Semaglutide + Pancreatitis: An AI-Assisted Signal Investigation

A complete pharmacovigilance signal investigation using 8 live data sources and 4 disproportionality measures. Every number was computed from current FDA, NLM, and WHO databases — not recalled from training data.

March 2026 | 8 data sources | 10 min read | Run it yourself

Semaglutide + pancreatitis is a confirmed statistical signal at POSSIBLE causality. All four disproportionality measures exceed their thresholds. The signal is already acknowledged in FDA-approved labeling (Section 5.2). Causality cannot be upgraded to PROBABLE without rechallenge data, which is unavailable for this drug-event pair.

Semaglutide is a GLP-1 receptor agonist approved for type 2 diabetes (Ozempic) and obesity (Wegovy). GLP-1 receptor agonists as a class have been associated with pancreatitis since exenatide's approval in 2005. The mechanism is biologically plausible: GLP-1 receptor activation stimulates pancreatic beta cells, and sustained stimulation may increase ductal pressure.

Pancreatitis is listed in semaglutide's FDA-approved labeling under Section 5.2 (Warnings and Precautions) and has been reported in postmarketing surveillance including cases of necrotizing pancreatitis. The boxed warning, however, is for thyroid C-cell tumors — a different safety concern.

This investigation quantifies the signal strength using four independent statistical measures, assesses causality using two validated algorithms, and compares semaglutide to liraglutide (a GLP-1 RA with longer market history) to evaluate whether this is a drug-specific or class-wide signal.

Data Sources

All data retrieved via NexVigilant Station MCP tools. Contingency table from OpenVigil reflects the full FAERS database (~20 million reports).

We computed four standard pharmacovigilance disproportionality measures from the same 2x2 contingency table. Each applies a different statistical correction, so agreement across all four provides stronger evidence than any single measure alone.

Consensus: Strong signal. All four measures exceed their thresholds with narrow confidence intervals. Chi-squared = 10,245 (p < 0.0001). This is not a marginal finding — the reporting rate for semaglutide + pancreatitis is approximately 7x the background rate.

What each measure tells us: PRR gives the raw proportional departure. ROR adjusts for event prevalence. IC measures information surprise (Shannon entropy). EBGM applies Bayesian shrinkage toward a prior — the most conservative correction. When all four agree, the signal is robust to methodological choice.

Statistical signal detection identifies disproportionate reporting but does not establish causation. We applied two validated causality algorithms to assess whether semaglutide plausibly caused the reported pancreatitis cases.

Why not PROBABLE? The Naranjo scale awards points for positive rechallenge (did pancreatitis recur when semaglutide was restarted?). This data is unavailable — clinicians rarely rechallenge with a drug suspected of causing pancreatitis. Without rechallenge, the maximum achievable Naranjo score is capped below the PROBABLE threshold. This is a data gap, not an exoneration.

To determine whether this signal is specific to semaglutide or shared across the GLP-1 receptor agonist class, we ran the same analysis on liraglutide (Victoza) — a GLP-1 RA approved since 2010 with 7 more years of post-market reporting history.

Interpretation: Both drugs produce a strong pancreatitis signal. Liraglutide's signal is approximately 2.5x stronger in magnitude, consistent with 7 additional years of accumulated reports (3,219 vs 2,068) rather than a fundamentally different safety profile. Both drugs share the same mechanism (GLP-1 receptor agonism), the same target organ (pancreas), and the same labeling section (5.2).

Conclusion: This is a class effect, not a semaglutide-specific finding. Any new GLP-1 receptor agonist entering the market should be monitored for pancreatitis from day one.

The FDA-approved labeling for semaglutide acknowledges pancreatitis across multiple sections:

The divergence between clinical trial incidence (0.1%) and FAERS signal strength (PRR 6.93) illustrates why post-market surveillance matters. Clinical trials are designed for efficacy, not safety signal detection — they are too short, too small, and exclude the populations most likely to experience rare adverse events.

A traditional signal assessment might stop at "PRR exceeds threshold — signal detected." This multi-source investigation revealed seven additional findings that change how the signal should be managed:

1. Trial vs post-market perspective divergence. Clinical trials show 0.1% incidence. FAERS shows PRR 6.93. Both are correct — they measure different populations over different timeframes. Neither alone tells the full story.
2. Causality capped at POSSIBLE by data gap. Both Naranjo and WHO-UMC assess POSSIBLE. This is not a weak result — it reflects the absence of rechallenge data, which clinicians rarely generate for serious adverse events.
3. High cascade potential. With 2,068 reports and growing, this signal affects prescribing decisions for millions of patients on GLP-1 agonists. Regulatory action (labeling change, REMS) would cascade widely.
4. Convergence unverified. We assessed a single time window. A quarterly PRR series from 2018-2026 would reveal whether the signal is stable, growing, or declining. This is the highest-priority next step.
5. Rechallenge void blocks escalation. Rechallenge data (did pancreatitis recur when semaglutide was restarted?) is the single piece of evidence that would upgrade causality from POSSIBLE to PROBABLE. It likely exists in medical records but is not in FAERS.
6. Class effect confirmed by comparator. Liraglutide shows the same signal pattern at 2.5x magnitude. Same mechanism, same organ, same labeling section. This is GLP-1 class pharmacology, not semaglutide-specific.
7. 104 published papers corroborate. PubMed returns 104 pharmacovigilance papers mentioning semaglutide as of 2026. Literature corroboration across independent research groups strengthens the biological plausibility assessment.

Semaglutide + pancreatitis is a confirmed signal at POSSIBLE causality.

The disproportionality is unambiguous (4/4 measures, chi-squared 10,245). The causality is limited by rechallenge data availability, not by contradictory evidence. The signal is a GLP-1 class effect, confirmed by liraglutide comparison. It is already acknowledged in FDA labeling.

Recommended actions: (1) Compute quarterly PRR trend 2018-2026 to assess signal trajectory. (2) Search for rechallenge case reports in published literature. (3) Monitor for potential REMS or labeling updates as GLP-1 prescribing volume increases with obesity indications.

Every finding in this article was computed from live data using NexVigilant Station. You can reproduce it in two ways:

Interactive Demo

Step through all 6 investigation steps in your browser. Each step calls the live Station API and shows the raw result.

Run the demo

Connect Your AI Agent

Add NexVigilant Station to Claude, and ask it to investigate any drug-event pair. No API key required.

MCP Server URL: mcp.nexvigilant.com/mcp

Then try: "Investigate metformin and lactic acidosis"

Connect Your AI

Methodology

This investigation used NexVigilant Station (146 public tools across 20 configurations). Data was queried from FDA FAERS via openFDA API, OpenVigil France for contingency tables, NLM DailyMed for approved labeling, NLM RxNav for drug identity resolution, and NLM PubMed for literature search. Disproportionality measures (PRR, ROR, IC, EBGM) were computed by NexVigilant's pharmacovigilance calculation engine. Causality was assessed using the Naranjo ADR Probability Scale and the WHO-UMC causality assessment system. All computations are deterministic and reproducible.

Limitations

FAERS is a spontaneous reporting database subject to reporting bias, under-reporting (estimated 90-95%), and confounding by indication. Disproportionality measures detect statistical associations, not causal relationships. This analysis should be interpreted alongside clinical evidence, not as a standalone causal claim.

Disclosure

NexVigilant is an independent pharmacovigilance technology company. This research was generated using our own tools to demonstrate their capabilities. We have no financial relationship with the manufacturers of semaglutide or liraglutide.