The research team, led by Yongming Wu and colleagues at West China Hospital of Sichuan University, pooled data from thirty studies that had examined the relationship between preoperative serum CEA and hidden nodal spread. The analysis was registered prospectively in PROSPERO, followed PRISMA reporting standards, and searched PubMed, Embase, Web of Science and the Cochrane Library from their inception through 18 March 2026. Rather than relying on a single study’s experience, the investigators used random-effects meta-analysis with restricted maximum likelihood estimation, a statistical framework designed to accommodate the genuine heterogeneity that arises when different hospitals, patient populations and laboratory assays are combined into one evidence base.

The headline numbers are striking. When CEA was treated as a dichotomous variable, meaning patients were classified as either above or below a laboratory cut-off value, the pooled univariate odds ratio for occult lymph node metastasis was 4.82, with a 95 percent confidence interval of 3.66 to 6.36. In plain terms, patients with an elevated preoperative CEA had nearly five times the odds of harboring undetected lymph node involvement compared with those whose levels were normal. When the researchers adjusted for other clinical factors in multivariate models, the association weakened but remained robust, with an adjusted odds ratio of 2.74. The fact that the signal survives statistical adjustment for competing variables is what elevates this finding from a statistical curiosity to something a clinician might actually act upon.

The analysis went further by treating CEA as a continuous variable, asking whether each incremental rise in the marker corresponded to a measurable increase in risk. It did. Every 1-ng/mL increase in preoperative CEA was associated with higher odds of occult nodal metastasis, with a multivariate odds ratio of 1.14 per unit increase and remarkably tight confidence intervals of 1.10 to 1.18. Notably, the heterogeneity statistic for this continuous-variable multivariate analysis was low, at I² of just 10.33 percent, suggesting that this dose-response relationship was unusually consistent across the included studies. The prediction interval, which estimates where the true effect would fall in a new study, remained above the null value, reinforcing confidence that the association is real rather than an artifact of pooling.

Perhaps the most surgically relevant insight came from separating hidden metastases by their anatomical location. Occult disease in N2 nodes, the mediastinal stations that sit centrally in the chest and whose involvement dramatically changes staging and treatment, showed the strongest association with elevated CEA, with a univariate odds ratio of 5.14 and low heterogeneity. Occult N1 disease, involving the hilar and lobar nodes closer to the tumor, was also associated with elevated CEA but with more modest and more variable effect estimates. For thoracic surgeons, this distinction is consequential. N2 involvement discovered after the fact can trigger adjuvant therapy, revision of the surgical plan, or both, and knowing preoperatively which patients carry this risk could reshape how invasive the initial nodal evaluation should be.

Subgroup analyses stratified by clinical stage, tumor size, histological type, CEA cut-off value and country revealed significant positive associations in most strata where enough studies existed to permit pooling. Meta-regression identified histological type as a particularly influential moderator, explaining 51.3 percent of the between-study heterogeneity, with country explaining a further 36.5 percent. The signal was especially pronounced in lung adenocarcinoma, the most common histological subtype of non-small cell lung cancer and one that is known biologically to express CEA more readily than squamous cell carcinoma. Publication year, by contrast, explained none of the heterogeneity, indicating that the association has not strengthened or faded as diagnostic technology has evolved over the decades covered by the included cohorts.

Skeptics of meta-analysis often worry that a single influential study could be propping up the entire result, or that journals have quietly buried studies that found nothing. The authors addressed both concerns directly. Leave-one-out analyses, in which each study is removed in turn and the pooled estimate recalculated, produced odds ratios ranging only from 4.54 to 5.17, a narrow band that speaks to the stability of the primary result. Egger’s regression test, a standard probe for funnel plot asymmetry caused by selective publication, returned a p-value of 0.819, indicating no detectable publication bias. These robustness checks do not guarantee the finding is unassailable, but they place it on considerably firmer footing than many biomarker associations that have entered surgical folklore without equivalent scrutiny.

The clinical context explains why this matters. Preoperative staging of lung cancer currently leans on computed tomography and, increasingly, positron emission tomography combined with CT, yet both modalities have limited sensitivity for nodal involvement, particularly in small tumors where involved nodes may be normal in size and only faintly metabolically active. Current guidelines therefore recommend invasive mediastinal staging for many patients, but uptake varies widely, and even endobronchial ultrasound-guided sampling can miss micrometastatic disease. A cheap, universally available serum test that flags patients at nearly fivefold elevated odds of hidden spread could help prioritize who receives the most thorough invasive staging, who warrants careful intraoperative lymph node dissection, and who might be candidates for neoadjuvant strategies in the future.

There are important caveats, and the authors are candid about them. Every cohort included in the meta-analysis came from East Asian populations, which raises genuine questions about external validity, since baseline CEA levels, smoking patterns, tumor genetics and histological distributions differ across populations. The meta-regression finding that country itself was a significant moderator underscores this concern rather than resolving it. The authors explicitly call for external validation in other populations before the association can be considered universally applicable. They also emphasize that CEA alone is neither sensitive nor specific enough to serve as a standalone staging test; instead, they advocate for combined predictive models that integrate CEA with imaging characteristics, clinical variables and potentially other biomarkers to improve preoperative prediction of nodal involvement.

Even with those limitations, the study represents a meaningful step toward precision in early lung cancer management. CEA is already measured routinely in many thoracic oncology programs, costs pennies compared with PET imaging, and carries none of the risks of invasive staging procedures. The meta-analysis, funded by the Postdoctoral Fellowship Program of the China Postdoctoral Science Foundation and the Postdoctoral Research Fund of West China Hospital, converts a scattered literature into a quantified, graded association with transparent uncertainty estimates, including prediction intervals that acknowledge what a future study might reasonably find. If external validation in Western and other populations confirms the signal, the humble preoperative blood draw could become a standard triage tool, ensuring that the patients whose scans look clean but whose blood tells a different story receive the surgical thoroughness their disease demands.

Subject of Research: Association between preoperative serum CEA and occult lymph node metastasis in clinical stage I non-small cell lung cancer

Article Title: Association between preoperative carcinoembryonic antigen and occult lymph node metastasis in clinical stage I non-small cell lung cancer: a meta-analysis

Article References: Association between preoperative carcinoembryonic antigen and occult lymph node metastasis in clinical stage I non-small cell lung cancer: a meta-analysis. (n.d.). https://doi.org/10.1186/s12885-026-17151-3

Image Credits: AI Generated

DOI: 10.1186/s12885-026-17151-3

Keywords: carcinoembryonic antigen, occult lymph node metastasis, non-small cell lung cancer, meta-analysis, lung adenocarcinoma, biomarker, cancer staging, thoracic surgery, N2 metastasis, predictive marker, PET/CT, serum tumor marker