The Forensic Audit Thesis — in one page
Where Few Are Looking
Last updated 10-15-2026, 04:47 PM PT
11%
of 53 “landmark” cancer papers were confirmed in Amgen’s own attempts
Begley & Ellis, Nature, 2012 (a commentary; the authors chose papers describing something new)
20–25%
of Bayer’s 67 in-house target-validation projects had published data completely in line with Bayer’s own findings
Prinz et al., NRDD, 2011, as summarized in open-access reviews; Begley & Ellis give about 25%
25.4%
total quotation error rate in a pooled analysis of studies of medical journal articles (major errors 11.9%, minor 11.5%; minor includes secondary citations)
Jergas & Baethge, PeerJ, 2015
$28B
per year (estimate) — US spend on preclinical research that is not reproducible
Freedman et al., PLOS Biology, 2015
This is the problem. Below is where everyone stands on solving it.
The market map — firms we reviewed, plotted
Two axes separate what these firms do. Subject on the vertical — patents versus scientific evidence. Method on the horizontal — expert opinion versus systematic reproducible process. The placements are our own judgment from each firm’s public materials, not the firms’ claims. By “citation-chain audit” we mean tracing each load-bearing claim in a target’s evidence to its primary source and tagging each link supported, partial, contradicted or missing. “Subject” is what a firm’s main product analyzes; “Method” is whether its output rests on an individual expert’s judgment or a documented, repeatable process. The firms shown are examples drawn from our April 2026 review of what each publishes about its own services, not a complete census.
IP × Expert Judgment
Bespoke IP counsel — case-by-case, not a productized service
IP × Systematic
PatSnap
Derwent
Questel
Cipher
Innography
Patent and IP analytics. Some of these platforms also index scientific literature (PatSnap’s site cites 260M+ scientific papers); in our review (April 2026) we found none marketing a citation-chain audit of the cited evidence.
Science × Expert Judgment
Alacrita
Technical and commercial due-diligence consultants (Alacrita’s site describes “technical and commercial assessments”). Expert review; in our review (April 2026) we found no named, reproducible citation-chain method described on their site.
Science × Systematic
None found.
We found no firm marketing a named, reproducible citation-chain audit service for biotech M&A (our review, April 2026; not exhaustive). Providers of systematic literature reviews and evidence synthesis work nearby; they summarize a body of literature rather than audit whether a target’s own citations support its claims.
Apparent white space
Method
Expert judgment →
← Systematic methodology
Subject
↑ IP / legal
↓ Scientific evidence
Published research on the problem
11%
Amgen
Only 6 of 53 “landmark” cancer papers were confirmed. The papers were deliberately chosen as describing something new.
Begley & Ellis, Nature, 2012 (a commentary)
20–25%
Bayer
Of Bayer’s 67 in-house target-validation projects, about this share had published data completely in line with Bayer’s own findings.
Prinz et al., NRDD, 2011, as reported in Begley & Ellis, Nature, 2012
Most
Ioannidis
An essay arguing that, under common research conditions, most published findings may be false.
PLOS Medicine, 2005 (an essay)
$28B
US Spend
Estimated per year on preclinical research that is not reproducible, in the US alone.
Freedman et al., PLOS Bio, 2015
25.4%
Journal quotations
Total quotation error rate in a pooled analysis of studies of medical journal articles. Major errors, where the cited source does not accord with the claim, were 11.9%; minor errors 11.5%.
Jergas & Baethge, PeerJ, 2015
31%
Press releases
Of 100 oncology press releases from 8 large pharmaceutical companies (2011 to 2016), only 31 included the size of the study findings.
Qunaj et al., JAMA Oncology, 2018
94% vs 51%
Antidepressant trials
Share of trials that looked positive in the published literature (94%) versus in the FDA’s own analysis (51%).
Turner et al., NEJM, 2008
One deal, in brief
AbbVie acquires Stemcentrx — announced April 2016, closed June 2016
AbbVie paid about $5.8 billion upfront for Stemcentrx and its lead drug Rova-T, a DLL3-targeted antibody-drug conjugate for small cell lung cancer. In December 2018 the Phase 3 TAHOE trial stopped enrolling after an independent monitoring committee recommended it, citing shorter overall survival on Rova-T than on the topotecan comparison arm. AbbVie estimated a net impact of approximately $4 billion (8-K, January 2019) and ended the Rova-T program in 2019.
→ Read the case, built from AbbVie’s filings and the original papers
~$4B
Net impact estimate, Jan 2019 8-K
The pain — why this is not fixable with current tools
Many non-scientist biotech CEOs have no routine way to audit their own evidence chain before investors, boards, or acquirers probe it. Senior scientific advisors evaluate plausibility. Patent tools map IP. In our review, neither walks the citation chain backward to verify that every link holds.
In our view the problem sits upstream of the diligence tools we reviewed.