materially different comparators separate.
- 02
Report between-study variance and prediction intervals when estimable; do not replace interpretation with a p-value threshold.
- 03
Convert relative effects to absolute effects only with a defensible baseline risk, exposure contrast and time horizon.
- 04
Preplanned sensitivity checks cover risk of bias, replacement food, BMI or other mediators, lag periods, outcome definitions, follow-up, sponsor involvement and overlapping study families.
06 · Living-update rule
Every check is logged. Only meaningful evidence moves the answer.
Operational surveillance runs daily. A new record is screened against the same criteria; the public conclusion changes only after eligibility, appraisal and human review. Silence is not an update: an unchanged conclusion is logged too.
Scientific
An eligible source can materially change an effect estimate, certainty or decision boundary.
Integrity
A correction, expression of concern, retraction or substantive formal discussion appears.
Methods
A protocol rule changes; the amendment is dated and explained before reuse.
Editorial
A material limitation, funding relationship or data/code status was missing or misstated.
Known limits of v0.8
- Not prospectively registered.
- Final database strategies are not yet information-specialist reviewed or fully executed.
- The 21 source cards are curated, not a deduplicated dual-screened corpus.
- Independent dual extraction, risk-of-bias adjudication and external review remain pending.
Amendment log
0.8 · 13 August 2026 — made English and Italian improvement summaries separate required receipt fields; finalization fails if either is missing or undersized, and each protocol renders only its matching language. · 0.7 — bound editorial completion to the exact successful automated receipt. A failed control or later same-day rerun now invalidates the prior editorial state until that specific receipt is finalized. · 0.6 — added an atomic command-driven editorial finalizer with explicit counts and yes/no declarations. · 0.5 — added a date-matched editorial receipt; stale completion is exposed as pending. · 0.4 — changed the operational date field to Create Date and added a one-day rolling overlap. · 0.3 — added complete MEDLINE records and hard completeness checks. · 0.2 — added the first Entry Date query and machine-readable receipt. · 0.1 — first public draft.