Minimizing the risks of stereotactic brain biopsy in suspected central nervous system lymphoma: a retrospective database study

Abstract

Background

Stereotactic brain biopsy is widely used to confirm the diagnosis of central nervous system (CNS) lymphoma; however, few studies have investigated the perioperative factors that impact the procedure’s clinical utility and outcome.

Method

We retrospectively reviewed all patients at our institution who underwent brain biopsy and were confirmed to have CNS lymphoma diagnosis in years 2001–2021. Clinical history, imaging, intraoperative data, and post-biopsy outcomes were obtained. Associations between perioperative factors and postprocedural morbidity/mortality were analyzed using univariate and multivariate models.

Results

CNS lymphomas were diagnosed in 75 patients via 79 separate biopsies, of which 12 (16%) had a prior diagnosis of lymphoma (CNS or systemic). Postoperative 30-day mortality was 8.9%. Intraoperative bleeding was noted in 6 patients (7.6%), postoperative hemorrhage in 4 patients (5.1%). We identified several factors leading to poor outcomes. Biopsy of deep-brain structures (thalamus and basal ganglia) was associated with higher rates of postoperative hemorrhage (17.6% vs. 1.6%, p = 0.030) and discharge to hospice (35.3% vs. 9.7%, p = 0.018). Biopsy at multiple depths, a modifiable risk factor, was associated with worse survival after multivariate analysis (p = 0.036). In patients with a history of systemic diffuse large B-cell lymphoma (DLBCL), all brain biopsies diagnosed DLBCL; several patients with non-DLBCL were diagnosed with other pathologies including glioblastoma.

Conclusion

Several variables were associated with poor outcomes after biopsy for suspected CNS lymphoma. Our findings support avoiding biopsy of multiple depths unless critical. These factors should be considered and further risk stratification for patient selection is necessary to optimize the clinical utility of brain biopsy.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00432-026-06458-0.

Introduction

Central nervous system (CNS) lymphoma is a rare variant of extranodal non-Hodgkin lymphoma. Primary CNS lymphoma (PCNSL) originates from the CNS parenchyma, leptomeninges, cranial nerves, and spinal cord and accounts for up to 2% of all primary CNS tumors in adults (Han and Batchelor 2017; Ostrom et al. 2021). Immunodeficiency is a relevant risk factor for development of CNS lymphoma, but most cases develop in immunocompetent hosts (Bhagavathi and Wilson 2008). Secondary CNS lymphoma refers to systemic non-Hodgkin lymphoma that has disseminated to the CNS. Median overall survival of PCNSL in population studies remains poor; however, for those receiving chemotherapy, it often now exceeds 5 years (Mendez et al. 2018; Therkelsen et al. 2023). In contrast, secondary CNS lymphoma continues to have a poor median survival of 3–5 months (Boehme et al. 2007; El-Galaly et al. 2018). Relapses are frequent, and remissions are often short-lasting. Therefore, appropriate diagnosis is critical to determine prognosis and management.

The diagnostic methods of primary CNS lymphoma include neuroimaging, ophthalmologic exam, cerebral spinal fluid (CSF) cytology, and stereotactic biopsy (Grommes et al. 2019). Magnetic resonance imaging (MRI) with gadolinium contrast has a high sensitivity but low specificity in identifying CNS lymphoma (Scott et al. 2013). Additionally, the empiric use of corticosteroids may reduce abnormal enhancement and vasogenic edema, further interfering with MRI interpretation (Gametchu 1987; Yim et al. 2020). CSF cytology can provide definitive diagnosis when positive but has a low sensitivity, partially due to the overlapping morphology of neoplastic and inflammatory lymphocytes. It may be more useful in cases with leptomeningeal involvement, cord involvement, or CSF dissemination (Hegde et al. 2005; Perske et al. 2011). In contrast, the diagnosis of SCNSL relies on contrast-enhanced neuroaxis imaging and CSF analysis to detect leptomeningeal involvement.

Given the diagnostic difficulties, stereotactic brain biopsy remains the diagnostic criterion standard when radiographic findings and CSF analysis are inconclusive (Hoang-Xuan et al. 2015). However, brain biopsies have known morbidity, including intracranial hemorrhage and neurological deficit, and risk of postoperative death (Hall 1998; Malone et al. 2015). A recent retrospective study analyzing patients with presumed CNS relapse from systemic lymphoma reported up to 8.3% biopsy-related morbidity and 10% 30-day mortality in patients, which is alarming (Brown et al. 2022). However, those risks for the general CNS lymphoma patients are unknown. In this study, we aim to determine the perioperative risk factors associated with poorer outcomes in all patients with suspected CNS lymphoma to optimize the utilization of this procedure.

Methods

Study design

In August 2022, we retrospectively reviewed all brain biopsies at the University of Michigan, Michigan Medicine, performed between January 1, 2001, and December 31, 2021. The study was approved by the University of Michigan Institutional Review Board, which deemed patient consent unnecessary. We used the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist (Checklists STROBE 2008).

Inclusion criteria

Patients were identified using Current Procedural Terminology (CPT) codes corresponding to stereotactic brain biopsy during the specified time frame. All patients, regardless of age or gender, were included if they had a tissue diagnosis of central nervous system (CNS) lymphoma via biopsy. We additionally identified five patients with a prior history of systemic lymphoma who underwent stereotactic brain biopsy to rule out CNS lymphoma but were found to have alternative pathologic diagnoses. These patients were excluded from the primary analysis and instead included in the last segment of Results section to inform the discussion on the diagnostic utility of brain biopsy in patients with prior lymphoma.

We reviewed perioperative factors, including clinical history, preoperative imaging, systemic workup, biopsy techniques, and intraoperative complications. Immunocompromised status was defined as the presence of active human immunodeficiency virus infection or being on immunomodulators for autoimmune disease or organ transplantation. Significant cardiac history is defined as history of acute coronary syndrome (ACS), congestive heart failure (CHF), arrhythmia s/p pacemaker implantation.

Exclusion criteria

Patients with a prior diagnosis of systemic lymphoma who did not have a brain biopsy confirming CNS lymphoma were excluded. These patients are described separately in the results section.

Outcome measures

Primary outcomes were postoperative morbidity and 30-day mortality from the day of surgery. Morbidity was determined by postoperative complications, prolonged intensive care unit stay (> 2 days), and discharge to hospice. Specifically, postoperative hemorrhage as one of the postoperative complications was defined as intracranial hemorrhage occurring within 30 days from the initial brain biopsy that resulted in neurologic changes or changes in management requiring further intervention. After a biopsy, on our routine postoperative CT scan, it is common to identify a small incidental radiographic finding of hemorrhage that did not cause any clinical changes. Incidental radiographic findings of small hemorrhages observed on routine postoperative CT scans that did not lead to clinical changes were excluded from the analysis. Clinically significant hemorrhage was defined as causing new neurological deficits (e.g., new focal neurological deficit, new-onset seizure) or as intraoperative hemorrhage that resulted in a non-diagnostic biopsy. Secondary outcomes included overall survival (OS), defined as the number of months from brain biopsy to the date of death, censored at the date of last contact using the Kaplan-Meier method. Performance status was not included as part of the outcomes due to high data incompletion rate (77.3%).

Statistical analysis

Descriptive statistics were used to summarize the baseline characteristics of the study cohort. Results were reported as frequency (percentage, %) or median (interquartile range, IQR). Univariate analysis used either Student-t test or Wilcoxon signed-rank test to compare continuous variables (depending on variable distribution), while Fisher’s exact test was used to compare categorial variables. Relative risks (RRs) with 95% confidence intervals (CIs) were calculated for the categorical variables. Age was dichotomized at the cohort median for RR calculation. Statistically significant covariates from the univariate analysis were selected into a multivariate logistic regression to test independent association with postoperative mortality. We used the Firth logistic regression method to minimize small-sample biases inherent to rare events (Firth 1993; Heinze and Schemper 2002). The results were reported as odds ratios with 95% CI. The impact of covariates on OS was evaluated using the multivariate Cox proportional hazard model, reported as hazard ratios (HRs) and 95% CI. Kaplan-Meier curves of OS were generated. Statistical significance was set at two-sided p-value of < 0.05. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC).

Results

Patient characteristics

A total of 75 patients with biopsy-proven CNS lymphoma were included. Four of the 75 patients had initial nondiagnostic biopsies and therefore underwent a repeat biopsy at our institution (repeat biopsy consistent with lymphoma), making the total distinct biopsy count 79. Twelve of the 75 patients (16%) had a prior history of lymphoma. Table 1 describes the demographic and clinical characteristics of the study cohort, stratified by history of prior lymphoma. The median follow-up time was 13.9 months (IQR 2.5–40.5). Twelve patients (16%) were clinically immunocompromised, 13 (17.3%) had chronic steroid use for a median of 59 (IQR 25–185) days. Among the patients who had chronic steroid use, 84.6% continued steroid therapy until the biopsy, 15.4% discontinued it at least 15 days before the procedure. At the time of initial presentation, 17.3% of patients were on anticoagulation, and 30.7% were on antiplatelet therapy. Patients received varying degrees of systematic workup (CT chest-abdomen-pelvis obtained in 74.7% of the study cohort, PET scan 22.8%, MRI spine 29.1%, CSF cytology 55.7%).

Table 1.

 Demographics and clinical characteristics by history of prior lymphoma, N (%) or median (IQR)

	Overall (N = 75)	History of lymphoma
		Yes (N = 12)	No (N = 63)
Median age in years	65.0 (50.2–72.7)	60.8 (45.2–67.4)	65.8 (54.0–73.4)
Sex
Female	35 (46.7)	5 (41.7)	30 (47.6)
Male	40 (53.3)	7 (58.3)	33 (52.4)
Race
White	66 (88.0)	12 (100.0)	54 (85.7)
Black	5 (6.7)	0 (0.0)	5 (7.9)
Other	2 (2.7)	0 (0.0)	2 (3.2)
Unknown	2 (2.7)	0 (0.0)	2 (3.2)
Prior lymphoma type
DLBCL		5 (41.7)
PCNSL		3 (25.0)
NHL, non-specified		2 (16.7)
Follicular lymphoma		1 (8.3)
Tonsil lymphoma		1 (8.3)
Hx of other cancer	15 (20.0)	2 (16.7)	13 (20.6)
Hx of hypertension	32 (42.7)	2 (16.7)	30 (47.6)
Hx of diabetes	11 (14.7)	3 (25.0)	8 (12.7)
Hx of DVT/PE	12 (16.0)	3 (25.0)	9 (14.3)
Significant cardiac history	6 (8.0)	0 (0.0)	6 (9.5)
Immunocompromised	12 (16.0)	3 (25.0)	9 (14.3)
Anticoagulation use	13 (17.3)	4 (33.3)	9 (14.3)
Antiplatelets use	23 (30.7)	2 (16.7)	21 (33.3)
Chronic steroid use	13 (17.3)	0 (0.0)	13 (20.6)

DLBCL, diffuse large B-cell lymphoma; DVT, deep vein thrombosis; Hx, history; IQR, interquartile range; NHL, non-Hodgkin lymphoma; PCNSL, primary central nervous system lymphoma; PE, pulmonary embolism

Preoperative imaging findings

Brain MRIs were obtained in 76 of 79 biopsies at the time of initial evaluation. MRI characteristics are summarized in Online Resource 1, stratified by biopsy diagnosis. Tumors presented as a single lesion in 42 (53.2%) and multifocal in 37 (46.8%) cases. Among the multifocal cases, 22 (59.5%) involved deep brain structures. Most lesions were hyperintense on T2-weighted images rather than iso/hypointense (68.6% vs. 31.4%). All lesions enhanced after intravenous contrast; 34.2% enhanced homogenously, 51.3% heterogeneously, and 15.8% peripherally. Diffusion restriction was seen in most lesions (61.8%). Concomitant hemorrhage and necrosis were uncommon but present (9.6% and 6.9%, respectively).

Biopsy outcomes

Clinical outcomes of brain biopsies are summarized in Table 2. Eight out of 75 patients had secondary CNS lymphoma, the reminder patients (67) had PCNSL. Most tissue diagnoses were diffuse large B-cell lymphoma (DLBCL) (N = 69, 87.3%). Non-DLBCL diagnoses included 3 Epstein-Barr virus–positive lymphoma, 2 lymphoplasmacytic lymphoma, 1 follicular lymphoma, and 4 gliosis/nondiagnostic. All nondiagnostic cases underwent repeat brain biopsies at our institution, which uniformly turned out to be DLBCL and therefore were included in the study cohort. Two of 4 (50%) patients with an initial nondiagnostic biopsy received steroids 8–9 days preoperatively; this is comparable to patients with diagnostic biopsies 39/70 (54.2%). Changes in clinical management occurred in most cases after a diagnostic biopsy (68/79, 86.1%). Of the patients who received new treatment, chemotherapy was started in 41 (51.9%), radiation therapy was started in 10 (12.7%), and both chemotherapy and radiation therapy in 13 (16.5%).

Table 2.

 Summary of brain biopsy clinical outcomes, n (%) or median (IQR)

Clinical outcomes per biopsy	Overall (N = 79)
Primary CNS Lymphoma (PCNSL)	67 (84.8)
Secondary CNS Lymphoma (SCNSL)	8 (10.1)
Biopsy diagnosis
DLBCL	69 (87.3)
EBV associated	3 (3.8)
LPL	2 (2.5)
Follicular lymphoma	1 (1.3)
Gliosis/nondiagnostic	4 (5.1)
Change in management	68 (86.1)
Started chemotherapy	41 (51.9)
Started radiation therapy	10 (12.7)
Started chemotherapy + radiation therapy	13 (16.5)
Postoperative complications	13 (16.5)
Hemorrhage	4 (5.1)
Stroke	2 (2.5)
Cranial infection	1 (1.3)
Other (multiple seizures, respiratory failure, UTI, PE, c. diff.)	10 (12.7)
Prolonged ICU stay	5 (6.3)
Neurological function in 30 days
Improved	13 (18.6)
Same	40 (57.1)
Worse	17 (24.3)
Discharge
Home	41 (51.9)
Rehab	25 (31.6)
Hospice	12 (15.2)
Postoperative mortality in 30 days	7 (8.9)
Overall mortality	48 (60.8)
Overall survival (months)	13.1 (2.9–35.0)

DLBCL, diffuse large B-cell lymphoma; EBV, Epstein-Barr virus; LPL, lymphoplasmacytic lymphoma; PE, pulmonary embolism; UTI, urinary tract infection

The overall postoperative complication rate was 16.5%. Complications included 4 (5.1%) postoperative hemorrhages, 2 (2.5%) ischemic strokes, 1 (1.3%) intracranial infection, and 10 (12.7%) others that were less directly related to the procedure itself (3 seizures, 1 pulmonary embolism, 1 urinary tract infection, 3 Clostridioides difficile infection, 1 acute respiratory failure secondary to pneumonia, 1 prolonged intubation due to low cranial nerve deficits). Only 41 (51.9%) were directly discharged to home after the biopsy; 12 patients (15.2%) were discharged to hospice and 25 patients (31.6%) were discharged to rehabilitation. The overall 30-day mortality was 7 (8.9%). Of those 7 mortalities, 3 were directly related to biopsy complications (2 developed postoperative hemorrhage, 1 developed generalized tonic colonic seizure and meningoencephalitis). The other 4 mortalities were related to patient and family eliciting hospice within 7 days of biopsy and passing away within 30 days from the procedure.

Perioperative factors impact postoperative morbidity and mortality

Associations between postoperative complications and perioperative factors are summarized in Online Resource 2. Postoperative hemorrhage developed after 4 biopsies, 3 were of the deep-brain structures (basal ganglia or thalamus); this biopsy site is associated with significantly higher risk of postoperative hemorrhage than other sites (17.6% vs. 1.6%, RR = 10.94, 95% CI = 1.62–73.85, p = 0.030), and is associated with a higher percentage of hospice discharge after the procedure (35.3% vs. 9.7%, RR = 3.65, 95% CI = 1.35–9.42, p = 0.018). More importantly, a total of 42 (53.2%) patients had disease involvement of the deep-brain structure, yet only 17 underwent biopsy of the deep brain lesions. The remaining 16 patients were biopsied at alternative foci of their multifocal lesions; among these patients, only 1 (5.9%) had postoperative hemorrhage in the setting of thrombocytopenia. History of anticoagulation or antiplatelet use was not associated with an increased risk of postoperative hemorrhage.

We also recorded the presence of intraoperative bleeding during the biopsy, which is defined as return of fresh blood on the biopsy needle that was documented by surgeons in the operative notes. These bleeding incidents were typically stabilized by warm irrigation, and replacement of the stylet. Interestingly, of the 6 patients who had documented intraoperative bleeding, only 1 (16.7%) developed clinically significant postoperative hemorrhage on imaging (RR = 4.06, 95% CI = 0.59–22.6) (Fig. 1). Despite this, intraoperative bleeding was evaluated for its association with other postoperative outcomes. Based on univariate analysis, patients were more likely to be discharged to hospice if they had intraoperative bleeding (RR = 4.06, 95% CI = 1.31–9.61, p = 0.042). We conducted a post-hoc analysis to assess for correlations between perioperative factors and increased risk of intraoperative bleeding. None of the variables reached statistical significance, however, 5/6 (83.3%) patients with intraoperative bleeding were over age 65 and had a prior history of hypertension.

Fig. 1.

A Non-contrasted head CT showing R basal ganglionic hemorrhage following stereotactic brain biopsy of lesion suspected of primary CNS lymphoma. Preoperative MRIs with and without contrast demonstrate a heterogeneously enhancing lesion with surrounding T2/FLAIR (fluid-attenuated inversion recovery) hyperintensity and diffusion restriction. B T1; C T1+; D T2; E FLAIR; F diffusion-weighted image. The patient underwent two biopsies: The first was nondiagnostic; the second biopsy 3 months later confirmed a diagnosis of diffuse large B-cell lymphoma, but the procedure was complicated by postoperative hemorrhage

We evaluated the association between perioperative risk factors and 30-day mortality. Based on univariate analysis (see Online Resource 3), biopsy of the deep-brain structures (p < 0.001), eloquent areas (p = 0.039), intraoperative bleeding (p = 0.008), and postoperative hemorrhage (p = 0.037) are correlated with a higher likelihood of death within 30 days. Particularly, patients who underwent deep-brain structure biopsy had a 30-day mortality rate of 35.3% in comparison to that of 14.3% in patients with multifocal lesion including the deep-brain structures but were biopsied elsewhere. Multivariate logistic regression adjusted for confounding factors (Table 3) further demonstrated that intraoperative bleeding was an independent predictor of 30-day mortality (odds ratio = 11.18, 95% CI = 1.15–109, p = 0.038).

Table 3.

 Multivariate logistic regression of 30-day mortality after brain biopsy (N = 79)

Variable	30-day mortality (yes/no)	Odds ratio (95% CI)	P value
Biopsy of deep-brain structures
Yes	6/11	14.51 (0.58, 365)	0.104
No	1/61
Biopsy of eloquent region
Yes	6/28	0.73 (0.025, 21.3)	0.854
No	1/43
Intraoperative bleeding
Yes	3/3	11.18 (1.15, 109)	0.038*
No	4/69
Postoperative hemorrhage
Yes	2/2	3.54 (0.25, 51.3)	0.353
No	5/70

Deep-brain structures include basal ganglia and thalamus

*p < 0.05, statistically significant

Perioperative factors and their association with long-term survival

The median overall survival time following biopsy was 13.1 months (IQR 2.9–35 months). In comparison to postoperative mortality, a separate set of factors dictate long-term prognosis. In multivariate analysis (Table 4), lower OS was independently associated with biopsy of deep-brain structures (HR = 4.99, 95% CI = 1.13–11.19, P = 0.013), multiple biopsy depths (HR = 2.14, 95% CI = 1.05–4.34, P = 0.036), intraoperative bleeding (HR = 4.14, 95% CI = 1.33–12.87, P = 0.014), older age at diagnosis (HR = 1.03, 95% CI = 1.00-1.06, P = 0.036), and immunocompromised status (HR = 3.27, 95% CI = 1.43–7.49, P = 0.005).

Table 4.

Multivariate analysis of overall survival after brain biopsy (N = 79)

Variable	Subgroup	Hazard ratio (95% CI)	P value
Age at diagnosis	–	1.03 (1.002, 1.06)	0.036*
Immunocompromised	Yes/No	3.27 (1.43, 7.49)	0.005*
Biopsy of deep-brain structures	Yes/No	4.99 (1.13, 11.19)	0.013*
Multiple depths	Yes/No	2.14 (1.05, 4.34)	0.036*
Intraoperative bleeding	Yes/No	4.14 (1.33, 12.87)	0.014*
Postoperative hemorrhage	Yes/No	2.22 (0.51, 9.73)	0.289

Deep-brain structures include basal ganglia and thalamus

*P < 0.05, statistically significant

Utility of brain biopsy in patients with a history of prior lymphoma

During the study period, 17 patients with a history of systemic (n = 13) or CNS (n = 4) lymphoma underwent brain biopsy for suspected CNS involvement. Of these patients, 12 (70.6%) had a new brain biopsy consistent with CNS recurrence, 2 (11.8%) showed glioblastoma, 1 (5.9%) showed progressive multifocal leukoencephalopathy (PMLE), and 2 (11.8%) were nondiagnostic with no repeat diagnostic procedure. Specifically, 5/5 (100%) diagnostic biopsies in patients with a history of non-CNS DLBCL showed brain DLBCL, and 3/3 (100%) biopsies in patients with a history of PCNSL showed recurrence. Meanwhile, neither of the above patients whose brain biopsies resulted in glioblastoma had a history of DLBCL or PCNSL. Within this “history of lymphoma” group, clinically significant complications occurred in 4 patients, 2 (16.7%) of whom died within 30 days post-op in the setting of significant medical co-morbidities. Detailed information for these patients is presented in Online Resource 4.

Discussion

We evaluated the perioperative factors associated with poor outcomes in patients undergoing brain biopsy for suspected CNS lymphoma. Interestingly, we have identified that intraoperative bleeding from the biopsy site, even if it does not result in postoperative hemorrhage on imaging, significantly correlates with a higher risk of discharge to hospice, 30-day mortality, and worse overall survival. It has previously been noted that postoperative hemorrhage is associated with poor outcomes (Hall 1998; Kaakaji et al. 2001; Malone et al. 2015). Despite this, it was the development of intraoperative bleeding and not postoperative hemorrhage that independently correlated strongest with overall worse outcomes in our cohort. This phenomenon may be due to intraoperative bleeding being a potential cause of increased edema or that it is inherently correlated with more aggressive tumor phenotype. Thus, identification of intraoperative bleeding, even if not seen on imaging, may imply the necessity of closer postoperative monitoring.

Our multivariate analysis of overall survival confirms the independent prognostic role of advanced age. Moreover, we identified that immunocompromised status, biopsy of the deep brain region (i.e., basal ganglia and thalamus), and multiple biopsy depths were independently associated with worse prognosis. We have also identified that biopsy of the deep brain structure is independently associated with higher risk of discharge to hospice after the procedure. Biopsy of the basal ganglia specifically is associated with a significantly higher risk of postoperative hemorrhage, which is consistent with prior data from Air et al. (2009) and Woodworth et al. (2006). Although this has been identified in multiple studies, the etiology of increased hemorrhage risk in deeper locations is not clear. This could be related to length of brain parenchyma that the biopsy needle must penetrate through, or due to the inherent vascularity or anatomic structure of these deeper regions. Ferreri et al. (2003) recognized in a previous study that age greater than 60 years, performance status, and disease involvement of deep regions are independently associated with worse survival in PCNSL patients (Ferreri et al. 2003). However, only 72% of their patient cohort underwent stereotactic biopsy, the rest were diagnosed via surgical resection, CSF cytology, and autopsy. Hence, whether biopsy of the deep structures contributed directly to overall survival is unclear based on their report. Our findings highlight the independent, iatrogenic risks associated with stereotactic biopsy of CNS lymphoma in addition to the known clinical risk factors. This is particularly relevant as patients with systemic hematological malignancy may have higher surgical risk than patients with isolated brain malignancy. Furthermore, though prior study has reported association between higher number of needle passes with higher incident of postoperative hemorrhage (Air et al. 2009), our study is the first to address the prognostic impact of biopsy at different depths. Therefore, we suggest that surgeons should avoid biopsy of multiple depths if possible as this is independently associated with worse outcomes. Furthermore, in cases where the intracranial disease is multifocal, avoiding biopsy of deep sites should be considered.

Our results also support prior findings that the use of anticoagulation or antiplatelet agents prior to admission does not increase the risk of postoperative hemorrhage complications (Malone et al. 2015). This indicates that withholding such medications preoperatively may be sufficient to reverse any adverse effects on hemostasis to maintain a low risk of hemorrhage.

The overall prognosis of CNS lymphoma with treatment is favorable to high-grade gliomas, hence the threshold for neurosurgeons to offer a biopsy is often lower (Mendez et al. 2018; Ostrom et al. 2016, 2021). However, with the risk profile of stereotactic biopsy in mind, it is worth investigating the utility of brain biopsy in patients with a prior history of lymphoma. In our study, 70% of biopsied intracranial lesions in patients with a prior history of lymphoma (systemic or CNS) were consistent with CNS lymphoma. More importantly, we found that patients with a history of systemic DLBCL who had diagnostic biopsies were all diagnosed with intracranial involvement (100%). This is consistent with the results published by Brown et al. (2022), where the authors identified that patients with a history of systemic DLBCL were almost always diagnosed with CNS relapse (97.2%). At the same time, a small percentage of patients in our study (11.8%) without prior systemic DLBCL were found to have glioblastoma on brain biopsy. This suggests that biopsy may be necessary in this patient cohort when CSF study and radiographic imaging are inconclusive. Together these findings are significant in reconsidering the diagnostic utility of biopsies in patients with systemic DLBCL and may suggest we minimize brain biopsy in such patient cohorts.

In many cases, systemic workup had not been completed prior to brain biopsy in these cases. CSF testing for fluid cytology, cell count, protein and glucose may lead to a diagnosis in some cases. These tests are especially helpful when leptomeningeal disease is present. The latter is also detected by contrasted total spine MRI (Michel et al. 2022). Signs of extracranial lymphoma should be searched for to help with diagnosis by less invasive means than brain biopsy. Hyun Suh et al. (2022) found older patients or those who are Epstein-Barr virus–positive were at increased odds of having extracranial lymphoma that could be detected by contrasted body CT or PET/CT. These studies may not detect testicular disease, and some advocate for testicular ultrasound in men over the age of 60 (O’Neill et al. 1995). Recently the utilization of liquid biomarkers for clinical diagnosis has increased. Specifically, the assessment of CSF cell free DNA to detect tumor specific mutations, such as MYD88 L265P mutation, or testing micro-RNAs can also be utilized to diagnose CNS lymphoma (Watanabe et al. 2019; Yamagishi et al. 2021; Zajdel et al. 2019; Zorofchian et al. 2018). CSF studies that are diagnostic may preclude the need for surgical biopsy, however, not every institution may have the capability to perform these studies, or the studies may require extensive amounts of time to complete. Nonetheless, CSF may allow for a less invasive approach for diagnosis in certain cases and its utilization is likely to increase over time as the capability to perform the studies increases.

Glucocorticoid use is thought to result in the disappearance of lymphoma cells, making diagnosis more difficult (Yim et al. 2020). The extent of impact of corticosteroids on pathologic diagnosis has varied in the literature. Some authors have found that short-term use with residual imaging findings does not frequently affect diagnostic yield (Binnahil et al. 2016; Bullis et al. 2020). Others have found the loss of lymphoma cells histologically as a result of steroid use may lead to incorrect diagnoses of demyelination or other inflammatory reactions (Barrantes-Freer et al. 2018; Choi et al. 2006; Lukas et al. 2018). In our patient cohort, the percentage of patients receiving steroids prior to brain biopsy is comparable in the non-diagnostic group versus diagnostic group (50% vs. 54.2%), supporting the former argument. Further study is needed to directly compare the non-diagnostic risk and treatment delay risk on the long-term outcome in patients with presumed CNS lymphoma.

Limitations

This is a retrospective review from a single institution and is limited by the small sample size, which inherently introduced bias and posed statistical challenges. We attempted to mitigate this by using penalized likelihood such as the Firth bias-corrected likelihood ratio. The study also lacks performance status as one of the prognostic markers, due to high rate of missing data in historical cohort. Additionally, the entire study cohort was evaluated by the same neuro-oncology team and biopsied by a select few surgeons. The consistency was helpful in reducing treatment-related confounders but could potentially limit the ability to generalize our results. Multicenter studies with a prospective design may provide more robust data to address to address the above shortcomings. Finally, there are many factors that come into play when it comes to long term survival for CNS lymphoma patients. Our overall survival analysis highlights the risk factors related to brain biopsy but certainly is not all-encompassing.

Conclusion

Stereotactic brain biopsy is an effective mean to obtain diagnosis of CNS lymphoma, but it is not without risk. Several intraoperative variables, including biopsy of the deep-brain structures, multiple depth biopsies, and intraoperative bleeding, are directly associated with higher postoperative morbidity and mortality in patients with suspected CNS lymphoma. These factors should be considered prior to attempting biopsy as the risk associated with brain biopsy can potentially outweigh its diagnostic benefit in some groups. Further risk stratification for patient selection is necessary to optimize the clinical utility of brain biopsy.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

The study was conceived and planned by AL, NS, EG, and WNA. Data were acquired by AL, NS, OS, SES, JAH, TH, BGT, and ASP. AL analyzed and interpreted the data and performed statistical analysis. The article was drafted by AL, SS, NS and revised by SS, YU, and WNA. Supervisory and administrative support were provided by WNA.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Data availability

The dataset supporting the conclusions of this article is included within the article and its additional files.

Declarations

Conflict of interest

The authors have no relevant financial or non-financial interests to disclose.

Ethical approval

This is an observational study. The University of Michigan Institutional Review Board issued an exemption of the study based on its design.

Consent to participate

Informed consent was waived by the University of Michigan Institutional Review Board based on the study design.

Consent for publication

Not applicable.