Skip to content

David Fraile Navarro

GLP-1 drugs like Ozempic are among the world's most prescribed — and in 2023 regulators opened suicide investigations after adverse-event databases lit up, even though the randomised trials showed no such risk.

Code Video Cerebral Valley

Core trick

Computes FAERS reporting ratios against alternative comparators to test whether a spontaneous-report safety signal survives comparator changes.

Tags

healthscience

Description

GLP-1 drugs like Ozempic are among the world's most prescribed — and in 2023 regulators opened suicide investigations after adverse-event databases lit up, even though the randomised trials showed no such risk. We reconciled that contradiction by rebuilding the entire evidence base as four strictly-separated streams — designed studies, spontaneous reports, mechanism, and an exploratory benefit arm — with formal risk-of-bias and GRADE certainty on the designed studies, measured error rates throughout, and tooling calibrated against known-answer benchmarks (AUROC 0.76). The result: the spontaneous "alarm" is unstable — its reporting ratio swings from 0.51 to 3.29 depending only on the comparator, and jumps exactly when the drug hit the news. That's the fingerprint of comparator choice and notoriety bias, not a real drug effect. The hopeful twist: reading the same database the same way, the signal points the other way for addiction — toward less drinking, less smoking, less craving across four substance classes (exploratory, but a genuine lead). Every piece — the dual-reviewer risk-of-bias engine, the GRADE harness, and the calibrated openFDA client — is open under Apache 2.0 and runs automatically: a reusable pipeline for evidence synthesis, ongoing drug-safety monitoring, and discovering new indications.
dafraile/glp1-neuropsych-safety★ 10 forksPythonApache-2.0Created Jul 8, 2026 · last push Jul 14, 2026

README excerpt

# GLP-1 RA neuropsychiatric safety — evidence triangulation

Triangulating the GLP-1 receptor agonist neuropsychiatric safety signal.
**Primary question:** In adults on GLP-1 RAs, does designed-study and mechanistic
evidence support, refute, or fail to explain the neuropsychiatric safety signal
(suicidality, self-harm, depression) seen in spontaneous adverse-event reports?

Four evidence streams, kept **strictly separate and never pooled together**:
- **Stream A** — designed-study synthesis of the neuropsychiatric **harm** question
  (RCTs + observational pharmacoepidemiology).
- **Stream B** — FAERS spontaneous-report disproportionality (openFDA), **comparison
  only**. Scope: **semaglutide × SMQ suicide/self-injury** (not class-level).
- **Stream C** — mechanistic plausibility, as a structured bidirectional evidence map.
- **Stream D** — **exploratory** designed-study synthesis of the addiction/reward
  **benefit** question, separate from the primary safety question.

Every headline number below is generated from
[`analysis_manifest.json`](analysis_manifest.json) (machine-built from the data files;
run `./reproduce.sh` to regenerate and verify). See
[`TRIANGULATION_SYNTHESIS.md`](TRIANGULATION_SYNTHESIS.md) for the integrated reading and
[`VALIDATION_PROTOCOL.md`](VALIDATION_PROTOCOL.md) for measured pipeline error rates.

## Headline result (calibrated)

Designed studies and current regulatory assessments **do not show an increased average
risk** of suicidal ideation or be

Team