Analytical Performance of Nanostring's Companion Diagnostic (CDx) Lymphoma Subtyping Test (LST) with Diffuse Large B-Cell Lymphoma Core Needle Biopsy Samples

Background: Diffuse large B-cell lymphoma (DLBCL) is an aggressive non-Hodgkin’s lymphoma with two distinct molecular cell-of-origin (COO) subtypes known as germinal center B-cell (GCB) or activated B-cell (ABC). DLBCL subtypes have been reported to be prognostic and potentially predictive of treatment benefit, underscoring the need for a precise and accurate CDx test (Roschewski, Nat Rev Clin Oncol, 2014). NanoString’s LST was developed to enable identification of the COO subtypes on the nCounter^® Dx Analysis System using formalin fixed paraffin embedded (FFPE) tissue specimens based on the previously developed Lymph2Cx gene expression profiling assay (Scott, Blood 2014). The test assigns the COO subtypes based on a Linear Predictor Score (LPS) that is calculated using a weighted sum of the gene expression. The LST is currently being used as a CDx in a phase III global clinical validation trial (clinicaltrials.gov NCT02285062) to identify newly diagnosed DLBCL patients of the ABC type who may preferentially respond to lenalidomide.  

The LST was previously reported to be a highly precise and accurate method of determining DLBCL subtypes (ABC and GCB) when using FFPE excisional/surgical biopsy tissue specimens as a test input (Wallden, JCO 2015). Although excisional biopsies are recommended for DLBCL diagnosis by current ESMO and NCCN guidelines, recent studies have reported excellent sensitivity and specificity in diagnosing malignant lymphomas by core needle biopsy (CNB) leading to increased usage of this method (Hu Q, Am J Clin Pathol, 2013, de Kerviler, Best Pract Res Cl Ha, 2012). The aim of the current study was to assess the analytical performance of the LST when using CNB samples as a test input.   

Methods: Recently archived CNB and small biopsy DLBCL samples were procured from Oregon Health & Science University (Portland, OR) (n=23), Hôpital Saint-Louis (Paris, France) (n=19), and Asterand Bioscience (n=12). Both nodal and extra-nodal tissue sites were included. Pathology review was performed on an H&E stained slide from each tissue sample to identify the area of DLBCL. Unstained slide mounted FFPE tissue sections (5 µm) were prepared from each sample for assay processing. The RNA extraction and LST procedures were the same as reported previously for surgical biopsy samples (Wallden, JCO 2015). The number of sections used for each RNA extraction was based on the tumor surface area (TSA) measured on the slide (typically 4-6 slides for TSA<10 mm²; 2-4 slides for TSA≥10 mm²). Multiple extractions were performed and subsequently tested independently for each sample (n=148 total). Additional variables tested in this study were user (n=2), reagent lot (n=2), and instrument (n=2). The reproducibility of the test outputs were evaluated using a linear mixed model (LMM) for LPS variance and concordance for subtype call. Principal component analysis (PCA) was performed on the gene expression profile to determine the major source(s) of variability. 

Results: Of the 54 total samples tested, 51 provided test results (failures were due to RNA degradation). With the 51 passing samples, an assay pass rate of 97% (144/148) was achieved. The total tissue area input into the test was correlated to RNA yield (R²=0.54). The LPS was highly reproducible across sample replicates (Figure 1) with a total assay standard deviation of 59 LPS units (<2% of the LPS range). This estimate was similar to surgical biopsy samples (76 LPS units). The subtype concordance was estimated as 96-98% (between users) where 2 of 51 samples with LPS scores close to the thresholds moved from ABC or GCB to unclassified, or vice versa. A PCA of the gene expression data demonstrated that the first principle component is highly correlated to the LPS providing further evidence that the biology underlying the LST is the major driver of algorithm gene expression in CNB. 

Conclusions: In this study we demonstrated robust performance of the LST using FFPE DLBCL CNB/small biopsy samples. Our data support that sufficient RNA yield and gene expression quality can be achieved from biopsy samples with measured TSA of 2 mm² or greater. We also show that the analytical performance of the LST is comparable between CNB/small biopsy samples and surgical biopsies and that the principal component of gene expression variation is LPS. The LST sample input requirements for the ongoing phase III trial will be updated to include CNB/small biopsy samples to accommodate clinical practice.

Disclosures: Storhoff: NanoString Technologies, Inc.: Employment , Other: Stock . Wallden: NanoString Technologies, Inc.: Employment , Other: Stock . Braziel: Oregon Health & Sciences University: Employment . Thieblemont: St. Louis Hospital, Paris, France: Employment . Hood: NanoString Technologies, Inc.: Employment , Other: Stock . Ravi: NanoString Technologies, Inc.: Employment , Other: Stock . Dennis: NanoString Technologies, Inc.: Employment , Other: Stock . Dowidar: NanoString Technologies, Inc.: Employment , Other: Stock . Danaher: NanoString Technologies, Inc.: Employment , Other: Stock . Dunlap: Oregon Health & Sciences University: Employment . Briere: St. Louis Hospital, Paris, France: Employment . de Kerviler: St. Louis Hospital, Paris, France: Employment . Ferree: NanoString Technologies, Inc.: Employment , Other: Stock , Patents & Royalties: NanoString Technologies, Inc. .

See more of: 622. Non-Hodgkin Lymphoma: Biology, excluding Therapy: Poster I
See more of: Non-Hodgkin Lymphoma: Biology, excluding Therapy
See more of: Oral and Poster Abstracts

<< Previous Abstract | Next Abstract >>

^*signifies non-member of ASH