Soon Ho Yoon MD, Presenter: Nothing to Disclose

Kyung Won Kim MD, Abstract Co-Author: Nothing to Disclose

Jin Mo Goo MD, PhD, Abstract Co-Author: Research Grant, Guerbet SA

Dong Wan Kim MD, PhD, Abstract Co-Author: Nothing to Disclose

Seokyung Hahn PhD, Abstract Co-Author: Nothing to Disclose

Figuring out intra- and inter-observer reproducibility in manual measurement of tumor burden on CT is essential for the interpreting treatment response in clinical practice as well as in clinical trials. We aimed to systematically review and meta-analyze intra- and inter-observer reproducibility in manual measurement of tumor burden according to the response evaluation criteria in solid tumors (RECIST) guideline on computed tomography (CT).

Two authors independently performed a literature search of the MEDLINE and EMBASE databases by using the search terms tumor, measurement, reproducibility and CT. We included studies in which observer reproducibility in manual measurement of target lesion was evaluated by experienced radiologists on CT in oncologic patients, with extractable outcomes including summary statistics on the percentage of relative difference between the measurements, intra-class correlation coefficient, and concordance correlation coefficient. The primary outcome was 95% limits of agreement (LOA) of relative measurement difference in unidimensional measurement of target lesion within observer and between observers derived from the Bland-Altman plot. Outcomes were pooled using a random-effect model.

Pooled 95% LOA for unidimensional measurement of single target lesion were (-18.0%, 16.2%) within observer and (-22.4%, 25.7%) between observers. Pooled 95% LOA for unidimensional measurements of the sum of multiple target lesions were (-9.8%, 13.1%) within observer and (-19.3, 19.5%) between observers. The reported intra-class correlation coefficients for the measurement of single lesion ranged from 0.94 to 0.99 within observer and from 0.79 to 0.99 between observers. The reported concordance correlation coefficients for single lesion ranged from 0.97 to 0.99 both within observer and between observers. There was a paucity of researches on observer reproducibility in measuring the interval change of tumor burden.

Pooled 95% LOA in measurement of single target lesion were over 20% between observers. Measurement reproducibility increased when single observer measured tumor burden and when tumor burden was assessed by the sum of multiple lesions.

Caution should be taken in measuring tumor burden on CT at specific time point, especially when different observers assess tumor burden on previous or follow-up CT scan.

Yoon, S, Kim, K, Goo, J, Kim, D, Hahn, S, Observer Reproducibility of Manual Measurement of Tumor Burden in RECIST Guideline: A Meta-analysis.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14008753.html