Analytical Validation of SKY92 for the Identification of High-Risk Multiple Myeloma

Multiple myeloma (MM) is an incurable plasma cell cancer with a large variability in survival. Patients MM classified as high risk by the SKY92 gene expression classifier are at of relapse and short Analytical validation assay was performed primary bone marrow specimens from 12 patients 7 reference line specimens. The results were 100% concordant and/or their expected result for sensitivity, specificity, microarray stability, RLT buffer stability. 90% specimen 96.4% intermediate precision, 80% to RNA For cell-line reproducibility, concordance least 92.9%, except one near-cut point specimen. clinical 100%, two Three independent laboratories ≥77.8% ≥92.9% experiments patient lines, respectively. Statistical acceptance thresholds developed Δ ≤1.48 (change score) SD ≤0.45 (SD across scores). Using Clinical Laboratory Standards Institute method choice (EP05-A2/A3), restricted maximum likelihood, observed values (0 1.14) SDs (0.22 0.31) passed criteria. Thus, we successfully present analytical prognostic molecular test individual MM. malignant disorder1Kyle R.A. Rajkumar S.V. myeloma.N Engl J Med. 2004; 351: 1860-1873Crossref PubMed Scopus (1216) Google Scholar, 2Kyle Therneau T.M. Offord J.R. Larson D.R. Plevak M.F. Melton 3rd, L.J., A long-term study prognosis monoclonal gammopathy undetermined significance.N 2002; 346: 564-569Crossref (1154) 3Kyle Incidence multiple Olmsted Counsty, Minnesota: trend over 6 decades.Cancer. 101: 2667-2674Crossref (127) 4Rajkumar Dimopoulos M.A. Palumbo A. Blade J. Merlini G. Mateos M.V. Kumar S. Hillengass Kastritis E. Richardson P. Landgren O. Paiva B. Dispenzieri Weiss LeLeu X. Zweegman Lonial Rosinol L. Zamagni Jagannath Sezer Kristinsson S.Y. Caers Usmani S.Z. Lahuerta J.J. Johnsen H.E. Beksac M. Cavo Goldschmidt H. Terpos Kyle Anderson K.C. Durie B.G. Miguel J.F. International Myeloma Working Group updated criteria diagnosis myeloma.Lancet Oncol. 2014; 15: e538-e548Abstract Full Text PDF (2075) Scholar accounts about 17% hematologic malignancies United States.5Siegel R.L. Miller K.D. Jemal Cancer statistics: 2019.CA Clin. 2019; 69: 7-34Crossref (11003) American Society estimates approximately 32,000 newly diagnosed cases 2019 >12,000 deaths MM.5Siegel heterogeneous disease increasing genetic changes each same patient, median overall survival between 4 10 years.6Kuiper R. Broyl de Knegt Y. van Vliet M.H. Beers E.H. der Holt el Jarari Mulligan Gregory W. Morgan Lokhorst H.M. Duin Sonneveld signature myeloma.Leukemia. 2012; 26: 2406-2413Crossref (144) More more therapies have been introduced landscape combat heterogeneity disease.7Rajkumar myeloma: 2018 update on diagnosis, risk-stratification, management.Am Hematol. 2018; 93: 981-1114Crossref (68) With 18 drugs currently approved US Food Drug Administration (https://www.cancer.gov/about-cancer/treatment/drugs/multiple-myeloma, last accessed September 8, 2020) that typically used combination, sometimes up four different mechanisms action, list possibilities long, generating complex decision tree. use stratification internationally recognized key opinion leaders guidelines tool making.8Greipp P.R. San Crowley Barlogie Bladé Boccadoro Child J.A. Avet-Loiseau Ludwig Powles Shimizu K. Shustik C. Tosi Turesson I. Westin staging system myeloma.J Clin 2005; 23: 3412-3420Crossref (1958) 9Avet-Loiseau Hulin Campion Rodon Marit Attal Royer Dib Voillat Bouscary D. Caillot Wetterwald Pegourie Lepeu Corront Karlin Stoppa A.M. Fuzibet J.G. Delbrel Guilhot F. Kolb Decaux Lamy T. Garderet Allangba Lifermann Anglaret Moreau Harousseau J.L. Facon Chromosomal abnormalities major factors elderly intergroupe francophone du myélome experience.J 2013; 31: 2806-2809Crossref (84) 10Paiva Cedena M.T. Puig N. Arana Vidriales M.B. Cordon Flores-Montero Gutierrez N.C. Martín-Ramos M.L. Martinez-Lopez Ocio E.M. Hernandez Teruel A.I. Rosiñol Echeveste Martinez Gironella Oriol Cabrera Martin Bargay Encinas Gonzalez Van Dongen Orfao Grupo Español Mieloma/Programa para Estudio la Terapéutica en Hemopatías Malignas (GEM/PETHEMA) Cooperative Study Groups: Minimal residual monitoring immune profiling patients.Blood. 2016; 127: 3165-3174Crossref (96) National Comprehensive Network unanimously agreed “a useful may be helpful selected estimate aggressiveness individualize treatment.”11Kumar S.K. Callander N.S. Liedtke Baljevic Campagnaro Castillo Chandler J.C. Cornell R.F. Costello Efebera Faiman Garfall Godby Holmberg Htut Huff C.A. Kang Malek Omel Raje Sborov Singhal Stockerl-Goldstein Tan Weber Johnson-Chilla Keller NCCN Guidelines Insights: Myeloma, Version 1.2020.J Natl Compr Canc Netw. 17: 1154-1165Crossref (50) In current study, profile classifier, SKY92, assesses evaluates 92 genes cells, validated. It provides binary classification into standard or improve prediction practice.6Kuiper Scholar,12Kuiper Broijl El W.M. Prediction high- low-risk based Staging System.Blood. 2015; 126: 1996-2004Crossref (63) Currently, does not direct any specific treatment strategy. studies 16 data sets, total 3418 cases, demonstrated accurate prognosticator high-risk standard-risk other markers.13Van Kuiper Best Chari Jakubowiak Levy J.B. Auclair Reece Siegel D.S. Stewart A.K. Trudel Vij Zimmerman Fonseca Prognostic its combination ISS cohort myeloma.Clin Lymphoma Leuk. 2017; 555-562Abstract (13) 14Van Ubels Dumee Bosman Waage Marker Is Validated Eight Datasets. European Hematology Association, Hague, Netherlands2016: P276Google 15Kuiper RNA-Seq Based Risk Stratification Validates High CoMMpass Trial. Netherlands2018: PF528Google 16Kuiper Levin M.D. Visser Hansson Velden A.W.G. Vermeulen Koenders Beverloo H.B. Stevens-Kroef Prognosis Elderly HOVON-87/NMSG-18 Revised SKY92-ISS. Netherlands2019: PS1374Google 17Van Jasielec Dytfeld L.C.M. Alonge Rosebeck Kandarpa McDonnell Griffith K.A. Predictive Gene Expression Profiling (GEP) Markers Confirmed Carfilzomib, Lenalidomide, Dexamethasone (KRd) Treated Newly Diagnosed (NDMM) Patients. Hematology, Washington, DC2014Crossref 18Wester Asselbergs Kersten M.J. Vellenga Marwijk Kooy Weerdt Minnema Lonergan Phase 2 carfilzomib, thalidomide, low-dose dexamethasone induction/consolidation diagnosed, transplant eligible myeloma, Carthadex trial: 2016.Blood. 128: 1141Crossref 19Kuiper M.-D. Annette Berna predictive performance combined R-ISS study.IMW Abstract. 710: 15-16Google 20Hofste op Bruinink Driessen Hajek Differential Effect Upfront Intensification Treatment Genetically Defined Groups - Combined Analysis ISS, del17p Scores EMN-02/Hovon-95 DC2018: 3186Google 21Van powerful identify both low cohorts: ASH 2015.Blood. 2970Crossref 22Chen Y.T. Valent Oliva Haferlach Chng W.-J. Larocca Validation Low Prognostication Retrospective, Multinational Cohort 155 Non-Trial IMW, Boston2019Abstract 23Sherborne A.L. Begum D.B. Price Johnson D.C. Ellis Smith Mirabella M.K. Kimber Jones Pawlyn Jenner M.W. Houlston R.S. Cook Russel Striha Collett Waterhouse Cairns D.A. Drayson Owen R.G. Davies F.E. G.J. Jackson G.H. Kaiser Identifying Ultra-High Integrated Molecular Genetic Profiling. DC2016Crossref 24Shah V. Sherborne Chowdhury Kendall behalf NCRI Haematology Studies GroupPredicting ultrahigh profiling: analysis XI trial patients.Leukemia. 2020; 34: 3091-3096Crossref Although preliminary (precision) published,25Van Huigh Precision Part Signature MMprofiler Assay. P282Google this reports outcome validity classifier. Regardless many successful results, fundamental importance patients, clinicians, regulatory reimbursement stakeholders because it demonstrates reliability technical level. Design these guided (CLSI), assessment standards, stability assessed [Palmetto GBA, MolDX Technical Assessment Summary Form (M00116, V2), https://www.palmettogba.com/palmetto/providers.nsf/files/moldx_technical_assessment_summary_form_(M00116).pdf/$file/moldx_technical_assessment_summary_form_(M00116).pdf, 15, 2020].26CLSIEvaluation Quantitative Measurement Procedures; Approved Guideline – Third Edition. CLSI document EP05-A3. Institute, Wayne, PA2014Google 27CLSIInterference Testing Chemistry, 3rd Edition EP07. PA2018Google 28CLSI EP12-A2User Protocol Evaluation Qualitative Test Performance; Second EP12-A2. PA2008Google 29CLSI I/LA28-A2Hewitt S.M. Robinowitz Bogen S.A. Gown Kalra K.L. Otis C.N. Spaulding Taylor C.R. Quality Assurance Control Implementation Immunohistochemistry Assays; Guideline-Second 30CLSI EP33 (ISBN 1-56238-927-0 [Print]; ISBN 1-56238-928-9 [Electronic]). PA2016Google 31Kuiper De Der A.W. EMC92/SKY92 revised robustly study.Haematologica. 102: 269-270Google Specimens listed Supplemental Table S1. De-identified fresh obtained consent under Institutional Review Board approval accordance Declaration Helsinki. Bone specimens, human heart cover score range risk, including near cut scores. various relevant standards (EP05-A3, EP07, EP012-A2, I/LA28-02, EP33) throughout. All analyses preplanned, quantitative well qualitative scores using final standardized after deliberate variations thereof determine robustness. Fresh aspirates collected heparin EDTA-containing collection tubes processed Ficoll density gradient (HISTOPAQUE-1077 H8889; Sigma-Aldrich, Zwijndrecht, Netherlands) obtain mononuclear followed immunomagnetic separation CD138-positive cells EasySep Human CD138 Positive Selection Kit (18357; Stemcell, Köln, Germany). extracted Qiagen AllPrep kits (Mini 80204 Micro 80284; Benelux BV, Venlo, Netherlands), quantity purity measured spectrophotometry (NanoDrop 2000; ThermoFisher Scientific, Breda, integrity Bioanalyzer (2100; Agilent, Amstelveen, Netherlands). cDNA prepared biotinylated 3′ IVT Plus hybridization wash stain kit (catalog numbers 902416 900720, respectively; Thermo Fisher, Fragmented copy reagents produce cocktails hybridized [based U133 2.0 GeneChip design (Thermo Fisher Singapore)], then scanned GCS3000Dx version (Dx2 instrument; Fisher). To ensure quality end result, control (QC) evaluated (Supplemental S2). If met, declared invalid excluded subsequent (in keeping routine protocols). generated uploaded Omnibus repository (https://www.ncbi.nlm.nih.gov/geo), available accession number GSE159289. calculated previously described.6Kuiper raw MAS532Wilson C.L. C.J. Simpleaffy: BioConductor package Affymetrix analysis.Bioinformatics. 21: 3683-3685Crossref (302) normalized, log2 transformed, mean variance normalized against fixed set probe sets multiplied weighting summed score, which defined when ≥0.7774, either (<0.7774) (≥0.7774). ordinal derived dimensionless score. Acceptance 83 representative Heamato Oncology Foundation Adults Netherlands (HOVON87)/Nordic (NMSG18) trial,32Wilson distribution (Figure 1A). Figure 1B shows relation added proportion class switching. Gaussian noise until 10% switched classification. corresponding switching 0.45 (C5-C95 around 0.7774 = 0.04 1.52) 1.48 exceeded. Agreement percentages following ILA28-A2E,29CLSI agreement (OA) (PHR + PSR), percentage wherein HR SR PHR PSR probabilities. limit detection varying input amounts triplicates three (RS2, RS3, RS4). 50, 100 (standard), 500 ng. QC, resulted did exceed threshold 2A 1), justifying amount ≥100 ng material provide robust SY92 result. training set6Kuiper (based CD38 flow cytometry separation) ≥80% purification, also QC cutoff testing. An impact minimum tumor content diluting freshly healthy volunteer white blood cells. Triplicates (RS3 RS4) purities 70%, 80%, 90%, 2B 1). Duplicates single run (Pa5 Pa7) 98%, 70% prespecified versus identified positively correlated linear R2 scores, RS3 (P 0.0008) RS4 0.0065) but impacting 2B). This further emphasized highly significant regression lines small β values. These justify chosen (≥80% purity) criterion assay.Table 1Summary ResultsStudySubstudySubparameterOverall agreement, %Observed ΔCriterionConclusionSensitivityMinimum input50 RNA500 RNA1001000.29, 0.16, 0.120.25, 0.61Δ < 1.48PassPassMinimum contentRS3 (70%–100%)Pa5 Pa7 (98% 70%)1001000.61 1.140.68 0.28Δ 1.48PassPassSpecificityInterfering substancesHeparin (3 concentration)1000.03, 0.01, 0.08Δ 1.48PassEDTA concentration)1000.18, 0.20, 0.22Δ 1.48PassHemoglobin (1 concentration)1000.02Δ 1.48PassFicoll concentration)1000.04Δ 1.48PassPrecisionRepeatabilityPooled S3)96.40.22SD 0.45PassIntermediate precisionPooled S3)0.28SD 0.45PassReproducibility linesAcross laboratories: RS1RS2RS3RS4RS5RS6RS7Across samples: 1-laboratory 2Laboratory 3Laboratory 2-laboratory 310010010010010055.310092.992.995.10.31 (Across samples)SD samplesAcross Pa5Pa8Pa10Pa11Pa12Across 310051.910057.410083.375.677.80.30 0.45PassStabilityMicroarray 3 lots24 months 4°C1000.00Δ 1.48PassPrimary sample stability24 hours room temperature marrow900.43Δ 1.48PassIntermediate stabilityPa8 Pa9 0–6 0–11 months80 (8/10) (3/3)0.76 0.14Δ 1.48PassPa5 0–1 5 0–3 months100 (5/5) (2/2)0.32 0.63Δ 1.48Pass Open table new tab Interfering substances can biological compounds intrinsic external reagent compounds. largely removed during workup levels processing could 33 experiments, interferants analyzed per guideline EP07-A3.27CLSIInterference (fold increase parentheses) spiked (RS3, 107/mL): heparin, 35.9 IU/mL (approximately 2×), 73.7 4×), 187 (11×); EDTA, 3.8 mg/mL 7.8 19.8 hemoglobin, mg/mL; 20%. (Table validating interfering do repeatability (within-run precision) precision (between-run EP05-A3 ILA28-A2 standard, respectively.26CLSIEvaluation Scholar,29CLSI Five (RS1 RS7) [RS5 RS6 (near point)] runs (different days). 252 assays pseudorandomized experimental order, where variables (three lots, Dx2 instruments, operators) duplicate. tests Shapiro-Wilk33Shapiro S.S. Wilk "An normality (complete specimens)".Biometrika. 1965; 52: 591-611Crossref testing, Grubbs34Grubbs Sample testing outlying observations.Ann Math Stat. 1950; 27-58Crossref outlier Brown-Forsythe35Brown Forsythe A.B. Robust equality variances.J Am Stat Assoc. 1974; 364-367Crossref (1701) heteroscedasticity thus evaluation mixed effect model, maximal likelihood. Repeatability duplicates all seven 0.27 (95% CI, 0.17–0.30) RS5, pooled 0.22 0.00–0.13). Intermediate variable [operator 0.07 0.00–0.13); 0.13 0.09–0.17); 0.01 0.00–0.06); instrument 0.00 0.00–0.04); 0.04–0.12)]; 0.28 0.25–0.30) 1, S3, 3A). largest 0.31 and, therefore, no exceeded maximally allowed 0.45, demonstrating consequently, imprecision stayed far below 10%. reproducibility standard26CLSIEvaluation Netherlands, Kingdom, States). RS4, RS7