Electronic RECIST workflow: named readers, second reads, and one standard operating procedure (SOP) for the audit trail

By Julio G. Martinez-Clark, Chief Executive Officer (CEO) of Amavita Research Services

At Amavita Research Services, we operate a Miami cardiovascular research site. I have operated a research site for 7 years, with zero Food and Drug Administration (FDA) Form 483 findings — the form the FDA issues when an inspection finds conditions that may violate its requirements. That record comes from making ownership explicit before the first scan, so a monitor or inspector can follow any measurement from image to filed result without asking who was responsible that day.

The inconsistency is usually a handoff, not a measurement

RECIST assesses whether a solid tumor has responded, stayed stable, or progressed, based on changes in selected target lesions over time. It depends on consistency: the same lesions, measured the same way, at every visit.

Inconsistency enters at the handoffs. One reader selects target lesions at baseline and another, without a documented handover, measures the follow-up. A borderline lesion is measured once, by whoever is available, with no record of why it was included. A correction is made in a spreadsheet or message thread and never reaches the system of record, so the audit trail shows the first value and the discussion lives elsewhere.

Moving to an electronic form solves none of that. Without a named owner it just produces a faster, better timestamped version of the same ambiguity. Decide who owns the read before you configure fields.

Name one trained reader per study

Assign one named reader per study, recorded in the SOP and delegation record before enrollment. Named means one accountable person whose training and approach set the standard — not the only person who can ever read.

That reader makes the baseline decisions everything else depends on: which lesions are target lesions, which are non-target, and how each will be measured. Document those choices in the workflow when they are made, not at the first follow-up visit.

Training must be specific. Before the first read, the named reader should complete the sponsor's protocol-specific training, review the imaging charter or manual, and complete a practice read if one is provided. File the record: if an inspector asks who was qualified to make the baseline selection, the answer should be a name and a dated record, not a department.

Require a second read for borderline lesions

A second read for every scan adds little for clear cases. A second read for no scan is how borderline errors become response errors. The SOP should define in advance which cases trigger an independent second read by a trained reader who has not seen the first measurements.

Define triggers such as a change close to a response or progression threshold, a lesion hard to measure because of shape, location, or image quality, a new lesion that would change the assessment, and any measurement that does not match the visual impression of the scan. Write them down before a real case tests them, so a second read is procedural, not personal.

Define what happens when the two reads disagree: readers confer and record a consensus, a third reader adjudicates, or the case is escalated to the sponsor or central reviewer, per the protocol. The resolution — both measurements, who conferred, what was decided, and why — belongs in the audit trail, not just the final number.

Write one SOP for capture, measurement, and filing

One SOP should cover the full path from image receipt to filed result. At a minimum, describe how images are received and checked, how lesions are selected at baseline, how each lesion is measured, how measurements are entered into the electronic system and into electronic data capture (EDC), the sponsor's database for trial data, if required, and how corrections are made.

Corrections are where audit trails most often break, so give them their own rule: never overwrite. Keep the original visible, enter the correction with reason, date, and the name of the person making it, and file any supporting document with it. If the system cannot preserve that history, know it before the study starts, not during monitoring.

Name the filing standard too: where the assessment lives, how images and measurement records link to it, how versions are named, and how long records are retained. A reader following the SOP should produce a record another trained person can reconstruct without interviewing them.

Train backups before you need them

A named reader without a trained backup is a single point of failure. Illness or departure then forces an untrained substitution at the moment consistency matters most.

Train at least one backup to the same standard, file the same evidence, and have the backup complete practice reads before a live assessment. The SOP should state when the backup may read and require a documented handover — baseline selections and prior adjudications — so the backup continues the same approach instead of starting a new one mid-study.

If a reader changes, record the change, date, reason, and handover. Sponsors and inspectors do not object to change — they object to change they have to discover.

Make the audit trail the record, not a byproduct

A complete audit trail lets an independent reviewer answer five questions without asking the site: who selected the lesions, who measured them, when each entry and change was made, what the prior value was, and which SOP version was in effect.

Test it in your own quality review: trace a completed assessment end to end, from image receipt to filing. If any step can only be explained by the person who did it, the workflow has a gap, even if the number is correct. Close it in the SOP or system configuration, and train the change. Named responsibility, written procedure, trained backup, and a record that stands on its own — the software is the last part of that chain, not the first.

What to settle before the next imaging study starts

Before the first scan, settle five items in writing: name the primary reader and a trained backup and file their training; document baseline lesion selection; define second-read triggers and the adjudication route; confirm the SOP covers capture, measurement, entry, correction, and filing and matches the workflow field for field; and test the audit trail on a practice case.

Do that once, up front, and assessments stop depending on memory and availability. Each read follows the same standard, borderline cases get a second pair of trained eyes by rule, and the audit trail tells the story on its own.