Peripheral hematological landscapes as biomarkers for detecting postoperative progression in glioblastoma multiforme: a multivariable risk scoring approach Authors: Ching-Chieh Tzeng, Meng-Wu Chung, Yin-Cheng Huang, Ko-Ting Chen, Ya-Jui Lin, Peng-Wei Hsu, Chi-Cheng Chuang, Kuo-Chen Wei, Cheng-Chi Lee Published: 2026 ABSTRACT PurposeGlioblastoma multiforme (GBM) is the most common malignant tumor with poor prognosis despite standard treatment. While various hematological parameters are prognostic for GBM survival, their potential in disease monitoring remains underexplored. Therefore, this study aimed to investigate the value of these parameters in monitoring GBM progression and to develop a clinically practical risk score model.MethodsWe retrospectively analyzed 193 newly diagnosed GBM patients who underwent surgery at Chang Gung Memorial Hospital. A total of 150 patients were included in the risk score model development and divided into progressive disease (PD) and non-PD groups, based on magnetic resonance imaging findings 6 months postoperatively. Parameters showing diagnostic potential for PD in univariate analyses were included in the multivariate analysis. Factors exhibiting independent effects were integrated into the risk score model. The remaining 43 patients underwent internal validation.ResultsSeventy-one patients were diagnosed with PD. In the univariate analyses, the white blood cell count (WBC), absolute neutrophil count (ANC), platelet count (PLT), and neutrophil-to-lymphocyte ratio (NLR) were significantly higher (p < 0.05) in the PD group. WBC, NLR, PLT, and resection status exhibited independent effects in multivariate analyses and were integrated into a four-point risk score model. A higher risk score correlated with significantly higher odds of PD (odds ratio = 2.57, p < 0.001). The model was reproducible in internal validation.ConclusionsThis study revealed that peripheral hematological parameters are associated with disease progression. The four-point risk score model demonstrated diagnostic utility for detecting disease progression at 6 months postoperatively.Supplementary InformationThe online version contains supplementary material available at 10.1007/s11060-026-05647-8. Introduction Glioblastoma multiforme (GBM) is the most common primary malignant tumor of the central nervous system, accounting for approximately 50% of all cases, and is associated with poor prognosis [1, 2]. Despite the standard Stupp regimen, the progression-free survival (PFS) of GBM is 6–8 months [3, 4]. For recurrent GBM, repeat surgery is beneficial when gross total resection is feasible [5]. Thus, the early detection of GBM recurrence is important. According to the National Comprehensive Cancer Network Guidelines for central nervous system cancers, postoperative surveillance for GBM should include brain magnetic resonance imaging (MRI) 3–6 weeks after concurrent chemoradiotherapy and every 2–3 months for the first 3 years. After the first 3 years, MRI should be performed every 2–4 months indefinitely [6]. However, compared with laboratory tests, brain MRI is less accessible and more costly, especially in some developing countries. Thus, developing a more convenient surveillance tool that enables more frequent monitoring of GBM would be valuable, and laboratory tests may effectively serve this purpose. Unlike many other cancers, GBM lacks a clinically established tumor marker for disease monitoring [7]. However, perturbations of immune organization, manifesting as alterations in peripheral blood cell counts, have been observed across diverse malignancies, including GBM [8]. The relationship between shorter survival times and altered peripheral blood cell counts in GBM has also been observed in many studies [9–11]. This study aimed to assess the potential of hematological parameters for detecting GBM progression and develop a clinically practical risk score model. Methods Study design and population This retrospective cohort study included patients who were newly diagnosed with GBM according to the 2021 World Health Organization classification of tumors of the central nervous system [12] and underwent surgical resection at Chang Gung Memorial Hospital between October 2011 and December 2024. Ethical approval for this study was provided by the Institutional Review Board of Chang Gung Memorial Hospital (IRB No. 202500824B0). This study was performed in line with the principles of the Declaration of Helsinki. The exclusion criteria were as follows: (1) age < 18 years; (2) patients who underwent biopsy only instead of maximal safe resection; (3) loss to follow-up or death within 6 months postoperatively; (4) lack of hemogram data or MRI scan at our time points of interest; (5) those who underwent a second operation for recurrence before the 6-month timepoint; and (6) other factors that could influence the hemogram data, including active infection or steroid use. Data collection Data were collected retrospectively from charts and image reviews. We used data from October 2011 to December 2022 for model development, whereas data from January 2023 to December 2024 were used for internal validation. Brain MRI review All patients underwent brain MRI preoperatively, postoperatively, and 3 and 6 months postoperatively. The extent of resection, i.e., gross total resection (GTR) or non-GTR, was determined on the basis of immediate postoperative brain MRI findings. The treatment response, including disease progression, was assessed using the Response Assessment in Neuro-Oncology criteria [13]. All brain MRI scans were independently reviewed by a senior neurosurgeon. Hematological parameters Hematological parameters were collected preoperatively, before the concurrent chemoradiotherapy (pre-CCRT), and 6 months postoperatively, including the white blood cell count (WBC), absolute neutrophil count (ANC), absolute lymphocyte count (ALC), platelet count (PLT), absolute monocyte count (AMC), neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), and monocyte-to-lymphocyte ratio (MLR). Statistical analysis The patients were divided into two groups, progressive disease (PD) and non-PD, based on whether disease progression was diagnosed on MRI within 6 months postoperatively. Intergroup univariate analyses were performed using independent t-tests for continuous variables and chi-squared tests for categorical variables. Point-biserial correlation analysis between binary categorical and continuous variables, along with linear regression with collinearity diagnostics between continuous variables, was performed to exclude overlapping factors. Variables with a p value < 0.05 in the univariate analyses were considered candidate variables and were introduced into the multivariate logistic regression analysis after ruling out substantial multicollinearity. In the point-biserial correlation analysis, an absolute correlation coefficient ≥ 0.7 was considered indicative of a high correlation. Collinearity analysis was performed using variance inflation factors (VIFs). Collinearity is possible when the VIF is ≥ 2 and severe when the VIF is ≥ 5. For better clinical applicability, the continuous variables were dichotomized and their associations with PD before and after dichotomization were compared. The cutoff values for continuous variables were determined using the Youden index via receiver operating characteristic (ROC) curve analysis. Finally, variables demonstrating independent effects were integrated into the risk score model. The risk model was evaluated using logistic regression analysis, with odds ratios calculated for each one-point increase in the score. Model discrimination was evaluated using ROC curve analysis and the area under the curve (AUC). Model calibration was assessed using the Brier score, calibration intercept, and calibration slope. For internal validation, the cutoff values and scoring algorithm derived from the training set were applied unchanged to the validation set without model refitting. All p-values were two-sided, with statistical significance set at p < 0.05. All statistical analyses were performed using SPSS Statistics 22 (IBM Corp.) and R software (version 4.5.0; R Core Team). Results Patient characteristics We reviewed 568 patients newly diagnosed with GBM at our institution during the study period. Among these patients, 375 were excluded based on the following exclusion criteria: age < 18 years, 8 patients; underwent biopsy only instead of maximal safe resection, 27 patients; loss to follow-up or death within 6 months postoperatively, 120 patients; lack of hemogram data or MRI image at the time points of interest, 160 patients; underwent a second operation for recurrence before the 6-month timepoint, 45 patients; and other factors including active infection or steroid use that could influence the hemogram data, 15 patients. In total, 193 patients were enrolled in this study, with 150 and 43 patients in the training and validation sets, respectively. A flowchart of patient selection is shown in Fig. 1. Fig. 1Flowchart of patient selection Flowchart of patient selection The demographic and imaging characteristics of patients in the training set are presented in Table 1. The average age (± standard deviation) was 58.3 (± 14.3), and the mean age (range) was 61.0 (21.0–85.0) years. Ninety-two patients (62.3%) were male, and 75 (50.0%) underwent gross total resection. Regarding tumor location, most of the GBMs were supratentorial (97.3%), and 31 cases (20.7%) involved midline structures. Within 6 months postoperatively, 71 patients (47.3%) exhibited disease progression. Table 1Baseline demographic characteristics, image features, resection status, and hematologic parametersPD (n = 71)Non-PD (n = 79)All (n = 150)P valueAge of diagnosis59.2 ± 13.257.5 ± 15.358.3 ± 14.30.460Gender0.626 Male45/71 (63.4%)47/79 (59.5%)92/150 (61.3%) Female26/71 (36.6%)32/79 (40.5%)58/150 (38.7%)Extent of resection0.002 Gross total resection26/71 (36.6%)49/79 (62.0%)75/150 (50.0%) Non-gross total resection45/71 (63.4%)30/79 (38.0%)75/150 (50.0%)Location0.122 Supratentorial71/71 (100.0%)75/79 (94.9%)146/150 (97.3%) Infratentorial0/71 (0.0%)4/79 (5.1%)4/150 (2.7%)Midline structures involvement0.347 With midline structures involvement17/71 (23.9%)14/79 (17.7%)31/150 (20.7%) Without midline structures involvement54/71 (76.1%)65/79 (82.3%)119/150 (79.3%)WBC, white blood cell; ANC, absolute neutrophil count; ALC, absolute lymphocyte count; AMC, absolute monocyte count; NLR, neutrophil-to-lymphocyte ratio; MLR, monocyte-to-lymphocyte ratio; PLT, platelet count; PLR, platelet-to-lymphocyte ratio Table 1Baseline demographic characteristics, image features, resection status, and hematologic parametersPD (n = 71)Non-PD (n = 79)All (n = 150)P valueAge of diagnosis59.2 ± 13.257.5 ± 15.358.3 ± 14.30.460Gender0.626 Male45/71 (63.4%)47/79 (59.5%)92/150 (61.3%) Female26/71 (36.6%)32/79 (40.5%)58/150 (38.7%)Extent of resection0.002 Gross total resection26/71 (36.6%)49/79 (62.0%)75/150 (50.0%) Non-gross total resection45/71 (63.4%)30/79 (38.0%)75/150 (50.0%)Location0.122 Supratentorial71/71 (100.0%)75/79 (94.9%)146/150 (97.3%) Infratentorial0/71 (0.0%)4/79 (5.1%)4/150 (2.7%)Midline structures involvement0.347 With midline structures involvement17/71 (23.9%)14/79 (17.7%)31/150 (20.7%) Without midline structures involvement54/71 (76.1%)65/79 (82.3%)119/150 (79.3%)WBC, white blood cell; ANC, absolute neutrophil count; ALC, absolute lymphocyte count; AMC, absolute monocyte count; NLR, neutrophil-to-lymphocyte ratio; MLR, monocyte-to-lymphocyte ratio; PLT, platelet count; PLR, platelet-to-lymphocyte ratio Baseline demographic characteristics, image features, resection status, and hematologic parameters WBC, white blood cell; ANC, absolute neutrophil count; ALC, absolute lymphocyte count; AMC, absolute monocyte count; NLR, neutrophil-to-lymphocyte ratio; MLR, monocyte-to-lymphocyte ratio; PLT, platelet count; PLR, platelet-to-lymphocyte ratio Univariate analyses Hematological parameters were compared across three time points (Fig. 2). Regarding the preoperative and pre-CCRT hematological parameters, significantly higher preoperative AMC (p = 0.041) and MLR (p = 0.048) were observed in the PD group than in the non-PD group. At 6 months postoperatively, the PD group exhibited higher WBC (p = 0.001), ANC (p < 0.001), NLR (p = 0.011), and PLT (p = 0.031) than those in the non-PD group; a higher PLR, with marginal significance (p = 0.066), was also observed in the PD group. No significant differences were observed in other hematological parameters. Age, sex, surgical resection rate, tumor location (supratentorial vs. infratentorial), and midline structure involvement were also analyzed. The only significant risk factor in univariate analyses was non-GTR (p = 0.002). Fig. 2Longitudinal trends of hematological parameters at three timepoints (preoperatively, pre-CCRT, and 6 months postoperatively). Values are presented as mean ± standard error. Preoperative AMC, MLR, and 6-month postoperative WBC, ANC, NLR, and PLT values are significantly higher in the disease progression group than in the control group. WBC, white blood cell; ANC, absolute neutrophil count; ALC, absolute lymphocyte count; NLR, neutrophil-to-lymphocyte ratio; PLT, platelet count; PLR, platelet-to-lymphocyte ratio; AMC, absolute monocyte count; MLR, monocyte-to-lymphocyte ratio; CCRT, concurrent chemoradiotherapy; * p < 0.05 Longitudinal trends of hematological parameters at three timepoints (preoperatively, pre-CCRT, and 6 months postoperatively). Values are presented as mean ± standard error. Preoperative AMC, MLR, and 6-month postoperative WBC, ANC, NLR, and PLT values are significantly higher in the disease progression group than in the control group. WBC, white blood cell; ANC, absolute neutrophil count; ALC, absolute lymphocyte count; NLR, neutrophil-to-lymphocyte ratio; PLT, platelet count; PLR, platelet-to-lymphocyte ratio; AMC, absolute monocyte count; MLR, monocyte-to-lymphocyte ratio; CCRT, concurrent chemoradiotherapy; * p < 0.05 Multivariate logistic regression analysis and model development In the point-biserial correlation analysis, no significant correlation was found between non-GTR and WBC_6m (r = 0.129, p = 0.115) or PLT_6m (r = 0.035, p = 0.668). Weak positive correlations were observed between non-GTR and ANC_6m (r = 0.193, p = 0.018) and between non-GTR and NLR_6m (r = 0.225, p = 0.006). The WBC, ANC, NLR, and PLT were included in the collinearity diagnostics, and severe collinearity was found between ANC and NLR (VIFs = 31.96 and 41.67, respectively). After removing ANC from the multivariate analysis, the VIFs of the WBC, NLR, and PLT decreased to 1.45, 1.34, and 1.15, respectively, showing no obvious collinearity. The WBC, NLR, PLT, and extent of resection were incorporated into the multivariate logistic regression model. According to the multivariate logistic regression and collinearity analyses, the risk score comprised the WBC, NLR, PLT, and resection status (GTR vs. non-GTR). Two models were developed in this study. Model 1 was developed using continuous hemogram data. In this model, resection status remained significantly associated with PD, whereas WBC, NLR, and PLT did not reach statistical significance individually. Although continuous data may preserve more statistical information, Model 2 dichotomized the hematological parameters to improve clinical applicability and establish an easy-to-use risk stratification tool. The cutoff values were 4750/µL for WBC, 4.2 for NLR, and 216,000/µL for PLT, as determined using the ROC-derived Youden index. After dichotomization, elevated WBC, elevated NLR, elevated PLT, and non-GTR were all significantly associated with PD. Therefore, model 2 was adopted as the basis of the simplified point-based risk score model. The details are presented in Table 2. Table 2Multivariate logistic regression analysis for the risk score modelOR95% CIP valueModel 1WBC(×10³/µL)1.2000.964–1.4930.103NLR1.1120.940–1.3160.215PLT(×10³/µL)1.0050.999–1.0110.104Non-GTR2.4411.206–4.9420.013Model 2WBC > 4750/µL2.4371.143–5.1950.021NLR > 4.22.5841.161–5.7520.020PLT > 216,000/µL2.8841.172–7.0940.021Non-GTR2.5321.230–5.2130.012WBC and PLT values in model 1 were rescaled prior to regression analysis to improve interpretability of odds ratios. WBC, white blood cell; NLR, neutrophil-to-lymphocyte ratio; PLT, platelet count; GTR, gross total resection Table 2Multivariate logistic regression analysis for the risk score modelOR95% CIP valueModel 1WBC(×10³/µL)1.2000.964–1.4930.103NLR1.1120.940–1.3160.215PLT(×10³/µL)1.0050.999–1.0110.104Non-GTR2.4411.206–4.9420.013Model 2WBC > 4750/µL2.4371.143–5.1950.021NLR > 4.22.5841.161–5.7520.020PLT > 216,000/µL2.8841.172–7.0940.021Non-GTR2.5321.230–5.2130.012WBC and PLT values in model 1 were rescaled prior to regression analysis to improve interpretability of odds ratios. WBC, white blood cell; NLR, neutrophil-to-lymphocyte ratio; PLT, platelet count; GTR, gross total resection Multivariate logistic regression analysis for the risk score model WBC and PLT values in model 1 were rescaled prior to regression analysis to improve interpretability of odds ratios. WBC, white blood cell; NLR, neutrophil-to-lymphocyte ratio; PLT, platelet count; GTR, gross total resection Model performance and fitness Patients were stratified using our risk score model, with a range of 0–4 points. The PD rates at 6 months across the 0–4-point range were 16.0% (4/25), 34.8% (16/46), 53.3% (24/45), 75% (21/28), and 100% (6/6), respectively. In the logistic regression analysis, a higher risk score was correlated with significantly higher odds of PD (odds ratio [OR] = 2.57, p < 0.001). The area under curve (AUC) of the risk score model was 0.741, with a Brier score of 0.215, indicating fair discrimination and predictive accuracy. The calibration intercept was − 0.001, and the calibration slope was 0.996, suggesting good calibration in the training set. The details are presented in Table 3. The ROC curve of the training set is shown in Fig. 3A. Compared with the logistic probability model using continuous hematological parameters, which showed an AUC of 0.711, the risk score model with dichotomized WBC, NLR, and PLT demonstrated slightly higher discrimination. The risk score also achieved higher discrimination than any individual component alone, including resection status. The AUCs of WBC, NLR, platelet count, and resection status were 0.633, 0.623, 0.592, and 0.627, respectively, as presented in Supplementary Fig. 1. Table 3Risk score model performance evaluation and calibrationScorePD (%)Non-PD (%)OR95% ClP valueAUCBrier scoreCalibration interceptCalibration slopeTraining set(n = 150)04 (16.0%)21 (84%)2.5741.767–3.750< 0.0010.7410.215-0.0010.996116 (34.8%)30 (65.2%)224 (53.3%)21 (46.7%)321 (75.0%)7 (25.0%)46 (100.0%)0 (0%)Validation set(n = 43)01 (16.7%)5 (83.3%)2.4201.297–4.5150.0050.7540.196-0.1810.93515 (33.3%)10 (66.7%)23 (33.3%)6 (66.7%)37 (87.5%)1 (12.5%)44 (80.0%)1 (20%)PD, progressive disease; OR, odds ratio; AUC, area under curve Table 3Risk score model performance evaluation and calibrationScorePD (%)Non-PD (%)OR95% ClP valueAUCBrier scoreCalibration interceptCalibration slopeTraining set(n = 150)04 (16.0%)21 (84%)2.5741.767–3.750< 0.0010.7410.215-0.0010.996116 (34.8%)30 (65.2%)224 (53.3%)21 (46.7%)321 (75.0%)7 (25.0%)46 (100.0%)0 (0%)Validation set(n = 43)01 (16.7%)5 (83.3%)2.4201.297–4.5150.0050.7540.196-0.1810.93515 (33.3%)10 (66.7%)23 (33.3%)6 (66.7%)37 (87.5%)1 (12.5%)44 (80.0%)1 (20%)PD, progressive disease; OR, odds ratio; AUC, area under curve Risk score model performance evaluation and calibration Training set (n = 150) Validation set (n = 43) PD, progressive disease; OR, odds ratio; AUC, area under curve Fig. 3(A) ROC curve of the risk score for the training set. (B) ROC curve of the risk score for the validation set. AUC, area under curve (A) ROC curve of the risk score for the training set. (B) ROC curve of the risk score for the validation set. AUC, area under curve Internal validation Patients in the validation set were stratified using the risk score model. The PD rates across the 0–4 point score range in the validation set were 16.7% (1/6), 33.3% (5/15), 33.3% (3/9), 7/8 (87.5%), and 4/5 (80%), respectively. In the logistic regression analysis, a higher risk score was also correlated with significantly higher odds of PD (OR = 2.42, p = 0.005), with an AUC of 0.754 and a Brier score of 0.196. The calibration intercept was − 0.181, and the calibration slope was 0.935. The details are presented in Table 3. The ROC curve of the validation set is shown in Fig. 3B. Discussion In this study, 71 patients were classified as PD and 79 as non-PD. Elevated inflammatory markers, including WBC, ANC, PLT, and NLR, were significantly associated with PD in univariate analyses. After addressing collinearity, WBC, NLR, PLT, and resection status were integrated into a simplified four-point risk score model. The model demonstrated moderate discriminative performance (AUC = 0.741) and was significantly associated with PD (OR = 2.57, p < 0.001). Importantly, internal validation showed comparable performance, supporting its potential utility as a clinically applicable tool for risk stratification and PD detection. Tumor microenvironment (TME) and immune system in GBM In GBM, the TME is highly heterogeneous and comprises extracellular matrix alterations, glioma stem cells, and diverse immune infiltrates [8, 14]. In the lymphoid compartment, malignancies are often characterized by lymphopenia along with functionally impaired T cells, which facilitate tumor progression. In GBM specifically, previous studies have reported elevated WBC and neutrophil count compared with those in healthy controls, correlating with increased systemic levels of pro-inflammatory cytokines, such as interleukin (IL)-6, IL-8, and IL-1β [8, 15]. Neutrophils and macrophages play a central role in systemic dysregulation. Neutrophils are recruited to the GBM site via strong chemokine gradients and tumor-derived cytokines. Once present, tumor-associated neutrophils (TANs) polarize into antitumor (N1-like) and pro-tumor (N2-like) states. In the absence of specific therapeutic modulation, TANs in GBM predominantly exhibit pro-tumor activities, including the promotion of angiogenesis and immunosuppression. Similarly, tumor-associated macrophages act as key regulators within the TME, fostering tumor growth by enhancing invasiveness and creating an immunosuppressive milieu through various inhibitory effects on other immune cells. These cytokine-driven shifts in hematopoiesis, specifically toward granulopoiesis at the expense of lymphopoiesis, are reflected in the NLR [16, 17]. Deng et al. suggested that the prognostic value of the NLR in GBM is primarily driven by chronic inflammation rather than lymphopenia alone [18]. Similarly, in the present study cohort, we observed a higher NLR in the PD group than in the non-PD group, despite no significant difference in the ALC. In addition, altered platelet profiles may contribute to the systemic landscape of GBM. Thrombocytosis in cancer is often driven by IL-6-induced thrombopoietin synthesis and the release of kynurenine, a tryptophan metabolite that shifts megakaryocyte-erythroid progenitor differentiation toward megakaryocytes. Elevated IL-6 levels and the resulting increase in the PLT have been consistently associated with a poor prognosis in GBM [19–22]. Clinical values of hematological parameters in GBM Prognostic values Regarding the clinical value of hematological parameters in GBM, most previous studies have focused on their prognostic value. Higher WBC [23, 24], ANC [22, 24], NLR [9, 10, 22, 25, 26], and PLT [22] were reported to be related to shorter survival, whereas the PLR tended to be higher in progressive disease across meta-analyses [22, 27]. Previous studies have reported inconsistent findings concerning the predictive capability of the ALC, but increasing evidence supports the relationship between treatment-related lymphopenia and poor prognosis [22, 28]. In a recent meta-analysis, a lower lymphocyte-to-monocyte ratio (LMR) correlated with poor overall survival but not progression-free survival [29]. These studies showed that hematological parameters correlated with glioma grade and provided potential diagnostic utility for GBM diagnosis [30, 31]. Dynamic change of hematological parameters Studies designed to directly evaluate the dynamic changes in peripheral hematological parameters in recurrence may provide insight into the disease state of recurrent GBM, but they remain scarce. In a cohort study involving four time points (preoperative, pre-radiotherapy, within or after concurrent chemoradiotherapy, and 4 weeks prior to or after recurrence), lower NLR, lymphocyte count, and PLT were found at recurrence than at the preoperative timepoint, potentially reflecting the importance of treatment-related myelosuppression on the peripheral blood cell count [18]. In another study that enrolled 41 patients with recurrent GBM, the matched-pair comparison revealed decreased values of neutrophils, lymphocytes, monocytes, and platelets and reduced LMR upon recurrence, while no significant difference was observed in the NLR and PLR. Additionally, survival analysis was performed in the recurrent cohort, and a poorer prognosis was associated with elevated values of neutrophils, monocytes, and platelets [32]. Recent studies have expanded beyond preoperative hematological parameters to explore the relationship between dynamic changes in these markers and the prognosis of GBM. Ma et al. reported that not only the preoperative NLR but also changes in the NLR were prognostic for progression-free survival in patients with gliomas [33]. Our previous research also yielded similar results, showing that a higher NLR after standard treatment than the preoperative value was an indicator of shorter PFS [10]. These findings suggest that longitudinal assessment of hematological parameters may reflect the evolving disease state and could potentially be associated with tumor progression or recurrence. Potential for detecting GBM progression Recently, the clinical value of non-invasive liquid biopsy as an adjuvant diagnostic tool in traditional imaging and histological studies for GBM has been broadly discussed and has shown promising results in disease monitoring. Despite their potential benefits, challenges remain that hinder their implementation in clinical practice, such as inadequate levels due to the blood–brain barrier and the lack of standardized protocols [34]. In comparison, peripheral hematological parameters are easier to assess. In addition to their prognostic value, their diagnostic value for disease monitoring remains underexplored and warrants further investigation. Evidence regarding the predictive value of hematological parameters for GBM progression is limited. Several studies have explored the utility of hematological parameters in differentiating disease progression from pseudoprogression in high-grade gliomas. Although not entirely equivalent, this distinction conceptually overlaps with our approach to differentiating PD and non-PD using these markers. Huang et al. analyzed the dynamic NLR of high-grade gliomas with early disease progression and pseudoprogression and found that a decreased post-treatment NLR was predictive of pseudoprogression [35]. Satragno et al. demonstrated that a higher NLR and systemic immune inflammation index (SII) (SII = platelet count × neutrophil count / lymphocyte count) were predictors of tumor progression, while the systemic inflammation response index (SIRI) (SIRI = neutrophil count × monocyte count / lymphocyte count) failed to achieve statistical significance [36]. Consistent with these findings, the WBC, ANC, NLR, and PLT were significantly associated with PD in the present study, whereas the AMC and MLR were not. Accordingly, we developed a four-point risk score model. Clinical implications Compared with other cancers that can be monitored using specific tumor markers, GBM surveillance relies primarily on regular MRI. Here, we developed a risk score model that showed potential utility for the short-term risk stratification of GBM progression. The simple four-factor risk score included the WBC, NLR, PLT, and resection status. Hematological parameters can be easily acquired through routine blood tests, making them applicable to clinical practice. The developed model showed fair discrimination and good calibration within the training set, and comparable performance in the validation set. However, the possibility of information loss after cutoff-based dichotomization of hematologic parameters must be taken into consideration. In addition, the cutoff values in this study were derived from a single-center cohort with a limited sample size and should be regarded as exploratory until further validated. Nevertheless, this cohort and the risk score model suggest that the hematologic parameters may have value in adjunctive PD detection and postoperative risk stratification before more accurate markers become available. Although replacing regular MRI with the risk score model would be inappropriate in GBM surveillance, it may help clinicians identify patients who could benefit from a more intensive follow-up, such as earlier MRI, potentially leading to adjustments in treatment strategies. Strengths and limitations Compared with our previous study [10], which investigated the prognostic value of the dynamic NLR, the present study focused on the correlation between several other hematological parameters and GBM progression at 6 months. To the best of our knowledge, this is the first study to investigate the potential value of hematological parameters in detecting GBM progression. Additionally, this study provides valuable information on the dynamic changes in hematological parameters in patients with GBM within 6 months postoperatively. Nevertheless, this study has some limitations. Bias was inevitable due to the retrospective design. Additionally, this was a single-center study with a limited sample size and a short follow-up duration (6 months). Although patients were excluded when the hematological parameters might have been influenced by confounding factors such as infection and medication, through a meticulous medical record review, ruling out all factors that might influence blood cell distribution is impossible. In addition, the exclusion of patients who died or underwent repeat surgery may have limited the representativeness of this cohort with respect to the broader GBM population. Even though leukocyte-related parameters showed prognostic value, we could not assess neutrophil phenotypic heterogeneity, including N1- and N2-like polarization states, using peripheral blood tests. Regarding pseudoprogression, the 2010 Response Assessment in Neuro-Oncology (RANO) criteria were developed to standardize the assessment of true progression and pseudoprogression. Critically, these criteria restrict true progression to new contrast enhancement outside of the radiation field within 12 weeks of completing radiotherapy [13]. In our study, tumor progression was primarily determined using MRI findings obtained within approximately 6 months after treatment, which may help reduce the diagnostic ambiguity between true progression and pseudoprogression. Moreover, advanced MRI techniques, including cerebral blood volume–based perfusion imaging, dynamic susceptibility contrast imaging, and magnetic resonance spectroscopy, may further enhance diagnostic accuracy and lesion characterization. In addition, positron emission tomography has emerged as a valuable adjunctive modality for detecting metabolically active tumor regions. Collectively, these imaging approaches provide complementary physiological and metabolic information that may improve evaluation of therapeutic response and disease progression [37]. Consequently, accurately distinguishing early tumor progression from pseudoprogression remains a significant clinical challenge, underscoring the need for longer follow-up periods and further validation through large-scale prospective studies. In some cases, differentiating between early progression and pseudoprogression may be difficult. Finally, since the hematological parameters incorporated into the risk score model were collected at the 6-month postoperative timepoint, which was close to the timing of MRI-based outcome assessment, this model should be interpreted as an adjunctive tool for concurrent risk stratification and PD detection rather than as a prospective predictive model. Conclusion This study developed a four-point risk score model, demonstrating moderate discriminatory ability for detecting disease progression within 6 months postoperatively in both the training and validation sets. Rather than replacing MRI surveillance, this model can serve as an adjunct to clinical practice. Prospective studies with larger cohorts, longer follow-up durations, and external validation are warranted to validate this risk score model. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2 Supplementary Material 3 Supplementary Material 3 Supplementary Material 4 Supplementary Material 4