The Virtual Inspector that reads every record.
A TMF-only virtual inspector that works through a structured, ten-block inspection workflow over your ingested extract — opening and reading every raw file, computing across every metadata and audit-trail row, and assembling a cited, human-reviewable inspection package. Not a cheaper review. A faster one.
Exhaustive coverage, run as a discipline
Panorama TMF QC replaces sampling with full-population machine review. Every record is enumerated, every raw file is opened, and every conclusion is bound to cited evidence with reconciling denominators.
Full-population read — no sampling
Machine execution removes the human reading-time constraint, so every metadata row, every audit row, and every raw file is reviewed. Sampling is never a substitute for reading.
Rules-driven, ten-block workflow
A deterministic engine works the same ten sequential blocks every time — explicit thresholds, expectedness rules, and branch logic — so results are repeatable and inspection-ready, not improvised.
Cited, human-reviewable outputs
Every claim carries its citation, counts, and denominators; every absence carries its documented search. The Block 10 package is delivered for human signoff — outputs support ALCOA+, they don't replace your judgment.
Not another cheaper review. A faster one.
Traditional TMF review samples because human reading time is the scarce resource. Panorama removes that constraint — so risk stops deciding whether a record is examined and starts deciding only the order, depth, and escalation of the read.
Sampling, because people don't scale
- A fraction of records are opened; the rest are inferred.
- Coverage is bounded by available reviewer hours.
- Findings depend on which sample a reviewer happened to pull.
- Re-running the review means re-spending the hours.
Read everything, allocate depth by risk
- Every accessible raw file is opened and read.
- Risk repurposed to order, depth, and escalation — never to skip.
- Files that fail to open or read are itemized, not silently dropped.
- Denominators reconcile: total files = read + failed.
The value is calendar-time compression to a defensible conclusion — not headcount savings on a review that was already incomplete.
Ten sequential work blocks
Each block produces reusable, cited outputs that become inputs to the next. Sequence comes before content, and citation comes before conclusion — so impressions are always anchored to established criteria, chronology, and expectedness.
Navigation foundation
Identify the trial, inventory the sources, and map the extract — metadata, audit trail, raw files, criteria artifacts, blinding posture, and cutoff anchors.
Expected-record framework
Define what records should exist from the TMF plan, taxonomy, protocol, site/country lists, and milestones — before testing what does.
Trial chronology
Reconstruct when key events occurred and when each record should have existed, establishing the timeline conclusions are measured against.
Whole-population screening
Compute across the full population for missing, inconsistent, inaccessible, late, misclassified, duplicated, or untraceable records.
Prioritize the review
Order the read and assign depth tiers — critical, high, standard — so the highest-impact conclusions stabilize earliest.
Targeted raw-file review
Open and read the raw content — confirming it supports the metadata, expectedness, entity assignment, version, and chronology, with visual review where needed.
Audit-trail lifecycle
Reconstruct each record's lifecycle from audit history — creation, replacement, reclassification, deletion, supersession — across the full audit population.
ALCOA+ assessment
Test metadata, raw files, and audit trail together against all ten ALCOA+ attributes, from attributable through traceable.
Expand & classify findings
Expand confirmed observations to test isolated vs. systemic, then classify by impact — with serious-breach candidates tagged for human attention.
Produce inspection outputs
Assemble the final package — coverage statements, evidence packages, findings, limitations, and an urgent-findings extract — for human review.
What the inspection covers
The TMF is reviewed by document family, not document number — grouped to the sponsor's own filing declaration where a verified index crosswalk exists.
Trial design & governance
Protocol, amendments, IB, monitoring/data/statistics plans, randomization and delegation.
Regulatory & ethics approvals
Authorizations, IRB/IEC approvals, amendment and consent approvals, version and scope alignment.
Site startup & activation
Selection, qualification, delegation, training, and green-light evidence ahead of first activity.
Participant protection & consent
ICF versions, translations, approvals, and re-consent — aligned to protocol and amendments.
Safety management
Safety plans, SUSAR/SAE communications, IB updates, and traceability to affected documents.
Investigational product / device
Shipment, receipt, storage, accountability, returns, and controlled blinding-sensitive records.
Monitoring, deviations & CAPA
Oversight reports, issue logs, deviations, and corrective actions traced through to closure.
Vendor & CRO oversight
Qualification, transfer-of-obligation, oversight evidence, and deliverable reconciliation.
Data, statistics & CSR
Database lock, SAP alignment, randomization/blinding, and clinical study report chronology.
Closeout, archive & retention
Closeout evidence, final accountability, archive inventory, and long-term retrievability.
Inspection time compression estimator
A rough, illustrative comparison of elapsed reading time: a manual full read versus Panorama's machine read of the same population. Adjust the inputs to your extract.
Illustrative estimate only. Manual figure assumes one reviewer reading every page at the chosen rate; Panorama figure assumes continuous processing at the selected mode's throughput. Real elapsed time depends on extract size, file legibility, hardware, and the depth tiers actually triggered. This estimator models reading time, not the full ten-block workflow.
Built for defensible conclusions
The design choices that separate Panorama from a generic AI summarizer — each one in service of an inspection-ready, reconstructable result.
Complete record representation
Every record is counted in every denominator and reconciled in every coverage identity — even records that fail to open become evidence, never gaps hidden by sampling.
Citation before conclusion
Reading everything licenses nothing on its own. Every stored claim carries its citation, counts, and denominators; every absence carries its documented search.
Human signoff, always
Final regulated conclusions remain with your reviewers. Panorama makes the package complete, cited, and checkable — the human keeps the decision.
Rules-driven engine
Deterministic engines apply explicit thresholds and branch logic over the extract, so the same input produces the same, repeatable, auditable output.
Privacy Blind & data protected
Participant identifiers are never transcribed and the trial blind is protected end to end — sensitive content is cited by location and structure, not reproduced.
Reconstructability focus
The standard is whether an independent reviewer could reconstruct trial conduct from the TMF — grounded in ICH E6(R3), EU 536/2014, EMA, FDA, and APAC expectations.
See Panorama TMF QC against your own extract
Request a pilot, or walk through the ten-block workflow with our team and decide where full-population review fits your inspection-readiness program.
Panorama TMF QC reviews an already-ingested TMF extract; it is not a live-eTMF, electronic-signature, or source-data-verification tool. Outputs support ALCOA+ assessment and inspection readiness and are intended for human review — they do not constitute a regulatory determination. Mission Viejo, CA · info@wcssrg.com