Serum CEA, CA 19 − 9, and FDG PET/CT SUVmax in relation to histological subtype and overall survival in stage IV lung cancer

This retrospective study investigated the association of serum carcinoembryonic antigen (CEA) and carbohydrate antigen 19 − 9 (CA 19 − 9) levels with 18 F-fluorodeoxyglucose positron emission tomography/computed tomography (FDG PET/CT) findings, histological subtype, and overall survival in patients with stage IV lung cancer. This is an observational, hypothesis-generating analysis; no diagnostic accuracy metrics (sensitivity, specificity, or ROC/AUC) were computed. A total of 217 patients with stage IV lung cancer were included in the analysis. Clinical data collected comprised tumor size, histopathological subtype, maximum standardized uptake value (SUVmax) obtained from FDG PET/CT, and serum CEA and CA 19 − 9 levels. Correlations among these variables and their relationships with survival outcomes were analyzed using IBM SPSS, and graphical visualizations were generated using Python. Statistical significance was established at P < 0.05. The cohort comprised 217 patients (196 males and 21 females; median age, 64 years [IQR, 56–72]), including 79 (36.4%) with adenocarcinoma, 94 (43.3%) with squamous cell carcinoma (SCC), and 44 (20.3%) with small cell lung cancer (SCLC). Median CEA was 3.78 ng/mL (IQR, 2.25–7.25), median CA 19 − 9 was 13.95 U/mL (IQR, 7.09–21.50), and median SUVmax was 14.40 (IQR, 9.30–18.80). Median serum 19 − 9, CEA/CA 19 − 9 ratio, and SUVmax did not differ significantly across histological subtypes (Kruskal–Wallis P = 0.942, 0.983, 0.838, and 0.326, respectively). Tumor size differed significantly across subtypes (Kruskal–Wallis P = 0.024), with the highest median value observed in the SCLC group. Median overall survival did not differ significantly across histological subtypes (Kruskal–Wallis P = 0.341; log-rank P = 0.312) or between sexes (Mann–Whitney U P = 0.942; log-rank P = 0.544). Apart from the mathematically expected correlations between the CEA/CA 19 − 9 ratio and its component biomarkers, Spearman correlation analysis identified only weak unadjusted associations, none of which remained statistically significant after Benjamini–Hochberg FDR correction. To avoid multicollinearity arising from the simultaneous inclusion of 19 − 9, and their arithmetically derived ratio, two separate multivariable Cox proportional hazards regression models were constructed. Older age was the only independent adverse prognostic factor in both models (aHR ≈ 1.03 per year, P = 0.001); serum biomarkers, SUVmax, tumor size, sex, and histological subtype were not independently associated with overall survival. In this retrospective stage IV lung cancer cohort, serum CEA, CA 19 − 9, and PET/CT SUVmax did not differ significantly across histological subtypes and were not independently associated with overall survival after adjustment for the covariates available in the multivariable Cox models. Older age was the only independent adverse prognostic factor identified in both models. These findings should not be interpreted as demonstrating an absence of prognostic value of these parameters in stage IV lung cancer in general, because several established clinical and molecular prognostic factors were unavailable for adjustment. The results are exploratory and hypothesis-generating and require validation in prospectively characterized cohorts.