Comment: Prediction of Survival in Surgically Treated Glioblastoma Multiforme Utilizing DTI and Contrast-Enhanced MRI

Letter to: Prediction of Survival in Surgically Treated Glioblastoma Multiforme Utilizing DTI and Contrast-Enhanced MRI Indian J Radiol Imaging 2026; 36(01): 048-058 DOI: 10.1055/s-0045-1807738 Prediction of Survival in Surgically Treated Glioblastoma Multiforme Utilizing DTI and Contrast-Enhanced MRI We read with great interest the article by Thomas et al.[ 1 ] We commend the authors for addressing the clinically important question of preoperative imaging biomarkers for survival prediction in glioblastoma (GBM). However, several methodological, data integrity, and statistical concerns warrant discussion. First, the study does not account for several well-validated independent predictors of survival in GBM, namely, age at diagnosis, Karnofsky Performance Status, extent of resection (EOR), and O⁶-methylguanine-DNA methyltransferase promoter methylation status.[ 2 ] Notably, no postoperative volumetric analysis was performed to objectively quantify EOR or residual tumor volume, making it untenable to assume that maximal safe resection was achieved across all cases, including those involving the eloquent cortex. Second, the diffusion tensor imaging acquisition protocol utilized only six diffusion-encoding directions for fractional anisotropy (FA) estimation. Contemporary literature recommends a minimum of 25 to 30 directions, with >64 directions preferred for clinical research, as fewer directions introduce angular undersampling and FA estimation error with zero degrees of freedom for noise correction.[ 3 ] Inter-observer variability for tumor volume and manual region of interest placement on FA maps was not reported, raising concerns regarding reproducibility. Third, there are some data-recording concerns. In Table 1 the mean tumor volumes appear to have a very large standard deviation with upper limits of the tumor volume approaching total intracranial volume, which seems biologically implausible.[ 1 ] This concern is compounded by the data in Table 2, where surviving patients are reported to harbor a median tumor volume of almost 80% of the intracranial volume. This may reflect a data recording, data-coding, or unit-conversion error.[ 1 ] Fourth, the follow-up strategy employed by the author needs explanation. As per our understanding, the survival and death data need to be time-coded, viz., 6 months, 1 year, and 2 years. Keeping a 6-month survivor in the survival group and a patient who died at 1 year in the dead group makes the groups heterogenous and renders any comparison between them incorrect. Fifth, there is a significant statistical concern. Firstly, the Methods section states that Pearson's correlation coefficient would be used; however, three key variables—tumor volume, perilesional edema volume, and FA value—were demonstrated to be non-normally distributed by the Shapiro–Wilk test.[ 1 ] Despite this, Table 3 employed Pearson's correlation to assess the relationship between the FA value and mortality.[ 1 ] Spearman's rank correlation should have been applied in the setting of non-normal distributions. Also, it is not clear how the mortality data were used as a continuous variable for the application of Pearson's correlation test. Mortality can be analyzed either as a time-to-event variable or as a categorical variable based on event status at a specific follow-up time point.[ 4 ] Finally, with 30 events and more than 10 predictors entered into the multivariate Cox model, the study violates the widely accepted rule of at least 10 to 15 events per variable, rendering the model severely underpowered.[ 5 ] The resulting estimates are unreliable, as evidenced by the hazard ratio of 45.9 (95% CI: 3.01–703.23) for microhemorrhage exceeding 25% in Table 5—a confidence interval spanning over 700 units, which is indicative of model instability.[ 1 ] Penalized regression methods, such as LASSO or ridge Cox regression, or restriction of model entry to two or three biologically prespecified variables, would have been more appropriate. We sincerely appreciate the initiative, novelty, and efforts made in this study. However, we hope the authors will address the concerns raised. We also suggest that a future prospective study with outcome assessment at a fixed follow-up duration be planned, taking into consideration the points raised above. Authors' Contribution R.C. and S.Z. conceptualized and collected the data and wrote the manuscript. V.K. extended administrative support, supervised the project, and proofread the manuscript. Publication History Article published online: 13 August 2026 © 2026. Indian Radiological Association. This is an open access article published by Thieme under the terms of the Creative Commons Attribution-NonDerivative-NonCommercial License, permitting copying and reproduction so long as the original work is given appropriate credit. Contents may not be used for commercial purposes, or adapted, remixed, transformed or built upon. (https://creativecommons.org/licenses/by-nc-nd/4.0/) Thieme Medical and Scientific Publishers Private Limited A-13A, Graphix Tower 1, 6th floor, Sector 62, Noida 201309, Uttar Pradesh, India Letter to this article: Reply to the Comment: Prediction of Survival in Surgically Treated Glioblastoma Multiforme Utilizing DTI and Contrast-Enhanced MRI Indian J Radiol Imaging eFirst DOI: 10.1055/s-0046-1827196