Tumor markers

Protocols for sample collection and processing

Availability of clinical specimens (e.g., serum, plasma, tissue) that have been collected, processed, and stored following validated standard operating procedures (SOP) that take account of potentially confounding factors, including those described below, is essential. ^, The panel should include samples from patients with a variety of malignant and nonmalignant conditions. Those from cancer patients should ideally include samples from all stages of disease and especially from patients with early-stage disease. Specimens from biobanks may be appropriate provided their provenance is well documented.

Analytical/technical validation

Each specific assay that will be used in the clinic must be developed and validated, having regard to the requirements described below (see the section on “Analytical Requirements”). Metrological requirements were established at the Stockholm Conference in 1999 and have recently been updated. ^, These incorporate assessments in which components of biological variation of the marker are taken into account where feasible. Detailed guidance on analytical requirements for tumor markers is available for serum-based immunoassays and immunohistochemical tests. ^,

Clinical validation

Defining the specific context in which the tumor marker is to be used is an essential first step that will identify the population to be used in the clinical validation step (i.e., the type of specimens required from carefully defined groups of subjects), the informative statistics, and the acceptability criteria. ^, Clinical validation should be reported in terms of diagnostic accuracy for tests intended for use in diagnosis, as described in Tables 33.8 and 33.9 . Pitfalls in clinical validation include study bias, overfitting of data, and multiplicities (statistical problems that may occur—for example, with disease subset analysis or the use of several different endpoints). The risk of these can be decreased by strict adherence to a written protocol and reporting of all planned steps and all work completed. The EGTM has proposed a four-phase model for tumor marker monitoring trials that is analogous to those used for the investigation of new drugs. Biomarker kinetics and correlation with tumor burden are assessed in phase I, while in phase II, the ability of the marker to identify, exclude, and/or predict a change in disease status is evaluated. In phase III the effectiveness of intervention based on trends in tumor marker concentration is assessed by measuring patient outcome in randomized trials, while phase IV involves clinical evaluation of long-term effects following incorporation of the tumor marker into routine clinical care. The first two phases are also applicable to tumor markers intended to be used solely for screening or diagnosis.

Demonstration of clinical value

Strong evidence that the test will benefit the patient (e.g., by increasing overall survival or leading to fewer invasive procedures or hospital visits) is now mandatory for introduction of new tests into most health care systems. Ideally this should be high-level evidence from a prospective or retrospective randomized clinical trial or systematic review of the literature.

Achieving regulatory approval

Requirements vary in different countries but in the United States involve either clearance approval by the US Food and Drug Administration (FDA) and evaluation in a Clinical Laboratory Improvement Amendments (CLIA)-certified laboratory or by a laboratory-developed test (LDT) pathway.

Postmarket evaluation

Analytical performance of the test in the routine clinical setting should be subject to continued evaluation, best achieved through implementation of rigorous IQC and PT procedures. Ongoing clinical audit to ensure that the marker retains its clinical value and that it is being used in the setting in which it was originally validated is also essential. When a newly introduced tumor marker replaces an earlier one, it is essential to ensure that the first test is discontinued.

Screening for cancer

Screening for disease differs fundamentally from diagnosis or “case finding” because the aim of screening is to detect unsuspected disease in asymptomatic subjects (see Table 33.3 ). The World Health Organization (WHO) criteria for screening programs originally proposed by Wilson and Jungner in 1968 remain valid. Table 33.10 compares how well these criteria are met by a long-established national screening program for choriocarcinoma in high-risk women and by proposed screening programs for prostate cancer. High sensitivity and high specificity of the test for the target disease are essential. Also fundamental to the success of the choriocarcinoma screening program is the high prevalence of the condition (estimated as 3 to 10%) in the high-risk population being screened, yielding a positive predictive value of approximately 92%. The important effect of prevalence of disease on positive predictive value is further illustrated in Table 33.11 .

The relevant test parameters described in Table 33.8 are depicted schematically in Fig. 33.2 . Diagnostic accuracy is established by studying test performance at different decision (cutoff) concentrations in appropriately selected diseased and disease-free populations. The relationship between these parameters is most conveniently depicted graphically by a receiver operating characteristic (ROC) plot ( Fig. 33.3 ). Changing the decision concentration selected alters both sensitivity and specificity but cannot improve both concurrently. These parameters define the characteristics of the test and, together with the prevalence of the disease in the population studied, yield the positive predictive values ( Box 33.1 ). The low prevalence of the majority of cancers, even in age groups most at risk, together with the low sensitivity and specificity of most tumor markers, generally precludes their use in screening. However, successful population screening for early colorectal cancer using a fecal immunochemical test for hemoglobin (FIT) is routine in some countries. The efficacy of screening patients at high risk of HCC and ovarian cancer is being actively investigated, and there is continuing debate about whether to screen for prostate cancer. The possibility of implementing genetic screening for some conditions is also being considered. These issues are considered in more detail in the organ-specific sections in this chapter.

Diagnosis of cancer

The same limitations of lack of sensitivity and specificity that apply to the use of tumor markers for screening also apply to diagnosis and case finding (see Table 33.3 ). Tumor markers are not helpful for diagnosis in patients with nonspecific symptoms and cannot replace a biopsy for establishing the primary diagnosis of cancer. However, in carefully selected undiagnosed patients who are at high risk of malignancy, raised concentrations of tumor markers may be informative and facilitate diagnosis (e.g., when the patient is unable or unwilling to undergo more invasive testing such as colonoscopy). In general, the higher the serum tumor marker value, the greater the likelihood of malignancy, but conversely it is essential to remember that results within the reference interval never necessarily exclude malignancy. The NACB recommends that requests for panels of tumor marker measurements be actively discouraged, together with inappropriate requests for prostate-specific antigen (PSA) in women or CA125 in men, because these are unlikely to lead to improved patient outcomes.

Assessing prognosis

Prognostic markers predict the likely outcome of disease with respect to risk of relapse or disease progression (see Table 33.3 ). They are generally most useful at the time of initial diagnosis, when a marker of good prognosis is suggestive of prolonged survival and/or the possibility of cure. A marker of poor prognosis indicates an increased probability of early recurrence. Prognostic markers can help identify patients with aggressive tumors that require further treatment (e.g., with systemic adjuvant chemotherapy following surgery) while minimizing the risk of overtreatment of patients with indolent disease that is unlikely to progress. It is important to note that prognostic markers provide a probability estimate of outcome for a heterogeneous population of patients and that no prognostic marker can accurately predict outcome for an individual patient.

Although many tumor markers have been reported to have prognostic significance, with more than 22,000 references relating to cancer prognostic markers in PubMed, fewer than 10 are regularly used in clinical practice. Possible reasons for this include inadequate study design, use of inappropriate statistical tests and overoptimistic reporting, failure of a marker to provide independent prognostic information adding to that from established prognostic factors (e.g., tumor size, grade, and lymph node involvement), and inadequate validation. Most prognostic markers are measured in tissue, but as indicated in Table 33.3 , measurement of AFP, hCG, and LDH following primary surgery in patients with nonseminomatous germ cell tumors (GCT) is essential. These markers provide strong independent prognostic information complementing that of conventional prognostic factors and have been well validated. Additionally, although prognostic markers may be of limited use unless there is a therapy that can change the course of the disease, some patients and caregivers may find prognostic information helpful.

Prediction of treatment response

The heterogeneous nature of cancer means that even cancers of the same histologic type may vary widely in response to a particular treatment, and for most cancers only a subset of patients will respond (see Table 33.3 ). Predictive markers that can identify those who will respond are therefore critical in treatment planning so that treatment can be offered to patients who will benefit and so alternative treatment can be offered to those who will not, also sparing them unnecessary side effects. With the development and use of new and costly molecularly targeted treatments, predictive markers (most of which are measured in tissue rather than serum) are increasingly both a clinical and an economic necessity. Several important predictive markers for breast cancer are already well established in clinical use. Further promising predictive markers are now available for breast cancer, colorectal cancer, glioma, melanoma, and NSCLC, while others are being investigated. (See the sections on Use of markers in specific malignancies.)

Monitoring response during and/or shortly after treatment

While the use of predictive tumor markers is increasingly important, the main application of most serum tumor markers is in monitoring the course of disease. For most diagnosed cancer patients, it is helpful to measure the relevant tumor marker pretreatment in order to provide a baseline for subsequent interpretation, and for some cancers, measurements made during and immediately after treatment are also desirable. In general, a decrease in marker concentrations to within expected population (i.e., “normal”) limits following treatment is a favorable sign. In some cases subnormal or absent post-treatment values should be achieved (e.g., for PSA postprostatectomy or thyroglobulin following thyroid ablation). Measurement of AFP and hCG is mandatory for GCTs prior to surgical excision because the pretreatment concentration is required to calculate the half-life of tumor marker decline following chemotherapy. Similarly, in ovarian cancer, measurement of CA125 and the extent of its decline during primary treatment may provide helpful prognostic information. (See the sections on specific cancers.)

Monitoring disease post-treatment to detect progression or recurrence

Serial monitoring of tumor marker concentrations post-treatment provides early indication of disease recurrence, often months before there are clinical signs and symptoms. Whether early detection is of benefit depends crucially on whether a rising tumor marker concentration will prompt clinical action (e.g., early ultrasound, CT or magnetic resonance imaging [MRI] scanning), whether an alternative treatment is available for the individual patient, and/or whether that treatment can be implemented without scan evidence of progression if none is available. If none of these apply and if the patient is not enrolled in a clinical trial, the patient should be made aware of the potential implications of having the test done, and whether tumor marker monitoring is likely to be of benefit should be seriously considered and discussed. ^,

Specimen timing

Timing of specimens within the day for tumor marker measurement is not usually critical because there is little evidence of diurnal variation for most markers. A pretreatment specimen is always helpful when interpreting subsequent results. In some cases, specific procedures may cause transient release of some tumor markers. Specimens should therefore be collected in advance of some procedures (e.g., CEA before colonoscopy or PSA before prostatic biopsy or digital rectal examination [DRE]) or after a suitable time interval after the procedure (e.g., CA125 following abdominal surgery). Awareness of benign conditions that may cause transient increases in tumor marker concentrations ( Table 33.14 ) and avoidance of inappropriate timing, if possible (e.g., measurement of CA125 in a woman who is menstruating, PSA in a man with an active urinary tract infection, or CA19-9 in a subject with cholestasis), reduce the risk of misinterpreting spuriously raised results that may cause undue distress to the patient and also decrease confidence in laboratory testing.

Specimen handling

Serum or plasma is usually (but not always) equally appropriate for tumor marker measurements, although gel tubes may not be suitable for some assays. Requirements should always be checked in the product information supplied with the reagents and/or from other sources. The common tumor markers are generally stable, but serum or plasma should be separated from the clot and stored at 4 °C (short-term) or below −30 °C as soon as possible, following relevant guidelines where available. For longer-term storage, specimens should be stored at −70 °C. Heat treatment (e.g., to deplete serum complement components or to inactivate human immunodeficiency virus [HIV]) should be avoided, particularly for hCG (which may dissociate at increased temperature to form its free α- and β-subunits) and PSA. Potential influence of transit time on analyte results should be considered when samples are exposed to increased temperatures. Standardized conditions of specimen collection and fixation are crucial for immunohistochemical analyses.

Assay validation

Prior to their introduction into routine clinical practice, both immunoassays and immunohistochemical methods must be validated as described above by defined and well-characterized protocols that meet regulatory guidelines (e.g., FDA approval in the United States and CE marking in Europe). Individual laboratories should verify analytical performance prior to introduction into routine use. Internationally recognized guidelines for the performance of immunohistochemical tests should be adopted where these are available. If appropriate high-quality reference materials are not available, it is essential that methods for immunohistochemistry are described in detail.

Internal quality control

As for any laboratory tests, robust procedures for IQC should be established. Within-run variability less than 5% and between-run variability less than 10% should be readily achievable for automated tumor marker methods. Some newer techniques may perform significantly better than this, although manual and/or research assays may be less precise. Appropriate action should be taken immediately if an assay run fails to meet objective criteria for assay acceptance so no potentially erroneous results are reported. Criteria for acceptance should be predefined and preferably based on logical criteria such as those of Westgard. The number of IQC specimens included per run should allow identification of an unacceptable run with a given probability appropriate to the clinical application. Given the long-term monitoring involved in cancer care, assay stability should be ensured over prolonged periods. Laboratories should have in place procedures and acceptance criteria for assessment of lot-to-lot variation which may adversely affect clinical outcomes.

Quality control (QC) material not provided by the method manufacturer is preferable because kit controls may provide an overly optimistic impression of performance as they are unlikely to be commutable with patient serum. ^, At least one authentic serum matrix control from an independent source should be included in addition to any QC materials provided by the method manufacturer. Negative and low positive controls should be included for all tumor markers and should include concentrations close to important decision points (e.g., 0.1 and 3 or 4 μg/L for PSA; 4 to 7 μg/L for AFP; 5 U/L for hCG). The broad concentration range should be covered, and ideally a high concentration control should occasionally be included to check the accuracy of dilution, whether manual or onboard.

Proficiency testing/external quality assessment

PT specimens should ideally be prepared from authentic patient sera, which for tumor markers may require dilution of high-concentration patient sera into a normal serum base pool. ^, As for IQC, specimens prepared by spiking purified analyte into serum base pools is likely to provide an overly optimistic impression of between-method performance. PT specimens should be commutable with patient specimens to ensure valid between-method comparisons. ^, Concentrations should assess performance over the working range and should include assessment of linearity on dilution, baseline security, and stability of results over time.

It is the responsibility of the PT provider to ensure that specimens are stable in transit. The target values (usually consensus means for heterogeneous analytes such as the tumor markers) should be accurate and stable as demonstrated by assessment of their accuracy (e.g., recovery of known amounts of added analyte), stability (i.e., reproducibility on repeat distribution of the same pool), and linearity on dilution (i.e., by issuing different dilutions of the same patient specimen). Because tumor markers are often monitored over long periods, regular assessment of reproducibility and stability of results over time is highly desirable. Reproducibility of results at low concentration is particularly important for AFP and hCG in GCTs and PSA in patients following prostatectomy because treatment may be instituted solely on the basis of a small increase in tumor marker concentration.

Occasional specimens should ideally be distributed to assess whether interference is observed in different methods (e.g., from biotin, heterophilic and other antibodies, high-dose hooking [see section below on “Clinically Relevant Interferences”]). Evaluation of interpretation and technical results is required for PT of immunohistochemical tests and can also be provided for quantitative tests through interpretative exercises and surveys of practice. These can make a powerful contribution to national audit by highlighting differences in reference intervals, reporting practice, and interpretation of clinical results, particularly when the ethos of the PT scheme is educational rather than regulatory.

Standardization

Major international efforts continue to be directed toward encouraging manufacturers to calibrate their methods accurately in terms of the established relevant International Standard (IS) or Reference Reagents ( Table 33.15 ). While availability of such standards does not guarantee improvement in between-method agreement, a reference material (provided its commutability can be demonstrated) provides a benchmark against which the accuracy of calibration can be assessed. Following adoption of the first IS for PSA by most providers of PSA assays, the mean between-laboratory CVs observed in the UK National External Quality Assessment Service (UK NEQAS) PT scheme for PSA decreased by approximately twofold, from more than 20% in 1995 to approximately 9.5% in 2005. Encouraging use of equimolar methods for PSA (i.e., PSA methods that recognize free and complexed PSA equally well) contributed to this improvement.

Improved understanding of what is being measured in an assay for a heterogeneous tumor marker is critical if improved comparability is to be achieved, as illustrated by the number of hCG-related molecules in Table 33.15 . Some further progress has been made by organizing collaborative workshops to identify the more clinically appropriate antibody specificities, ^, and discussions are in progress about how to improve between-method comparability for the complex CA tumor markers for which no ISs have been established (see Table 33.15 ). It is helpful if manufacturers provide clear information about the specificity of the antibodies used in their methods and data on cross-reactivity that is readily comparable with that of other methods so users are aware of the differences.

Despite these efforts to improve comparability, the molecular heterogeneity of most tumor markers means that results obtained using different methods are not interchangeable, and considerable care is required in the interpretation of serial results obtained in more than one method.

Clinically relevant interferences

Tumor marker measurements are subject to the same interferences as all immunoassays (see Chapter 26 ), but clinical biochemists need to be aware of several that are of particular relevance.

High-dose “hooking.”

Because tumor marker concentrations can range over several orders of magnitude, protocols enabling identification of high-dose “hooking” are essential to minimize the risk of reporting erroneously low results, particularly in patients for whom markers are being measured for the first time. An example is shown in Fig. 33.4 . The risk of “hooking” can be reduced by using solid-phase antibodies of higher binding capacity, by using sequential assays that include a wash step, and by assaying specimens at two dilutions. Methods vary in their vulnerability to this interference.

Postanalytical requirements

Brief clinical information about the source of the suspected or diagnosed malignancy and the treatment stage (e.g., preoperative, postoperative, prechemotherapy) is highly desirable. This information should be recorded both in the laboratory computer and on the laboratory report, which should include cumulative and, if possible, graphical reporting of serial results because trends in tumor marker results are generally more informative than single results. Such reporting facilitates interpretation of results and can also help to identify occasional errors in requesting or in the laboratory (e.g., incorrect sample identification or missampling on an analyzer). Cumulative or graphic display of results also highlights unexpected results (e.g., sudden changes that are out-of-accord with the clinical picture) that require confirmation and further investigation. Brief comments relating to interpretation of the analytical results (e.g., whether or not an increase is likely to be clinically significant) and helpful advice about the frequency of monitoring and the need for confirmatory specimens are also desirable.

Urgent results that may be required for immediate patient management should be identified by the reporting biochemist so as to ensure that they reach the relevant clinician promptly (e.g., by telephone if appropriate). These include tumor marker results that can be used to diagnose advanced disease in critically ill but treatable patients (e.g., AFP in hepatoblastoma, hCG in choriocarcinoma, AFP and hCG in nonseminomatous GCTs, and PSA in men with advanced prostate cancer that may respond to endocrine therapy). The consequences of failure to ensure communication of such results directly to the clinician responsible for the care of the patient can be severe. ^,

In view of the method-related differences discussed above, the method used should be stated on the clinical report so any changes in method are readily identifiable. If there has been a method change, it is highly desirable that the laboratory indicates whether this is likely to influence interpretation of the trend in results. There should be a defined protocol if methods are changed, and the likely effect should be communicated to clinical users prior to the change. Managing the change may necessitate analyzing the previous specimen by the new method or by requesting a further specimen to reestablish the baseline and/or confirm the trend in marker concentrations. If the results are likely to be significantly affected by the change in method, as should be clear from the initial validation, it may be desirable to run old and new methods in parallel for a defined changeover period, an approach that also helps clinicians become accustomed to the new values.

Reference intervals should be derived using an appropriate healthy population and should be specific to the method used (see Chapter 9 ). They are usually most relevant for pretreatment of cancer patients. Subsequently, the patient’s individual baseline results provide the most important reference point for most marker results, and application of reference intervals derived in healthy populations can be misleading. For example, at least 6 weeks postprostatectomy, a confirmed PSA concentration of 2.0 μg/L suggests persistent or progressive disease, although this concentration is well within reference intervals for healthy individuals. Provided baseline marker concentrations are well established in post-treatment of diagnosed patients, sustained increases even within the reference interval or other decision limits may be significant and should be treated as possible relapse, provided the measurement procedure is the same.

Reporting critical increases in tumor marker concentrations, taking into account the analytical performance of the test, biological variation (where possible), and the individual reference intervals, helps to contribute to an earlier diagnosis of relapse. The percentage increase or decrease that constitutes a significant change should be defined and should take account of analytical and biological variation, ^, as well as the expected rate of change in benign and malignant conditions and the time between samples. For tumor markers, differences in the magnitudes of their biological variation contribute significantly to these percentages. ^, A confirmed increase or decrease of ±25% is frequently considered to be of clinical significance, but more work is required in this area.

Laboratories should be able to provide calculated tumor marker half-lives or doubling times for markers for which these are relevant (e.g., AFP and hCG). Half-lives are defined as the time to 50% reduction of circulating tumor marker concentration following complete removal of tumor tissue. Their calculation may be irrelevant if a 50% reduction does not represent a significant change. (See the section on “Germ Cell Tumors” for calculation details.)

Proactive provision of a high-quality tumor marker service helps to encourage good communication between laboratory and clinical staff and is likely to encourage appropriate use of tumor marker tests and early identification of any results that are not in accord with the clinical picture and require investigation. An example of a laboratory report that fulfills many of these requirements is shown in Fig. 33.5 .

Bladder cancer

Approximately 712,000 Americans were living with bladder cancer in 2017, with 81,400 new cases expected to be diagnosed in 2020. Smoking can affect the risk of bladder cancer. The most common symptom of bladder cancer is intermittent hematuria, which is present in 80 to 85% of patients. Presenting symptoms may also include voiding problems or dysuria. Transitional cell carcinomas (TCCs) account for the majority of bladder cancers, but adenocarcinomas, squamous cell carcinomas, and sarcomas also occur. Diagnosis is usually established by cystoscopic evaluation. Cytology of cells shed into the urine is very effective in identifying high-grade bladder cancers but misses many papillary urothelial neoplasms of low malignant potential. The majority of bladder cancer patients are diagnosed with nonmuscle invasive tumors. Primary treatment is complete surgical resection, usually by transurethral resection with or without intravesical treatments with bacille Calmette-Guérin immunotherapy or intravesical chemotherapy. ^, Radiotherapy may also be required. Even when resection is considered to be complete, there is a high risk of recurrence, and 50 to 70% of patients will develop tumor recurrence within 5 years, depending on the stage of disease. Lifelong surveillance is therefore required.

Breast cancer

Breast cancer is the most common cancer in women worldwide, affecting 10 to 12% of women. In symptomatic women, the main presenting features include a lump in the breast, nipple change, or discharge and skin contour changes. More than 270,000 new cases are likely to be diagnosed in the United States in 2020, with approximately 42,000 deaths and nearly 3 million women living with the disease. Worldwide the incidence appears to be increasing, but in some Western countries, mortality rates are declining, This decrease has been attributed to earlier detection by systematic screening with mammography, greater awareness among women of early signs of breast cancer, and the availability of adjuvant treatment for newly diagnosed cases. While 5-year relative survival rates increased from 75.2% in 1975 to 89.7% in 2003, survival rates have not changed significantly since then. For the approximately 60% of breast cancer patients in the United States who have localized disease at diagnosis (i.e., confined to the primary site), the 5-year survival rate is 98.6%, compared with 25.9% for the 6% of patients who have distant metastases at diagnosis.

Primary treatment for localized breast cancer is either breast-conserving surgery and radiation or mastectomy. ^, Following primary treatment, most women with invasive breast cancer receive systemic adjuvant therapy such as chemotherapy, hormone therapy, or immunotherapy, or a combination of these. Depending on estrogen receptor (ER) status and other factors, not all patients with breast cancer require adjuvant treatment.