TUMOR MARKERS

Tumor markers are indicators of cellular, biochemical, molecular, or genetic alterations by

which neoplasia can be recognized. These surrogate measures of the biology of the cancer
provide insight into the clinical behavior of the tumor. This is particularly useful when the
cancer is not clinically detectable. The information provided may

Be diagnostic and distinguish benign from malignant disease

Correlate with the amount of tumor present (so-called tumor burden)

Allow subtype classification to more accurately stage patients

Be prognostic, either by the presence or absence of the marker or by its concentration

Guide choice of therapy and predict response to therapy

The ideal tumor marker has three defining characteristics:

1.

The marker is expressed exclusively by the particular tumor

2.

Collection of the specimen for the tumor marker assay is easy.

3.

The assay itself is reproducible, rapid, and inexpensive.

Currently, there is no one marker that fulfills all these criteria for any cancer, nor is there any
specific cancer in which there are biomarkers that completely describe its behavior.
Tumor markers fall into three broad categoriesproteins, genetic mutations, and epigenetic
changes ( Box 29-2 ). All three may be found in the tumor tissue itself. Tumor markers found
in body fluids, particularly blood and urine, have the greatest potential for clinical application
because of the ease of access to these fluids for analysis and because repeated sampling
allows in vivo monitoring of the malignancy for such features as disease progression or
recurrence, metastasis, and response to therapy.
Box 29-2
Potential Nonprotein Tumor Markers
RNA-Based Markers
Overexpressed/underexpressed transcripts
Regulatory RNA (e.g., micro-RNA)
DNA-Based Markers
Single-nucleotide polymorphisms (SNPs)
Chromosomal translocationsbcr-abl (Philadelphia)
Changes in DNA copy number
Microsatellite instability
Epigenetic changes (e.g., differential promoter region methylation)
From Ludwig JA, Weinstein JN: Biomarkers in cancer staging, prognosis and treatment
selection. Nat Rev Cancer 5:845-856, 2005.

Rather than provide an exhaustive review of all tumor markers, this section outlines the major
categories of tumor markers and focuses on evidence for the tumor markers currently in
clinical use.
Protein Tumor Markers

Proteins were the first type of tumor marker identified and hence are considered the so-called
classic tumor markers. However, despite decades of research, few are in clinical use. Those
routinely used are in general limited by poor sensitivity and specificity. Their concentrations
in serum or plasma generally correlate with tumor burden inasmuch as they are shed from the
expanding neoplasm.
Carcinoembryonic Antigen

Carcinoembryonic antigen (CEA) is probably the most studied cancer tumor marker and is
predominantly used clinically in patients with cancer of the colon and rectum. It is an
oncofetal protein that is normally present during fetal life but can be seen in low
concentration in healthy adults. Structurally, it is a glycoprotein with a molecular weight of
200 kd and is a component of the glycocalyx, located on the luminal side of the cell
membrane of normal epithelial intestinal cells. CEA is a member of a large family of proteins
that are related to the immunoglobulin gene superfamily. The molecule itself is secreted into
the circulation and is also found in the mucous secretions of the stomach, small intestine, and
biliary tree. Although its exact function is unknown, CEA has been shown to be involved in
cell adhesion and is able to inhibit apoptosis induced by loss of anchorage to the ECM.
Testing

Immunoassay kits allow determination of serum CEA levels accurately, reproducibly, and
relatively inexpensively. Normal serum levels are less than 2.5 ng/mL, borderline if 2.5 to 5.0
ng/mL, and elevated if greater than 5.0 ng/mL. Borderline levels occur with benign disorders
such as inflammatory bowel disease, pancreatitis, cirrhosis, and chronic obstructive
pulmonary disease, and smoking can also increase CEAthe upper limit of normal in
smokers is considered 5 ng/mL.
Screening

CEA is not useful as a screening test because of its low sensitivity in early-stage disease
elevated CEA levels occur in only 5% to 40% of patients with localized disease.
Prognosis

Elevated CEA levels reflect the burden of tumor present. The degree of CEA elevation
correlates with increasing stage of disease, and therefore CEA levels have prognostic value.
Preoperative serum CEA is an independent predictor of survivalthe higher the preoperative
serum level, the poorer the prognosis. This effect persists even after patients are stratified for
resectability and extent of local tumor invasion. Five-year survival is significantly worse in
patients with elevated preoperative CEA levels than in those with a normal preoperative CEA
level. Furthermore, 5-year survival is higher in patients whose elevated preoperative CEA
normalized postoperatively. Finally, patients with elevated preoperative CEA levels have
higher recurrence rates than do those with normal CEA levels.

Monitoring

The most common application of CEA is to monitor patients for recurrent disease. CEA is
most sensitive for hepatic or retroperitoneal metastasis and relatively insensitive for local,
pulmonary, or peritoneal involvement. About 75% of patients with recurrent colorectal cancer
have an elevated serum CEA level before the development of symptoms. However, the
pattern or magnitude of the rise in CEA levels is of no value in distinguishing localized
recurrence from distant disease. Because elevations of CEA may be transient, repeat
measurement is performed as confirmation of the trend. A confirmed rising trend in CEA
prompts evaluation for recurrent disease.
Because CEA reflects tumor burden, it is useful in monitoring response to chemotherapy in
patients with metastatic cancer. An elevated CEA level is an independent factor associated
with poor survival and progression on 5-fluorouracil chemotherapy in patients with
metastatic colorectal cancer. Patients with advanced cancer whose CEA levels fall during
chemotherapy survive significantly longer than do patients whose CEA levels do not change
or increase.
-Fetoprotein

-Fetoprotein (AFP) is used for the detection and management of HCC. It is an oncofetal
antigen that consists of a single-chain polypeptide with a molecular weight of 700 kd. Levels
are elevated in the fetus, decrease sharply after birth, and are increased during pregnancy. It is
synthesized by hepatocytes and endodermally derived gastrointestinal tissues.
Testing

AFP is measured with immunoassay kits, either enzyme-linked immunoassays or

radioimmunoassays. The upper limit of normal for a healthy, nonpregnant adult is less than
25 ng/mL. Ten percent to 20% of HCCs do not have detectable levels of AFP. Levels are also
raised in nonseminomatous testicular cancer, for which it is a valuable tumor marker (see
discussion later). Twenty percent of patients with gastric or pancreatic cancer and 5% of
patients with colorectal or lung cancer have significant elevations (>5 ng/mL) in serum AFP
levels. Elevated levels are also seen in hepatitis, inflammatory bowel disease, and cirrhosis.
Screening

AFP has an estimated sensitivity of 25% to 75%, a specificity of 76% to 94%, and a positive
predictive value of 9% to 50%. However, note that the sensitivity and specificity vary with
the cutoff value chosen. If the cutoff is set at 20 ng/mL, the sensitivity and specificity are
30% and 87%, respectively, but if raised to 100 and 400 ng/mL, the sensitivity and specificity
vary from 72% to 56% and 70% to 94%, respectively.
The combination of AFP and ultrasound improves the efficacy of screening. One surveillance
study of 1125 patients with HCV reported a sensitivity of 100% with a combination of AFP
and ultrasound versus a sensitivity of 75% for AFP alone and 87% for ultrasound alone.[36]
Cost-effectiveness analysis is used to calculate the cost of each additional life year gained in
terms of quality-adjusted life years (QALYs). A QALY less than $50,000 is considered costeffective. In the United States, recent studies suggest that surveillance of patients with HCVrelated cirrhosis with a combination of AFP and an imaging modality (either ultrasound or
computed tomography) would gain QALYs at acceptable cost. [46] [47] [48]

Prognosis

The AFP concentration reflects tumor size, with levels higher than 400 ng/mL being
associated with larger tumors. As a result, it has been shown that AFP correlates with stage
and prognosis. The rate of increase, expressed as AFP doubling time, has also been associated
with poorer prognosis.
Monitoring

AFP has been shown to decline after resection or ablation. After complete resection, AFP
levels should drop and remain at less than 10 ng/mL. Shirabe and colleagues[40] found that in
patients with HCC whose preoperative AFP level was higher than 100 ng/mL and
postoperative AFP did not fall below 20 ng/mL, early recurrence within the first
postoperative year should be strongly suspected. In patients whose AFP levels do normalize
postoperatively, a subsequent rise in AFP over the course of serial serum measurements has
been found to be the best indicator of recurrent disease. It was the first measured abnormality
in 34% of these patients. However, in some patients who had elevated serum levels of AFP
with their original HCC, postoperative levels of AFP were unreliable in detecting recurrence.
Five (12%) patients did not have elevated serum levels despite the presence of recurrent
disease.
Tumor regrowth after chemoembolization does not correlate with rate of increase in AFP or
tumor burden.
AFP levels usually decline in response to effective chemotherapy. Monitoring of AFP
therefore avoids prolonged use of ineffective and potentially toxic chemotherapy.
Carbohydrate Antigen 19-9

Carbohydrate antigen 19-9 (CA 19-9) is widely used as a serum marker for pancreas cancer,
but its use is limited to monitoring response to therapy, not as a diagnostic marker. It is a
mucin-type glycoprotein expressed on the surface of pancreatic cancer cells and was initially
detected by monoclonal antibodies raised against colon cancer cell lines in a mouse model.
The CA 19-9 epitope is normally present within the biliary tree. Biliary tract disease, both
acute and chronic, can elevate serum CA 19-9 levels.
Testing

CA 19-9 is detected with an immunoassay, and the upper limit of normal for a healthy adult is
37 U/mL. Sensitivities of CA 19-9 in the diagnosis of pancreatic cancer range from 67% to
92%, with specificities ranging from 68% to 92%. The utility of CA 19-9 as a diagnostic
marker is limited in a number of ways. First, patients with negative Lewisa blood group
antigen cannot synthesize CA 19-9, and therefore it is not used as a serologic marker in these
individuals, who make up about 10% of the population. Second, patients with benign biliary
tract disease can have levels up to 400 U/mL, with 87% having concentrations higher than 70
U/mL. Significant numbers of patients with pancreatitis, either acute or chronic, also have
elevated levels. Third, besides pancreatic cancer, CA 19-9 levels are also elevated in patients
with other cancers, including those of the biliary tree (95%), stomach (5%), colon (15%),
liver (HCC, 7%) and lung (13%). For colorectal cancer, CA 19-9 levels add little clinically
useful information to determination of CEA levels.

Screening

CA 19-9 is not useful as a screening modality because of its low sensitivity in early-stage
disease. With increasing levels of CA 19-9, the diagnosis of pancreatic cancer becomes more
accurate. When a cutoff level of 100 U/mL is used, a number of studies have demonstrated
that although sensitivity ranges from 60% to 84%, specificity for pancreas cancer is 95% or
greater. Levels higher than 1000 U/mL are almost diagnostic of pancreatic cancer. Because of
its frequent elevation in benign biliary tract disease, CA 19-9 is not useful in distinguishing
benign from malignant distal common bile duct strictures.
Prognosis

In patients with pancreatic cancer who have CA 19-9 detectable in their serum, the level has
been shown to correlate with tumor burden. For example, higher CA 19-9 levels typically
correlate with higher tumor stage, and more than 95% of patients with unresectable disease
have levels higher than 1000 U/mL. Of patients who undergo curative resection, those whose
CA 19-9 levels returned to normal survived longer than those whose levels fell but never
normalized.
Monitoring

Serial measurement of CA 19-9 is used to monitor response to therapy. A rise in CA 19-9

after curative resection has been shown to precede clinical or computed tomographic
evidence of recurrence by 2 to 9 months. In patients with unresectable/metastatic disease,
failure of CA 19-9 levels to fall with chemotherapy reflects poor tumor response. However, in
both settings, the lack of alternative effective therapies limits the utility of serial monitoring
of CA 19-9.
Prostate-Specific Antigen

Prostate-specific antigen (PSA) is a serine protease that is formed in the prostatic epithelium
and secreted into the prostatic ducts. Its function is to digest the gel that is formed in seminal
fluid after ejaculation. Under normal circumstances, only small amounts of PSA leak into the
circulation. With enlargement of the gland (e.g., in patients with benign prostatic hyperplasia
[BPH]) or distortion of its architecture, serum PSA levels increase. Thus, PSA is considered a
tissue-specific rather than a prostate cancerspecific markerpatients who have undergone
curative radical prostatectomy, as well as females, have no detectable PSA.
Testing

PSA is detected with an immunoassay. Besides BPH, other instances in which serum PSA
levels may be elevated include prostatitis, prostatic massage, prostatic biopsy, and digital
rectal examination. Initial studies set the upper limit of normal for PSA at 4 ng/mL, with
levels greater than 10 ng/mL being suspicious for malignancy and levels of 4 to 10 ng/mL
being indeterminate. Since then, it has been found that the upper limit of the normal range of
PSA increases with age. The limit is 2.5 ng/mL for men aged 40 to 49 years, 3.5 ng/mL for
those 50 to 59, 4.5 ng/mL for those 60 to 69, and 6.5 ng/mL for men 70 years and older. The
rate of increase in PSA in a normal 60-year-old is 0.04 ng/mL/yr.
Expressing PSA relative to prostatic volume and time has also helped discriminate cancer
from benign conditions in which the PSA level is less than 10 ng/mL but greater than the

upper limit of normal for the patient's age. PSA density is defined as the ratio of PSA to
prostatic volume, as measured by transrectal ultrasound or magnetic resonance imaging.
Higher PSA densities are more suggestive of malignancy than BPH because the amount of
PSA released per gram of prostate cancer is significantly greater than that released from
normal prostatic tissue.
The ratio of free to total PSA has also been found to improve the specificity of prostate
cancer diagnosis in the PSA range of 4 to 10 ng/mL. The PSA slope (also known as PSA
velocity) is the rate of change in the concentration of PSA over time. For individuals with
initial levels lower than 4.0 ng/mL, a PSA slope of greater than 0.75 ng/mL/yr is considered
significant; for patients whose baseline level is higher than 4.0 ng/mL, a slope greater than
0.4 ng/mL/yr is considered significant.
Screening

PSA is widely used as a screening tool for prostate cancer because it enables early detection
and diagnosis of this disease. At present, two large randomized trials for prostate cancer, one
in Europe (The European Randomized Study of Screening for Prostate Cancer) and the other
in the United States (The Prostate, Lung, Colorectal and Ovary Cancer), are aiming to answer
the question of whether screening reduces mortality. Results are expected in 2008.
Screening detects prostate cancer earlier. However, much concern has been raised about the
risk of overdiagnosis. Autopsy studies have found that prostate cancer can be found in 55% of
men in their 5th decade of life and 64% in their 7th decade, thus indicating that a significant
proportion of these cancers are not lethal. Only one in eight screening-detected cancers is
likely to kill its host if left untreated.
Monitoring Response to Therapy

After operative resection, the PSA level is expected to normalize after 2 to 3 weeks. In
patients whose PSA level remained elevated 6 months after radical prostatectomy, recurrent
disease eventually developed. In contrast, it takes 3 to 5 months for PSA to normalize after
radiotherapy. However, failure of the PSA level to normalize after radiotherapy also predicts
relapse. A rise in serum PSA is usually the first sign of either local recurrence or metastatic
progression. In patients with advanced disease, PSA levels are also used to monitor response
to systemic therapy.
Carbohydrate Antigen 125

Carbohydrate antigen 125 (CA 125) is a carbohydrate epitope on a glycoprotein carcinoma

antigen. It is present in the fetus and in derivatives of the coelomic epithelium, including the
peritoneum, pleura, pericardium, and amnion. In healthy adults, CA 125 has been detected by
immunohistochemistry in the epithelium of the fallopian tubes, endometrium, and
endocervix. However, neither adult nor fetal ovarian epithelium expresses CA 125.
Testing

CA 125 levels are measured with an immunoassay, with the upper limit of normal set at 35
U/mL. Elevated levels are detected in 80% of patients with ovarian cancer. In patients with
ovarian masses, an elevated CA 125 level has a sensitivity of 75% and a specificity of
approximately 90% for malignancy. It is also detectable in a high percentage of patients with

cancer of the fallopian tube, endometrium, and cervix, as well as in nongynecologic

malignancies of the pancreas, colon, lung, and liver. Benign conditions in which CA 125 is
elevated include endometriosis, adenomyosis, uterine fibroids, pelvic inflammatory disease,
cirrhosis, and ascites. As for CA 19-9 in patients with pancreatic cancer, CA 125 is an adjunct
to diagnosis rather than being diagnostic by itself.
Screening

Alone, CA 125 is not useful as a screening tool for ovarian cancer because of its poor
specificity. However, the United Kingdom Collaborative Trial of Ovarian Cancer Screening is
evaluating the effectiveness of CA 125 in postmenopausal women. In this study, women
classified as high risk according to their CA 125 level are further screened with transvaginal
ultrasound.
Prognosis

Patients with elevated CA 125 levels at the time of diagnosis have a worse prognosis than
patients with normal levels do. Absolute levels of CA 125 do not clearly correlate with tumor
stage, although with increasing stage, greater percentages of patients have elevated CA 125
levels50% of stage I patients, 70% of stage II patients, 90% of stage III patients, and 98%
of stage IV patients.
Monitoring Response to Therapy

CA 125 is of value in monitoring the disease course. Partial or complete response to therapy
is associated with a decrease in CA 125 levels in more than 95% of patients. Increasing levels
of CA 125 correlate with disease recurrence and precede clinical or imaging evidence of
recurrence by a median of 3 months. When rising CA 125 levels are used as an indication for
second-look laparotomy, recurrent disease is found approximately 90% of the time.
CA 125 levels in peritoneal fluid may be more sensitive than serum levels. Thus, in patients
whose serum CA 125 level normalizes during therapy, peritoneal fluid CA 125 levels may
better be able to distinguish patients with residual disease from those without. The upper limit
of normal for peritoneal fluid CA 125 is 200 U/mL.
-Fetoprotein and Human Chorionic Gonadotropin in Testicular Germ Cell Tumors

Nonseminomatous testicular cancers comprise several different histologic types, including

embryonal carcinoma, syncytiotrophoblasts (choriocarcinoma), yolk sac tumors, and
teratomas. Marker expression can be predicted on the basis of the predominant histologic
typehuman chorionic gonadotropin (HCG) is detected in more than 90% of
choriocarcinomas, whereas AFP is expressed by 90% to 95% of yolk sac tumors, 20% of
teratomas, and 10% of embryonal carcinomas.
Diagnosis

Of patients with proven nonseminomatous testicular germ cell tumors, about 50% will have
elevated serum levels of HCG and 60% will have elevated AFP, with either marker being
elevated in 90% of cases. Determination of both marker levels is very important because
nearly half these tumors secrete only one of these substances. In addition to the high rate of
marker positivity, there have been very few cases of spuriously elevated serum levels of HCG

or AFP in patients without testicular cancer. The presence of a testicular tumor in

combination with an elevated level of AFP or HCG is suggestive of testicular cancer, without
being diagnostic. Elevated levels of these markers in a man younger than 40 years without
signs of a testicular tumor may indicate extratesticular germ cell cancer.
Prognosis

An absolute AFP concentration greater than 500 ng/mL or an HCG level higher than 1000
ng/mL is predictive of a poor prognosis. These tumor markers are useful in identifying
biologically distinct categories of morphologically similar tumors. In one study considering
pretreatment levels of AFP and HCG, 92% of patients with normal levels of both markers
achieved complete remission as compared with just 26% of those with elevated AFP only,
46% of those with elevated HCG only, and 35% of those with elevations in both. Similarly,
when comparing groups of patients with similar disease burdens, those with elevated marker
levels have a worse prognosis than do those with normal marker levels.[41]
Monitoring

In the majority of patients with nonseminomatous germ cell tumors, tumor marker levels
correlate with response to chemotherapy. The rate of marker decline (half-life), calculated
from weekly determinations after initiation of chemotherapy, can be used for early
identification of patients who will respond poorly to chemotherapy. Half-lives longer than 3.5
days for HCG or longer than 7 days for AFP suggest that very aggressive therapy is required,
such as high-dose chemotherapy in combination with stem cell transplantation. However,
there is a significant percentage of patients whose levels of tumor markers fall despite failure
of their tumors to regress with therapy.
After completion of primary therapy, increasing marker concentrations, even in the absence
of other features of recurrence, may lead to salvage chemotherapy. Therefore, it is important
to exclude false-positive results. The HCG level needs to be measured in urine, where the
concentration is generally similar to that in serum; however, interfering substances are not
excreted into urine. Intensive chemotherapy may induce hypogonadism with associated HCG
levels of up to 5 to 10 IU/L. It can be differentiated from relapse by measurement of
luteinizing hormone and follicle-stimulating hormonesimilar to the postmenopausal state in
women, levels higher than 30 to 50 IU/L indicate that HCG is derived from the pituitary.
DNA-Based Markers

Specific mutations in oncogenes, tumor-suppressor genes, and mismatch repair genes can
serve as biomarkers. These mutations may be germline, such as the ret proto-oncogene of
MEN 2 and the APC gene of FAP, or somatic mutations, such as the occurrence of p53
mutations in a wide variety of tumors. Chromosomal abnormalities such as the 9:22
translocation that creates the bcr-abl oncogene are also useful biomarkers. Specific singlenucleotide polymorphisms have been identified that are associated with increased risk for
specific cancers, and haplotype assessment has been shown to predict susceptibility to several
cancers, including prostate, breast, lung, and colon cancer. Paik and coworkers[42] described
an algorithm to predict the likelihood of distant recurrence in patients with node-negative,
tamoxifen-treated breast cancer based on the expression of 21 genes in tumor tissue.
Epigenetic Changes

Testing for epigenetic changes is still at an early discovery stage and has not yet reached the
clinic. However, it has great potential for a number of reasons. First, DNA assays for aberrant
methylation are easier and more sensitive than those for point mutations. Second, cancerspecific DNA methylation patterns can be detected in tumor-derived free DNA in the
bloodstream and in epithelial tumor cells shed into the lumen. This ease of access to sample
medium may facilitate efforts at detection and monitoring of cancer. Third, DNA-methylation
profiles are more chemically and biologically stable than RNA or most proteins. As a result,
they may be more reliably detected in diverse biologic fluids.
Methylation biomarker studies have been performed in a variety of cancers, including breast,
esophageal, gastric, colorectal, and prostate cancer. Sources of the DNA have included
plasma/serum, urine, sputum, and saliva. A number of general observations have been made.
Targeted biologic fluid sources of DNA, such as urine for bladder cancer, tended to give
higher clinical sensitivities than serum or plasma analysis did. In contrast, the specificity of
plasma or serum detection of tumor-specific markers was found to be extremely high
approximately 100%. Combining DNA methylation assays may complement existing
screening methods with high sensitivity but low specificity, such as PSA in prostate cancer.
The use of panels of methylation targets in these studies improved the clinical sensitivity of
the assay.
Potential Applications

1.

Early detection. Although abnormal epigenetic silencing of genes can occur at any
time during carcinogenesis, it appears to occur most frequently early in the
transformation process. Aberrant crypt foci that contain preneoplastic hyperplastic
colonic epithelial cells have been found to demonstrate abnormal methylation in
promoter regions of genes involved in abnormal activation of the Wnt signaling
pathway.[43] Detection of abnormal methylation patterns in histologically normal
cells may emerge as a useful marker for assessment of cancer risk.

2.

Predict response to therapy. Methylation of specific genes can be linked to the

biologic behavior of the tumor. A number of studies have reported associations
between DNA methylation markers and response to chemotherapy. The most
extensive work has been done on CpG hypermethylation of the O[6]-methylguanine
DNA methyltransferase (MGMT) gene, which appears to confer sensitivity to
various alkylating chemotherapeutic agents. MGMT methylation was associated with
prolonged survival in glioma patients treated with carmustine and in patients with
large diffuse B-cell lymphoma who were treated with cyclophosphamide as part of
multidrug regimens.[44] Widschwendter and associates[45] studied the correlation
between methylation profiles and hormone receptor status in breast cancer. In
particular, they found that methylation of the ESR1 gene and the PGR gene was the
best predictor of progesterone and estrogen receptor status, respectively.
Furthermore, ESR1 methylation outperformed hormone receptor status as a predictor
of clinical response in patients treated with tamoxifen. Individual methylation
markers, such as the E-cadherin promoter, have also been linked to breast cancer
metastasis.

3.

Prognostication. Abnormal methylation of combinations of genes has been

associated with a poor outcome.

On an opposite note, loss of methylation is increasingly being recognized as an important

event in carcinogenesis.[32] Hypomethylated CpG islands have been associated with the

activation of nearby genes. For example, hypomethylation of the promoter for the
cancer/testis antigen CAGE correlates with increased expression of the gene and is found in
premalignant lesions of the stomach.[46] Similar instances of demethylated promoters
activating their downstream genes have been found in numerous other cancers, including
those of the colon, pancreas, liver, uterus, lung, and cervix.[32] In a recent study of ovarian
carcinogenesis, hypomethylation of centromeric and juxtacentromeric satellite DNA was
found to be increased in tumors of advanced stage or high grade, and this strong
hypomethylation was an independent marker of poor prognosis.[47] Furthermore, genomewide hypomethylation has also been detected in cancer cells and may contribute to genomic
instability.[43]
DNA-methylation profiles in which both hypermethylation and hypomethylation are
examined may provide greater insight into tumor behavior than possible with either profile
alone.
RNA-Based Markers

RNA-based markers have been identified in the context of global mRNA expression by highthroughput technologies. These microarrays (gene chips) allow us to measure the
expression of 30,000 to 40,000 human genes in a single experiment. Statistical modeling then
allows selection of groups of genes, or so-called fingerprints, that best distinguish disease
states. For example, a variety of studies have used gene expression profiles in breast cancer
cells to identify previously unrecognized molecular subtypes associated with differences in
survival,[48] more accurately predict outcome,[49] and predict response to neoadjuvant therapy.
[50]
The level of expression of genes responsible for the metabolism of chemotherapeutic
agents has been used to predict clinical response to these agents in lung and colon cancer.
Similar methods have been applied to many other cancers. Though very promising, these
assays will require extensive multi-institutional validation before they can be used in routine
practice.
Proteomic Profiling

Proteomics is the study of all the proteins expressed by the genome. Ultimately, genetic
mutations are manifested at the protein level and involve derangements in protein function
and communication within diseased cells and within their microenvironment. Execution of
the disease process occurs through altered protein function. Protein tumor biomarkers are
thought to be low-abundance proteins (concentrations in the nanomolar range) that are shed
from tumor cells or from the tumor-host interface into the circulation. Detection and
measurement of these proteins provide information about the clinical behavior of the cancer.
Proteomic profiling using mass spectrometry technologies generates complex fingerprints of
ion peaks corresponding to protein concentrations, which can be correlated with disease
states. Numerous studies using samples of blood (plasma or serum), urine, and pancreatic
juice have demonstrated the feasibility of this technology for discovery of biomarkers and
early detection of ovarian, breast, prostate, and pancreatic cancer. Identification of
reproducible protein signatures of specific diseases has the potential to achieve much higher
diagnostic sensitivity and specificity than possible with currently available biomarkers.
Proteomic profiling lacks a standardized methodology and remains time and labor intensive.
For the moment, these technologies are not ready for routine clinical use. Their principal role
is in discovery of protein biomarkers. Candidate biomarkers discovered through this process

can be validated with standard immunometric techniques after the development of specific
antibodies.
Clearly, the future holds great promise for the greater use of biomarkers in the clinical
management of patients with cancer ( Table 29-8 ). We expect that combinations of tumor
markers, including combinations of different types of tumor markers, will be developed and
then incorporated into formal staging criteria. There will also be further delineation of the
role of tumor markers in predicting the response to biologic and other types of therapy.
Table 29-8 -- Biomarkers and Biologically Targeted Therapies
CANCER

BIOMARKER

THERAPY

Breast

Estrogen receptor,
progesterone receptor

Tamoxifen/aromatase
inhibitors

Lymphoma

CD20

Rituximab

Chronic myelogenous
leukemia (CML)

bcr-abl

Imatinib

Gastrointestinal stromal tumor c-kit

(GIST)

Imatinib

Nonsmall cell lung cancer

EGFR mutation

Gefitinib

Breast

HER2/neu

Trastuzumab

From Ludwig JA, Weinstein JN: Biomarkers in cancer staging, prognosis and treatment
selection. Nat Rev Cancer 5:845-856, 2005.
Biomarker expression is increasingly being used, independent of formal staging criteria, to
decide which patients receive biologically targeted therapies.
EGFR, epidermal growth factor receptor.