Gadolinium-based contrast media does not improve the staging of neuroblastoma image-defined risk factors at diagnosis.

Document Type


Publication Date



DOI: 10.1002/pbc.30724


BACKGROUND: Neuroblastoma risk stratification relies on prognostic risk factors and image-defined risk factors (IDRFs). Evaluating neuroblastoma typically involves magnetic resonance imaging (MRI) with gadolinium-based contrast media (GBCM, "contrast"). However, there are concerns regarding adverse effects and cost of GBCM. We aimed to assess the impact of intravenous GBCM on interobserver agreement for neuroblastoma staging on baseline MRI.

PROCEDURE: We reviewed baseline MRI scans of 50 children with abdominopelvic neuroblastomas confirmed by histopathology. Duplicate sets of images were created, with post-contrast T1-weighted sequences removed from one set. Four pediatric radiologists independently analyzed the scans in a randomized manner. They recorded primary tumor size, presence of IDRFs, and metastatic lesions. Agreement among the reviewers was measured using kappa and Fleiss kappa statistics.

RESULTS: Mean age of included children was 3.3 years (range: 0.01-14.9 years), and 20 [40%] were females. Mean tumor size was 5.7 cm in greatest axial diameter. Pre-contrast versus post-contrast MRI showed excellent agreement for tumor measurement. Overlapping confidence intervals (CIs) were seen in nearly all categories of interobserver agreement on the presence or absence of individual IDRFs, with agreement ranging from poor to substantial, regardless of the presence of contrast. The overall interobserver agreement on the presence of at least one IDRF was substantial with contrast (kappa = .63; 95% CI: .52-.75) and moderate without contrast (kappa = .5; 95% CI: .39-.61); although the overlapping CIs suggest a lack of meaningful difference. Similarly, interobserver agreement on the presence or absence of individual sites of metastatic disease ranged from poor to substantial. The interobserver agreement on the overall determination of presence of metastatic disease was fair with contrast (kappa = .49; 95% CI: .38-.61) and moderate without contrast (kappa = .71; 95% CI: .59-.826).

CONCLUSIONS: Contrast does not improve tumor size measurement or radiologist agreement on the presence or absence of IDRFs or metastatic disease in children with newly diagnosed neuroblastoma.

Journal Title

Pediatric blood & cancer





First Page


Last Page


MeSH Keywords

Child; Female; Humans; Infant, Newborn; Infant; Child, Preschool; Adolescent; Male; Contrast Media; Gadolinium; Neuroblastoma; Magnetic Resonance Imaging; Risk Factors; Retrospective Studies


neuroblastoma; radiology; surgery

Library Record