Why CMR protocol consistency is the hardest multi-site problem
CMR protocol consistency across multi-site trial imaging is the hardest unpaid line item: site-level drift accumulates before the first read, and core-lab read precision only holds when intake is cohort-tagged from the start.
Why site-level drift accumulates
Every site runs a slightly different stack. Different scanner vendor, different software build, different SOP version, different technologist comfort with free-breathing cines. Field-strength variations, local anesthesia protocols, and PACS tag schemas that do not reconcile compound the drift. By the time a study lands at the core-lab, the protocol tag in the DICOM header no longer reliably matches the trial's intended acquisition parameters — even when the technologist at acquisition believed they were running the right protocol.
What core-labs actually see
A non-trivial share of multi-site studies arrive with protocol tags the core-lab cannot reconcile. The reader derives the intended sequence from the cine stack, the metadata, and a site note — reviewer time spent on what the trial sponsor assumed intake had locked down. When tag drift is large enough, the reader has to escalate rather than sign. Read precision is not the bottleneck. Intake integrity is.
Where a shared QA rubric helps
A rubric every site runs at intake — same motion-blur grade, same signal-homogeneity criterion, same gating failure categorization — collapses the language the core-lab hears. When a borderline cine is parked, the reason in the park queue is the same wording at site A as at site D. The reader stops reconciling terminology and starts reading.
How cohort-tagged intake collapses read variance
Slotting studies into cohort at intake — by protocol, by field strength, by acquisition date — gives the reader a comparison set. Drift within a cohort becomes visible; drift between cohorts becomes a structural problem the trial sponsor can address. Without cohort tags at intake, every cross-site variance is suspected of being the same finding artefact.
What a pilot looks like
Two sites, one shared protocol, one rubric, one intake path. The metrics the pilot reports on are read variance across sites, after-hours page rate per site, and time-to-intake median. The pilot collapses the language the core-lab hears and surfaces which site-level acquisition practices still drift. It is the cheapest test of whether the protocol will hold at five sites or fifty.
Related reading: The 7T CMR workflow, end to end and The overnight radiologist burden no CMR program budgets for.
Ready to standardize a multi-site CMR protocol for an upcoming trial? Tell us what you are putting together.