Serum extracellular vesicle-associated SNORD54 adds diagnostic information to CEA and CYFRA21-1 in non-small cell lung cancer: a case-control development study.

BACKGROUND

Blood-based biomarkers that complement established serum markers may improve molecular classification of non-small cell lung cancer (NSCLC). We evaluated whether SNORD54 and SNORD41 measured in a serum extracellular vesicle (EV)-enriched fraction provided incremental discrimination beyond carcinoembryonic antigen (CEA) and cytokeratin 19 fragment 21-1 (CYFRA21-1).

METHODS

This single-center case-control biomarker-development study included serum samples from patients with histologically or cytologically confirmed NSCLC and contemporaneous healthy controls. Supporting expression evidence was assessed in paired lung adenocarcinoma and lung squamous cell carcinoma datasets and in 48 paired formalin-fixed, paraffin-embedded NSCLC specimens. Serum analyses included gene-specific cohorts and a common complete-case cohort with SNORD54, SNORD41, CEA, and CYFRA21-1 measurements. Logistic regression models were compared using paired DeLong tests. Model parsimony, adjustment for age, sex, and smoking, locked Stage I-II and Stage I subgroup analyses, U6-normalization sensitivity analyses, and 2,000-resample bootstrap internal validation were performed.

RESULTS

The complete-case cohort comprised 98 patients with NSCLC and 79 healthy controls. Adding SNORD54 to CEA and CYFRA21-1 increased the AUC from 0.851(95% confidence interval [CI], 0.795-0.907) to 0.951 (95% CI, 0.920-0.982; ΔAUC, 0.100; 95% CI, 0.050-0.150; P<0.001). At the development-derived cutoff, the three-marker model achieved 85.7% sensitivity and 94.9% specificity. SNORD41 did not improve the SNORD54-containing model (ΔAUC, 0.0003; P=0.480; likelihood-ratio P=0.953) and increased both Akaike and Bayesian information criteria. With coefficients and the cutoff fixed from the overall cohort, the three-marker model achieved AUCs of 0.904 (95% CI, 0.843-0.964) for Stage I-II disease and 0.882 (95% CI, 0.804-0.960) for Stage I disease. SNORD54 remained informative after adjustment for age, sex, and smoking. The optimism-corrected AUC was 0.948. U6 Ct values differed between diagnostic groups; associations of both snoRNAs with case status persisted in sensitivity models accounting for U6 Ct.

CONCLUSIONS

SNORD54 measured in a serum EV-enriched fraction provided incremental discrimination beyond CEA and CYFRA21-1 in this NSCLC case-control cohort. The parsimonious SNORD54 + CEA + CYFRA21-1 model retained discrimination in locked early-stage subgroup analyses and showed limited apparent optimism after bootstrap correction. Independent validation in cohorts that include benign pulmonary nodules and other relevant lung diseases, together with standardized EV processing and validated RNA normalization, will define the clinical transportability of this three-marker panel.