RocSite's engine reads a published trial, compares what it concluded to what its own primary endpoint actually showed, and flags the gap — then quote-verifies both sentences against the source and runs an adversarial pass that tries to refute the flag. Below are real cases that survived. The words are the papers' own.
A clinical trial declares, in advance, the one main question it's testing — its primary endpoint. If that result comes back negative (no real difference), the honest conclusion is "it didn't work."
Spin is when the paper claims a benefit anyway — by leaning on a post-hoc subgroup, switching to a secondary outcome, or wording the abstract to sound positive. Our engine reads the paper, pulls the claim and the actual result, and shows them side by side. Hover any underlined term for a definition.
These are methodological flags, not accusations of misconduct — outcome-reporting spin is common and often unintentional. We quote each paper's own published words; judge for yourself, and read the source.
Each case above was surfaced by an automated audit, quote-verified against its source, and run through an adversarial "try to refute this" pass before earning a place here. This is a sample — the engine has 6.2 million papers indexed and over 1,200 more candidate trials queued.
Every flag shows the paper's own two sentences — the claim and the actual result — each matched verbatim to the source. No paraphrase, no judgment call.
6.2M papers indexed and growing. Pointed at a full literature base — or a sponsor's own trial portfolio — the engine flags conclusion-vs-result gaps no human review could cover.
For journals, HEOR, regulators, and due-diligence teams: surface spin, outcome-switching, and overstated claims before they propagate.
Run a paper, a portfolio, or a body of literature through the RocSite engine and get signed, sourced integrity flags you can defend.
Request an Evidence Integrity Score →