Validation of 4-marker protein panel for the early detection of lung cancer using PLCO samples

Background: We have previously demonstrated that a protein panel consisting of ProSFTPB, CEA, CA125 and CYFRA21.1 may improve lung cancer risk assessment and has potential to define eligibility for computed tomography screening. Herein, we aimed to validate the classifier performance of the 4- marker protein panel using pre-diagnostic serum samples from the PLCO cohort. We additionally explored the additive value of diacetylspermine (DAS) with the 4-marker protein panel for identifying lung cancer cases. Methods: ProSFTPB, CEA, CA125 and CYFRA21.1 levels were measured in baseline sera of 537 lung cancer cases (76 SCLC/461 NSCLC) diagnosed within 6 years of baseline blood draw and 3772 cancer-free controls using bead-based immunoassays. DAS was measured using ultrahigh performance liquid chromatography mass spectrometry. Samples were analyzed in a double-blinded randomized fashion. Results: Overall classification performance (receiver operating characteristic area under the curve (ROAUC)) of the 4-marker panel for delineating lung cases diagnosed within 1 year and 1 to 2 years of baseline blood draw from cancer-free controls was 0.78 (95% CI: 0.74-0.82) and 0.73 (95% CI: 0.68-0.78), respectively. Classification performances of the 4-marker panel amongst lung cancer cases diagnosed within 1 year of baseline blood draw stratified into adenocarcinoma, squamous cell carcinoma and small cell lung cancer subtypes yielded ROAUCS of 0.78 (95% CI: 0.72- 0.85), 0.76 (95% CI: 0.69-0.83) and 0.79 (95% CI: 0.68-0.90), respectively. Sub-analyses adjusting for smoking status yielded comparable ROAUC point estimates. Comparison of the 4-marker performance amongst non-NLST and NLST eligible lung cancer patients diagnosed within 1 year of baseline blood draw in comparison to matched cancer-free controls resulted in ROAUCs of 0.71 (95% CI: 0.63-0.79) and 0.74 (95% CI: 0.69-0.80), respectively. Analyses evaluating the additive classifier performance of serum DAS with that of the 4-marker protein panel revealed statistically significant improvement (McNemar Exact Test 2-sided p < 0.05) in sensitivity at high specificity derived from youden index for SCLC and squamous cell carcinoma cases diagnosed within 2 years from baseline blood draw, respectively, in comparison to the 4-marker protein panel alone. Conclusions: We have validated the performance of the 4-marker panel for early detection of lung cancer in the PLCO pre-diagnostic cohort. We further demonstrate that DAS can complement the 4-marker protein panel and identify more SCLC and squamous cell carcinoma cases.