Glossary

Plain-language definitions for terms used across the Oncology Failure Atlas.

OFA: Oncology Failure Atlas. A structured atlas of 11,525 terminated and withdrawn oncology trial records, built from public sources only.
MDIP™: Mapped Discontinuation Intelligence Patterns. MDIP™ is OFA's discontinuation-pattern taxonomy for mapping public stop-reason language from terminated or withdrawn oncology trials into structured, comparable categories (e.g. Recruitment / Enrollment, Sponsor Strategic / Portfolio, Efficacy / Futility, Safety / Toxicity, Funding / Resource, Unclear / Not Specified, Unclear Why Stopped — Reported). MDIP™ categories are discontinuation-pattern categories, not biological mechanisms. They support research navigation and comparison, but they do not determine definitive root cause, clinical meaning, regulatory interpretation, investment significance, or future outcome.
Mapped Discontinuation Intelligence Patterns: Long form of MDIP. OFA's discontinuation-pattern taxonomy for mapping public stop-reason language into structured, comparable categories. Not a biological mechanism, not a prediction.
Discontinuation Pattern: A category from the MDIP™ taxonomy describing the publicly reported reason a trial was terminated or withdrawn. A pattern, not a cause.
Harmonized Mechanism Class: The product-facing mechanism label used across OFA. Source: public.trial_mechanism_harmonization.harmonized_mechanism_class, joined to trials by nct_id. Mechanism filtering and mechanism distributions use this field — never the legacy raw public.trials.mechanism_class.
trial_mechanism_harmonization.harmonized_mechanism_class: The harmonized mechanism source-of-truth. Replaces the raw mechanism_class field for all product-facing mechanism views, filters, and exports.
public.trials.mechanism_class (legacy / raw): A legacy, raw, and contaminated field on the base trials table. Present in storage for provenance but excluded from product surfaces because it mixes naming conventions, formulations, and aliases. Do not use as a product-facing mechanism.
Confidence labels: Indicate how clearly the reported public stop-language supports the assigned MDIP discontinuation-pattern category. They are classification-confidence labels only. They do not indicate clinical certainty, regulatory certainty, scientific validity, or probability of future outcomes.
RecurSignal™: A historical recurrence signal (0–100) reflecting how closely a trial's discontinuation pattern resembles previously documented patterns in the same MDIP™ category. Not a prediction, probability, forecast, risk score, or clinical/investment recommendation.
why_stopped: The free-text field on ClinicalTrials.gov where sponsors describe why a trial was terminated or withdrawn. It is the primary signal for Phase 2 deterministic MDIP™ v4 classification.
Deterministic classification: Rule-based classification that always produces the same output for the same input. Phase 2 MDIP™ v4 classification is deterministic and auditable from the source why_stopped language — no LLM was used.
UNCLEAR_WHY_STOPPED_REPORTED: Phase 2 MDIP™ v4 category for trials where a why_stopped value is reported but is too ambiguous to map to a specific discontinuation-pattern category. A transparent bucket, not a hidden failure.
UNCLEAR_NOT_SPECIFIED: Phase 2 MDIP™ v4 category for trials where no why_stopped value is reported. A transparent bucket, not a hidden failure.
RECRUITMENT_ENROLLMENT: Discontinuation driven by insufficient enrolment, slow accrual, or inability to retain participants.
SPONSOR_STRATEGIC_PORTFOLIO: Discontinuation tied to publicly reported sponsor, business, or pipeline/portfolio decisions, not to a stated safety or efficacy signal.
FUNDING_RESOURCE: Discontinuation tied to loss of funding, grant termination, or sponsor financial constraints.
SAFETY_TOXICITY: Discontinuation following a reported safety signal, adverse event pattern, or toxicity concern.
INVESTIGATOR_SITE_OPERATIONAL: Discontinuation driven by investigator departure, site closure, or other operational/site-level issues.
EFFICACY_FUTILITY: Discontinuation following an interim analysis, futility finding, or lack of demonstrated clinical benefit.
REGULATORY_ADMINISTRATIVE: Discontinuation tied to regulatory action, IRB/EC decisions, or administrative changes.
MANUFACTURING_SUPPLY: Discontinuation tied to drug supply, manufacturing, or product-availability issues.
COVID_EXTERNAL_SHOCK: Discontinuation explicitly attributed to COVID-19 or another external shock disrupting trial conduct.
FORMULATION_PK_OPTIMIZATION: Discontinuation to allow reformulation, dose optimization, or pharmacokinetic refinement before continuing development.
PubMed NONE: No PubMed match was found by the current matching process for that trial record. It does not prove absence of publication, scientific evidence, or related literature.
OpenFDA NO_MATCH: No OpenFDA match was found by the current matching process for that trial record. It does not prove absence of safety evidence, adverse-event reporting, regulatory history, or drug-specific evidence.
CrossRef: A supplementary positive DOI-link layer used to identify publication DOI links connected to trial records where the current matching process finds them. A missing CrossRef link does not prove absence of publication or evidence.
RecurSignal v1: Deterministic historical recurrence signal calculated across all 11,525 records using MDIP category, trial phase, and classification confidence. Capped at 5–95. Not a prediction, probability, forecast, risk score, or clinical/investment recommendation.
PubMed linked signal: An OFA-resolved link between a trial record and a PubMed entry (DIRECT_NCT or TITLE_SIMILARITY). A linked signal is a navigation aid, not a literature review.
OpenFDA linked signal: An OFA-resolved link between a trial record and an OpenFDA entry (EXACT or FUZZY drug-name match). A linked signal is a navigation aid, not a regulatory or safety determination.
CrossRef DOI link: A supplementary DOI-to-NCT link surfaced from CrossRef where available. Absence of a CrossRef link does not mean no publication exists.
No linked signal currently available: OFA could not resolve a link to the source for that trial via the current matching process. It is not evidence that the underlying publication, safety record, or regulatory action does not exist.
Evidence Pack: The structured snapshot of OFA-filtered analytics sent to the OFA Research Brief Generator: active filters, record counts, top MDIP™ patterns, top harmonized mechanisms, phase distribution, linked-signal availability, representative records, and limitations. Built entirely from OFA data — no external sources are added.
OFA Research Brief Generator: The Claude-assisted workflow that drafts a human-reviewable core research brief from an OFA evidence pack. Server-side only via the claude-research-hub Supabase Edge Function. Output is a research draft, not a scientific, clinical, regulatory, investment, medical, or treatment conclusion.
Core Research Brief: The concise Claude-assisted output generated by the OFA Research Brief Generator from an OFA evidence pack. It includes pattern summary, knowledge gaps, hypotheses with ofa_evidence_basis, research questions, validation plan, and safety notes. It is a human-reviewable research draft, not a scientific, clinical, regulatory, investment, medical, or treatment conclusion.
Claude-assisted research draft: A core research brief drafted by Claude from an OFA evidence pack. Includes pattern summary, knowledge gaps, hypotheses (each with an ofa_evidence_basis), research questions, validation plan, and safety notes. Requires expert human review before any downstream use.
ofa_evidence_basis: Required field on every Claude-drafted hypothesis. Cites the specific evidence-pack facts (e.g. record counts, NCT IDs, top patterns) the hypothesis is grounded in.
Human review required: Every Claude-assisted brief is an evidence-bound draft. It must be reviewed by a qualified expert before being used in research, regulatory, clinical, or investment contexts.

Oncology Failure Atlas · MDIP™ and RecurSignal™ are proprietary. No clinical or investment advice.