Aim: Assessment of magnetic resonance imaging (MRI) in glioblastoma can be challenging. For patients with recurrent glioblastoma managed on the CABARET trial, we compared disease status assessed at hospitals and subsequent blinded central expert radiological review.
Methods: MRI results and clinical status at specified time points were used for site and central assessment of disease status. Clinical status was determined by the site. Response Assessment in Neuro-Oncology (RANO) criteria were used for both assessments. Site and central assessments of progression-free survival (PFS) and response rates were compared. Inter-rater variability for central review progression dates was assessed.
Results: Central review resulted in shorter PFS in 45% of 89 evaluable patients (n = 40). Median PFS was 3.6 (central) versus 3.9 months (site) (hazard ratio 1.5, 95% confidence interval 1.3-1.8, P < 0.001). Responses were documented more frequently by sites (n = 16, 18%) than centrally (n = 11, 12%). Seven of 120 patients continued on trial without site-determined progression for more than 6 months beyond the central review determination of progression. Of scans reviewed by all three central reviewers, 33% were fully concordant for progression date.
Conclusion: While the difference between site and central PFS dates was statistically significant, the 0.3-month median difference is small. The variability within central review is consistent with previous studies, highlighting the challenges in MRI interpretation in this context. A small proportion of patients benefited from treatment well beyond the centrally determined progression date, reinforcing that clinical status together with radiology results are important determinants of whether a therapy is effective for an individual.
Keywords: bevacizumab; carboplatin; clinical trial; glioblastoma; magnetic resonance imaging.
© 2017 John Wiley & Sons Australia, Ltd.