Skip to main content
TRUSTWORTHY (audit-first build) Trials: 2 · all passed 6-gate audit · k ≥2
Built post-cleanup using the audit-first pipeline. Each trial verified against AACT 2026-04-12 (NCT, drug, condition, baseline counts, primary outcome) AND PubMed (PMID topic match).
Methodology: FINAL_INTEGRITY_REPORT_V2.md · audit log · main index

Pacritinib in MF + cytopenia

RapidMeta · audit-first build · 2 included trials · Pacritinib in Myelofibrosis · Data: AACT 2026-04-12 + PubMed E-utilities

PICO

PopulationAdults randomised in trials registered on ClinicalTrials.gov for Myelofibrosis.
InterventionPacritinib (AACT-verified intervention name in each trial).
ComparatorActive comparator or placebo as registered on AACT.
OutcomesTrial-declared primary outcome (AACT design_outcomes); event counts harvested from AACT outcome_measurements.
DesignInterventional RCTs (post-2010 start date), Phase 2/3 or Phase 3/4.

Eligibility (6-gate audit)

  1. GATE-A — NCT exists in AACT 2026-04-12 snapshot.
  2. GATE-B — Drug pattern "pacritinib" present in AACT interventions for the NCT.
  3. GATE-C — Condition pattern "myelofibrosis" present in AACT conditions.
  4. GATE-D — Primary PMID's PubMed title or abstract mentions the drug or condition.
  5. GATE-E — AACT baseline_counts reports ≥2 per-arm participant rows.
  6. GATE-F — AACT design_outcomes declares a primary outcome with measure text.

Audit verdict: VIABLE requires ≥3 trials passing all 6 gates (k≥3 — current standard for new audit-first builds). Older builds use k≥2.

Pre-specified primary outcomes (per trial)

Statistical methods

Inverse-variance random-effects pooling on the log-OR scale (Woolf estimator from AACT event counts), DerSimonian-Laird τ² with Hartung-Knapp-Sidik-Jonkman (HKSJ) variance correction and tk-1 critical value (Cochrane Handbook v6.5 conventions). Single-trial entries report Wald 95% CI on the log scale.

PRISMA-style screening

Records identified (AACT × PubMed)8
Records excluded (gate failure)6
Records included (all 6 gates pass)2

Exclusions by first failing gate

Gaten
D_pmid_topic_match6

Screening is fully deterministic: every record is auto-flagged by the gate that first fails. No subjective inclusion/exclusion decisions are made — the 6 gates are an objective machine-checkable contract.

NCT04884191 NCT04884191 · PMID 34551507 · 2022

Phase 2 Study: An Open-Label, Randomized, Phase 2 Dose-Finding Study of Pacritinib in Patients With Primary Myelofibrosis, Post-Polycythemia Vera Myelofibrosis,

Primary outcome (AACT): Spleen Volume Reduction Response (≥ 35%)

EventsNo eventTotal N
Intervention05252
Control25355
PAC326 NCT02055781 · PMID 29522138 · 2018

Pacritinib Versus Best Available Therapy to Treat Patients With Myelofibrosis and Thrombocytopenia

Primary outcome (AACT): Spleen Volume Reduction

EventsNo eventTotal N
Intervention1193104
Control2105107

Forest plot — per-trial OR + pooled estimate

Trial log-OR (95% CI) NCT04884191 (no event data) PAC326 6.21 [1.34, 28.74] Pooled (k=1) OR 6.21 [1.342, 28.743]

Pooled estimate (REML-DL + HKSJ, Cochrane v6.5)

OR 6.21 [1.342, 28.743]
k = 1 · method = single-trial · τ² = — · Q = — · I² = —%

1. Heterogeneity

Requires k≥2 trials with event data.

2. Prediction interval

Requires k≥3 trials (Cochrane v6.5).

3. Egger's test

Requires k≥3.

4. Leave-one-out sensitivity

Requires k≥3.

5. PRISMA 2020 status

IdentificationAACT × PubMed intersection (deterministic)
Screening6-gate audit (machine-checkable)
Eligibilityk=2 trials passed all 6 gates
Includedk=1 with extractable event counts

6. Risk of Bias (audit-first scope)

Audit-first builds do not apply RoB 2.0 domain coding — that requires full-text trial reading. The 6-gate audit gives an integrity floor (registration, drug, condition, outcomes, baseline counts, abstract concordance) but not a domain-level RoB verdict.

For RoB 2.0 coding, see the matching curated *_REVIEW.html for this topic if one exists in the flagship portfolio.

7. GRADE certainty (auto-derived)

Certainty: Very low

Auto-derivation considers k and I² only; full GRADE requires RoB, indirectness, publication bias judgment, and effect-magnitude scoring (not applied in audit-first builds). Triggers: k=1 (imprecision).

8. Reproducibility fingerprint

sha256[:16] = c1cf89b6ada3f3ad

Bit-stable hash of {sorted NCT list, pooled OR + 95% CI}. Reproduce by re-running scripts/add_topic_autodiscover.py + generate_topic_html.py against the same AACT snapshot.

Audit-first build

This review was generated by the 6-gate audit-first pipeline after the 16-round portfolio cleanup. Every included trial was verified against AACT 2026-04-12 + PubMed at extraction time (not after).

Limits of this build