Case Studies

The warning was already in the early paper.

Assets that failed in the clinic, traced back to the pivotal early paper. Each is a timeline: the defect the Rigor Engine flagged, then the Phase 3 failure it foreshadowed years later. You didn’t need the readout to distrust them.

Request a demo

Three assets, three trails that ended in failure

Different programs, different defects. Each early paper carried concrete, machine-checked problems that floored it at 1★, years before the clinic confirmed the risk.

DLL3 antibody-drug conjugate

Asset withheldsmall-cell lung cancer

2017Early paper

Phase 1 · first-in-human · The Lancet Oncology · Not retracted

1★

2 years later

2019 Failed

Phase 3 · TAHOE

Stopped early for worse survival; the asset was discontinued.

What the engine caught in the 2017 paper

seriousEfficacy rests on a surrogate endpoint with no demonstrated link to clinical benefit
seriousTwo cited references that resolve to nothing in any registry
lowA result sentence duplicated verbatim between the Abstract and the Results

Full report and asset name on a demo

PDGFRα monoclonal antibody

Asset withheldadvanced soft-tissue sarcoma

2016Early paper

Phase 2 · randomized · The Lancet · Not retracted

1★

4 years later

2020 Failed

Phase 3 · ANNOUNCE

Failed to confirm the survival benefit; accelerated approval was withdrawn.

What the engine caught in the 2016 paper

seriousA statistically impossible result: a reported confidence interval that excludes its own point estimate
mediumReported percentages that do not match their own counts

Full report and asset name on a demo

anti-sclerostin monoclonal antibody

Asset withheldosteogenesis imperfecta

2024Early paper

Phase 2b · ASTEROID · Journal of Bone and Mineral Research · Not retracted

1★

same trial

2024 Failed

Phase 2b · primary analysis

The headline did not survive its own primary endpoint; the program was redesigned before advancing.

What the engine caught in the 2024 paper

seriousThe primary endpoint did not reach significance, so the efficacy claim leans on a secondary
seriousThe placebo arm was dropped mid-study
mediumParticipant counts that do not reconcile

Full report and asset name on a demo

None of these papers was retracted. The engine caught every defect from the paper text itself.

We grade the paper, not the outcome

The engine never sees the trial result. Beside disease- and target-matched controls that succeeded, the failed assets sit at 1★, the controls higher, except where a success was itself reported poorly.

Failed assets

1.0★ all three

Each floored by a concrete, text-level defect in the early paper.

Matched controls

2.92★ avg

6 disease/target-matched programs that succeeded.

The tell that it’s honest: A Phase 3 success that still scored low; the engine grades the paper, not the outcome. The engine never sees the trial outcome. It measures evidence quality in the text and lets the reader draw the risk conclusion.

What “early paper” means

The pivotal Phase 1/2 or target-engagement paper a diligence team read before committing. The later phase records the outcome only; it is never what the engine reviews.

How the star works

A deduction model: every paper starts at 5★ and loses stars for concrete, named findings. A single critical defect can floor it at 1★, so one serious problem is never averaged away.

The honest limit

Rigor is not outcome. A well-reported study can still fail; a sloppy paper can still describe a drug that works. The claim is narrow: the engine surfaces the defects that make early data unsafe to bank on. A triage signal, not a verdict.

See the full reports

Schedule a demo to see every asset, finding, and live Rigor Record

For BD, licensing, and diligence teams: we walk the full case studies with you.

Request a demo See a full sample Rigor Record

Case Studies

The warning was already in the early paper.

Request a demo

Three assets, three trails that ended in failure

Different programs, different defects. Each early paper carried concrete, machine-checked problems that floored it at 1★, years before the clinic confirmed the risk.

DLL3 antibody-drug conjugate

Asset withheldsmall-cell lung cancer

2017Early paper

Phase 1 · first-in-human · The Lancet Oncology · Not retracted

1★

2 years later

2019 Failed

Phase 3 · TAHOE

Stopped early for worse survival; the asset was discontinued.

What the engine caught in the 2017 paper

seriousEfficacy rests on a surrogate endpoint with no demonstrated link to clinical benefit
seriousTwo cited references that resolve to nothing in any registry
lowA result sentence duplicated verbatim between the Abstract and the Results

Full report and asset name on a demo

PDGFRα monoclonal antibody

Asset withheldadvanced soft-tissue sarcoma

2016Early paper

Phase 2 · randomized · The Lancet · Not retracted

1★

4 years later

2020 Failed

Phase 3 · ANNOUNCE

Failed to confirm the survival benefit; accelerated approval was withdrawn.

What the engine caught in the 2016 paper

seriousA statistically impossible result: a reported confidence interval that excludes its own point estimate
mediumReported percentages that do not match their own counts

Full report and asset name on a demo

anti-sclerostin monoclonal antibody

Asset withheldosteogenesis imperfecta

2024Early paper

Phase 2b · ASTEROID · Journal of Bone and Mineral Research · Not retracted

1★

same trial

2024 Failed

Phase 2b · primary analysis

The headline did not survive its own primary endpoint; the program was redesigned before advancing.

What the engine caught in the 2024 paper

seriousThe primary endpoint did not reach significance, so the efficacy claim leans on a secondary
seriousThe placebo arm was dropped mid-study
mediumParticipant counts that do not reconcile

Full report and asset name on a demo

None of these papers was retracted. The engine caught every defect from the paper text itself.

We grade the paper, not the outcome

Failed assets

1.0★ all three

Each floored by a concrete, text-level defect in the early paper.

Matched controls

2.92★ avg

6 disease/target-matched programs that succeeded.

What “early paper” means

The pivotal Phase 1/2 or target-engagement paper a diligence team read before committing. The later phase records the outcome only; it is never what the engine reviews.

How the star works

A deduction model: every paper starts at 5★ and loses stars for concrete, named findings. A single critical defect can floor it at 1★, so one serious problem is never averaged away.

The honest limit

See the full reports

Schedule a demo to see every asset, finding, and live Rigor Record

For BD, licensing, and diligence teams: we walk the full case studies with you.

Request a demo See a full sample Rigor Record