Combination of Biomarkers-IGM by Logistic Regression Improves Diagnostic Accuracy in Hepatocellular Carcinoma

Abstract

Background and aim Circulating immune complexes of alpha-fetoprotein (AFP) and SCCA (squamous cell carcinoma antigen) with IgMs (AFP-IgM and SCCA-IgM, respectively) represent a promising new class of serum markers with diagnostic and prognostic value in the management of patients with hepatocellular carcinoma (HCC). The enhancement of diagnostic indexes for cancer detection achieved with combination of biomarkers by linear logistic regression has been demonstrated in several simulation studies on multiple diagnostic assays. The aim of the study was to evaluate the improvement of the diagnostic accuracy by combination of SCCA-IgM and AFP-IgM compared with the diagnostic accuracy of free AFP in sera of patients with HCC and cirrhosis. Patients and methods Serum samples from 81 patients with HCC (mean age ± SD: 63 ± 8 years; male patients 73%; HCV infected 52%; without any therapeutic treatment 50%) and 82 patients with cirrhosis (mean age ± SD: 53 ± 9 years; male patients 66%; HCV infected 68%) were collected. The logistic regression parameters were calculated on a previously published data set of 50 patients with HCC and 50 patients with cirrhosis, where the distribution of serum levels of SCCA-IgM and AFP-IgM in HCC patients was significantly different from that in patients with cirrhosis (Mann-Whitney U test, p<0.01). Serum levels of SCCA-IgM and AFP-IgM were assessed in parallel using the Hepa-IC and Hepa AFP-IC kits and AFP was determined by an automatic analyzer (ADVIA Centaur®, Siemens Diagnostics, Italy). Results The patients were stratified according to sex (male), HCV infection (positive) and therapeutic treatment (negative) to apply a linear logistic regression model using regression parameters calculated from a data set of a reference population. A subgroup of 30 patients with HCC and 41 patients with cirrhosis was analyzed. The diagnostic accuracy measured as the area under the ROC curve (AUC) for the SCCA-IgM and AFP-IgM assays was roughly the same, with a weak supremacy of AFP-IgM (AUC = 0.62), and was comparable with that of free AFP (AUC = 0.64). The gain in diagnostic accuracy achieved with the combination of SCCA-IgM and AFP-IgM levels by linear logistic regression was 14% (AUC = 0.71) compared with the diagnostic accuracy of AFP-IgM and 11% compared with that of free AFP. With the estimation of the partial AUC (pAUC0.3), an alternative measurement of AUC that takes into consideration only specificity rates with clinical relevance (specificity >70%), the highest improvement in accuracy for HCC detection was obtained, with an increase of up to 62% compared to pAUC0.3 of AFP-IgM and up to 23% compared to pAUC0.3 of free AFP. Conclusion The results demonstrate that the combination by linear logistic regression of biomarkers-IgM immune complexes improves the diagnostic accuracy for HCC detection using regression parameters calculated in a reference population with defined clinical characteristics. These results support the usefulness of devices based on a panel of non-overlapping biomarkers-IgM complexes, such as multi-marker biochips, increasing the sensitivity and maintaining a high specificity for HCC diagnosis compared with conventional single-marker assays.