C the Signs

Artificial intelligence is claimed throughout, and no learned component is described anywhere the company publishes. The home page and What We Do page say the platform uses AI to detect cancer early, scan whole populations and predict tumor origin with 94 percent accuracy, and the research page refers to an AI model flagging cancers years before diagnosis in Mayo Clinic records.

What the same pages describe is a curated knowledge base: more than 1,000 validated signs, symptoms and risk factors assessed against national and local guidelines, with recommendations drawn from those guidelines, and the privacy policy refers to "our algorithms" for population risk identification and questionnaire triage.

No page, publication or abstract the company lists describes a model whose parameters are estimated from data, a training set or out of sample validation of a learned model; functions of this kind can be performed by weighted rules. None of this says whether the product works, which is a question for the evidence.

How oversight works is published for each route, with one part missing. In the consultation tool the GP receives a risk assessment and recommended pathway and makes and owns the referral. In the patient services, the privacy policy states that questionnaire triage to the GP, to secondary care or to advice only is an automated decision with no human oversight, with a right to human review on request, and the What We Do page says patients can be booked straight into diagnostic tests without seeing a doctor. What is not published is the threshold at which a patient triage result goes to a clinician before action, or how missed cases are monitored across practices, and that is the missing part.

The home and What We Do pages say "C the Signs identifies 99 out of every 100 patients with cancer - helping ensure no one is missed when every moment counts". The sentence concedes misses in its own figure, so it reads as the outcome the product aims to support rather than a promise that it never misses. The 99 percent figure itself is a performance claim. Other lines make the clinician the subject ("helping clinicians never miss the one case that keeps them up at night").

The approach is described in general terms with nothing identified. The company describes more than 1,000 validated signs, symptoms and risk factors assessed against national and local guidelines, reading of structured and unstructured record data, tumor specific and pan-cancer models for population risk identification, and tumor origin prediction, and says it maps to SNOMED CT, ICD and CPT coding.

No architecture, method, training or validation description, version history or update practice is published, and the UKCA Class I declaration is self-certified, so no reviewed technical file is public either. Guideline based recommendations change when national guidance changes, and nothing says how or when the knowledge base is updated or how customers are told.

The model layer is the company's own, and the rest of the chain is described without names. The privacy policy describes risk identification and triage run by "our algorithms", with no outside model provider mentioned, and says data is held in a UK database under the company's control. It also says the company uses external suppliers for CRM and analytics, some outside the United Kingdom and EEA, and names none of them; no subprocessor list is published. Because the tool reads the full GP record and writes coded entries back into it across whole counties, the parties that can reach that data matter.

Named deployments with dated outcome figures, published mostly as conference abstracts by company authors. The research page reports an ESMO 2025 evaluation across GP practices in Somerset ICB (a 36 percent improvement in stage shift, a 10 percent cancer conversion rate from referrals against 6 percent nationally, and early stage diagnoses via urgent pathways rising from about 60 to 80 percent); a JCO Oncology Practice abstract finding detection rates rising from 58.7 to 66.0 percent without more referrals; an ASCO 2025 health economic assessment of an evaluation in Suffolk and North East Essex commissioned by Health Innovation East (808 percent return on investment, 272 patients diagnosed earlier, 1.5 million pounds a year); Mayo Clinic record analyses (26 percent of five cancers flagged earlier than physician diagnosis); an endometrial self-assessment pathway (time to specialist review from about two months to five days); and an ovarian cancer real world study in the Journal of Cancer Policy.

The designs are before and after or retrospective, company staff are authors throughout, and most are abstracts without full methods. The 99 percent sensitivity headline counts only patients the tool assessed and publishes no denominator.

Retention and de-identification are stated; use for model development is not. The privacy policy says the company acts as processor for GP practices and as controller for population risk identification, trial eligibility and questionnaire triage, lists the patient data it holds as controller (identifiers, NHS number, contact details and all health information in the GP record), and says patient information is kept for the duration of the contract with the practice and then anonymized; user data is kept up to seven years after an account ends, website inquiries 24 months and technical data four years.

Anonymized data is used for statistics and publication. Nothing says whether record data read by the tool, or the coded entries it writes back, is used to build or tune its algorithms. Nothing is published on safety engineering, such as how a wrong recommendation or triage decision is detected and handled as a safety event.

The United States rule is invoked, and the business associate position is not stated. The What We Do page lists HIPAA and HITECH compliance alongside UK Data Protection Act and GDPR compliance and NHS Data Security and Protection Toolkit compliance. Because the company invokes the US rule, the grade reads the US posture: no business associate status, agreement or template is published.

The UK documentation is substantive and is recorded as context: the privacy policy sets out the company's role as processor for GP practices and as controller for population risk identification, trial eligibility and questionnaire triage, the data held, lawful bases, retention and international transfers.

Recognized certifications are named in the company's own material without the artifacts. The What We Do page lists Cyber Essentials Plus certification, ISO 27001 and ISO 20000 certification, NHS Data Security and Protection Toolkit compliance, NHS Digital Technology Assessment Criteria passed, HSCN approved user status and IM1 approved supplier status.

No certificate, certifying body, scope, date or trust center is published, so the scope and currency of the ISO certificate are for the buyer to confirm. The privacy policy describes UK hosted servers and need to know staff access in general terms.

A named non-US regime with a class stated. The What We Do page lists the product as a UKCA Class I medical device, alongside compliance with NHS clinical risk management standards DCB0129 and DCB0160 and the NHS Digital Technology Assessment Criteria. Class I is self-declared by the manufacturer, so no body has reviewed the product's clinical performance. The positioning question is scope.

The consultation tool supports a GP's referral decision, but the patient services triage people to secondary care or advice by automated decision with no human oversight (privacy policy) and can book patients directly into diagnostic tests without a doctor (What We Do page), and the page claims tumor origin prediction at 94 percent accuracy. No page says whether those uses sit inside the declared Class I intended purpose. No FDA position is stated; the same page says the governance is designed for both the UK and US, and the site's terms say it is directed at UK residents.

Evaluation within the group the product is most at risk of failing is published, in abstract form. The company's research page lists a disparities analysis of early cancer detection in areas of high deprivation (4th NCI Symposium on Cancer Health Disparities) and an AACR 2025 abstract reporting reduced diagnostic burden with improved detection and equity, and the What We Do page describes the system as designed for all genders, ages and backgrounds with multilingual patient resources and self-referral by SMS, phone or email.

Deprivation matters because late presentation and fewer consultations are the route by which a consultation based tool misses people. The work is in conference abstracts rather than full papers, no performance figures are published compared across ethnic, age, sex or deprivation groups, and no bias testing method or monitoring for drift as guidelines change is described. The company names NHS clinical risk management (DCB0129 and DCB0160) compliance, which governs clinical safety process rather than model behavior.

A correction route the affected patient can exercise against the company, covering the automated decision itself. The privacy policy says that when a patient identified by the company's algorithms completes a questionnaire, the company triages them (referral back to the GP, to secondary care, or advice only) by automated decision with "no human oversight", and that the patient has the right to obtain human intervention on that decision, to express their point of view and to contest it, by contacting the company.

The company is the controller for this processing, so the route runs to it rather than through the GP practice. The route covers the patient triage services; in the consultation tool the GP makes and owns the referral decision. No warranty, indemnity or remediation commitment attaches to a recommendation, and no error rate is published with a denominator (the 99 percent sensitivity figure gives none).

Named bidirectional integration with the major English GP record systems, running in production across whole integrated care boards. The What We Do page states the platform is fully integrated with EMIS, SystmOne and Vision, is an NHS Digital IM1 approved supplier (IM1 is the NHS interface mechanism whose approved supplier list is public), reads structured and unstructured record data, and writes back to the medical record as coded entries, with FHIR and HL7 support named.

The news page reports rollout to every GP practice in Northamptonshire (December 2025), across Suffolk and North East Essex and into Norfolk and Waveney, and the ESMO 2025 evaluation covers practices across Somerset ICB. Because assessments and safety netting are coded into the record, the output is visible to later clinicians; whether a coded entry is marked as generated by the system is not stated. Outside NHS primary care no integration is named.

A hosted service with United Kingdom residency committed and the cross-border path named in part. The privacy policy says all information held within the business is stored in a database under the company's control in the United Kingdom, on secure servers in the United Kingdom, and that some suppliers, including the newsletter provider and some data analytics providers, are outside the United Kingdom and EEA, with standard contractual clauses or Data Privacy Framework membership as safeguards.

The service reaches clinicians through the GP record systems and a web application, and the company names HSCN approved user status. Tenant isolation is not described, and the suppliers outside the United Kingdom are not named, so which data crosses to them is left open.

No price is published, and how the service is bought is discoverable. The terms of use say the services are free of charge to individual health practitioners, with payment made by NHS funded bodies, including GP practices, or other organizations, and the news page shows purchase by integrated care boards and cancer alliances across whole counties.

A health economic assessment presented at ASCO 2025, from an evaluation commissioned by Health Innovation East, reports an 808 percent return on investment and 1.5 million pounds in annual savings; that is a value claim, not the price, and a commissioner still cannot size the contract before asking.

Coverage is named across three routes, with outcome evidence behind the consultation tool and part of the rest. Sold for early cancer detection, the product assesses risk across more than 100 cancer types in NHS general practice consultations, offers patient self-assessment with direct booking to tests in some pathways, and scans practice populations to invite people at higher risk.

Evidence covers general practice in several integrated care boards, an endometrial self-assessment pathway and retrospective analyses of Mayo Clinic records in the United States; the population scanning route and the tumor origin prediction carry figures without published method. Every live deployment named is in the NHS, built around NICE and local guidelines and English record systems.