In Reply We thank Karajannis et al for their comments on our Original Investigation1 and bringing up what is the quintessential question for many of us as both clinical researchers and practicing pediatric oncologists: What evidence provides us with justification to change standard practice?
We are in complete agreement that exploratory, unplanned, or hypothesis-generating subgroup analyses are generally insufficient and should be confirmed, ideally in the setting of a prospective randomized clinical trial. However, it is incorrect that the analysis of carboplatin efficacy within medulloblastoma groups in the Children’s Oncology Group ACNS0332 trial1 was either exploratory or hypothesis-generating. At the time of initial study planning nearly 20 years ago, the molecular groups of medulloblastoma were not understood. In 2011, Northcott and colleagues2 described 4 distinct molecular variants of medulloblastoma and, in collaboration with 3 independent groups who observed similar findings, established an international consensus describing the 4 molecular subgroups of medulloblastoma,3 which has been widely shared and adopted by the World Health Organization in 2016.4 According to the 2012 international consensus, group 3 medulloblastoma was identified as having a terrible prognosis,3 thus generating the hypothesis that this was the group most in need of better therapy.