MARKERS OF CANCER

Cancer can be thought of as a disease of thebody, a disease of organs, and a disease of cells(Kahn 1994):

Since the clinical expression of cancer beginswith cellular growth, cancer has been definedpathologically as a cellular disease. Cliniciansare also aware that cancer is a disease oforgans. Cell growth yields masses, whichthen invade the organ and disrupt its function.Cancer may also be defined as a diseaseprocess of the body in that the body’sresponse to the presence of the tumor affectsthe presentation and behavior of the patient.The usefulness of the laboratory in cancermanagement must reflect an understanding ofthe changes in the genes, cells, organs, andbodily responses to cancer.

A disease of the bodyThe response of the body to cancer is, in large

part, mediated by cytokines released by inflam-matory cells reacting to the cancer. Certain of thecytokines generate a persistent acute phase response,with its effects upon plasma proteins and circulatingblood cells, and others are responsible for the consti-tution symptoms so typical of cancer, such as fatigueand weight loss (Tisdale 1997). The body alsoreacts to cancer by mounting an immune response tolatent and neo-antigens expressed by the cancercells. This contributes to the chronic inflammatoryresponse and can result in the development ofautoimmune disorders including a number of neuro-logic paraneoplastic syndromes (Dropcho 1998). Ingeneral, the activation of inflammatory and immuneresponses in cancer does not produce laboratoryfindings that are specific for cancer.

Other systemic consequences of cancer resultfrom the release of bioactive substances fromtumors. Hematologic abnormalities caused by can-cer include cytophilias resulting from the secretionof hematopoietic growth factors by tumors andanemia due to the persistent inflammatory state, poornutrition, and bleeding. Coagulation disorders are

also frequently present in patients with cancer.Hemorrhagic conditions are sometimes due tothromobocytopenia, arising as an autoimmuneprocess, but more frequently are a consequence ofchronic activation of the plasminogen system(Carroll and Binder 1999). Thrombotic states arecharacterized by localized thrombosis and dissemi-nated intravascular coagulation. They are caused bythe release of tissue factor and cancer procoagulantfrom cancer cells (Falanga and Rickles 1999,Gordon and Mielicki 1997) and by the thrombocyto-sis and increased plasma concentrations of clottingfactors (especially, fibrinogen and factor VIII)caused by the persistent inflammatory state. Withthe exception of cancer procoagulant, the substancesresponsible for these hematologic abnormalities arenot specific for cancer so they have little role in thelaboratory diagnosis of cancer. The laboratory eval-uation of these processes is, however, a criticalaspect of the care of the cancer patient.

Profound systemic effects are also caused by thehypersecretion of hormones by endocrine tumors.Similar effects are caused by peptide hormonessecreted by some tumors of nonendocrine organs inwhat are referred to as the endocrine/metabolicsyndromes of cancer (Odell 1997). In many cases ofthese syndromes, the hormone that is secreted is thenormal product of a minor population of hormone-secreting cells within the tissue of origin of thecancer, such as the secretion of erythropoietin byrenal carcinomas. In some cases, however, thehormone arises from a cancer in a tissue that is notbelieved to contain cells that elaborate the hormone,such as the secretion of erythropoietin by cerebellarhemangioblastomas. It is possible that in tumors ofthis sort, the hormone is expressed as a normalparacrine cytokine within the organ and that itassumes the status of a hormone simply by itsoverexpression by the tumor (Odell 1997).

A disease of organsCancers generate masses of tumor cells that dis-

place and compress neighboring normal tissueleading to injury of the tissue and loss of tissuefunction. The extent of the effects on the nearby

Cancer 11-1

Chapter 11CANCER

© 2001 Dennis A. Noe

tissue depend, in part, upon the size of the tumor.While still small, tumors tend to have little or noeffect upon the surrounding tissue. As they grow insize, they affect a greater and greater amount ofnormal tissue. Even more important than sizethough is location. Cancers that are positioned tocollapse or obstruct a tubular structure or largevessel will produce effects long before a tumorgrowing far from any vital structure. For instance,because it produces obstruction, a squamous carci-noma of the bronchus will cause much more in theway of clinical effects than a comparably sizedadenocarcinoma in the periphery of the lung.

Another determinant of the extent of the effectsof a tumor upon the surrounding tissue is the patho-biologic stage of the tumor. Before a cancer hasbegun to spread, all of its local effects are due to thevolume of space it occupies. Once it becomesinvasive, cancer cells infiltrate normal tissue, dis-rupting the parenchymal microarchitecture and ob-structing the microvasculature. This is very injuri-ous to the normal tissue and inevitably leads to anaccelerated loss of tissue function. When a cancerbecomes metastatic, it spreads to regional lymphnodes, where it causes local mass effects, and todistant sites, where it causes injury and loss offunction in the seeded organs.

The interplay of tumor location and pathobio-logic state is well illustrated by prostate cancer.Most prostate cancers arise in the posterior portionof the prostate, away from the urethra, and expandwithout causing any obstructive symptoms. If theyare not detected early by clinical screening proce-dures, they will become invasive within the pelvis,in which case they can cause bladder dysfunction,ureteral obstruction with hydronephrosis, and localnerve and blood vessel damage usually manifestingas impotence. Prostate cancer can present withsymptoms of metastatic disease if the tumor metasta-sizes early and the local spread has not producedsymptoms. Bone pain, lymphadenopathy, and con-stitutional signs of cancer, most notably weight loss,constitute the usual symptoms of metastatic disease.In contrast, those cancers that develop more anteri-orly tend to cause symptomatic urethral obstructionearly in their evolution, often before they are widelyinvasive or metastatic.

Because of interindividual variability in the timecourse and pattern of tissue injury and function losscaused by tumors, markers of injury and functionhave a limited role in the diagnosis of cancer.

Function markers are extremely useful, though, inproviding quantitative indices of the functional statusof involved organs.

A disease of cellsCancer cells differ from normal cells in an

immense number of ways, many of which are ofcentral importance in the study of the biology ofcancer (Hanahan and Weinberg 2000). Those dif-ferences that are clinically measurable can be used ascellular markers of cancer. These markers can beplaced into two broad categorizes. The firstcategory consists of markers that are present ascomponents of whole cells. The laboratory mea-surement of these markers requires a cell sample.The second category consists of marker substancesthat are present in the body fluids as a result ofhaving been secreted by living cancer cells orreleased from dead cancer cells. These markers aremeasured in body fluids, most often in plasma, anddo not require access to a cell sample.

Cancer cells. Cancer cells are identified bystudies of cellular phenotype and cellular genotype.Of the studies of phenotype, the characterization ofmorphology by microscopy is the most important. Itremains the most specific of all laboratory markersof cancer. For this reason, the microscopic exami-nation of cells or tissue obtained by biopsy of atumor is considered essential for the diagnosis ofcancer. Morphologic characteristics of interest in-clude the cytologic features of individual cells, themicroarchitecture of the cell population, and theinvasion of normal tissue by the cells. Biopsytechniques that provide tissue specimens usuallyallow all of these characteristics to be evaluated.The cytologic features of some cancers are highlyspecific and allow for the diagnosis of cancer basedon the examination of cells obtained by scrape oraspiration biopsy.

Genetic abnormalities are frequently demonstra-ble in cancer cells (Table 11.1). Many of the abnor-malities that have been found are present soconsistently among tumors of the same type that theyprobably represent genetic changes involved in theinception and malignant progression of that tumortype (e.g., Ried et al. 1996). Large-scale structuralabnormalities of the chromosomes such as deletionsand translocations can be studied by chromosomeanalysis (see Chapter 10). Chromosome analysis byG-banding is performed on tumor cells taken directlyfrom biopsy material or on tumor cells grown in

Cancer 11-2

short-term culture. Direct preparations tend to haveonly a few metaphase cells for study. Tumor cellcultures usually yield a satisfactory number ofmetaphase cells but culturing of the cells can selectagainst the more dysfunctional cancer cells in thetumor population and can select for normal cellsco-inoculated from the biopsy material (Ketter et al.1996). This decreases the sensitivity of the method.Karyotyping by fluorescence in situ hybridizationsuffers from the same problems. The use of fluores-cence in situ hybridization for the detection ofdefined chromosome abnormalities, on the otherhand, has many advantages: it can be performed oninterphase nuclei, so tumor cell culture is not neces-sary, and paraffin-embedded biopsy materialprepared for histologic study can serve as the sourceof cells. It has proven to be an informative andaccurate method for the directed cytogenetic analysisof solid tumors (Werner et al. 1997) and hemato-logic malignancies (Kearney 1999).

Chromosome deletions and translocations canalso be studied in fresh, frozen, and even paraffin-embedded material using a variety of moleculardiagnostic techniques (see Chapter 10). Considerthe t(9;22) translocation that results in the Philadel-phia chromosome of chronic myelogenous leukemia(Faderl et al. 1999). It consistently arises fromtranslocation of the Abelson oncogene, abl, onchromosome 9 to the breakpoint cluster region, bcr,on chromosome 22. The presence of the bcr/ablfusion DNA sequence can be demonstrated bySouthern blot hybridization or preferential PCRamplification. The translocation can also be demon-strated by detection of either the bcr/abl fusionprotein or the bcr/abl fusion gene transcriptionproduct. The fusion mRNA is detected using North-ern blot hybridization or reverse transcription-PCRspecific for the bcr/abl mRNA sequence. Deletionsand translocations that show extensive heterogeneityin the site of the chromosome abnormality are notreadily studied using molecular diagnostic methods.Gene amplification is another genetic abnormality

contributing to the development and progression ofcancer. Amplification usually involves a region ofDNA larger than one gene and can produce a struc-tural alteration large enough to be detected bychromosome analysis either as double minutes(separate small DNA fragments) or as a homogene-ously staining region (DNA segment inserted into achromosome). Smaller amplifications can bedemonstrated using quantitative modifications offluorescence in situ hybridization, blot hybridization,and PCR amplification (Crotty et al. 1994).

Cancer-associated point mutations of oncogenesand tumor suppressor genes may be restricted toonly a few base positions in the gene, such as rasoncogene mutations, but more typically they occur atdiverse bases in the gene, as seen with the p53 tumorsuppressor gene (Loda 1994). Allele-specific oligo-nucleotide hybridization is most commonly used todetect point mutations when the mutations in a geneconsist of only a few recognized nucleotide substitu-tions. That technique is not applied to genes thathave a large catalog of nucleotide substitutionsbecause of the large number of probes that have tobe used. A different technique, single strand confor-mational polymorphism analysis, is used then(Nollau and Wagener 1997). In this technique, thegene’s DNA is amplified by PCR using labeledprimers or labeled nucleotides (so that all of theamplification products are labeled). The productsare digested using restriction endonucleases toproduce fragments smaller than 200 bp, denatured toyield single-stranded DNA, and electrophoreticallyseparated in nondenaturing polyacrylamide gels.Single-stranded DNA fragments form secondarystructures that depend upon the base composition ofthe fragment; the structure can vary as a result ofeven a single base alteration. The different singlestrand structures migrate with different speeds in thegel leading to their separation. Mutated DNA isrecognized by the abnormal migration position of itssingle strand structure. Mutations at any position inthe gene can be detected using this method.

Cancer 11-3

Table 11.1Some Genetic Alterations Seen in Cancer Alteration Example Cancer

Chromosome deletion del 8p12 prostate carcinoma

Chromosome translocation t(15;17) acute promyelocytic leukemia

Gene amplification erbB2 breast carcinoma

Gene point mutation K-ras pancreatic carcinoma

A different way in which genetic alterations canbe used in the laboratory evaluation of cancer is todetect a genomic change that is not abnormal initself but which indicates a monoclonal origin of thetumor cells. The foremost example of this is in theevaluation of lymphoid cancer where monoclonalityis revealed by demonstrating an identical V(D)Jrearrangement in all of the cancer cells. Homogene-ity of the immunoglobulin heavy chain gene rear-rangement is found in tumors of B lymphocytelineage, and of the T cell receptor α or β genes intumors of T cell lineage (Scarpa and Achille 1997).

Marker substances. The substances that serveas markers for cancer originate in cancer cells andenter the circulation following secretion from livingcells or release from dead cells (Figure 11.1).Secreted substances enter the plasma directly anddistribute in the extracellular fluids. They are re-moved by systemic processes. Substances that arereleased from dead cells enter the extracellular spacein and around the tumor and are either catabolizedlocally or are removed in the lymph. Lymph-bornemarker substances eventually enter the plasma,distribute in the extracellular fluids, and are elimi-nated. Marker substances released from cancerousblood cells are unusual in that they have directaccess to the plasma. Marker substance concentra-tions can be measured in any of the body fluidstransited by the marker substance. The most com-monly studied fluid is plasma. Other fluids, such aspericardial fluid and peritoneal fluid, are studiedwhen tumor involvement of the respective mesothe-lial linings is suspected. Urine may be studied if themarker substance or a metabolite of the substance is

eliminated by renal clearance; concentrations in theurine are higher than in the plasma thereby allowingfor easy analyte measurement.

The concentration of a marker substance in theplasma and extracellular fluids depends upon the rateof entry of the substance into the fluids and its rateof systemic elimination. Except in advanceddisease, when many body functions are compro-mised, the systemic elimination of marker substancesis fairly constant. That means that the predominantvariable in the plasma concentration of markersubstances is the rate of entry into the body fluids.For secreted substances, the rate of entry is deter-mined by the individual cell rate of substance synthe-sis and secretion and by the number of cells, i.e. thesize of the tumor. The individual cell secretion rateusually varies from cell to cell within a tumor due tothe greater than normal degree of inter-cell pheno-typic variability found in cancer. The rate alsovaries depending upon where the cancer cells are intheir malignant evolution. Some marker substanceswill be expressed in early in the evolution of thecancer, when the cancer cells are fairly well differ-entiated, and not later, when the cells show poordifferentiation. Other marker substances will beexpressed in poorly differentiated cancer cells andnot in well differentiated cells. Still other markersubstances will be expressed throughout the pheno-typic evolution of the cancer.

For released substances, the entry rate is deter-mined by the individual cell content of the markersubstance, by the number of cells, and by theturnover rate of the cells. As with secreted markersubstances, released marker substances show

Cancer 11-4

Figure 11.1 A model of the disposition of marker substances for cancer.

localcatabolism

plasmaextracellular

fluids

systemicelimination

lymphPLASMA AND

EXTRACELLULAR FLUIDS

LOSS

CANCER CELLSdeath

clonalevolution

loss of markersnew markers

localextracellular

pool

variability in the individual cell expression of markersubstance within a tumor and variability in the levelof expression of the substance at different stages inthe malignant evolution of the cancer.

The clinical application of a marker substancedepends upon two forms of tissue specificity: aspecificity for cancerous tissue as opposed tononcancerous tissue and a specificity for a particulartissue or organ. Tissue specificity is achieved byutilizing substances that arise predominantly incancers of the tissue of interest. Such substances areusually identified in one of two ways. One approachis to identify substances that show specificity for thenormal tissue hoping that they will also be specificfor cancer of the tissue. Not infrequently, however,the candidate substances are less tissue specific incancer. One reason for this is that cancer cells ofthe tissue may express the substance at a lower levelthan normal tissue cells. Another reason is thatcancer cells from other tissues, especially embryo-logically related tissues, may express the substanceat higher levels than normal. The second approachis to identify substances expressed in cancer of thetissue of interest hoping that some will be specificfor the cancer. This has often been done by inocu-lating animals with human tumor cells to raiseantibodies to cancer cell substances. The substancesso identified are typically referred to as tumorantigens (Sell 1980). In this approach, a lack oftissue specificity of the candidate substance can bedue to expression of the substance in cancers ofother tissues or even expression of the substance inother normal tissues.

As regards cancer specificity, in the few waysthat the genome of a cancer cell differs from thenormal genome, there is the potential for the produc-tion of a substance that is truly specific for the

cancer producing it. The bcr/abl fusion proteinproduced by the t(9;22) translocation of chronicmyelogenous leukemia is an example of such asubstance. Unfortunately, no substances of this sorthave yet been found that achieve clinically measur-able concentrations in the body fluids.

The marker substances currently in clinical useare products of that portion of the cancer genomeshared with the normal cell genome and, therefore,have the potential for being produced by normalcells (Table 11.2). That means that none of them isabsolutely specific for cancer. The degree of cancerspecificity they do attain depends in large part uponthe relative specificity of the attribute of cancerreflected in the laboratory measurement of the sub-stance. These attributes include: monoclonality,altered expression of cellular constituents andproducts, alteration in the dynamics of substancerelease, and increased cell turnover. Of these attrib-utes, monoclonality is the most specific for cancer.

As mentioned previously, monoclonality can bedemonstrated in B lymphocyte cancer by showinghomogeneity of the V(D)J rearrangement of theimmunoglobulin heavy chain gene in the cancerousB lymphocytes. Monoclonality can similarly beimplied from the demonstration of structuralhomogeneity of the immunoglobulins secreted bycancerous B lymphocytes and plasma cells (Keren1999).

Alpha-fetoprotein is an example of a markersubstance that is expressed at much higher levelsthan normal in certain cancers (Abelev and Eraiser1999). Alpha-fetoprotein is an albumin-like plasmaprotein expressed at high levels in the yolk sacduring early embryonic development and in the liverduring late embryonic and early fetal development(Deutsch 1991). In mature liver cells it is normally

Cancer 11-5

Table 11.2Some Selected Marker Substances of CancerClass Marker substance Cancer

Cellular constituents carcinoembryonic antigen colorectal carcinoma

CA-125 ovarian carcinoma

Secreted products

Hormones β-chorionic gonadotropin germ cell tumors, choriocarcinoma

calcitonin medullary carcinoma of the thyroid

Enzymes prostate-specific antigen prostate carcinoma

Plasma proteins immunoglobulin multiple myeloma, B cell leukemia

alpha-fetoprotein germ cell tumors, primary liver cancer

expressed at extremely low levels or not at all, sothat it is present in only trace concentrations in theplasma. Its expression in cancers that have cells ofyolk sac lineage (germ cell tumors) or of hepatocytelineage (primary liver cancers) results in markedlyelevated plasma concentrations. Its specificity forcancer is fairly high but increased plasma concentra-tions are sometimes seen in conditions associatedwith extensive hepatocellular injury, such as acuteviral hepatitis and liver metastases. Elevated plasmaconcentrations can also be seen in cirrhosis,presumably as a result of alpha-fetoprotein expres-sion in regenerative hepatocytes.

Carcinoembryonic antigen (CEA) and prostatespecific antigen (PSA) are examples of markersubstances that are expressed at somewhat higherthan normal levels in certain cancers (Hammarström1999, Stenman et al. 1999, respectively). CEA isan adhesive protein found in the glycocalyx of themicrovillar surface of intestinal cells. PSA is aserine protease secreted by prostate epithelial cellsinto seminal fluid. Both proteins are expressed inmature tissue. They are normally present in lowconcentration in the plasma, lower than expectedbased on the level of expression of the proteins. Theprobable reason for this is that, following epithelialcell death, the majority of the released substance islost into the respective organ lumen, CEA into thebowel lumen and PSA into the prostatic ducts, ratherthan entering the local extracellular space where it isavailable for lymphatic uptake.

Cancer cells express these marker substances athigher than normal levels. This cellular over-expression and the increased turnover typical ofcancer cells increase the plasma concentrations ofthe substances. In addition, derangement of theglandular architecture in cancer contributes to theelevated plasma concentrations. Many of the glandsdo not have ductular connections. All of the markersubstance released by cell turnover in such glandsenters the extracellular space, with no loss into alumen, and ends up contributing to the circulatingpool of the substance. The same is true of cancercells not forming glands. Because these markersubstances are expressed in normal tissue, theirplasma concentrations are increased in conditionscausing injury to the respective organs. For CEA,injury to embryologically related organs that expressthe substance at low levels, such as the pancreas, canalso elevate its plasma concentration. Benign prolif-erative disorders also lead to increased marker

substance concentrations. Such behavior is charac-teristic of all marker substances that depend upondifferential expression for their specificity. It is themajor limitation to the specificity of such markersubstances.

Hormones are also thought of as tumor markersbut they are generally not at all specific for cancer.Cancers of the endocrine organs only infrequentlysecrete hormones; hypersecretion of hormonesinstead usually indicates benign disorders, eitherhyperplasia or an adenoma. Calcitonin secretion bymedullary carcinoma of the thyroid is one exception(Giuffrida and Gharib 1998). Calcitonin is secretedby the thyroid C-cells, also called parafollicularcells. Proliferative disorders of C-cells are rare inthe general population but are always present inindividuals with one of the familial C-cell syn-dromes: multiple endocrine neoplasia types 2a,multiple endocrine neoplasia types 2b, and familialmedullary carcinoma of the thyroid. In these syn-dromes, C-cell disease is invariably malignant(medullary carcinoma of the thyroid) or premalig-nant (C-cell hyperplasia). Calcitonin hypersecretionis usual in both malignant and premalignant diseasemaking plasma calcitonin a sensitive and specificmarker substance for C-cell cancer. Another hor-mone that, outside of pregnancy, is specific forcancer is β-chorionic gonadotropin which markschoricarcinomatous elements in germ cell tumors(Abelev and Eraiser 1999).

DIAGNOSIS

Markers of cancer are used in every aspect ofmedical care for cancer (Hayes et al. 1996). Theyare employed in screening for cancer and are veryimportant in monitoring for post-therapeutic recur-rence of disease. Markers have a lesser role inestablishing the diagnosis of cancer; because of itssuperior reliability, pathologic examination is almostalways considered necessary for the diagnosis ofcancer. However, markers, especially cellularmarkers, can be useful in supporting a microscopicdiagnosis of cancer, in elucidating the cellularlineage of the cancer, and in classifying therapeuticsubgroups. Prognosis in cancer relates most directlyto tumor type and to anatomic markers of invasive-ness and metastatic spread but considerable effortsare currently being made, and some success has beenachieved, in identifying markers that aid in refiningprognostic classification.

Cancer 11-6

ScreeningThe rationale for clinical screening programs is

the detection of treatable disorders prior to theirbecoming clinically manifest. In consideration ofthe criterion for being treatable, this translates, at aminimum, into the detection of cancer before it hasbecome metastatic and often it translates into thedetection of cancer while it is still limited to theorgan of origin. In order to achieve this end, thescreening study must have acceptable diagnosticperformance during some portion of the early phaseof the growth of the cancer—that is, it must have ascreening window—and the screening program mustbe designed so as to assure that the screening studyis performed at least once during that screeningwindow. The likelihood of a screening markerachieving the first goal depends upon the naturalhistory of the cancer and the pattern of the marker’sexpression within the context of that natural history.The likelihood of a screening program achieving thesecond goal depends upon the schedule on which thescreening study is performed. These points will beillustrated by a consideration of screening forprostate cancer using PSA. PSA will also serve asthe illustrative marker in several subsequent sectionsof this chapter. There is no small degree of contro-versy surrounding the use of PSA as a screeningmarker (Woolf and Rothemich 1999, Svetec andThompson 1998, Moss and Melia 1998). The readeris strongly urged to refer to the cited articles as wellas to more recent articles in the literature to appreci-ate the many viewpoints in this controversy.

The natural history of prostate cancer, accordingto the model proposed by Stenman et al. (1999), isdepicted in Figure 11.2. There is an initial phase ofintraepithelial neoplasia that can arise as early as thethird or fourth decade of life (Franks 1954) butwhich may appear later in life. During this phase,which is of uncertain duration (in the figure, aduration of 15 years is shown), the cancer has afairly slow rate of grow. Once the tumor becomeslocally invasive it experiences a period of relativelyrapid growth. When the cancer reaches a size atwhich vascularization is needed, its growth ratelessens with doubling times estimated to be on theorder of two to three years (Schmid et al. 1993).This rate of growth persists until the cancermetastasizes.

Before prostate cancer becomes metastatic, therate of entry of PSA into the body fluids appears tobe directly related to the size of the cancer. A direct

relationship between size and PSA entry rate is alsofound for benign prostatic hyperplasia (BPH). Themagnitude of the entry rates differs considerablybetween the two diseases, however. Stamey et al.(1987) estimate that, on average, normal prostatecontributes to plasma PSA concentration at a rate of0.2 µg/L/g, BPH at a rate of 0.6 ug/L/g, andprostate cancer at a rate of 2 µg/L/g. Statisticalmodeling of the large data set in the report byCollins et al. (1993) indicates that BPH increasesplasma PSA concentrations with a median value of0.13 µg/L/g and that normal prostate makes only anextremely small contribution (Noe, unpublished;based on a model of direct proportionality betweentissue volume and plasma PSA concentration and alognormal distribution of PSA values). Modeling ofthe data in Partin et al. (1990) yields median valuesof 0.14 µg/L/g and 2.64 µg/L/g, respectively, forBPH and prostate cancer (Noe, unpublished).

In men with prostate cancer but otherwisenormal prostates, prostate cancer would be detect-able as soon as the plasma PSA concentrationreached the limit of detection of the assay used tomeasure PSA concentration. A desirable value forthis limit is 0.2 µg/L (Stenman et al. 1995). Basedon a median value of 2.64 µg/L PSA concentrationrise per 1 g of cancer tissue, half of the men wouldhave detectable PSA concentrations while theircancers weighed 0.075 g or less. If the standarddeviation in the lognormal distribution of therelationship between PSA concentration and tissuemass in prostate cancer is similar to that in BPH,ninety-five percent of the cancers would be detect-able while their mass was less than 0.22 g. This iswell shy of the mass associated with capsular

Cancer 11-7

Figure 11.2 A model of the development of prostatecancer.

0 10 20 30 40 50 60 70 80

Years of development

0.01

0.1

1

10

100

Tum

or d

iam

eter

(mm

)

invasion (3 to 10 g, Stenman et al. 1999), indicatingthat most cancers could be detected while they wereorgan limited. But would they be detected? Thatdepends upon the screening program schedule. If,for example, the schedule called for a single PSAdetermination at the age of 50 years, prostatecancers would be detected only in those men whohad a cancer that had already reached a detectablesize. They would represent only a fraction of theindividuals who would eventually have symptomaticprostate cancer. If, on the other hand, the PSAconcentration were measured yearly after the age of50 years, almost every cancer would be detectedbefore capsular invasion. This is so because it takesabout 13 years for a prostate cancer to grow from0.22 g to 3 grams (see Figure 11.2; 0.22 g and 3 gequate to diameters of 7.5 and 18 mm, respectively).Thus, many screening studies would be performedon every man harboring a prostate cancer while thecancer was detectable and pre-invasive. An optimalscreening schedule is one in which the screeningstudy is performed once during the screeningwindow. If a sensitivity of 0.95 were to be consid-ered the minimal acceptable screening performance,the screening window for PSA would be 13 years,from a tumor mass of 0.22 g to a mass of 3 g.Performing the screening study on a schedule ofonce every 13 years would therefore be optimal. Toaccount for possible lapses in the regularity ofobtaining a screening study, a PSA determinationevery 10 years might be recommended. A problemwith this analysis is, of course, that it applies to onlya few men, those with normal prostate glands. By

the time they reach their fifties, an age at whichstarting to screen for prostate cancer is reasonablebased on the natural history of the disease, most menhave developed some degree of BPH. This meansthat there is an appreciable background plasma PSAconcentration. Figure 11.3 shows populationfrequency distributions for plasma PSA concentra-tion. The PSA concentrations in men without cancerare due entirely to the presence of BPH; the PSAconcentrations in men with cancer are due to cancerand concomitant BPH (Partin et al. 1990). There isa considerable overlap in the frequency distributionsof the two populations. Consequently, screening forprostate cancer in the male population as a wholenecessitates a tradeoff between screening sensitivityand specificity. For unscheduled, or one-time,screening among men 50 to 59 years old, the trade-off is as shown in Figure 11.4 (for a similar analysisof a different data set read the article by Meigs et al.1996). To determine the critical screening value forthe plasma PSA concentration in this setting, thethreshold likelihood ratio for followup must becalculated. This is done using the formula,

threshold likelihood ratio for followup =

(1 − prevalence) P[rejection]prevalence (1 −P[rejection])

where P[rejection] is the threshold posteriorprobability for rejection of the diagnosis of prostate

Cancer 11-8

Figure 11.3 Reference frequency distributions for plasmaPSA concentration. The curves represent lognormalfrequency distributions fit to the data reported by Catalonaet al. (1994) for men 50 to 59 years old.

0 5 10 15 20

Plasma PSA concentration (µg/L)

0

0.1

0.2

0.3

0.4Fr

eque

ncy Cancer free

Cancer

0 0.2 0.4 0.6 0.8 1

Specificity

0

0.2

0.4

0.6

0.8

1

Sens

itivi

ty

Figure 11.4 ROC curves for screening for prostate cancerusing plasma PSA concentration. The squares representthe data reported by Catalona et al. (1994) for men 50 to 59years old. The continuous line is the curve constructed fromlognormal frequency distributions of the data.

cancer. Using the value derived for the thresholdlikelihood ratio for followup, the critical screeningvalue is determined from the likelihood ratio curve(shown in Figure 11.5).

To calculate the threshold likelihood ratio forfollowup, the prevalence of prostate disease in theclinical population must be estimated and the thresh-old posterior probability for rejection of the diagno-sis of prostate cancer must be selected. In theirstudy, Catalona et al. (1994b) found that, in menaged 50 to 59 years, the prevalence of prostatecancer is approximately 0.04. This value serves asthe prevalence estimate. Any of a number of valuescould be selected for the threshold posteriorprobability for rejection of the diagnosis. Forinstance, it might be felt to be reasonable to rejectthe diagnosis if the posterior probability of cancer isless than 0.06 (one and one-half times theprevalence). Substituting this threshold probabilityinto the formula yields a threshold likelihood ratio of1.5. The corresponding critical value for the plasmaPSA concentration is 4.0 µg/L. Using the modeledreference frequency distributions (Figure 11.3) and acritical value 4.0 µg/L, the sensitivity of the screen-ing study is predicted to be 0.67. Catalona et al.report a sensitivity of 0.75 in their study sample atthis critical value (Catalona et al. 1994b). They alsoreport that approximately 70 percent of the cancersdetected using this critical value were confined to theprostate and thus highly treatable. The specificity atthis critical value is predicted to be 0.74 and wasfound by Catalona et al. (1994b) to be 0.63.

Typically, a screening program gives betterperformance than unscheduled screening: betterspecificity can be had at a comparable rate of detec-tion of organ-limited cancer or an improved rate ofdetecting organ-limited cancer can be obtained at acomparable specificity. Consider the case in whichthe specificity of the PSA screening program in menaged 50 to 59 years is targeted to be 0.75. Thisspecificity is the same as that predicted in theforegoing discussion of unscheduled PSA screening.For a single performance of the study, the stipulatedspecificity is obtained using a critical PSA concen-tration of 4.0 µg/L. In a screening program, thestudy is performed multiple times so, in general,using the same critical value would result in a loweroverall specificity. However, overall specificityremains nearly the same as single-study specificitywhen the repeated measurements are highly corre-lated, as they would be in a PSA screening programbecause the PSA concentration in each individualreflects the mass of BPH in that individual and themass changes only slowly over time. Therefore, acritical value of 4.0 µg/L can be used for the screen-ing program. The schedule of screening studies iscalculated based on the target detection rate.Consider, for example, a fairly ambitious targetdetection rate, 0.9. At the stipulated critical value,this detection rate is found to correspond to a cancermass of 1.74 g. (This calculation is based on anassumption made earlier, that the standard deviationin the lognormal distribution of the relationshipbetween PSA concentration and tissue mass incancer is similar to that in BPH, and takes intoaccount the concomitant presence of BPH). Thismass is about half that at which prostate cancer firstshows capsular invasion, which is 3 g (Stenman etal. 1999), so the screening window is approximatelyone doubling time which, for tumors of this size, istwo to three years (Schmid et al. 1993). Therefore,a cautious screening schedule would consist of aPSA determination every two years.

Improving screening performance. The per-formance of plasma PSA concentration as a screen-ing study for prostate cancer is not outstanding.This can be appreciated by comparing the ROCcurve for PSA concentration (Figure 11.4) with theROC curve for a truly excellent diagnostic markersuch as ferritin concentration, which is used to detectiron deficiency in adults with anemia (Figure 3.14).The area under the ROC curve for PSA con-centration is 0.78 while that for ferritin

Cancer 11-9

Figure 11.5 The likelihood ratio of prostate cancer in men50 to 59 years old as a function of their plasma PSAconcentration constructed from lognormal frequency distri-butions (Figure 11.3). Also shown is a reference line for alikelihood ratio of 1.

0 5 10 15 20

Plasma PSA concentration (µg/L)

0.01

0.1

1

10Li

kelih

ood

ratio

of p

rost

ate

canc

er

concentration is 0.95. Clearly, it would be better ifthe screening performance for PSA concentrationwere more like the performance of ferritinconcentration.

One way to improve the discriminatory power ofa screening marker is to reduce the preanalytic andanalytic components of the variability in themeasurement of the marker. In the case of PSAconcentration, preanalytic variability could bereduced, for example, by stipulating that bloodspecimens be obtained only after a 2 day abstinencefrom sexual activity, as ejaculation is associated witha 24-hour elevation in plasma PSA concentration(Simak et al. 1993). Because digital rectal examina-tion can sometimes raise the plasma PSA concentra-tion (Bruel et al. 1992), it is already recommendedthat blood specimens be drawn before the physicalexamination. There are many ways to reduce theanalytic variability of a laboratory study, but themost important way is to use a more specificmethod of measurement. Immunoassays are used tomeasure PSA concentration. Some of the availableassays utilize polyclonal antibodies which can beexpected to recognize various epitopes on the PSAmolecule. Standardization in terms of utilizingmonoclonal antibodies to a single epitope couldreduce analytic variability and improve inter-laboratory transportability of study results (Stenmanet al. 1995).

Another way to improve screening performanceis to take account of known sources of biologicvariability in the measurement of the screening

marker. As discussed in Chapter 6, such sources ofvariability typically include gender, age, and race.Age has been shown to contribute to the variabilityof PSA concentrations. Figure 11.6 shows the fre-quency distribution for PSA concentration as afunction of age, as modeled by Oesterling et al.(1993). The frequency distribution shifts rightward,i.e., toward higher values of PSA concentration,with increasing age. Much of this age effect is dueto an increase in prostate size in men over the age of55 years; the increase being attributable to progres-sive BPH (Collins et al. 1993). However, there isalso an independent contribution of age alone(Oesterling et al. 1993, Collins et al. 1993).

The frequency distribution of PSA concentrationin patients with prostate cancer also shifts rightwardwith increasing age. The magnitudes of the right-ward shifts of the two distributions are nearly equal.Consequently, the separation between the two distri-butions remains constant and the ROC curve forscreening for prostate cancer using PSA concentra-tion retains roughly the same shape and area underthe curve (Catalona et al. 1994b). Note, however,that the PSA values associated with given points onthe ROC curve change with increasing age. Forinstance, Catalona et al. (1994b) found that thesensitivity and specificity achieved with a PSAconcentration of 4.0 µg/L in men aged 50 to 59 areapproximately 0.75 and 0.65, respectively. In menaged 60 to 69, this screening performance is foundwith a PSA concentration of 4.5 µg/L, and in men70 years and older with a PSA concentration of 5.0µg/L.

The age-specific frequency distributions shownin Figure 11.6 were derived from a population ofwhite males living in Minnesota. Race is anothersource of biologic variability in PSA concentration,so the distributions cannot be applied to men of otherraces. In particular, on a decade-by-decade basis,median PSA concentrations in black men are compa-rable to those in white men, but the upper limits ofthe normal range are much higher in black men(Morgan et al. 1996).

A more explicit way in which to account for thecontribution of hyperplastic prostate tissue to theplasma PSA concentration is to adjust for the size ofthe prostate. Prostate volume can be measured clini-cally by transrectal ultrasound. The method isrelatively easy to perform but unfortunately suffersfrom appreciable measurement variability (Catalonaet al. 1994a). Figure 11.7 shows the reference range

Cancer 11-10

Figure 11.6 Plasma PSA concentration as a function ofage. The dots represent the observed data. The lines arecontours of the frequency distribution as calculated basedon a lognormal statistical model of PSA values. ReprintedOesterling JE, Jacobsen SJ, Chute CG, Guess HA, GirmanCJ, Panser LA, and Lieber MM. 1993. Serum prostate-specific antigen in a community-based population of healthymen. JAMA 270:860.

for PSA concentration as a function of prostate sizeas modeled by Oesterling et al. (1993). The lines inthe figure can be used like linear discriminantfunctions for screening classification. For instance,to achieve a screening specificity of 0.90, PSAconcentration and prostatic volume measurementpairs above the “90 percentile” line would beconsidered screen-positive and those below would beconsidered screen-negative. Alternatively, the PSAconcentration and prostatic volume measurementscan be used to derive a discriminant score that isused to classify result pairs. The most appropriatealgebraic form for the discriminant score is a linearequation. The algebraic form that is actually usedclinically, however, is the ratio of PSA concentra-tion to prostate volume, referred to as the PSAdensity. Evaluation of this ratio in a modest numberof men indicates that a ratio of 0.05 has a sensitivityof 0.95 for cancer confined to the prostate and aspecificity of 0.70; a ratio of 0.10 has a sensitivityof 0.80 and a specificity of 0.95 (Benson et al.1992). This is far better screening performance thanthat achieved by the PSA concentration. The speci-ficity estimates are consistent with the data reportedby Oesterling et al. (1993) in their large study ofPSA concentration and prostate volume (Figure11.7; the 75th percentile line approximates a PSAdensity of 0.06 and the 95th percentile line approxi-mates a PSA density of 0.11). The sensitivity esti-mates are similar to those found for PSA density in alarge study reported by Catalona et al. (1994a): 0.71and 0.90 for PSA densities of 0.05 and 0.10,

respectively. The specificity estimates reported byCatalona et al. (1994a) are lower than those reportedby Benson et al. (1992). However, the study byCatalona et al. (1994a) was confined to men withPSA concentrations above 4.0 µg/L or with abnor-mal digital rectal examinations, resulting in a strongbias toward higher values of PSA density amongmen with BPH, and thereby leading to significantunderestimation of the specificity. (Unfortunately,this design limitation is not widely appreciated andthe low specificity estimates found in the study arecited as evidence that the PSA density is not a betterscreening study than PSA concentration).

It is sometimes possible to increase the clinicalinformation provided by a laboratory measurementby interpreting the result in the context of a physiol-ogic model. An example of this would be the inter-pretation of arterial blood gas measurements using aquantitative model of acid-base metabolism. Model-based interpretation can also be used to improve thediscriminatory power of a screening study. Incancer screening, a model of tumor growth is alogical choice. As a tumor grows, the amount ofmarker substance released by it usually increases;although this will not be so if the growth resultsfrom, or is accompanied by, malignant evolution ofthe cancer with reduction in the expression of themarker. As the rate of entry of the marker into thecirculation increases, the plasma concentration of themarker increases. If the marker is followed overtime, the steady increase in its concentration can bedemonstrated and the presence of the growing tumorinferred. An advantage of this approach is that anupward trend in marker concentration may be detect-able at concentrations of the marker that are belowwhat would otherwise be the critical value for themarker. Hence, a cancer can be detected when it issmaller. A disadvantage of the approach is that anindividual must be tested a number of times over aninterval of time long enough to allow for a trend inmarker concentration to be detected.

In screening for prostate cancer, the trend inPSA concentrations over time is called the PSAvelocity. It is usually evaluated as the average of twoconsecutive measurements of the rate of change ofthe PSA concentration. Studies of the variability ofPSA concentration in healthy men and in men withBPH indicate that an interval of at least a yearbetween concentration determinations is desirablewhen computing the rate of change in the concentra-tion (Carter et al. 1995, Kadmon et al. 1996). That

Cancer 11-11

Figure 11.7 Plasma PSA concentration as a function ofprostate volume. The dots represent the observed data.The lines are contours of the frequency distribution ascalculated based on a lognormal statistical model of PSAvalues. Reprinted Oesterling JE, Jacobsen SJ, Chute CG,Guess HA, Girman CJ, Panser LA, and Lieber MM. 1993.Serum prostate-specific antigen in a community-basedpopulation of healthy men. JAMA 270:860.

means that three PSA determinations equally spacedover two years is the minimum testing schedule.

A nonzero PSA velocity is expected in men withprostate cancer but also in men with BPH, as BPH isalso a progressive process. Because the rate ofrelease of PSA from hyperplastic prostate tissue isusually small and that from cancerous prostate tissueis generally large, the PSA velocity would beexpected to be lower in men with BPH than in menwith tumors. However, in men with PSA concentra-tions less than 4 µg/L, the PSA velocity in men withprostate cancer is similar to that in men with BPH(Carter et al. 1992). A dramatic increase in PSAvelocity is found in most men with prostate canceronly after their PSA concentration exceeds 4 µg/L(Carter et al. 1992). It appears, therefore, that PSAvelocity does not improve the lead time for detectionof prostate cancer when compared to simple screen-ing based on a critical PSA concentration of 4 µg/L.

Alternative markers. The approaches to im-proving screening performance considered in thepreceding paragraphs do not address the issue that isoften at the heart of the failure of a plasma substanceto serve as a satisfactory screening marker which isthat the would-be marker does not have a highdegree of cancer specificity. Markers that are notspecific for cancer cannot perform well in a settingin which there is a high prevalence of benigndisease. In such circumstances it is necessary toidentify a screening marker that is more cancerspecific. Usually this entails finding another markersubstance altogether but, in some cases, a certainform of the original marker can be identified that ismore highly cancer specific. The cancer specificityof the form may arise from its preferential synthesisin cancer cells or from some cancer-specific post-translational modification of the marker. No plasmasubstance has yet been found to be more specific forprostate cancer than PSA.

There is, however, a form of PSA that appearsto be somewhat more cancer specific, protein-boundPSA (Polascik et al. 1999). PSA exists in the circu-lation in four forms: (1) proPSA (Mikolajczyk et al.1997), which is the enzymatically inactive precursorform of PSA, (2) active PSA, (3) nicked PSA,which is the product of the action of nicking enzymeon either proPSA or active PSA, and (4) otherdegradation products of PSA (Hilz et al. 1999,Charrier et al. 1999). In tissue (Jung et al. 2000)and at the time of release into the circulation, all ofthese forms of PSA exist as free species. In the

circulation, active PSA is rapidly inactivated bybinding to the plasma protease inhibitors, α2-macro-globulin and α1-antichymotrypsin. Essentially all ofthe circulating active PSA is inhibitor-bound. Bothforms are cleared from the circulation by the liver.The α2-macroglobulin-bound form is removed veryrapidly (half-life of minutes) while the α1-anti-chymotrypsin-bound form is removed slowly (half-life of days). Consequently, even though active PSAis more avidly bound by α2-macroglobulin than byα1-antichymotrypsin, the α1-antichymotrypsin-boundform is the predominant plasma form. Nicked PSAbinds very slowly to α2-macroglobulin and not at allto α1-antichymotrypsin. As a result, it is present inplasma predominantly in the free form. The otherdegradation products, as well as proPSA, are presententirely in the free state.

It has been found that the percentage of PSApresent in the free state in plasma is lower inprostate cancer than in BPH. This finding meansthat, in prostate cancer, a greater fraction of thecirculating PSA is present in the α1-antichymo-trypsin-bound form. One way that this could comeabout is if the PSA entering the circulation fromcancerous tissue were enriched in the active form.Then, of that not rapidly eliminated by binding toα2-macroglobulin, most would bind to α1-anti-chymotrypsin. Alternatively, the release of humankallikrein 2, the prostatic enzyme that activatesproPSA (Rittenhouse et al. 1998), could be greaterfrom cancerous tissue than from BPH, resulting in agreater fraction of the PSA being activated andending up bound to α1-antichymotrypsin. It iscurrently not known if these or other mechanismsexplain the clinical findings (Stenman et al. 1999).

Screening for a genetic predisposition tocancer. The development and evolution of cancerappear to result from an accumulation of geneticlesions within the cell of origin (Kinzler and Vogel-stein 1996). These lesions include gain-of-functionmutations of oncogenes, loss-of-function mutationsof tumor-suppressor genes, and loss-of-functionmutations of DNA repair genes (Lynch et al. 1997).When a lesion with oncogenic potential is present asa germline mutation, the total number of somaticgenetic lesions that need to occur for the develop-ment of cancer is fewer. Consequently, there is anincreased probability that the individual bearing themutation will develop cancer. This predispositionmay be predominantly to cancer of a particular organor it may be to cancers of several different organs.

Cancer 11-12

Laboratory screening for germline mutations thatpredispose to cancer is an appealing idea but screen-ing for a mutation is only of practical value whenfour conditions are met: (1) the mutation must conferan appreciable risk of cancer, (2) screened individu-als must have a significant probability of harboringthe mutation, (3) the screening laboratory study musthave acceptable diagnostic performance, and (4)there must be something that can done to substan-tially reduce the risk of developing invasive cancerin individuals who have the mutation (Ponder 1997).

Unless the risk of cancer is increased to anunacceptable level by the presence of an oncogenicmutation, individuals will not be willing to undergoa test for the mutation, especially if the test is expen-sive. If the baseline risk for a cancer is high, suchas with the most common cancers, a modestmutation-related increase in the risk of the cancermay yield an level of risk that is unacceptable tomost individuals. Thus, for these cancers, evenmutations with relatively minor oncogenic effectsmay be desirable targets for screening. If thebaseline risk for a cancer is low, a mutation needs tohave a major oncogenic effect to be considered forscreening because only a dramatic mutation-relatedincrease in the risk of the cancer will result in aunacceptable level of risk for the cancer.

The prevalence of oncogenic germline mutationsis very low. Screening random individuals for oneof these mutations would therefore be inefficientand, if the screening test did not have perfect speci-ficity, there would be a large number of false-positive results. To minimize these problems,screening studies are performed only on individualswho have an increased likelihood of having themutation as indicated by a family history of the typeor types of cancer associated with the mutation.

Screening for oncogenic mutations relies primar-ily on molecular diagnostic studies. These studiesnormally have excellent specificity. They are, how-ever, usually limited to the detection of well-characterized mutations. Consequently, the diag-nostic sensitivity of a study will be very high amongmembers of families with well-characterizedmutations but will be it will be very low amongindividuals who come from families with novelmutations. The overall sensitivity of a study will bedetermined by the proportion of families who have apredisposition to cancer due to well-characterizedmutations. For those cancers in which the sensitiv-ity of the laboratory screening test is unsatisfactory,

the practical solution is to use the family history asthe screening tool.

Screening for a predisposition to cancer isclearly justified when there is an intervention thatcan be undertaken to lessen or eliminate the risk ofdeveloping the cancer. This may mean something assimple as a change in lifestyle designed to lessen theexposure to a carcinogen implicated in the develop-ment of the cancer but, more often, it means prophy-lactic removal of the organ or organs likely todevelop the cancer. In the absence of interventionoptions, screening can be useful for identifying thoseindividuals who should undergo intensified cancersurveillance. This usually consists of early institu-tion of cancer screening and shorter screening inter-vals to account for the tendency for an early age ofappearance in mutation-related cancers and for theirrapid evolution into invasive and metastatic disease.

Colorectal cancer is an example of a cancer thathas hereditary forms. Two of these are familialadenomatous polyposis and hereditary nonpolyposiscolorectal cancer (HNPCC). HNPCC is responsiblefor approximately 5 percent of the cases of colorec-tal cancer. It can be caused by loss-of-functionmutations in a number of DNA mismatch repairgenes; about half of the cases are caused bymutations in hMLH1 and hMSH2 (Boland 2000,Lynch and de la Chapelle 1999). Mutations in eitherhMLH1 or hMSH2 result in an 80 percent probabil-ity of developing colorectal cancer and, in females, a40 to 60 percent probability of developing endome-trial cancer (Vasen et al. 1996). Carriers ofmutations in hMSH2 also have an increased risk ofcancer of a number of other organs. There are twoclinical approaches available to reduce the burden ofthis very high probability of developing cancer. Inthe first approach, the risk of developing invasivecancer is lowered by beginning a program of annualcolonoscopic surveillance when an individual at riskfor HNPCC is in his or her twenties (Lynch andLynch 2000). This program addresses the earlierage of onset of the hereditary form of the cancer.The program also accounts for the preference forproximal colon (colonoscopy) and the shorteradenoma-to-carcinoma transition time (annual proce-dures). In the second approach, prophylactic subto-tal colectomy is performed, reducing the risk ofdeveloping cancer by removing most of the tissue atrisk (Lynch and Lynch 2000).

Individuals at increased risk for a predispositionto colorectal cancer are identified by a family history

Cancer 11-13

of colorectal cancer among first-degree relatives.Because of the high rate of cancer developmentamong individuals with HNPCC, kindreds affectedby this disease often have multiple family members,across several generations, who have had nonpolypo-sis colorectal cancer or endometrial cancer.Additionally, a history of onset of the disease before50 years of age is common. These findings havebeen formalized as the Amsterdam criteria for theclinical diagnosis of HNPCC. When the Amsterdamcriteria are met, family members can assume theycome from an affected kindred. The question forthem is if they are carriers of the causative mutation.This question can be answered currently only inthose families with mutations in genes known tocause an appreciable fraction of the cases ofHNPCC. Half of the cases are caused by mutationsin hMLH1 or hMSH2 and laboratory studies areavailable for mutation analysis of these genes. Anindividual who is found to carry a mutation isadvised to enter a program of intensive cancersurveillance. If it is felt to be unlikely that he or shewill be able to comply with the demands of such aprogram, the option of a subtotal colectomy may beoffered. Family members who are found not tocarry a mutation can simply be advised to enroll in aroutine program of screening for colorectal canceronce they reach the age of 50 years. Members offamilies in which HNPCC is caused by othermutations cannot have their individual cancer riskdetermined; they are left knowing only that theyhave a 50:50 chance of carrying a causativemutation. Intensive cancer surveillance is recom-mended for all family members in this circumstance.

When the Amsterdam criteria are not met, thereis a possibility that a familial clustering of cancersresults from a different form of hereditary colorec-tal cancer—one with a lower rate of development ofcancer—or else the clustering could be a chanceaggregation of sporadic tumors. Mutation analysisof hMLH1 or hMSH2 would be informative in thoseindividuals in whom the result was positive, but anegative result would not exclude the possibility ofHNPCC, as half of the kindreds with this disorderhave a different genetic basis, nor would it excludethe possibility of other familial cancer predisposi-tions. Clinical management in this situation isguided by the epidemiologic finding that, in indi-viduals who have two or more first-degree familymembers with colorectal cancer, the risk of develop-ing colorectal cancer is multiplied about 2.75 times

(Fuchs et al. 1994). Because colorectal cancer iscommon, there being a lifetime risk of 0.06 in thegeneral population, even a modest increase in risk issignificant and it is usually felt prudent to offerintensive cancer surveillance as a clinical option.

Individuals with colorectal cancer who have afamily history of colorectal cancer in a first-degreerelative, who have previously had colorectal canceror endometrial cancer, who are younger than 55years, or who have a tumor with characteristicpathologic findings, are candidates for examinationof their tumor tissue for microsatellite instability andmutation analysis (Jass 2000). Mutations of theDNA mismatch repair enzymes that cause HNPCCresult in insertion and deletion mutations at microsat-ellite sequences, referred to as microsatellite insta-bility (Boland 2000). Hypermethylation of thepromoter region of hMLH1, which is not associatedwith a familial predisposition to cancer, also causesmicrosatellite instability but occurs only in elderlypatients (Jass 2000). In cases in which tumors showmicrosatellite instability, histochemical staining ofthe tumor tissue or laboratory mutation analysis ofwhite cells from the patient can be used to identifyindividuals with germline mutations in hMLH1 orhMSH2. Demonstration of such a mutation estab-lishes a diagnosis of HNPCC. Given this finding,first-degree family members should undergo evalua-tion of their carrier status.

DiagnosisBecause of the serious medical and psychological

implications of being diagnosed with cancer, thediagnosis must be established with certainty. Withvery few exceptions, this means that microscopicexamination of the cancerous tissue is required. Themicroscopic examination may take the form of surgi-cal pathologic review of a biopsy specimen, cytopa-thologic examination of cells exfoliated from thelesion, or hematologic evaluation of blood or bonemarrow. Despite the generally high specificity ofmicroscopic examination, it is not usual for thediagnosis of cancer to be based solely on the micro-scopic findings. To achieve even greater diagnosticspecificity, clinical, imaging, and laboratory findingsare also considered. Typically, it is only if there isconsistency in all of the findings that the diagnosis ofcancer is made.

Whenever possible, diagnosis includes classifica-tion of the cancer in terms of its cell type. Classifi-cation is usually quite easy given the location and

Cancer 11-14

microscopic appearance of a tumor but, occa-sionally, it is necessary to perform special studies inorder to make an assignment as to the cell type(Table 11.3). Electron microscopy can be used toextend the resolution of the microscopic examinationand special stains can be employed to identify certaincellular contents. Cell-type specific enzymes can bedetected by enzyme histochemistry and enzymes andother proteins can be demonstrated by immunocyto-chemistry. Characteristic chromosomal abnormali-ties can be detected by microscopic techniques suchas fluorescence in situ hybridization (FISH) or bymolecular biologic analysis, such as Southern blothybridization or polymerase chain reaction (PCR).Chromosomal abnormalities have proven to beespecially useful markers in the classification ofhematologic malignancies (Frizzera et al. 1999).Among solid tumors, the t(11;22) translocation ofEwings sarcoma has been found to be a reliableclassification marker.

MANAGEMENT

The management of a patient with cancerincludes establishing a prognosis, selecting appropri-ate therapy, and clinical monitoring. Markers ofcancer are used in all of these aspects of care.

Prognosis and predictionIn cancer medicine, prognosis can be thought of

as an expression of the probability that the cancerwill progress. This probability is usually expressedas the likelihood of progression within a specificperiod of time, say a certain number of years.Obviously, for many tumors, the prognosis dependsupon whether therapy is undertaken; it is possible

for the prognosis of the untreated cancer to bedismal while the prognosis with treatment may begood. To account for the effects of therapeuticintervention on prognosis, the concept of predictionhas been developed. Prediction is the probabilitythat the cancer will respond to a specified therapy.

Prognosis depends largely upon the type ofcancer and upon the extent and location of spread ofthe cancer at the time of presentation. These andother features of a cancer that relate to the prognosisin individual patients are referred to as prognosticfactors. Prognostic factors that relate to the type oftumor include histopathologic attributes that corre-late with tumor proliferation rate, aggressiveness, ormetastatic potential; genetic alterations that relate tothe stage of malignant evolution of the cancer; andcell and marker substance findings that relate tophenotypic features of the cancer associated withdisease progression. The extent of spread of acancer is assessed from clinical examination, tumorbiopsy or surgical removal, and imaging studies.Prognosis usually worsens in the order: in situcancer, tumor confined to the organ of origin, localspread of tumor beyond the organ of origin, involve-ment of regional lymph nodes, and distant metastaticspread. The prognostic importance of each of thesecategories varies somewhat among cancer types asdoes the prognostic implication of the exact extent ofintra-organ and local spread. As an aid to the clini-cian, staging systems have been devised that list thevarying criteria of tumor spread and order theirimportance. The most popular of these are the TNM(Tumor Node Metastasis) staging systems. A crucialuse of staging designations beyond that of establish-ing a prognosis is to aid in identifying patients inwhom the extent of tumor spread is such that

Cancer 11-15

Table 11.3Cell type-specific Cellular Constituents of Use in the Classification of CancerCellular constituent Technique Example Cancer

Membrane protein immunocytochemistry leukocyte common antigen lymphoma, lymphocytic leaukemia

Secretory granules, electron microscopy premelanosome melanoma granule contents

Cytoskeletal protein electron microscopy desmosomes squamous carcinoma

Cytosolic enzyme enzyme histochemistry chloroacetate esterase acute myelogenous leukemia

Cytosolic protein immunocytochemistry myoglobin rhabdomyosarcoma

Chromosomes FISH t(11;22) Ewings sarcomaPCRSouthern blot hybridization

local-regional therapy (surgery and radiotherapy)may be effective. In prostate cancer, for instance,radical prostatectomy is recommended only forpatients with stage T1 (clinically inapparent tumorneither palpable nor visible by imaging), T2 (tumorconfined within prostate), or T3a (unilateral extra-capsular extension of tumor through the prostatecapsule) cancer that is N0 (no regional node metasta-sis) or NX (regional lymph nodes cannot beassessed) and M0 (no distant metastases) or MX(presence of metastasis cannot be assessed).

Predictive factors are clinical and laboratoryfeatures that allow accurate prediction of theresponse to a therapy; they include cell and markersubstance findings that relate to the molecular targetof therapy or to features of the cancer associatedwith resistance to therapy. Because predictionapplies to a specific therapy, the predictive factorsfor a cancer may vary among the therapies availablefor treatment of the cancer. For example, overex-pression of erbB2 in breast cancer is not a usefulpredictive factor for response to endocrine therapybut it is a predictive factor for response to trastuzu-mab (Yamauchi et al. 2001). The predictive factorsin use for a particular therapy can also be expectedto change over time as our understanding of themolecular and cell biology of cancer allows for thedevelopment of new and more accurate markers. Itis worthwhile noting that although some prognosticfactors have a role only in prognosis and somepredictive factors only in prediction, often the samemarker may serve as a prognostic and a predictivefactor (Hayes et al. 1998).

Prognostic and predictive classification can beeither dichotomous or quantitative. Dichotomousclassification is based upon the use of a criticalvalue, or combination of values (for combinationtesting), that identify the prognostic or predictivecategory in which an individual likely belongs. Ifthe measured value of the prognostic or predictivefactor exceeds the critical value, he or she isassigned to one category and if the measured value isless than the critical value, he or she is assigned toanother category. The probability of belonging tothe assigned category can be calculated using Bayesformula (as discussed in Chapter 5); when there aretwo classification categories,

P[post]=

prevalence $ FCC1

prevalence $ FCC1 + (1 − prevalence)(1 − FCC2)

where prevalence is the prevalence of the assignedcategory and the performance characteristics FCC1

and FCC2 are the fractions of patients correctingclassified in each of the two categories. Dichoto-mous classification provides less information aboutthe individual than quantitative classification, inwhich the probability of an individual belonging to aparticular prognostic or predictive group is calcu-lated using the likelihood ratio associated with themeasured value of the prognostic or predictivefactor; for two classification categories,

P[post] =

prevalence $ likelihood ratio

prevalence $ likelihood ratio + (1 − prevalence)

As discussed in Chapter 3, both of these classifi-cation approaches can be extended to take intoaccount multiple classification categories and combi-nations of factor values. An example of this hasbeen reported by Partin et al. (1997). The authorspresent tables that list the probability of the patho-logic stage of prostate cancer, as determined using aquantitative classification approach, in patients withlocalized disease. Four pathologic stages are consid-ered: organ-confined disease, isolated capsularpenetration, seminal vesicle involvement, and pelviclymph node involvement. The probabilities dependupon three prognostic factors: the plasma PSAconcentration, the TNM stage, and the Gleasonscore (a histologic grading system for assessingaggressiveness of prostate cancer). These sameauthors have also generated a nomogram (Figure11.8) based on the same three prognostic factors thatcan be used to calculate the probability of recurrenceof cancer within five years following radicalprostatectomy (Kattan et al. 1998).

Ex vivo drug sensitivity testing. The predictionof response to therapy based on plasma markers andtumor cell predictive factors can be very useful inplanning chemotherapy for a patient but, until all ofthe cellular factors that confer susceptibility to anyparticular drug are known, there will remain uncer-tainty in the prediction. Directly testing the suscep-tibility of living tumor cells to drugs—somethingsimilar to in vitro antimicrobial susceptibility testing,but for cancer cells—could potentially circumventthis problem and allow for highly reliable individual-ized chemotherapy. Unfortunately, the very processof studying cancer cells while keeping them alive exvivo can lead to alterations in the cells or preferentialsurvival of unrepresentative cells, thereby lessening

Cancer 11-16

the reliability of the result of the study. Additionalproblems with ex vivo studies include variability inthe tumor cell sampling process due to intra-tumoralcancer cell heterogeneity and loss of the stromal andvascular setting that defines the tumor's microenvi-ronment in the patient. Despite these obstacles, anumber of ex vivo drug sensitivity assays have beendeveloped and new ones continue to be developed(Bellamy 1992, Cree and Kurbacher 1997) although,to date, none of the assays has been shown to bereliable enough for clinical use.

MonitoringClinical monitoring of a cancer patient consists

of monitoring the patient’s tumor and monitoring thephysiologic status of the patient. Tumor monitoringmay be undertaken to evaluate the response totherapy, to detect the recurrence of a cancer follow-ing successful therapy, or to assess the progress ofan established tumor. Monitoring is accomplishedprimarily by clinical examination, imaging studies,

and, for some tumor types, by serial measurement ofthe plasma concentration of a marker substance. Inleukemias, monitoring relies on serial counts ofcancer cells in blood. A number of techniques arecurrently being developed for the detection of micro-metastatic carcinoma cells in blood and bone marrowand of cancer cell DNA in plasma (Pantel and vonKnebel Doeberitz 2000). Laboratory studies basedon these techniques will allow much greater analyticsensitivity in the monitoring of tumors.

Conventional treatment has aimed for eradicationof cancer but newer approaches may produce growthcontrol rather than cell death. Therapy directed atsuppressing tumor growth may be expected to stabi-lize the plasma concentration of a tumor marker. Ifthe intent of therapy is to reduce the number ofcancer cells, with the hope of totally eliminating thecancerous clone, the plasma concentration of atumor marker will be expected to decline over time.The rapidity of the decline will be determined by therate of cell loss in the tumor (in response to the

Cancer 11-17

Figure 11.8 Nomogram for calculating the probability of prostate cancer recurrence within 5 years following radicalprostatectomy. Reprinted from Kattan MW, Eastham JA, Stapleton AMF, Wheeler TM, and Scarpino PT. 1998. A preopera-tive nomogram for disease recurrence following radical prostatectomy for prostate cancer. J Natl Cancer Inst 90:766.

therapy) and by the plasma half-life of the marker(Bidart et al. 1999). In the extreme, when a tumorhas been surgically excised in its entirety, the plasmaconcentration of the marker will fall exponentiallyuntil a new steady-state is reached. As a rule-of-thumb, this requires five or more half-lives of themarker. If the marker is highly tumor-specific, theplasma concentration of the marker will fall to zero(that is, below the lower limit of quantification).The same is true for a marker that is highly organ-specific if the entire organ is removed as a surgicalapproach to treatment of a tumor. In both cases, ifthe plasma concentration of the marker remainsmeasurable, it can be inferred that the tumor has notbeen completely excised. These considerationsapply, for instance, to PSA following radicalprostatectomy for prostate cancer. PSA has a plas-ma half-life of two to three days and becomes unde-tectable in the plasma by 21 days (seven to tenhalf-lives) after a successful prostatectomy (Bidart etal. 1999). Persistence of a measurable concentrationof PSA in the plasma indicates the presence of resid-ual tumor. It is important, when evaluating thesuccess of a surgical tumor excision, not to measurethe marker concentration soon after surgery whilethe marker present in the circulation at the time ofsurgery has still not been completely cleared fromthe plasma. Similarly, the plasma concentration of amarker should not be measured too early after treat-ment by chemotherapy when evaluating the anti-tumor effect of the therapy because the markerconcentration will not have had time to decline to itsnew steady-state. It is even possible to see earlyincreases in the plasma concentration of a marker ifthe marker is released into the circulation fromdying cancer cells. It is also possible to see modest,transient increases in marker concentration afterseveral months of therapy, if the chemotherapeuticagents reduce the clearance rate of the marker due toliver dysfunction. This happens with CEA occasion-ally following chemotherapy of colorectal cancer.

Patients may be monitored following therapy inorder to detect recurrence of their cancer if earlydetection of tumor recurrence and institution oftherapy will alter the outcome. This is not unlikeperiodic screening for the appearance of the canceralthough the monitoring program may have a differ-ent schedule and dissimilar critical values from thoseused in screening because of the higher probabilityof disease and because prior therapy can alter theperformance of a plasma marker in detecting

cancerous tissue. In the case of monitoring PSAfollowing radical prostatectomy, the removal of allhyperplastic (and normal) prostate tissue eliminatesthe high background plasma PSA concentrations thatmake the interpretation of PSA concentration in thescreening setting so difficult.

In general, when using a tumor-specific markeror an organ-specific marker following excisionsurgery of the organ, tumor recurrence is indicatedby the return of measurable plasma concentrations ofthe marker. The rapidity with which this happensdepends upon the amount of residual tumor fromwhich the recurrent tumor arises and upon the rate ofgrowth of the tumor. The amount of residual tumoris always hoped to be zero, but it can be as large asthat associated with a marker concentration justbelow the limit of quantification. Because of thepossibility that, if present, the amount of residualtumor is just shy of that needed for detection by themarker, monitoring is begun soon after the comple-tion of therapy, when even a small amount of tumorgrowth will have produced a measurable plasmamarker concentration. This allows for the earliestpossible detection of recurrent disease. The subse-quent monitoring schedule is based on clinicalexperience with the time of appearance of residualtumor following the particular therapy employed. Ifrecurrences tend to appear soon after therapy, inten-sive post-therapy monitoring is appropriate. Ifrecurrences usually arise sporadically for yearsfollowing therapy, monitoring may be less frequentbut will need to continued for many years.

Most markers used in cancer monitoring are nottumor-specific or even organ-specific, so the detec-tion of tumor recurrence requires identifying aplasma marker concentration, or concentrationprofile, that can reliably be distinguished from thebackground marker concentration arising from thenon-tumor or non-organ sources of the marker.There are three approaches that can be used in thissetting: (1) establishment of a marker concentrationthat serves as a critical value above which thediagnosis of tumor recurrence is made, (2) use of asignificant difference rule, or (3) use of a clinicalmonitoring rule such as the 22s rule and the 41s rulewhich can be used to evaluate consecutive monitor-ing measurements when the findings from themeasurements are of concern but do not exceed thesignificance criterion for a concentration difference.Figure 11.9 (identical to Figure 5.2) shows CEAconcentrations in a patient being monitored for

Cancer 11-18

recurrent breast cancer after surgery (data fromWinkel et al. 1982). The plasma concentration ofthe marker was measured monthly and for the first13 months the concentration was fairly constant.This suggests that the concentrations during thatinterval represent the background CEA concentrationin this patient. The CEA concentration at month 14is 3.5 µg/L, which appears to be an increase abovebackground but does not exceed the population-basedcritical value of 7.4 µg/L. The critical value wasdetermined using postoperative CEA concentrationsin patients who, after long-term follow-up, didn'texperience a tumor recurrence. If the intraindividualvariability in CEA concentration in this particularpatient is used as the basis for the calculation of asignificant difference in marker concentration,

significant difference m 2 2varindiv

a value of 0.51 µg/L is found (varindiv is 0.032µg2/L2). This formula for a significant difference isbased on a specificity of 0.95, which is probably notappropriate given the gravity of the diagnosis here; aspecificity of 0.999 is more justified, in which case,

significant difference m 3.3 2varindiv

At this level of specificity, a significant difference inCEA concentration is 0.83 µg/L for this patient.This value is less than the observed difference of 2.1µg/L, so the CEA concentration at month 14 repre-sents a significant increase. Because an increase inCEA concentration can result from conditions affect-ing any tissue that releases CEA into the circulation(for example, pneumonia or hepatitis), it cannot beassumed that the increase seen in month 14 necessar-ily implies tumor recurrence. In this way the

significant difference rule differs from thepopulation-based critical value approach where CEAconcentrations above the critical value specificallyimply tumor recurrence and not disorders of otherCEA-producing tissues. Because tumor recurrenceis one of the possible causes of a significant increasein CEA concentration, clinical evaluation of thepatient may be initiated. The continued increases inCEA concentration seen over the subsequent monthsin this patient are convincing evidence of tumorrecurrence. Indeed, she developed clinical signs oftumor soon after the last monitoring specimen wasobtained.

REFERENCES

Abelev GI and Eraiser TL. 1999. Cellular aspects ofalpha-fetoprotein reexpression in tumors. SeminCancer Biol 9:67.

Bellamy WT. 1992. Prediction of response to drugtherapy of cancer. A review of in vitro assays. Drugs44:690.

Benson MC, Whang IS, Pantuck A, King K, Kaplan SA,Olsson CA, and Cooner WH. 1992. Prostate specificantigen density: a means of distinguishing benignprostatic hypertrophy and prostate cancer. J Urol147:815.

Bidart J-M, Thuillier F, Augereau C, Chalas J, Daver A,Jacob N, Labrousse F, and Voitot H. 1999. Kineticsof serum tumor marker concentrations and usefulnessin clinical monitoring. Clin Chem 45:1695.

Boland CR. 2000. Molecular genetics of hereditarynonpolyposis colorectal cancer. Ann N Y Acad Sci910:50.

Bruel J, Binder K, Block T, and Hartung R. 1992. Effectof digital rectal examination on serum concentration ofprostate-specific antigen. Eur Urol 21:195.

Carroll VA and Binder BR. 1999. The role of the plasmi-nogen activation system in cancer. Semin ThrombHemost 25:183.

Carter HB, Pearson JD, Metter EJ, Brant LJ, Chan DW,Andres R, Fozard JL, and Walsh PC. 1992. Longitu-dinal evaluation of prostate-specific antigen levels inmen with and without prostate disease. JAMA267:2215.

Carter HB, Pearson JD, Waclawiw Z, Metter EJ, ChanDW, Guess HA, and Walsh PC. 1995. Prostate-specific antigen variability in men without prostatecancer: effect of sampling interval on prostate-specificantigen velocity. Urology 45:591.

Catalona WJ, Richie JP, deKernion JB, Ahmann FR,Ratliff TL, Dalkin BL, Kavoussi LR, MacFarlaneMT, and Southwick PC. 1994a. Comparison ofprostate specific antigen concentration versus prostatespecific antigen density in the early detection of

Cancer 11-19

0 5 10 15 20

Time (months)

0

2

4

6

8

10

12C

EA (µ

g/L)

population-basedcritical value

Figure 11.9 Postoperative plasma CEA concentrations in apatient with breast cancer.

prostate cancer: receiver operating characteristiccurves. J Urol 152:2031.

Catalona WJ, Hudson MA, Scardino PT, Richie JP,Ahmann FR, Flanigan RC, deKernion JB, Ratliff TL,Kavoussi LR, Dalkin BL, Waters WB, MacFarlaneMT, and Southwick PC. 1994b. Selection of optimalprostate specific antigen cutoffs for early detection ofprostate cancer: receiver operating characteristiccurves. J Urol 152:2037.

Charrier JP, Tournel C, Michel S, Dalbon P, and JolivetM. 1999. Two-dimensional electrophoresis ofprostate-specific antigen in sera of men with prostatecancer or benign prostate hyperplasia. Electrophoresis10:1075.

Collins GN, Lee RJ, McKelvie GB, Rogers ACN, andHehir M. 1993. Relationship between prostate specificantigen, prostate volume and age in the benignprostate. Br J Urol 71:445.

Cree IA and Kurbacher CM. 1997. Individualizingchemotherapy for solid tumors—is there any alterna-tive? Anticancer Drugs 8:541.

Crotty PL, Staggs RA, Porter PT, Killeen AA, andMcGlennen RC. 1994. Quantitative analysis inmolecular diagnostics. Hum Pathol 25:572.

Deutsch HF. 1991. Chemistry and biology of alpha-fetoprotein. Adv Cancer Res 56:253.

Dropcho EJ. 1998. Neurologic paraneoplastic syndromes.J Neurol Sci 153:264.

Faderl S, Talpaz M, Estrov Z, and Kantatjian HM. 1999.Chronic myelogenous leukemia: biology and therapy.Ann Intern Med 131:207.

Falanga A and Rickles FR. 1999. Pathophysiology of thethrombophilic state in the cancer patient. SeminThromb Hemost 25:173.

Franks L. 1954. Latent carcinoma of the prostate. J PathBact 68:603.

Frizzera G, Wu CD, and Inghirami G. 1999. The useful-ness of immunophenotypic and genotypic studies inthe diagnosis and classification of hematopoietic andlymphoid neoplasms. An update. Am J Clin Pathol111(Suppl 1):S13.

Fuchs CS, Giovannucci EL, Colditz GA, Hunter DJ,Speizer FE, and Willett WC. 1994. A prospectivestudy of family history and the risk of colorectalcancer. N Engl J Med 331:1669.

Giuffrida D and Gharib H. 1998. Current diagnosis andmanagement of medullary thyroid carcinoma. AnnOncol 9:695.

Gordon SG and Mielicki WP. 1997. Cancer procoagulant:a factor X activator, tumor marker and growth factorfrom malignant tissue. Blood Coagul Fibrinolysis8:73.

Hammarström S. 1999. The carcinoembryonic antigen(CEA) family: structures, suggested functions andexpression in normal and malignant tissues. SeminCancer Biol 9:67.

Hanahan D and Weinberg RA. 2000. The hallmarks ofcancer. Cell 100:57.

Hayes DF, Bast RC, Desch CE, Fritsch Jr H, KemenyNE, Jessup JM, Locker GY, Macdonald JS, MennelRG, Norton L, Ravdin P, Taube S, and Winn RJ.1996. Tumor marker utility grading system: a frame-work to evaluate clinical utility of tumor markers. JNatl Cancer Inst 88:1456.

Hayes DF, Trock B, and Harris AL. 1998. Assessing theclinical impact of prognostic factors: when is “statisti-cally significant” clinically useful? Breast Cancer ResTreat 52:305.

Hilz H, Noldus J, Hammerer P, Buck F, Luck M, andHuland H. 1999. Molecular heterogeneity of free PSAin sera of patients with benign and malignant prostatetumors. Eur Urol 36:286.

Jass JR. 2000. Pathology of hereditary nonpolyposiscolorectal cancer. Ann N Y Acad Sci 910:62.

Jung K, Brux B, Lein M, Rudolph B, Kristiansen G,Hauptmann S, Schnorr D, Loening SA, and Sinha P.2000. Molecular forms of prostate-specific antigen inmalignant and benign prostatic tissue: biochemical anddiagnostic implications. Clin Chem 46:47.

Kadmon D, Weinberg AD, Williams RH, Pavlik VN,Cooper P, and Migliore PJ. 1996. Pitfalls in interpret-ing prostate specific antigen velocity. J Urol155:1655.

Kahn SB. 1994. Neoplasia. In Noe DA and Rock, eds.Laboratory medicine. The Selection and Interpretationof Clinical Laboratory Studies. Williams & Wilkins,Baltimore.

Kattan MW, Eastham JA, Stapleton AMF, Wheeler TM,and Scardino PT. 1998. A preoperative nomogram fordisease recurrence following radical prostatectomy forprostate cancer. J Natl Cancer Inst 90:766.

Kearney L. 1999. The impact of the new fish technologieson the cytogenetics of haematological malignancies.Br J Haematol 104:648.

Keren DF. 1999. Procedures for the evaluation ofmonoclonal immunoglobulins. Arch Pathol Lab Med123:126.

Ketter R, Zwergel T, Romanakis K, Unteregger G,Ziegler M, Zang KD, and Wullich B. 1996. Selectiontoward diploid cells in prostatic carcinoma derived cellcultures. Prostate 28:364.

Kinzler KW and Vogelstein B. 1996. Lessons fromhereditary colon cancer. Cell 87:159.

Loda M. 1994. Polymerase chain reaction-based methodsfor the detection of mutations in oncogenes and tumorsuppressor genes. Hum Pathol 25:564.

Lynch HT and de la Chapelle A. 1999. Genetic suscepti-bility to non-polyposis colorectal cancer. J Med Genet36:801.

Lynch HT, Fusaro RM, and Lynch JF. 1997. Cancergenetics in the new era of molecular biology. Ann NYAcad Sci 833:1.

Cancer 11-20

Lynch HT and Lynch JF. 2000. Hereditary nonpolyposiscolorectal cancer. Semin Surg Oncol 18:305.

Meigs JB, Barry MJ, Oesterling JE, and Jacobsen SJ.1996. Interpreting results of prostate-specific antigentesting for early detection of prostate cancer. J GenIntern Med 11:505.

Mikolajczyk SD, Grauer LS, Millar LS, Hill TM, KumarA, Rittenhouse HG, Wolfert RL, and Saedi MS.1997. A precursor form of PSA (pPSA) is a compo-nent of the free PSA in prostate cancer serum.Urology 50:710.

Morgan TO, Jacobsen SJ, McCarthy WF, Jacobson DJ,McLeod DG, and Moul JW. 1996. Age-specific refer-ence ranges for serum prostate-specific antigen inblack men. N Engl J Med 335:304.

Moss SM and Melia J. 1998. Screening for prostatecancer: the current position. Br Med Bull 54:791.

Nollau P and Wagener C. 1997. Methods for detection ofpoint mutations: performance and quality assessment.Clin Chem 43:1114.

Odell WD. 1997. Endocrine/metabolic syndromes ofcancer. Semin Oncol 24:299.

Oesterling JE, Jacobsen SJ, Chute CG, Guess HA,Girman CJ, Panser LA, and Lieber MM. 1993. Serumprostate-specific antigen in a community-basedpopulation of healthy men. JAMA 270:860.

Pantel K and von Knebel Doeberitz M. 2000. Detectionand clinical relevance of micrometastatic cancer cells.Curr Opin Oncol 12:95.

Partin AW, Carter HB, Chan DW, Epstein JI, OesterlingJE, Rock RC, Weber JP, and Walsh PC. 1990.Prostate specific antigen in the staging of localizedprostate cancer: influence of tumor differentiation,tumor volume and benign hyperplasia. J Urol143:747.

Partin AW, Kattan MW, Subong ENP, Walsh PC, WojnoKJ, Oesterling JE, Scardino PT, and Pearson JD.1997. Combination of prostate-specific antigen, clini-cal stage, and Gleason score to predict pathologicalstage of localized prostate cancer. A multi-institutionalupdate. JAMA 277:1445.

Polascik TJ, Oesterling JE, and Partin AW. 1999.Prostate specific antigen: a decade of discovery—whatwe have learned and where we are going. J Urol 162:293.

Ponder B. 1997. Genetic testing for cancer risk. Science278:1050.

Ried T, Knutzen R, Steinbeck R, Blegen H, Schrock E,Heselmeyer K, du Manoir S, and Auer G. 1996.Comparative genomic hybridization reveals a specificpattern of chromosomal gains and losses during thegenesis of colorectal tumors. Genes ChromosomesCancer 15:234.

Rittenhouse HG, Finlay JA, Mikolajczyk SD, and PartinAW. 1998. Human kallikrein 2 (hK2) and prostate-specific antigen (PSA): two closely related, but

distinct, kallikreins in the prostate. Crit Rev Clin LabSci 35:275.

Scarpa A and Achille A. 1997. Molecular techniques inhematopathology. Leuk Lymphoma 26 (Suppl 1):77.

Schmid HP, McNeal JE, and Stamey TA. 1993. Clinicalobservations on the doubling time of prostate cancer.Eur Urol 2:60.

Sell S. 1980. Cancer Markers: Developmental and Diag-nostic Significance. Humana Press, New York.StameyTA, Yang N, Hay AR, McNeal JE, Freiha FS, andRedwine E. Prostate-specific antigen as a serummarker for adenocarcinoma of the prostate. N Engl JMed 317:909.

Simak R, Madersbacher S, Zhang ZF, and Maier U.1993. The impact of ejaculation on serum prostatespecific antigen. J Urol 150:895.

Stamey TA, Yang N, Hay AR, McNeal JE, Freiha F, andRedwine EA. 1987. Prostate-specific antigen as aserum marker for adenocarcinoma of the prostate.New Engl J Med 317:909.

Stenman U-H, Leinonen J, and Zhang W-M. 1995.Standardization of PSA determinations. Scand J ClinLab Invest 55 Suppl 221:45.

Stenman U-H, Leinonen J, Zhang W-M, and Finne P.1999. Prostate-specific antigen. Semin Cancer Biol9:83.

Svetec D and Thompsom IM. 1998. PSA screening-current controversy. Ann Oncol 9:1283.

Tisdale MJ. 1997. Biology of cachexia. J Natl CancerInst 89:1763.

Vasen HFA, Wijnen JT, Menko FH, Kleibeuker JH, TaalBG, Griffioen G, Nagengast FM, Meijers-HeijboerEH, Bertario L, Varesco L, Bisgaard ML, Mohr J,Fodde R, and Khan PM. 1996. Cancer risk in familieswith hereditary nonpolyposis colorectal cancerdiagnosed by mutation analysis. Gastroenterology110:1020.

Werner M, Wilkens L, Aubele M, Nolte M, Zitzels-berger H, and Komminoth P. 1997. Interphase cytoge-netics in pathology: Principles, methods, andapplications of fluorescence in situ hybridization(FISH). Histochem Cell Biol 108:381.

Winkel P, Bentzon MW, Statland BE, Mouridsen H, andSheike ). 1982. Predicting recurrence in patients withbreast cancer based on cumulative laboratory results.A new technique for the application of time seriesanalysis. Clin Chem 28:2057.

Woolf SH and Rothemich SF. 1999. Screening forprostate cancer: the roles of science, policy, andopinion in determining what is best for patients. AnnuRev Med 50:207.

Yamauchi H, Stearns V, and Hayes DF. 2001. When is atumor marker ready for prime time? A case study ofc-erbB-2 as a predictive factor in breast cancer. J ClinOncol 19:2334.

Cancer 11-21