Trifecta of science: originality, controversy, confirmation.

Abstract

Significant contributions to the scientific literature provide a new result or insight, challenge or disconfirm accepted knowledge, or replicate and extend an important finding from previous research. In Moxon-Emre et al, we find all three. The new insight is the apparent importance of the radiotherapy boost as compared with the craniospinal radiation (CSR) dose in modeling neurocognitive outcome. The challenge to accepted knowledge comes in the form of the absence of an effect on neurocognitive outcome of reduced CSR dose. Finally, the replication refers to the rate of posterior fossa (PF) mutism and its impact on neurocognitive outcome. While this study contributes in several other ways, I will limit my comments to these three findings. Much of the relationship between radiation therapy in medulloblastoma (MB) and neurocognitive outcome (often in the form of IQ scores) can be captured by a simple formulation of dose volume interaction in which high dose and high volume of brain irradiated produces the worst outcomes (lower IQs) and low dose to smaller volumes of brain produces the best outcomes. Intermediate states produce intermediate outcomes. Parenthetically, it is acknowledged that this formulation omits the likely disproportionate contributions of specific structures and neural systems supporting neurocognitive function and development, which is an important line of research under active investigation in several programs in both humans and in animal models. Variation in overall dose is a function of CSR dose and boost dose, while variation in volume has increasingly become a matter of the volume of brain in the PF included in the boost, since CSR volume is a relative constant. Modern radiation therapy protocols have decreased the CSR dose for average-risk MB, and substantial progress has been made through such technological advances as 3D conformal therapy, intensity-modulated radiation therapy, and proton beam therapy in decreasing the volume of tissue in the PF exposed to radiation. In applying this simple algorithm to the current study, we would expect the three groups to be ordered in the following manner, with the first having the best outcome and the last, the worst outcome: (a) reduced CSR–tumor bed (TB) boost; (b) reduced CSR-PF boost or standard CSR-TB boost; (c) standard CSR-PF boost. In Figure 1A of Moxon-Emre et al, full-scale IQ (FSIQ) trajectories seem to support some but not all of these conjectures. The group receiving the least radiation to the least volume of brain does, indeed, have the most favorable outcome. However, there is not the predicted sorting of the other three groups (ie, they all have similar outcomes defined by nearly identical slopes). As the authors point out, this is somewhat anomalous and hard to reconcile with published studies claiming a neurocognitive benefit to reduced dose CSR compared with standard dose CSR (in this case 18.00 to 23.40 GY and 30.6039.40 Gy, respectively). The authors correctly raise doubts as to the reliability of the findings from one particular group in their study, the standard CSR-TB boost group, because of the small number of subjects in this group (n 7 for FSIQ). This group had somewhat higher baseline (intercept) IQ, much larger standard deviation in the intercept, and larger SE of the slope which casts further doubt on the findings from this particular group. Still, the absence of a difference in outcome between the reduced CSR-PF boost and the standard CSR-PF boost group is puzzling. They have similar means, similar intercepts, and similar slopes. Focusing on these two groups, Table 1 in Moxon-Emre et al hints at a possible explanation in the form of differences in aggregate risk indicators between these two groups with that of the standard CSR-PF group being more favorable. The standard CSR-PF compared with the reduced CSR-PF group had a lower percentage receiving chemotherapy (80.4% v 94.4%), a lower rate of mutism (13.7% v 32.1%), a lower percentage of hydrocephalus requiring diversion (35% v 50%), and older mean age at diagnosis (7.89 v 6.54 years). Their models apparently included mutism and hydrocephalus, but not age and chemotherapy. This raises the question, then, as to whether, aside from CSR dose, other factors influencing outcome were more favorable in the standard CSR-PF groups thereby offsetting the CSR dose advantage of the other group. It is also worth noting that version of IQ test can also introduce unknown variance across cohorts followed longitudinally. This is not to find fault with this particular study since this is a vexing problem for everyone working in this field. It merely reminds us of the limited precision of our outcome instruments, the extraordinary complexity of the developmental phenomena under investigation, and why judgements as to the preponderance of evidence are cautiously made. This study also illustrates the challenges posed by modern methods in radiation oncology capable of heterogeneous dose distributions that are not easily characterized for correlation with neurocognitive outcome. One approach that we proposed several years ago uses integral biologic effective dose, a radiobiologic metric that models the physical radiation dose and biologic effect. Similarly, equivalent uniform dose can be calculated from dose-volume histograms and has been used in one study to demonstrate optimized dose to the tumor while sparing structures presumed to be critical to neurocognitive functioning. These radiobiologic metrics and other approaches that JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L VOLUME 32 NUMBER 17 JUNE 1