Tumour markers — role in detection and management of cancer

Molecules that show the presence of cancer or give information about how it will behave, for example its likelihood of progressing or responding to treatment.

An ideal marker would have high positive and negative predictive values; be specific to one tumour and not raised in benign disease; reflect tumour burden; have a short half-life; use a cheap, standardised, automated assay with clear reference limits; and be validated in a large prospective trial. No marker meets all of these.

Screening is limited: poor sensitivity for early disease, poor specificity and the low prevalence of cancer give low positive predictive values.

Faecal occult blood (now faecal immunochemical tests) for colorectal cancer is the best validated: four trials showed a 16% relative reduction in colorectal cancer deaths.

PSA: the mortality benefit is unclear and overdiagnosis is substantial; in the European trial, 1,410 men had to be screened and 48 extra cancers treated to prevent one death. Men should be told the risks and benefits before testing.

CA-125 with transvaginal ultrasound did not reduce ovarian cancer deaths and is not recommended outside trials.

AFP with ultrasound every 6 months in cirrhosis: guidance differs, and some bodies recommend ultrasound alone.

Diagnosis: markers cannot make a primary diagnosis, but can help separate benign from malignant disease. CA-125 above 35 U/mL in a postmenopausal woman with a pelvic mass should prompt referral to a gynaecological oncologist; in cirrhosis, a liver mass over 2 cm with AFP above 200 µg/L and typical imaging suggests hepatocellular carcinoma. Treatment generally needs histological proof, with rare exceptions such as hCG in choriocarcinoma.

Staging and prognosis: LDH, AFP and hCG in testicular germ cell tumours, LDH in lymphoma, and tissue markers (uPA/PAI-1, gene-expression tests) in node-negative breast cancer.

Predicting treatment response: oestrogen and progesterone receptors (hormone therapy) and HER2 (trastuzumab) in breast cancer, KRAS (anti-EGFR antibodies) in colorectal cancer, and EGFR mutation (EGFR kinase inhibitors) in lung cancer.

Surveillance after curative treatment:

CEA after colorectal cancer surgery, every 2–3 months for at least 3 years in stage II–III disease if the patient could have further treatment; a rise of at least 30%, confirmed about a month later, is significant.

AFP and hCG after non-seminomatous germ cell tumours, the closest to an ideal marker.

PSA after radical prostatectomy (a rise to 0.2 µg/L or more is biochemical recurrence) or after radiotherapy (a rise of 2 µg/L or more above the nadir).

CA-125 after ovarian cancer and CA 15-3 after breast cancer remain contested: earlier treatment on a rising CA-125 did not improve survival.

Monitoring treatment of advanced disease: falling levels go with response and rising levels with progression. Expect transient spikes in the first weeks of chemotherapy, and never judge response on markers alone.

Benign causes of raised markers: CEA in smokers, inflammatory bowel disease, hepatitis, cirrhosis and pancreatitis; CA-125 with peritoneal irritation, endometriosis, pelvic inflammatory disease and hepatitis; PSA with prostatitis and benign prostatic hyperplasia; AFP with hepatitis and liver injury. Smoking, alcohol and COPD can also raise markers.

Liver or kidney disease changes clearance; for example, CEA stays high when the liver cannot clear it.

A single value is unreliable: follow trends, using the same assay and laboratory. Space serial tests about 3 months apart, repeating an abnormal result in 2–4 weeks; after surgery, wait 5–6 half-lives before judging the fall. Very high levels can read falsely low (the hook effect), so retest on dilution.

Markers can rise with age without cancer, so interpret raised values with caution. PSA rises with age; age-specific cut-offs have been proposed but are not widely endorsed.

Before a PSA test, explain its possible benefits and harms; and base decisions on trends and the whole clinical picture rather than a single value.