Castor’s EMR-to-EDC extraction pipeline converts medical records into structured EDC data. Every field reviewed by a trained abstractor before it enters your database.
$48.75 / chart
30 min / chart
Trusted by medical device companies, pharma sponsors, and outcomes research teams worldwide
From PMCF registries to oncology RWE to prospective natural history studies, Castor handles your clinical data end to end.
EU MDR and FDA 522 post-market requirements demand structured patient-level data over years of follow-up. Castor runs PMCF registries built for regulatory submission, with field-level audit trails and CDISC export.
In practice
A leading orthopedic sports medicine device company ran a multi-site PMCF registry across EU and US sites, tracking patient-reported outcomes against implant performance over 24 months.
An energy-based device sponsor used Castor’s AI extraction pipeline to abstract post-procedure records for an FDA 510(k) PMCF study, cutting abstraction cost by over 70%.
A dermatology diagnostics company structured historical biopsy records and imaging metadata for a multi-year PMCF cohort across pediatric and adult populations.
Send automated patient surveys at 1, 3, and 5-year follow-up intervals without manual coordination. Castor’s ePRO solutions support multi-language delivery across mobile, tablet, and web. PMCF studies on the platform have collected more than 434,000 patient survey data points to date.
In practice
A top-5 global outcomes research firm ran a multi-country retrospective RWE evaluation of a first-line oncology biologic with Castor, abstracting treatment patterns and outcomes in a real-world population outside trial conditions.
A specialty CRO used Castor to structure multi-site chart review records for an MS therapy study spanning neurological outcomes across European treatment centers.
Castor’s clinical data management team delivered database lock and define.xml submission packages for a 12-country HEOR study, reducing lock-to-delivery time to under two weeks.
Castor supports direct-to-patient (D2P) data collection for natural history studies, prospective disease registries, drug registries, and external control arms (ECAs).
In practice
A rare disease sponsor structured an ongoing natural history registry with Castor to build an external control arm for an upcoming FDA single-arm submission, capturing longitudinal patient records prospectively over 36 months.
A biotech sponsor built a prospective disease registry for a small-population indication with Castor’s D2P capabilities, capturing longitudinal patient data to characterize natural disease progression and support future regulatory submissions.
An academic medical center ran a multi-institution longitudinal cohort with Castor, combining prospective patient follow-up with retrospective record abstraction in a single unified database.
From PMCF registries to oncology RWE to prospective natural history studies, Castor handles your clinical data end to end.
EU MDR and FDA 522 post-market requirements demand structured patient-level data over years of follow-up. Castor runs PMCF registries built for regulatory submission, with field-level audit trails and CDISC export.
In practice
A leading orthopedic sports medicine device company ran a multi-site PMCF registry across EU and US sites, tracking patient-reported outcomes against implant performance over 24 months.
An energy-based device sponsor used Castor’s AI extraction pipeline to abstract post-procedure records for an FDA 510(k) PMCF study, cutting abstraction cost by over 70%.
A dermatology diagnostics company structured historical biopsy records and imaging metadata for a multi-year PMCF cohort across pediatric and adult populations.
Send automated patient surveys at 1, 3, and 5-year follow-up intervals without manual coordination. Castor’s ePRO solutions support multi-language delivery across mobile, tablet, and web. PMCF studies on the platform have collected more than 434,000 patient survey data points to date.
In practice
A top-5 global outcomes research firm ran a multi-country retrospective RWE evaluation of a first-line oncology biologic with Castor, abstracting treatment patterns and outcomes in a real-world population outside trial conditions.
A specialty CRO used Castor to structure multi-site chart review records for an MS therapy study spanning neurological outcomes across European treatment centers.
Castor’s clinical data management team delivered database lock and define.xml submission packages for a 12-country HEOR study, reducing lock-to-delivery time to under two weeks.
Castor supports direct-to-patient (D2P) data collection for natural history studies, prospective disease registries, drug registries, and external control arms (ECAs).
In practice
A rare disease sponsor structured an ongoing natural history registry with Castor to build an external control arm for an upcoming FDA single-arm submission, capturing longitudinal patient records prospectively over 36 months.
A biotech sponsor built a prospective disease registry for a small-population indication with Castor’s D2P capabilities, capturing longitudinal patient data to characterize natural disease progression and support future regulatory submissions.
An academic medical center ran a multi-institution longitudinal cohort with Castor, combining prospective patient follow-up with retrospective record abstraction in a single unified database.
Castor’s AI does the heavy lifting. A trained human reviewer closes the loop before any value enters your EDC. Every extracted field is traceable back to its source document.
cost reduction vs. manual abstraction
faster per chart (30 min → 6 min)
error rate after human validation
Visual Audit Trail — source to EDC
Patient presented on 12-Mar-2024 with confirmed diagnosis of NSCLC. PD-L1 expression >50%. Primary treatment initiated with pembrolizumab monotherapy.
Index date: 12-Mar-2024
ECOG performance status: 1. Stage at diagnosis: IVB. Histology: adenocarcinoma. EGFR/ALK/ROS1: wild-type.
ECOG PS: 1
Tumor proportion score (TPS): 72%. PD-L1 IHC 22C3 pharmDx assay. Sufficient biopsy material confirmed.
PD-L1 TPS: 72%
AI extraction requires de-identified records. Remote SDV requires the original identifiable source. Castor solves both with a dual-pathway architecture, so neither obligation is sacrificed for the other.
Protected health information is redacted at the edge, in memory, before the record reaches the AI extraction layer. The de-identified version exists only for the duration of the extraction process and is never stored.
PHI redacted in memory using edge computing, not stored on any server in de-identified form
AI operates only on the temporary de-identified version of the record
Extraction completes and the redacted version is discarded — no PHI exposure in the AI layer
HITL validation occurs against the original source (see Pathway B)
The original identifiable source document is stored in a 21 CFR Part 11 compliant vault, accessible only to authorized personnel (CRAs, auditors, monitors) under role-based access control.
Original records preserved intact, unmodified, in a 21 CFR Part 11 environment with full audit trail
CRAs and monitors access the original source for remote SDV without requiring a site visit
Bounding-box links in the Visual Audit Trail connect directly to the relevant passage in the vaulted original
Satisfies ICH GCP E6(R3) requirements for original source document access during oversight activities
Regulatory basis: ICH GCP E6(R3) Annex 1 requires that original, identifiable source documents remain accessible for audit and SDV purposes. Castor’s Pathway B ensures this obligation is met while Pathway A ensures AI extraction never touches PHI directly. The two pathways operate independently but are linked through the Visual Audit Trail, giving auditors a complete chain of custody from source document to EDC field.
Use our tools to manage your own study, or partner with our experts to do it for you. We take care of the technicalities so you can focus on the science.
Load records from EHR exports, CSV, or direct HL7 feeds. Castor maps incoming data to your eCRF structure with configurable field-level transformation rules.
Programmatic edit checks flag inconsistencies at entry. Queries route to the responsible party with a full response thread, tracked in the audit log.
Export in ODM, CDASH, and CDISC formats. Audit trails are captured at field level with timestamps and user attribution for PMCF, RWE, and regulatory submissions.
Lock at the patient, site, or study level. Freeze individual fields or entire forms. All lock actions are logged with reason and user identity.
For teams without dedicated DM capacity, Castor’s expert team handles your full data management program from DMP through database lock and submission package delivery.
Castor’s team writes and maintains your DMP, aligned to ICH E6(R3) and your study protocol, covering all abstraction, coding, and query processes.
Adverse events and concomitant medications are coded against current MedDRA and WHO Drug dictionaries by certified medical coders, with the full coding log delivered at lock.
Castor’s team manages the full query lifecycle from issue through resolution, with weekly status reports and a target query-to-resolution turnaround of under five business days.
Full lock process including final edit check sweep, audit trail review, define.xml preparation, and delivery of a CDISC-compliant submission-ready dataset.
Three programs that moved from manual abstraction to AI-validated extraction with Castor.
Regulators worldwide require ongoing post-market clinical data. EU MDR Article 61 mandates it for CE-marked devices. FDA post-market surveillance rules demand similar rigor. Spreadsheets and manual processes cannot produce audit-ready data or meet reporting timelines.
Annex XIV compliant
A top-5 global outcomes research firm ran a retrospective real-world evidence evaluation of a first-line oncology biologic with Castor, abstracting patient records to assess treatment patterns and outcomes in a real-world population outside clinical trial conditions. The study covered multiple countries with CDISC-ready output for the pharma sponsor.
Country RWE study
Oncology biologic
A cardiac AI company used Castor to structure echocardiogram data from clinical records for training a heart failure AI model. Castor’s abstraction workflow converted unstructured imaging metadata and clinical notes into labeled, validated training datasets, supporting the model’s subsequent FDA 510(k) clearance.
AI training dataAI training data
510(k) cleared AI
“The eCRF builder is the most configurable we’ve evaluated. Branching logic and edit checks are genuinely easy to set up, and the audit trail holds up under scrutiny.”
“We ran a 700-patient PMCF registry on Castor. The query workflow and multi-site access made coordinating across 12 hospitals manageable. Nothing else we evaluated came close.”
“Our data management team used Castor’s bulk import and CDISC export capabilities for a multi-country RWE study. Clean data, clean lock. That’s all we needed.”
Castor meets the rigorous standards of the FDA, EMA, and notified bodies. With a complete audit trail for every data point and every login, your data is always submission-ready and fully traceable.
Information security management system, certified annually
Business Associate Agreement available, US patient data compliant
EU data residency, DPA available, SCCs for international transfers
Electronic records and signatures, validated for FDA regulatory submissions
FDA 2024 Real-World Data Guidance alignment: Castor’s Visual Audit Trail and source document vault satisfy FDA’s 2024 guidance on traceability requirements for real-world data used in regulatory submissions. Every extracted value links to its source passage, and every human validation decision is timestamped and attributed — giving sponsors a complete, inspection-ready chain of custody from raw EMR data to final EDC entry.
Castor’s AI extraction achieves a 0% error rate after the human validation layer is applied. The AI extracts values and flags each field for human review. A trained reviewer approves, edits, or rejects every extracted data point before it enters the EDC. The Visual Audit Trail links each approved value to its source location in the original medical record, creating a fully traceable chain of custody. This combination of AI throughput and human sign-off is what makes the 0% figure defensible to a regulator or auditor, not just to an internal metrics dashboard.
After AI extraction, every field appears in a review queue alongside a bounding-box view of the source document. The reviewer sees the extracted value and the exact passage it came from, side by side. One click approves the extraction. If the value is wrong, the reviewer edits it directly and the correction is logged with a timestamp and user identity. The corrected or approved value is then written to the EDC. No data enters the EDC until a human has reviewed and taken action on it. The full review log, including the original AI suggestion and the human response, is retained in the audit trail.
Yes. Castor uses a dual-pathway approach. Pathway A uses a privacy-by-design approach during AI extraction: protected health information is processed in memory and never stored in de-identified form. Pathway B vaults the original identifiable source document in a 21 CFR Part 11 compliant environment, accessible only to authorized CRAs and auditors under role-based access control. The Visual Audit Trail links the EDC field back to the exact passage in the original vaulted document, so a CRA performing remote SDV can verify the source without traveling to site. This architecture satisfies ICH GCP E6(R3) requirements for original source document access while keeping PHI out of the AI processing layer.
The built-in tools give your team full control: bulk imports, edit checks, query workflows, and CDISC exports managed by your staff. The managed DM service means Castor’s own team handles your DMP, MedDRA coding, query resolution, and database lock. Many teams start self-managed and bring in the expert team for database lock and final dataset delivery. Both paths produce audit-ready data with a complete 21 CFR Part 11 trail.
