Signal management · PASS · benefit–risk
A safety signal shouldn't wait six months for an answer.
Run the same pre-specified protocol across every connected real-world data source in parallel. Aggregate-only. Reproducible. Inspectable. The infrastructure regulators are already asking you to use.
SIGNAL MANAGEMENT
From disproportionality to a refined RWD answer
EPIDEMIOLOGY IN PV
PASS & PAES without five-SOW orchestration
BENEFIT–RISK
Incidence, background rates, subgroup attributions
QPPV & REGULATORY
Evidence an authority can re-run, not just read
01 · The shift
Signal evaluation is the slowest step in pharmacovigilance. It doesn't have to be.
THE OLD CADENCE
- Disproportionality hit in spontaneous reports. Signal raised.
- Three CROs engaged — three CDMs, three analytic specs, three contracts.
- Six-to-twelve months of scoping, programming, QC and reconciliation.
- Answers arrive heterogeneous. Prescribing continues in the meantime.
WITH UNISON
- Safety scientist opens a pre-specified protocol template.
- Protocol compiles to UQL and fans out across connected biobanks.
- Per-source aggregate results return with provenance and heterogeneity visible.
- Signal refinement, background-rate comparison, subgroup cut — same day, same protocol.
02 · What Safety teams can run
Five workflows, one query surface.
SIGNAL REFINEMENT
Validate a spontaneous-report signal in RWD
Move from disproportionality to an incidence-based picture. Exposed vs unexposed, time-at-risk, latency, seriousness — with a population denominator, not a reporting ratio.
OUTPUT
Incidence rates · IRRs · time-to-event · subgroup breakdownsBACKGROUND RATES
Expected event rates in the target population
Observed-over-expected without a standing study. Refresh the background rate every time the label, the indication or the population shifts.
OUTPUT
Age-/sex-standardised rates · O/E with CIs · per-source detailPASS / PAES
Post-authorisation safety & efficacy studies
Pre-specified protocol, regulator-agreed analyses, reproducible from the UQL artefact. Heterogeneity you can inspect, not a black box you have to defend.
OUTPUT
PASS report · methods artefact · per-source transparencyDRUG UTILISATION
DUS, channelling and off-label use
Characterise who is actually being prescribed the product. Detect channelling. Quantify off-label patterns. Keep the utilisation picture current between PSURs.
OUTPUT
DUS tables · channelling metrics · indication mixBENEFIT–RISK REFRESH
Keep the benefit–risk picture current
When the evidence base moves — new indication, new population, new safety finding — re-run the whole benefit–risk package. Not a new study, a re-executed artefact.
OUTPUT
Incidence · attributable risk · subgroup breakdowns03 · Pre-specified, executed across the federation
From signal to regulator-ready evidence, without the five-SOW orchestration.
01
Pre-specify
Protocol written against OMOP concepts. Exposure, outcome, cohort, time-at-risk — locked before execution.
02
Compile
Unison compiles the protocol to UQL — a deterministic, aggregate-only contract regulators can read.
03
Federate
UQL fans out across connected real-world data sources. Data never leaves the custodian.
04
Inspect
Per-source counts, rates and heterogeneity are visible. Differences are something you can explain.
05
Replay
Every result is backed by a replayable UQL artefact. A regulator re-runs, not just reads.
A DAY IN SAFETY
"Is the acute hepatic injury rate elevated in the on-label population?"
A signal-management lead needs an incidence-based refinement before the next safety board. Open the template. Lock the parameters. Let the federation run.
Pre-specified
SAP locked before execution
Executed
federated across 4 biobanks
Inspected
per-source heterogeneity visible
Replayable
UQL artefact, regulator-ready
# unison · safety workspace
> "Incidence of acute hepatic injury · on-label cohort · 180d TAR · vs matched comparator."
→ Template: signal-refinement · compiled to UQL
→ Federated across 4 biobanks · aggregate-only
Exposed n 62,189
Comparator n 188,402
Incidence / 1 000 PY
Exposed 3.8 (95% CI 3.3–4.3)
Comparator 2.1 (95% CI 1.9–2.3)
IRR 1.82 (95% CI 1.55–2.14)
I² 22% (low heterogeneity)
# replayable artefact · uql://query/sig-7a12
04 · The regulatory shape
Aligned to the direction of travel in regulator guidance.
Pre-specification & replay
Protocol locked before execution. Every result tied to a replayable UQL artefact a regulator can re-run on the same data.
Aggregate-only by design
UQL cannot express a patient-level query. Exfiltration is structurally impossible — not merely prohibited by policy.
Federated, custodian-respecting
The analysis travels to the data. Custodians keep control. Every mapping decision is attributable, logged and reversible.
Standards & fit:· OMOP CDM-native· Cyber Essentials Plus· CFR 21 Part 11-ready· EHDS-aligned· DARWIN-EU compatible· GVP-aware
05 · What changes for the function
From "commission a study" to "execute the protocol."
SIGNAL-TO-ANSWER LATENCY
months → days
Refine a signal inside the window where prescribing can still respond.
HETEROGENEITY
black box → inspectable
Per-source results, visible. Differences are a finding, not a defect.
EVIDENCE SHELF-LIFE
one-off study → living artefact
Re-run on new data; the protocol — and the audit trail — stays intact.
Scope a pilot with your safety team in 2 weeks