A double-blind clinical study using serum samples of patients with histologically confirmed diagnosis of NSCLC was carried out at the First Sechenov University in Moscow, RF. During the clinical study, a group of medical researchers at the Sechenov University did not know the identification of the samples prior to the data analysis. The blinding was performed by third party specialists of the external independent laboratory of the Federal Service for Surveillance for Health Care of the Russian Federation. This research design provides a high level of internal credibility and avoids any bias, randomness or confusion. As a result, the team of independent experts did not produce the identification of the samples until the data had been disclosed.

The study was approved by the local ethics committee of Sechenov First Moscow State Medical University (Sechenov University); protocol №07–17 dated 13.09.2017. The study was conducted in accordance with good clinical practice and the Declaration of Helsinki. All samples from patients included in the study have signed voluntary informed consent to participate in the studies, collecting biological material, clinical data and sharing anonymous data were collected according to approval by the Local Ethics Committee of Sechenov First Moscow State Medical University. A total of 55 patients with histologically confirmed LC and 56 healthy control subjects were included in the analysis. The serum samples were collected at the clinical center of Sechenov University from the study participants after the night fasting and were taken to the clinical laboratory.

Statistical analysis of the results, analysis of the ROC-curves, area under curve (AUC) measurement, calculation of the test’s diagnostic characteristics, such as Sensitivity (Sen), Specificity (Sp) and Test Accuracy (Acc) to identify early stages of NSCLC were conducted using the MedCalc Software. The results were considered statistically significant at p< 0.05. The weighted kappa (k) coefficient demonstrate how much one diagnostic test is consistent with another evaluation method taken for true judgment “Gold Standard” (16), was used to further assess the efficacy of the marker application. Kappa coefficients are being interpreted as indices of the test quality in evidence-based medicine. The exact values of the weighed kappa-coefficient demonstrate a significant difference between two diagnostic methods: an agreement between the two estimation methods is considered bad if 0< k< 0.20 and good if k > 0.81.

All serum samples were taken before treatment. This study included 304 blinded by reputable third-party certification company serum samples from 141 histologically verified NSCLC patients with known TNM histopathological classification and staging performed according to AJCC eighth edition (17), 30 patients with COPD and 133 healthy control subjects that have been tested by two independent expert teams in different clinical laboratories. The study included two groups of samples from different sources. Among the 141 NSCLC samples, 44 (31.2%) were at Stage I, 23 (16.8%) at Stage II a-b and 74 (52%) at Stage III a-c. Blind group 1 consisted of 209 serum specimens, including 100 healthy control subjects, Stage I (N=44), Stage II (N=23), and Stage III a-c (N=42) of NSCLC patients. The majority of serum samples in this group (69%) were from patients with early stages of NSCLC (N=67). Serum samples from healthy control subjects (N=100) in group 1, as well as from patients with NSCLC before treatment (N=62) were obtained by venous puncture and were centrifuged and stored at –85°C until assayed according to the standard approved protocol of the First Sechenov University (“Sechenov University”) Moscow, Russia, for prospective research study. Some samples (N=79) from group 1 were obtained from Bio Specimen Bank Precision For Medicine, Inc. (USA). Another blind group 2 of samples consisted of 95 serum samples, including healthy control samples (N= 33), 32 samples from NSCLC patients (IIa-IIIa-c) and 30 samples from patients with COPD was obtained and described by the Institute of Clinical Chemistry (Frankfurt, Germany). Most COPD and NSCLC patients (80%) were chain smokers, with at least 10 years of smoking history. The average age for healthy volunteers was 53 years old (95% CI: 44–85 years), 68 years old (95% CI: 24-93) for NSCLC patients and 66 years old for COPD patients. All patients were Caucasian. Most NSCLC patients (84%) were identified through LDCT screening and the rest 16% were clinically identified. Tumor marker levels were measured on entering the study, without any pre-analytical processing of serum samples. The following variables were assessed in all the patients: age, sex, race, tumor origin, histological type and tumor stage. Quantitative determination of tumor marker levels was determined by commercial kits ELECSYS СА-125, ELECSYS СА 19-9, ELECSYS CYFRА 21-1, ELECSYS SCC (COBAS, Roche Diagnostics GmbH, Germany, EU)] on multifunctional immunoassay analyzer MODULAR ANALYTICS E170 Elecsys (COBAS, Roche Diagnostics GmbH, Germany, EU) in electrochemiluminescence mode. Quantitative measurement of the CEA, CA 15–3 and NSE levels was carried out using ELISA kits CA15-3-IFA-BEST, CEA-IFA-BEST, NSE-IFA-BEST (“VEKTOR-BEST”, Novosibirsk, RF) by the Tecan Spark reader (Germany, EU) using originally collected serum samples. We considered the upper limits of normality to be 5 ng/ml for CEA, 2.5 ng/ml for CYFRA 21-1, 2 ng/ml for SCC, 37 U/ml for CA19.9, 25 ng/ml for NSE, 35 U/ml for CA-125, 30 U/ml for CA 15-3, and 5000 U/ml for CA-62. The serum level of the CA-62 cancer antigen was measured using the chemiluminescent set of reagents CLIA-CA-62 (JVS Diagnostics LLC, RF) on the Tecan Spark reader (Germany, EU) using thermally treated serum aliquots remaining from other cancer marker measurements. All serum samples from groups 1 and 2 were centrifuged (1300 g, 10 minutes) in SST vacutainer tubes with clot activator and separating gel. After that, serum samples were frozen at -86°C before use. Following the measurements for other cancer markers, for CA-62 level’s evaluation all tested serum samples were inactivated by thermal treatment (+56°C, for 30 minutes) using standard protocols for blood draw (18). Serum samples were collected under an IRB-approved protocol from Federal licensed/registered facility following GMPs.

The diagnostic characteristics of various tumor markers CA-62, CEA, CA 15-3, NSE, SA-125, CA 19-9, CYFRA 21–1 and SCC were compared for 304 serum samples of patients, including 141 NSCLC (Stage I-III), 133 samples from healthy volunteers and 30 samples from COPD patients in a blind clinical study. The majority (54.5%) of NSCLC patients had squamous cell carcinoma, 42% had adenocarcinoma and 3.5% had large cell carcinoma. Almost half (48%) of NSCLC patients had early stages of the disease, while 52% had advanced stages of lung cancer. Table 1 shows the descriptive statistics of the studied samples for different NSCLC stages, healthy control subjects and patients withCOPD, median and mean (SD, CI 95%) levels of tumor markers CA-62, CEA, CA 15-3, NSE, CA-125, CA-19-9, CYFRA 21-1, and SCC.

Tumor markers expression levels obtained for CEA (4.2 - 5.1 ng/mL), CA 15-3 (17–27 U/mL), CA-125 (21–28 U/mL), CA 19-9 (12–37 U/ml), NSE (8–22 ng/mL), CYFRA 21-1 (1.9 – 4.35 ng/mL) and SCC (0.9 – 1.7 ng/mL) for NSCLC patients correlated well with literature data for early and advanced stages of non-small cell lung cancer (19–24). Unlike all TM studied in this paper, CA-62 glycoprotein indicated the highest level of expression at Stage I of NSCLC (12,822 U/mL), which remained very high at more advanced stages of cancer: Stage II (8,772 U/mL) and Stage III (10,390 U/mL). Other tumor markers demonstrated lesser sensitivity: CEA (48.2%) > CYFRA 21-1 (44%) > CA-125 (22%) > CA 15-3 (19%) > SCC (18.4%) > CA 19-9 (10%) > NSE (9%).

Using the standard cutoff values recommended by the commercial test kits manufacturers for ECLIA Elecsys CA-125, CA 19-9, CYFRA 21–1 and SCC, ELISAs CA15-3-IFA-BEST, CEA-IFA-BEST, NSE-IFA-BEST and CLIA-CA-62 were calculated their diagnostic parameters for all NSCLC stages (I-III). The results are presented in Table 2 . A ratio of serum samples studied with elevated biomarkers within a group of NSCLC patients with early stage (N=67) and advanced stages (N=74), as well as healthy controls subjects and patients with COPD (N=163) are presented in Figure 1 .

A comparison of the studied tumor markers sensitivity of NSCLC early stages (I and II) detection revealed the following pattern: CA-62 (92%) > CEA (37%) > CYFRA 21-1 (9%) and SCC (9%) > NSE (4.5%) > CA-125 (3%) > CA-15-3 (1.5%) > CA-19-9 (1%). The highest sensitivity of NSCLC early stages detection (92%) was observed for CA-62 marker as compared to CEA, CYFRA 21–1 and SCC tumor markers, which are already used in the diagnosis and monitoring of LC (9, 24). At stages III a-c with the lesion of bifurcation lymph nodes or nodules of the mediastinum, the sensitivity of markers is more comparable to the CA-62 cancer antigen and decreases in the following order: CA-62 (100%)< CYFRA 21-1 (75%)< CEA (55%)< CA-125% (36%)< CA 15-3 (35%)< SCC (28%)< NSE (24%)< CA 19-9 (20%).

Figure 1 shows the percentage of patients (%) in a cohort of early-stage NSCLC (N=67) in a cohort with advanced stages of NSCLC (N=74), in a cohort of healthy control subjects and patients with COPD (N=163). In a cohort of COPD patients, the use of recommended cutoffs for each TM revealed insignificant false positive results ranging from 1.2% for CA 15–3 and NSE markers to 5.5% for CEA ( Figure 1 ).

The majority (83%) of patients with COPD had the median concentration of CA-62 carcinoma-specific marker (2,606 U/ml) and CEA (2.3 ng/ml), which were at the same level as healthy controls. Only 3% of the COPD patient’s cohort (4/30) and healthy control subjects (1/133) showed a slight increase in the CA-62 level at the standard cutoff value of 5000 U/ml at 97% Specificity. For carcinoembryonic antigen CEA only 6% of COPD patients demonstrated an elevated marker’s level using a 5 ng/ml cutoff value at 93.8% specificity. Healthy control subjects (N=133) and COPD patients (N=30) showed a normal median expression of CA-62 cancer antigen (2,667 U/mL and 2,664 U/mL) and CEA antigen (2.2 ng/mL and 2.4 ng/mL). We have not found any difference between samples from Bio Specimen Bank Precision For Medicine, Inc. (USA) and samples collected in the hospitals in Russia: the median, an average of each cohort of patients had similar value for samples collected in the USA and in Russia. At the same time, for all other markers medians studied, the medians of healthy volunteers (N=133) and COPD patients (N=30) were in the same range as the early stages of NSCLC ( Table 1 ).

The present study was designed as a retrospective double-blind format, which implies a predetermined imbalance in disease prevalence compared to the general population. Specifically, the group of patients with verified diagnosis of cancer was significantly over-represented relative to real-world prevalence, while the control group was designed to be of comparable size to ensure adequate statistical power. While the case-to-control ratio does not reflect population prevalence, it is a standard and appropriate design for estimating the essential diagnostic properties (sensitivity, specificity, AUC) of a biomarker, independent of prevalence. We also note that no statistical feature selection or model-based tuning was applied to optimize the marker combination post hoc, which reduces the likelihood of performance inflation due to overfitting.

However, this design does not allow the direct calculation of Positive Predictive Value (PPV) and Negative Predictive Value (NPV), as both are highly dependent on the true prevalence of the disease in the target screening population. Therefore, including PPV and NPV estimates based on the current dataset may therefore mislead readers unfamiliar with the methodological nuances of diagnostic test evaluation. Accurate assessment of these values requires modeling based on population-level prevalence data and would be more appropriate in a prospective study setting.

It is noteworthy to mention that in contrast to the mucins and other tumor markers, which expression is directly proportional to the tumor growth, the marker CA-62 has very high Sensitivity (91-96.4%) at the early stages of NSCLC ( Figure 2 ).

The diagnostic ROC curves were generated according to the histological TNM classification using a graphical application of MedCalc statistical software (Belgium, EU). Comparison of ROC curves for the most prospective tumor markers, such as CEA, CYFRA 21-1, SCC and CA-62 for the entire NSCLC patient’s cohort against all healthy controls and COPD patients ( Figure 3 ) showed a valid difference in the areas under the curve between CA-62 (AUC =0.981) and other TM (area difference AUC = 0.21 - 0.7). The areas under the diagnostic ROC curves were lower and rather expected for other tumor markers: CEA (AUC=0.84) > CYFRA 21-1 (AUC=0.753) > SCC (AUC=0.682). The ability to differentiate NSCLC from a cohort of COPD patients and healthy controls was more significant for CA-62 marker (AUC =0.981) with a diagnostic sensitivity of 95% at 97% specificity ( Figure 3 ). Figure 4 shows the best sensitivity and specificity for detecting early-stage NSCL for CA-62 and CEA markers, with top highest AUC=0.973 for CA-62 and AUC=0.852 for CEA, correspondingly. The CYFRA 21–1 and SCC tumor markers demonstrated insignificant AUC=0.598 and AUC=0.517, respectively.

Using an elevated CA-62 marker cutoff value of 5600 U/ml provides a lowest number of false positives, a high level of specificity (100%) that is required for clinical detection of early stages of lung cancer at screening of patients at risk (heavy smokers over 50). It was derived empirically from the diagnostic ROC curve of NSCLC patients versus cohort of COPD patients ( Figure 5 ). A comparison of the AUC difference of ROC curves for CA-62 and CEA confirmed that the CA-62 marker is the most accurate in identifying early stages of NSCLC in asymptomatic patients. Other tumor markers studied, such as CA-125, CA 15-3, CA 19-9, NSE did not have sufficient sensitivity and specificity to diagnose early stages of lung cancer in asymptomatic patients. However, a combination of various biomarkers allows achieving a thorough discrimination between benign and malignant neoplasms better than any single TM.

There are a few mathematical methods to combine TM. The first method is rather simple and based on averaging of the values for all 3 TM after their normalization. For normalizing all the markers, we used one reference cutoff value of the selected TM (CEA, cutoff 5 ng/ml). In this case, all sample values for CYFRA 21–1 were multiplied by 2 and divided by a 1000 for CA-62 marker.

To plot the ROC curve for the combination of CA-62, CEA and CYFRA 21–1 tumor markers using defined parameters were implemented the following mathematical techniques:

1) A conversion of three independent parameters, such as CA-62, CEA and CYFRA 21–1 values, to one arbitrary value corresponding to the principles-based developed testing approach in accordance with the following mathematical transformations:

(a) Samples were assigned the values equal to:

Therefore, samples were considered as “positive” if: 3TM Value > 5 U/ml, and “negative” otherwise. This approach allows the samples to have only one parameter to be considered. The ROC curves were plotted in accordance with the TNM classification. Diagnostic characteristics of CEA CA-62, CYFRA 21-1, and a combination of 3TM (CA-62, CEA and CYFRA 21-1) were evaluated based on their sensitivity and specificity, test accuracy, and curves were compared for different IVD methods. The level of significance was set at P<0.001.

The second method is a linear regression analysis, which is measured as the probability of a binary outcome (e.g. disease or no disease) given the variables used (25). It allows calculating the probability of cancer in a particular patient. For example, in a patient 62 years of age, with CYFRA 21-1 1ng/ml, CEA 4.2 ng/ml, CA-62–6200 U/ml and a positive LDCT scan the probability of having lung cancer is 72%. This is very important for both the doctor and a patient because there is no other way to assess the relative weight of all the variables. A regression model can compare the predicted results with the actual condition, which allows also generating a ROC curve ( Figure 6 ). The only difference is that the specificity and sensitivity stem from the combination of all the variables used in the model. Such mathematical model can also assess the significance of each variable used. For example, in combining cancer markers, the value of the markers is highly significant (very low p) whereas age might be not that important (p > 0.05), in which case, we can exclude age as a variable.

We studied a number of various TM combinations (CEA, CYFRA 21-1 & SCC; CEA, CA-125 & CYFRA 21-1; CEA, CA -125 & CA 15-3; CEA, CA-62, CYFRA 21-1 & SCC; CEA,CA-62,CYFRA 21-1 &CA-125; CA 15-3, CA-62 &CYFRA 21-1; CEA, CA-62 and CYFRA 21-1) aiming to find the utmost effective TM panel for early-stage LC detection. Obtained results confirm that a combination of biomarkers selected produces better results for Stage I & Stage II detection than CA-62 cancer marker alone (AUC 0.990 vs 0.973) and other previously proposed panel of markers containing PENK, pro-SP, hGH and CA15-3 (AUC = 0.785) (17). Therefore, adding a cytokeratin-19 fragment CYFRA 21–1 to a combination panel of CA-62 and CEA allows increasing specificity by eliminating false positive results of COPD patients. That improves the diagnostic value of the selected TM panel (CA-62, CEA and CYFRA 21-1) (Table 3). The results obtained demonstrated that a combination of two different N- and O-glycoproteins with a 19-citrate fragment (CEA, CA-62 and CYFRA 21-1) allows achieving 100% specificity, 93% sensitivity, and 94% test accuracy for either mathematical method.

A comparative analysis of the data presented on various TM and their panels used in lung cancer diagnosis worldwide showed that the CA-62 marker and its panel (CA-62, CEA and CYFRA 21-1) has the highest diagnostics characteristics as compared to other laboratory diagnostic methods for estimating lung malignancies currently available.

LDCT remains resource-intensive, requiring specialized equipment, trained technicians, radiologists, and physical infrastructure. Its implementation is often limited by availability and scheduling bottlenecks. In contrast, CA-62 testing is based on a simple blood draw, compatible with existing venous sampling workflows, including mobile and remote settings.

The CA-62 immunoassay, while novel, is built on established immunoassay platforms and does not require costly imaging or capital-intensive infrastructure. Unlike LDCT, which incurs fixed high costs per scan, the cost of CA-62 testing decreases significantly with scale, due to relatively stable reagent use and automation potential. We estimate that the direct cost per test can be reduced to below USD 20 in high-throughput settings, which is comparable to other standard IVD immunoassays (PSA, CEA).

While false positives are a known concern in screening, our modeling shows that incorporating CA-62 as a pre-screening filter reduces the number of patients requiring LDCT by a factor of 15–16 times, based on calculated PPV from real-world prevalence, sensitivity, and specificity.

In addition to LDCT, the use of tumor markers for monitoring the treatment of LC patients and monitoring their condition after reaching remission has been widely implemented as for evaluating the effectiveness of chemotherapy performed, and for the detection of relapse and dynamic surveillance of patients with LC (8, 13–16). Examples of such markers include CEA, NSE, SCC, CYFRA-21, and CA-125, but all these individual biomarkers are not sensitive enough to identify a reliable direct relationship between disease progression and their elevation. The limitation on TM use for LC detection is due to their low sensitivity (15-40%) at detecting early stages of cancer (21, 22).

The results of the blind clinical study described in this article showed the most significant diagnostic characteristics using the carcinoma-specific marker CA-62 for the detection of NSCLC (stages I-III): 96% sensitivity at 97% specificity, compared to other tumor markers studied by us or their combinations described in the literature ( Table 3 ). The best sensitivity values for early-stage detection (I and II) NSCLC were observed for CA-62 (92%) among all TMs: CEA (37%), CYFRA 21-1 (9%), SCC (9%), NSE (4.5%), CA-125 (3%), CA 15-3 (1.5%), and CA 19-9 (1%) ( Figure 1 ). The increased CA-62 marker cutoff value (5,600 U/ml) allowed the elimination of false positives, reaching high specificity (100%) at 89% sensitivity required for clinical detection of early-stage lung cancer.

An additional evaluation of the marker application efficacy includes the calculation of the weighted kappa coefficient (k), which is clearly demonstrates the difference between various diagnostic methods, exactly which biomarker or their combination classifies patients more reliably due to the lesser likelihood of a random coincidence of the test results with the histopathological findings (17). The exact values of the weighed kappa-coefficient demonstrate a significant difference between two diagnostic methods: an agreement between the two estimation methods is considered bad if 0< k< 0.20 and good if k > 0.81. When evaluating the detection effectiveness of NSCLC patients from the entire cohort, the values of kappa coefficients were distributed as follows: CA-62 (k=0.92) > CYFRA 21-1 (k=0.44) > CEA (k=0.43) > SCC (k=0.19). Similar results were obtained for the evaluation of the similar criteria in the cohort of patients with early stages of NSCLC: CA-62 (k=0.88) > CYFRA 21-1 (k=0.09) > CEA (k= 0.37) > SCC (k= 0.09). Therefore, from the entire panel of tumor markers CA-62 is the only marker that demonstrated high correlation with histopathology results in terms of recognition of malignant processes in nodules found incidentally on LDCT-scan or in the presence of pathological pulmonary changes, including infiltrates.

It is important to note that unlike mucins and other tumor markers, which are increased proportionally to the tumor growth, the marker CA-62 is significantly elevated from Stage I and demonstrated the highest diagnostic characteristics in detecting early stages (I & II) of NSCLC (Sen=91- 96.4%) ( Figure 2 ).

This is undoubtedly that CA-62 has a great advantage over other tumor markers with much lower sensitivity (7.1% – 48.2%) for detecting early stages of lung cancer. In particular, the biomarkers CYFRA 21-1 (Sen=44%), SCC (Sen=18.4%), and CA 19-9 (Sen=7.1%) do not have sufficient sensitivity to diagnose asymptomatic lung cancer, however at the same time they have high enough specificity to differentiate between the benign and malignant lung neoplasms.

Another approach for eliminating false positive results consists in combining highly sensitive cancer antigen CA-62 with other biomarkers. Therefore, an addition of cytokeratin 19 CYFRA 21–1 to glycoproteins CA-62 and CEA increases the specificity of early-staged NSCLC detection by eliminating false positive results, which significantly improves the diagnostic value of the tumor marker signature (CA-62, CEA & CYFRA 21-1): 100% Specificity, 93% Sensitivity and 94% test accuracy ( Table 3 ), highlighting the added value of CA-62 as a complementary marker that enhances both early detection and diagnostic precision. Using various mathematical approaches (using an average of biomarkers combination or regression analysis) allows achieving high diagnostic characteristics of chosen expanded TM combination (CEA, CA-62 & CYFRA 21-1) and getting a probability estimation of having LC using regression.

The results of the clinical study demonstrated that using the biomarkers signature (CA-62, CEA and CYFRA 21-1) allows increasing the Specificity of CT diagnostics for patients with suspicious changes on the tomogram, improving the interpretation of visualized localized focus, and improving the accuracy of differential diagnosis at detecting early stages of LC up to 94%. These findings suggest that CA-62 not only confirms the prior utility of CEA and CYFRA 21–1 but also extends their effectiveness into earlier disease stages, offering new opportunities for pre-symptomatic identification of high-risk individuals.

The unique feature of highly glycosylated N-glycoprotein CA-62 consists in its expression in large concentration on the cell membrane of transformed maligned stem cells from the onset of carcinogenesis, long before any clinical symptoms appearance (10, 11). The Sensitivity, Specificity and CLIA-CA-62 Test accuracy for the entire cohort of NSCLC patients were high at 96%, 97%, and 96.7%, which is higher than the sum of other tumor markers and their panels presented in Table 3 , which are currently used in LC diagnosis. Other lung cancer diagnostic methods have either a low sensitivity of 25-40% to 80%, depending on the stage of disease, or low specificity values. The Molina R. et al. (14) group’s studies have demonstrated that successful TM combinations have enough diagnostic sensitivity to detect LC. Clinical studies conducted by the authors on 802 patients demonstrated that using of a combination of several known markers (CEA, CYFRA 21–1 and SCC) for NSCLC and (ProGRP, NSE, CEA or CYFRA 21-1) for small-cell lung cancer (SCLC) allow reaching Sen=80% in LC patients with Sp = 92% at Stages I-III. Studies by Kyoko Okamura et all (13) confirmed the high diagnostic value of CEA and CYFRA 21–1 tumor markers for differentiating 655 patients with primary LC from 237 COPD patients. The sensitivity and specificity of CEA for LC detection was 69% and 68%, respectively, compared to 43% for CYFRA 21–1 at Sp=89%. The combined use of these two markers revealed a more significant result compared to individual TM at Sen=33% and Sp=95% for LC with a prevalence rate of 51% of patients at risk. Later studies of the Molina et al. (15) identified six serum TM (CEA, CA 15-3, SCC, CYFRA 21-1, NSE and progastrin releasing peptide ProGRP) that were associated with the histological subtypes of LC: NSCLC and SCLC. The authors showed that combined use of the 6 tumor markers described above showed better diagnostic characteristics for NSCLC detection as compared to the individual characteristics of tumor markers: better sensitivity, specificity PPV and NPV (87.1%, 82%, 85.3% and 84.3%, respectively. Jinlin Sun et al. (30) investigated another interesting combination of TM for NSCLC detection: CEA, secretory sialoprotein osteopontine (OPN) and secreted protein Dickkopf-1 (DKK1), which showed promising results. The sensitivity and specificity of the (CEA and OPN) combination for NSCLC (AUC=0.920) (95% CI, 0.875-0.964) were 87.5% and 86.67%, respectively. The area under the ROC curve of CEA and DKK1 for NSCLC was AUC=0.912 (95% CI, 0.866-0.958) with sensitivity and specificity 92.5% and 76.67%, respectively. It is important to clarify that the direct comparisons of different methods used are limited due to population heterogeneity, different assays, and statistical handling.

Over the past l0 years, a significant amount of research data has accumulated on other promising biomarkers for lung cancer detection: autoimmune antibody signature (Autoantibody signature, AAB) (31), micro-RNA (32) and pro-surfactant Protein B (Pro-SFTPB) (33). The presence of such biomarkers in plasma or serum of patients could be an independent predictor of lung cancer and may be a valuable addition to existing lung cancer screening models.

Finally, the results obtained confirm that the combined use of TM, in particular CEA, CA-62 and CYFRA 21–1 could be a useful strategy for improving of the integrated LC risk evaluation in a high-risk group. Reducing the proportion of false positives in initial LDCT scans, as well as reducing LC hyper diagnosis using a more effective evaluation of tumor aggressiveness, is an important and as yet unachieved clinical challenge.