Table 1 Definition of the different risk factors likely to triggering adjudications. List of terms used as predefined risk factors and their corresponding definitions

1.Different scans at baseline

LI and BICR selected scans with different SeriesUID as DICOM Tag.

(e.g. same imaging modality but with different reconstruction parameters)

2. Different scan(s) at follow up

LI and BICR selected at least one different scan with different SeriesUID as DICOM Tag during patient follow up. Their measures were extracted from different scans at at least one time point after baseline.

3. Different number of target lesions

Readers selected a different total number of target lesions in the tumor burden.

4. New lesions

If any reader reported one or more lesion(s) that was/were not recorded on the previous time point.

5. Different target Lesions

LI and BICR each selected a tumor burden that does not contain exactly the same lesions. Note: Both readers may have selected the same number of target lesions, however not the same ones.

6. Non-measurable lesions

LI and BICR measured the same Target Lesion but did not have the same perception of lesion boundaries (due to its complexity). We adopted part of RECIST 1.1 definition for non-measurable lesions [24] as lesions likely to lead to non-reproducible measurements. The labelling of non-measurable lesions is done on baseline data only.

7. Progressive non-target lesion

A patient is classified as having progressive disease by non-target lesions when either LI, BICR or both declared progression based on non-target lesions. (According to RECIST, unequivocal progression of non-target lesions has a very specific definition which is supposed to be an “extremely rare” event.)