Performance Characteristics of the BluePrint® Breast Cancer Diagnostic Test

Abstract

The analytical performance of a multi-gene diagnostic signature depends on many parameters, including precision, repeatability, reproducibility and intra-tumor heterogeneity. Here we study the analytical performance of the BluePrint 80-gene breast cancer molecular subtyping test through determination of these performance characteristics. BluePrint measures the expression of 80 genes that assess functional pathways which determine the intrinsic breast cancer molecular subtypes (i.e. Luminal-type, HER2-type, Basal-type). Knowing a tumor's dominant functional pathway can help allocate effective treatment to appropriate patients. Here we show that BluePrint is a highly precise and highly reproducible test with correlations above 98% based on the generated index and subtype concordance above 99%. Therefore, BluePrint can be used as a robust and reliable tool to identify breast cancer molecular subtypes.

Figure 1

Repeatability and precision of BluePrint FFPE indices. Chart showing BluePrint indices (y-axis) of four clinical samples (S1, red, BluePrint Basal-type; S2 green, BluePrint HER2-type; S3 black, BluePrint Luminal-type; S4, blue, BluePrint Luminal-type) in duplicate (run1, run2) over a 20-day period (x-axis). Each dot represents a single breast cancer sample for which total RNA underwent BluePrint microarray laboratory processing and analysis. FFPE = formalin-fixed paraffin embedded. Categorical concordance was 100%.

Figure 2

Over-time reproducibility of BluePrint FFPE indices. Chart showing over-time index measurements from January 2012 through January 2015 of three clinical control samples covering the three subtypes. A. C1, measuring Luminal-type index n = 1639, B. C2, measuring HER2-type index n = 1534, C. C3, measuring Basal-type index n = 1594. Concordance for all samples overtime was 100%. Each dot represents a single breast cancer sample for which total RNA underwent BluePrint microarray laboratory processing and analysis.

Figure 3

Inter-laboratory reproducibility of BluePrint FFPE – comparison of BluePrint indices between two Agendia sites Amsterdam (location 1) and Irvine (location 2) (N = 97). A. BluePrint Luminal-type indices, B. BluePrint HER2-type indices, C. BluePrint Basal-type indices, D. BluePrint indices comparison for the final assigned subtype (Luminal-, HER2- and or Basal-type). BluePrint subtype concordance was 100%. Each dot represents a single breast cancer sample for which total RNA underwent BluePrint microarray laboratory processing and analysis.

Figure 4

Reproducibility of BluePrint between two independent isolations, comparison of indices between isolation 1 and isolation 2 using different sections of the tumor (N = 46). A. BluePrint Luminal-type indices, B. BluePrint HER2-type indices, C. BluePrint Basal-type indices, D. BluePrint indices comparison for the final assigned subtype (Luminal-, HER2- and or Basal-type). BluePrint subtype concordance was 100%.

Figure 5

Comparison study: agreement of BluePrint FFPE between the gold standard array (array 1) and the new array (array 2) (N = 98). A. BluePrint Luminal-type indices, B. BluePrint HER2-type indices, C. BluePrint Basal-type indices, D. BluePrint indices comparison for the final assigned subtype (Luminal-, HER2- and or Basal-type). BluePrint subtype concordance was 99%. Each dot represents a single breast cancer sample for which total RNA underwent BluePrint microarray laboratory processing and analysis.

Figure 6

Comparison study: agreement of BluePrint FFPE between gold standard scanner (Scanner 1) and the upgraded scanner (Scanner 2) (N = 80). A. BluePrint Luminal-type indices, B. BluePrint HER2-type indices, C. BluePrint Basal-type indices, D. BluePrint indices comparison for the final assigned subtype (Luminal-, HER2- and or Basal-type). BluePrint subtype concordance was 100%. Each dot represents a single breast cancer sample for which total RNA underwent BluePrint microarray laboratory processing and analysis and was scanned on both scanners.

Figure 7

Kaplan–Meier curve of 345 microarRAy-prognoSTics-in-breast-cancER (RASTER) study patients for 5-year distant recurrence-free interval (DRFI) assessed using BluePrint FF (A) and FFPE (B) tissues. Kaplan–Meier curves plotted for DRFI show comparable clinical performance of BluePrint Luminal-type (blue lines), HER2-type (green lines) and Basal-type (red lines) patients in matched FF and FFPE in a series of 345 early-stage breast cancer patients. P-values of significance are assessed using log-rank test and showed on the graphs. FF = Fresh Frozen, FFPE = formalin-fixed paraffin embedded.