RECIST 1.1 in Practice: How Independent Imaging Review Protects Oncology Endpoints

RECIST 1.1, the Response Evaluation Criteria in Solid Tumors, is the standard for measuring tumor response in oncology trials, classifying outcomes as complete or partial response, stable disease or progressive disease. Because applying it involves judgment, blinded independent central review, often with double reads and adjudication, keeps response endpoints consistent and regulatory-grade.

RECIST 1.1 fits on a few pages. Measure the target lesions, add up the diameters, compare with baseline, apply the thresholds. Anyone reading it for the first time comes away thinking the hard part of oncology imaging must be somewhere else.

The hard part is that every step in that sentence contains a decision. Which lesions are measurable. Which five to follow. Where the lesion ends and the surrounding tissue begins. Whether the new finding on the liver is a metastasis or an artifact of a different reconstruction. The criteria are precise; their application is not automatic.

RECIST 1.1 Explained

RECIST 1.1 was published by Eisenhauer and colleagues in the European Journal of Cancer in 2009, revising the original 2000 criteria. It provides a standardized way to quantify tumor burden from imaging and to classify how that burden changes under treatment.

The framework separates lesions into two groups at baseline. Target lesions are measurable lesions selected for quantitative follow-up: a maximum of five in total, and no more than two per organ. Non-target lesions are everything else that is disease but is not being measured, followed qualitatively as present, absent or unequivocally progressing.

Measurability has thresholds. A non-nodal lesion is measurable at 10 mm or more in longest diameter on CT. Lymph nodes are measured on the short axis and are considered target-eligible at 15 mm or more; nodes between 10 and 15 mm are pathological but non-target; below 10 mm they are considered normal. Target lesion measurements are summed, using longest diameter for non-nodal lesions and short axis for nodes, to produce the sum of diameters that response is calculated from.

The four response categories (CR / PR / SD / PD)

Best overall response is then derived across the whole sequence of timepoints, subject to confirmation requirements where the protocol specifies them. It is a derived variable, which means an error at any single timepoint can propagate into the patient-level result.

Where RECIST Gets Hard in Practice

Five situations account for most disagreement between readers.

Target lesion selection. The criteria say to select the largest lesions that are reproducibly measurable. Reproducibility is a judgment, and two readers can build two defensible target sets from the same baseline scan. Once selected, that set is followed for the rest of the study, so a divergence at baseline propagates through every subsequent timepoint.

Measurement boundaries. Deciding where a lesion ends against adjacent tissue, atelectasis or an adjacent structure is where most millimeters are gained and lost. On a lesion close to a threshold, a few millimeters decide the response category.

The nadir reference for progression. Progression is measured against the smallest sum recorded during the study, not against baseline. Readers who anchor on baseline out of habit systematically under-call progression.

Non-target progression. Unequivocal progression of non-target disease can drive PD on its own, without any target lesion measurement. Unequivocal is doing considerable work in that sentence.

New lesions. A new lesion means progression regardless of what the target lesions are doing. Distinguishing a genuinely new lesion from one that was always there but visible only on a thinner reconstruction requires care, and it is exactly the kind of decision that varies by reader.

How Blinded Independent Central Review (BICR) Protects Endpoints

Blinded independent central review addresses this by routing every scan through the same reading process, performed by readers who are independent of the investigational sites and blinded to treatment assignment.

Read design is the part that matters most and gets discussed least. A single read is fast and carries one reader's judgment into the endpoint. A double read with adjudication has two readers work independently, and where their response assessments diverge, a third adjudicator selects between them. That design produces a result that does not depend on which reader happened to be assigned, which is the definition of reproducibility for this kind of endpoint.

Consistency also has to be maintained over time. Reader training against the specific criteria, prespecified qualification thresholds, and periodic blinded re-testing with predefined image sets during the study are what stop reader drift over an eighteen-month enrollment period.

The honest caveat belongs here. A meta-analysis by Amit and colleagues in the European Journal of Cancer in 2011 compared local evaluation with central review of progression across oncology trials and found a strong correlation between the two treatment effect estimates, with a correlation coefficient of 0.947, concluding that local evaluation showed little evidence of systematic bias in well-blinded randomized studies. BICR earns its cost where bias can be directional: open-label trials, single-arm studies with no comparator to cancel reader noise, response-rate endpoints supporting accelerated approval, and smaller studies where a few reclassified patients move the result.

RECIST vs iRECIST vs Other Criteria

RECIST 1.1 was designed for cytotoxic and targeted agents, where tumor shrinkage is the expected signal of activity. Immunotherapy broke that assumption. Some patients show apparent growth or new lesions before responding, a pattern known as pseudoprogression, which RECIST 1.1 classifies as progression and removes from treatment. iRECIST, published by the RECIST Working Group in The Lancet Oncology in 2017, addresses this with a two-step progression assessment.

Under iRECIST, an initial finding of progression is recorded as unconfirmed progression, iUPD. Confirmation requires a follow-up assessment four to eight weeks later showing further worsening, which becomes confirmed progression, iCPD. If the follow-up scan does not confirm, the patient continues on treatment and can still be recorded as responding. Clinically stable patients can therefore stay on therapy through an apparent progression rather than being censored out of it.

Other frameworks apply to disease settings where solid tumor measurement does not fit: Lugano and LYRIC in lymphoma, PERCIST for PET-based metabolic response, RANO and RANO-BM in neuro-oncology, PCWG3 in prostate cancer, mRECIST in hepatocellular carcinoma, Cheson and IMWG in hematological disease. Each carries its own reader training requirement, which is why matching reader specialty to the criteria in the protocol is a feasibility question rather than a preference.

Banook reads against RECIST 1.1, iRECIST, Lugano, LYRIC, PERCIST, mRECIST, Cheson, irRECIST, IMWG, CHOI, PCWG3, Macdonald, RANO, RANO-BM, INRC, WHO and tumor volumetrics, drawing on a panel of more than 500 board-certified specialists worldwide, each with at least ten years of experience, including oncologists, radiologists, neuroradiologists and nuclear medicine physicians. Reader consistency is maintained through statistical monitoring of performance, consensus sessions, retraining and phantom-based bias checks, and each study is governed by an imaging review charter setting out the independent central read procedure, reader management, reading methodology and schedule. The group has supported more than 350 imaging studies across oncology indications including lung, breast, colorectal, prostate, pancreatic, hepatocellular, gastric, ovarian, renal cell, melanoma, glioblastoma, head and neck, lymphoma and multiple myeloma.

The wider case for centralized imaging is covered in Why Central Imaging Has Become a Strategic Asset in Clinical Trial Execution, and multicenter endpoint quality in How Central Imaging Enhances Endpoint Quality in Multicenter Clinical Studies

👉 Discover Banook Central Imaging solutions or contact our team of imaging experts

Frequently Asked Questions

What is RECIST 1.1?

RECIST 1.1 is the Response Evaluation Criteria in Solid Tumors, version 1.1, published by Eisenhauer and colleagues in the European Journal of Cancer in 2009. It standardizes how tumor burden is measured on imaging and how change under treatment is classified, using up to five target lesions with no more than two per organ.

What are the RECIST response categories?

Four. Complete response, the disappearance of all target lesions with pathological nodes reduced below 10 mm on short axis. Partial response, at least a 30% decrease in the sum of diameters from baseline. Progressive disease, at least a 20% increase from the smallest recorded sum with an absolute increase of at least 5 mm, or unequivocal non-target progression, or a new lesion. Stable disease, which is neither.

What is BICR in oncology trials?

Blinded independent central review is the interpretation of trial imaging by readers who are independent of the sites and blinded to treatment assignment, applying the response criteria uniformly across the study. It commonly uses a double read with adjudication so that the result does not depend on which reader assessed the case.

RECIST vs iRECIST: what is the difference?

iRECIST adapts RECIST 1.1 for immunotherapy, where pseudoprogression can make a responding patient look like a progressing one. It introduces unconfirmed progression, iUPD, which must be confirmed on a scan four to eight weeks later to become confirmed progression, iCPD. Clinically stable patients can continue treatment through an unconfirmed progression.

Why use independent review for RECIST?

Because applying the criteria involves judgment at several points: target lesion selection, measurement boundaries, the nadir reference for progression, and what counts as a new or unequivocally progressing lesion. Independent blinded review applies one convention across the study. Its value is greatest in open-label and single-arm trials, and in response-rate endpoints, where reader bias cannot cancel out between arms.