Laboratory Medicine ›› 2018, Vol. 33 ›› Issue (7): 590-596.DOI: 10.3969/j.issn.1673-8640.2018.07.003

• Orginal Article • Previous Articles     Next Articles

Combined determination of multiple tumor markers for the diagnosis of primary lung cancer

ZHANG Haichen1, WANG Hao2, SONG Yunxiao1, MA Jin3()   

  1. 1. Department of Clinical Laboratory,Shanghai Xuhui Central Hospital,Shanghai 200031,China
    2. Department of Clinical Laboratory,Jiangxi Mental Hospital,Nanchang 330029,Jiangxi,China
    3. School of Public Health,Shanghai Jiaotong University School of Medicine,Shanghai 200025,China
  • Received:2018-02-01 Online:2018-07-30 Published:2018-07-27

Abstract:

Objective To investigate the role of combined determination of multiple tumor markers in the diagnosis of lung cancer,and to establish appropriate strategies of tumor marker determinations. Methods A total of 280 patients with primary lung cancer(PLC) and 455 patients with benign pulmonary disease(BPD) were enrolled. Serum alpha-fetoprotein(AFP),carcinoembryonic antigen(CEA),carbohydrate antigen(CA)50,CA242,CA125,CA15-3,CA19-9,cytokeratin 19 fragment(CYFRA 21-1),CA72-4,squamous cell carcinoma antigen (SCC-Ag),neuron-specific enolase(NSE) and tumor specific growth factor(TSGF) were determined. Receiver operating characteristic(ROC) curve was used to evaluate diagnosis performance. Exploratory factor analysis(EFA) and Logistic regression model were used to evaluate the role of combined determination. Results The levels of CEA,CA15-3,CA72-4,CA242,CYFRA 21-1,CA125,CA19-9,CA50,SCC-Ag,NSE and AFP in PLC group were higher than those in BPD group (P<0.01),except for TSGF. ROC curve analysis showed that the areas under curves (AUC)of CA15-3,NSE,CA125,CEA,CYFRA 21-1 and CA72-4 in PLC group were ≥0.7,while the AUC of TSGF,CA19-9,CA242,AFP,CA50 and SCC-Ag were <0.7. After excluding TSGF,EFA showed that there were 4 independent potential factors in PLC group. The independent potential factors were evaluated by comprehensive evaluation in order to give PLC predictive model,which was made up of CA125,CA19-9,CYFRA 21-1,NSE,SCC-Ag,CEA and CA15-3. The diagnosis performance of PLC predictive model and the combined determination of 11 tumor markers had no difference (Z=1.744,P=0.081). The PLC predictive model had better diagnosis performance comparing to that of any single tumor marker determination,having the AUC of 0.831,the sensitivity of 70.7% and the specificity of 83.7%,and the positive likelihood ratio (+LR) and negative likelihood ratio (-LR) were 4.35 and 0.35,respectively. The optimal cut-off value was -0.958 8. Conclusions Comparing to single tumor marker determinations,the PLC predictive model reached a balance between sensitivity and specificity. The diagnosis performance of PLC predictive model is not decreased with the reduction of tumor marker number.

Key words: Tumor marker, Primary lung cancer, Exploratory factor analysis, Logistic regression analysis, Receiver operating characteristic curve

CLC Number: