In this issue of the AJNR, Graves et al (page 613) describe serial 3D proton MR spectroscopic (MRS) imaging and conventional postcontrast T1-weighted MR imaging in 18 adult patients with gliomas prior to and after MR-targeted gamma knife radiosurgery. Their main results summarized below were in general agreement with many prior single-voxel MRS studies in that increased choline (Cho) and decreased N-acetylaspartate (NAA) were evident in areas of recurrent tumor, whereas Cho was reduced in patients who responded to therapy and who underwent stable follow-up MR examinations. The authors adopted a rigorous definition with a contralateral brain reference for tumor-suggestive MRS voxels as those in which normalized Cho (nCho) was elevated and normalized NAA (nNAA) was decreased by at least 2 SD of the corresponding normal metabolite distribution. As experience increases, analyses of sensitivity, specificity, and area under the receiver-operator characteristic curve for this definition of tumor-suggestive MR spectra, and for threshold (cutpoint) values of nCho and the Cho/NAA ratio in gamma-knife treated cases, with histologic correlations, would be enlightening.
Changes at 1 and approximately 5 months in the median values of nCho, nNAA, and the Cho/NAA ratio within the gamma knife target were tabulated for patients with radiographic (MR) outcomes compatible with stable disease, local recurrence contiguous with the target, or recurrence remote from the target. A trend toward reduced median nCho was seen within the gamma knife target in the first postoperative month in all outcome groups, and at 5 months for patients with stable disease and local radiographic recurrence. In light of prior reports regarding changes in Cho and NAA in irradiated gliomas and in normal brain, and of radiobiological data regarding normal glial and neuronal radioresistance, the authors suggested that changes in nCho and nNAA levels within the gamma knife target could reflect changes in glial and neuronal cell densities after cell death induced by high-dose radiation. Similarly, a trend toward a progressive drop in the median Cho/NAA ratio within the gamma knife target from baseline to 5 months postoperatively was observed in the stable and in the remote recurrence groups. An explanation for this serial fall in Cho/NAA suggested by the authors may be a corresponding progressive drop in the fraction of viable neoplasm as a consequence of high-dose radiation, perhaps of greater clinical significance than individual nCho and nNAA levels. Although confirmation of these hypotheses with scheduled serial biopsies in patients with glioma would be impractical, serial biopsies and direct correlation of nCho, nNAA, and Cho/NAA with: 1) histologic measurements of tumor fraction, necrotic fraction, and mitotic index; 2) markers of apoptosis and angiogenesis required for neoplastic growth; and 3) levels of oncogene and suppressor expression may be insightful in future studies with animal models.
Regarding radiographic recurrence, high values of median Cho/NAA were evident in regions of new contrast enhancement for both local and remote recurrence groups at 1 and 5 months postoperatively. Biopsy confirmed residual/recurrent glioma in six patients with local or remote radiographic recurrence. Presuming that gliomas in the patients who underwent biopsy were most refractory to treatment, it is of interest that Cho/NAA values on final follow-up spectra were all greater than median 5-month postoperative values for new regions of contrast enhancement. The observation that in nine cases an MR spectroscopic abnormality preceded a coincident increase in postcontrast enhancement by 1 to 2 months suggests an expanding role for MRS in postoperative monitoring.
Although not stated explicitly in the article, the authors suggest definitions for metabolic outcomes to therapy in their narration of the number of patients in whom at least one tumor-suggestive voxel was found over the 3D MRS field of view. A possible definition for a metabolic response to gamma knife therapy would be the absence of any tumor-suggestive MRS voxels in a volume of interest, either within the target or the surrounding brain. Conversely, metabolic residual/recurrent disease could be defined as the presence of at least one tumor-suggestive voxel in a volume. Two questions arise relative to the metabolic response of gliomas to gamma knife irradiation, given the apparent conceptual paradox of a focal irradiative treatment for an often refractory and diffusely infiltrating neoplasm. First, speaking to the effect of gamma knife irradiation on the targeted tissue only, one may ask, “Does the number of patients with no tumor-suggestive MRS voxels within the MR-derived target at final follow-up differ from that at pre-gamma-knife baseline?” The reported data suggest a trend at final follow-up toward fewer patients with at least one tumor-suggestive voxel inside the gamma knife target (11 vs four patients). With a larger sample size, one may consider a contingency table with mutually exclusive states (metabolic response vs metabolic residual/recurrence) in the rows (baseline pre-gamma-knife) and columns (final follow-up) amenable to analysis with McNemar's matched χ2 method that may demonstrate a significantly improved metabolic outcome within the target.
Speaking to a possible lack of metabolic response to gamma knife irradiation outside of the MR-derived target, one may ask an analogous second question, “Does the number of patients with no tumor-suggestive MRS voxels outside of the MR-derived target at final follow-up differ from that at pre-gamma-knife baseline?” The reported data suggest a similar number of patients with metabolic residual/recurrent disease outside of the target at pre-gamma-knife baseline and final follow-up (10 vs 11 patients). If contingency tables or other analyses were to demonstrate a significant metabolic response within, but no change outside of, the MR-derived target, on average, a logical extension might be to combine MRS and MR findings in the formulation of the gamma knife target, as was done with Graves et al's patient 13. Similar analyses may prove insightful for necrosis-suggestive MRS voxels determined with lipid/lactate resonances.
The many strengths of Graves et al's article include: high clinical relevance for medical centers conducting or contemplating a gamma knife service; a heterogeneous study population with respect to glioma type, grade, and prior therapy typical of recruitment into emerging treatment techniques; flexible and reproducible lesion sampling with small MRS voxels localized in three dimensions; gamma knife–treating physicians blinded to MRS results to reduce study bias; and straightforward definitions of MRS tumor-suggestive voxels and radiographic response to gamma knife irradiation. The few limitations of their article include: a small study sample that precluded statistical analyses to confirm or refute significant changes after radiosurgery in metabolite levels both within and outside of gamma knife targets, and both within (intra-) and between (inter-) radiographic outcome groups; unavailable post-gamma-knife histologic outcome for 12 of 18 patients; and unreported (or unreferenced) information regarding patient tolerance and technical success of combined MR and 3D MRS examinations.
As more patients are recruited, subgroup analyses may suggest different optimal gamma knife targeting algorithms or doses for patients with different histologic grades of glioma, or for those who have received prior cytotoxic therapy. Finally, further studies might extend the reporting of other outcome measures such as survival curves, and perhaps perform randomization to study arms that include or exclude gamma knife irradiation.
- Copyright © American Society of Neuroradiology