Back to Insights

Most imaging attention in early-phase oncology goes to the response call. The call with the most immediate consequence for a patient is progression, because it is the one that ends treatment.

Three Routes, and Only One Is the Sum

Under RECIST 1.1 a patient progresses in one of three ways. The sum of target lesion diameters rises at least 20% above the smallest sum recorded, with an absolute increase of at least 5 mm. Non-target disease progresses unequivocally. Or a new lesion appears.

Most of the conversation about measurement focuses on the first. In practice the other two are where judgment concentrates. Whether non-target growth is unequivocal is a reader's call, and whether a small finding on a follow-up scan is a new lesion is often a reader's call too. RECIST anticipates this: an equivocal new lesion is followed, and progression is declared only once it is confirmed.

Why It Is the Call Most Exposed to Error

Measurement variability between careful readers is a known property of RECIST, and the 20% line is narrower than the 30% line. A borderline patient can cross it on measurement noise. Add the judgment calls on non-target disease and new lesions, and progression becomes the assessment where two qualified readers are most likely to disagree.

The cost of disagreement is asymmetric in time. A disputed response can be adjudicated later without harming anyone. A disputed progression has usually already changed the patient's treatment by the time anyone disputes it.

Under Immunotherapy, the First Progression Is Often Not the Last Word

Under immunotherapy, tumors can enlarge and new lesions can appear before disease shrinks or resolves. Patients are told they progressed and come off therapy, and sites are not consistently trained to recognize the pattern.

iRECIST was written for this. It treats a first progression as unconfirmed — iUPD — and asks for confirmation at the next assessment, four to eight weeks later, before it becomes confirmed progression, iCPD. New lesions are recorded and followed rather than ending the assessment.

A confirmation rule only protects a patient if it is applied before the treatment decision. In most trials, the decision is made at the site, on the local read.

Who Makes the Call, and When

The investigator at the site sees each scan first and decides whether a patient continues. Central review, where a trial has it, is frequently organized around the endpoint rather than the treatment decision, so it can arrive after the patient has already come off study.

That sequence is reasonable for an endpoint. It is not designed to protect a patient from an unconfirmed call. An independent read before treatment stops is a different service with a different clock.

What an Independent Progression Read Needs

A read fast enough to matter, against a turnaround the sponsor and the sites agree before the first patient enrolls.
The full history: side-by-side comparison against prior time points, by the same reader who has followed that subject throughout, so today's scan is compared on the same basis as the last one.
The charter's rules, written in advance: which progression routes the independent read covers, how iUPD and iCPD are handled, and what happens to a new lesion that is equivocal.
A defined path when the local read and the independent read disagree — who decides, how quickly, and what the site is told.
And a record. Every measurement event, the reader's identity and qualifications, and the reason for any change, in a 21 CFR Part 11 audit trail an inspector can follow.

What It Costs to Get It Wrong, in Both Directions

A false progression takes a patient off a therapy that may have been working and puts an event in the progression-free survival curve that should not be there. In a small study, one or two of those can change how the data reads.

A missed progression keeps a patient on a failing therapy longer than necessary and delays the next decision. Neither error is rare enough to ignore, and neither is visible in a response table.

Five Questions to Put to Whoever Reads Your Study

  1. Who confirms progression before a patient comes off treatment — the site alone, or an independent reader?
  2. How quickly does an independent progression read come back, and is that written into the agreement?
  3. Is the same reader assessing a subject at every time point?
  4. How does the charter handle iUPD, iCPD and equivocal new lesions?
  5. When the local and independent reads disagree, who decides, and how fast?

GenPhase.ai is an imaging core lab for oncology sponsors, built on ONIX AI. Sponsors start with imaging eligibility confirmation, confirmation of progression or recurrence before treatment stops, an exploratory endpoint read, or an independent review of scans from a completed study. Send us the protocol and the enrollment timeline. Our clinical team typically responds within one business day, and you will have a scoped plan and pricing within five business days.

Sources: Eisenhauer EA, et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 2009;45:228–247. Seymour L, et al. iRECIST: guidelines for response criteria for use in trials testing immunotherapeutics. Lancet Oncol 2017;18:e143–e152.