Summary

Tumour markers are mainly measured in clinical laboratories. A distinction should be made between tumour markers and tumour-specific markers. Utmost caution should be exercised in the interpretation of tumour-marker results because of analytical interference and overlap with benign conditions. The use of a tumour marker for early detection of malignancy is challenging because of its low positive predictive value. However, with an understanding of the performance characteristics of tumour markers and correct interpretation of the results, tumour markers are useful in clinical management of diseases.

摘要

腫瘤指標的檢測常被濫用作篩檢潛在的癌症。事實上腫瘤指標如同許多的臨床檢驗一樣有其限制，在某些非癌病的狀況下也可以升高，而正常並不代表就一定沒有癌症。腫瘤指標的最主要功能是用來評估癌症的治療成果和追縱是否復發，只有少數的腫瘤指標在特殊狀況下可用以輔助診斷。

---

Introduction

A tumour marker is a biochemical substance occurring in blood, urine or body tissues which are usually present at very low levels in healthy individuals but may reach high levels in the presence of malignancy. Elevation of these substances in malignancy is usually derived from tissues normally producing that marker or from embryogenically closely related tissues. The potential uses of tumour markers include screening for and diagnosis of malignancy, determining prognosis, monitoring response to therapy and detecting recurrence.  

Many commercial medical laboratories provide different combinations of serum tumour-marker tests for cancer surveillance. This practice by the commercial medical laboratories may lead to the misconception that serum tumour markers can be used reliably to screen for and diagnose malignancies. In fact, this liberal practice is not substantially supported by scientific evidence.¹

An ideal tumour marker should be (i) tested positive only when malignancy is present; (ii) tested positive early in the development of malignancy; (iii) tumour specific; and (iv) able to reflect the extent of disease. Unfortunately, no existing tumour marker has met this ideal. This fact is due to the lack of desired specificity and sensitivity of tumour markers in general. Sensitivity is the ability of a test to correctly label an individual as having the disease. Most tumour markers have poor sensitivity because they become elevated in advanced malignancy only; and such test is unable to detect asymptomatic disease in the early stage. Specificity is the ability of a test to correctly classify an individual as free of the particular disease when the result is negative. Most tumour markers are non-specific for a single cancer; they are found in different tumours of the same tissue type. In fact, even tumour markers that are considered highly sensitive and specific may have low predictive values, given the low prevalence of the cancer in general.² For a cancer with prevalence of 50 per 100,000, a screening test would require a sensitivity of at least 90% and a specificity of more than 99.6% to achieve a positive predictive value of 10% (the minimum positive predictive value set by most epidemiologists to support a screening test).³ Most tumour markers should, therefore, not be used as a “cancer screening tool”. In the United States, the only tumour marker which has been approved for screening is the prostate specific antigen for prostate cancer. It is important to understand the sensitivity, specificity, clinical indication, limitation and potential pitfall of each commonly used tumour markers. This article describes nine tumour markers that are commonly encountered in general practice by family doctors, drawing attention to the merits and limitations of each test.

Alpha-foetoprotein

Alpha-foetoprotein (AFP), structurally related to albumin, is the major serum protein produced during foetal life and its level regresses after birth. Higher serum levels are found in infancy and pregnancy. The primary malignancies associated with AFP elevations are hepatocellular carcinoma (HCC), hepatoblastoma and nonseminomatous germ cell tumour. HCC is one of the most common cancers in Hong Kong. About 8% of our population are chronic carriers of the hepatitis B virus (HBV) and they are at risk of developing cirrhosis and HCC.⁴ Surveillance of HCC in HBV carriers remain controversial.⁵ Although there is still no strong evidence that screening for HCC improves survival, many physicians advocate screening patients with either serum AFP or ultrasonography of the liver, or both.⁶ The Asia Pacific Guideline recommends surveillance for high-risk HBV-infected persons (males, those aged >40 year old, cirrhotics, and/or having a positive family history of serious liver disease) using ultrasonography and serum alpha-foetoprotein, every three to six months.⁷ Screening the high risk groups would ensure that both acceptable assay sensitivity and cost effectiveness of the AFP test are achieved. Apart from screening, measurements of the serum AFP levels are useful in the evaluation of patients who present with liver mass. Nevertheless, it is important to note that 20% of patients with HCC have normal AFP levels so that a negative test result is not totally reliable.⁸ Serial measurements of serum AFP are also used for monitoring cancer treatment progress and detecting recurrence.

Case 1

A 30-year-old HBV carrier presented with jaundice and deranged liver function (alanine transaminase 150 IU/L; alkaline phosphatase 100 IU/L; total bilirubin 50 µmol/L). The serum AFP level was 30µg/L (reference value <10µg /L).

Interpretative comment:

In this case, the patient had chronic viral hepatitis B infection. Apart from malignancy, mildly to moderately elevated AFP levels (10-500µg /L) may also be seen in acute and chronic hepatitis, cirrhosis, biliary tract obstruction and alcoholic liver disease. A raised level at one point in time should therefore not be relied on its own to diagnose cancer. Changes in levels over time should clinically be more useful.⁹ A sustained rise in AFP is suggestive of HCC, while stable or decreasing results make it less likely. To distinguish HCC from other causes of elevated AFP, research has focused on a tumour-specific subtype of the alpha-foetoprotein which is detectable via lectin binding and isoelectric focusing. The percentage of this lectin-binding AFP is increased in HCC with high specificity.¹⁰

Case 2

A two-month-old baby girl presented to a private paediatrician for a routine postnatal check and was found to have an abdominal mass. Computed tomography scan showed the presence of an intrinsic mass within the left lobe of the liver. Hepatoblastoma, an uncommon malignant liver neoplasm occurring in infants and children, was suspected. She was admitted to hospital for further management. Initial laboratory testing showed AFP level of 1,600µg /L (age-related reference range: 16–2,000µg /L). As hepatoblastoma is expected to have a much higher AFP level, the apparently normal level was not compatible with the clinical picture. Therefore, the laboratory re-analyzed the original sample with 100-fold dilutions and revealed the true AFP level of 1,000,000 µg/L.

Interpretative comment:

The initial AFP result was “normal” within the age-related reference range for infants. AFP is higher in infants than in adults. This apparently normal AFP level was indeed an analytical artifact caused by high dose “hook effect”. Hook effects in immunoassays have been recognised as a major pitfall for many years and occur mostly with sandwich immunoassay methods. Sandwich immunoassay is a method using two antibodies, which bind to different sites on the antigen. A “sandwich” is formed by the capture antibody, the antigen being measured in between, and the signal antibody. After a washing step, only the sandwich complexes remain, and the signal antibody is directly proportional to the antigen concentration. If an extremely high antigen concentration is present, the capture and signal antibodies are separately saturated with the antigen, thus preventing a sandwich formation. Thus, the antigen concentration measured is falsely low. Although less common nowadays with assay reformulation by manufacturers, the problem cannot be totally eliminated and still occurs from time to time. The phenomenon most commonly occurs in tumour-marker assays–such as AFP, prostate specific antigen, human chorionic gonadotropin where the pathological range of possible analyte concentration spans several orders of magnitude. An undetected hook effect can lead to false negative result and delay recognition of a tumour. High clinical suspicion is required when the laboratory data is not compatible with clinical presentations. Good communication between clinician and the laboratory is therefore important so that samples with clinically incompatible tumour-marker concentrations can be identified and re-analysed.¹¹

Carcinoembryonic antigen

Carcinoembryonic antigen (CEA) acquired its name because it was first identified in the foetal colon and in colorectal cancer. Besides colorectal cancer, it can be seen elevated in adenocarcinoma originating from other sites including gastric, pancreatic, breast and ovary. Elevated CEA levels are also found in benign conditions including cirrhosis, inflammatory bowel disease, pancreatitis, bronchitis, emphysema and renal disease. Levels may be elevated in healthy smokers.¹²

CEA is a frequently used test but is disappointing. The marker does not have a role for screening of colorectal cancer because its sensitivity and specificity are not high enough, particularly for the early stages of disease. So, in populations with low prevalence of colorectal cancer, there are many false-positive and false-negative results. 

CEA is of better use in determining prognosis and detecting relapse in patients after intended curative treatment of the colorectal cancer. Studies have shown that CEA levels can start to rise, up to 10 months ahead of clinical features of disease.¹³ Earlier detection of recurring disease may have survival benefits. 

Case 3

A 54-year-old man had metastatic colorectal adenocarcinoma and received combination chemotherapy. The pre-treatment CEA value was 500 µg/L. During follow-up, a further rise in CEA level (700 µg/L) was observed despite partial tumour response verified by abdominal computed tomography scan.

Interpretative comment:

An initial rise in CEA level during effective chemotherapy in colorectal cancer patients may not always indicate progression of disease but may be a transient CEA surge in patients responding to chemotherapy. Chemotherapy-induced CEA surge is found in 10% of patients with metastatic colorectal cancer. The rise can persist for 2 months before it recedes to a level below the baseline.¹⁴ An early rise in tumour markers after initiation of treatment should, therefore, be serially followed up to differentiate a true from a transient rise.

Human chorionic gonadotropin (hCG)

Human chorionic gonadotropin (hCG) is a heterodimeric glycoprotein secreted into the circulation by trophoblastic cells. Gestational trophoblastic disease (GTD) is a rare complication of pregnancy associated with raised hCG levels. Serum hCG levels can be used to screen for GTD in patients with molar pregnancy as well as to monitor response to therapy. Serum hCG and/or AFP levels are also elevated in germ cell tumours and are useful in determining prognosis at the time of diagnosis and also for long term monitoring. On the other hand, persistent raised levels of hCG may be found in women in the absence of pregnancy or any evidence of tumour. This can be due to production of hCG by the pituitary gland in peri-menopausal or post-menopausal women.¹⁵ The pituitary hCG production can be excluded by showing that the hCG production is suppressed by a 3-week treatment with high dose progesterone oral contraceptive pills.¹⁶

Case 4

A 25-year-old woman was incidentally found to have a positive hCG pregnancy blood test. However, ultrasound test, dilation and curettage and laparoscopy did not reveal any pregnancy or ectopic pregnancy. The positive hCG persisted and a presumptive diagnosis of gestational trophoblastic disease was made. She was given chemotherapy for assumed malignancy.

Interpretative comment:

This is a “phantom hCG” case. “Phantom hCG” refers to false positive hCG results in patients who were treated for assumed gestational trophoblastic disease with surgery and/or chemotherapy, without subsequent falls in their hCG levels. The false immunoreactivity was usually due to heterophile antibodies in the serum that interfere with the hCG assay.¹⁷ Such heterophile antibody is present due to previous exposure to animals (e.g. mouse, rabbit etc) or monoclonal antibody treatments. There are several ways to detect heterophile antibody interference, –such as repeating the test with another immunoassay method, adding blocking reagent, and detecting positive hCG signals in blood but not in urine because the heterophile antibody does not cross the glomerular membranes into the urine.¹⁸ Chemical Pathologists could be consulted if in doubt.

Prostate specific antigen (PSA)

Prostate specific antigen (PSA) is a tissue specific serine protease present in prostatic epithelial cells and seminal fluid. The largest proportion of PSA in the serum is complexed with a1-antichymotrypsin. Whilst serum total PSA levels are elevated in prostate cancer, there has been controversy surrounding the use of PSA in screening for prostate cancer due to the concern of over-diagnosis of clinically insignificant cancers.¹⁹ Although prostate cancer is unequivocally lethal in some patients, most men die with other conditions rather than of prostate cancer.²⁰ Some believe PSA screening may cause more harm than good because it may lead some men to get treated for prostate cancers that would never have caused them problems, and the treatment itself can have significant side effects. There are two randomized, controlled trials designed to determine the effect of screening on prostate cancer mortality. The conclusion of the recently published first report of Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial was that after 7 to 10 years of follow-up, the rate of death from prostate cancer was very low and did not differ significantly between screening group receiving PSA and digital rectal examination and control group receiving usual care.²¹ The other trial, European Randomized Study of Screening for Prostate Cancer (ERSPC), reported that PSA screening without digital rectal examination was associated with an absolute reduction in prostate-cancer mortality of about 7 deaths per 10,000 men after 9 years of follow-up. However, there was no reduction in overall mortality.²²

Besides prostate cancer, elevated serum PSA levels may be found in a number of non-malignant conditions such as benign prostate hyperplasia, prostatitis, trauma, ejaculation etc, giving the test a low specificity. PSA levels also tend to increase with age. (Table 1)²³ The incidence of prostate cancer in Hong Kong is lower than those in most Western countries. Coupled with the fact that specificity of PSA is not very high, PSA screening in Hong Kong is likely to result in a substantial false positive rate. On the other hand, about 20% of men with prostate cancer have serum total PSA levels within the normal range.²⁴ Clinicians should provide information on the potential benefits and possible harms of routine PSA screening to allow patients to make informed choice about PSA testing.²⁵

Serum total PSA levels are useful for the diagnosis of prostate cancer when used in conjunction with the history, clinical examination, and complementary investigations (such as ultrasound and prostatic biopsy). The positive predictive value is 20 to 30 percent for total PSA levels > 4 µg/L and increase to 50 percent for total PSA levels > 10 µg/L.24 In patients with total PSA values between 4 and 10 µg/L, measuring the percentage of free (uncomplexed) PSA relative to the total PSA in men can help to improve the positive predictive value of PSA testing. PSA can also be used to monitor the clinical course of patients diagnosed with prostate cancer. After radical prostatectomy, the serum level falls with a half-life of approximately 2.5 days but can be many months following radiotherapy. After the initiation of hormonal therapy, PSA levels decrease over a period of three to four months.²⁶

Case 5

A 55-year-old man was referred for prostate cancer screening. Digital rectal examination was normal and the total PSA level was 6 µg/L. 

Interpretative comment:

The use of % free PSA is recommended as an aid in distinguishing men with prostate cancer from benign prostatic hypertrophy when the total PSA level is within the range of 4-10 µg/L.1 When less than 10% of PSA is free, the positive predictive value for prostate cancer increases to 55% and prostate biopsies should be performed. On the other hand, if the % fPSA is greater than 25%, the probability of prostate cancer is less than 10%.²⁷

Case 6

A 45-year-old man presented with frequent urination, urgency and painful ejaculation. Ultrasound showed prostate enlargement and biochemical correlation was suggested by the radiologist. The serum total PSA level was 30 µg/L.

Interpretative comment:

PSA has been reported to be markedly raised in acute urinary tract infection with levels up to 100 µg/L.²⁸ The PSA levels would return to baseline within eight weeks of symptom resolution. In this case, the patient was diagnosed to have acute prostatitis and treated with antibiotics. The PSA level returned to normal afterwards.

Cancer antigen (CA) 15-3

The cancer antigen (CA) 15-3 is a glycoprotein that is present on the apical surface of normal epithelial cells. It is associated with breast cancers although elevated levels can also be found in other malignancies including colon, ovarian and liver cancers. The degree of rise in CA 15-3 correlates with the disease stage: high levels correlate with metastatic disease and, consequently, poorer prognosis. C0A 15-3 may also be elevated in normal individuals and benign disorders of the breast and liver. The primary application of CA 15-3 is in the monitoring of treatment and recurrence in patients with breast cancer.²⁹

Cancer antigen (CA)-125

The cancer antigen (CA)-125 antigen is expressed in many tissues such as the endometrium, endocervix and peritoneum. It is consistently elevated in ovarian cancer and variably elevated in other intra-abdominal malignancies. A number of non-malignant gynaecologic conditions, such as endometriosis, pelvic inflammatory diseases and benign ovarian cysts, are associated with increased serum CA-125 levels. A raised serum CA-125 level may also be found in non-gynaecologic conditions such as hypothyroidism, poorly controlled diabetes mellitus, chronic liver disease and renal disease. The likelihood that elevated CA-125 levels are associated with malignancy increases in post-menopausal women.³⁰

The main role of CA-125 is its role as a marker for assessing response to cancer treatment. An increase in levels during or after treatment is a strong predictor of future disease progression. While the use of CA-125 in screening asymptomatic women for ovarian cancer is controversial, its measurement in patients with pelvic masses may help to differentiate malignant from benign lesions. It can also be used to screen for ovarian cancer in women with strong family history of ovarian cancers.³¹

Case 7

A 50-year-old woman requested screening for ovarian cancer. She has no family history of ovarian cancer and physical examination was normal.

Interpretative comment:

At present, most expert groups advise against CA-125 being used for screening for ovarian cancer in asymptomatic women.1 The National Cancer Institute of USA has expressed the following view: “There is insufficient evidence to establish that screening for ovarian cancer with serum tumour markers such as CA-125, transvaginal ultrasound or pelvic examinations would result in a decrease in the mortality from ovarian cancer. Based on solid evidence, routine screening for ovarian cancer would result in many diagnostic laparoscopies and laparotomies for each ovarian cancer found.”³²

Several large ovarian cancer screening trials are currently in progress to determine whether screening with CA-125 and transvaginal sonography improves survival and whether it is cost effective.^{33, 34}

Epstein-Barr viral serology

Nasopharyngeal cancer (NPC) has the highest incidence in Southern Chinese population, with an incidence rate ranging from 15 to 50 per 100,000.³⁵ NPC is closely associated with the Epstein-Barr virus (EBV), and elevated titres of IgA antibody to the EBV viral capsid antigen (EBV IgA-VCA) and EBV early antigen (EBV IgA-EA) are conventionally measured in patients with suspected NPC. Although EBV IgA-VCA serology is a sensitive marker of NPC, it has a high false-positive rate and may be elevated in normal individuals. On the other hand, EBV IgA-EA serology remains highly specific for NPC, with a specificity of over 95%, but its low sensitivity precludes its use as a population screening tool.³⁶ Since NPC shows strong familial aggregation, recent studies suggest it is reasonable to screen first-degree relatives of NPC patients with EBV serology to achieve early detection.^{37, 38} In one study, a combined panel of EBV serology and DNA test can increase the predictive value, in the diagnosis of and screening for NPC.³⁹

Case 8

A 34-year-old man presented with epistaxis and cervical lymphadenopathy. The EBV IgA-VCA and IgA-EA serology were both negative.

Interpretative comment:

About 10% of patients with NPC do not have a raised EBV IgA VCA titre.³⁶ It is important not to rely solely on the test to screen high-risk or symptomatic patients. A biopsy of the nasopharynx should be the gold standard for NPC diagnosis.

Other tumor markers

Squamous cell carcinoma antigen (SCCA) is a marker for squamous cell cancers, which can occur in the cervix, head and neck, lung, and skin. Levels of SCCA can be used as an aid to stage the carcinoma and to determine the response to treatment. It is not suitable for screening purposes due to a lack in clinical sensitivity and specificity.⁴⁰

Cancer antigen (CA) 19-9 is elevated primarily in pancreatic and biliary tract cancers but also in other malignancies. Benign conditions such as inflammatory conditions of the hepatobiliary system, biliary obstruction, thyroid diseases and pancreatitis also result in CA 19-9 elevations. It is used mainly for the differential diagnosis of pancreatic cancer.

Case 9

A patient presented with obstructive jaundice due to an impacted stone in the common bile duct with cholangitis, where a CA19-9 level of 61,800 U/mL prompted suspicion of a malignant cause. After treatment, the CA19-9 returned to a normal level. One year post-operatively, neither abdominal ultrasound nor CT-scan showed any sign of intra-abdominal malignancy.

Interpretative comment:

High levels of CA19-9 can be caused by benign obstructive jaundice. CA19-9 has little value as a tumour marker in obstructive jaundice. Elevated values should be repeated after relief of jaundice.⁴¹

Conclusion

Most tumour markers have low sensitivity and specificity for cancer, which limit their ability to be used as sole diagnostic tests. They should not be interpreted alone but in the context of other clinical findings and investigations. Ultimately, advances in cancer genetics and proteomics may lead to the development of tumour markers with higher sensitivity and specificity in the future. 

Key messages

1. Tumour marker is a biochemical substance which is usually present at very low levels in healthy individuals but may reach high levels in the presence of malignancy.
2. The potential uses of tumour markers include screening, diagnosis, determining prognosis, monitoring response to therapy and detecting recurrence.
3. Most tumour markers have poor sensitivity because they become elevated in advanced malignancy only; and such test is unable to detect asymptomatic disease in the early stage.
4. No tumour marker is specific for cancer as it can also be elevated in some non-cancerous conditions.
5. Even tumour markers that are considered highly sensitive and specific may have low predictive values, given the low prevalence of the cancer in general.
6. Most tumour markers should, therefore, not be used as a “cancer screening tool”.

---

---

References