Mol Diagn Ther 2009; 13 (6): 367-373

REVIEW ARTICLE 1177-1062/09/0006-0367/$49.95/0

2009 Adis Data Information BV. All rights reserved.

Molecular Classification of Cancers of

Unknown Primary Site
F. Anthony Greco1 and Mark G. Erlander2
1 Sarah Cannon Cancer Center, Nashville, Tennessee, USA
2 bioTheranostics, Inc., San Diego, California, USA

Contents

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
1. Molecular Profiling in Human Cancers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368
2. Molecular Profiling in CUP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 370
3. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372

Abstract Cancer of unknown primary site is a common metastatic cancer, diagnosed in about 50 000 patients per
year in the US. The diagnosis, classification, and management of patients with carcinoma of unknown
primary site has been difficult and frustrating. The therapy has usually been empiric for the majority of
patients, and their prognosis has been poor.
Molecular classification of metastatic cancers with known primary sites has been accurate (7689%), as
reported from several studies. Molecular profiling of initial biopsy specimens has tremendous potential as a
test to diagnose the site of tumor origin in patients with unknown primary cancer. Several retrospective
studies of molecular profiling assays have provided indirect validation of the accuracy of primary site
prediction, based on correlations with clinicopathologic features. One additional study of initial diagnostic
biopsies in unknown primary cancer patients, where primary tumor sites were identified months to years
later, has provided more direct validation of the accuracy of molecular classification (the primary tumor sites
of 15 of 20 patients were correctly predicted). The ability to diagnose and classify unknown primary cancer
more precisely would allow for more site-specific or targeted therapy, and likely improve patient outcomes.
Several clinical studies are in progress or planned to test this concept.

Cancer of unknown primary site (CUP) is a common me-
tastatic cancer, accounting for about 5% (50 000 patients) of all
advanced cancer each year in the US.[1] These patients are
unique in that they have no clinically apparent primary cancer
to explain the source of their metastasis. CUP is not a single
neoplasm with a defined natural history, but a very hetero-
geneous group with variable clinical features and histopatho-
logies. Diagnosis, classification, and management of these
patients has been difficult and frustrating.
The definition of CUP by clinicians and investigators has not
been the same over the years. All cell types are included in the
definition but, with rare exceptions, the lineage (carcinoma,
lymphoma, sarcoma, or melanoma) can be defined. The large
majority of CUP patients have carcinoma. Although the anato-
mical primary tumor site is not found at the time of diagnosis,
the clinicopathologic evaluation of the patients has frequently
varied and, therefore, studies in CUP have often been difficult
to interpret and compare. This difficulty is understandable
since the evaluation of these patients has changed over the years
as diagnostic technology has improved. A standard CUP de-
finition has been published[1] but remains dynamic and con-
tinues to evolve. Unless an anatomical primary tumor site is
identified at the time of diagnosis, the patients by our definition
have CUP, regardless of the immunohistochemistry (IHC)
368
stains or molecular profile predictions. Until these important
diagnostic tests are clinically validated in CUP with compari-
sons between outcome data and known primary cancers, the
primary site in CUP remains only a prediction rather than
an absolute identification of the origin of the cancers in these
patients.
Over the past 30 years, some patients have been classified
into subsets according to clinical features and classic histo-
pathology. A minority of patients (20%) are within these several
subsets that are more treatable and have a better prognosis than
the group as a whole with CUP.[2] The majority of CUP patients
have carcinoma and most of these are adenocarcinomas. Al-
though the pathogenesis and biology of CUP is poorly under-
stood, most patients harbor a clinically undetectable or occult
primary site, which gives rise to the detectable metastasis.
Several autopsy series have documented occult primary sites
in about 75% of all patients examined, most commonly in the
lung, pancreas, kidney/adrenal gland, liver/bile duct, and colon/
rectum.[3] Since the majority of CUP patients have a small,
clinically undetectable primary site, they represent a large
number of different cancers. Diagnosis of the primary tumor
site at the time of clinical presentation has not been possible for
most patients. IHC has been helpful in suggesting or predicting
the possible primary sites in some patients by the demonstra-
tion of various markers in the tumor cells.[4] However, there are
only a few IHC markers specific enough to confidently identify
the primary site. In most patients with CUP, the primary site
remains speculative, even after extensive IHC study of the
tumor with multiple markers. Diagnosis of the primary site
is a critical issue, since therapy is more likely to be effective if
site-specific therapy is administered, rather than empirical
single regimens for all patients. The therapy for many advanced
cancer patients has improved, and the types of therapies are
different for many of these specific cancers. Therefore, accurate
classification and diagnosis of the type of cancer present in
CUP patients is important.
Pathologic evaluation specifically, standard light micro-
scopy and IHC stains has been one of the essential steps in
predicting the possible primary site of origin in CUP.[4] Al-
though there is no question that IHC has revolutionized cancer
pathology, studies reporting qualitative metrics of the success
of IHC to identify the primary site in CUP are rather scant. A
major reason for this is the difficulty of documenting the pri-
mary tumor site. A commonly cited study regarding the ability
of IHC to diagnose the primary site is by DeYoung and Wick,[5]
in which there was a reported 67% success rate in identifying
the primary site by using IHC. However, data were not dis-
closed, and also the study was reported 9 years ago. Since then,
2009 Adis Data Information BV. All rights reserved.
Greco & Erlander
the repertoire of cell-type IHC markers has substantially in-
creased. The issue of validation of IHC in reference to the
primary site in patients with CUP remains an important con-
sideration. Most of the IHC markers have been validated for
patients with known primary cancer. Frequently, it is assumed
that if an IHC marker is present in CUP, this indicates the
specific primary cancer derived from validation studies of the
marker in known cancer types. In our experience, a definitive
primary site diagnosis based on IHC findings is not common in
patients with CUP, and in one small autopsy series, histo-
chemical or IHC stains were helpful in revealing the diagnosis
of the primary tumor site (verified at autopsy) in only four of
27 patients.[6]
The amount of tissue available for IHC workup is often
limited, given the increased use of core needle biopsies and fine
needle aspirates to obtain diagnostic material. Therefore,
sometimes a complete IHC workup is difficult and can lead the
IHC analysis down an incorrect line of investigation. Each IHC
marker has a sensitivity and specificity for a given cancer type,
and these statistical parameters can change, depending on the
number and type/subtype of different tumors examined. Fi-
nally, IHC is a subjective technique requiring an experienced
pathologist for interpretation. Therefore, there is a need for
other classification methods. In particular, gene expression
profiling may compliment IHC methods because it is quanti-
tative and can classify tumors that IHC has difficulty with, and
it uses small amounts of tissue (3001000 cells) to generate a
prediction for cancer classification.
Molecular profiling is a relatively new technology with tre-
mendous potential in CUP classification and diagnosis.[7] There
is no other large group of cancer patients with an obscure or
undefined site of origin. In the setting of CUP, the appropriate
use of molecular profiling may provide critical information
relevant to the specific type of cancer present, particularly when
considered with the clinical and pathologic features. Accurate
classification and diagnosis will likely improve the outcome of
many CUP patients, since more appropriate, tailored therapy
would be administered to each patient.
1. Molecular Profiling in Human Cancers
The invention of microarrays about 15 years ago enabled the
simultaneous quantification of whole-genome gene expression
and has led to unprecedented advancements in genomic-related
computational technologies, a continued unraveling of cancer-
activated pathways, and discovery of diagnostic, prognostic,
and drug response signatures in a multitude of cancer types.[8,9]
Mol Diagn Ther 2009; 13 (6)
Molecular Classification of CUP 369

With regard to cancer classification/diagnosis, Golub et al.[10]
demonstrated that gene expression profiling could be used to
diagnostically discriminate between acute myeloid leukemia
and acute lymphoblastic leukemia. This landmark study de-
monstrated for the first time that cancer types could be classi-
fied by measuring the differential expression of specific gene
sets. Subsequently, Ramaswamy et al.[11] and Su et al.[12]
demonstrated that numerous cancer types could be classified by
gene expression profiling.
Since these studies, there have been numerous additional
reports demonstrating the ability to use gene expression to
classify human tumors.[13-21] This approach has been successful
because the 400-odd different cell types in the body have dif-
ferent gene expression profiles, which reflect their distinct
cellular functions (e.g. hepatocytes versus breast luminal epi-
thelium), and cancers developing from these different cell types
retain some of their unique cell-type differences within their
gene expression pattern. Furthermore, it has been shown
through several validation studies that metastatic cancers retain
expression patterns found in their respective primary cancers
and that poorly differentiated and undifferentiated tumors of a
given cancer type often retain expression patterns observed in
their respective well-differentiated tumors.[21]
A summary of selected gene expression studies in human
cancers of known primary tumor sites is illustrated in table I.
From a technical point of view, a clinically viable test needs to
be compatible with formalin-fixed, paraffin-embedded tissues,
given that nearly all tumor samples are preserved in this fash-
ion. Second, the test should include the ability initially to
complete laser microdissection. CUP tumors by definition are
predominantly metastatic, growing in other organs/tissues;
therefore, it is of paramount importance that the sample pro-
cess is highly enriched for the tumor cells. This point cannot
be overemphasized, given that these gene expression tests are
not measuring tumor specific markers but, rather, expression
dynamics in relation to cell lineage, as described above. Third,
the test needs to be compatible with low numbers of tumor cells
(300500) since, in practical terms, many samples are core
needle biopsies or fine needle aspirates. Finally, the test needs to
be able to discriminate between a large number of tumor types,
given that CUP can arise from many primary sites. The number
of tumor types recognized by these various assays varies from 6
to 39 (table I).
In the majority of clinical settings in patients with a known
primary cancer site, the diagnosis is secured by the clinical
findings and the use of traditional pathologic methods. Gene
expression profiles for diagnosis are not necessary. In CUP, the
potential role for gene expression profiling in classifying the
primary tumor site is apparent. In the past decade, several
molecular-based clinical assays for tumor classification in CUP
have been developed, based on the fundamental tenet that ef-
fective treatment for cancers relies on knowledge of the primary
anatomical origin/site of the tumor. In parallel with these ef-
forts has been the emergence of biologic-targeted therapies that
inhibit specific signal transduction pathways active in parti-
cular cancer types (e.g. human epidermal growth factor re-
ceptor-2 [HER2] and estrogen receptor expression in breast
cancer, epidermal growth factor receptor [EGFR] mutation in

Table I. Summary of selected gene expression studies in human cancers of known primary sites
Original study No. of samples Gene expression Tissue No. of cancers Accuracy Commercial entity
platform compatibility classified (%)
Tothill et al.[14] Training set: 229 tumors; Microarray Frozen 14 89 None
validation set the same but using (10 500 cDNAs)
leave-one-out cross-validation
Talantov et al.[15] Training set: 260 tumors; Real-time RT-PCR FFPE 6 plus other 76 www.veridex.com
validation set: 48 tumors (10 genes) cancers
Rosenfeld et al.[16] Training set: 253 tumors; Real-time RT-PCR FFPE 22 86 www.rosettagenomics.com
validation set: 83 tumors (48 micro-RNAs)
Monzon et al.[17] Validation set: 477 tumors Microarrays Frozen 15 89 www.pathworkdx.com
(1550 genes)
Pillai et al.[22] Validation set: 352 tumors Microarrays FFPE 15 89 www.pathworkdx.com
(1550 genes)
Ma et al.[21] Training set: 481 tumors; Real-time RT-PCR FFPE 39 87 in www.biotheranostics.com
validation set: 119 tumors (92 genes) 32 tumor
types
cDNA = complementary DNA; FFPE = formalin-fixed, paraffin-embedded; RT = reverse transcriptase.

2009 Adis Data Information BV. All rights reserved. Mol Diagn Ther 2009; 13 (6)
370
non-small-cell lung cancer, and KRAS mutation in colorectal
cancer). This has led to the prediction that oncologists in the
future will make treatment decisions largely based on the mole-
cular makeup (mutational status, pathway activation, etc.) of the
patients cancer, irrespective of the primary site of origin. Cur-
rently, however, knowledge of the type and primary site of cancer
is essential in determining the therapy and prognosis of patients.
The initial validation of gene expression profiling assays was
done in known human primary tumors. The various assays are
all quite accurate (7689%) at independently determining the
tumor type by these methods (table I). Classification or diag-
nosis of the particular tumor type in CUP biopsy specimens has
been more difficult to validate, as the primary site or specific
type of cancer is not known and usually never becomes known
in these patients. Based on the ability of molecular profiling to
accurately identify both primary tumors and metastatic sites in
patients with known primary cancer, it would seem logical to
assume a similar accuracy rate for these assays in metastatic
sites in CUP. However, extension of these findings to CUP
assumes that knowing the site of origin translates into improved
patient outcomes that are analogous to improvement observed
from treatments based on known tumors. This assumption
argues that (a) CUP is not an entirely specific biology per se[23]
but, rather, a collection of unspecified sets of cancer types
which, given the correct technology, can be sorted out; and (b)
once the cancer type is identified, the patient will be treated and
will benefit similarly to a patient with a known primary tumor.
There is merit in this argument, since it has been previously
recognized that certain subsets of CUP have a favorable prog-
nosis[2] (e.g. isolated axillary node adenocarcinoma in women
usually represents occult breast cancer). More recently, CUP in
association with a molecular profile or IHC prediction of colo-
rectal cancer appears to have a clinical course and response to
treatment consistent with metastatic colorectal cancer.[19,24]
On the other hand, CUP is in several ways different from
known advanced cancers. The absence of a clinically detectable
primary tumor (even though most have occult small primary
tumors proven by autopsy studies) is obvious. In addition, in
CUP patients proven to have a specific primary tumor at autop-
sy, the pattern of metastasis is frequently atypical or different
from known primary cancer (e.g. bone metastasis is more
frequent than liver metastasis in occult pancreatic carcinoma;
lymph node and lung metastases are more frequent than bone
metastasis in occult prostate carcinoma). CUP may have a
unique genetic profile, although no data yet support this pos-
sibility. Patients with CUP and clinically undetectable pri-
maries (occult primaries) may also have different molecular
profiles than patients with the same known overt primaries.
2009 Adis Data Information BV. All rights reserved.
Greco & Erlander
Therefore, validation of the usefulness and accuracy of mole-
cular profiling in CUP biopsy samples is important. Once this
validation is acceptable in CUP, then the clinical use of a mole-
cular profiling assay would be critical to help determine the type
of cancer and therapy for each patient. Molecular classification
would become an essential element of the evaluation and classi-
fication of CUP patients.
2. Molecular Profiling in Cancer of
Unknown Primary Site
Few would disagree that accurate determination of the pri-
mary tumor site in patients with CUP would be helpful. This
would provide a specific diagnosis and help formulate a prog-
nosis for the patient, as well as specific site-directed treatment.
The group of patients with CUP represents an ideal setting to
employ an accurate and reproducible molecular classification
system for diagnosis. There have been several retrospective
studies evaluating the possible usefulness of molecular profiling
in the diagnosis or classification of the primary tumor site in
CUP patients.[14,15,18-20,25,26] All of these data provide some
validation of the usefulness and accuracy of molecular profiling
in predicting the primary tumor site. However, the validation
(except for one study thus far[26]) has been indirect and based on
correlation with clinical features, response to therapies, and
pathology (including IHC stains). In the retrospective studies,
the inclusion criteria were variable from study to study, in-
cluding IHC-predicted or resolved cases,[15,18] limited IHC
with mainly clinical correlations[19,20,25] and actual identifica-
tion of latent primary tumor sites.[26] The various studies did
not have the same inclusion criteria, leading to some difficulty
in interpretation when comparing studies. Each of these studies
will now be reviewed briefly (table II).
Tothill et al.[14] tested a complementary DNA microarray on
the initial diagnosis biopsy specimens from 13 patients with
CUP. The predictions of the primary tumor site of origin were
compared with clinicopathologic features, and the most likely
primary site was decided by a medical oncologist. In 11 of 13
patients (85%), their molecular classification prediction was
consistent with the most likely primary site as determined by the
clinical and pathologic data. The actual anatomical primary
tumor site was not verified in any of these patients.
Talantov et al.[15] included 33 patients with CUP in their
total tumor samples from 449 patients in their validation study
of their reverse transcriptase (RT)-PCR assay for known pri-
mary cancers. Their assay was capable of identifying six tumor
types, and the assay results were consistent in 17 of 22 patients,
with the prediction of the primary site made by IHC.
Mol Diagn Ther 2009; 13 (6)
Molecular Classification of CUP 371

Table II. Summary of gene expression profile validation studies in cancers of unknown primary site (CUP)
Study Assay Validation of assay Validation of assay
by indirect correlationa by direct correlationb
Tothill et al.[14] Microarray[14] 13 patients, Not done
11 predicted (85%)
Talantov et al.[15] RT-PCR[15] 22 patients, Not done
17 primaries predicted (77%)
Varadhachary et al.[19] RT-PCR[15] 120 patients, Not done
63 primaries predicted (61%)
Monzon et al.[25] Microarray[17] 21 patients, Not done
16 primaries predicted (76%)
Bridgewater et al.[20] Microarray[27] 21 patients, Not done
18 primaries predicted (86%)
Horlings et al.[18] Microarray[27] 38 patients, Not done
29 primaries predicted (76%)
Greco et al.[26] RT-PCR[21] Not applicable, 20 patients,
latent primary site known 15 predicted accurately (75%)
a Clinical and/or pathologic (immunohistochemistry) features only; no primary tumor site documented.
b Primary tumor site of origin later definitely identified.
RT = reverse transcriptase.

Varadhachary et al.[19] used this same RT-PCR assay on
biopsies from 120 patients with CUP. In 63 of 120 patients
(61%), a primary site was predicted, and the clinicopathologic
features and response to treatment were compatible with the
predicted primary site in the majority of patients. A colorectal
cancer profile was suggested as an important group of CUP
patients to recognize, since their responses to treatment utili-
zing colorectal-specific chemotherapy were similar to those of
known colon cancer patients.
Bridgewater et al.[20] used a microarray assay on biopsies
from 21 CUP patients. These primary site predictions were
correlated with clinical and pathologic features. The primary
site prediction was clinically feasible in 18 of 21 patients, and
in 12 patients the clinical management would have been influ-
enced by the assay results.
Horlings et al.[18] studied a microarray assay on samples
from 38 CUP patients. Twenty-two of 38 tumor samples could
not be classified by IHC, and in 14 a tissue of origin was pre-
dicted (64%) by the microarray assay that was consistent with
the clinicopathologic features. The assay was in agreement with
15 of the 16 patients primaries (94%) predicted by IHC. The
anatomical primary tumor site was not confirmed in any patient.
Monzon et al.[25] evaluated a microassay-based expression
profile assay on fresh-frozen biopsy specimens from 21 patients
with CUP. The primary site was predicted in 16 of 21 patients
(76%) and was indeterminate in five (24%). The results in six of
seven gastrointestinal predictions (five colorectal and one
pancreatic) were consistent with the clinical differential in
these patients. Correlations with the assay results in the other
nine patients were not reported in the abstract.
The validation of the accuracy of molecular profile assays in
predicting the primary site in CUP patients in these five retro-
spective studies (one in abstract form only) was indirect, with
correlations based on clinical and pathologic features. The
question remains how accurate the assays are in predicting or
diagnosing the actual primary tumor site. Greco et al.[26] re-
cently reported, in abstract form, a direct retrospective vali-
dation study in 20 CUP patients who later had their primary
sites discovered (median 10 months later, range 254 months
later following initial diagnosis of CUP). The inclusion criteria
required a primary anatomical tumor site identification at least
8 weeks or longer following the initial diagnosis of CUP. Their
initial diagnostic biopsies were evaluated by an RT-PCR assay
capable of identifying 32 tumor types. In 15 of 20 specimens
(75%) from these patients, the primary tumor was accurately
predicted, providing direct validation of this assay. This is very
encouraging, but additional direct validation studies would be
helpful in providing confirmation and reassurance of the use-
fulness of molecular profiling assays in classifying the primary
tumor site in CUP patients. The value of these assays will likely
be even more clinically helpful in primary site prediction when
used in concert with clinical and IHC data for each patient.
In summary, the primary site prediction for CUP has been
examined by several of the molecular tests. Given the occult,

2009 Adis Data Information BV. All rights reserved. Mol Diagn Ther 2009; 13 (6)
372
clinically undetectable nature of the primary sites in most CUP
patients, it is extremely difficult to verify the accuracy of mole-
cular assays or IHC stains in predicting the true primary site.
The studies of CUP by Tothill et al.,[14] Talantov et al.,[15]
Monzon et al.,[25] Varadachary et al.,[19] Bridgewater et al.,[20]
and Horlings et al.[18] were all based on indirect comparisons
with clinical and IHC correlations. It is important to note that
these were simply correlations or indirect validation and may
not have reflected the true accuracy of the molecular assay or,
for that matter, IHC. The actual primary tumor site was not
identified in any of these patients. To obtain a more direct
validation of the accuracy of molecular profiling to predict the
primary tumor site in CUP, patients with latent primary tumor
sites were evaluated to test the RT-PCR assay, as previously
reported by Ma et al.[21] As mentioned, this assay accurately
predicted the primary tumor site or origin in 15 of 20 CUP
patients from their initial diagnostic biopsies. This was direct
validation in patients with primary tumor sites that were later
found, and convincing evidence that gene expression profiling
can add value in the clinical evaluation of CUP. All of these
studies provide support for this concept.
3. Discussion
The molecular classification of CUP continues to evolve.
These data specifically evaluating the accuracy of primary site
prediction are difficult to generate, since most patients never
have verifiable primary sites found. Most of the molecular
profile validation studies available in CUP are indirect evidence
with clinicopathologic correlations (table II). The only direct
validation (latent primaries were known) was reported by
Greco et al.[26] and was accurate in 75% of the patients. Iro-
nically, the majority of patients with CUP do have small,
clinically undetected or occult primary sites, as documented in
the past from autopsy series. However, most of these patients
never have a clinically detectable primary site found during
their lifetime. In those rare patients whose latent primary sites
have later been discovered, the accuracy of a molecular pro-
filing assay in predicting the primary was very reasonable.
Additional data from this group may be forthcoming. Corre-
lation of molecular assay primary site prediction from initial
biopsies with the actual primary site found at postmortem ex-
amination would also be of great interest.
The accuracy of molecular profiling in classifying tumors
with known primary sites is now established, and data in CUP,
although much less robust and often indirect, are encouraging.
Clinical trials in CUP patients are now critical to determine if
molecular profile classification or diagnosis of the specific tu-
2009 Adis Data Information BV. All rights reserved.
Greco & Erlander
mor type will improve patient outcomes after site-specific or
tailored therapy. Preliminary data do suggest that CUP with
colorectal molecular profile signatures responds similarly to
known advanced colorectal cancer treated with colorectal site-
specific therapy.[19] However, the number of patients is too
small to feel confident that outcomes in these colorectal profile
CUP patients will routinely parallel those in patients with
known colorectal cancer. The ability to classify and diagnose
patients more precisely allows appropriate clinical trials to be
designed to answer the ultimate question of whether CUP pa-
tient outcomes will be improved. Several studies are now in
progress and others are planned.
The optimal CUP study design would require molecular
profiling of the biopsy specimen and random allocation of
patients to either a standard empiric systemic regimen or to site-
specific or tailored systemic therapy, as determined by the
assay. Other trials prospectively treating patients according to
the molecular profile classification will also be of value. If
molecular classification is proven to impart the same prognosis
for CUP patients as for those with corresponding known pri-
mary cancer, then an advance will have been achieved for CUP
patients.
The ideal management of CUP patients in the immediate
future will likely be based upon accurate diagnosis or classifi-
cation of the primary tumor site. Perhaps more importantly, the
stage is set for a more precise molecular understanding of each
patients tumor, which will likely result in the recognition of a
specific individualized therapeutic approach. This evolving role
of gene expression profiling of human cancers may be more
important in determining therapy for patients in the future than
knowledge of their primary tumor site of origin.
Acknowledgments
Dr Greco has received consultancy fees and speakers honoraria from
bioTheranostics, Inc. Dr Erlander is an employee of bioTheranostics, Inc.
and holds stock options in the company.
References
1. Greco FA, Hainsworth JD. Cancer of unknown primary site. In: DeVita Jr VT,
Hellman S, Rosenberg M, editors. Cancer: principles and practice of oncology.
8th ed. Philadelphia (PA): Lippincott, 2008: 2363-87
2. Hainsworth JD, Fizazi K. Treatment for patients with unknown primary
cancer and favorable prognostic factors. Semin Oncol 2009; 36: 44-51
3. Pentheroudakis G, Golfinopoulos V, Pavlidis N. Switching benchmarks in
cancer of unknown primary: from autopsy to microarray. Eur J Cancer 2007;
43: 2026-36
4. Oien K. Pathological evaluation of unknown primary cancer. Semin Oncol
2009; 36: 8-37
Mol Diagn Ther 2009; 13 (6)
Molecular Classification of CUP
5. DeYoung BR, Wick MR. Immunohistologic evaluation of metastatic carci-
nomas of unknown origin: an algorithmic approach. Semin Diagn Pathol
2000; 17: 184-93
6. Al-Brahim N, Ross C, Carter B, et al. The value of postmortem examination
in cases of metastasis of unknown origin-20 year retrospective data from a
tertiary care center. Am Diagn Pathol 2005; 9: 77-80
7. Bender RA, Erlander MG. Molecular classification of unknown primary
cancer. Semin Oncol 2009; 36: 38-43
8. Li X, Quigg RJ, Zhou J, et al. Clinical utility of microassays: current status,
existing challenges and future outlook. Curr Genomics 2008; 9: 466-74
9. Sotiriou C, Piccart MJ. Taking gene-expression profiling to the clinic: when will
molecular signatures become relevant to patient care? Nature Rev Cancer
2007; 7: 545-53
10. Golub TR, Slonim DK, Tamayo P, et al. Molecular classification of cancer:
class discovery and class prediction by gene expression monitoring. Science
1999; 286: 531-7
11. Ramaswamy S, Tamayo P, Rifkin R, et al. Multiclass cancer diagnosis using
tumor gene expression signatures. Proc Natl Acad Sci U S A 2001; 98: 15149-54
12. Su AI, Welsh JB, Sapinoso LM, et al. Molecular classification of human car-
cinomas by use of gene expression signatures. Cancer Res 2001; 61: 7388-93
13. Bloom G, Yang IV, Boulware D, et al. Multi-platform, multi-site, microarray-
based human tumor classification. Amer J Path 2004; 164: 9-16
14. Tothill RW, Kowalczyk A, Rischin D, et al. An expression-based site of origin
diagnostic method designed for clinical application to cancer of unknown
origin. Cancer Res 2005; 65: 4031-40
15. Talantov D, Baden J, Jatkoe T. A quantitative reverse transcriptase-
polymerase chain reaction array to identify metastatic carcinoma tissue of
origin. J Mol Diag 2006; 8: 320-9
16. Rosenfeld N, Aharonov R, Meiril E, et al. MicroRNAs accurately identify
cancer tissue origin. Nature Biotech 2008; 26: 462-9
17. Monzon FA, Lyons-Weiler M, Buturovic LJ, et al. Multicenter validation of a
1,550-gene expression profile for identification of tumor tissue of origin. J Clin
Oncol 2009; 27: 2503-8
18. Horlings HM, van Laar R, Kerst J-M, et al. Gene expression profiling to
identify the histogenetic origin of metastatic adenocarcinomas of unknown
primary. J Clin Oncol 2008; 26: 4435-41
2009 Adis Data Information BV. All rights reserved.
373
19. Varadhachary G, Talantov D, Raber M, et al. Molecular profiling of
carcinoma of unknown primary and correlation with clinical evaluation.
J Clin Oncol 2008; 26: 4442-8
20. Bridgewater J, van Laar R, Floore A, et al. Gene expression profiling
may improve diagnosis in patients with carcinoma of unknown primary.
Br J Cancer 2008; 98: 1425-30
21. Ma X-J, Pate R, Wang X, et al. Molecular classification of human cancers using
a 92-gene real-time quantitative polymerase chain reaction array. Arch Path
Lab Med 2006; 130: 465-73
22. Pillai R, Deeter R, Rigl CT, et al. Validation of a microarray-based gene
expression test for tumors with uncertain origins using formalin-fixed
paraffin-embedded (FFPE) specimens [abstract no. e22015]. J Clin Oncol
2009; 27 (15S) [online]. Available from URL: http://meeting.ascopubs.org/
cgi/content/abstract/27/15S/e22015 [Accessed 2009 Sep 13]
23. Pentheroudakis G, Briasoulis E, Pavlidis N. Cancer of unknown primary site:
missing primary or missing biology? Oncologist 2007; 12: 418-25
24. Varadhachary GR, Raber MN, Matamoros A, et al. Carcinoma of unknown
primary with a colon-cancer profile-changing paradigm and emerging
definitions. Lancet Oncol 2008; 9: 596-9
25. Monzon FA, Henner WD, Medeiros F. Use of microarray-based 1550-gene
expression profile-of carcinoma of unknown primary (CUP) [abstract no.
313]. Gastrointestinal Cancers Symposium, American Society of Clinical
Oncology; 2009 Jan 15-17; San Francisco (CA)
26. Greco FA, Spigel DR, Yardley DA, et al. Unknown primary cancer (UPC):
accuracy of tissue of origin prediction by molecular profiling [abstract no.
11070]. J Clin Oncol 2009; 27 (15S) [online]. Available from URL: http://
meeting.ascopubs.org/cgi/content/abstract/27/15S/11070 [Accessed 2009
Sep 13]
27. van Laar RK, Ma XJ, de Jong D, et al. Implementation of a novel microarray-
based diagnostic test for cancer of known primary. Int J Cancer 2009; 125:
1390-7
Correspondence: F. Anthony Greco, MD, Sarah Cannon Research Institute,
250 25th Avenue North, Suite 110, Nashville, TN 37203, USA.
E-mail: fgreco@tnonc.com
Mol Diagn Ther 2009; 13 (6)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.