""" Proof: "Daily low-dose aspirin reduces the risk of recurrent non-fatal myocardial infarction in patients with prior cardiovascular disease, per the Antithrombotic Trialists' Collaboration meta-analysis." Claim type: compound / causal. The claim uses causal language ("reduces the risk"), so per the Proof Engine causal rule it is decomposed into an association sub-claim (SC1) and a causation sub-claim (SC2). Generated: 2026-05-20 """ import os import sys PROOF_ENGINE_ROOT = os.environ.get("PROOF_ENGINE_ROOT") if not PROOF_ENGINE_ROOT: _d = os.path.dirname(os.path.abspath(__file__)) while _d != os.path.dirname(_d): if os.path.isdir(os.path.join(_d, "proof-engine", "skills", "proof-engine", "scripts")): PROOF_ENGINE_ROOT = os.path.join(_d, "proof-engine", "skills", "proof-engine") break _d = os.path.dirname(_d) if not PROOF_ENGINE_ROOT: raise RuntimeError("PROOF_ENGINE_ROOT not set and skill dir not found via walk-up from proof.py") sys.path.insert(0, PROOF_ENGINE_ROOT) from scripts.verify_citations import verify_all_citations from scripts.computations import compare, apply_verdict_qualifier from scripts.proof_summary import ProofSummaryBuilder # --------------------------------------------------------------------------- # 1. CLAIM INTERPRETATION (Rule 4) # --------------------------------------------------------------------------- CLAIM_NATURAL = ( "Daily low-dose aspirin reduces the risk of recurrent non-fatal myocardial " "infarction in patients with prior cardiovascular disease, per the " "Antithrombotic Trialists' Collaboration meta-analysis." ) CLAIM_FORMAL = { "subject": ( "Daily low-dose aspirin and recurrent non-fatal myocardial infarction in " "patients with prior cardiovascular disease, as assessed by the " "Antithrombotic Trialists' (ATT) Collaboration meta-analyses" ), "is_time_sensitive": False, "sub_claims": [ { "id": "SC1", "property": ( "The ATT Collaboration meta-analysis found daily low-dose aspirin " "associated with a reduction in non-fatal myocardial infarction " "among patients with prior cardiovascular disease" ), "operator": ">=", "threshold": 2, "operator_note": ( "SC1 (association) holds when at least 2 verified quotations from " "the ATT meta-analysis publications report that low-dose aspirin / " "antiplatelet therapy is associated with a reduction in non-fatal " "myocardial infarction (or the coronary-event composite that " "includes non-fatal MI) among patients with prior cardiovascular " "disease." ), }, { "id": "SC2", "property": ( "The ATT Collaboration meta-analysis pools randomized controlled " "trials, giving the observed reduction a causal (not merely " "associational) basis" ), "operator": ">=", "threshold": 2, "operator_note": ( "SC2 (causation) holds when at least 2 verified quotations " "establish that the ATT meta-analysis pools randomized controlled " "trials. Random allocation removes confounding, so a reduction " "observed consistently across pooled RCTs supports a causal " "interpretation -- this meets the Proof Engine's gold-standard " "criterion for SC-causation (RCTs / controlled experiments)." ), }, ], "compound_operator": "AND", "operator_note": ( "The claim uses causal language ('reduces the risk'), so it is decomposed " "into an association sub-claim (SC1) and a causation sub-claim (SC2); both " "must hold for a PROVED verdict (compound_operator AND). The claim is " "explicitly attributed to 'the Antithrombotic Trialists' Collaboration " "meta-analysis', so the proof evaluates it against the ATT Collaboration's " "own publications: the 2002 BMJ collaborative overview of antiplatelet " "therapy in high-risk patients and the 2009 Lancet individual-participant-" "data meta-analysis of aspirin in primary and secondary prevention. " "'Patients with prior cardiovascular disease' is interpreted as the " "secondary-prevention / high-risk population the ATT analyses studied " "(acute or previous occlusive vascular disease). 'Recurrent non-fatal " "myocardial infarction' is interpreted as non-fatal MI events occurring " "within that established-disease population; the ATT analyses include an " "explicit 'previous myocardial infarction' patient category in which any " "subsequent non-fatal MI is unambiguously recurrent. 'Low-dose aspirin' is " "interpreted as the 75-150 mg/day range the ATT 2002 analysis found 'at " "least as effective as higher daily doses'. Source-count thresholds are " "set to 2 (not the default 3): by construction the claim restricts the " "evidence base to one named source program (the ATT Collaboration " "meta-analysis), which produced exactly two landmark meta-analyses on this " "question -- both peer-reviewed in top-tier journals (BMJ, Lancet), each " "pooling large randomized-trial datasets, and funded by public/charitable " "bodies (UK Medical Research Council, British Heart Foundation, Cancer " "Research UK, EC Biomed) with no aspirin-manufacturer funding. No conflict " "of interest favoring the claim was identified." ), } # --------------------------------------------------------------------------- # 2. FACT REGISTRY (report IDs -> proof-script keys) # --------------------------------------------------------------------------- FACT_REGISTRY = { "B1": {"key": "sc1_att2002_mi", "label": "SC1 - ATT 2002 BMJ: non-fatal MI reduced by one third in high-risk patients"}, "B2": {"key": "sc1_att2002_dose", "label": "SC1 - ATT 2002 BMJ: low-dose aspirin (75-150 mg) at least as effective as higher doses"}, "B3": {"key": "sc1_att2009", "label": "SC1 - ATT 2009 Lancet: aspirin reduces serious vascular events in secondary-prevention trials"}, "B4": {"key": "sc2_att2002", "label": "SC2 - ATT 2002 BMJ: inclusion restricted to randomised trials"}, "B5": {"key": "sc2_att2009", "label": "SC2 - ATT 2009 Lancet: meta-analysis of individual participant data from randomised trials"}, "A1": {"label": "SC1 verified source count", "method": None, "result": None}, "A2": {"label": "SC2 verified source count", "method": None, "result": None}, } # --------------------------------------------------------------------------- # 3. EMPIRICAL FACTS (grouped by sub-claim via the sc1_ / sc2_ key prefix) # --------------------------------------------------------------------------- _PMC_2002 = "https://pmc.ncbi.nlm.nih.gov/articles/PMC64503/" _PUBMED_2009 = "https://pubmed.ncbi.nlm.nih.gov/19482214/" _SRC_2002 = ("Antithrombotic Trialists' Collaboration. Collaborative meta-analysis of " "randomised trials of antiplatelet therapy for prevention of death, " "myocardial infarction, and stroke in high risk patients. BMJ " "2002;324:71-86 (full text via PubMed Central, PMC64503).") _SRC_2009 = ("Antithrombotic Trialists' (ATT) Collaboration. Aspirin in the primary and " "secondary prevention of vascular disease: collaborative meta-analysis of " "individual participant data from randomised trials. Lancet " "2009;373:1849-60 (abstract via PubMed, PMID 19482214).") # Snapshot fallback (Rule 2). PubMed Central and PubMed periodically block or # throttle automated fetches, so each citation carries an offline snapshot of # the page captured on 2026-05-20. The verifier uses the fallback chain # live -> snapshot -> Wayback, so a transient live-fetch failure cannot change # the verdict. The snapshots are public-domain government pages, stored as # separate files to keep this script readable. _PROOF_DIR = os.path.dirname(os.path.abspath(__file__)) def _load_snapshot(fname): fpath = os.path.join(_PROOF_DIR, fname) try: with open(fpath, encoding="utf-8") as f: return f.read() except FileNotFoundError: return None _SNAPSHOT_2002 = _load_snapshot("snapshots/pmc64503.html") _SNAPSHOT_2009 = _load_snapshot("snapshots/pubmed19482214.html") _SNAPSHOT_DATE = "2026-05-20" empirical_facts = { # ---- SC1: association (aspirin reduces non-fatal MI in prior-CVD patients) ---- "sc1_att2002_mi": { "quote": ( "Overall, among these high risk patients, allocation to antiplatelet " "therapy reduced the combined outcome of any serious vascular event by " "about one quarter; non-fatal myocardial infarction was reduced by one " "third, non-fatal stroke by one quarter, and vascular mortality by one " "sixth" ), "url": _PMC_2002, "source_name": _SRC_2002, "snapshot": _SNAPSHOT_2002, "snapshot_source": "public:pre_fetched", "snapshot_fetched_at": _SNAPSHOT_DATE, }, "sc1_att2002_dose": { "quote": ( "Aspirin was the most widely studied antiplatelet drug, with doses of " "75-150 mg daily at least as effective as higher daily doses." ), "url": _PMC_2002, "source_name": _SRC_2002, "snapshot": _SNAPSHOT_2002, "snapshot_source": "public:pre_fetched", "snapshot_fetched_at": _SNAPSHOT_DATE, }, "sc1_att2009": { "quote": ( "In the secondary prevention trials, aspirin allocation yielded a " "greater absolute reduction in serious vascular events" ), "url": _PUBMED_2009, "source_name": _SRC_2009, "snapshot": _SNAPSHOT_2009, "snapshot_source": "public:pre_fetched", "snapshot_fetched_at": _SNAPSHOT_DATE, }, # ---- SC2: causation (the meta-analysis pools randomised controlled trials) ---- "sc2_att2002": { "quote": ( "Randomised trials of an antiplatelet regimen versus control or of one " "antiplatelet regimen versus another in high risk patients" ), "url": _PMC_2002, "source_name": _SRC_2002, "snapshot": _SNAPSHOT_2002, "snapshot_source": "public:pre_fetched", "snapshot_fetched_at": _SNAPSHOT_DATE, }, "sc2_att2009": { "quote": ( "collaborative meta-analysis of individual participant data from " "randomised trials" ), "url": _PUBMED_2009, "source_name": _SRC_2009, "snapshot": _SNAPSHOT_2009, "snapshot_source": "public:pre_fetched", "snapshot_fetched_at": _SNAPSHOT_DATE, }, } # --------------------------------------------------------------------------- # 4. CITATION VERIFICATION (Rule 2) # --------------------------------------------------------------------------- citation_results = verify_all_citations(empirical_facts, wayback_fallback=True) # --------------------------------------------------------------------------- # 5. COUNT VERIFIED SOURCES PER SUB-CLAIM # --------------------------------------------------------------------------- COUNTABLE_STATUSES = ("verified", "partial") sc1_keys = [k for k in empirical_facts if k.startswith("sc1_")] sc2_keys = [k for k in empirical_facts if k.startswith("sc2_")] n_sc1 = sum(1 for k in sc1_keys if citation_results[k]["status"] in COUNTABLE_STATUSES) n_sc2 = sum(1 for k in sc2_keys if citation_results[k]["status"] in COUNTABLE_STATUSES) # --------------------------------------------------------------------------- # 6. PER-SUB-CLAIM EVALUATION (Rule 7 - compare(), no eval()) # --------------------------------------------------------------------------- sc1_holds = compare(n_sc1, ">=", CLAIM_FORMAL["sub_claims"][0]["threshold"], label="SC1 (association): verified ATT sources") sc2_holds = compare(n_sc2, ">=", CLAIM_FORMAL["sub_claims"][1]["threshold"], label="SC2 (causation): verified ATT sources") # --------------------------------------------------------------------------- # 7. COMPOUND EVALUATION # --------------------------------------------------------------------------- n_holding = sum([sc1_holds, sc2_holds]) n_total = len(CLAIM_FORMAL["sub_claims"]) claim_holds = compare(n_holding, "==", n_total, label="compound: all sub-claims hold") # --------------------------------------------------------------------------- # 8. COI FLAGS (Rule 6) - per sub-claim # --------------------------------------------------------------------------- # The ATT Collaboration is an independent academic consortium funded by the UK # Medical Research Council, British Heart Foundation, Cancer Research UK and the # EC Biomed Programme -- no aspirin-manufacturer funding. Its 2009 Lancet paper # concluded that aspirin is "of uncertain net value" in primary prevention, # demonstrating willingness to report null/negative findings. No conflict of # interest favoring the claim was identified for either sub-claim. sc1_coi_flags = [] sc2_coi_flags = [] # --------------------------------------------------------------------------- # 9. ADVERSARIAL CHECKS (Rule 5) - counter-evidence searched during research # --------------------------------------------------------------------------- adversarial_checks = [ { "question": ("Does the ATT meta-analysis actually report a reduction in " "non-fatal myocardial infarction, or is the attribution wrong?"), "verification_performed": ( "Fetched and read the ATT 2002 BMJ meta-analysis (PMC64503) and the " "ATT 2009 Lancet meta-analysis abstract (PubMed 19482214); searched for " "the verbatim findings on non-fatal MI and secondary prevention." ), "finding": ( "The attribution is accurate. The 2002 BMJ ATT meta-analysis explicitly " "states that, among high-risk patients, 'non-fatal myocardial infarction " "was reduced by one third'. The 2009 Lancet ATT meta-analysis reports " "that in secondary-prevention trials aspirin reduced serious vascular " "events and coronary events. No mismatch between the claim and the named " "source was found." ), "breaks_proof": False, }, { "question": ("Does the bleeding harm of aspirin negate or contradict the " "reduction in non-fatal MI?"), "verification_performed": ( "Searched for counter-evidence on aspirin's bleeding harms and net " "benefit in secondary prevention (gastrointestinal / extracranial " "bleeding meta-analyses)." ), "finding": ( "Aspirin does increase major extracranial and gastrointestinal bleeding " "-- a well-documented, separate harm. This does not contradict the " "claim, which concerns the non-fatal-MI outcome specifically, not net " "clinical benefit. The ATT 2002 paper itself reports that in the " "high-risk categories 'the absolute benefits substantially outweighed " "the absolute risks of major extracranial bleeding'. Bleeding is a " "treatment trade-off, not evidence against the MI-reduction finding, so " "it does not break the proof." ), "breaks_proof": False, }, { "question": ("Is the secondary-prevention benefit of aspirin contradicted by " "more recent evidence from the modern statin / revascularization " "era?"), "verification_performed": ( "Searched 2020-2025 reviews and meta-analyses that question aspirin in " "secondary prevention, including recent commentary and a 2025 systematic " "review/meta-analysis of aspirin for secondary prevention of MI." ), "finding": ( "Recent commentary debates whether aspirin's marginal benefit is as " "large in the modern treatment era and notes P2Y12 inhibitors as " "alternatives. However, a 2025 systematic review/meta-analysis still " "found aspirin reduced recurrent events by roughly a fifth in secondary " "prevention, and no source claims the ATT finding of reduced non-fatal " "MI was wrong or reversed. The direction of effect is confirmed by " "modern evidence; only the magnitude and its role relative to newer " "therapies are debated. This does not break a proof scoped to the ATT " "meta-analysis finding." ), "breaks_proof": False, }, { "question": ("Does 'low-dose' aspirin specifically -- not antiplatelet " "therapy in general -- carry the non-fatal MI reduction?"), "verification_performed": ( "Checked the ATT 2002 BMJ dose analysis and the scope of the 2009 " "Lancet meta-analysis." ), "finding": ( "Supported. The ATT 2002 meta-analysis found 'doses of 75-150 mg daily " "at least as effective as higher daily doses' and identified aspirin as " "'the most widely studied antiplatelet drug'; the 2009 ATT meta-analysis " "is explicitly an analysis of 'low-dose aspirin'. The 2002 paper notes " "the effects of doses below 75 mg are 'less certain', but standard " "low-dose aspirin (75-150 mg) is squarely within the supported range, " "so this does not break the proof." ), "breaks_proof": False, }, { "question": ("Could 'recurrent' non-fatal MI be unsupported because the ATT " "meta-analysis pooled mixed prior-CVD populations rather than " "only prior-MI patients?"), "verification_performed": ( "Reviewed the ATT meta-analysis patient categories in the 2002 BMJ and " "2009 Lancet reports." ), "finding": ( "The ATT meta-analysis explicitly includes a 'previous myocardial " "infarction' patient category among its high-risk groups (six " "secondary-prevention prior-MI trials in the 2009 analysis), in which a " "subsequent non-fatal MI is by definition recurrent. The proof " "interprets 'recurrent non-fatal MI in patients with prior CVD' as " "non-fatal MI events within the established-CVD secondary-prevention " "population the ATT analysed -- an interpretation documented in the " "operator_note that does not overstate the source -- so this does not " "break the proof." ), "breaks_proof": False, }, ] # --------------------------------------------------------------------------- # 10. VERDICT # --------------------------------------------------------------------------- if __name__ == "__main__": any_unverified = any( cr["status"] != "verified" for cr in citation_results.values() ) any_breaks = any(ac.get("breaks_proof") for ac in adversarial_checks) is_disproof = CLAIM_FORMAL.get("proof_direction") == "disprove" # Per-sub-claim COI gate (Rule 6) sc1_confirmed_keys = {k for k in sc1_keys if citation_results[k]["status"] in COUNTABLE_STATUSES} sc1_coi_favorable = {f["source_key"] for f in sc1_coi_flags if f["direction"] == "favorable_to_subject" and f["source_key"] in sc1_confirmed_keys} sc1_coi_unfavorable = {f["source_key"] for f in sc1_coi_flags if f["direction"] == "unfavorable_to_subject" and f["source_key"] in sc1_confirmed_keys} sc1_coi_majority = max(len(sc1_coi_favorable), len(sc1_coi_unfavorable)) if sc1_coi_flags else 0 sc1_threshold = CLAIM_FORMAL["sub_claims"][0]["threshold"] sc1_coi_override = n_sc1 >= sc1_threshold and sc1_coi_majority > n_sc1 / 2 sc2_confirmed_keys = {k for k in sc2_keys if citation_results[k]["status"] in COUNTABLE_STATUSES} sc2_coi_favorable = {f["source_key"] for f in sc2_coi_flags if f["direction"] == "favorable_to_subject" and f["source_key"] in sc2_confirmed_keys} sc2_coi_unfavorable = {f["source_key"] for f in sc2_coi_flags if f["direction"] == "unfavorable_to_subject" and f["source_key"] in sc2_confirmed_keys} sc2_coi_majority = max(len(sc2_coi_favorable), len(sc2_coi_unfavorable)) if sc2_coi_flags else 0 sc2_threshold = CLAIM_FORMAL["sub_claims"][1]["threshold"] sc2_coi_override = n_sc2 >= sc2_threshold and sc2_coi_majority > n_sc2 / 2 any_coi_override = sc1_coi_override or sc2_coi_override # This is a causal-decomposition compound claim, not a contested-qualifier # claim, so the contested-qualifier branch must not fire. is_contested_qualifier = "qualifier" in CLAIM_FORMAL.get("operator_note", "").lower() if any_breaks: base_verdict = "UNDETERMINED" elif any_coi_override: base_verdict = "UNDETERMINED" elif is_contested_qualifier and sc1_holds and not sc2_holds: base_verdict = "DISPROVED" elif not claim_holds and n_holding > 0: base_verdict = "PARTIALLY VERIFIED" elif claim_holds: base_verdict = "DISPROVED" if is_disproof else "PROVED" elif not claim_holds and n_holding == 0: base_verdict = "UNDETERMINED" else: base_verdict = "UNDETERMINED" # defensive fallback verdict = apply_verdict_qualifier(base_verdict, any_unverified) builder = ProofSummaryBuilder(CLAIM_NATURAL, CLAIM_FORMAL) for fid, info in FACT_REGISTRY.items(): if not fid.startswith("B"): continue ef_key = info["key"] ef = empirical_facts[ef_key] cr = citation_results.get(ef_key, {}) sub_claim = "SC1" if ef_key in sc1_keys else "SC2" builder.add_empirical_fact( fid, label=info["label"], source_name=ef["source_name"], source_url=ef["url"], source_quote=ef["quote"], sub_claim=sub_claim, ) builder.set_verification( fid, status=cr.get("status", "unknown"), method=cr.get("method", "full_quote"), coverage_pct=cr.get("coverage_pct"), fetch_mode=cr.get("fetch_mode", "live"), credibility=cr.get("credibility", {}), ) builder.set_extraction( fid, value=cr.get("status", "unknown"), value_in_quote=cr.get("status") in COUNTABLE_STATUSES, quote_snippet=ef["quote"][:80], ) sc1_fact_ids = [fid for fid, info in FACT_REGISTRY.items() if fid.startswith("B") and info["key"] in sc1_keys] sc2_fact_ids = [fid for fid, info in FACT_REGISTRY.items() if fid.startswith("B") and info["key"] in sc2_keys] builder.add_computed_fact( "A1", label="SC1 verified source count", method=f"count(verified sc1 citations) = {n_sc1}", result=n_sc1, depends_on=sc1_fact_ids, sub_claim="SC1", ) builder.add_computed_fact( "A2", label="SC2 verified source count", method=f"count(verified sc2 citations) = {n_sc2}", result=n_sc2, depends_on=sc2_fact_ids, sub_claim="SC2", ) builder.add_cross_check( description="SC1: independent ATT meta-analysis publications consulted", fact_ids=sc1_fact_ids, n_sources_consulted=len(sc1_keys), n_sources_verified=n_sc1, sources={k: citation_results[k]["status"] for k in sc1_keys}, independence_note=( "SC1 draws on two distinct ATT meta-analyses published seven years " "apart in different journals (BMJ 2002, Lancet 2009) using different " "methods (tabular data vs individual participant data) and overlapping " "but non-identical trial sets. B1 and B2 are two distinct findings " "(the non-fatal-MI reduction and the low-dose equivalence) from the " "2002 BMJ paper; B3 is from the independent 2009 Lancet meta-analysis." ), coi_flags=sc1_coi_flags, agreement=sc1_holds, ) builder.add_cross_check( description="SC2: independent ATT meta-analysis publications consulted", fact_ids=sc2_fact_ids, n_sources_consulted=len(sc2_keys), n_sources_verified=n_sc2, sources={k: citation_results[k]["status"] for k in sc2_keys}, independence_note=( "SC2 is confirmed by both ATT publications independently: the 2002 BMJ " "paper restricts inclusion to 'Randomised trials', and the 2009 Lancet " "paper is a 'collaborative meta-analysis of individual participant data " "from randomised trials'. Two separate publications each document the " "randomized-controlled-trial basis." ), coi_flags=sc2_coi_flags, agreement=sc2_holds, ) builder.add_sub_claim_result( id="SC1", n_confirming=n_sc1, threshold=CLAIM_FORMAL["sub_claims"][0]["threshold"], holds=sc1_holds, ) builder.add_sub_claim_result( id="SC2", n_confirming=n_sc2, threshold=CLAIM_FORMAL["sub_claims"][1]["threshold"], holds=sc2_holds, ) for ac in adversarial_checks: builder.add_adversarial_check( question=ac["question"], verification_performed=ac["verification_performed"], finding=ac["finding"], breaks_proof=ac["breaks_proof"], ) builder.set_verdict(base_verdict, any_unverified=any_unverified) builder.set_key_results( n_holding=n_holding, n_total=n_total, claim_holds=claim_holds, sc1_verified_sources=n_sc1, sc2_verified_sources=n_sc2, ) print("=" * 70) print("PROOF: Daily low-dose aspirin and recurrent non-fatal MI (ATT)") print("=" * 70) print(f"SC1 (association) verified ATT sources: {n_sc1} " f"(threshold {CLAIM_FORMAL['sub_claims'][0]['threshold']}) -> holds={sc1_holds}") print(f"SC2 (causation) verified ATT sources: {n_sc2} " f"(threshold {CLAIM_FORMAL['sub_claims'][1]['threshold']}) -> holds={sc2_holds}") print(f"Sub-claims holding: {n_holding}/{n_total}") print(f"Adversarial checks that break the proof: " f"{sum(1 for ac in adversarial_checks if ac['breaks_proof'])}") print(f"\nVERDICT: {verdict}") print("=" * 70) builder.emit()