Magnetic resonance assessment of sinusoidal obstruction syndrome after neoadjuvant chemotherapy for colorectal liver metastases is not reproducible

Background Sinusoidal obstruction syndrome (SOS) due to chemotherapy can cause severe hepatotoxicity, leading to impaired outcome in patients with colorectal cancer. A previous study introduced gadoxetic acid-enhanced magnetic resonance imaging (Gd-EOB-MRI) to diagnose SOS. Purpose To assess the reproducibility of Gd-EOB-MRI-based SOS diagnosis and its relationship with response to chemotherapy and long-term outcome. Material and Methods Twenty-six Gd-EOB-MRI scans of patients undergoing chemotherapy for colorectal liver metastases (CRLM) were retrospectively analyzed. Three radiologists, blinded to clinical data, independently scored presence and severity of SOS on a 5-point scale (0, definitely not present to 4, definitely present). Patients with a score >= 3 were considered SOS+. Inter-observer agreement between readers was assessed with kappa statistics. Response (RECIST 1.1.), occurrence of new CRLM during follow-up (hepatic progression) and overall survival (OS) were compared between patients with and without SOS. Results The inter-observer agreement of SOS scores was poor, with quadratic kappas of 0.17-0.40. For the binary outcome of SOS+ (confidence level [CL] 3-4) vs. SOS- (CL 0-2) agreement was poor, with kappas of 0.03-0.37. Median follow-up was 24 months (range 4-44 months). Response and OS between patients with and without SOS did not differ significantly for any of the readers. Conclusion Inter-observer agreement for the diagnosis of SOS on Gd-EOB-MRI is poor. No significant correlation with relevant outcomes was found for any of the readers. Therefore, MRI for SOS diagnosis might be less useful than previously reported. Other techniques should be explored to accurately diagnose SOS in absence of histological confirmation.