SCIENTIFIC METHOD
A transparent framework for identifying, verifying, classifying and interpreting scientific literature in ketogenic and metabolic research.
METHOD PRINCIPLES
The methodology is designed to keep bibliographic identification, evidence classification and scientific interpretation as separate steps. A paper is first identified and verified, then classified by study design and clinical area, and only afterwards interpreted in relation to the question it can actually answer.
DISCOVERY
PubMed is monitored for recent and historical literature related to ketogenic therapies, nutritional ketosis, VLCKD/VLEKT, modified Atkins approaches, ketone supplementation, ketone metabolism and clinically relevant metabolic interventions. Broad discovery is followed by a more conservative relevance filter.
VERIFICATION
Before a publication is promoted into the Scientific Library, its PMID is used as the primary bibliographic identifier. DOI and PMC links are taken from the PubMed record associated with that PMID. If PubMed does not provide a verified DOI, no DOI link is displayed.
Inclusion requires a direct relationship with ketogenic or metabolic interventions, ketone physiology, or a clinical field in which these approaches are being investigated. Mere mention of ketosis, ketone bodies or low-carbohydrate diets is not sufficient when the relationship to the research question is incidental.
Human randomized trials, other clinical interventions, observational studies, systematic reviews and meta-analyses, guidelines or consensus documents, case reports, reviews and preclinical or mechanistic studies are kept distinct. Randomization in an animal experiment is not treated as a randomized clinical trial.
A publication may be assigned to more than one clinical area when the connection is direct and scientifically meaningful. The purpose of the classification is navigation and evidence mapping, not to imply that a treatment is established for every area in which a paper appears.
SCIENTIFIC LIBRARY
The Scientific Library is the organized evidence layer. Records are deduplicated by PMID, DOI and normalized title, classified by evidence type and mapped to relevant clinical areas. Inclusion supports navigation; it is not an endorsement of a study’s conclusions.
LATEST EVIDENCE
Latest Evidence provides a rolling view of recently indexed PubMed literature. It is intentionally broader and more time-sensitive than the Scientific Library. Appearance in Latest Evidence does not by itself imply promotion into the structured Library.
Associations are not treated as causation. Statistical significance is not automatically equated with clinical importance. Preclinical findings are not generalized to humans without supporting human evidence. Conclusions are interpreted in proportion to study design, population, comparator, intervention and outcome quality.
Research Articles are intended to explain what a study adds, what can reasonably be inferred from it and which questions remain unresolved. They are not designed as translations of abstracts. Methodological detail is retained when it changes interpretation; otherwise the emphasis is placed on scientific meaning and limits.
Recent literature surveillance runs on a recurring schedule, while bibliographic integrity can be re-audited independently. Historical coverage has been extended retrospectively, but the database remains subject to correction when identifiers, classifications or source metadata are improved.
The website is designed for scientific evidence navigation and interpretation. It is not a clinical guideline, a treatment protocol or a substitute for individualized medical judgment. Inclusion of a publication does not establish efficacy, safety or clinical recommendation.
OPERATIONAL PROTOCOL
The parameters below describe the current production workflow as of 22 September 2026. They are published so that the retrieval and curation logic can be independently reproduced and audited.
Recent-literature surveillance uses PubMed through NCBI E-utilities and runs daily at 03:17 UTC. The discovery window is 365 days (maximum 500 PubMed records per run); Latest Evidence displays the most recent 90 days, limited to 100 records and sorted by publication date.
The scheduled historical backfill runs weekly on Sunday at 04:41 UTC. It moves backwards from 2024 toward 1970 in three-year windows, retrieves up to 3,000 PubMed identifiers per window and promotes at most 400 eligible records per run. Earlier historical literature has been added through separate audited backfill passes.
(
"ketogenic diet"[Title/Abstract] OR "ketogenic diets"[Title/Abstract]
OR "ketogenic therapy"[Title/Abstract] OR "ketogenic metabolic therapy"[Title/Abstract]
OR "nutritional ketosis"[Title/Abstract] OR "very low calorie ketogenic diet"[Title/Abstract]
OR "very-low-calorie ketogenic diet"[Title/Abstract] OR "very low energy ketogenic therapy"[Title/Abstract]
OR "modified Atkins diet"[Title/Abstract] OR VLCKD[Title/Abstract] OR VLEKT[Title/Abstract]
OR "exogenous ketone"[Title/Abstract] OR "exogenous ketones"[Title/Abstract]
OR "ketone ester"[Title/Abstract] OR "ketone esters"[Title/Abstract]
OR "ketone salt"[Title/Abstract] OR "ketone salts"[Title/Abstract]
OR (("beta-hydroxybutyrate"[Title/Abstract] OR "ketone bodies"[Title/Abstract])
AND (ketogenic[Title/Abstract] OR "nutritional ketosis"[Title/Abstract] OR "ketogenic therapy"[Title/Abstract]))
)A retrieved record is retained only when the title contains at least one core term (ketogenic, nutritional ketosis, modified Atkins, VLCKD, VLEKT, exogenous ketone, ketone ester or ketone salt), or when the title contains beta-hydroxybutyrate or ketone bodies and the title/abstract also contains a ketogenic-context term. This second step intentionally makes the public dataset narrower than the discovery query.
Clinical areas are assigned through a declared keyword map. A keyword match scores 6 points in the title, 2 in MeSH terms and 1 in the abstract. Broad areas require either a title match or a score of at least 3. Up to three areas are retained when their score is at least 2 and no more than four points below the top-scoring area. A new record is eligible for structured-Library promotion only when at least one mapped area is valid and the top confidence score is at least 2.
Duplicates are checked by PMID, normalized DOI and normalized title. Before promotion, the record is re-fetched from PubMed and the bibliographic identity is verified. DOI and PMC identifiers are accepted only from the identifier list belonging to that PubMed record, rather than from cited references. Records with no valid mapped area or insufficient classification confidence are not promoted automatically.
Evidence type is assigned using PubMed publication types, MeSH indexing and conservative title/abstract rules. The operational order is: systematic review/meta-analysis; guideline/consensus; case report/case series; preclinical/mechanistic evidence; randomized clinical trial; other clinical intervention; observational human study; review; other. Strong animal or in-vitro signals take precedence over incidental human wording so that animal experiments are not mislabeled as clinical trials.
PubMed is treated as the bibliographic source of record. Metadata and identifiers can be re-audited independently of discovery. If a correction, retraction or expression of concern is identified during re-audit, the affected Library record is reviewed and may be corrected, annotated or withdrawn as appropriate. This is a quality-control policy, not a claim of real-time retraction surveillance.
Recent PubMed-indexed literature in ketogenic and metabolic research, presented separately from the curated Scientific Library.
View latest publications