Tumor Markers: What Cancer Blood Tests Can and Cannot Say

Few lab results carry as much emotional weight as a tumor marker. A single number on a blood draw — a CA-125, a PSA, a CEA — can feel like a verdict on whether you have cancer, whether your treatment is working, or whether the cancer is coming back. And few lab results are as widely misunderstood, by patients and sometimes by the clinicians ordering them.

The honest summary of this entire page fits in three sentences. Tumor markers are genuinely useful for following a cancer that has already been diagnosed — tracking treatment response and watching for recurrence. With a small number of specific exceptions, they are poor screening tests, and drawing a "cancer panel" on a healthy person with no symptoms mostly manufactures anxiety, follow-up procedures, and cost without saving lives. Knowing which situation you are in — monitoring versus screening — is the single most important thing to understand before you read any tumor-marker number.

Table of Contents

  1. What a Tumor Marker Is
  2. Why These Are Not Screening Tests
  3. The Markers One by One
  4. What Tumor Markers Are Genuinely Good At
  5. Reading a Mildly Elevated Result
  6. The New Generation: ctDNA and Liquid Biopsy
  7. When a Panel Makes Sense
  8. Costs and Access
  9. Living With Surveillance
  10. Research Papers and References
  11. Connections

What a Tumor Marker Is

A tumor marker is a substance — almost always a protein or glycoprotein — that can be measured in blood and that tends to rise when certain cancers are present. Cancer cells overproduce these molecules, or the body produces them in response to a tumor, and some of that excess spills into the bloodstream where a lab can quantify it.

The central caveat of the entire field belongs in the first paragraph, not the fine print: almost every tumor marker is also produced by normal tissue and rises in benign, non-cancerous conditions. CA-125 is made by the lining of the abdomen and pelvis, so endometriosis, fibroids, menstruation, and even a first-trimester pregnancy raise it. PSA is made by every prostate, and an enlarged or inflamed prostate makes more of it. CEA rises in smokers. CA 19-9 rises whenever bile ducts are irritated or blocked, cancer or no cancer. Liver regeneration after hepatitis pushes up AFP.

That is why a tumor marker result is never, by itself, a diagnosis of cancer — and, just as important, a normal tumor marker is never proof that cancer is absent. Many early cancers shed little or no marker; some people genetically cannot produce certain markers at all. A marker is one data point that only means something in context: your symptoms, your imaging, your history, and — most powerfully — the trend of the same marker in the same person over time.

Because assay methods differ between manufacturers, reference ranges vary from lab to lab, and results from different labs are not always directly comparable. When a marker is being followed over months or years, clinicians try to keep the testing at the same laboratory on the same assay platform, so that a change in the number reflects a change in you rather than a change in the machine.

Why These Are Not Screening Tests

This is the section that saves people the most grief, so it is worth working through with real numbers rather than hand-waving.

A screening test hunts for disease in people who feel fine. For a screening test to help rather than harm, it needs extraordinary specificity — a very low false-positive rate — because in a healthy population, the disease it is hunting is rare. This is where tumor markers fail, and the arithmetic shows why.

Consider CA-125 as an ovarian-cancer screen. Ovarian cancer in postmenopausal women is, thankfully, uncommon: on the order of 1 case per 2,500 women per year. Meanwhile roughly 1–2% of perfectly healthy women run a CA-125 above the standard 35 U/mL cutoff because of benign gynecologic and abdominal conditions. Now screen 10,000 healthy postmenopausal women. You can expect only about 4 new ovarian cancers in a year — but somewhere between 100 and 200 elevated results. Even if the test caught every single cancer (it does not — early-stage tumors often shed little CA-125), more than 95 of every 100 positive results would be false alarms. Each false alarm buys an ultrasound, often a repeat draw, sometimes a CT, sometimes a surgery — performed on a woman who was healthy when she walked in.

This is not a thought experiment. It has been tested in two of the largest cancer-screening trials ever run:

Read those two results again, because they contain the hardest and most counterintuitive lesson in screening: finding some cancers a little earlier is not automatically the same as saving lives. A marker often rises only when a tumor is already substantial; the lead time gained can be too short to change the outcome, while the false positives generate real, measurable harm in people who never had cancer.

The same logic is why ordering a "tumor marker panel" — CEA, CA 19-9, CA-125, AFP and friends, drawn together on a healthy person as a cancer check — is discouraged by essentially every major oncology and laboratory-medicine guideline, including the ASCO and NACB guidance cited at the bottom of this page. Stack five tests that are each 95–98% specific and the odds that a healthy person clears all of them shrink with every tube of blood. The panel does not meaningfully rule cancer in or out; it mostly generates at least one mildly abnormal number to chase.

The honest exceptions — where blood-marker screening has a real, established role:

The Markers One by One

Reference cutoffs below are typical laboratory values; your lab's printed range governs your result. For every marker, the pattern to internalize is the same: what cancer it follows, what benign conditions raise it, and what its blind spots are.

PSA (prostate-specific antigen) — prostate

The most-ordered tumor marker in the world, and the only one with trial evidence for population screening — with the heavy caveats discussed above. A common cutoff is 4.0 ng/mL, though age-adjusted ranges and PSA velocity matter more than any single threshold. Benign prostatic enlargement, prostatitis, recent ejaculation, cycling, and urinary procedures all raise PSA. After treatment for prostate cancer, PSA becomes a superb monitoring tool — after surgical removal of the prostate it should be undetectable, and any confirmed rise is investigated. Full details on the dedicated PSA Test page.

CEA (carcinoembryonic antigen) — colorectal and others

CEA's established job is monitoring colorectal cancer: a baseline before surgery, then serial measurements (commonly every 3–6 months for about five years) to watch for recurrence, per ASCO guidance. Typical cutoff is about 3 ng/mL — but smokers run higher baselines, and up to roughly 5 ng/mL can be unremarkable in a smoker. CEA also drifts up in benign liver disease, inflammatory bowel disease, pancreatitis, hypothyroidism, and diverticulitis, and it rises in several non-colorectal cancers (lung, breast, stomach, pancreas, medullary thyroid), which is exactly why it cannot localize anything. It is never a screening test. See the site's CEA Test page.

CA-125 — ovarian

Typical cutoff 35 U/mL. Genuinely valuable for monitoring known ovarian cancer through chemotherapy and afterward, and useful as one input (in the ROMA and RMI calculators) when a pelvic mass has already been found. As a standalone screen in healthy women it fails, for the reasons and with the trial evidence given above. Endometriosis, fibroids, menstruation, pregnancy, pelvic inflammatory disease, cirrhosis with ascites, heart failure, and any inflammation of the abdominal lining can raise it; premenopausal women have more of these benign causes, which makes their elevations especially hard to interpret. The dedicated CA-125 page covers it in depth.

CA 19-9 — pancreatic and biliary

Typical cutoff 37 U/mL. Used to follow pancreatic and bile-duct cancers through treatment. Its two famous traps: first, any biliary obstruction raises it — a gallstone blocking a duct can push CA 19-9 into the hundreds with no cancer anywhere; pancreatitis and cholangitis do the same. Second, the Lewis-negative blind spot: roughly 5–10% of people lack the Lewis blood-group enzyme needed to synthesize the CA 19-9 antigen and will show a near-zero CA 19-9 even with advanced pancreatic cancer. A normal CA 19-9 therefore never rules the disease out.

AFP (alpha-fetoprotein) — liver and germ cell

A fetal protein that normally falls to under about 10–20 ng/mL after infancy. It rises in hepatocellular (liver) cancer and in nonseminomatous germ cell tumors of the testis and ovary — and, benignly, in pregnancy and in regenerating liver tissue during hepatitis flares or cirrhosis. In a high-risk liver (cirrhosis, chronic hepatitis B) it earns a real surveillance role alongside ultrasound, as described above; a markedly elevated AFP with characteristic imaging can even establish a liver-cancer diagnosis without biopsy. In testicular cancer, AFP is part of formal staging and is followed after treatment. See the AFP Test page.

Beta-hCG — germ cell tumors and pregnancy

The pregnancy hormone, under 5 mIU/mL in non-pregnant adults. It is a cornerstone marker for testicular and ovarian germ cell tumors and for gestational trophoblastic disease (molar pregnancy, choriocarcinoma), where it tracks disease burden closely enough to guide chemotherapy cycle by cycle. Obviously, pregnancy must always be excluded first in anyone who can be pregnant. A notorious lab artifact deserves mention: "phantom hCG," where interfering (heterophile) antibodies in a person's blood produce a persistent false-positive serum hCG — documented cases exist of patients receiving chemotherapy for a cancer that was never there. A urine hCG (interfering antibodies do not pass into urine) and testing on a different assay platform unmask it. Details on the Beta-hCG page.

CA 15-3 and CA 27.29 — breast

Two assays measuring the same protein (MUC-1); typical cutoffs are around 30 U/mL and 38 U/mL respectively. They can help track metastatic breast cancer during treatment alongside imaging. What they are not for — and this surprises many survivors — is routine surveillance after treatment for early-stage breast cancer: ASCO's long-standing surveillance guidance recommends against routine tumor markers (and routine scans) in asymptomatic early-stage survivors, because detecting a relapse a few months before symptoms appear has not been shown to improve survival or quality of life, while the monitoring itself generates false alarms. Benign elevations occur in liver disease, hypothyroidism, and benign breast conditions.

Chromogranin A — neuroendocrine tumors

A protein stored in neuroendocrine cells, used to follow carcinoid and other neuroendocrine tumors. Its dominant real-world problem is not cancer at all: acid-suppressing drugs. Proton-pump inhibitors (omeprazole, esomeprazole, pantoprazole) reliably drive chromogranin A up — often several-fold — by stimulating the stomach's neuroendocrine cells, and the effect lingers; the drug generally must be held for around two weeks (under medical guidance, since rebound reflux is real) before the level is interpretable. Atrophic gastritis, kidney impairment, and heart failure also raise it. Reference ranges are strongly assay-dependent, so trends on one platform are what count.

Thyroglobulin — thyroid, after surgery

Thyroglobulin is made by all thyroid tissue, so in a person with an intact thyroid it says nothing about cancer. Its power appears after total thyroidectomy (with or without radioactive iodine) for differentiated thyroid cancer: with the gland gone, circulating thyroglobulin should approach zero, and any measurable, rising level points to residual or recurrent thyroid tissue. The critical caveat: about one in four thyroid-cancer patients carries anti-thyroglobulin antibodies, which corrupt the immunoassay (typically falsely lowering it) — so antibodies must be measured with every thyroglobulin level, and in antibody-positive patients the antibody trend itself becomes the surrogate marker. Broader thyroid blood work is covered on the Thyroid Panel page.

LDH and beta-2 microglobulin — lymphoma and myeloma context

Two markers that are not cancer-specific at all, yet formally built into cancer staging. LDH (lactate dehydrogenase) is an enzyme present in nearly every cell; it rises with rapid cell turnover of any cause — a strenuous workout, hemolysis, a heart attack — but in lymphoma it is part of the International Prognostic Index, in testicular cancer it is one of the three official serum staging markers (with AFP and beta-hCG), and in melanoma it informs staging of advanced disease. Beta-2 microglobulin is shed by most nucleated cells and cleared by the kidneys; in multiple myeloma it anchors the International Staging System (below 3.5 mg/L favors stage I, above 5.5 mg/L defines stage III), and it carries prognostic weight in CLL and lymphoma — but kidney impairment alone raises it, which staging systems must and do account for.

What Tumor Markers Are Genuinely Good At

After a section of caveats, the genuine value deserves equal airtime. In a person with a diagnosed cancer that expresses a marker, that marker becomes a cheap, repeatable, nearly painless window into disease burden. Three jobs, in order of strength of evidence:

  1. Tracking treatment response. A CA-125 falling with each chemotherapy cycle, a CEA dropping after colorectal surgery, an hCG collapsing during germ-cell-tumor chemo — these falling curves confirm, between scans, that treatment is working. A marker that fails to fall, or reverses, prompts earlier imaging and an earlier change of plan. In germ cell tumors the markers are so informative that treatment decisions are formally built around them.
  2. Establishing a baseline and prognosis at diagnosis. Pre-treatment marker levels feed directly into staging and risk models — LDH in lymphoma's IPI, beta-2 microglobulin in myeloma's ISS, the AFP/hCG/LDH trio in testicular cancer — and a very high presurgical CEA flags colorectal patients who need closer follow-up.
  3. Recurrence surveillance — with an honest asterisk. Serial markers after curative treatment can detect relapse months before symptoms or scans. Whether that earlier detection improves survival depends on whether an earlier-found relapse can be treated more effectively — and the answer differs by cancer. The colorectal-CEA literature is the fairest test case: in the randomized FACS trial, intensive CEA and CT follow-up roughly tripled the fraction of recurrences caught early enough for potentially curative surgery (about 7% versus 2% with minimal follow-up), yet total deaths were not measurably reduced (Primrose et al., JAMA 2014). Guidelines still recommend CEA surveillance — catching an operable liver recurrence genuinely changes some individual lives — but the honest framing is that surveillance offers a chance at a better outcome, not a guarantee, and for some cancers (early breast cancer, above) marker surveillance has failed the test outright and is not recommended.

One practical rule ties all three jobs together: a marker is only worth measuring when you know what you will do with the answer. Oncologists order markers on a schedule tied to decision points. Extra draws between those points add anxiety without adding decisions.

Reading a Mildly Elevated Result

The most common tumor-marker situation in real life is not dramatic: a number a little above the reference range, often found on a test that arguably should not have been ordered. Before the spiral starts, three facts:

None of this means dismissing results. A markedly elevated marker, a rising trend, or any elevation paired with symptoms (weight loss, bleeding, a mass, new pain) deserves a proper workup without delay. The point is that the workup should be proportionate to the whole picture, not to the fear the word "tumor" on a lab slip produces.

The New Generation: ctDNA and Liquid Biopsy

The classic protein markers above are being joined by something categorically different: tests that detect fragments of DNA shed by tumor cells into the blood (circulating tumor DNA, ctDNA) — the "liquid biopsy." Tumor DNA carries cancer-specific mutations and, crucially, cancer-specific chemical (methylation) patterns that can even hint at which organ the DNA came from. Three distinct uses are at three very different levels of maturity:

The best-known multi-cancer early detection (MCED) test is Galleri, built on methylation analysis of cell-free DNA. The honest scorecard, from its published validation (Liu et al. 2020; Klein et al. 2021, Annals of Oncology): specificity is excellent — about 99.5%, meaning roughly 5 false positives per 1,000 people screened, dramatically better than any protein marker. Overall sensitivity across cancers was about 51%, but heavily stage-dependent: roughly 17% for stage I, 40% for stage II, 77% for stage III, 90% for stage IV. When a signal was detected, the predicted organ of origin was right about 89% of the time. In its first prospective return-of-results study in US adults, around 1% received a "cancer signal detected" result, and roughly 4 in 10 of those turned out to have cancer — a strikingly high positive predictive value for a screening test, and simultaneously a reminder that most positives still trigger a cancer hunt in someone without cancer.

The unresolved questions are exactly the ones the CA-125 story taught us to ask. Stage-I sensitivity is modest, so the test is weakest precisely where early detection matters most. Some cancers it finds early are aggressive ones where lead time may not change outcomes; others may be slow-growing ones that would never have caused harm (overdiagnosis). A negative result may falsely reassure people into skipping proven screening. And no MCED test has yet demonstrated that using it reduces cancer mortality — the only endpoint that ultimately justifies screening healthy people. That is what the NHS-Galleri trial in England — about 140,000 participants, the largest MCED trial ever run — was designed to answer; after an interim look, the NHS opted to await the trial's final results rather than roll the test out, with final data expected in 2026. Meanwhile a related question — using MCED in people who already have symptoms — showed encouraging accuracy in the SYMPLIFY study (Nicholson et al., Lancet Oncology 2023), where the tested population's higher cancer rate flips the predictive math in the test's favor.

Fair summary: liquid biopsy is the most credible attempt yet at a blood test for early cancer — genuinely better engineering than the protein-marker panels it will likely replace — and it is still, today, an unproven screening strategy that a healthy person can reasonably decline while the mortality evidence is pending, or pursue with clear eyes about cost and uncertainty.

When a Panel Makes Sense

Everything above argues against marker panels in healthy, asymptomatic people. There are, however, situations where drawing several markers together is exactly right:

The common thread: a panel earns its place when the person is no longer "general population" — symptoms, findings, or genetics have already raised the stakes — and when each marker on the requisition has a specific question attached to it.

Costs and Access

Individual protein tumor markers are inexpensive as lab tests go: typical US cash prices run roughly $25–$150 per marker, and direct-to-consumer lab services sell most of them without a physician visit. That accessibility cuts both ways — it is precisely how healthy people end up holding an unexplained CA 19-9 of 41 at 11 p.m. with no one to interpret it.

Living With Surveillance

For cancer survivors, tumor markers stop being an abstraction and become a rhythm of life: the draw, the wait, the portal refresh. There is a word for the dread that builds before every surveillance test — "scanxiety" — and it applies to blood draws as much as scans. If the week before each CEA feels like holding your breath, you are not being irrational; you are having the normal response to a genuinely high-stakes number, and virtually everyone in surveillance feels some version of it.

Some things that help, drawn from survivorship practice:

A tumor marker is, in the end, a tool with a narrow and genuine competence: following a known cancer through treatment and afterward. Used there, it is one of oncology's quiet workhorses. Used as a fortune-teller for the healthy, it mostly tells frightening fortunes that do not come true. Knowing the difference is how you make these numbers serve you instead of the other way around.

Research Papers and References

Key peer-reviewed sources for the claims on this page — the major screening trials (PLCO, UKCTOCS, ERSPC), the tumor-marker practice guidelines, the CEA-surveillance randomized trial, the AFP-surveillance meta-analysis, and the primary multi-cancer early-detection (Galleri) literature. Each citation links to the publisher's full record via DOI.

  1. Duffy MJ. Tumor Markers in Clinical Practice: A Review Focusing on Common Solid Cancers. Medical Principles and Practice. 2013;22(1):4–11.
  2. Sturgeon CM, Duffy MJ, Stenman UH, Lilja H, et al. National Academy of Clinical Biochemistry Laboratory Medicine Practice Guidelines for Use of Tumor Markers in Testicular, Prostate, Colorectal, Breast, and Ovarian Cancers. Clinical Chemistry. 2008;54(12):e11–e79.
  3. Locker GY, Hamilton S, Harris J, Jessup JM, et al. ASCO 2006 Update of Recommendations for the Use of Tumor Markers in Gastrointestinal Cancer. Journal of Clinical Oncology. 2006;24(33):5313–5327.
  4. Buys SS, Partridge E, Black A, et al. Effect of Screening on Ovarian Cancer Mortality: The Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Randomized Controlled Trial. JAMA. 2011;305(22):2295–2303.
  5. Jacobs IJ, Menon U, Ryan A, Gentry-Maharaj A, et al. Ovarian Cancer Screening and Mortality in the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS): A Randomised Controlled Trial. The Lancet. 2016;387(10022):945–956.
  6. Menon U, Gentry-Maharaj A, Burnell M, Singh N, et al. Ovarian Cancer Population Screening and Mortality After Long-Term Follow-Up in the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS): A Randomised Controlled Trial. The Lancet. 2021;397(10290):2182–2193.
  7. Schröder FH, Hugosson J, Roobol MJ, Tammela TLJ, et al. Screening and Prostate-Cancer Mortality in a Randomized European Study (ERSPC). New England Journal of Medicine. 2009;360(13):1320–1328.
  8. Primrose JN, Perera R, Gray A, Rose P, et al. Effect of 3 to 5 Years of Scheduled CEA and CT Follow-up to Detect Recurrence of Colorectal Cancer: The FACS Randomized Clinical Trial. JAMA. 2014;311(3):263–270.
  9. Tzartzeva K, Obi J, Rich NE, Parikh ND, et al. Surveillance Imaging and Alpha Fetoprotein for Early Detection of Hepatocellular Carcinoma in Patients With Cirrhosis: A Meta-analysis. Gastroenterology. 2018;154(6):1706–1718.e1.
  10. Liu MC, Oxnard GR, Klein EA, Swanton C, Seiden MV, et al. Sensitive and Specific Multi-Cancer Detection and Localization Using Methylation Signatures in Cell-Free DNA. Annals of Oncology. 2020;31(6):745–759.
  11. Klein EA, Richards D, Cohn A, Tummala M, et al. Clinical Validation of a Targeted Methylation-Based Multi-Cancer Early Detection Test Using an Independent Validation Set. Annals of Oncology. 2021;32(9):1167–1177.
  12. Nicholson BD, Oke J, Virdee PS, Harris DA, et al. Multi-Cancer Early Detection Test in Symptomatic Patients Referred for Cancer Investigation in England and Wales (SYMPLIFY): A Large-Scale, Observational Cohort Study. The Lancet Oncology. 2023;24(7):733–743.

Live PubMed topic searches for areas evolving quickly:

  1. CA 19-9 and Lewis-antigen-negative pancreatic cancer
  2. Multi-cancer early detection and ctDNA screening
  3. Chromogranin A and proton-pump-inhibitor interference
  4. Tumor-marker surveillance in asymptomatic breast cancer survivors

Connections

Back to Table of Contents