TrialVyx Blog
Pharmacovigilance Signal Intelligence.
Methods, analysis, and regulatory perspective on drug interaction detection — written for pharmacovigilance scientists, not for marketing audiences.
Why Multi-Drug Interaction Signals Hide in Plain Sight in FAERS
Pairwise ROR analysis misses 40% of clinically significant DDI signals. Here's why the math fails for polypharmacy and what graph-based analysis finds instead.
All Articles
Technical writing from the TrialVyx team
Why Multi-Drug Interaction Signals Hide in Plain Sight in FAERS
Pairwise ROR analysis misses 40% of clinically significant DDI signals. Here's why the math fails for polypharmacy and what graph-based analysis finds instead.
AI-Assisted MedDRA Coding: Where It Helps and Where It Doesn't
Auto-coding with AI reduces ICSR processing time significantly — but certain AE narratives require human judgment that current models consistently get wrong.
The Limits of Disproportionality Analysis in Modern PV
ROR, PRR, and BCPNN were designed for single-drug monitoring. Post-market safety in 2025 requires handling polypharmacy regimens at scale.
91 Days: Measuring Signal Detection Lead Time Against FAERS Confirmations
How we measured detection lead time by retrospectively testing our model against 312 historically confirmed FAERS signals from 2019-2022.
FDA FAERS for Safety Scientists: What the Public Database Actually Contains
FAERS holds over 20 million adverse event reports but most safety scientists only query a fraction. A practical guide to what's in there and what isn't.
Reconciling FAERS and EudraVigilance: Where the Data Gaps Are
Running parallel signal detection across US and EU databases sounds straightforward — until you hit the MedDRA version mismatches and reporter-type bias.
Signal Management Bottlenecks: What PV Teams Actually Lose Time To
We surveyed 24 pharmacovigilance scientists about where signal management time actually goes. The answers weren't where most automation vendors focus.
Graph Neural Networks vs. Bayesian Networks for DDI Prediction: A Practical Comparison
Two competing computational approaches to multi-drug interaction prediction, compared on FAERS real-world data rather than benchmark datasets.
Automating ICSR Triage: A Case Study from a Generic Manufacturer's PV Department
How one PV team reduced ICSR classification time from 4.2 hours to 47 minutes per case through structured automation — and where they still needed human review.
What Can and Can't Be Automated in PSUR Preparation
Aggregate reports aren't just data compilation — they require scientific judgment. Here's a clear line between what's automatable and what regulatory agencies expect a human to have written.
The Sensitivity-Specificity Tradeoff in Pharmacovigilance Signal Detection
Maximizing signal sensitivity in FAERS produces noise that overwhelms PV teams. Here's how to set detection thresholds that balance early warning with workload reality.
REMS Programs and Post-Market Safety Monitoring: What Signal Detection Feeds Into
Risk Evaluation and Mitigation Strategies require ongoing safety signal monitoring. A practical overview of how signal detection data feeds REMS obligations.