Reply to Freyschlag et al

We appreciate the commentary of Freyschlag et al regarding our manuscript identifying a clinical neurocognitive phenotype varying on the basis of the mutation status of the isocitrate dehydrogenase (IDH) 1 gene.¹ We agree that identifying and differentiating the specific factor(s) (eg, tumor, edema) that cause neurocognitive dysfunction is challenging, just as identifying the tumor margin from edema can be challenging radiographically. The thoughtful commentary of Freyschlag et al elaborates on several interesting issues in studying neurocognitive function in this patient population.

Several demographic and clinical features, including age and KPS, are highly correlated with IDH1 mutation status. While it may initially seem like a good idea to simply control for these in a statistical model, given multicollinearity between the demographic and clinical features and the tumor mutation status, as well as neurocognitive outcome, this would lead to suppression and reduce the magnitude of the relationship between the predictors and the outcome. As for the specific concern regarding age differences across IDH1 groups, the standardized neurocognitive test scores are age adjusted—thus the difference in neurocognitive performance between the 2 patient samples is unlikely to be due to the difference in age.

Freyschlag et al correctly pointed out that patients with IDH1 wild-type (WT) tumors had smaller T1-weighted but similar fluid attenuated inversion recovery volumes compared with mutant (IDH1-M) tumors. They also had greater and more frequent neurocognitive dysfunction with highly significant associations between neurocognitive function and lesion size—whereas this association was absent in patients with IDH1-M tumors. We have recently published results from our analysis of the brain’s gray matter structural connectome in this patient sample that demonstrated patients with IDH1-WT tumor compared with IDH1-M tumor had lower degree and lower network efficiency in several medial frontal, posterior parietal, and subcortical regions.² Cognitive reserve mediated the relationship between network efficiency and cognitive dysfunction in patients with IDH1-M tumors but not IDH1-WT tumors. Further, tumor volume was a predictor of cognitive dysfunction in patients with IDH1-WT tumors but not IDH1-M tumors. Again, while ascribing specific causal effects is difficult, we argue that the more rapidly growing IDH1-WT tumor results in greater disruption in the organization of the brain’s network topology and greater cognitive dysfunction. Given the increasing interest in both the molecular characterization of tumors and quantitatively and objectively measuring neurocognitive function, we expect there will be larger patient cohorts in the future that will permit investigating the extent to which patient age mediates neuroplasticity, brain connectivity, and neurocognitive function within and between samples of patients with IDH1-WT and IDH1-M tumor.

Importantly, we disagree that the most likely “cause” of neurocognitive impairment is diffuse infiltration into subcortical networks in a manner similar to low-grade gliomas as suggested by Freyschlag et al. Numerous studies, including our own work,³ have documented worse neurocognitive function in patients with higher-grade tumors matched or controlled for lesion size. Similarly, stroke patients with similar lesion volume as brain tumor patients present with greater neurocognitive dysfunction. This underscores the rationale for postulating that lesion momentum, specifically more rapidly expanding (and potentially destructive) neuropathology within the fixed space of the cranial vault and its effects on compensatory neuroplasticity, is perhaps the most important contributor to neurocognitive dysfunction. Hence, efforts to slow tumor growth are warranted to minimize neurocognitive dysfunction and extend life.