Run an IVDR performance study end to end

An IVDR Annex XIII performance evaluation lives or dies on the quality of its data: paired measurements against a comparator, acceptance criteria applied consistently, specimens traceable from subject to result, and every value defensible when the notified body asks. Most EDCs treat all of this as customization. We treat it as the product.

Start from the protocol, not from a blank form

Upload the study protocol and the AI assistant drafts the eCRF, the visit schedule and the consent bindings for your data manager to review. What took a form-build vendor six weeks becomes a review meeting. Nothing is applied without a human decision, and every configuration change is versioned and audited.

Score performance while you capture

These are not abstract "edit checks" — they are the actual rules a quantitative self-test protocol needs, declared as configuration. From our own PT/INR study, as concrete defaults:

• Acceptance criteria at entry. Below the 2.0 INR switch point the pair must agree within ±0.4 absolute; from 2.0–4.5 within ±20% relative; up to 6.0 within ±25%; above, ±30% (the ISO 17593 banding). The pass/fail verdict is computed the moment both values exist and written back into the CRF — a failing pair is visible on day one, not at database lock.
• Repeatability rules. Two readings differing by ≥0.3 INR (either below 2.0) or ≥0.5 (both above) trigger an immediate, documented repeat-test prompt that the operator must acknowledge — the acknowledgement is stored and audited.
• Method comparison, live. Bias, Bland–Altman 95% limits of agreement, Theil–Sen robust regression and Pearson correlation for quantitative endpoints; PPA/NPA/OPA with Wilson 95% confidence intervals for qualitative ones — computed server-side on current data, exportable as CSV.
• Specimen chain of custody. Subject → specimen → result, with lifecycle status and site-scoped traceability.

Leave with data your statisticians can use immediately

Exports are standards-shaped from the start: CDISC SDTM-style domains (DM, SV, findings and events), ODM 1.3 XML carrying the full study metadata plus clinical data in one file, real Excel workbooks and CSV. Explicitly-missing values keep their coded reason (measurement failed / not applicable / not asked / unknown / not done), so "not done" never masquerades as "not entered". A readiness score — weighted 40% form completion, 30% SDV coverage, 30% signatures — summarizes the study file at any moment, with open queries listed beside it.

Because we run these studies ourselves, the platform ships with the parts an IVD study actually needs — and our team can design the study build and the statistical outputs with you.