Quantitative, longitudinal measurements of response to immunotherapy for glioblastoma using multiparametric MRI in a long-term survivor treated with autologous tumor lysate–loaded dendritic cell vaccine: an illustrative case Authors: Archith Rajan, Steven Brem, Daniel Gewolb, Ajay Kumar, Laiz Laura de Godoy, Arati Desai, MacLean P Nasrallah, Timothy Lucas, Laurie A Loevner, Suyash Mohan, Sanjeev Chawla Published: 2026 ABSTRACT AbstractBACKGROUNDNeurosurgeons are in the forefront of developing clinical trials that harness the immune system to combat glioblastoma. Autologous tumor lysate–loaded dendritic cell vaccine (DCVax-L) is a novel immunotherapy for glioblastoma. An unmet need is to develop a sensitive, specific, feasible, clinically relevant, neuroimaging biomarker that accurately reflects response to immunotherapy.OBSERVATIONSThis illustrative case involves a 63-year-old female who survived for 64 months following near-total resection of a glioblastoma and enrollment in a clinical trial (NCT00045968) using DCVax-L. A multivariate logistic regression model was used to assess treatment response longitudinally at 11 time points measuring predictive probability (PP) of tumor progression based on diffusion tensor imaging– and perfusion-weighted imaging–derived parameters within the contrast-enhancing region. Lesions were considered true progression (TP) if the PP was ≥ 50% and pseudoprogression (PsP; predominant treatment effects) if the PP was < 50 %. These findings were subsequently validated by histopathological analysis during resection for tumor recurrence.LESSONSThese findings support the use of multiparametric MRI to monitor the effects of immunotherapy in patients with glioblastoma. Such multiparametric prediction models using noninvasive, advanced neuroimaging may be helpful to differentiate TP from PsP in longitudinal assessment of treatment response. https://thejns.org/doi/10.3171/CASE26655 Keywords: biomarker, dendritic cell vaccine, glioblastoma, immunotherapy, multiparametric MRIABBREVIATIONS: CL = coefficient of linear anisotropy, CP = coefficient of planar anisotropy, CS = coefficient of spherical anisotropy, DCVax-L = autologous tumor lysate–loaded dendritic cell vaccine, DSC = dynamic susceptibility contrast, DTI = diffusion tensor imaging, FA = fractional anisotropy, KPS = Karnofsky Performance Status, MD = mean diffusivity, mpMRI = multiparametric MRI, mRANO = modified RANO, PP = predictive probability, PsP = pseudoprogression, PWI = perfusion-weighted imaging, RANO = response assessment in neuro-oncology, rCBV = relative cerebral blood volume, TP = true progression Neurosurgeons are in the forefront of developing novel immunotherapeutic approaches to the treatment of glioblastoma, using various neurosurgical techniques that “take the fight to the tumor,”1 for example, viral oncolytics,1,2 convection-enhanced delivery,1,3 implanted reservoirs for CAR-T cell therapy,1,4,5 or focused ultrasound.1,6 Vaccination with tumor-pulsed dendritic cells amplifies neoantigens and shows promise as an effective novel therapy for glioblastoma7,8 and other cancers.9 A large, multicenter, phase 3, prospective, externally controlled cohort trial showed extension of overall survival for both newly diagnosed and recurrent glioblastoma; remarkably, there was a meaningful increase in the long-term tail of the survival curves,7 with an increase of 5-year survival rate to 13% compared to 5% with temozolomide alone,10,11 consistent with an immune memory mechanism.7 Apart from consecutive stereotactic biopsies of the tumor microenvironment to discern changes in the immune cellular landscape between “responders” and “nonresponders,”12 there is a strong, unmet need to develop real-time, noninvasive neuroimaging that can assess response to therapy, provide guidance to clinical decision-making to evaluate efficacy while a trial is in progress, and differentiate true progression (TP) from pseudoprogression (PsP) due to the infiltration of T cells, necrosis, and local inflammation at the tumor site.13 Because long-term survival is strongly correlated with delay or prevention of recurrence,14 the difference between TP and PsP is not trivial and is a major cause of patient anxiety, termed “scanxiety,” underscoring the need for a robust model of response assessment.15 During the past decade, our group has developed, refined, and validated a multiparametric MRI (mpMRI) model that integrates diffusion and perfusion metrics.16–18 This model has proven useful in distinguishing between TP and PsP in patients with glioblastoma, including those receiving cellular immunotherapy.19,20 In this retrospectively analyzed report, we describe in detail the longitudinal mpMRI features of a long-term survivor (overall survival 64 months), the distinct imaging features of a durable response, while addressing the challenges of interpreting immunotherapy-induced response patterns over this long follow-up period. Illustrative Case The clinical milestones and mpMRI evaluations of the long-term survivor are summarized in Fig. 1. In brief, a 63-year-old female, previously healthy, presented with intermittent numbness of the palate, transient speech arrest, dysphagia, and a sensation of a pulsating tongue. Initially treated for presumed seizures, she became aphasic with left facial weakness. An emergency MRI study revealed a 3.1-cm, necrotic, heterogeneously enhancing mass in the left lateral pre- and postcentral gyri (Fig. 2). She underwent a near-total resection and tumor tissue was harvested for enrollment in the autologous tumor lysate–loaded dendritic cell vaccine (DCVax-L) trial (NCT00045968). The pathology report revealed the histological and molecular features of an IDH-wildtype glioblastoma with a positive MGMT methylated promoter. The EGFR was amplified; the tissue was negative for EGFRvIII. Postoperatively, her speech improved, her Karnofsky Performance Status (KPS) score was 90%, and after initiating chemoradiation therapy (Stupp protocol),10 the patient was enrolled in the DCVax-L trial. She received a total of 10 intradermal injections of DCVax-L in the arm (0.15 ml per injection) over a period of 27.5 months. She also completed 12 cycles of adjuvant temozolomide with stable neurological deficits, a KPS score of 90%, and without MRI evidence of progression. By month 37 postoperatively, MRI showed a slowly increasing T2-FLAIR signal with small areas of heterogeneous enhancement in the left periatrial white matter (Fig. 3). There was no elevation of relative cerebral blood volume (rCBV), favoring treatment-related change (PsP) over TP. Study schema for serial imaging analysis. CER = contrast-enhancing region; GBM = glioblastoma; DOD = date of death; M = month; WK = week. A: Axial scans acquired 4 days prior to the initial resection, revealing a heterogeneously enhancing mass with a cystic region/necrosis within the posterior left frontal lobe (arrow). B: Axial scans acquired 1 month postresection showing postsurgical changes with minimal residual tumor (arrow). C: Histopathological H&E stain obtained from the initial resection, confirming the diagnosis of glioblastoma with > 50% viable tumor along with a subset of multinucleated cells and areas of myxoid background. A: Serial axial images showing the time course of changes in the contrast-enhancing region (CER; arrow) and T2-FLAIR signal abnormalities. Blue arrows indicate the contrast-enhancing region. A consistent decrease in the lesion size associated with DCVax-L is demonstrated, with an absence of measurable lesion from month (M) 27 to M37. This is followed by a recurrence and progression of the tumor after M41. B: Longitudinal changes in contrast-enhancing region volumes are expressed as percentage change with respect to the baseline. A steady decrease in the volume associated with the immunotherapy treatment was observed, followed by a steady increase linked to tumor progression. Blue arrows indicate time points of DCVax-L administration, and orange arrows indicate time points of repeat resection. Yellow triangles represent evaluable serial MRI time points (n = 8), and pink triangles indicate three time points at which the lesion showed no visible contrast enhancement. All MR images were acquired on a 3T MR scanner with a ­12-channel, phased-array head coil. In addition to standard sequences, additional diffusion tensor imaging (DTI) and dynamic susceptibility contrast (DSC) perfusion-weighted imaging (PWI) sequences were acquired using previously described methods.16–18,21 DTI data were analyzed for computing various pixel-wise, quantitative parametric maps such as mean diffusivity (MD), fractional anisotropy (FA), coefficient of linear anisotropy (CL), planar anisotropy (CP), and spherical anisotropy (CS). A previously validated multiparametric logistic regression model16 consisting of FA, CL, and rCBVmax from contrast-enhancing regions of tumor was applied to estimate the predictive probability (PP) of TP at each time point using the following regression equation: where β0 = −16.17, β1 = 194.01, β2 = −285.65, and β3 = 1.21. Lesions were considered TP (predominantly viable tumor) if the PP was ≥ 50% and PsP (predominantly treatment effects) if the PP was < 50%. The accuracy of the mpMRI model was evaluated by comparing its MR findings (TP versus PsP) at each of the follow-up points with a “ground truth” comparator established by either histopathological analysis or the modified response assessment in neuro-oncology (mRANO)22 and RANO 2.023 criteria. Tumor specimen was available twice, and we used previously reported criteria to establish TP (> 25% of the section containing malignant viable cells) or PsP (< 25% viable malignant cells in the context of treatment-related changes).24,25 For the other MRI readings (n = 6), we compared the mpMRI prediction model with the updated mRANO and RANO 2.0 criteria,22,23 the current gold standard for response assessment in neuro-oncology. The overall survival was 64 months from the initial resection, 60 months from initiation of the DCVax-L vaccine. Longitudinal follow-up MR images (n = 11) were acquired over 57 months, with the baseline imaging (month 0) obtained 3 months after initial resection and 2 weeks prior to the first DCVax-L dose (Fig. 1). The longitudinal trajectory of the contrast-enhancing region volume over eight evaluable time points is shown in Fig. 3B. Advanced imaging metrics at a key time point preceding the second resection (month 42), along with the corresponding histopathological stain at month 48, are shown (Fig. 4). Longitudinal percentage changes in diffusion and perfusion metrics are shown in Fig. 5A, as well as the corresponding multiparametric model–derived PP values in Fig. 5B. A: DTI- and DSC-PWI–derived metrics at M42, the imaging time point immediately preceding the second resection (M48). Blue and white arrows indicate the contrast-enhancing region. Multiparametric MRI–based predictive model (rCBVmax = 4.14, FA = 0.14, CL = 0.05) from the contrast-enhancing region suggested a diagnosis of TP with a PP of 59.76%. B: Pathological H&E stain obtained from the second resection at M48 (left), confirming recurrent/residual glioblastoma with infiltration (40% viable tumor), palisading necrosis, and giant cells. A subsequent stain obtained from the resection at M53 (right) also showed recurrent/residual glioblastoma (30% viable tumor) with giant cells, abundant macrophages, and gliosis. PC = postcontrast. A: Percentage changes in DSC-PWI– and DTI-derived metrics at follow-up periods from baseline (M0). B: The time course of the PP of tumor progression is plotted. Lesions were considered TP if the PP was ≥ 50% and PsP if the PP was < 50%. Quantitative analyses for M27–M37 were not performed due to the absence of contrast-enhancing lesion. The dashed red line denotes a 50% PP value, which we propose as a threshold value to differentiate between PsP (< 50%) and TP (≥ 50%). During the administration of DCVax-L (months 2–27), serial imaging demonstrated a continual decrease in the volume of the contrast-enhancing mass, resulting in complete disappearance by month 27, 1 month prior to the final vaccine dose at month 28 (Fig. 3). By the 6th month, the volume of the contrast-enhancing region had dropped 74.6% from baseline. No measurable contrast-enhancing lesion was seen at subsequent time points (months 31 and 37), with the complete response stable for approximately 10 months overall. At month 41, however, contrast enhancement reappeared in a similar location of the tumor, expanding steadily until month 57, ultimately reaching a five- to sixfold expansion (567%) of the original baseline volume (Fig. 3). While individual imaging metrics showed limited stand-alone interpretability, there were consistent longitudinal trends that correlated with clinical status. During the early treatment window (baseline to the 6th month), rCBVmax showed a steady decline, followed by a sustained increase concomitant with completion of the DCVax-L vaccine (months 41–57). Concurrently, the FA and CL increased over this time, with larger changes (> 50%) observed after the final dose. The parameters MD and MDmin demonstrated reductions during the same time intervals (Fig. 5A). The mpMRI prediction model accurately classified the disease state as PsP or TP at each of the evaluable time points (n = 8). During DCVax-L administration (baseline to the 6th month), PP values remained < 50% (for baseline, month 2, month 4, and month 6, the respective PP values were 2.3%, 28.9%, 7.7%, and 9.5%), consistent with treatment effect. The low PP values at months 4 and 6 were harbingers of subsequent radiographic resolution. Conversely, from the onset of tumor recurrence (month 41), the PP values consistently exceeded 50%, suggesting a threshold value indicating TP (months 41, 42, 51, and 57, the PP respective values were 94.1%, 59.8%, 99%, and 99.5%). TP was validated by histological confirmation of tumor specimens at the second (month 48) and third (month 53) resections (Fig. 3B). Informed Consent The necessary informed consent was obtained in this study. Discussion Observations This illustrative case involves a 63-year-old female who survived for 64 months following near-total resection of glioblastoma and enrollment in a clinical trial (NCT00045968) using DCVax-L. A multivariate logistic regression model was used to assess treatment response longitudinally at 11 time points measuring the PP of tumor progression based on DTI- and PWI-derived parameters within the contrast-enhancing region. Lesions were considered TP, if the PP was ≥ 50% and PsP (predominant treatment effects) if the PP was < 50%. These findings were subsequently validated by histopathological analyses from the tumor specimens collected during the follow-up tumor resections. During the entire 64-month course of this patient, a long-term survivor (> 5 years), treated with DCVax-L immunotherapy, the mpMRI-based prediction model correctly classified new enhancing lesions as either PsP or TP (recurrence). Differentiating PsP from TP constitutes a major decision point in patient management and prognosis; physicians can choose to start an alternative therapy, transition to a new clinical trial, consider a repeat resection, or stay the course if there is evidence of PsP and the expectation that the lesion will resolve with time or with the aid of steroids, bevacizumab or laser interstitial thermal therapy.26 The ability to have a sensitive, specific, reproducible, clinically relevant, neuroimaging biomarker would greatly accelerate the introduction of novel therapies, particularly immunotherapy, in which standard metrics such as progression-free survival can be confounded by immune infiltrates.7 The phase 3 DCVax-L trial took > 15 years to complete, at a cost of hundreds of millions of dollars, with the main endpoint of overall survival.7 The ability to have an mpMRI prediction model could provide dynamic feedback, to make “go/stop” decisions within a clinical trial and inform physicians to enhance vaccine trials with a combination of immunomodulatory agents, such as cytokines,27 checkpoint inhibitors,2,28 toll-like receptor agonists,29 or tumor treating fields.30The addition of DCVax-L to standard of care opens new avenues in the management of glioblastoma, which is a highly lethal disease with a nearly 100% recurrence rate and poor overall survival.7,8 Immune approaches have been limited, in part, because of the extraordinary heterogeneity of antigens in glioblastoma, an immunologically “cold” tumor.7 DCVax-L is a personalized active immunotherapy as use of the patient’s own tumor lysate primes the dendritic cells, enabling the T cells to target the spectrum of antigens expressed by the patient’s tumor.7,8 To eliminate solid tumors with immunotherapy, it is important that the CD4+ T cells engage with the CD8+ T cells on the same dendritic cells, forming a three-cell cluster (triad) that reprograms otherwise ineffective, dysfunctional CD8+ T cells to become effective, cytotoxic cells that eliminate cancer cells.31 With advances in nucleic acid sequencing and bioinformatics, the next generation of dendritic cell vaccines could be personalized formulations incorporating neoantigens arising from genomic alterations; immune monitoring at the population and single-cell level is currently in development.32 The ideal assay to differentiate PsP from TP and distinguish responders from nonresponders would be to perform serial, longitudinal, stereotactic biopsies and interrogate the immune microenvironment,12 but this would not be feasible for a long-term survivor and carries a concern for sampling errors or risk of injury. We propose a testable hypothesis: using the mpMRI model, a PP ≥ 50% reflects TP and, conversely, values < 50% reflect PsP (Fig. 5B). To validate this hypothesis, larger, multicenter clinical trials are needed. There are additional limitations, particularly the reliance on a single illustrative case to suggest a potential paradigm shift in monitoring immunotherapy response in glioblastoma. The immune system and the biological response to glioblastoma growth are a dynamic process, and without the full spectrum of “omic” analysis from genetic, biochemical, and cellular data,12,33the MRI data remain a key surrogate. The earliest changes in oncogenesis are invisible to even the most sensitive imaging modality.33,34 Other quantitative mpMRI methods are being developed.35 The integration of mpMRI with other technologies such as artificial intelligence/machine learning,36 cell-free DNA/“liquid biopsy,”37 and amino acid–based PET scanning26,38,39 can complement mpMRI. Future work will determine how scalable and generalizable the approach is to the management of patients with glioblastoma and stratification for clinical trials. Until then, RANO 2.0,23,40 with its inherent limitations, remains the gold standard for MRI assessment of responses to novel therapies, including immunotherapy. Furthermore, the patient had favorable prognostic features, including MGMT promoter methylation, a high KPS score, near-total resection, completion of chemoradiation therapy (temozolomide), and the absence of comorbidities, so the long-term survival could be due to a resilient, physiological immune system41 rather than the vaccine effect per se. However, our patient had these favorable genetic, molecular, and clinical features throughout her illness, so the key takeaway message is the correlation of the PP values with the clinical course, as cycle of response to progression was heralded by the rise in the PP value. The search for a real-time, rapid, reliable imaging biomarker could transform clinical trial development.15 Large, contemporary, population-based studies have estimated the 5-year survival for glioblastoma to be around 4%;11by contrast, in the multicenter trial for the dendritic cell vaccine, the 5-year survival rose to 13%.7 The immunosuppressive microenvironment of glioblastoma acts as an “immunostat” consisting of cytokines, immune-checkpoint proteins, and other factors that dampen or disable antitumor immune responses. Overcoming this barrier is critical to developing an effective and broadly generalizable vaccine applicable to most patients. However, the mpMRI scale highlighted in this case presentation could be an important tool in the armamentarium of the neuro-oncologist. One of the great achievements of modern medicine has been the development of antiretroviral therapy, which transformed the outlook for a patient with HIV infection from 85% mortality in 5 years to a life expectancy approaching that of the general population.42 The development of effective combination therapies was guided by readily available blood-based biomarkers (CD4 counts, quantitative HIV RNA testing viral load).42 Until liquid biopsies are developed further, there should be an effort to validate advanced neuroimaging techniques to monitor, predict, and better assess response to novel therapies, particularly immunotherapy. Lessons We report a patient with IDH-wildtype glioblastoma who survived for > 5 years. The patient underwent 11 MRI studies during this period and tissue analysis at the initial resection and at two follow-up resections for tumor progression. We evaluated an mpMRI prediction model developed to differentiate TP from PsP. We found that a PP value ≥ 50% reliably predicted tumor progression versus a value of 0%–50% consistent with PsP. Taken together, the mpMRI prediction model could be used as an imaging biomarker to stratify patients for clinical trials, be used in monitoring trials to determine go/stop decisions during immunotherapy monitoring, and guide informed decisions for repeat surgery. Acknowledgments Support for this study was provided by NIH/NCI to S.M. and S.C. (grant no. 1R01CA262584-01). Disclosures M. Nasrallah reported payment for an educational lecture from Servier outside the submitted work. S. Mohan reported research grants from Galileo CDS and Novocure outside the submitted work. Author Contributions Conception and design: Mohan, Rajan, Brem, Chawla. Acquisition of data: Mohan, Rajan, Brem, Gewolb, Kumar, Desai, Nasrallah, Lucas, Chawla. Analysis and interpretation of data: all authors. Drafting the article: Mohan, Rajan, Brem, Gewolb, Kumar, de Godoy, Loevner, Chawla. Critically revising the article: Mohan, Brem, Kumar, de Godoy, Desai, Nasrallah, Lucas, Chawla. Reviewed submitted version of manuscript: Mohan, Rajan, Brem, Gewolb, Kumar, de Godoy, Desai, Nasrallah, Loevner, Chawla. Approved the final version of the manuscript on behalf of all authors: Mohan. Statistical analysis: Mohan, Rajan, Chawla. Administrative/technical/material support: Mohan, Brem, Loevner, Chawla. Study supervision: Mohan, Chawla. Laboratory principal investigators: Mohan, Chawla. Supplemental Information Previous Presentations Portions of this work were presented at the American Society of Neuroradiology 2026 Annual Meeting, Austin, TX, May 17–21, 2026. Correspondence Suyash Mohan: Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA. suyash.mohan@pennmedicine.upenn.edu. References