Measurable Supratentorial White Matter Volume Changes in Patients with Diffuse Intrinsic Pontine Glioma Treated with an Anti-Vascular Endothelial Growth Factor Agent, Steroids, and Radiation

Discussion

According to published data from the Pediatric Brain Tumor Consortium, children with DIPG having >25% decrease in tumor volume (and ADC) after RT have a higher 6-month survival rate than those without such decreases, but the reasons are still unclear.⁶ Our data, with >50% volume decrease as the threshold, do not support this observation.

In our cohort, almost half of the patients experienced a volume reduction of >50% of the BS-L sometime during the course of their disease; yet, the ultimate outcome remains uniformly dismal. In this study, we showed that apart from the significant volume changes within the tumor lesion itself, measurable and non-negligible volume changes may also be induced in remote normal brain parenchyma, which is not directly targeted by RT; therefore, those changes likely develop in response to systemic medication used during treatment, particularly corticosteroids, which cause well-known, reversible pseudoatrophic changes in the brain.

In recent years, molecularly targeted treatments for adult high-grade gliomas have generated considerable interest.^7,8 Vandetanib, the anti-VEGF agent used in our clinical trial, is a tyrosine kinase receptor inhibitor that may inhibit VEGF receptor-2 tyrosine kinase activity and shows additional inhibitory activity against the Ret proto-oncogene tyrosine-protein kinase receptor and the epidermal growth factor receptor in isolated enzyme assays.⁸ By targeting VEGF, a promoter of angiogenesis that leads to the “haphazard” formation of deficient, new vessels in response to tissue hypoxia, anti-VEGF agents may indirectly decrease fluid leakage into the interstitial space, reducing peritumoral edema in synergy with corticosteroids by reducing the number of defective, leaky vessels.^9,10 The effect of anti-VEGF agents, including vandetanib, on normal vessels (and brain parenchyma) is unclear.

In the present study, the progressive decrease of the normalized ST-WM volume with time was evident. Similarly, although on a larger scale, normalized BS-L volume decreased until approximately 2 months following RT completion, after which BS-L volumes started to increase again, presumably due to commencing tumor recurrence. The BS-L responded well to treatment during Phase I. Normalized ADC values of the lesion decreased during this phase, likely indicating contraction of the intratumoral extracellular water compartments in response to synergistic effects of the various treatments. According to protocol requirements, vandetanib levels were kept constant in each patient stratum, contrary to dexamethasone, which was initially given at higher doses and gradually tapered off as the lesion size decreased and the patient's clinical signs and symptoms improved.² Vandetanib may not substantially reduce volume during Phase I because pharmacokinetic studies have shown that therapeutic blood levels of vandetanib are reached after approximately 45 days (6.5 weeks) in adults and, presumably, in pediatric patients as well.^2,11 Hence, by the eRT (6 weeks), vandetanib may not have reached a steady-state blood level that would be sufficient to substantially affect intratumoral edema and, thus, BS-L volume size. Hence, the pronounced volumetric changes in the BS-L during Phase I are likely due to a combination of tumor control by RT and the removal of vasogenic edema by dexamethasone.

Volume changes were also evident in ST-WM as early as Phase I, even though the supratentorial brain was largely outside the radiation field. On the basis of test simulations, approximately 20%–25% of the supratentorial brain (mainly the posterior temporal and the occipital lobes) in patients treated for DIPG receives radiation doses higher than 20 Gy. Therefore, volume decreases of the ST-WM reflect the effect of systemically administered dexamethasone on the normal cerebral parenchyma, especially because no regional differences in pseudoatrophic changes were seen between the “irradiated” and the “nonirradiated” brain regions. For the above-mentioned reasons, vandetanib is not expected to substantially affect ST-WM. The decreasing ADC values in ST-WM may indicate that water is removed, resulting in contraction of the interstitial space in the brain during Phase I. In our patients, no clinical or MR imaging finding suggested that the volume decrease in ST-WM would be indicative of any drug-induced myelinotoxicity.

As expected, the trend toward volume and ADC reduction of the BS-L continued in Phase II after RT completion, most probably due to the sustained effects of RT on the BS-L and the ongoing, potentially synergistic effect of dexamethasone and vandetanib. During this phase, dexamethasone was gradually reduced, and by the end of W14, most patients were taken off therapy. Therefore, the decreased rates may reflect decreasing corticosteroid use, suggesting that the delayed effects of RT and vandetanib may not be sufficient to sustain control of edema. Nevertheless, it is unclear which treatment component has the dominant effect on tumor volume reduction during Phase II. Normalized ST-WM volume also decreased at a slower rate, again, most likely due to weaker dexamethasone effects. The steady decrease of ADC values in ST-WM indicates further contraction of the interstitial water compartments in normal brain parenchyma.

In Phase III, BS-L volume progressively increased despite sustained vandetanib therapy, indicating early signs of tumor recurrence. ADC values gradually increased, suggesting a rebound of edema due to an increasing tumor burden.

Unlike in BS-Ls, volume in ST-WM minimally increased during Phase III at the group level. Because anti-VEGF was continuously given, the only factor inducing this change is probably the discontinuation of dexamethasone for most patients and slow remigration of water back to the ST-WM, leading to a modest re-expansion of the interstitial space. Such reversible brain atrophy due to medication-induced effects has been documented in clinical studies. For example, corticosteroid-induced brain volume changes occur in patients with multiple sclerosis, with researchers showing that brain fractional volumes return to baseline values 1–2 months after treatment.^12,13 Reversible changes in the brain also occurred in a cohort of patients with alcohol dependence under detoxification.^14,15 Similarly, reversal of pseudoatrophic brain changes and cognitive improvement have been associated with discontinuation of drug therapy in epilepsy.¹⁶ All studies accentuated the confounding dynamics of water shifts induced by drugs or other substances in the brain parenchyma, without a concomitant effect on histoarchitectural integrity.

Because the vandetanib dose level was fixed in each stratum, the only robust variable in patients was the presence or absence of dexamethasone. Our results show that corticosteroids may induce measurable volume changes not only in the BS-L but also in the ST-WM volume. This finding is quite remarkable because it clearly indicates that corticosteroids are powerful water-volume regulators not only in pathologic but also in normal tissues. The potential practical implication of this finding is that tumor volume reduction induced by drugs without a direct antitumoral effect (ie, corticosteroids and anti-VEGF drugs in our study) may be potentially misinterpreted as a partial response during treatment or follow-up when volume reduction exceeds 50%. This overestimation occurred in almost 50% of our patients. Hence, observed changes in DIPG “lesion volume” during treatment and the associated T2 normalization may not be reliable indicators of effective tumor control. Therefore, reliable appreciation of the actual antitumor effect (or lack thereof) of current or future antitumoral drugs in treating DIPG (or other diffusely infiltrative gliomas) in patients who receive corticosteroids may not be possible.

For the whole patient cohort, an overall mean 20% decrease of ST-WM volume was seen, though the ST-WM was largely unexposed to radiation. Assessing how much BS-L volume reduction was induced by RT itself is not feasible, but one may assume that at least 20% of the change is due to concurrent “drug therapy” as seen in the ST-WM. This percentage, however, may be even higher because steroids and anti-VEGF drugs may have a more robust effect on edematous tissue than on normal parenchyma. This notion is supported by our clinical observations suggesting that DIPGs presenting markedly increased T2 signal intensity at diagnosis respond better to therapy, likely due to the presence of more easily “treatable” intratumoral edema. We took advantage of the available data and calculated an “adjusted” volume decrease for the BS-L, after which only 2 patients had a volume reduction of >50%. However, only 4 patients had a volume increase of >25% at the ePFS. These data suggest that both the −50% and the +25% thresholds have poor correlation with clinical reality and are inadequate for partial response and progressive disease determinations. This finding is in keeping with reports showing that advanced MR imaging techniques, such as proton MR spectroscopy, show little, if any, improvement in the metabolic profile of lesions, even during apparent volume reduction of the tumor.^17,18

In DIPG, the T2 hyperintense area corresponds to the extent of vasogenic edema, which is induced by the tumor within through a mechanism (eg, development of leaky, angioneogenetic vessels) promoting the influx of excess water from the intravascular compartment to the intralesional interstitial compartment. In practical clinical terms, the T2 hyperintense area is likely an inaccurate representation of the actual tumor burden. Consequently, temporal variations of the T2 hyperintense lesion volume during the disease are probably inadequate to monitor therapy- or disease progression–related changes in tumor burden because non-negligible portions of those volume variations may reflect bulk water shifts to and from the lesion area. This situation would be particularly concerning in clinical trials.

The Response Assessment in Neuro-Oncology working group has developed response criteria in adult high-grade gliomas. Although most clinical trials adopt the Response Assessment in Neuro-Oncology criteria,¹⁹ the validity and appropriateness of these schemes for pediatric patients require further studies. Pediatric brain tumors present with unique challenges that may require adjustment of imaging criteria; however, a “one size fits all” strategy may not be the right answer.^19,20 It is unlikely that a universal, standard set of response criteria would be suitable for all types of pediatric brain tumors, including embryonal tumors, infiltrative gliomas, and focal gliomas, due to inherently different histoarchitectural features and tumor-host relationships. Advanced, quantitative MR imaging data may more directly capture key biologic or pathophysiologic tumor properties: indeed, perfusion and spectroscopic data reportedly have potential prognostic value in DIPG.^21,22 A recent study also demonstrated that diffusion measurements could be used as noninvasive biomarkers to differentiate distinct subgroups of DIPG and select patients for clinical trials.²³ Nonetheless, the possibility of incorporating advanced MR imaging techniques (perfusion imaging DTI, MR spectroscopy) in response assessment schemes for childhood brain cancer is still under evaluation.

Overall, shortcomings of anatomic MR imaging in reflecting actual tumor burden in cerebral neoplasms at diagnosis and during treatment are increasingly recognized, especially in cases of diffusely infiltrative high-grade gliomas, such as DIPG. Identifying surrogate biomarkers other than volume-based measures that can accurately predict or measure tumor response to treatment is crucial to effectively managing DIPG or other infiltrative tumors in clinical care and in drug trials.