Clinical development of a blood biomarker using apolipoprotein-A2 isoforms for early detection of pancreatic cancer

Abstract

Background: We have previously reported apolipoprotein A2-isoforms (apoA2-is) as candidate plasma biomarkers for early-stage pancreatic cancer. The aim of this study was the clinical development of apoA2-is.

Methods: We established a new enzyme-linked immunosorbent sandwich assay for apoA2-is under the Japanese medical device Quality Management System requirements and performed in vitro diagnostic tests with prespecified end points using 2732 plasma samples. The clinical equivalence and significance of apoA2-is were compared with CA19-9.

Results: The point estimate of the area under the curve to distinguish between pancreatic cancer (n = 106) and healthy controls (n = 106) was higher for apoA2-ATQ/AT [0.879, 95% confidence interval (CI): 0.832-0.925] than for CA19-9 (0.849, 95% CI 0.793-0.905) and achieved the primary end point. The cutoff apoA2-ATQ/AT of 59.5 μg/mL was defined based on a specificity of 95% in 2000 healthy samples, and the reliability of specificities was confirmed in two independent healthy cohorts as 95.3% (n = 106, 95% CI 89.4-98.0%) and 95.8% (n = 400, 95% CI 93.3-97.3%). The sensitivities of apoA2-ATQ/AT for detecting both stage I (47.4%) and I/II (50%) pancreatic cancers were higher than those of CA19-9 (36.8% and 46.7%, respectively). The combination of apoA2-ATQ/AT (cutoff, 59.5 μg/mL) and CA19-9 (37 U/mL) increased the sensitivity for pancreatic cancer to 87.7% compared with 69.8% for CA19-9 alone. The clinical performance of apoA2-is was blindly confirmed by the National Cancer Institute Early Detection Research Network.

Conclusions: The clinical performance of ApoA2-ATQ/AT as a blood biomarker is equivalent to or better than that of CA19-9.

Keywords: Apolipoprotein A2-isoform; Blood biomarker; Carbohydrate antigen 19-9 (CA19-9); Early detection of pancreatic cancer.

Fig. 1

Selection criteria and eligibility requirements for enrollment of plasma samples in clinical investigations and the apoA2-i ELISA kit for the in vitro diagnostic (IVD) test. A Healthy individuals in the cutoff setting test (n = 2000) and in the cutoff evaluation test (n = 400). B Healthy individuals in the clinical performance test for evaluating the apoA2-i ELISA kit for IVD (n = 106). C Pancreatic cancer in the clinical performance test for evaluating the apoA2-i ELISA kit for IVD (n = 106). D IPMNs in the clinical performance test for evaluating the apoA2-i ELISA kit for IVD (n = 30). E Chronic pancreatitis in the clinical performance test for evaluating the apoA2-i ELISA kit for IVD (n = 10)

Fig. 2

Configuration of the apoA2-i ELISA kit for research use only (RUO) and in vitro diagnostic (IVD) use (A, B). Distributions and discrimination performances of apoA2-i and CA19-9 in pancreatic cancer, high-risk individuals, and healthy controls (C–G). A The ELISA IVD configuration comprises two ELISAs for measuring apoA2-AT and apoA2-ATQ. Both assays for apoA2-AT and apoA2-ATQ use Pan-apoA2 monoclonal antibody as the solid phase antibody. The assay for apoA2-AT uses apoA2-AT-specific rabbit polyclonal antibody for the HRP-conjugated antibody. The assay for apoA2-ATQ uses ApoA2-ATQ-specific monoclonal antibody for the HRP-conjugated antibody. B The ELISA RUO kit comprises two ELISAs (classical type). The assay for apoA2-AT uses apoA2-AT-specific rabbit polyclonal antibody for the immobilized antibody and Pan-apoA2 monoclonal antibody for the HRP-conjugated antibody. The assay for apoA2-ATQ uses Pan-apoA2 polyclonal antibody for the immobilized antibody and ApoA2-ATQ-specific monoclonal antibody for the HRP-conjugated antibody. (C) Distribution of apoA2-ATQ/AT [APOA2-i Index; (apoA2_ATQ/AT) = $\sqrt{(\text{apoA}{2}_{\text{ATQ}}\ast \text{apoA}{2}_{\text{AT}})}$] in the plasma of healthy controls, each stage of pancreatic cancer, IPMNs, and chronic pancreatitis. Black bars: median values. The asterisks indicate significance on the Wilcoxon rank sum test compared to healthy controls (*p < 0.05; **p < 0.01; ***p < 0.001). D Distribution of CA19-9 levels in plasma of healthy controls and each stage of pancreatic cancer. Black bars: median values. The asterisks indicate significance on the Wilcoxon rank sum test compared to healthy controls (*p < 0.05; **p < 0.01; ***p < 0.001). E Primary end point: ROCs and AUCs of apoA2-ATQ/AT and CA19-9 (red line: apoA2-ATQ/AT, blue dashed line: CA19-9) between healthy controls and all stages of pancreatic cancer. F ROCs and AUCs of apoA2-ATQ/AT and CA19-9 (red line: apoA2-ATQ/AT, blue dashed line: CA19-9) between healthy controls and stage I pancreatic cancer. G ROCs and AUCs of apoA2-ATQ/AT and CA19-9 (red line: apoA2-ATQ/AT, blue dashed line: CA19-9) between healthy controls and TS1 (tumor size ≤ 2 cm) of pancreatic cancer

Fig. 3

Validity of the cutoff value of apoA2-ATQ/AT, combination assay with apoA2-ATQ/AT and CA19-9, and positive predictive value (PPV) and negative predictive value (NPV) of apoA2-ATQ/AT and CA19-9. A Kernel density estimation of apoA2-ATQ/AT in the cutoff setting test. ApoA2-ATQ/AT with specificity of 95% is 59.5 μg/mL (dashed line: 95% specificity). B Kernel density estimation of apoA2-ATQ/AT in the cutoff estimation test. The specificity of apoA2-ATQ/AT of 59.5 μg/mL is 95.8% (dashed line: apoA2-ATQ/AT 59.5 μg/mL). Two-dimensional scatter plots of apoA2-ATQ/AT and CA19-9 in healthy controls and pancreatic cancer (C) (blue dots: healthy controls, red crosses: pancreatic cancer) and in each stage of pancreatic cancer (D) (black triangles: stage I, purple circles: stage II, purple squares: stage III, and gray crosses: stage IV pancreatic cancer) [black dashed lines: cutoff values of apoA2-ATQ/AT (59.5 μg/mL) and CA19-9 (37 U/mL)]. PPV and NPV of apoA2-ATQ/AT (E) and CA19-9 (F) in the population according to the prevalence of pancreatic cancer calculated by Bayesian estimation (red line: PPV, blue dashed line: NPV)