RSNA: Dodging common RECIST pitfalls

CHICAGO—Response Evaluation Criteria In Solid Tumors (RECIST) was designed to improve assessment of tumor response to treatment, but extracting RECIST data accurately can present a challenge to those new to the process.

Richard G Abramson, MD, of the Vanderbilt-Ingram Cancer Center in Nashville, Tenn., and colleagues looked at the pitfalls in RECIST data extraction, and will be offering practical tips in a poster presentation at the annual meeting of the Radiological Society of North America (RSNA).

Before the show, Health Imaging reached out to Abramson to get a preview of the presentation. Check out his responses below:

Health Imaging: What are the most common errors with extracting RECIST data?

Richard Abramson: In our quality improvement review we discovered errors in four major categories: (1) selecting target lesions at baseline, (2) reassessing target lesions, (3) reassessing nontarget lesions, and (4) assessing for new lesions. The most common errors, committed mostly by readers fairly new to RECIST, were designating a lesion as a target lesion when it was not unequivocally a metastasis and premature assignment of "progressive disease" (PD) to a nontarget lesion on follow-up.

HI: How can researchers avoid some of the more common pitfalls and maintain accuracy in tumor measurement?

RA: Tried-and-true rules for maximizing accuracy and reproducibility when measuring tumors -- rules that apply both for research and for routine clinical practice -- include zooming in to make measurements, measuring on the same contrast phase at every follow-up, and trying to have the same person do the measuring each time. From our quality improvement review, we can also suggest the following pearls for extracting RECIST data: (1) if there is a reasonable possibility that a lesion on a baseline scan could be something other than a metastasis, then classify it as a nontarget lesion, (2) if a change in a nontarget lesion could reasonably be attributed to something other than disease progression, don't call it PD, and (3) display the current scan and the nadir (often the baseline) scan side-by-side to avoid missing slowly worsening disease.

HI: You’ve mentioned that RECIST cannot totally eliminate subjective interpretation. Why?

RA: RECIST was designed to make response assessment as objective and reproducible as possible, but the standardized, quantitative aspect of RECIST applies only to "measureable" disease.  As we all know, many cancer lesions are inherently "nonmeasureable" -- examples would include lymphangitic tumor spread in the lungs or leptomeningeal carcinomatosis or infiltrative liver metastases -- and for those lesions, RECIST defaults to qualitative assessment.  RECIST 1.1 has tried to impart some additional guidance for nonmeasureable disease, especially with regard to assessment of nontarget lesions at follow-up, but at the end of the day the reader may be forced into evaluating overall disease burden subjectively, especially when there are no lesions that qualify for target lesion designation.

Check out the full presentation, titled “Pitfalls in RECIST Data Extraction for Cancer Clinical Trials: Lessons Learned from a Quality Improvement Review of a Cancer Imaging Support Laboratory”, in person at RSNA (Sunday, Dec. 1, 12:30 – 1:00 p.m.).