Increased glioblastoma risk in patients with myeloproliferative neoplasms and myelodysplastic syndromes

Abstract

Targeted therapies significantly extend the survival of patients with myeloproliferative neoplasms (MPNs) and myelodysplastic syndromes (MDS). This prolonged lifespan introduces an unintended clinical hazard. These patients face an elevated risk of developing primary central nervous system (CNS) tumors. The internal relationship between them remains to be explored. We integrated the Global Burden of Disease (GBD) and the Cancer Incidence in Five Continents (CI5) datasets to conduct spatial clustering mapping using Moran’s I. Vector autoregression and Granger causality analysis were employed to examine the longitudinal incidence of these diseases in Sweden from 1990 to 2021. Bibliometric profiling was utilized to identify research frontiers, and two-sample Mendelian randomization (MR) was used to determine genetic correlation. Geospatial mapping indicated a dually dominant burden of MPN + MDS (MPN and MDS, not MPN/MDS) and CNS tumors in developed regions. Longitudinal data from Sweden exposed a strictly unidirectional predictive trajectory exclusive to females. Myeloid malignancies reliably preceded CNS tumors. Bibliometric analysis demonstrated a remarkable interdecadal convergence on neuroinflammation and the tumor microenvironment. Significantly, MR confirmed that MPN (odds ratio [OR] = 1.84, 95% confidence interval [CI]: 1.22–2.76) and MDS (OR = 2.54, 95% CI: 1.62–3.96) have definite positive effects on GBM susceptibility. We established the possible comorbidity relationship between MPN + MDS and GBM, and raised the myeloid-neuro-immune crosstalk theory. It lays a reliable foundation on the interactions between MPN + MDS and GBM for future research, and offers insights for building multidisciplinary team collaboration.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s10238-026-02296-z.

Introduction

Myeloproliferative neoplasms (MPN), which include essential thrombocythemia (ET), polycythemia vera (PV), and primary myelofibrosis (PMF), are clonal hematopoietic stem cell disorders characterized by excessive production of mature myeloid cells driven by activating mutations. In contrast, myelodysplastic syndromes (MDS) are a heterogeneous group of clonal disorders defined by ineffective hematopoiesis, bone marrow dysplasia, and peripheral cytopenias. While these entities share some overlapping genetic features, they differ fundamentally in their pathophysiology and clinical presentation.

Although progress in targeted therapies has significantly prolonged the life expectancy of patients with ) MPN + MDS (MPN and MDS, not MPN/MDS), this survival advantage has been accompanied by a concurrent increase in the risk of independent primary solid tumors (IPSTs) [1, 2]. Nonhematological solid tumors have now become a significant cause of mortality in this patient population [2]. Particularly, central nervous system (CNS) tumors impose a disproportionately high health burden due to their extremely high lethality and morbidity rates [3]. When CNS tumors coexist with MPN + MDS, the inherent hemostatic dysfunctions, including both thrombotic and hemorrhagic tendencies, associated with these hematological disorders present formidable challenges and substantial risks during conventional neurosurgical and chemo radiotherapeutic treatments [4]. Therefore, it is of urgent clinical significance to elucidate the epidemiological risks between these two distinct entities.

From a physiological standpoint, this hematologic and neurologic comorbidity may be fundamentally rooted in the abnormal myeloid clonal expansion and the persistent chronic inflammatory state that are central to MPN pathogenesis. The hyperactivated JAK-STAT signaling pathway creates a strong proinflammatory microenvironment [5, 6]. It may potentially undermine the integrity of the blood-brain barrier (BBB) and subsequently establish a highly permissive oncogenic niche for primary CNS tumor development [7, 8]. While a Swedish population-based cohort study previously detected alarming signals, indicating a significantly elevated risk of brain tumors among MPN patients (hazard ratio [HR] = 2.8), and the large-scale genome wide association studies (GWAS) confirmed the existence of shared genetic susceptibility loci between gliomas and various non-glioblastoma malignancies, several crucial uncertainties remain[9, 10]. This signifies whether prevalent spatial clustering effects of this comorbidity or a genuine causal relationship at the genomic level. These fundamental questions have, to date, remained entirely unaddressed.

We designed a multidimensional investigation into the MPN + MDS and CNS tumor comorbidity, among which glioblastoma (GBM) represents the most aggressive primary malignancy [11]. Our study integrated datasets from the Global Burden of Disease (GBD) and Cancer Incidence in Five Continents (CI5). We applied spatial autocorrelation to define macroscopic geographical clustering patterns, and utilized Granger causality testing to establish chronological sequence of this joint disease burden. Because of the growing number of publications, bibliometrics has been widely recognized as a popular research method [12, 13]. Subsequent bibliometric analysis mapped the evolving research frontiers and thematic hotspots. Mendelian randomization (MR) then provided genetic framework and established a strict causal relationship at the genomic level. The resulting evidence defines a new clinical model for hematopoietic and neurological comorbidity, thus providing groundbreaking insights for future targeted interventions.

Method

Data source and search strategy

Epidemiological data

Macroscopic epidemiological datasets for the present investigation were obtained from the GBD study and the CI5 repositories. Specifically, GBD related metrics were accessed via the GBD Results Tool (http://ghdx.healthdata.org/gbd-results/). By searching with the keywords “Myelodysplastic, myeloproliferative, and other hematopoietic neoplasms”, we extracted epidemiological statistics stratified by age and sex for both target pathologies, which cover 204 countries and territories spanning the period from 1990 to 2021 [14]. Dataset retrieval concluded exactly on August 31, 2025. Moreover, the CI5 database provides highly standardized cancer registration data. We utilized the latest Volume XII from CI5. This specific dataset enabled supplementary calibration and rigorous internal validation. The registry provides statistical indices spanning 2013 to 2017. These metrics originate from 460 local registries across 65 countries. We downloaded the data for the corresponding codes based on the matching keywords in the GBD database.

Literature source and search strategy

The Web of Science Core Collection (WoSCC) serves as the preeminent and comprehensive global citation repository, encompassing highly cited and authoritative journals across diverse scientific fields [15]. The Science Citation Index Expanded (SCIE) edition within WoSCC was selected as the primary data source. We initially queried the WoSCC for MPN + MDS and CNS tumor intersections. The search yielded only 18 records largely compromised by irrelevant studies. Therefore, we abandoned this initial approach and formulated a revised search strategy. The final chronological parameter for document retrieval was restricted between January 1, 2005, and February 15, 2026. Inclusion criteria exclusively targeted original English language “Articles”.

To minimize bias arising from daily database updates, all literature mining procedures were systematically carried out on a single day (February 19, 2026). The resulting metadata were exported in Plain Text format for subsequent computational processing.

GWAS data

The most suitable genome wide association study (GWAS) datasets for MPN and MDS were respectively identified from the GWAS catalog database. Subsequently, the GWAS datasets for MPN and MDS respectively were derived from OpenGWAS and GWAS Catalog, and the GWAS datasets for GBM were retrieved from OpenGWAS and Finngen R11.

Statistical analysis

Geographic stratification and joint incidence patterns of MPN + MDS and CNS tumors

To conduct a systematic assessment of the macroscopic spatial heterogeneity inherent in this joint disease dually dominant model, a stratification protocol based on quartiles was employed. Specifically, the regional incidence rates of both pathologies were separately divided into four distinct levels of disease burden: low (< 25th percentile), lower intermediate (25th to 50th percentile), higher intermediate (50th to 75th percentile), and peak burden (> 75th percentile) [16].

Geographical regions demonstrating the same stratification levels for both MPN + MDS and CNS incidence were categorically labeled as “Consistent units”, which are referred to as dually dominant. In contrast, areas where the epidemiological burden level of MPN + MDS exceeded that of CNS tumors were designated as " MPN + MDS dominant units,” while the opposite geographical situation was classified as “CNS dominant units.” Through this comparative classification framework, the entire global research area was integrated into three distinct geographical phenotypes.

Spatial autocorrelation and comorbidity analysis

We constructed a hybrid spatial weight matrix to evaluate the global comorbidity and clustering patterns. The overall spatial autocorrelation of incidence rates was evaluated using the Global Moran’s I statistic [17]. Mathematical expression is defined as:

(denotes the total number of nations, and represent the incidence rates in regions  i and  j, constitutes the mean incidence, and signifies the spatial weight matrix element)  

We also generated pseudo-P values utilizing Monte Carlo simulations with 999 random permutations to ascertain statistical robustness. The Local Indicators of Spatial Association (LISA) pinpointed the exact topographical loci of spatial clusters [18]. The equation is given below:

( represents the sample variance)

Grounded on spatial significance testing, the geographical typologies were stratified into High-High (H-H) hotspots, Low-Low (L-L) cold spots, High-Low (H-L) spatial outliers, Low-High (L-H) spatial outliers, and nonsignificant regions. We mathematically synthesized a spatial comorbidity index. This metric multiplied the age standardized incidence rates (ASIR) of the two pathologies. Pearson correlation analysis was deployed to evaluate the linear dependencies among MPN + MDS incidence, CNS tumor incidence, and the derived comorbidity index. The two sided indicated statistical significance.

Time Series and Granger causality analysis

Vector autoregression (VAR) models integrated longitudinal incidence datasets from Sweden to delineate the chronological precedence and directional predictive dynamics between the MPN + MDS and CNS tumor ASIRs [19]. We then performed Granger causality assessments stratified by sex. Bootstrap resampling framework ensures robust statistical inferences, which is crucial for avoiding spurious causal relationships[20]. The resampling process constructs empirical P value distributions, yielding the 95% confidence intervals (CIs). Sensitivity analyses confirmed the temporal stability of the identified associations. The analytical design includes the setting of a 7-year temporal lag(lag_7) and the truncation of the recent 5-year observational window(drop_last5). For better visual representation of the data, the resultant P values were logarithmically transformed, with a threshold set at , indicating statistical significance.

Bibliometric and knowledge mapping analysis

VOSviewer (version 1.6.20) was utilized to build robust co-occurrence networks that included core authors, national institutions, and keywords. The dataset size dictated the customized truncation thresholds: the minimum occurrence frequency was set at 20 for the larger MPN + MDS dataset (n = 5034), while it was proportionally adjusted to 10 for the CNS dataset (n = 1961). The specific search formula is detailed in Supplementary Fig. 4. In the generated network mapping, the node sizes were configured to be directly proportional to their occurrence frequencies, and the edge thickness represented the intensity of collaboration or co-occurrence. Simultaneously, the Overlay Visualizations function was employed to intuitively track the interdecadal changes in research hotspots. Moreover, the bibliometrix R package generated global country collaboration maps and trend topics, thus revealing the current landscape of national academic research, alongside the developmental trajectories of scientific interests.

Genetic causal interrogations

In strict accordance with the core principles of Mendelian randomization (MR), three GWAS datasets for MPN and MDS were initially screened to identify genome wide significant single nucleotide polymorphisms (SNPs) with P-value less than 5 × 10⁻⁸. In cases where no such SNPs were detected, a less stringent threshold of P < 5 × 10⁻⁶ was employed as an alternative criterion. Subsequently, linkage disequilibrium (LD) analysis on the selected SNPs was applied by a threshold of kb < 1000 and r² > 0.01. The remaining independent SNPs (nsnp) were then queried for their respective phenotypic traits in the PubMed and OPENGWAS databases to avoid potential confounding factors.

The inverse variance weighting (IVW) method served as the primary analytical approach. The “TwoSampleMR” R package executed this specific algorithm. Sensitivity analyses incorporated heterogeneity testing and horizontal pleiotropy examinations. These rigorous evaluations validated the positive results.

Result

Global spatial distribution and joint incidence patterns

The 2021 GBD datasets mapped the global MPN + MDS burden (Fig. 1Aa). Greece reported the highest ASIR at 20.61573 per 100,000 population. Mali (15.91340) and Norway (15.61429) followed closely. For CNS tumors, a different hierarchy emerged (Fig. 1Ab). Norway led the cohort with an ASIR of 14.96124. Denmark (14.16192) and Monaco (13.91034) occupied the subsequent ranks.

Fig. 1.

Global distribution of the incidence of MPN + MDS and CNS tumors, and characteristics of their joint incidence patterns. (Aa) Incidence of MPN + MDS based on GBD 2021 data. (Ab) Incidence of CNS tumors based on GBD 2021 data. (Ac) Geographical dominant types of the joint disease burden based on GBD 2021 data. (Ad) Incidence of MPN + MDS based on CI5 XII data. In both GBD and CI5, colors tending toward purple indicate higher incidence rate grades, while colors tending toward the lighter end of the scale indicate lower grades. Additionally, in CI5, gray denotes countries for which no statistical data are available. (Ae) Incidence of CNS tumors based on CI5 XII data. (Af) Geographical dominant types based on CI5 XII data. The joint incidence patterns are defined by the specific incidence levels of the two diseases. Red indicates co-dominant areas/countries, purple indicates areas/countries with MPN + MDS predominance, and orange indicates areas/countries with CNS predominance

CI5 Volume XII registry provided a complementary global perspective. Its MPN + MDS incidence map identified Denmark (15.66445) as the primary hotspot (Fig. 1Ad). Malta (14.57193) and Australia (13.30366) completed the top tier. The CI5 CNS map highlighted that Norway peaked again at 7.54046, positioned ahead of Lithuania (7.30633) and Latvia (7.25977) (Fig. 1Ae).

The joint incidence maps from GBD and CI5 Volume XII demonstrate a remarkable global spatial overlap in the incidence rate quartiles, and dually dominant countries account for a significant proportion (Supplementary Fig. 1A). These consistent geographical units carry a heavy burden of disease. Supplementary Tables S1 and S2 catalog the specific nations across three distinct classification patterns. They mainly concentrate in developed regions, including North America, Western Europe, and Australasia (Figs. 1Ac and 1Af). This distinct synchronization in disease burden levels implies the existence of shared environmental exposures, convergent genetic susceptibilities, or consistent regional diagnostic capabilities.

CNS dominant areas localized primarily in East Asia, while MPN + MDS dominant areas were sporadically distributed across South America, the Middle East and Eastern Europe. These divergent patterns emphasize the regional heterogeneity of this joint disease burden. Spatial maps identify geographic distribution overlap.

Decomposition of incidence drivers

To verify the robustness of the identified spatial phenotypes, an intersection analysis across the two databases successfully demarcated 14 highly congruent “Consistent units” regions (Supplementary Fig. 1B). A subsequent decomposition analysis of the incidence determinants within these specific regions indicated that population aging and demographic growth were the fundamental factors driving the overall increase in the burden (Supplementary Fig. 2A). Correlation analyses further confirmed that both population aging (r = 0.7275, P = 0.004) and population growth (r = 0.8110, P < 0.001) showed a highly synchronized positive correlation between the effect sizes of the two pathologies (Supplementary Fig. 2B). Epidemiological transitions did not yield statistically significant associations. These findings strongly suggest that fundamental demographic aging and growth are the most crucial drivers, leading to the concurrent increase in the comorbidity burden across highly vulnerable areas.

Spatial autocorrelation and comorbidity analysis

The scatter plots are divided into four quadrants: High-High indicates high domestic incidence accompanied by high incidence in neighboring countries; High-Low indicates high domestic incidence with low neighboring incidence; Low-Low represents low domestic and neighboring incidence; and Low-High indicates low domestic incidence alongside high neighboring incidence. The Global Moran’s I statistic validated a pronounced positive spatial autocorrelation for both malignancies on a global scale (GBD: MPN + MDS I = 0.5206, CNS I = 0.7136; CI5: MPN + MDS I = 0.185, CNS I = 0.4462). Furthermore, Pearson correlation matrices unveiled robust positive linear associations not only between the incidence rates of MPN + MDS and CNS tumors (GBD r = 0.489; CI5 r = 0.533), but also between these individual baseline rates and the synthesized spatial comorbidity index (r = 0.859–0.884) (Fig. 2A and B). To preclude stochastic artifacts, the analysis incorporated rigorous Monte Carlo permutation tests. The distributions represent the null hypothesis generated via 999 random spatial permutations. The observed Moran’s I values fell entirely outside these simulated random distributions (Fig. 2C). This confirms a highly significant and systematic geographical clustering for both malignancies worldwide. Every dataset achieved strict statistical significance at .

Fig. 2.

Spatial autocorrelation and correlation analysis of the incidence of MPN + MDS and CNS tumors. (A, B) Moran’s I scatter plots and paired disease correlation matrices of GBD and CI5 data. The triangular heatmaps illustrate the Pearson correlation coefficients among MPN + MDS incidence, CNS incidence, and the comorbidity index, deeper red or blue shading indicated stronger correlations. (C) Density distributions of simulated Moran’s I alongside observed statistical values. The vertical red lines indicate the actual observed statistics based on real-world data. The significant deviation of the red lines from the simulated distributions, confirms that the observed spatial clustering is not a random occurrence. (D) Country counts across different LISA spatial cluster categories. (E) Global Local Indicators of Spatial Association cluster maps of MPN + MDS and CNS tumors utilizing GBD and CI5 data. Red regions denote high-incidence spatial hotspots for the diseases, while blue regions denote relatively safe, low-incidence clusters. Grey areas indicate regions lacking statistical significance or where data is missing

Subsequent LISA cluster analyses precisely pinpointed the geographical hotspots and cold spots of the disease burden. Sweden was uniformly identified across both databases as a statistically significant “High-High” clustering epicenter (Fig. 2D and E). Extensive spatial mapping reveals a substantial comorbidity overlap between these two distinct malignancies. Their global epidemiological trajectories synchronize.

Longitudinal epidemiological dynamics within the Swedish cohort

Spatial mapping previously isolated Sweden as a primary dual incidence hotspot. Detailed epidemiological profiling isolates this specific national dataset. Demographic stratification exposes a synchronization between MPN + MDS and CNS tumors across successive age brackets (r = 0.62, P < 0.001). Both malignancies reach the peak within older adult demographics (Supplementary Fig. 3A).

Joinpoint regression models trace three decades of historical incidence fluctuations (Supplementary Fig. 3B). Longitudinal dynamics demonstrate alternating periods of escalation and stabilization for both disease categories. Biological sex influences disease susceptibility. Male cohorts bear a heavier burden across most advanced age groups for both diseases (Supplementary Fig. 3 C and 3D).

Predictive modeling projects future standardized incidence rates until 2036. Bayesian algorithms forecast a gradual future stabilization following historical peaks (Supplementary Fig. 3E and 3 F).

Temporal trends and Granger causality analysis in Sweden

Given the consistent categorization of Sweden as a “High-High” hotspot, we extracted the longitudinal ASIR data from 1990 to 2021 to formed the analytical baseline (Fig. 3A). The curves map incidence trajectories for both diseases. Granger models then evaluated the sequential direction (Fig. 3B). The results revealed a unidirectional predictive dynamic exclusive to females. MPN + MDS incidence exerted a statistically significant Granger causal effect on CNS tumor occurrence within this female population (-log₁₀(P) > 1.301, P < 0.05). In contrast, no such significant associations were detected in the male cohort, nor in the reverse CNS predicting MPN + MDS scenario.

Fig. 3.

Temporal trends and Granger causality analysis of MPN + MDS and CNS tumors in Sweden. (A) Longitudinal incidence trajectories of MPN + MDS and CNS tumors from 1990 to 2021 stratified by sex. (B) Granger causality assessments indicating sequential prediction directions. (C) Primary analysis P values with Bootstrap 95% confidence intervals. (D) Frequency distribution profiles of Bootstrap resampling P values. (E) Significance frequency and proportion of Granger causal outcomes. (F) Robustness checks and sensitivity analyses incorporating temporal lag adjustments and observational window truncations

A bootstrap resampling framework validated the unidirectional association exclusive to females (Fig. 3C). Female myeloid to brain prediction models generate distinct P value distribution profiles (Fig. 3D). Iteration spreads reveal asymmetry. P values accumulate near zero thresholds. Figure 3E extracted the proportion of significant outcomes. The frequency of Granger causal significance in the female ‘myeloid to brain’ direction significantly exceeded the 5% baseline threshold. Collectively, these multiple lines of evidence confirm that this represents a robust statistical causal signal rather than a result of random error.

Subsequent sensitivity analyses confirmed that the significance persisted even after adjusting lag_7; however, this significance was diminished upon the exclusion of drop_last5 (Fig. 3F). This attenuation highlights the contemporary signature of this comorbidity correlation: over the past decade, the widespread implementation of advanced neuroimaging and next generation sequencing (NGS) has significantly reduced underdiagnosis. Modern targeted therapies substantially extend MPN patient overall survival, providing a sufficient time frame for the clinical manifestation of CNS tumors. Competing mortality risks previously concealed concurrent neurological malignancies.

Global research landscapes and knowledge mapping

Supplementary Fig. 4 A and 4B illustrate respectively the literature screening process in both diseases. Global productivity maps and global collaboration networks identify United States and China as dual core hubs in both research domains (Supplementary Fig. 5 A and 5B). Table 1 identifies top individual contributors. Guillermo Garcia-Manero completely monopolizes MPN + MDS publication metrics (TP = 56). United States academics capture nine top 20 positions. William F. Elmquist commands CNS output (TP = 8). Chinese researchers overwhelm sixteen corresponding elite ranks. In the field of MPN + MDS, leading oncological institutions like MD Anderson Cancer Center project massive academic influence. Capital Medical University dominates CNS tumor literature. Total publications reach 46, securing primary institutional rank. However, American institutions, especially Massachusetts General Hospital, maintained a substantial leading position in core academic impact indicators, particularly the average citations per publication (Table 2). The United States held absolute preeminence in both total publications (TP = 1,922) and total citations (TC = 112,339). CNS tumor publications display contrasting results. China overtakes United States research volumes (TP = 664), manifesting remarkable research vigor (Table 3).

Table 1.

The top 20 productive authors within the respective fields of MPN+MDS and CNS tumor research from 2005 to February 15, 2026. This assessment specifically utilized three key bibliometric metrics: total publications (TP), total citations (TC), and average citations per publication (TC/TP)

	MPN+MDS						CNS tumor
	Rank	Author name	Country	TP	TC	TC/TP	Rank	Author name	Country	TP	TC	TC/TP
Authors	1	Garcia-Manero, Guillermo	The United States	56	5384	96.14	1	Elmquist, William F.	The United States	8	261	32.63
	2	Maciejewski, Jaroslaw P.	The United States	52	4939	94.98	2	Chen, Qianxue	China	8	213	26.63
	3	Li, Xiao	China	51	1081	21.20	3	Chen, Xin	China	8	157	19.63
	4	Haferlach, Torsten	Germany	47	2615	55.64	4	Deckert, Martina	Germany	8	138	17.25
	5	Cazzola, Mario	Italy	43	4643	107.98	5	Montesinos-Rongen, Manuel	Germany	8	138	17.25
	6	Levine, Ross L.	The United States	42	6113	145.55	6	Zhang, Rui	China	7	108	15.43
	7	Guo, Juan	China	42	865	20.60	7	Liu, Ying	China	6	342	57.00
	8	Kantarjian, Hagop	The United States	38	7547	198.61	8	Zhang, Li	China	6	301	50.17
	9	Abdel-Wahab, Omar	The United States	38	6395	168.29	9	Sun, Qian	China	6	216	36.00
	10	Ogawa, Seishi	Japan	38	1603	42.18	10	Wang, Shu	China	6	153	25.50
	11	Tefferi, Ayalew	The United States	37	3647	98.57	11	Zhang, Wei	China	6	142	23.67
	12	Chang, Chunkang	China	37	685	18.51	12	Zhang, Jian	China	6	112	18.67
	13	Sekeres, Mikkael A.	The United States	35	2394	68.40	13	Brunn, Anna	Germany	6	105	17.50
	14	Ebert, Benjamin L.	The United States	34	7001	205.91	14	Liu, Baohui	China	6	81	13.50
	15	Verstovsek, Srdan	The United States	34	3787	111.38	15	Li, Xiang	China	6	76	12.67
	16	Malcovati, Luca	Italy	33	3412	103.39	16	Wang, Wei	China	6	60	10.00
	17	Zhang, Zheng	China	33	666	20.18	17	Wang, Yu	China	6	15	2.50
	18	Fenaux, Pierre	France	32	2479	77.47	18	Chen, Jian	China	5	240	48.00
	19	Boultwood, Jacqueline	The United Kingdom	31	3333	107.52	19	Zhang, Yi	China	5	200	40.00
	20	Haferlach, Claudia	Germany	31	1387	44.74	20	Gao, Lun	China	5	198	39.60

Abbreviations: MPN= myeloproliferative neoplasms, MDS= myelodysplastic syndromes, CNS tumor= central nervous system tumor

Table 2.

The top 20 productive institutions within the respective fields of MPN+MDS and CNS tumor research from 2005 to February 15, 2026. This assessment specifically utilized three key bibliometric metrics: total publications (TP), total citations (TC), and average citations per publication (TC/TP)

	MPN+MDS						CNS tumor
	Rank	Institution	Country	TP	TC	TC/TP	Rank	Institution	Country	TP	TC	TC/TP
Institutions	1	Univ Texas MD Anderson Canc Ctr	The United States	186	15701	84.41	1	Capital Med Univ	China	47	599	12.74
	2	Mem Sloan Kettering Canc Ctr	The United States	171	18855	110.26	2	Fudan Univ	China	39	726	18.62
	3	Cleveland Clin	The United States	119	7038	59.14	3	Mem Sloan Kettering Canc Ctr	The United States	37	3531	95.43
	4	Mayo Clin	The United States	111	6669	60.08	4	Southern Med Univ	China	34	471	13.85
	5	Harvard Univ	The United States	101	14665	145.20	5	German Canc Res Ctr	Germany	31	692	22.32
	6	Harvard Med Sch	The United States	94	3996	42.51	6	Shanghai Jiao Tong Univ	China	31	653	21.06
	7	Univ Pavia	Italy	92	10968	119.22	7	Harvard Med Sch	The United States	28	1596	57.00
	8	Dana Farber Canc Inst	The United States	91	8632	94.86	8	Nanjing Med Univ	China	28	756	27.00
	9	Univ Tokyo	Japan	87	6812	78.30	9	Massachusetts Gen Hosp	The United States	24	3014	125.58
	10	Shanghai Jiao Tong Univ	China	85	1779	20.93	10	Univ Texas MD Anderson Canc Ctr	The United States	24	2208	92.00
	11	H Lee Moffitt Canc Ctr & Res Inst	The United States	79	3861	48.87	11	Soochow Univ	China	24	779	32.46
	12	Washington Univ	The United States	77	8564	111.22	12	Chinese Acad Sci	China	24	513	21.38
	13	Hannover Med Sch	Germany	77	4596	59.69	13	Univ Calif San Francisco	The United States	23	2037	88.57
	14	Univ Chicago	The United States	74	4706	63.59	14	Mayo Clin	The United States	23	976	42.43
	15	Johns Hopkins Univ	The United States	72	5190	72.08	15	Shandong Univ	China	23	473	20.57
	16	Univ Freiburg	Germany	70	3046	43.51	16	China Med Univ	China	23	440	19.13
	17	Univ Florence	Italy	68	6061	89.13	17	Univ Calif Los Angeles	The United States	22	1619	73.59
	18	NCI	The United States	68	3890	57.21	18	Harvard Univ	The United States	22	1444	65.64
	19	Univ Washington	The United States	67	5308	79.22	19	Johns Hopkins Univ	The United States	22	1250	56.82
	20	Northwestern Univ	The United States	66	2438	36.94	20	Univ Toronto	Canada	22	709	32.23

Abbreviations: MPN= myeloproliferative neoplasms, MDS= myelodysplastic syndromes, CNS tumor= central nervous system tumor

Table 3.

The top 20 productive countries within the respective fields of MPN+MDS and CNS tumor research from 2005 to February 15, 2026. This assessment specifically utilized three key bibliometric metrics: total publications (TP), total citations (TC), and average citations per publication (TC/TP)

	MPN+MDS					CNS tumor
	Rank	Country	TP	TC	TC/TP	Rank	Country	TP	TC	TC/TP
Countries	1	The United States	1922	112339	58.45	1	China	664	14172	21.34
	2	China	837	13988	16.71	2	The United States	595	29287	49.22
	3	Germany	536	31275	58.35	3	Germany	156	7199	46.15
	4	Italy	473	27475	58.09	4	Japan	106	4242	40.02
	5	Japan	472	17632	37.36	5	Italy	105	5828	55.50
	6	The United Kingdom	368	27904	75.83	6	The United Kingdom	88	5151	58.53
	7	France	357	23012	64.46	7	Canada	85	2313	27.21
	8	Spain	204	9810	48.09	8	France	74	4739	64.04
	9	Canada	156	7601	48.72	9	South Korea	60	4076	67.93
	10	Netherlands	143	7833	54.78	10	Spain	60	2980	49.67
	11	Brazil	129	2675	20.74	11	Brazil	51	942	18.47
	12	South Korea	126	2537	20.13	12	Australia	45	2944	65.42
	13	Switzerland	119	6417	53.92	13	Poland	40	1063	26.58
	14	Sweden	116	9548	82.31	14	Taiwan	38	1491	39.24
	15	Australia	116	6171	53.20	15	Netherlands	33	1250	37.88
	16	Austria	108	8855	81.99	16	India	28	502	17.93
	17	Denmark	92	4365	47.45	17	Sweden	27	1272	47.11
	18	Belgium	90	10845	120.50	18	Switzerland	27	914	33.85
	19	Czech Republic	90	2344	26.04	19	Iran	22	374	17.00
	20	Greece	80	4446	55.58	20	Belgium	20	850	42.50

Abbreviations: MPN= myeloproliferative neoplasms, MDS= myelodysplastic syndromes, CNS tumor= central nervous system tumor

Bibliographic networks revealed distinct thematic landscapes. MPN data and MDS data segregated into three clear clusters (Fig. 4A). Red nodes grouped primary clinical syndromes. Green nodes isolated key genetic mutations including JAK2. Blue nodes represented treatment resistance. Meanwhile, CNS literature formed five specialized sectors (Fig. 4B). The purple module centered entirely on glioblastoma. Green nodes captured neuroinflammation. Yellow nodes detailed cellular invasion. Remaining network sections highlight critical prognostic biomarkers.

Fig. 4.

Keyword co-occurrence networks and chronological evolution of research frontiers. (A) Keyword co-occurrence network of MPN + MDS research with frequencies higher than 20. (B) Keyword co-occurrence network of CNS tumor research with frequencies higher than 10. (C) Overlay visualization of MPN + MDS research keywords. (D) Overlay visualization of CNS tumor keywords. Node sizes directly reflect occurrence frequencies. The node colors indicate the average publication year of the keywords, transitioning from blue to red. The color bar displayed in the lower right corner of the subplots

Keyword co-occurrence profiling revealed the interdecadal shifts of research frontiers. For the MPN + MDS group, the focus of research has vigorously shifted from traditional cytogenetics to precision diagnostics and genomic profiling, as exemplified by keywords such as “mutation” and “next generation sequencing“(Fig. 4C). CNS tumor group transcended basic apoptosis research and aggressively pivoted towards the remodeling of the immune microenvironment, as emphasized by terms like “neuroinflammation,” “microglia,” and “tumor microenvironment“(Fig. 4D).

More notably, the trend topic analysis deciphered a distinct convergence of core therapeutic targets (Supplementary Fig. 5 C and 5D). The specific disease entities “myelodysplastic syndromes (MDS)” and “glioblastoma (GBM)” emerged as highly prominent and persistent primary themes within their respective groups. The strong chronological convergence of MDS and GBM as research centers points towards potential connection.

Genetic causal association between MPN and MDS with GBM

As depicted in Fig. 5A, GCST90000032 was utilized as the exposure instrument for MPN (P = 0.004), whereas GCST90042662 was designated for MDS (P = 0.008, after the deliberate exclusion of the insignificant GCST90043940 dataset).

Fig. 5.

Two-sample mendelian randomization analysis evaluating the genetic causal associations between MPN and MDS with GBM. (A) Selection and significance screening of GWAS datasets for MPN and MDS exposures. (B) Analytical framework assessing the causal effect of MPN on GBM. (C) Analytical framework assessing the causal effect of MDS on GBM. Red text highlights rs7705526 as a shared instrumental variable across both models. (D) Forest plot illustrating the final causal estimates and corresponding confidence intervals

Quality control pipelines evaluate structural integrity (Fig. 5B and C). Heterogeneity remains absent (P = 0.1568). Analytical procedures eliminate horizontal pleiotropy concerns (P = 0.9107), fully meeting the core methodological assumptions of the MR framework. Notably, the genetic variant rs7705526 emerged as a shared IV across both causal models. Supplementary Table S3 catalogs the clinical annotations and specific gene consequences for selected variants, ensuring the exclusion of confounding pleiotropic factors.

The resultant forest plots provided compelling genetic evidence (Fig. 5D). MPN significantly increased the risk of developing GBM (odds ratio [OR] = 1.84, 95% CI: 1.22–2.76) MDS genetic predispositions trigger even heavier pathogenic consequences (OR = 2.54, 95% CI: 1.62–3.96).

Discussion

The observed spatial overlap may be partially driven by confounding due to shared demographic and healthcare-related factors. To minimize such bias, we utilized two independent large-scale global databases (GBD and CI5) and performed rigorous cross-validation. Importantly, our Mendelian randomization analysis provides genetic-level evidence supporting a genuine biological relationship that extends beyond mere ecological correlation.

Our multidimensional assessments clarify that these paired malignancies not only exhibit highly synchronized spatial clustering and demographically influenced trajectories on a global scale but also display a profound chronological correlation specifically within the female population. Global spatial distributions reveal consistent overlaps. The findings highlight a compelling consistent pattern: the geographical areas with high incidences of both MPN + MDS and CNS tumors show significant converge, mainly concentrating in developed regions including North America, Western Europe, and Australasia. Spatial configurations physically manifest latent diagnostic and demographic double hits.

Decomposition algorithms isolate primary epidemiological drivers. Demographic aging and population growth propel parallel disease burden escalation in these consistent regions. Global life expectancy continually extends. Elderly populations face prolonged carcinogen exposure alongside natural physiological decline. These cumulative effects significantly drive up the overall cancer incidence [21].

These marked regional disparities also strongly reflect socioeconomic influences. Development levels directly dictate disease detection capabilities. Highly developed countries have more comprehensive cancer registries and better access to advanced healthcare resources. Conversely, the seemingly low incidence rates reported in some underdeveloped regions may merely conceal the actual underlying disease burden [22, 23].

MPN + MDS incidence significantly elevates CNS tumor risk. This specific dynamic exists exclusively among females. Notably, this association heavily depends on contemporary epidemiological data from the most recent period of five years. Survival outcomes reveal unexpected epidemiological paradoxes. Advanced targeted therapies like Janus kinase (JAK) inhibitors, have significantly prolonged the overall survival of MPN patients. Paradoxically, this overall reduction in competing mortality risks provides patients with sufficient time to be exposed to and subsequently develop independent primary or synchronous CNS tumors [2].

Integrating these relationships with the established shared genetic susceptibility loci between gliomas and various pan cancer entities, we thus aroused myeloid-neuro-immune crosstalk paradigm into the specific MPN and GBM comorbidity landscape [10, 24, 25]. Chronic MPN and MDS frequently terminate in acute myeloid leukemia (AML). AML exhibits striking pathobiological similarities with glioblastoma (GBM) [26]. Isocitrate dehydrogenase (IDH) mutations represent signature shared genetic drivers. Mutant IDH enzymes act as extreme initiators driving MPN + MDS transformation into AML. Identical IDH aberrations characterize primary glioma pathogenesis [27–29]. The key driving event in MPN pathogenesis is the abnormal hyperactivation of the JAK-STAT signaling pathway. Mutated pathways force malignant peripheral myeloid cell proliferation. Moreover, the common TERT SNP is a well-established determinant of inter-individual telomere length as repeatedly identified in GWAS. In MPN, the germline predisposition toward longer telomeres extends the replicative lifespan of hematopoietic stem cells, creating a permissive state for the acquisition of driver mutations and subsequent clonal expansion. In GBM, although the same germline variant contributes to overall longer telomeres, somatic TERT promoter mutations often arise as a convergent mechanism to maintain telomere homeostasis during gliomagenesis. Thus, this germline TERT variant may serve as a shared “genetic soil” for both diseases. Clonal hematopoiesis simultaneously triggers a severe cytokine storm [30]. The peripheral circulation accumulates high levels of interleukin-6 (IL-6), tumor necrosis factor-α (TNF-α), and interleukin-1β (IL-1β) [5, 31]. This systemic inflammation directly attacks the tight junctions of cerebral capillary endothelial cells, thereby causing a pathological increase in blood-brain barrier (BBB) permeability [7, 32].

Along with BBB leakage, abnormally mutated peripheral monocyte/macrophage populations, together with macromolecular proinflammatory mediators, can freely infiltrate the central nervous system parenchyma [33]. Core chemokine receptor CXCR4 dictates this specific cellular trafficking. CXCR4 gradients actively guide abnormal peripheral stem cells and monocytes across compromised physical barriers [11]. In this compartment, activated microglia naturally reside within this specific brain compartment, which are inherently part of the myeloid macrophage lineage [34, 35]. Microglia engage in extensive crosstalk with the invading peripheral immune cells [36]. Together, they are likely reprogrammed into a tumor promoting phenotype, such as M2 polarization. This exact transformation creates a highly favorable oncogenic niche. It simultaneously builds a robust angiogenic microenvironment perfectly suited for highly aggressive GBM [33, 37].

When synthesizing these insightful findings from the perspective of holistic integrative medicine, the development of an interdisciplinary, comprehensive management strategy becomes an urgent clinical necessity. The conventional, fragmented, and siloed single-discipline paradigm of care is fundamentally inadequate to navigate the intricate complexities of this “hematologic-neurologic” comorbidity. Throughout routine hematological surveillance, clinicians ought to maintain a heightened index of suspicion for any neurological manifestations. However, the inherent dependence on statistical modeling in global epidemiological databases may inadvertently underestimate the actual comorbidity burden in resource limited countries.

This study focused exclusively on the association between MPN and GBM, without systematically exploring other tumor types, which represents an inherent limitation of our observational exploration to find the causal relationship. Nevertheless, we recognize that investigating potential links between MPN and other malignancies is of considerable value. Future studies will address this by expanding the tumor spectrum and performing more comprehensive association analyses.

Conclusion

MPN + MDS and CNS tumors represent formidable global public health challenges. Our finding could serve as a reference for the development of public health strategies. It lays a solid foundation for future research into the interactions between MPN + MDS and the CNS, particularly in the context of glioblastoma, and also offers insights for building multidisciplinary team collaboration.

Competing Interests

No potential conflict of interest was reported by the authors.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

MPN

Myeloproliferative neoplasms

MDS

Myelodysplastic syndromes

CNS

Central nervous system

GBM

Glioblastoma

HR

Hazard Ratio

CI

Confidence interval

GBD

Global Burden of Disease

CI5

Cancer Incidence in Five Continents

GWAS

Genome-Wide Association Studies

BBB

Blood-Brain Barrier

LISA

Local Indicators of Spatial Association

MR

Mendelian Randomization

TC

Total Citations

TP

Total Publications

Author contributions

HY contributed to the methodology, conceptualized and supervised the overall study, participated in the methodology design and software-based formal analysis, and critically reviewed and revised the manuscript.SH curated the relevant data, performed the formal analysis using the analytical software, and drafted the original manuscript and contributed to the revision. LH and WL funded the research. All authors read and approved the final manuscript.

Funding

None.

Data availability

The GBD data used in this study is publicly available at http://ghdx.healthdata.org/. The CI5 data used in this study is publicly available at CI5: Cancer Incidence in Five Continents. The WoS literature source data that support the findings of this study are available from Web of Science. Researchers can freely access all analyzed genome wide association datasets directly through the official FinnGen and IEU OpenGWAS repositories.

Declarations

Competing interests

The authors declare no competing interests.

Consent for publication

Not applicable.

Ethics approval

Original primary study investigators previously secured all necessary institutional ethical approvals and baseline informed patient consent. This current study conducts secondary analyses utilizing completely anonymized public summary statistics. Formal institutional review board exemption therefore applies.