High-Resolution Magic Angle Spinning Metabolomic Profiling of IDH-Wild-Type Glioblastoma Reveals a Composite Surgical Sampling Signature Shaped by Clinical and Anatomical Tumor Features

Background/Objectives: Tissue-based metabolomic readouts in IDH-wild-type glioblastoma may be strongly shaped by how tumor tissue is surgically accessed and sampled. We aimed to determine whether, and to what extent, surgical sampling context structures the HRMAS metabolic landscape, and to disentangle sampling-related contributions from clinico-anatomical confounders. Methods: We retrospectively analyzed 99 patients with de novo IDH-wild-type glioblastoma (35 biopsy-only, 64 resection: 40 gross-total, 21 near-total, 3 subtotal), yielding 166 HRMAS spectra and 47 quantified metabolites (nmol/mg). Patient-level profiles were compared using PCA, metabolite-wise testing, pathway-level aggregation (10 pathways), and variance partitioning by PERMANOVA, both unadjusted and adjusted for age, WHO PS, deep-seated location, midline involvement, multifocality, MGMT methylation, and eloquent area. Sensitivity analyses included clinico-anatomically restricted subgroups, 15 canonical metabolite ratios, and Probabilistic Quotient Normalization. Intratumoral heterogeneity was assessed in 44 multi-sampled patients. Results: Biopsy-only and resection-derived cases separated along PC1 in unsupervised PCA (62.6% variance; p < 0.001), with 42/47 metabolites differing after FDR correction. However, the surgical group explained only 2.6% of the total variance (PERMANOVA p = 0.026), and this share was no longer significant after confounder adjustment (p = 0.39). Clinico-anatomical restriction progressively attenuated the effect (42/47 → 1/47 significant metabolites). Ratio-based and PQN analyses showed a residual compositional difference beyond scaling (13/15 ratios; 16/47 metabolites). Intratumoral heterogeneity was greater in resections and preserved in an n-matched analysis (p = 0.020). Conclusions: The apparent biopsy-versus-resection metabolic difference is largely a composite signal reflecting clinico-anatomical patient selection with a smaller tissue-composition contribution. Biopsy-only and resection-derived specimens should not be pooled uncritically in tissue-based metabolomic studies of glioblastoma.