Automated CT volumetry of pulmonary metastases: the effect of a reduced growth threshold and target lesion number on the reliability of therapy response assessment using RECIST criteria.

The purpose of this study was to evaluate the reproducibility of CT-volumetric tumour response assessment of pulmonary metastasis using variable volume change thresholds (VCT) and target lesions with response evaluation criteria in solid tumours (RECIST). Fifty consecutive patients with pulmonary metastases undergoing follow-up multislice CT under chemotherapy were assessed for response to chemotherapy with modifications to RECIST: (1) decreasing the percentual VCT for diagnosis of tumour response (range = 70%-20%), (2) reducing the number of target lesions (range = 1-5). Continuous and categorical observer agreements were tested by Bland and Altman and extended (kappa(e)) or non-weighted kappa (kappa) and correlated with percentual VCT to predict observer agreement. A total of 202 metastases were evaluated (average volume = 522.4 mm(3) +/- 902.4 mm(3)). General agreement on treatment response was very high (kappa(e) = 0.93-1), but was reduced with VCT < 35% (kappa(e) < 0.95). Kappa correlation with VCT values was strong (r=0.94-0.96; p< or =0.0002). Average confidence decreased significantly at VCT < 45% (p < 0.01) and agreement on stable disease at VCT < 35% (kappa(e) < 0.95; p < 0.01). Reduction of target lesions (n < 3; VCT = 35%) resulted in decreased reader confidence (for n = 1: kappa = 0.49; p < 0.05). Agreement for evaluation of treatment response was robust using VCT > or =35% and > or =3 metastases. This may translate into shortening of follow-up intervals or enable for response assessment with tumours displaying minimal volume change.