CluePoints

Statistical monitoring and learned models power core CluePoints modules. SMART supports anomaly detection, IQD investigates discrepancies in trial listings, and IMC ranks dictionary code suggestions from clinical terms. These functions depend on data driven analysis rather than fixed reporting alone. The broader suite also carries issue management, site oversight, medical and safety review, audit records, and the workflow for acting on findings.

Staff confirm queries and coding decisions within that environment. The model therefore supplies an essential analytical engine inside a larger trial management and quality oversight product, with human review and operational records retaining value around the automated investigation.

IQD gives data managers supporting evidence before they confirm a discrepancy, edit the proposed query, and submit it to Veeva EDC. IMC similarly presents ranked suggestions that coding staff accept or reject before finalization. Each workflow records decisions for inspection. However, IQD also couples its greater than 99% recall claim with an assurance that all material issues are caught. That completeness assertion is stronger than the stated metric establishes.

Review of surfaced findings does not reveal an omission that never reached the team. The controls are concrete, while stopping thresholds and a defined process for discovering and recovering missed findings remain unspecified.

SMART is the named statistical engine underlying CluePoints' anomaly detection history. IMC uses deep learning trained on MedDRA, WHODrug, and verbatim terms from more than 100 clinical trials, generating ranked suggestions from historical coding patterns and dictionary structures. IQD uses large language models for cross-domain discrepancy investigation and refines its behavior through user interaction as studies change. Those descriptions identify approaches and some training inputs.

The production model families and suppliers behind IQD, current algorithm versions, full training population composition, and approval schedule for model replacements remain unspecified. Dictionary upversioning support is distinct from identifying a change to the model itself.

CluePoints identifies its own SMART engine and describes IMC's deep learning approach, but does not name the language model suppliers used by IQD. The privacy notice sets contractual obligations for subprocessors and requires customer consultation when a new processor can access personal data. It also mentions AI services used internally by employees, which is a different scope from identifying the product's model chain.

The hosting provider remains unnamed, despite a commitment that stored information stays encrypted and inaccessible to that provider. A complete named processor register, model routing map, and provider specific retention and training terms remain unspecified.

IQD claims at least 80% precision and more than 99% recall across at least 100 production studies, including co-development with GSK. IMC separately claims up to 99% MedDRA accuracy, 95% WHODrug accuracy, and about a 50% reduction in manual coding effort. Those figures describe current modules, but a dated test design, evaluated term or discrepancy counts, reference adjudication method, and uncertainty estimates remain unspecified.

The number of production studies does not supply those missing denominators. Earlier SMART research and the historical FDA collaboration concern different technology and do not establish performance of today's query detection or coding release.

Clinical trial information is pseudonymized, with the customer retaining the link between a patient identifier and an individual. CluePoints may also anonymize information by removing identifying characteristics and use deidentified or aggregated trial data to develop, test, and maintain its software. IQD adapts through user interaction, so the distinction between operational data use and product improvement matters.

Clinical and user information is retained according to business purposes and legal requirements rather than a fixed schedule. Specific training restrictions, deletion timing, and downstream model retention remain unclear, as do technical hallucination controls and procedures for investigating an AI safety event.

The October 2026 privacy notice identifies CluePoints as a processor for customer clinical trial data under GDPR. Clinical information can include patient identifiers, age, date of birth, ethnicity, and health data, with sponsors retaining responsibility for the processing. Customers are instructed not to upload direct identifiers such as names, social numbers, or biometric data, and CluePoints does not hold the key connecting a patient identifier to a name.

Subprocessors receive corresponding contractual data protection obligations, with customer consultation when they can access personal data. This provides substantive non-US service privacy coverage; a readable US BAA and its applicable product scope remain unspecified.

The trust center offers a SOC 2 Type II report covering security, availability, and confidentiality controls, along with independent penetration testing certificates and a disaster recovery report. Separate test certificates are listed for modules including IQD, IMC, CMP, MSR, and SPOT. Access requires business details and confidentiality acceptance; the SOC report additionally requires a signed access letter that is reviewed before release.

The current reporting period and full system scope therefore remain behind that process. ISO 27001 certification is still in progress. A general self-service heading does not make the SOC artifact directly retrievable without requesting access.

The software supports clinical research data oversight, medical coding, and records of trial quality decisions. IQD retains query actions in an environment described as consistent with 21 CFR Part 11, while IMC preserves suggestions, approvals, and rejections for inspection. These are electronic record and traceability functions rather than a claim that the software independently diagnoses or treats a patient.

CluePoints' 2016 FDA research agreement involved SMART and data driven identification of anomalous trial sites. That historical research relationship does not constitute device clearance for current IQD or IMC modules, and no current medical device authorization is claimed for their stated use.

Coding staff remain responsible for accepting or rejecting IMC suggestions, and IQD records the evidence, confirmation decision, and query action associated with a finding. Dictionary conventions and human validation remain authoritative in the coding workflow. These mechanisms govern trial work and preserve accountability for individual actions. They do not establish comparative model performance across demographic groups, therapeutic areas, regions, or difficult clinical terminology.

IQD's continuing adaptation through user interaction also lacks a specified model release approval process or acceptance thresholds. Historical SMART research is separate from a current governance framework for the newer agentic and deep learning modules.

IMC keeps the suggested code and the coding team's acceptance or rejection, while IQD records supporting evidence, confirmation, and the query sent to the study system. These histories give teams a route to inspect and challenge an individual output. The privacy notice separately supports correction of personal information and assistance to customers handling data rights requests; that does not create an output accuracy remedy for a coding or discrepancy error.

An output warranty, customer indemnity tied to model results, and a contractual obligation to correct affected trial decisions remain unspecified. The advertised performance percentages also lack the evaluation counts needed to interpret an error commitment.

IQD can submit editable query text directly to Veeva EDC after a data manager confirms a discrepancy. The resulting audit history follows detection, evidence review, confirmation, and query action. The wider monitoring platform also ingests clinical, trial management, and safety data, while IMC fits into existing coding workflows. Veeva establishes a named research data connection with an action written into that system. It does not establish bidirectional exchange with a patient EHR. A current named clinical record integration, its read and write permissions, and the production scope of any chart writeback remain unspecified.

CluePoints operates as hosted software for trial teams, with account access, secured servers, backup and restore processes, and encryption for stored data and communications. Its privacy terms allow transfers to US and UK affiliates and certain subcontractors, supported by contractual transfer arrangements. The hosting provider is described as holding encrypted data without access to its contents, but is not named in the product privacy notice.

A committed country for trial storage, regional model inference boundaries, customer tenancy design, and a private deployment option remain unspecified. Transfer mechanisms address handling obligations without establishing that all processing stays within one jurisdiction.

Sponsors and CROs can purchase a configurable suite covering central monitoring, site oversight, safety review, coding, and query detection. IQD offers a demonstration or a pilot designed around the customer's study environment, while IMC offers a product demonstration and discussion of the coding operation. Consulting and services sit alongside the software portfolio. This makes the purchasing route and possible module scope understandable before contact.

Subscription rates, charges per study or user, pilot fees, minimum commitments, and the cost of adding modules remain unspecified, so the available delivery options do not establish an upfront budget.

CluePoints serves large pharmaceutical companies, midsize sponsors, and CROs managing trials for other organizations. IQD targets data managers performing complex cross-domain listing review, while IMC supports coding of adverse events and concomitant medications. The suite can operate across studies and sponsor portfolios, with GSK named in IQD's development and at least 100 production studies included in its performance claim.

Those references establish trial use without identifying the therapeutic areas, phases, age groups, or regional populations represented in the evaluation. Comparable performance across specialties and the broader medical and safety review portfolio therefore remains unspecified.