Scrutinise the evidence for any medicine.

AI evidence synthesis, built for medicine. For those who need to know when the evidence moves.

Upadacitinib (Rinvoq) · Atopic Dermatitis · Weekly claim register · Cycle W19 2026 Updated this cycle

Claim	Verdict	Movement	Score
Efficacy
UPA 30mg achieves superior EASI-75 vs placebo in moderate-to-severe AD	Definitive	— Stable	88%
IGA 0/1 clear or almost-clear skin achieved at week 16	Definitive	— Stable	89%
Rapid itch relief within 2 weeks is superior to dupilumab	Likely	↑ Strengthening	82%
Sustained EASI-75 response maintained at week 52	Likely	— Stable	73%
Comparative Efficacy
UPA 30mg achieves deep response superior to dupilumab at week 16	Likely	↑ Strengthening	85%
EASI-75 response rate is non-inferior to tralokinumab at week 16	Likely	— Stable	79%
Head-to-head superiority over lebrikizumab on itch NRS at week 12	Uncertain	— Stable	52%
Safety
Serious infection rate is not elevated vs dupilumab over 52 weeks	Definitive	↑ Strengthening	86%
Acne incidence is higher with UPA 30mg vs dupilumab in adolescents	Likely	— Stable	74%
Class-level Safety
Long-term cardiovascular risk is not elevated vs dupilumab in AD patients	Uncertain	— Stable	48%
Malignancy incidence is not elevated above background in the AD indication	Doubtful	↓ Weakening	31%

How it works

A four-agent pipeline built to scrutinise the evidence.

Four AI agents working in sequence — retrieval, analysis, adversarial challenge, and calibrated verdict. Every score is traceable back to a real DOI.

4 Agent Pipeline

Retriever

Pulls the full evidence corpus across four databases — including papers that contradict the claim. Every retrieved DOI is logged before the next agent runs.

Analyst

Reads each retrieved paper against a structured rubric: study design, sample size, effect size, follow-up duration, and comparator fidelity.

Archie — Sceptic

The adversarial agent. Actively tries to falsify the claim — probing for confounding, publication bias, and methodological limits. Archie's objections must be explicitly overruled to advance the score.

Moderator

Weighs the Analyst's summary against Archie's objections and renders the final calibrated verdict — with written rationale, DOI chain, and evidence-tier breakdown.

Standard AI vs Scrutinise

General-purpose AI is not built to stress-test scientific evidence.

ChatGPT and Claude are built to provide fluent answers. Scrutinise is built to challenge the evidence and be wrong less often where it matters most.

− General-purpose AI

+ Scrutinise

Hallucination — fabricates citations, invents trial names, conflates study designs.

Every verdict traces to a real DOI. Papers contradicting the claim are retrieved and scored alongside supporting ones.

No structured critique — no adversarial step to challenge a claim or stress-test the conclusion.

Archie the Sceptic is built into every score. The adversarial agent actively tries to falsify each claim before the verdict is rendered.

No audit trail — no DOI chain, no retrieval log, no record of what was reviewed. Undefendable in a payer or HTA challenge.

Full corpus log per score: papers retrieved, scored, excluded with reason, and Archie's sceptic rationale — timestamped.

Stale training data — cutoff months or years before you ask the question.

Corpus retrieved live every cycle. The score reflects the literature as it stands this week.

Who it's for

Three tiers of access. Each built for a different depth of need.

Pharma and biotech organisations have different functions requiring different depths of evidence access. Scrutinise is structured around three tiers that match the way decisions are actually made.

Signal digest

Passive surveillance

Time-constrained professionals who need to stay current without actively monitoring the literature. Personalised to their asset and therapeutic area, delivered every Sunday.

Weekly signals for monitored assets and claims
Plain-language explanation of what moved and why
Material mover alerts between cycles when a high-priority paper lands

Evidence terminal

Full platform access

Working scientists and analysts who need the full platform — claim-level scoring detail, confidence trajectories over time, paper-level verdicts, and structured data exports.

Full claim library with five-factor scoring and Archie's sceptic detail
Paper-level evidence review for significant publications
Competitive landscape by therapeutic area and endpoint

Custom intelligence

Bespoke analysis

Strategic analysis commissioned for a specific need — a competitor matrix before a product launch, a KOL brief for a congress, or a full evidence assessment ahead of a major investment decision.

Competitive intelligence matrices across drugs and endpoints
KOL profiling from high-certainty claim citation patterns
Conference abstract scoring and trial readout briefings

What Scrutinise delivers

Scientific intelligence delivered when you need to know.

Weekly · Per medicine

Claim register

Every scientific claim behind your medicine, scored on a five-tier verdict scale and flagged for movement. Organised by cluster — efficacy, safety, comparative, class-level, real-world.

ID	Claim	Verdict	Movement	Score
EFF-04	UPA 30mg achieves deep response superior to dupilumab at week 16	Likely	↑ +7pp	85%
CCT-04	Malignancy incidence is not elevated above background in AD	Doubtful	↓ −5pp	31%
EFF-05	Rapid itch relief within 2 weeks superior to dupilumab	Likely	—	76%

Monthly · Per indication

Competitive intelligence

A hub-and-spoke matrix scoring all active comparators against a reference medicine across the endpoints that matter in clinical and payer conversations. Signal movers flagged each cycle.

Medicine	EASI-75	Deep resp.	Serious AEs	Itch relief
UPA 30mgRinvoq	78% H2H	85% ↑	54%	70%
LebrikizumabEbglyss	22% ↓	28%	48%	45%
NemolizumabNemluvio	44%	38%	52%	72% ↑

On request · Per cluster

Evidence gap report

A structured analysis of every open evidence gap in your medicine's claim landscape — classified by type, prioritised by HTA and commercial relevance, with a recommended study design for each.

High priority · Class contamination

Long-term CV and malignancy — JAK class in AD

No H2H long-term safety data vs dupilumab. Primary vulnerability in PBAC and payer conversations. Study required: long-term OLE reporting CV and malignancy by indication.

Medium priority · Methodological

NMA ranking instability — deep response endpoint

NMA pooled across heterogeneous trials. Score directionally correct but wider CIs than H2H. One further head-to-head trial would convert this to a robust lead.

Pre and post-conference · Per meeting

Conference brief

A rapid scan of every major congress — EADV, ECCO, ACC, ASCO, and others — identifying the abstracts and late-breaking data most likely to shift your evidence landscape before the papers publish.

High impactEADV 2026 · Late-breaking

Nash 2025 integrated safety analysis — 27,000 patient-years across all UPA indications

High impactEADV 2026 · Oral abstract

JADE Compare extension — abrocitinib long-term H2H data vs dupilumab at 104 weeks

WatchEADV 2026 · Poster

Lebrikizumab real-world registry data — itch response in biologic-naive patients

On demand · Per indication

Researcher Profiling

A better way to rank and engage KOLs not around publication count — but their proportional contribution to what the existing evidence establishes. Know exactly when and how to engage them around the evidence.

Upadacitinib · Atopic dermatitis · Quality-weighted evidence ranking · 312 scored papers

#	Researcher	Papers	Share of quality evidence	Contribution
01	Emma Guttman-Yassky Icahn School of Medicine · New York	34 papers	26%	Consensus
02	Jonathan Silverberg George Washington University	28 papers	19%	Consensus
03	Andreas Wollenberg Ludwig Maximilian University · Munich	19 papers	13%	Contrarian
04	Kristian Reich Translational Research in Inflammatory Skin Diseases · Hamburg	16 papers	10%	Consensus
05	Marjolein de Bruin-Weller University Medical Centre Utrecht	11 papers	7%	Consensus

Top 5 researchers account for 75% of quality-weighted evidence. 58 researchers contributed at least one scored paper.

Request access

Scrutinise the evidence for any medicine.

A four-agent pipeline built to scrutinise the evidence.

General-purpose AI is not built to stress-test scientific evidence.

Three tiers of access. Each built for a different depth of need.

Scientific intelligence delivered when you need to know.

We work with a small number of pharmaceutical teams at a time.