Primary central nervous system lymphoma is one of the most unforgiving tumors a neurosurgeon can encounter. It accounts for roughly three percent of all intracranial tumors, it is aggressive, and surgery is rarely more than a biopsy, because the real weapons against it are high-dose chemotherapy regimens built around methotrexate. Yet this cancer has a peculiar weakness: it is often exquisitely sensitive to corticosteroids, the humble anti-inflammatory drugs given to almost every patient with a brain tumor to ease swelling and neurological symptoms. A dose of dexamethasone can shrink a contrast-enhancing lymphoma lesion on MRI within days, and in some documented cases the malignant B-cells vanish so completely from a biopsy sample that the diagnosis becomes impossible to establish. That same sensitivity, however, may carry information that clinicians have never systematically measured. A new multicenter study from Austria, published in the Journal of Neuro-Oncology, offers the first structured, prospective description of how these tumors behave radiologically after routine steroid therapy, and it raises a striking possibility: the direction in which a lymphoma moves on MRI after steroids may itself be a predictor of how long the patient lives.
The research team, led by Florian Scheichel of University Hospital St. Pölten with colleagues from the Medical University of Innsbruck and Kepler University Hospital in Linz, confronted a methodological puzzle that has plagued the field for decades. Almost all existing data on steroid response in primary CNS lymphoma come from retrospective series with wildly inconsistent information about steroid dose, duration, and timing. In those studies, patients received steroids before their diagnostic imaging for variable reasons, and the intervals between scans and drug administration were uncontrolled. Some patients may have already regressed before their first scan; others may have progressed after an initial regression that nobody captured. The result was a literature in which reported response rates ranged from nearly ninety percent of tumors showing no change to nearly sixty percent showing progression, numbers so divergent that no reliable picture of steroid sensitivity could be drawn. The Austrian group designed their study to break this cycle by exploiting a window that every patient passes through but that had never been systematically imaged: the days immediately after biopsy, when the diagnosis is already secured and steroids are given routinely for symptom relief.
The prospective design was elegant in its simplicity. Patients with suspected primary large B-cell lymphoma of the central nervous system, who were scheduled for stereotactic or endoscopic biopsy at three Austrian neurosurgical centers, were enrolled after informed consent. Patients with systemic lymphoma, immunocompromise, or Epstein-Barr virus positivity were excluded, and the final diagnosis had to be confirmed as primary large B-cell CNS lymphoma. After surgery, all centers administered a comparable steroid protocol: an initial bolus of 40 milligrams of dexamethasone, followed by 16 milligrams daily, with the dose halved every three days. Critically, the team planned an additional thin-slice MRI five to ten days after surgery, timed to compare directly with the preoperative navigational MRI used to plan the biopsy. Patients were excluded if that second scan would have delayed definitive cancer treatment, a safeguard that cost the study thirteen of its forty screened candidates but protected the integrity of the remaining data. Eighteen patients formed the prospective volumetric cohort, ten women and eight men, none of whom suffered surgical complications that would have confounded the volume measurements.
The volumetric analysis was performed with 3D Slicer, an open-source platform for medical image computing, by segmenting the contrast-enhancing lesion on both the preoperative and post-steroid scans and calculating relative volume change. The results revealed a spectrum of behavior far wider than the word regression suggests. In 83.3 percent of cases the tumor shrank after steroids, with a median volume reduction of 40 percent, but the individual responses ranged from a barely perceptible one percent shrinkage to a dramatic 77 percent collapse. Meanwhile, in 16.7 percent of patients, the tumor did something counterintuitive: it grew despite the steroid therapy, with a median progression of 46 percent and one case expanding by 94 percent. Mean tumor volume fell from 16.1 cubic centimeters before surgery to 11.9 cubic centimeters after steroids across the cohort. The median interval between scans was seven days, and the median cumulative postoperative dexamethasone dose was 112 milligrams, with nearly all patients receiving between 104 and 140 milligrams.
Perhaps the most clinically consequential negative finding was what did not explain this heterogeneity. The researchers tested whether the degree of tumor volume change correlated with the cumulative steroid dose, the dose normalized to body surface area, or the duration of therapy, and found no evidence for any of these relationships. The Kendall-tau correlation between volume change and cumulative dose was a negligible 0.072 with a p-value of 0.694, and duration fared no better. In other words, a tumor that melted away under a standard steroid taper was not simply receiving more drug than one that expanded; the difference appears to reside in the biology of the malignant clone itself. There was, however, a suggestive signal involving lactate dehydrogenase, or LDH, the serum enzyme long known to be an adverse prognostic marker in both systemic lymphomas and primary CNS lymphoma. In the prospective cohort, higher preoperative LDH levels showed a trend toward correlation with worse radiological response, and when the analysis was expanded to a combined cohort of 31 patients, that correlation became statistically significant, with higher LDH predicting less tumor shrinkage on two-dimensional measurement.
That combined cohort was the study’s second analytical layer. Because eighteen patients were too few for meaningful survival statistics, the team merged their prospective data with a carefully filtered retrospective cohort of thirteen patients from their own earlier 2021 multicenter study. The inclusion criteria were strict: retrospective patients needed at least two preoperative MRI scans with steroid therapy initiated no more than two days before the first scan and continuing or uninterrupted between scans, ensuring that any observed change genuinely reflected the drug’s effect rather than uncontrolled timing. For this combined group, response was quantified as the change in the largest cross-sectional tumor area on an axial MRI slice, a two-dimensional measure that was validated against the volumetric results in the prospective patients with an intraclass correlation coefficient of 0.87, indicating excellent agreement between the two methods. In the combined cohort, 74.2 percent of tumors regressed and 25.8 percent progressed, and the two groups proved comparable in age, tumor size, Karnofsky performance status, involvement of deep brain structures, and treatment type, meaning the response pattern was not simply a proxy for having a bigger or more advanced tumor.
The survival analysis delivered the study’s headline result. Patients whose tumors regressed after steroids had a median overall survival of 31.4 months, while those whose tumors progressed despite steroids survived a median of just 3.9 months, a difference that reached statistical significance in Kaplan-Meier analysis with a p-value of 0.015. A sensitivity analysis excluding the three patients who had received steroids before biopsy preserved the finding, and a stratification using RECIST-style thresholds, at least 30 percent regression, stable disease, or at least 20 percent progression, showed a consistent trend, with median survival not reached in the regression group, 11.1 months for stable disease, and 3.9 months for progression. The authors then turned to Firth-corrected Cox regression, a statistical method designed for small samples with few events, and built a multivariable model including radiological response, age, Karnofsky performance status, and whether induction therapy contained high-dose methotrexate. In that model, tumor progression despite steroids carried a hazard ratio of 3.37 for death, with a p-value of 0.02 and a 95 percent confidence interval of 1.21 to 9.35, suggesting that radiological response is an independent prognostic factor rather than a mere echo of known predictors.
The biological interpretation is where the finding becomes genuinely provocative. Corticosteroids kill lymphoid cells through glucocorticoid receptor-mediated apoptotic pathways, and the sensitivity of a given lymphoma to this mechanism may reflect broader properties of drug responsiveness in the malignant clone. If a tumor cannot be pushed into apoptosis by a potent glucocorticoid, the reasoning goes, it may also respond poorly to methotrexate and the other cytotoxic agents that form the backbone of CNS lymphoma therapy. Progression despite steroids would then be a visible, early, and essentially free signal of multidrug resistance, available from scans that are already part of routine care. This would align with the rare but well-documented historic cases in which steroids alone produced prolonged, occasionally years-long remissions, presumably in patients with extreme intrinsic sensitivity. The authors are careful to frame their conclusions as exploratory: the sample was small, the survival analysis necessarily incorporated retrospective data, and confounders such as salvage therapies and treatment heterogeneity could not be fully controlled. Overall survival was chosen over progression-free survival precisely because follow-up strategies across centers were inconsistent.
The study also carries a practical warning that neurosurgeons and emergency physicians ignore at their patients’ peril. Preoperative steroids can render biopsies inconclusive, with histological disappearance of malignant B-cells reported in up to 52 percent of pretreated cases and threefold increases in non-diagnostic biopsies described in prior work. Guidelines therefore recommend avoiding steroids before tissue acquisition whenever clinically feasible, yet patients frequently arrive at specialized centers already steroid-treated because of neurological deterioration earlier in their pathway. The new data add a dimension to this dilemma: the same steroid exposure that risks obscuring the diagnosis may also, if properly documented with serial imaging, yield prognostic information that could eventually help stratify patients into risk-adapted treatment protocols. The Austrian team emphasizes that larger prospective multicenter studies are needed to validate the survival finding and to unravel the molecular mechanisms that determine why one patient’s lymphoma dissolves under dexamethasone while another’s grows through it. If those mechanisms can be identified, the humble steroid challenge, long viewed as a diagnostic nuisance, could be reframed as one of the cheapest prognostic tests in neuro-oncology.
Subject of Research: Radiological response of primary central nervous system lymphoma to corticosteroid therapy and its prognostic value for overall survival
Article Title: Radiological response of primary central nervous system lymphoma after corticosteroid therapy and its predictive value on overall survival: a multicenter study
Article References: Scheichel, F., Pinggera, D., Rossmann, T., Popadic, B., Aspalter, S., Schön, V., Woehrer, A., Dorostkar, M. M., Dorfer, C., Freyschlag, C. F., & Marhold, F. (2026). Radiological response of primary central nervous system lymphoma after corticosteroid therapy and its predictive value on overall survival: a multicenter study. Journal of Neuro-Oncology, 179(2), Article 63. https://doi.org/10.1007/s11060-026-05773-3
Image Credits: AI Generated
DOI: 10.1007/s11060-026-05773-3
Keywords: primary CNS lymphoma, corticosteroid therapy, MRI, overall survival, dexamethasone, lactate dehydrogenase, prognostic factors, volumetric analysis, methotrexate, neuro-oncology, Kaplan-Meier analysis, drug sensitivity
Cite Scienmag News
APA MLA Chicago
Nathaniel Bowman. (October 4, 2026). How a Brain Lymphoma Responds to Steroids on MRI May Predict Survival. Scienmag. https://scienmag.com/how-a-brain-lymphoma-responds-to-steroids-on-mri-may-predict-survival/
Nathaniel Bowman. “How a Brain Lymphoma Responds to Steroids on MRI May Predict Survival.” Scienmag, 4 October 2026, https://scienmag.com/how-a-brain-lymphoma-responds-to-steroids-on-mri-may-predict-survival/. Accessed 4 October 2026.
Nathaniel Bowman. “How a Brain Lymphoma Responds to Steroids on MRI May Predict Survival.” Scienmag. October 4, 2026. https://scienmag.com/how-a-brain-lymphoma-responds-to-steroids-on-mri-may-predict-survival/
Copy citation Download RIS
Tags: brain lymphoma response to steroidscorticosteroid effects on malignant brain B-cellscorticosteroid sensitivity in brain tumorscorticosteroid therapydexamethasonedrug sensitivityKaplan–Meier analysislactate dehydrogenasemethotrexateMRIMRI biomarkers for CNS lymphoma prognosisMRI imaging changes after steroid treatmentMRI prediction of survival in primary central nervous system lymphomamulticenter study on lymphoma MRI responseneuro-oncologyoverall survivalpredicting patient outcomes in primary brain lymphomaprimary CNS lymphomaprognostic factorsprognostic value of MRI in CNS lymphomaradiological behavior of CNS tumorsrole of dexamethasone in brain tumor managementsteroid-induced tumor shrinkage in brain cancervolumetric analysis
