Diagnostic yield of postmortem brain examination following premortem brain biopsy for neoplastic and nonneoplastic disease

Abstract

Medical autopsies have decreased in frequency due in part to advances in radiological techniques and increased availability of molecular and other ancillary testing. However, premortem diagnosis of CNS disease remains challenging; while ∼90% of brain tumor biopsies are diagnostic, only 20%–70% of biopsies for presumed nonneoplastic disease result in a specific diagnosis. The added benefits of performing an autopsy following surgical brain biopsy are not well defined. A retrospective analysis was performed of patients who underwent brain biopsy and autopsy at Brigham and Women’s Hospital from 2003 to 2022. A total of 135 cases were identified, including 95 (70%) patients with primary CNS neoplasms, 16 (12%) with metastatic tumors, and 24 (18%) with nonneoplastic neurological disease. Diagnostic concordance between biopsy and autopsy diagnosis was excellent both for primary CNS neoplasms (98%) and metastatic tumors (94%). Conversely, patients with nonneoplastic disease received definitive premortem diagnoses in 7/24 (29%) cases. Five (21%) additional patients received conclusive diagnoses following autopsy; 8 (33%) received a more specific differential diagnosis compared to the biopsy. Overall, autopsy confirmed premortem diagnoses or provided new diagnostic information in 131/135 (97%) cases, highlighting the value in performing postmortem brain examination in patients with both neoplastic and nonneoplastic diseases.

INTRODUCTION

Despite the many benefits of performing medical autopsies, it is currently estimated that fewer than 5% of all hospital deaths in the United States receive a postmortem examination (1, 2). While an autopsy may not be required to determine cause and manner of death in some hospital-based cases, additional clinical information can be obtained and there is significant research and educational value in confirming premortem diagnosis and documenting extent and progression of disease. For cases in which the premortem diagnosis is less clear, obtaining tissue samples at the time of autopsy for histology, molecular diagnostics, and other ancillary testing (e.g. microbiological cultures) may yield a definitive diagnosis. This is particularly useful for neurologic diseases of uncertain etiology, for which premortem diagnosis remains especially challenging.

Compared to the well-established role of CNS biopsy in neoplastic disease, where diagnostic yield approaches 100% (3, 4), there is a lack of consensus regarding the sensitivity, specificity, clinical indications, and timing of CNS biopsy in patients with idiopathic neurologic decline. Previous studies have explored the utility of CNS biopsy in patients with HIV/AIDS (5–7), neurodegenerative disorders (8, 9), and adult and pediatric patients with both insidious and rapid unexplained neurological deterioration (10–18). In these groups, the diagnostic yield of biopsies varied from 20% to 72%; however, brain autopsy findings were rarely examined, and if present, only represented a subset of the studied cohorts.

Whether the diagnostic ambiguity of premortem CNS biopsy results from sampling bias or other factors requires additional clarity. To address this issue, we performed a retrospective review of all patients that received at least 1 premortem CNS surgical biopsy and subsequent brain autopsy at a large academic tertiary care facility in Boston, Massachusetts over a 20-year period from 2003 to 2022. To our knowledge, there are currently no published studies with the primary aim of evaluating CNS biopsies in both neoplastic and nonneoplastic disease in the context of postmortem neuropathological examination.

MATERIALS AND METHODS

This study was approved by the Mass General Brigham Institutional Review Board, under a waived consent protocol. A systematic retrospective study was performed at Brigham and Women’s Hospital ([BWH], Boston, MA), via review of electronic reports. Cases were included in the study if a brain autopsy report was generated between January 2003 and December 2022 and a prior brain biopsy surgical pathology report was available for review from 1987 to 2022, either from a procedure performed at BWH or from slides generated at an outside institution but submitted to BWH for consultation. Age, sex, premortem CNS biopsy diagnoses, relevant prebiopsy imaging features, ancillary test results, and final neuropathology autopsy diagnosis were extracted from surgical biopsy report(s), autopsy report, and review of available electronic medical records.

Cases were classified according to final autopsy diagnosis as either primary CNS neoplasm, metastatic disease, or nonneoplastic CNS disease. Tumor diagnoses reflect the 5th edition (2021) of the WHO Classification of Tumors of the Central Nervous System (19). Tumors previously classified as glioblastoma prior to 2021 based on histological features but lacking IDH1/IDH2 testing are designated as high-grade glioma, NOS. Histological concordance and discordance were defined as similar (e.g. within tumor class such a glioma or metastatic carcinoma) or different findings (e.g. nonspecific reactive changes vs high-grade glioma), respectively, reported at biopsy and autopsy. Both histologically concordant and discordant cases were further subdivided depending on whether a definitive diagnosis was rendered. In cases without a definitive diagnosis, the presence of a more specific differential diagnosis was assessed by review of the pathologist comment section of the biopsy and autopsy reports. Figure graphs were prepared using Microsoft Excel. Statistical significance was calculated used chi-square test with a p value < 0.05.

RESULTS

From 2003 to 2022, a total of 135 patients (37% female; median age 58 [range: 22–87] years at the time of death) were identified with brain autopsy following premortem CNS biopsy (Fig. 1). Primary CNS neoplasms were most common (n = 95, 70%), including 47 glioblastoma-IDH-wildtype, 28 high-grade glioma, NOS, 8 astrocytoma, IDH-mutant, 2 oligodendroglioma, IDH-mutant and 1p/19q co-deleted, 1 diffuse midline glioma, H3 K27-altered, 3 meningioma, 1 schwannoma, 1 adamantinomatous craniopharyngioma, 1 pituitary adenoma, and 3 primary CNS lymphomas. Metastatic tumors (n = 16, 12%) included 5 lung adenocarcinomas, 3 breast adenocarcinomas, 3 renal cell carcinomas, 2 lung squamous cell carcinomas, and 3 hematologic malignancies. Nonneoplastic disease (n = 24, 18%) included 8 inflammatory/infectious cases, 5 demyelinating, 2 infarction, 2 vascular malformation, 2 cerebral amyloid angiopathy, 2 with findings of neurodegenerative disease, 1 with findings associated with chronic seizure activity, 1 with radiation injury, and 1 with findings associated with toxic injury.

Figure 1.

Case overview by final autopsy diagnosis. The pie chart on the right shows percentages of the overall cohort consisting of primary CNS neoplasms (blue), metastatic disease (gray), and nonneoplastic disease (green). Specific diagnoses within each category are listed on the left.

Primary CNS neoplasms were found to be diagnostically concordant in 93/95 (98%) cases, in which the biopsy findings were morphologically similar or identical to the autopsy. Two surgical cases yielded nondiagnostic biopsies and were subsequently diagnosed at autopsy as glioblastoma, IDH-wildtype, CNS WHO grade 4 and high-grade glioma, NOS with a primary leptomeningeal gliomatosis pattern (Table 1). A total of 11/86 (13%) gliomas differed in histologically relevant grading features at biopsy versus autopsy (Table 2). In 3 cases of glioblastoma, IDH-wildtype CNS WHO grade 4, biopsies lacked vascular proliferation and necrosis. In 2 cases, a definitive diagnosis could be made by molecular features (i.e. TERT promoter mutation and/or polysomy 7 with monosomy 10), and vascular proliferation and/or necrosis were subsequently identified at autopsy. In the other case, definitive molecular features were absent, but necrosis was identified at autopsy, confirming the diagnosis. Two cases of astrocytoma, IDH-mutant increased from grade 2 at biopsy to grade 4 at autopsy due to the presence of mitoses, vascular proliferation, and necrosis. Similarly, 1 case of oligodendroglioma, IDH-mutant, and 1p/19q-codeleted increased from grade 2 at biopsy to grade 3 at autopsy due to the presence of mitoses, vascular proliferation, and necrosis. Five additional cases lacked molecular features required for definitive classification using the WHO 5th edition criteria, and were all retrospectively classified in this study as high-grade glioma, NOS. One case lacked mitoses, vascular proliferation, or necrosis on biopsy but was found to have all 3 at autopsy. The 4 other cases had varying numbers of mitoses at biopsy without vascular proliferation or necrosis and were subsequently found at autopsy to have vascular proliferation in 2 cases and necrosis in all 4 cases. In 7 of these cases, histologic differences were likely a result of tumor progression, in which the tumors did not exhibit MRI contrast enhancement prior to initial biopsy; however, over months to years many of these tumors subsequently exhibited enhancement on interval imaging correlating with higher grade histology. In 4 cases, the lack of higher grade histological features on initial biopsy may have been due to sampling. Three of these patients presented with enhancing lesions on MRI but lacked vascular proliferation or necrosis in the biopsy tissue that was subsequently identified at autopsy. In the remaining case, the patient survived less than a month following initial biopsy, and molecular features of glioblastoma were detected by chromosomal microarray.

Table 1.

Cases with nondiagnostic biopsy and autopsy examination resulting in definitive diagnosis

	Biopsy diagnosis	Autopsy diagnosis
Primary CNS neoplasms (n = 2)	Mildly hypercellular brain tissue with some atypical cells	High-grade glioma, NOS with primary leptomeningeal gliomatosis pattern
	Gray and white matter without diagnostic abnormalities	Glioblastoma, IDH-wildtype, CNS WHO grade 4
Metastasis (n = 1)	White matter with vacuolization, gliosis, and axonal injury	Diffuse large B-cell lymphoma (systemic primary)
Nonneoplastic disease (n = 5)	Minute fragment of paucicellular tissue	Spinal cord with radiation injury
	Vessels with extensive acute and chronic inflammation	Cytomegalovirus encephalitis
	T-cell inflammatory infiltrate involving leptomeninges and cerebellar cortex	Powassan encephalitis
	Cerebellar parenchyma with Bergmann gliosis and loss of Purkinje neurons	Powassan rhombencephalitis
	Cerebellar cortex with focal reactive changes	Cerebral amyloid angiopathy

Table 2.

Gliomas with changes in final diagnosis or grading at autopsy

Initial biopsy diagnosis and high-grade histological features *	MRI findings prior to initial biopsy (subsequent imaging findings if available)	Pertinent molecular features associated with initial biopsy tissue	Interval between initial biopsy and autopsy (months)	Autopsy diagnosis and high-grade histological features *
High-grade glioma, NOS; scattered mitoses, no vascular proliferation or necrosis	Nonenhancing	EGFR gain but not amplification (CISH)	21	High-grade glioma, NOS; mitoses, vascular proliferation, and necrosis
High-grade glioma, NOS; scattered mitoses, no vascular proliferation or necrosis	Nonenhancing (enhancement on repeat imaging)	No molecular testing	22	High-grade glioma, NOS; mitoses and necrosis
Infiltrating glioma, NOS; no mitoses, vascular proliferation, or necrosis	Enhancing	No EGFR amplification (CISH)	17	High-grade glioma, NOS; mitoses, vascular proliferation, and necrosis
High-grade glioma, NOS; scattered mitoses, no vascular proliferation or necrosis	Minimal enhancement	No molecular testing	45	High-grade glioma, NOS; mitoses, vascular proliferation, and necrosis
High-grade glioma, NOS; rare mitoses, no vascular proliferation or necrosis	Nonenhancing (enhancement on repeat imaging)	1p/19q retained (FISH); EGFR gain but not amplification (CISH)	20	High-grade glioma, NOS; mitoses and necrosis
High-grade glioma, NOS; scattered mitoses, no vascular proliferation or necrosis	Enhancing	IDH1 R132H negative (IHC); focal TP53 loss (aCGH); negative for IDH1/IDH2 mutations (Oncomap NGS)	10	Glioblastoma, IDH-wildtype, CNS WHO grade 4; mitoses and necrosis
Astrocytoma, IDH-mutant, CNS WHO grade 2; no mitoses, vascular proliferation, or necrosis	Nonenhancing (enhancement on repeat imaging)	IDH1 R132H mutated (IHC); 1p/19q retained (FISH); partial 10q loss (including PTEN), 13q loss (including RB1), and 22q loss (including NF2)(aCGH)	15	Astrocytoma, IDH-mutant, CNS WHO grade 4; mitoses, vascular proliferation, and necrosis
Glioblastoma, IDH-wildtype, CNS WHO grade 4; no mitoses, vascular proliferation, or necrosis	Nonenhancing	IDH1 R132H negative (IHC); ATRX expression lost (IHC); polysomy 7, monosomy 10, monosomy 22, chromosome 13 loss	<1	Glioblastoma, IDH-wildtype, CNS WHO grade 4; mitoses and vascular proliferation
Astrocytoma, IDH-mutant, CNS WHO grade 2; no mitoses, vascular proliferation, or necrosis	Nonenhancing (enhancement on repeat imaging)	EGFR gain but not amplification (CISH); 1p/19q retained (FISH); IDH1 R132H positive (IHC); ATRX expression lost (IHC)	144	Astrocytoma, IDH-mutant, CNS WHO grade 4; mitoses, vascular proliferation, and necrosis
Glioblastoma, IDH-wildtype, CNS WHO grade 4; scattered mitoses, no vascular proliferation or necrosis	Nonenhancing	IDH1 R132H negative (IHC); EGFR p. A289T, PTEN p. G129R, TERT C228T, polysomy 7, monosomy 10, and 1–2 copy loss of CDKN2A/CDKN2B, negative for IDH1/IDH2 mutations (Oncopanel NGS)	42	Glioblastoma, IDH-wildtype, CNS WHO grade 4; mitoses, vascular proliferation, and necrosis
Oligodendroglioma, IDH-mutant, and 1p/19q-codeleted, CNS WHO grade 2; no mitoses, vascular proliferation, or necrosis	Nonenhancing (enhancement on repeat imaging)	1p/19q co-deletion (FISH), IDH1 R132 mutant (Oncomap NGS)	222	Oligodendroglioma, IDH-mutant, and 1p/19q-codeleted, CNS WHO grade 3; mitoses, vascular proliferation, and necrosis

^*Tumor names and grades reflect diagnostic criteria from the 5th edition (2021) of the WHO Classification of Tumors of the Central Nervous System.

Abbreviations: aCGH, array comparative genomic hybridization; CISH, chromogenic in situ hybridization; FISH, fluorescence in situ hybridization; IHC, immunohistochemistry; NGS, next-generation sequencing.

Metastatic tumors involving the CNS were diagnostically concordant in 15/16 (94%) cases between biopsy and autopsy, including all cases of lung, breast, and renal cell carcinoma. One patient with a history of testicular diffuse large B cell lymphoma presented with changes in mental status and was found to have multiple lesions on MRI with evidence of obstructive hydrocephalus. Brain biopsy at the time of ventriculo-peritoneal shunt placement only showed white matter with vacuolization, gliosis, and axonal injury; autopsy confirmed involvement of lymphoma in multiple brain areas (Table 1).

Nonneoplastic CNS disease showed the lowest rate of diagnostic concordance between biopsy and autopsy with 7/24 (29%) cases. This included 2 cases of arteriovenous malformation, and 1 case each of cerebral amyloid angiopathy, progressive multifocal leukoencephalopathy, cerebral mucormycosis, Alzheimer disease-type pathology, and nonnecrotizing granulomatous leptomeningitis. A definitive diagnosis was made at autopsy in 5/17 (29%) of the remaining cases (Table 1), including 2 cases of Powassan encephalitis, and 1 each of cytomegalovirus encephalitis, cerebral amyloid angiopathy, and spinal cord with radiation injury. In the remaining 12/135 (9%) cases, both the biopsy and the autopsy findings were reported descriptively with a more specific differential diagnosis included in the autopsy report in 8/12 (67%) cases (Table 3). This included 4 cases with inflammatory/infectious patterns without specific etiologic agents identified, 3 demyelinating lesions of unclear etiology, 3 infarctions of unclear etiology, 1 with findings associated with chronic seizure activity, and 1 with findings associated with toxic injury. More specific differential diagnoses were provided in all cases of demyelinating lesions, infarction, and toxic injury in the absence of a specific clinicopathologic entity, while the differential for inflammatory/infective lesions remained as similarly broad as the biopsy diagnoses for most cases.

Table 3.

Overview of nondiagnostic cases

Disease category	Histologic concordance between biopsy and autopsy specimens	Cases with more specific differential diagnosis following autopsy examination
Inflammatory/infective, n (%)	4/4 (100)	1/4 (25)
Demyelinating, n (%)	2/3 (66)	3/3 (100)
Infarction, n (%)	2/3 (66)	3/3 (100)
Findings associated with chronic seizure activity, n (%)	1/1 (100)	0/1 (0)
Findings associated with toxic injury, n (%)	0/1 (0)	1/1 (100)

DISCUSSION

In this study, we systematically examined a large series of autopsy cases with the goal of determining the utility of postmortem examination following premortem surgical biopsy for both neoplastic and nonneoplastic CNS disease. Overall, we found both high histological concordance (85%) between biopsy and autopsy and high diagnostic yield (91%). Autopsy led to a specific diagnosis in 8/20 (40%) cases following a nondiagnostic biopsy and provided a more specific differential diagnosis in 8/20 (40%) cases. Thus, autopsy confirmed premortem diagnoses (as well as document progression to higher grade tumors), or provided new diagnostic information in 131/135 (97%) cases.

Our findings of excellent diagnostic yield (100%) and histologic concordance for neoplastic disease (97%) (Table 4) are consistent with the current literature. In patients with discrete brain lesions, multiple factors have been shown to affect diagnostic yield of CNS biopsies including anatomic location and volume of the lesion (20), number of tissue pieces collected (21), performing an intraoperative histologic smear (3), and surgeon experience (20). In our study, only 2 patients with primary CNS neoplasms had nondiagnostic biopsies. One patient presented with an enhancing lesion that was clinically suspected to represent a demyelinating process; however, biopsies did not show abnormalities in gray or white matter, and autopsy demonstrated extensive involvement of the commissural white matter by glioblastoma (Fig. 2A). Biopsy from the second patient showed mildly hypercellular brain tissue with some atypical cells, and autopsy showed primary leptomeningeal gliomatosis broadly involving the neuroaxis (Fig. 2B). In both cases, the nondiagnostic biopsies could be attributed to sampling error, which was overcome by comprehensive evaluation at autopsy.

Table 4.

Diagnostic yield and concordance of brain autopsy with surgical biopsies for neoplastic and nonneoplastic disease

	Neoplastic disease	Nonneoplastic disease	p value *
Diagnostic yield of brain autopsy following surgical biopsy, n (%)	111/111 (100)	12/24 (50)	<0.00001
Concordance between autopsy and surgical biopsy diagnoses, n (%)	108/111 (97)	7/24 (29)	<0.00001

^*p value from chi-square test. p < 0.05 considered statistically significant.

Figure 2.

Representative histology from cases with nondiagnostic biopsy followed by definitive diagnosis at autopsy. (A) Biopsy showed fragments of gray and white matter without diagnostic abnormalities (left, H&E, 10× objective), while autopsy revealed diagnostic features of glioblastoma, IDH-wildtype, CNS WHO grade 4 (right, H&E, 4× objective). (B) Biopsy showed mildly hypercellular brain tissue with some atypical cells (left, H&E, 10× objective), while autopsy revealed high-grade glioma, NOS with a primary leptomeningeal gliomatosis pattern (middle, H&E, 4× objective; right, H&E, 4× objective). (C) Biopsy showed vessels with extensive acute and chronic inflammation (left, H&E, 10× objective), and autopsy showed necrosis (middle, H&E, 4× objective) and scattered CMV-positive cells (right, CMV immunohistochemistry [IHC], 40× objective), consistent with cytomegalovirus encephalitis. (D) Biopsy showed cerebellar cortex with focal reactive changes (left, H&E, 10× objective), and autopsy showed cerebral amyloid angiopathy (middle, H&E, 10× objective) confirmed by beta-amyloid IHC (right, beta-amyloid IHC, 10× objective).

Conversely, in nonneoplastic CNS disease, we found that diagnostic yield of autopsy was moderate (50%) and histologic concordance was low (29%) (Table 4). In our cohort, all cases of nonneoplastic CNS disease with a diagnostic premortem biopsy demonstrated specific diagnostic histological features including visible microorganisms, abnormal blood vessels, or accumulation of protein aggregates. These results also reflect those reported in the literature. For example, Javedan et al (8) reviewed biopsies in 50 patients with idiopathic progressive cognitive decline, 10 of which were diagnostic (20%) including 6 cases of CNS infection (identified by immunohistochemistry, special stains, and/or visible microorganisms) and 1 case of Alzheimer disease with visible neurofibrillary tangles and senile plaques. In a more recent study, Lockhart et al (11) reviewed 331 biopsies for unknown neurological disease. Seventy-one percent of the biopsies were diagnostic, most commonly due to infection, cerebral amyloid angiopathy, and demyelination.

We identified several cases of nonneoplastic CNS disease where autopsy examination resulted in a definitive diagnosis following a nondiagnostic premortem biopsy. These cases included 2 patients with Powassan virus infection and 1 patient with cytomegalovirus (CMV) infection. The diagnoses of Powassan encephalitis/rhombencephalitis would not have been possible without procurement of tissue at the time of autopsy for analysis by the Centers for Disease Control and Prevention (CDC) (22). In the case of CMV encephalitis, characteristic viral cytopathic inclusions were sparse on autopsy and absent on the small amount of biopsy tissue, which showed extensive necrosis and inflammation involving brain parenchyma and blood vessels lead to the impression of vasculitis on biopsy (Fig. 2C). In a patient with a history of squamous cell carcinoma, the clinical and radiologic suspicion favored recurrent malignancy, while the biopsy showed minute fragment of paucicellular tissue; autopsy revealed spinal cord with radiation-associated injury. Finally, a patient who presented with a cerebellar infarct, initially showed cerebellar cortex with focal reactive changes on biopsy, while autopsy showed cerebral amyloid angiopathy (Fig. 2D). The combination of additional sampling and appropriate ancillary testing was essential for making these nonneoplastic diagnoses.

While this study has many strengths, including 20 years of data from a large academic medical center with a dedicated neuropathology service, there are some important limitations to consider. Inclusion of cases from a single-center limits generalizability to other practice settings with different patient populations, which may have greater or fewer resources available for surgical and autopsy diagnosis. Reliance of surgical pathology and autopsy reports also limits the ability to standardize the diagnoses and collection of histological data; however, this is somewhat mitigated by the limited number of neuropathologists who rendered the original diagnoses. Finally, lack of molecular data for older cases limits the ability to classify some tumors using the current WHO diagnostic criteria.

With the availability of continuously improving CNS imaging methods, neurosurgical techniques, and less invasive testing using CSF or blood, many patients with neurological illness are successfully diagnosed and appropriately treated. Autopsy serves an important quality control function by confirming premortem diagnoses, which occurred in 115/135 (85%) of cases in this cohort, including all cases of primary CNS neoplasms. Increased tumor grade or presence of higher grade histological features was identified in 11/86 (13%) gliomas at autopsy compared to biopsy, which in most cases could be accounted for by tumor progression, while under sampling may have accounted for tumor grade discrepancy in a small number of patients (n = 3). A subset of our patients with high-grade gliomas were consented for tissue collection at the time of autopsy, which has tremendous utility in the study of treatment effect, both for standard of care therapy and investigational agents including oncolytic viruses (23, 24). The lack of definitive diagnosis following autopsy did occur in 12 (9%) cases in this cohort, however, valuable information was still obtained by ruling out potential alternative diagnoses. While it is not feasible to perform an autopsy for every patient with neurological disease, this study does suggest that valuable information can be gained from performing autopsies on patients with premortem CNS biopsies and should be considered in this setting. We therefore recommend offering brain autopsy for any patient who has undergone a neurosurgical procedure, demonstrated abnormal neuroradiological findings, exhibited acute or chronic neurological symptoms (e.g. sensory, motor, or cognitive), or has an unexplained cause of death.

FUNDING

None declared.

CONFLICT OF INTEREST

The authors have no duality or conflicts of interest to declare.