Yes. Carcinoembryonic antigen (CEA) fluctuates during chemotherapy, and it does so for reasons that have little to do with whether the treatment is working. Levels can fall as CEA-producing tumor cells die, rise briefly as dying cells release stored protein, drift with liver inflammation, or jump because a unrelated infection is active. Roughly a quarter to a third of colorectal cancers never produce meaningful CEA at all, so a flat reading in those patients reveals nothing either way. The number matters far less than its direction over several weeks and the clinical picture around it.
The biology behind a moving number
CEA is a glycoprotein normally made during fetal gut development and largely silenced after birth. Colorectal adenocarcinoma cells frequently reactivate the gene, then anchor the protein to their surface and shed it into blood. Serum CEA therefore samples the total shedding burden of CEA-expressing cells plus whatever leaks from normal tissue. The test measures protein concentration, not living tumor mass, and the two diverge whenever shedding rates change faster than cell counts do.
The liver clears circulating CEA, with a half-life measured in the range of a few days to roughly a week depending on the assay and the patient. That clearance window explains why a real treatment response appears as a gradual slope over two to three cycles rather than an immediate drop. It also explains why hepatic dysfunction slows the fall: a liver busy with inflammation or chemotherapy-related injury removes less CEA per unit time, so the same tumor burden reads higher. Kidney function and body composition shift the number modestly as well.
Chemotherapy adds a second dynamic. Cytotoxic agents kill tumor cells, and dying cells can release intracellular and surface-bound CEA before clearance catches up. A transient bump in the days after an infusion is biologically plausible for this reason. Tumor lysis of large burden can produce the same effect, and it is a recognized confounder in CEA interpretation during active treatment. The practical consequence is that a single rise timed to an infusion cycle carries less weight than a rise sustained across cycles.
Assay variation deserves more attention than it usually gets. Different platforms use different antibodies and calibrators, and the same sample can differ by several nanograms per milliliter between methods. Serial monitoring should stay on one platform and one laboratory whenever possible. A change smaller than roughly 30 percent between two draws is close to assay noise for many methods, though the exact analytical variation is method-specific and should be read from the reporting lab's own specification.
What the numbers usually look like
Reference ranges depend on the assay and on smoking status. Many laboratories use an upper limit near 3 ng/mL for nonsmokers and near 5 ng/mL for smokers, so treat those as typical rather than universal. The eligible population is also smaller than most people assume: only a minority of colorectal cancers shed enough CEA to be trackable. In the rest, the marker is uninformative regardless of how often it is drawn.
Interpretation is anchored to the patient's own baseline rather than the population cutoff. A patient whose tumor marker sat at 40 ng/mL before treatment and now reads 12 has a meaningfully different situation from someone who started at 2 and now reads 12. Percentage change from a validated pretreatment value carries more signal than absolute level, and clinicians commonly watch for sustained trends rather than isolated points. A decline of 50 percent or more across early cycles is generally read as favorable, while a sustained rise suggests resistance or progression, subject to confirmation.
| Scenario | Typical CEA direction | What it usually reflects | Confidence |
|---|---|---|---|
| Pretreatment baseline high, falling steadily | Down 30-70% over 2-3 cycles | Response to cytotoxic therapy | Moderate to high |
| Pretreatment baseline high, transient rise after infusion | Brief spike, then resume decline | Dying-cell release, assay timing | Moderate |
| Pretreatment baseline high, sustained rise across cycles | Up progressively | Resistance or progression | Moderate, needs imaging |
| Pretreatment baseline normal, stays normal | Flat | Uninformative marker | Low |
| Rise with fever, cough, or diarrhea | Up, often sharply | Infection or inflammation | Low for cancer |
| Rise with new jaundice or right-upper-quadrant pain | Up | Hepatic involvement or biliary obstruction | Moderate |
The table's "confidence" column is the part most thin explainers omit. CEA is a supportive test, not a standalone verdict, and imaging plus clinical assessment remain the reference standard for response.
The counterintuitive part: a rising CEA can be good news briefly
The most professionally important caveat is that a short-term rise does not equal treatment failure. Dying tumor cells release antigen, so a spike in the first days to two weeks after an infusion can coincide with a tumor that is actually shrinking. The same pattern appears in tumor lysis and in flares of inflammation. A single high reading taken during that window is close to uninterpretable, which is why sampling schedule matters as much as the number itself.
The reverse error is equally common. A falling CEA feels reassuring, but it cannot distinguish a shrinking tumor from reduced shedding by a tumor that has stopped expressing the gene, a phenomenon seen as cancers evolve under treatment. Discordance between a falling marker and enlarging lesions on imaging is well recognized and should prompt re-evaluation, not relief. Neither direction is authoritative in isolation.
Where fluctuations mislead
Several edge cases reliably generate confusing numbers:
Concurrent infection, especially pneumonia or cholangitis, can push CEA and related markers upward independent of tumor activity. Liver inflammation from chemotherapy, viral hepatitis, or alcohol raises readings by slowing clearance and by releasing antigen from injured tissue. Smoking raises baseline CEA modestly and chronically, which shifts every subsequent comparison. Benign biliary obstruction and pancreatitis elevate the marker without malignant progression. A different laboratory or assay platform can produce a step change that looks like a trend. Prior surgery, especially resection with anastomosis, alters baseline anatomy and shedding patterns.
Red flags that warrant prompt clinical contact, not interpretation of a single number: new jaundice or dark urine; fever with rigors; new right-upper-quadrant pain; worsening abdominal distension; unexplained weight loss despite treatment; or a CEA that doubles within a few weeks on repeat testing with the same assay.
Those warnings are not about the number itself. They describe clinical events that happen to move the number and need evaluation on their own terms.
Reading a trend properly: the timeline
Fluctuations are best understood as a time series. The table below shows the typical sequence a clinician watches, and where each reading fits.
| Time point | What is being measured | Common interpretation | Practical note |
|---|---|---|---|
| Pre-treatment | Baseline CEA on a defined assay | Establishes whether the marker is trackable | Without this, later changes are uninterpretable |
| Days 3-14 after first infusion | Early post-treatment value | Transient rise possible from cell death | Rarely actionable alone |
| Before cycle 3 | Trend across two cycles | Sustained 30-50% decline suggests response | Same lab, same platform |
| Before cycle 4-6 | Continued trend | Plateau or rise suggests resistance | Pair with imaging |
| Mid-treatment scan | CEA plus imaging | Discordance requires review | Marker never overrides imaging |
| End of planned regimen | Final trend plus imaging | Confirms overall trajectory | Baseline shift may follow surgery |
| Surveillance after treatment | Serial values every few months | New sustained rise triggers workup | Confirm with repeat and imaging |
The intervals are typical anchors for colorectal treatment monitoring, but they vary by regimen and institution. The key principle is the same everywhere: compare like with like, on schedule, on one platform.
What to actually do with a fluctuating value
Concrete steps, in order:
Confirm the assay and laboratory match the prior draw. Ask the reporting lab directly if the platform changed. Compare percentage change, not raw difference, against the validated pretreatment baseline. A rise from 2 to 3 ng/mL is noise; a rise from 40 to 60 is a trend. Repeat an isolated abnormal value within two to four weeks before drawing conclusions. Single spikes are uninterpretable. Note the timing relative to the last infusion and any recent illness. A reading during a febrile illness should be set aside. Pair every marker trend with imaging and clinical assessment. If the marker and imaging disagree, imaging wins and the case gets reviewed. Do not request extra CEA draws between scheduled assessments. More frequent sampling adds noise and anxiety without adding signal.
The reconciliation point matters. Some sources advise waiting for a 30 percent change, others a 50 percent change, and others simply "two consecutive rises." These thresholds exist for different purposes: 30 percent approximates analytical noise for some assays, while 50 percent is a response threshold borrowed from imaging criteria. A defensible working rule is to treat anything under 30 percent as within noise, a 30-50 percent sustained change as worth a repeat, and a sustained change above 50 percent as worth escalating to the care team with imaging review. No single threshold replaces clinical judgment.
Limits of what a CEA can tell you
CEA cannot localize disease, cannot distinguish viable tumor from scar tissue, and cannot confirm that a treatment is working in a patient whose tumor never produced the marker. It performs poorly as a screening test precisely because benign conditions overlap with malignant ranges. Sensitivity and specificity vary by cutoff and population, and the marker is not validated for diagnosis on its own. Guideline bodies recommend it for monitoring in colorectal cancer, not as a standalone decision tool, and not as a substitute for imaging.
The deeper limitation is conceptual. A blood concentration reflects shedding, clearance, and assay behavior at one moment. A tumor is a changing population of cells with shifting gene expression. Trying to read treatment response from a single protein level is like inferring traffic from one car's speedometer. Trends help, context helps more, and imaging plus clinical assessment remain the reference standard against which the marker is judged.
For patients, the useful takeaway is narrow and practical. Ask whether your tumor ever produced a trackable CEA, keep every draw on the same platform, expect that fluctuations happen, and route decisions through your oncology team rather than a single result. For clinicians, the discipline is the same one that governs any biomarker: interpret the trend, not the point, and never let a number outrank the patient in front of you.