Every signal cites its source.
Only 1–10% of adverse drug events are reported to the FDA. Traditional signal detection takes 6–18 months. Skippy Pharmacovigilance runs the full MGPS pipeline across 7 authoritative sources — deterministically, traceably, and 21 CFR Part 11 compliant. Zero generative text.
Signal detection and PSUR extract
Full DPA suite on a known literature-confirmed signal — and automated PSUR period summary. Same output shape as production.
Full DPA suite: MGPS, ROR, PRR, IC/BCPNN, vigiRank, Weibull TTO — plus evidence-base cross-reference across 7 sources.
Six endpoints. Full lifecycle.
Every capability is a deterministic endpoint — no generative inference, no ambiguous output. Embed signal detection, causality scoring, and PSUR generation directly in your safety system.
Full DPA suite against FAERS + 6 evidence-base sources. Returns MGPS EB05, ROR, PRR, IC/BCPNN, MaxSPRT, vigiRank composite, and Weibull TTO onset classification per drug-reaction pair.
2×2 contingency table with stratified analysis across age, sex, country, and indication subgroups. Mantel-Haenszel pooling, EB shrinkage, and Weber masking detection. Full EM convergence guaranteed.
10-item Naranjo algorithm scored per drug-reaction pair. Returns numeric score, causality category (definite ≥9 / probable 5–8 / possible 1–4 / doubtful ≤0), and hash-chained audit record.
ICH E2B(R3) XML full round-trip import and export. CIOMS Form I plain-text intake. Fellegi-Sunter duplicate detection on every submission. Quarantine workflow for suspect duplicates.
Structured PSUR generation from signal queue, case data, and benefit-risk MCDA. EMA GVP Module VII compliant. Regulatory clock tracking for submission deadline enforcement. PDF + XML output.
Continuous ingestion from EMA EudraVigilance, MHRA Yellow Card, Health Canada MedEffect, PMDA, TGA, and Swissmedic. Returns emerging signals not yet in FAERS with cross-jurisdictional risk flags.
Five deterministic steps.
From raw ICSR intake to regulatory-ready export — no freeform text, no generative inference, every number derived from a peer-reviewed formula.
E2B(R3) XML and CIOMS Form I cases enter the ICSR pipeline with Fellegi-Sunter duplicate detection. Seven evidence-base sources populate Neo4j continuously — DailyMed, ClinicalTrials.gov, PubMed, FAERS, SIDER, DGIdb, ChEMBL.
Full DPA suite runs: ROR, PRR, IC/BCPNN, full MGPS/EBGM, MaxSPRT. Naranjo causality scored per drug-reaction pair. Duplicate detection applied before counting — no inflated signal scores.
Each candidate signal is traced through 7 finding sources with contradiction detection. SCCS, SSA, Weibull TTO, and TreeScan run as epidemiological validation passes.
vigiRank composite score + MGPS EB05 + EB shrinkage rank the signal queue. Stratified DPA surfaces subgroup heterogeneity. Reviewers act with role-gated decisions and electronic signatures.
CIOMS I/V PDF, E2B(R3) XML, and CSV outputs generated from templates — no freeform text, no generative inference. PSUR builder structures reports to EMA GVP Module VI.
Publication-grade epidemiology.
Every method implemented from its primary literature reference. No approximations. Each module is open to audit at every layer.
DuMouchel 1999 two-component gamma-Poisson shrinkage with full EM convergence. 15× precision improvement over simplified pipelines. Reports EB05, EBGM, ROR, PRR, IC/BCPNN, and MaxSPRT with Weber masking detection.
Per-stratum ROR across age, sex, country, and indication subgroups. Empirical Bayes shrinkage applied within each stratum. Mantel-Haenszel pooling produces a single adjusted MH-ROR across all strata.
Multi-criteria signal triage: IC025 + temporal trend + geographic spread + case informativeness + reviewer attention weighting. Caster et al. 2017 demonstrated 2.5× improvement in triage performance over DPA alone.
Each case serves as their own control, eliminating time-invariant between-person confounding. Gold standard for vaccine safety surveillance. Reports incidence rate ratio (IRR) with confidence intervals.
Fits Weibull distribution to time-to-onset data. Shape parameter κ classifies onset: early (κ < 0.8), random (κ ≈ 1), or delayed (κ > 1.2). WHO-UMC method with full parameter reporting.
Bernoulli hierarchical scan over the full MedDRA hierarchy: PT → HLT → HLGT → SOC. Detects class-level AE clusters that individual PT-level DPA would miss. Reports log-likelihood ratio with adjusted p-value.
Detects drug-AE signals from prescription ordering patterns without a comparator drug. Asymmetry in the sequence of drug A before drug B identifies probable adverse drug reactions. Reports adjusted sequence ratio.
Hash-chained immutable audit trail on every write. Electronic signatures with printed name, reason, and timestamp. RBAC: viewer, reviewer, manager, admin. GDPR retention management. Tamper-evident case chains ready for FDA inspection.
ICH E2B(R3) XML full round-trip import and export. CIOMS Form I plain-text intake. Full case schema: demographics, suspect and concomitant drugs, reactions, seriousness, outcome, reporter, causality, and stratum keys.
Fellegi-Sunter probabilistic record linkage across all ICSR fields. Weighted matching on patient demographics, drug, reaction, reporter, and date. Prevents inflated signal counts from duplicate spontaneous reports.
Neo4j graph integrates DailyMed, ClinicalTrials.gov, PubMed (20 specialties), FAERS, SIDER, DGIdb, and ChEMBL. Every signal is traceable to source literature, labels, and trials. Contradiction detection across sources.
Unified view spanning trial AEs (ClinicalTrials.gov), labeled events (DailyMed), and post-market spontaneous signals (FAERS). Tracks AE evolution from early clinical development through post-approval pharmacovigilance.
Not a workflow tool. A scientific engine.
Oracle Argus, ArisGlobal LifeSphere, and Veeva Vault Safety manage case workflow. Skippy adds the scientific layer they lack — lineage-traced provenance, gold-standard disproportionality analysis, and a compliance chain auditable line-by-line.
Every signal resolves to a citation chain: DailyMed label → ClinicalTrials.gov arm → PubMed abstract → FAERS spontaneous report. Auditors can walk the evidence chain node-by-node. No black box. No hallucination.
DuMouchel 1999 two-component gamma-Poisson shrinkage with full EM convergence. Benchmarked at 8 true signals vs. 121 on a 2,601-pair stress test. Zero false positives on all 8 known positive controls.
Every output is deterministic, template-driven, and traceable. No LLM inference. The same query returns the same answer every time — citable in regulatory submissions without post-processing.
Hash-chained immutable audit trail on every write. Electronic signatures with printed name, reason, and timestamp. RBAC with four defined roles. Ready for FDA inspection on day one.
Six jurisdictions. Continuous ingestion.
Signals that emerge in one regulatory jurisdiction before appearing in FAERS give pharmacovigilance teams a 3–9 month head start. Skippy monitors all six major feeds in parallel — surfacing cross-jurisdictional risk before it reaches the US market.
European Medicines Agency's pharmacovigilance database. Primary source for EU post-market safety data — signals here often precede FDA FAERS by months.
UK Medicines and Healthcare products Regulatory Agency spontaneous reporting scheme. Post-Brexit divergence makes UK signals an independent cross-reference from EU data.
Canadian adverse reaction database. Demographically distinct population with different prescribing patterns — valuable for surfacing AEs that are age- or sex-stratified.
Pharmaceuticals and Medical Devices Agency adverse drug reaction database. Japanese population pharmacogenomics diverge from Western populations — critical for metabolism-dependent AEs.
Therapeutic Goods Administration Database of Adverse Event Notifications (DAEN). Southern hemisphere population with distinct seasonal and environmental exposure patterns.
Swiss Agency for Therapeutic Products vigilance database. Neutral jurisdiction with distinct reporting culture — frequently surfaces signals missed by larger national databases.
Rofecoxib's cardiovascular risk signal was present in FAERS data years before market withdrawal. Traditional signal detection timelines — 6–18 months — are not a technical constraint. They are a methods constraint.
Full MGPS, multi-source cross-reference, and vigiRank composite scoring compress the signal-to-action timeline. Every confirmed signal is traceable to the exact reports, labels, and literature that produced it — before it reaches a reviewer's queue.
“We will not submit a safety report we cannot defend line-by-line to an FDA inspector.”
Common requirement from pharmacovigilance leads at large pharma
Skippy Pharmacovigilance is built for that requirement. Every signal detection run generates a tamper-evident audit record with the exact data sources, statistical methods, thresholds used, and reviewer identity — hash-chained and ready for inspection.
Pharma, biotech, and regulatory agencies
ICH E2B(R3) · GVP Module VI · FDA 21 CFR Part 11
ICH E2B(R3) full round-trip ICSR import and export. EMA GVP Module VI PSUR structure. FDA 21 CFR Part 11 electronic records and signatures. Regulatory clock tracking for PSUR/PBRER deadline enforcement.
HIPAA for de-identified EHR data integration. GDPR retention management for EU patient data. SOC 2 Type II certified. Every detection run generates a tamper-evident audit record with data sources, methods, and thresholds — ready for inspection.
Seven authoritative sources.
Every signal is cross-referenced against all seven. Contradiction detection fires when sources disagree — surfacing ambiguity explicitly, not silently dropping it.
Spontaneous adverse event reports — primary signal source
FDA-approved drug labels — labeled event cross-reference
Trial-phase AEs — pre-market signal history
20 medical specialties — literature evidence
Side effect resource — validated adverse effect database
Drug-gene interactions — mechanism cross-reference
Bioactivity data — target and mechanism evidence
Other Skippy medical products
Real-time drug-drug interaction checking with CYP enzyme mechanism, panel risk scoring, and verified alternatives — at the point of prescribing.
Hard verification gate for any AI-generated medical claim. Calibrated confidence, source lineage, and an immutable audit record on every call.
CPIC-grounded genotype-guided dosing for 21,000+ gene-drug pairs — personalized medication safety at the point of prescribing.
Pharmacovigilance Decision Support · Regulatory Compliance. Skippy Pharmacovigilance is a signal detection and pharmacovigilance documentation platform. All signal assessments, causality ratings, and PSUR outputs are generated deterministically from structured data and peer-reviewed statistical methods — not generative AI inference. Outputs are intended for review by qualified pharmacovigilance professionals before regulatory submission. 21 CFR Part 11 compliant. HIPAA-ready.
See Pharmacovigilance in your safety program
We work with large pharma, biotech, and CROs running post-market surveillance. Let's talk about your signal detection and PSUR problem.