Build a robust evidence foundation for your most critical decisions.
Adcurare reads every document behind an asset as one evidence system, recomputes the key numbers, and returns a brief your experts can verify — every finding linked to its source.
Scientific evidence intelligence for pharma. One platform, two teams: business development and R&D.
Illustrative. See real runs
Recognition
Built by a research team recognized for exactly this work: rigorous, reproducible science.
Alzheimer’s Insights AI Prize
Alzheimer’s Disease Data Initiative · $1M global agentic-AI competition
NIH Replication Prize
NIH Common Fund · For advancing reproducible science
One real-world case.
A withdrawn asset. We ran automated quality control on its pivotal early paper, blind to the later trial result: it flags where the evidence package does not reconcile and routes the survival claim to a gating analysis. Every finding links to its source.
Olaratumab + doxorubicin
Eli Lilly · NCT01185964 · advanced soft-tissue sarcomaObjective response rate of 18.2% reported with 95% CI 29.6–29.8 — an interval that excludes its own point estimate. Statistically impossible.
Retrospective, outcome-blinded review using only evidence available at the 2016 decision point: the headline statistics do not reconcile, so the survival estimate needs independent reconstruction before it can be underwritten.
The Lancet · accelerated approval
Failed to confirm the survival benefit; accelerated approval was withdrawn.
Cross the false claims off your list.
Every card is a real Rigor Engine run on a positive-looking early paper.
Surrogate endpoint, no benefit link; two phantom citations
A 95% CI that excludes its own point estimate
Efficacy rests on a bone-density surrogate; placebo arm dissolved mid-study
Same target, same disease; no critical findings
One reconciled count discrepancy, nothing disqualifying
Clean design, complete reporting
3 flagged weak here later failed in Phase 3 or lost approval.
Two numbers that don't hold up.
In this labeled synthetic study: the ITT denominator doesn't reconcile across documents, and the primary-endpoint P value doesn't recompute — so the headline result isn't significant as reported.
Intention-to-treat denominator is inconsistent across documents
Source claim
The intention-to-treat population comprised 812 randomized participants.
“…the intention-to-treat population comprised 812 randomized participants (406 per arm)…”
Results, paragraph 1; Abstract
Detected inconsistency
Abstract and Results state 812 (406/arm); the CONSORT diagram and Table 1 both sum to 806. The 6-participant gap is never reconciled.
Abstract/Results: 406 + 406 = 812 CONSORT: 404 + 402 = 806 Δ = 6 participants (unreconciled)
Recommended follow-up
Which count is correct for the ITT population, and where are the 6 participants accounted for?
Reported primary-endpoint P value does not reconstruct from the stated statistic
Source claim
The primary endpoint was met (P = 0.048).
“…a statistically significant difference was observed for the primary endpoint (χ²(1) = 3.29, P = 0.048).”
Results, Table 2
Detected inconsistency
χ²(1) = 3.29 gives a two-sided P ≈ 0.070, not 0.048. As reported, the primary endpoint does not clear P < 0.05 — it is not significant.
P(χ²(1) > 3.29) = 0.070 (two-sided) Reported: P = 0.048 P = 0.048 would require χ²(1) ≈ 3.91
Recommended follow-up
Provide the exact 2×2 table and test used for the primary endpoint.
Every Rigor Record contains
- 01Executive summary
- 02Principal claims
- 03Evidence sources examined
- 04Findings organized by materiality
- 05Source passage and calculation
- 06Confidence and verification status
- 07Potential decision impact
- 08Recommended follow-up questions
- 09Methods and limitations
- 10Versioned audit trail
Concise. Robust. Elegant.
A rigor review should surface the finding that could change the decision, give you the evidence to defend it, and spare you everything else.
Numerical and statistical consistency
Recompute reported statistics and check denominators, confidence intervals, P values, and tables against the underlying data.
Protocol-to-publication concordance
Compare protocols, SAPs, registries, and publications to surface deviations from what was prespecified.
Claim-to-evidence verification
Test whether each consequential claim is actually supported by the evidence presented.
Reporting and ethics completeness
Flag reporting omissions, missing disclosures, and ethics or registration gaps, without treating omission as misconduct.
Reproducibility and decision-risk assessment
Assess whether the analysis can be reproduced and where the decision risk concentrates.
The depth behind a single Rigor Record.
Reported statistics are recomputed from the source, not taken on trust. One deliverable a domain expert can check line by line, in place of reconciling the documents by hand.
- 8
- dimensions of rigor scored, from scientific premise to reporting transparency
- 3×
- independent reviewer passes, reconciled by a synthesis pass
- 5
- evidence sources read as one system: publication, protocol, SAP, registry, CONSORT
- 100%
- of findings linked to their source passage, with a confidence and a materiality
Where Adcurare earns its keep.
For business development
Know what the evidence is worth before you license the asset.
- In-licensing and acquisition diligence
- Clinical-trial evidence review
- Competitive and landscape evidence assessment
- Data-room and claim verification
- Portfolio and indication prioritization
- Decision-ready evidence reports
For R&D
Know if the science is solid enough to build on.
- Foundational-evidence validation before you commit a program
- Protocol, SAP, and registry-to-publication concordance
- Reproducibility and statistical review
- Internal prepublication and data-package QC
- Evaluation of AI-generated scientific work
- Systematic comparison of competing claims
A reproducibility research team, and your evidence handled with care.
The team
Built by researchers recognized for exactly this work.
Adcurare is built by a research team working on reproducibility and scientific rigor. The standard we hold your evidence to is the one our own work is measured by.
- Alzheimer’s Insights AI Prize — Alzheimer’s Disease Data Initiative. $1M global agentic-AI competition.
- NIH Replication Prize — NIH Common Fund. For advancing reproducible science.
Security & data handling
How we handle your evidence.
- Encrypted in transit and at rest.
- An NDA and DPA are available before any documents are shared.
- Your evidence is never used to train models.
- You control retention, deletion, and export.
- Subprocessors named on request.
No SOC 2 or HIPAA certification claimed. A security overview is shared under NDA.
Every finding is a candidate for human review, not a final verdict. See our responsible-verification commitments.
A private, defined pilot on one real decision.
- Scope
- One real decision — an in-licensing asset, a program, or a manuscript — agreed up front.
- Turnaround
- About two weeks from data access to a readout with your team.
- Success metrics
- Set with you, so results are judged against a bar you set.
Get started
Bring decision-grade verification into your workflow.
For business development
See Adcurare on a live in-licensing or diligence decision: the pivotal trials behind an asset, a data room, or a competing evidence base. Start with a demo; we scope a private pilot when it is a fit.
For R&D
See Adcurare on the science you are deciding whether to build on: a protocol, an analysis, a data package, or a scientific-agent workflow. Start with a demo; scope a pilot on a real decision.
Prefer to look first? See a sample Rigor Record.