Original Articles
An Algorithmic Approach to the Brain Biopsy—Part I
B. K. Kleinschmidt-DeMasters, MD; Richard A. Prayson, MD
● Context.—The formulation of appropriate differential diagnoses for a slide is essential to the practice of surgical
pathology but can be particularly challenging for residents
and fellows. Algorithmic flow charts can help the less experienced pathologist to systematically consider all possible choices and eliminate incorrect diagnoses. They can
assist pathologists-in-training in developing orderly, sequential, and logical thinking skills when confronting difficult cases.
Objective.—To present an algorithmic flow chart as an
approach to formulating differential diagnoses for lesions
seen in surgical neuropathology.
Design.—An algorithmic flow chart to be used in teaching residents.
Results.—Algorithms are not intended to be final diag-
nostic answers on any given case. Algorithms do not substitute for training received from experienced mentors nor
do they substitute for comprehensive reading by trainees
of reference textbooks. Algorithmic flow diagrams can,
however, direct the viewer to the correct spot in reference
texts for further in-depth reading once they hone down
their diagnostic choices to a smaller number of entities.
The best feature of algorithms is that they remind the user
to consider all possibilities on each case, even if they can
be quickly eliminated from further consideration.
Conclusions.—In Part I, we assist the resident in learning
how to handle brain biopsies in general and how to distinguish nonneoplastic lesions that mimic tumors from true
neoplasms.
(Arch Pathol Lab Med. 2006;130:1630–1638)
OVERVIEW TO THE USE OF THE
ALGORITHMIC APPROACH
It is critical in surgical pathology to develop appropriate
differential diagnostic lists for the microscopic lesion you
are seeing on your patient, whether you are a resident,
fellow, or senior pathologist in practice. Trainees in surgical pathology, however, have the additional task of trying to adopt the thought processes of their mentors. These
come only through long practice of the specialty—how
does an experienced pathologist rationally approach a
tough case?
All too often, senior surgical pathologists find it difficult
to break down for new trainees the multiple individual
steps they took in their own mind to arrive at the final
correct diagnosis. With experience, the sequential thinking
process becomes so rapid that it may be almost rote for
senior diagnosticians to arrive at a correct final diagnosis.
Indeed, they may no longer be consciously aware of what
algorithm they did utilize. So when residents ask how the
attending surgical pathologist decided the diagnosis on a
given slide was ‘‘adenocarcinoma,’’ the response may be
‘‘because it looks like it!’’
The following algorithmic approach to brain biopsy
specimens has been formulated to assist residents in their
early years of training (Figure 1). The approach may also
be helpful, however, for pathologists in practice at hospitals where there is a relatively low volume of brain biopsies every year. The formula is especially helpful at the
time of frozen section.
These diagrammatic charts attempt to parse out the critical individual questions one needs to ask oneself when
approaching a diagnostically challenging central nervous
system (CNS) case (Figure 1).
Flow diagrams cannot replace the necessary reading by
trainees of reference neuropathology books, and several
neuropathology texts of particular usefulness to residents
are suggested.1–8 Unfortunately, however, you cannot use
a textbook effectively if you are not able to narrow down
the differential diagnoses to a limited number of entities.
Hence these flow charts are meant to be tear-out ‘‘cheat
sheets’’ that will direct the user to the correct spot in reference books where one can read about diagnostic choices
in greater depth.
Accepted for publication April 13, 2006.
From the Departments of Pathology, Neurology, and Neurosurgery,
University of Colorado Health Sciences Center, Denver (Dr Kleinschmidt-DeMasters); and the Department of Anatomic Pathology,
Cleveland Clinic Foundation, Cleveland, Ohio (Dr Prayson).
The authors have no relevant financial interest in the products or
companies described in this article.
Reprints: B. K. Kleinschmidt-DeMasters, MD, University of Colorado
Health Sciences Center, Pathology, Neurology, and Neurosurgery, 4200
E Ninth Ave, B216, Denver, CO 80262 (e-mail: bk.demasters@
uchsc.edu).
Algorithmic flow diagrams are also useful in that they
can steer the user away from incorrect diagnoses. Finally,
the use of an algorithm can underscore cases in which
outside consultation may be necessary.
We have no illusions that this is a perfect algorithmic
chart. But we do share with you what we are using for
our own trainees and note that it seems to work for us in
daily practice. We hope, in fact, that other surgical pathologists and neuropathologists will be motivated to im-
1630 Arch Pathol Lab Med—Vol 130, November 2006
See also p 1602.
Algorithmic Approach to Brain Biopsy: Part I—Kleinschmidt-DeMasters & Prayson
Figure 1. An algorithmic approach to use
when first encountering biopsy material from
the central nervous system at the time of frozen or permanent section. CNS indicates central nervous system.
prove on it, add to it, edit it, indeed write all over it, to
meet their own needs at their own training institutions or
hospitals.
In Part I, we provide a very basic ‘‘get in the game, get
in the ballpark’’ flow chart that asks questions that assist
in separating out normal, reactive, and neoplastic processes (Figure 1). The reader will note that we have spent most
of our efforts at getting to the point of determining whether something is or is not a neoplasm and relatively little
time at the end discussing how to sort out various types
of primary brain tumors from each other. In part, this is
because the most serious error one can make in neuropathology is to overcall a nonneoplastic process (such as
demyelinating disease) as a neoplasm. But it is also an
acknowledgment that diagnosing specific tumor types
does not lend itself readily to an algorithmic flow chart.
We recognize that at this point in the thinking process, the
user will have available a variety of excellent neuropathology texts1–8 that can be used to further narrow down
the diagnosis. Our goal is to get the reader to the correct
chapter(s) in the text(s).
In Part II of ‘‘An Algorithmic Approach to the Brain
Arch Pathol Lab Med—Vol 130, November 2006
Biopsy’’ (in this issue of ARCHIVES), we will provide 2
separate algorithms. The first will be directed at working
through granulomatous disease of the CNS. In the second
algorithm in Part II, we will tackle the point in the flow
diagram for ‘‘Are there abnormal blood vessels or macrophages present?’’ Once again in Part II, we will place
our emphasis on nonneoplastic CNS entities, some of
which are easily confused with tumors.
PART I—FROZEN AND PERMANENT SECTIONS
Algorithm n. A step-by-step procedure . . . for solving a
problem in a finite number of steps.
1. Where Am I?
Normal cortex, especially the cerebellar granular cell
layer, deep layers of temporal or parietal cortex with
prominent normal neuronal satellitosis, and mesial temporal cortex, as well as several other selected sites in the
CNS such as infundibular stalk, posterior pituitary gland,
pineal gland, and peripheral nervous system (PNS) dorsal
root ganglia, can be mistaken for neoplastic entities.
Neuroanatomy is difficult, easily forgotten if you don’t
Algorithmic Approach to Brain Biopsy: Part I—Kleinschmidt-DeMasters & Prayson
1631
use it often, and daunting to some pathologists. Make
yourself a simple box of 7⫹ normal sections from the next
autopsy you perform, taken from various sites in the CNS;
keep it in your office or in the frozen section room as a
handy reference and comparison.
2. Could This Tissue Be Normal?
Neurosurgical specimens, unlike almost any other biopsy specimens, are often not fully representative of the
lesion or may not be in the lesion at all, despite neuroimaging guidance. The neurosurgeon may insist that he
or she is ‘‘there’’ but even millimeters difference in some
lesions can determine whether or not diagnostic material
is received by the pathologist. Your job at frozen section
(FS) is to identify lesional tissue consistent/compatible
with the clinical and neuroimaging features of the case,
not to render a specific diagnosis. Do not let the neurosurgeon’s badgering make you ‘‘over call’’ tissue as the
lesion, only to discover the next morning on permanent
sections that what you had was normal cortex and the
neurosurgeon had been superficial to the lesion.
There are a few variations on how to handle the actual
tissue sent for FS. One of the issues is whether you should
perform a touch preparation (TP), perform a crush preparation (CP), or go directly to FS. Both touch and crush
preparations yield rapid results, and in many instances a
touch or squash preparation is sufficient to make the diagnosis, obviating the need for freezing the tissue. An advantage of TPs is that the TP and FS can be performed on
the same tissue sample, with the TP yielding excellent cytologic detail and the FS further revealing architectural
features of the process. Some things, such as metastatic
lesions and pituitary adenomas, can be diagnosed on TP
alone. One can use both TP and FS on the same case at
the time of intraoperative consultation in order to optimize their complementary information. We do rapid
‘‘snap’’ FS (using either liquid nitrogen or a dry ice/ethanol bath mixture in a vacuum flask) to minimize ice crystal artifacts in the permanent section. Although it takes a
little longer, the use of hematoxylin-eosin stain at the time
of FS is very helpful for many CNS/PNS processes because it better reveals cytoplasmic qualities, especially eosinophilic glial cell cytoplasm, than does the polychrome
‘‘quick’’ stain. The CP technique is favored by many neuropathologists. There is nothing wrong with doing both
CP and TP on the same case if there is adequate tissue,
but obviously CP ⫹ FS on the same fragment cannot be
performed. If using only the CP for diagnosis on minute
fragments of tissue, you should not use all the tissue that
you are sent for the intraoperative consultation since nothing will be left for permanent section.
A frequent problem revolves around tiny stereotactic biopsy specimens and whether you should use the entire
single core you are sent, or only part, for the frozen section. Remember the goal: to leave the FS room with tissue
that you are confident will be lesional and representative.
Therefore, you might start by freezing half the specimen
you were sent, provided that it is not minuscule. If you
can make the diagnosis by using only half the sample,
that’s great. But if you cannot, then consult the neurosurgeon and explain that you will need to use all of the core
that was sent for the FS and ask whether you can be provided with more tissue at the same level for permanent section when you are done with the FS. Neurosurgeons usually can easily do this with no risk to the patient; if they
1632 Arch Pathol Lab Med—Vol 130, November 2006
are utilizing a side cutting needle, they simply turn the
bore of the needle. This additional tissue should be sent
out for permanent sections while you are still in the FS
suite so that you can supervise that the tissue is properly
handled and fixed in formalin. If, however, this second
half of the first core was not diagnostic, then inquire as to
whether they are going to send you a deeper or more
superficial core for a second FS. Often what neurosurgeons first send out as the ‘‘target’’ is not as good for our
pathologic analyses as T ⫹ 10, T ⫺ 20 (ie, specimens obtained deep or superficial to the target). This interaction
between you and the neurosurgeon at the time of intraoperative consultation can often prompt the neurosurgeon
to do a serial sampling approach that will achieve a proper
diagnosis for your mutual patient.
3. Could This Be a Process That Mimics a Neoplasm?
Are Macrophages Present? Stop! Macrophages are
usually not present in large numbers in low-grade tumors
unless the tumor has been preoperatively treated with radiotherapy or chemotherapy. Macrophages are often present in highly necrotic high-grade tumors but usually not
in large clusters or sheets. Rule in/out infarction or demyelinating disease at the time of permanent section, with
special stains such as Luxol fast blue–periodic acid-Schiff
for myelin, in conjunction with antineurofilament or Bielschowsky for axons. Macrophages can be further identified through the use of CD68 and HAM-56 immunostaining.
Are Neutrophils Present? Stop! Neutrophils are almost never present in untreated glial tumors, and almost
no type of CNS vasculitis is characterized by a predominantly neutrophilic infiltrate. Inquire whether tissue has
been sent for cultures to the microbiology laboratory. Rule
in/out bacterial, atypical bacterial, or toxoplasmosis abscesses at the time of permanent section, with special
stains, remembering to include Gram tissue stain, Fite,
and/or prolonged Grocott methenamine silver for Nocardia species.
Are Granulomas Present? Stop! Focal granulomatous
infectious diseases, particularly in immunocompetent
hosts, can mimic brain tumors. Inquire whether fungal
and mycobacterial cultures have been sent. Noninfectious
processes, notably neurosarcoidosis and germ cell tumors
of the CNS, can also cause granulomas. Rheumatoid arthritis and Wegener granulomatosis also can form granulomas but usually involve the dura mater. Rule in/out specific fungal or mycobacterial infections at the time of permanent sections, with special stains including Grocott methenamine silver, periodic acid–Schiff, acid-fast bacilli, etc.
Are Numerous Lymphocytes Present? Stop! Viral encephalitis seldom comes to biopsy anymore, after the advent of cerebrospinal fluid polymerase chain reaction testing. Nevertheless, if you are seeing perivascular lymphocytic cuffing and think you may also be seeing microglial
clusters, inquire about viral cultures and possible previous
cerebrospinal fluid polymerase chain reaction.
Usually you will be dealing with a noninfectious process containing lymphocytes. Unlike the situation with
neutrophils and granulomas in CNS biopsy specimens,
lymphocytes are fairly common in several types of primary and secondary CNS tumors (ganglion cell tumors,
pleomorphic xanthoastrocytomas, gemistocytic astrocytomas, germ cell tumors, melanomas, malignant lymphomas). But if you cannot recognize the additional, coexis-
Algorithmic Approach to Brain Biopsy: Part I—Kleinschmidt-DeMasters & Prayson
tent glioma/metastasis/lymphoma cell population in the
lesion, then also consider demyelinating disease, polyarteritis nodosa, primary CNS vasculitis, and infections such
as cysticercosis and angioimmunoproliferative lesion/
lymphomatoid granulomatosis at permanent section. The
only additional thing to do at the time of FS if lymphoma
remains in your differential diagnosis is to consider sending a portion of the fresh tissue for flow cytometry. In
particularly perplexing cases where viral etiologies are
still a consideration, frozen tissue set aside for potential
tissue polymerase chain reaction testing and tissue for
electron microscopy is also recommended, provided sufficient material is available.
When you come to this point in the algorithm, where
you are trying to determine whether something is a reactive process, consider doing a TP. They beautifully highlight macrophages, neutrophils, and even the histiocytes
in granulomas but equally importantly may bring out the
cytologic features of lymphoma better than FS does (depending on the quality of frozen sections at your institution).
Are Abnormal Blood Vessels Present? Stop! Vascular
malformations likely to come to biopsy include arteriovenous malformation and cavernous angioma (cavernoma). Arteriovenous malformations often include few of the
altered arteries; don’t mistake these for venous angiomas.
Cavernous angiomas are particularly problematic if they
have ruptured and parts of the lesion are obliterated.
These show both extensive surrounding hemosiderin deposition and dystrophic mineralization, something that is
infrequent in tumors with abnormal vasculature. Several
types of CNS/PNS tumors may have notoriously profuse
vasculature that simulates a vascular malformation, notably schwannoma and pilocytic astrocytoma.
If the vessels also show destruction of vessel walls or
permeation by inflammatory cells, consider vasculitis. Abnormally thickened vessels prompt consideration of hypertensive vasculopathy, cerebral amyloid angiopathy (frequently seen in clot evacuations from elderly patients), or
rarely, cerebral autosomal dominant arteriopathy with
subcortical infarcts and leukoencephalopathy.
4. Could This Be Reactive Gliosis?
Several types of reactive gliosis exist and look quite different (Figure 2, A through I; Figure 3, A through I).
Could This Be Subacute Gliosis? Subacute reactive
gliosis with numerous plump gemistocytic astrocytes can
be easily mistaken for a gemistocytic astrocytoma. Features of subacute reactive gliosis not seen in gliomas are
as follows:
1. Reactive astrocytes are evenly dispersed within the
tissue, at approximately equal distances from each other
(Figure 2, A). In infarctions (Figure 2, H and I; Figure 3,
B and C) and demyelinating disease, the reactive astrocytes will be in a mosaic with macrophages in between
them. In subacute reactive gliosis near an abscess (Figure
2, B and C) or other mass lesion (Figure 2, D through G),
there will be few, if any, macrophages, but the astrocytes
remain dispersed in the background neuropil. In contrast,
neoplastic astrocytoma cells usually manifest an irregular
distribution and more disorganization, often causing them
to cluster or touch.
2. Reactive astrocytes have long, tapering, starlike (astro) cytoplasmic processes (Figure 3, A), as opposed to
Arch Pathol Lab Med—Vol 130, November 2006
the short and irregular blunted processes seen in gemistocytic neoplastic astrocytes (Figure 3, D). This can be seen
variably on hematoxylin-eosin and best on glial fibrillary
acidic protein (GFAP) immunostaining. Although GFAP is
not a tumor stain, it nicely reveals the differences between
processes of reactive versus neoplastic astrocytes (Figure
3, E and F).
3. Reactive astrocytes show an absence of nuclear atypia and high cell density. Although radiation therapy (external beam or stereotactic) can cause significant cytologic
atypia in astrocytes (Figure 3, G through I), it does not
result in true hypercellularity to the degree of tumor, provided one distinguishes the astrocytes from the accompanying macrophages that can also be present (Figure 3,
H and I). Remember that with biopsies undertaken to answer the question of radiation necrosis versus residual
growing tumor (usually glioma), your clinical colleagues
are asking what is the major process that you see to account for the neuroimaging changes? They want to know
whether to hold therapy or add a more efficacious therapy. Hence, if what you are seeing is predominantly bland
necrosis devoid of cellular response associated with fibrinoid vascular necrosis and severely hyalinized and altered
blood vessels, the 6 funny-looking astrocytes that are present in the biopsy material are not the issue.
4. Neoplastic gemistocytic astrocytes, but not reactive
astrocytes, keep ‘‘bad company’’ (ie, they are interspersed
with smaller, initially less conspicuous neoplastic astrocytes that manifest angular hyperchromatic nuclei and nuclear anaplasia and pleomorphism—look for them!).
Could This Be Chronic Gliosis? Biopsies are often
undertaken on certain slow-growing cysts and tumors that
classically induce chronic gliosis at their perimeter; if the
neurosurgeon has inadvertently given you tissue from the
edge of the true lesion, this will be a difficult problem to
resolve. Common examples include Rosenthal fiber formation and chronic gliosis around craniopharyngiomas,
hemangioblastoma, and pineal cysts. The presence of Rosenthal fibers is not specific to tumors.
More often when you are having difficulty distinguishing gliosis from glioma it is because the neurosurgeon is
on the infiltrating edge of a low-grade—or even highgrade—glioma. This usually happens near cortical graywhite matter junctions, the very spot where the neurosurgeon often starts by taking the first superficial ‘‘bite’’ of
tissue for FS! Also remember that the neurosurgeon can
be taking tissue from 2 to 3 cm deep to the surface but
still be at a gray-white matter junction.
Stop! At this point check the clinical and neuroimaging
features in the chart or call the operating room.
If the neuroimaging shows an enhancing process, especially
ring-enhancing mass, and you are struggling with whether this
is gliosis or glioma, the neurosurgeons are not in the optimal location for the biopsy (‘‘they are in the yard, not the
house’’). You should request more tissue. This is the optimal use for stereotactic biopsies; by redirecting the depth
of the biopsy needle the neurosurgeon usually can obtain
the junction between viable and nonviable tissue, which is
the optimal area for diagnostic yield for us in pathology.
What if the Neuroimaging Suggests a Nonenhancing
Mass and You Have Difficulty Distinguishing Gliosis
From Glioma? Given that it is a mass, you are most likely dealing with a low-grade glioma. Work with the neurosurgeon to obtain another piece of tissue if at all pos-
Algorithmic Approach to Brain Biopsy: Part I—Kleinschmidt-DeMasters & Prayson
1633
1634 Arch Pathol Lab Med—Vol 130, November 2006
Algorithmic Approach to Brain Biopsy: Part I—Kleinschmidt-DeMasters & Prayson
sible. Volume is everything in this situation, and even additional small pieces of tissue cumulatively add up.
If the neuroimaging shows a nonenhancing mass, the
slight hypercellularity you are seeing may be representative of the lesion, but it still is not diagnostic for you. Remember your goal—to leave the FS room with clearly lesional tissue, even if you cannot be more specific about the
diagnosis.
If you do not have tissue you think is clearly lesional,
we advocate not calling what you are seeing ‘‘gliosis’’ in
this situation; some neurosurgeons, when they hear the
pathologist render a diagnosis of ‘‘gliosis,’’ will automatically assume that they are truly in a glioma but the pathologist is just too reticent to call it. Some will even cease
the operation and obtain no more tissue, assuming the
pathologist will ‘‘get it right and diagnose glioma in the
morning.’’ If the tissue is nondiagnostic, call it that, or call
it ‘‘white matter, mildly hypercellular, cannot diagnose tumor.’’
What if the Neuroimaging Suggests a Nonenhancing
Ill-Defined Process and You Are Having Difficulty Distinguishing Gliosis from Glioma? Few biopsies are undertaken for diffuse ill-defined white matter processes
that induce chronic gliosis, such as leukoencephalopathies,
but once again, if that is truly what is in the differential,
then tissue volume is everything. Stereotactic biopsies
have a notoriously low yield in this situation, and whenever we are consulted in advance, we make certain that
the clinicians understand the statistical likelihood of a
yield on the particular procedure they are considering.
Once this is explained, the neurosurgeons almost always
opt for an open biopsy.
Distinguishing chronic gliosis from glioma can still be
difficult at the time of permanent section, particularly if
what you have to work with is tiny fragments of tissue or
biopsy specimens from gray matter or the gray-white matter junction (which usually represent the infiltrating edge
of a tumor). You can only work with what you have as a pathologist. While this is almost a nonissue in the rest of surgical pathology, the addendum comment of ‘‘biopsies may
not be fully representative of the lesion’’ is all too frequent
on neuropathology reports. Remember also that we have
no truly definitive ‘‘tumor immunostains’’ in neuropa-
thology to use at the time of permanent section. The pattern of immunostaining for MIB-1, TP53, and GFAP, when
taken collectively and only in conjunction with the hematoxylin-eosin stain, is usually the most helpful. Nevertheless, this re-emphasizes the need to assist the neurosurgeon in obtaining diagnostic tissue at the time of surgery—optimal tissue sampling is the best special stain we
have in neuropathology.
5. It’s Not a Reactive Process, It’s Neoplastic!
On most biopsy specimens and permanent sections, it’s
a ‘‘quick trip’’ through the above flow-chart questions before you arrive at this juncture in the algorithm.
Could This Tumor Be Metastatic? Typical features of
metastases include sharp demarcation from adjacent brain
tissue, little or no individual tumor cell infiltration, tight
cell-to-cell cohesion of metastatic tumor cells, absence of
interspersed neuropil, little or no microvascular proliferation, and absence of pseudopalisading around geographic areas of necrosis. Less certain criteria seen occasionally
in both metastases and high-grade gliomas include large
zonal necrosis with preservation of tumor cells around
blood vessels and cytologic monotony. Gliosarcomas (GSs)
can especially simulate metastases because of their relatively sharp circumscription, but usually the abundant interspersed reticulin fibers and microvascular proliferation
present in GSs distinguish these from metastatic disease.
Could This Tumor Be a Lymphoma? Remember that
lymphomas may be primary or secondary and that lymphomas can involve a variety of different compartments
in the brain (often with unique features), including epidural space, dura/subdural space, nerve roots, leptomeninges, and parenchyma. With the exception of primary
low-grade B-cell lymphomas of the dura, most primary
CNS lymphomas are parenchymal. Within parenchyma
they may be well-defined discrete metastatic-like parenchymal masses or quite diffuse and infiltrative. They can
be unifocal or multifocal and with or without necrosis,
depending on the immune status of the host.
Could This Tumor Be a Diffuse Astrocytoma/Glioblastoma? Typical features of gliomas include infiltrative borders, less cell-to-cell cohesion than metastases, microcyst formation, calcification, and arcuate vasculature in
←
Figure 2. A, Chronic gliosis near an old bacterial brain abscess shows scattered reactive astrocytes that are relatively equidistant from each other
and possess prominent eosinophilic cytoplasm. Note the scant perivascular mononuclear cell infiltrates in the upper right-hand corner of the
photomicrograph (hematoxylin-eosin, original magnification ⫻400). B, The old bacterial abscess cavity wall shows a dense strip of blue collagen
(upper left), as well as intense chronic gliosis in the adjacent brain tissue (lower right of photomicrograph). Linear fibrosis is a ‘‘stop sign’’ that
usually negates a diagnosis of low-grade astrocytoma (Masson trichrome stain, original magnification ⫻200). C, The old bacterial abscess cavity
wall is surrounded by a hypercellular carpet of reactive astrocytes (lower right of photomicrograph) and is best illustrated by glial fibrillary acidic
protein (GFAP) immunostaining (GFAP immunostain with light hematoxylin counterstain, original magnification ⫻200). D, Milder chronic gliosis
is solicited when the instigating process is less focal or less severe. This example comes from an area adjacent to a primary central nervous system
lymphoma; note the relative preservation of the native background oligodendrocytes and fewer numbers of reactive astrocytes with plump eosinophilic cytoplasm (hematoxylin-eosin, original magnification ⫻400; compare with Figure 1, A). E, Patchy myelin pallor can be seen in cases of
primary central nervous system lymphomas or gliomas that infiltrate the white matter with relatively small numbers of tumor cells. Note the illdefined loss of Luxol fast blue staining; hypercellular lymphoma is present in perivascular locations (Luxol fast blue–periodic acid-Schiff, original
magnification ⫻200). F, The same area as Figure 1, D, with immunostaining for GFAP, shows the small number of stellate reactive astrocytes in
areas of mild chronic gliosis (GFAP immunostain with light hematoxylin counterstain, original magnification ⫻200). G, A high-power photomicrograph of the primary central nervous system lymphoma shows large numbers of tumor cells in perivascular locations; note strong immunoreactivity for the B-cell marker, CD20, in the inset (hematoxylin-eosin, original magnification ⫻600; immunostaining for CD20 with light hematoxylin
counterstain, original magnification ⫻600 [inset]). H, A focus of subacute infarction with hyperplastic reactive blood vessels and small numbers
of erythrocytes is present (upper left), with mild hypercellularity at the perimeter of the infarcted tissue (hematoxylin-eosin, ⫻100). I, The same
subacute infarction (Figure 1, H) immunostained for GFAP shows stellate, relatively equidistant spaced reactive astrocytes at the perimeter; note
astrocytes sending long tapering processes to nearby small blood vessels (GFAP immunostain with light hematoxylin counterstain, original magnification ⫻200).
Arch Pathol Lab Med—Vol 130, November 2006
Algorithmic Approach to Brain Biopsy: Part I—Kleinschmidt-DeMasters & Prayson
1635
1636 Arch Pathol Lab Med—Vol 130, November 2006
Algorithmic Approach to Brain Biopsy: Part I—Kleinschmidt-DeMasters & Prayson
certain lower-grade variants and microvascular proliferation and/or necrosis (with and without pseudopalisading)
in high-grade variants. Ultimately, none of these features
alone is absolute, and the diagnosis as usual rests with
cytologic features, coupled with supplemental immunohistochemical and electron microscopic verification.
Once the site of the biopsy is correctly identified (from
a combination of clinical and neuroimaging information,
as well as histologic features on the slide), location emerges
as one of the most important determinants for formulating
a differential diagnostic list. Tumors primary to the CNS/
PNS follow the ‘‘real estate principle’’—what is important
is ‘‘location, location, location.’’ For example, a spinal cord
biopsy specimen from an intramedullary mass that you
identify as a nonmetastatic tumor is most likely ependymoma or astrocytoma. In contrast, a nonmetastatic extramedullary, intradural spinal cord mass is likely to be a
meningioma or schwannoma.
The real estate principle holds so strongly that you
should make every effort to know the differential diagnoses for each site and exclude the common things (including unusual variants of common things) before diagnosing tumors low on the differential list for that site.
Although GFAP is one of the more reliable, more specific immunostains, even it should usually not be used in
isolation on difficult cases of high-grade tumors and without a ‘‘panel approach.’’ Stop! Some pancytokeratin cocktails cross react with GFAP, but fortunately CAM 5.2 does
not. Very high-grade gliomas may show very little GFAP
immunostaining but usually will have S100 protein or vimentin immunostaining, as these are expressed early in
neurodevelopment in the more undifferentiated, primitive
neuroglial cells recapitulated by some high-grade gliomas.
Stop! High-grade neoplasms often show immunohistochemical ‘‘infidelity,’’ and gliomas are no exception; class
III beta tubulin staining, formerly thought to be a good
marker of neuronal differentiation, has been found in the
neoplastic glial cells of high-grade gliomas.
Could This Be a Type of Primary CNS/PNS Neoplasm
Other Than Diffuse Astrocytoma/Glioblastoma—How
Can I Watch for This? Although all of neuropathologic
tumor classification cannot be condensed down to an algorithmic format, here are Stop! signs that suggest an easy
way to be on the lookout for unusual variants of primary
CNS neoplasm, as opposed to the more common, diffuse
fibrillary astrocytomas (World Health Organization grades
2, 3 [anaplastic astrocytoma], and 4 [glioblastoma]) that
usually affect cerebral hemispheric white matter and pons
in children and adults.
1. Nonpontine lesion in a person less than 18 years of
age—consider pilocytic astrocytoma, medulloblastoma,
ependymoma, ganglion cell tumor, craniopharyngioma.
2. Lesion in a child less than 2 years of age—consider
atypical teratoid/rhabdoid tumor, large cell medulloblastoma, choroid plexus carcinoma, desmoplastic infantile ganglioglioma/astrocytoma.
3. Lesion in a person with long-standing seizures—consider ganglion cell tumor, dysembroplastic neuroepithelial tumor, pilocytic astrocytoma.
4. Lesion within certain anatomic sites:
a. Ventricles—ependymoma, choroid plexus papilloma
or carcinoma, central neurocytoma, subependymoma, subependymal giant cell astrocytoma
b. Mesial temporal lobe—as above for seizures
c. Superficially located in cerebral hemisphere—pleomorphic xanthoastrocytoma, ganglion cell tumor,
desmoplastic infantile ganglioglioma/astrocytoma,
oligodendroglioma, as well as occasional meningiomas invading superficial cortex
d. Cerebellum—medulloblastoma, ependymoma, pilocytic astrocytoma, hemangioblastoma, atypical teratoid/rhabdoid tumor
e. Tectal plate/pineal gland—pilocytic astrocytoma;
pineocytoma, pineoblastoma, germ cell tumor, pineal cyst
f. Filum terminale—myxopapillary ependymoma,
paraganglioma, schwannoma (sometimes nerve root
and filum cannot be reliably distinguished as origin)
g. Spinal cord, intramedullary—ependymoma, pilocytic astrocytoma, ganglion cell tumor in addition to
diffuse astrocytoma
h. Sellar/suprasellar sites—pituitary adenoma, meningioma, craniopharyngioma, Rathke cleft cyst, germ
cell tumor
5. Lesions with large cysts and enhancing mural nodules
on neuroimaging—pilocytic astrocytoma, hemangioblastoma, ganglion cell tumors, pleomorphic xanthoastrocytoma.
6. Calcified lesion—macroscopic or microscopic—oligodendroglioma, ganglion cell tumor, pilocytic astrocy-
←
Figure 3. A, Reactive astrocytes possess long tapering ‘‘astral’’ processes. Note that these starlike cells occasionally can touch one another (GFAP
immunostain with light hematoxylin counterstain, original magnification ⫻600). B, A higher-power photomicrograph of the same area of subacute
infarction illustrated in Figure 1, H demonstrates the hyperplastic vasculature and hypercellularity within the necrotic nidus of the infarct (hematoxylin-eosin, original magnification ⫻200). C, In the case of a subacute infarction, the hypercellularity within the necrotic tissue is due to
macrophage influx, as highlighted by CD68 immunostaining for macrophages (CD68 immunostain with light hematoxylin counterstain, original
magnification ⫻600). D, In contrast to the modest hypercellularity seen in the chronic gliosis adjacent to an old bacterial abscess cavity (illustrated
in Figure 1, A), a gemistocytic astrocytoma shows greater cell density, even at low microscopic power. Nevertheless, prominent perivascular
nonneoplastic lymphocytic cuffing is often present in gemistocytic astrocytomas and may prompt confusion with an infectious process (hematoxylineosin, ⫻200, compare with Figure 1, A). E and F, GFAP immunostaining of high-grade astrocytomas illustrates that tumor cells have more blunted
and irregular cell processes, or even rounded contours, compared to the starlike reactive astrocytes seen in Figure 2, A (GFAP immunostain with
light hematoxylin counterstain, original magnifications ⫻600). G, Radiation necrosis seldom yields the intense macrophage response seen with a
subacute infarction and is associated with vessels showing characteristic fibrinoid vascular necrosis (hematoxylin-eosin, original magnification
⫻200). H, Radiation necrosis usually shows only scant perivascular mononuclear cell reaction, and a fibrous capsule does not develop, as it does
surrounding bacterial abscesses (hematoxylin-eosin, original magnification ⫻200). I, Higher-power photomicrograph of radiation necrosis demonstrates that occasional reactive astrocytes can show mild nuclear atypia; note, however, the low cell density and the paucity of cells with these
changes. This patient was radiated for a high cervical lymph node metastasis from melanoma; a small portion of cerebellum was within the
radiation portals and developed radionecrosis. Hence, we can be certain that the atypical astrocyte in this photomicrograph is not an isolated
residual astrocytoma cell (hematoxylin-eosin, original magnification ⫻600).
Arch Pathol Lab Med—Vol 130, November 2006
Algorithmic Approach to Brain Biopsy: Part I—Kleinschmidt-DeMasters & Prayson
1637
toma, pleomorphic xanthoastrocytoma, spinal cord meningiomas.
7. Lesion with copious vasculature—arcuate, staghorn,
capillary-sized—oligodendroglioma, central neurocytoma, ganglion cell tumor, dysembryoplastic neuroepithelial tumor, pilocytic astrocytoma, schwannoma,
hemangioblastoma, hemangiopericytoma, angiomatous
meningioma.
8. Lesion with highly pleomorphic cells—pleomorphic
xanthoastrocytoma, ganglion cell tumor, in addition to
high-grade diffuse astrocytoma.
9. Lesion with Rosenthal fibers or eosinophilic granular
bodies—ganglion cell tumor, pleomorphic xanthoastrocytoma, pilocytic astrocytoma, subependymoma.
We believe it is important to alert the reader to special
brain tumor types other than the diffuse astrocytoma
group (ie, astrocytoma grade 2, anaplastic astrocytoma
grade 3, glioblastoma multiforme grade 4) at the end of
the first algorithm. Some of these special types of ‘‘other’’
brain tumors can at least be suspected based on their tendency to occur at young ages or in certain anatomic locations in the brain or to be associated with cysts, calcification, mural nodules, or specific histologic features (Rosenthal fibers, eosinophilic granular bodies, etc). Several
of these ‘‘other’’ tumors possess features that can lead to
1638 Arch Pathol Lab Med—Vol 130, November 2006
erroneous interpretation of these as high-grade neoplasms.
Examples include the high cellularity of even grade 2 oligodendrogliomas and the extreme pleomorphism of tumor cells in pleomorphic xanthoastrocytomas and some
ganglion cell tumors. Some of these ‘‘other’’ tumors in
neuropathology possess features that have different prognostic implications in their given tumor type, such as the
vascular proliferation that is innocuous when seen in pilocytic astrocytomas but is a high-grade feature when seen
in diffuse astrocytomas.
The authors thank Ms Susan Peth for expert manuscript preparation and Ms Lisa Litzenberger for photographic expertise.
References
1. Burger PC, Scheithauer BW, Vogel FS, eds. Surgical Pathology of the Nervous System and Its Coverings. 4th ed. New York, NY: Churchill Livingstone;
2002.
2. Fuller GN, Goodman JC, eds. Practical Review of Neuropathology. Philadelphia, Pa: Lippincott Williams & Wilkins; 2001.
3. Kleihues P, Cavanee WK, eds. Pathology and Genetics of Tumours of the
Nervous System. 2nd ed. Lyon, France: IARC Press; 2000. World Health Organization Classification of Tumours.
4. Okazaki H, Scheithauer BW, eds. Atlas of Neuropathology. New York, NY:
Gower Medical Publishers; 1988.
5. Oppenheimer D, Esiri MM, eds. Diagnostic Neuropathology. Oxford, England: Blackwell Scientific Publications; 1989.
6. Prayson RA, ed. Neuropathology Review. Totowa, NJ: Humana Press; 2001.
7. Prayson RA, Cohen ML, eds. Practical Differential Diagnosis in Surgical
Neuropathology. 1st ed. Totowa, NJ: Humana Press; 2000.
8. Prayson RA, ed. Neuropathology: A Volume in the Foundations in Diagnostic Pathology Series. Philadelphia, Pa: Churchill Livingstone; 2005.
Algorithmic Approach to Brain Biopsy: Part I—Kleinschmidt-DeMasters & Prayson