FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W1575098853> ?p ?o ?g. }

• W1575098853 endingPage "445" @default.
• W1575098853 startingPage "438" @default.
• W1575098853 abstract "A 15-gene prognostic signature for early-stage, completely resected, non–small-cell lung carcinoma, (which distinguishes between patients with good and poor prognoses) was clinically validated in prior studies. To achieve operational efficiencies, this study was designed to evaluate the assay's performance in RNA-stabilized tissue as an alternative to the fresh-frozen tissue format originally used to develop the assay. The percent concordance between matched tissue formats was 84% (95% Wilson CI, 70%–92%), a level of agreement comparable to the inherent reproducibility of the assay observed within biological replicates of fresh-frozen tissue. Furthermore, the analytical performance of the assay using the RNA-stabilized tissue format was evaluated. When compared to an accredited reference laboratory, the clinical laboratory achieved a concordance of 94% (95% Wilson CI, 81%–98%), and there was no evidence of bias between the laboratories. The lower limit of quantitation for the target RNA concentration was confirmed to be, at most, 12.5 ng/μL. The assay reportable range defined in terms of risk score units was determined to be −4.295 to 4.210. In a large-scale precision study, the assay showed high reproducibility and repeatability. When subjected to a maximal amount of genomic DNA, a potential contaminant, the assay still produced the expected results. The 15-gene signature was confirmed to produce reliable results and, thus, is suitable for its intended use. A 15-gene prognostic signature for early-stage, completely resected, non–small-cell lung carcinoma, (which distinguishes between patients with good and poor prognoses) was clinically validated in prior studies. To achieve operational efficiencies, this study was designed to evaluate the assay's performance in RNA-stabilized tissue as an alternative to the fresh-frozen tissue format originally used to develop the assay. The percent concordance between matched tissue formats was 84% (95% Wilson CI, 70%–92%), a level of agreement comparable to the inherent reproducibility of the assay observed within biological replicates of fresh-frozen tissue. Furthermore, the analytical performance of the assay using the RNA-stabilized tissue format was evaluated. When compared to an accredited reference laboratory, the clinical laboratory achieved a concordance of 94% (95% Wilson CI, 81%–98%), and there was no evidence of bias between the laboratories. The lower limit of quantitation for the target RNA concentration was confirmed to be, at most, 12.5 ng/μL. The assay reportable range defined in terms of risk score units was determined to be −4.295 to 4.210. In a large-scale precision study, the assay showed high reproducibility and repeatability. When subjected to a maximal amount of genomic DNA, a potential contaminant, the assay still produced the expected results. The 15-gene signature was confirmed to produce reliable results and, thus, is suitable for its intended use. The number of new lung cancer cases in 2014 in North America is expected to be well above 225,000 (American Cancer Society, Surveillance Research, 2014, http://www.cancer.org/acs/groups/content/@research/documents/document/acspc-041776.pdf, last accessed January 13, 2015; Canadian Cancer Society, http://www.cancer.ca/en/cancer-information/cancer-type/lung/statistics/?region=pe, last accessed January 13, 2015) with non–small-cell lung carcinoma (NSCLC) accounting for 85% of these.1Molina J.R. Yang P. Cassivi S.D. Schild S.E. Adjei A.A. Non–small cell lung cancer: epidemiology, risk factors, treatment, and survivorship.Mayo Clin Proc. 2008; 83: 584-594Abstract Full Text Full Text PDF PubMed Scopus (2580) Google Scholar, 2Leighl N. Treatment paradigms for patients with metastatic non-small-cell lung cancer: first-, second-, and third-line.Curr Oncol. 2012; 19: S52-S58Crossref PubMed Scopus (103) Google Scholar The standard treatment for early-stage (I and II) NSCLC indicates adjuvant chemotherapy for stage II patients and high-risk stage IB patients.3National Comprehensive Cancer Network: NCCN Clinical Practice Guidelines in Oncology. Non-Small Cell Lung Cancer. Version 2.2014. Fort Washington, PA, National Comprehensive Cancer Network, 2014Google Scholar Because some stage II NSCLC patients with better prognosis could be spared the risks associated with adjuvant chemotherapy, and 30% to 40% of stage I patients with worse prognosis could benefit from it,4Nesbitt J.R. Putnam J.B. Walsh G.L. Walsh G.L. Roth J.A. Mountain C.F. Survival in early-stage non-small cell lung cancer.Ann Thorac Surg. 1995; 60: 466-472Abstract Full Text PDF PubMed Scopus (337) Google Scholar a prognostic biomarker for early-stage NSCLC could have high clinical utility. Recent research led to the development of a 15-gene prognostic signature that may serve this purpose. The signature was developed in a set of 62 early-stage NSCLC fresh-frozen (FF) tissue specimens with the aim to distinguish between patients with good and poor prognoses. The signature was subsequently validated in four large public NSCLC microarray data sets.5Zhu C.Q. Ding K. Strumpf D. Weir B.A. Meyerson M. Pennell N. Thomas R.K. Naoki K. Ladd-Acosta C. Liu N. Pintilie M. Der S. Seymour L. Jurisica I. Shepherd F.A. Tsao M.S. Prognostic and predictive gene signature for adjuvant chemotherapy in resected non-small-cell lung cancer.J Clin Oncol. 2010; 28: 4417-4424Crossref PubMed Scopus (357) Google Scholar Additionally, the prognostic value of the signature was confirmed in an independent cohort of 181 early-stage NSCLC cases using FF tissue specimens, where patients with a poor prognosis signature died at nearly twice the rate of those with a good prognosis.6Der S.D. Sykes J. Pintilie M. Zhu C.Q. Strumpf D. Liu N. Jurisica I. Shepherd F.A. Tsao M.S. Validation of a histology-independent prognostic gene signature for early-stage, non-small-cell lung cancer including stage IA patients.J Thorac Oncol. 2014; 9: 59-64Abstract Full Text Full Text PDF PubMed Scopus (195) Google Scholar To make the signature amenable to clinical use, this study investigated an RNA-stabilized tissue format as an alternative to the FF format, because the latter is known to pose significant operational challenges in the clinical setting. The results of this tissue platform agreement study are reported herein. Once the assay was found to yield concordant results in RNA-stabilized tissues, a series of analytical validation studies was performed in conformity with the Clinical Laboratory Improvement Amendments and the Clinical Laboratory Standards Institute guidelines.7CLSICLSI Document MM22. Microarrays for Diagnosis and Monitoring Infectious Diseases; Approved Guideline. Clinical and Laboratory Standards Institute, Wayne, PA2014Google Scholar, 8CLSICLSI document EP09-A2-IR. Method Comparison and Bias Estimation Using Patient Samples; Approved Guideline-Second Edition (Interim Revision). Clinical and Laboratory Standards Institute, Wayne, PA2010Google Scholar, 9CLSICLSI document GP29–A2. Assessment of Laboratory Tests When Proficiency Testing is Not Available; Approved Guideline-Second Edition. Clinical and Laboratory Standards Institute, Wayne, PA2008Google Scholar, 10NCCLSNCCLS document EP21-A. Estimation of Total Analytical Error for Clinical Laboratory Methods; Approved Guideline. NCCLS, Wayne, PA2003Google Scholar, 11NCCLSNCCLS document EP17-A. Protocols for Determination of Limits of Detection and Limits of Quantification; Approved Guideline. NCCLS, Wayne, PA2004Google Scholar The assay's accuracy, precision, sensitivity, specificity, and reportable range were examined. The analytical validation studies and their results, as well as the most pertinent quality control (QC) measures, are described herein. All data resulting from these studies can be found on the Gene Expression Omnibus repository website (http://www.ncbi.nlm.nih.gov/geo; accession number GSE63074). Matched FF and RNA-stabilized [RNAlater (RNAL) medium (Life Technologies, Foster City, CA)] tissues were obtained from 43 NSCLC patients for use in the tissue platform agreement study. RNA-stabilized tissues from 22 NSCLC patients were secured for use in the analytical validation studies. Tissues were obtained from both commercial (BioServe, Beltsville, MD) and research (Precision Therapeutics, Pittsburgh, PA) sources under protocols approved by Essex Institutional Review Board (commercial; Lebanon, NJ), Copernicus Group Independent Review Board (research; Durham, NC), and Western Institutional Review Board (research; Puyallup, WA). Patients provided written consent for the collection of tumor tissue at the time of surgery. Tissues included in these studies were required to be histologically confirmed as adenocarcinoma, squamous, or large cell carcinoma; to not have been previously treated with chemotherapy, targeted therapy, or radiation therapy; to contain a minimum of 20% tumor (as determined by evaluation of a hematoxylin and eosin−stained slide by a board-certified pathologist); to be collected and treated with RNAL medium within 1 hour of surgical resection; to remain in the solution for at least 24 hours at 4°C for adequate stabilization; and on receipt, to be removed from the reagent and frozen at −80°C until RNA extraction. Also, to ensure sufficient penetration of RNAL into tissues, a minimum ratio of RNAL to tissue volume of 10:1 was required with at least one dimension of tissue at 5 mm or less. Total RNA was isolated from each tissue specimen via a phenol–chloroform extraction from FF or RNAL tissues. Following isolation, the RNA was treated with DNase and purified. Specimens with a minimal concentration of 12.5 ng/μL and RNA Integrity Number not less than 3 proceeded to RNA amplification using the Ovation Pico WTA System V2 (NuGEN Technologies, San Carlos, CA). The resulting product was fragmented and labeled with the Encore Biotin Module (NuGEN Technologies). The cDNA targets were then hybridized to the HG-U133 Plus 2.0 GeneChips (Affymetrix, Santa Clara, CA). The microarray was processed and scanned on the GeneChip System 3000Dx version 2 (Affymetrix). The resulting microarray data were then subjected to a series of QC and processing steps as part of the following computational algorithm. First, reference-based robust multichip averaging (refRMA) normalization and summarization was performed using 86 reference arrays12Katz S. Irizarry R.A. Lin X. Tripputi M. Porter M.W. A summarization approach for Affymetrix GeneChip data using a reference training set from a large, biologically diverse database.BMC Bioinformatics. 2006; 7: 464Crossref PubMed Scopus (65) Google Scholar followed by the evaluation of housekeeping genes, spike in controls, background noise, percent present, and scaling factor. The risk score for array j was then calculated based on a fixed formula:sj=∑k=115γk(wjk−μk)/σk,(1) where wjk is the refRMA normalized gene expression value for array j, gene k, γk is a weighting coefficient associated with gene k, μk is the mean expression value for gene k in the reference set, and σk is the SD of the expression values for gene k in the reference set. The values of γk, μk, and σk are given in Table 1. Finally, the patient tumors were categorized as either high or low risk based on a pre-defined risk score cutoff of −0.1,5Zhu C.Q. Ding K. Strumpf D. Weir B.A. Meyerson M. Pennell N. Thomas R.K. Naoki K. Ladd-Acosta C. Liu N. Pintilie M. Der S. Seymour L. Jurisica I. Shepherd F.A. Tsao M.S. Prognostic and predictive gene signature for adjuvant chemotherapy in resected non-small-cell lung cancer.J Clin Oncol. 2010; 28: 4417-4424Crossref PubMed Scopus (357) Google Scholar{sj≤−0.10,low-risk categorysj>−0.10,high-riskcategory.(2) Table 1The Kolmogorov–Smirnov for Goodness-of-Fit Test P Value and the Coefficients by Gene and the Signature CoefficientsProbe setGene symbolFitted distributionKolmogorov-Smirnov testReference mean, μkReference SD, σkCoefficient, γk201243_s_atATP1B111.55 + 0.64tdf = 40.64711.4190.871−0.126208399_s_atEDN34.02 + 0.15tdf = 30.8034.0590.2860.076221591_s_atFAM64ANormal (5.20, 0.63)0.9525.2010.6360.261218881_s_atFOSL2Normal (9.02, 0.58)0.6939.0230.582−0.082202814_s_atHEXIM18.16 + 0.34tdf = 80.8098.1810.3960.313202490_atIKBKAP3.92 + 0.29tdf = 150.5543.9290.3100.032204584_atL1CAM3.24 + 0.15tdf = 30.0943.5620.7820.316204179_atMBNormal (6.21, 1.40)0.5026.2101.4080.301205386_s_atMDM28.36 + 0.40tdf = 30.6098.4960.7340.099206426_atMLANA3.50 + 0.27tdf = 180.6863.5130.2830.227210016_atMYT1L2.74 + 0.18tdf = 80.4922.7530.2070.354203001_s_atSTMN24.22 + 0.37tdf = 30.2234.4180.8080.112203147_s_atTRIM149.42 + 0.32tdf = 30.9679.4330.472−0.079202707_atUMPS4.71 + 0.31tdf = 30.5094.8130.5030.256219171_s_atZNF2364.60 + 0.36tdf = 70.9544.6170.4230.241 Open table in a new tab Each FF and RNAL tissue specimen was split into two pieces, creating two biological replicates per tissue format, per patient. Thus, for each patient, RNA was extracted from four tissue pieces (two FF, two RNAL), followed by microarray-based genomic profiling. The 15-gene signature was applied to each profile, generating a numerical risk score and a risk category (high, low) using methods previously described5Zhu C.Q. Ding K. Strumpf D. Weir B.A. Meyerson M. Pennell N. Thomas R.K. Naoki K. Ladd-Acosta C. Liu N. Pintilie M. Der S. Seymour L. Jurisica I. Shepherd F.A. Tsao M.S. Prognostic and predictive gene signature for adjuvant chemotherapy in resected non-small-cell lung cancer.J Clin Oncol. 2010; 28: 4417-4424Crossref PubMed Scopus (357) Google Scholar and outlined above. The level of agreement was evaluated within biological replicates of the FF tissue format, as well as between the tissue formats by comparing the averaged risk scores of the biological replicates for each format. Finally, the level of agreement between tissue formats was compared to the level of agreement within FF biological replicates. To determine the maximal amount of residual genomic DNA (gDNA) in the RNA that could result from protocol deviations during the DNase treatment step, total RNA was extracted from five NSCLC specimens. The following DNase treatment conditions were tested: operators, DNase incubation time, DNase concentration, and DNase reagent lot. Each extracted total RNA sample was separated into two aliquots. One aliquot was reverse transcribed using the High-Capacity cDNA Archive Kit (Life Technologies). The other aliquot was not reverse transcribed. Both aliquots were then tested by real-time quantitative PCR using the Eukaryotic 18S rRNA Endogenous Control and TaqMan Fast Universal PCR Master Mix (Life Technologies) on the ViiA-7 real-time quantitative PCR instrument (Life Technologies). The amount of residual gDNA present was ascertained using the following equation: gDNA% = 0.5Ct2−Ct1 × 100%, where Ct1 was generated from the reverse-transcribed aliquot and Ct2 was generated from the aliquot that was not reverse transcribed. The specimen with the highest percentage of gDNA was selected as the maximal scenario, which was 4.512 × 10−3%. gDNA was then evaluated as a potential interfering material in the assay. Two NSCLC specimens previously profiled by the assay, one high risk and one low risk, were selected. The following amounts of gDNA spike-ins were tested: 0×, 0.5×, and 1× the calculated maximal residual gDNA, as described above. For each specimen, five microarray replicate assays were generated at each of the spike-in concentrations. To ensure robust analytical performance of the assay, several QC steps were implemented along the assay process. For example, to mitigate the risk associated with insufficient tumor content in a specimen, a tumor content evaluation was performed by a pathologist for every specimen, immediately on arrival at the time of histological processing. The pathologist evaluated a hematoxylin and eosin-stained slide from the tissue specimen and determined the tumor cellularity (percentage of tumor present). Only specimens with tumor cellularity ≥20% proceeded to RNA extraction. This is an example of a QC measure at the pre-analytical stage. At the analytical stage, three QC metrics were selected to determine the success of amplification in this assay: cDNA yield, percent present, and poly-A results. The yield of cDNA was measured immediately after the amplification process. The positive controls, discussed below, had to generate a minimum cDNA yield of 5.0 μg (200 ng/μL in concentration). The same limit was also used for patient samples. The Affymetrix MAS5 QC algorithm was applied to assess whether a probe set was absent or present. The acceptable limits for percent present were determined to be 61.18% and 74.78%, based on 148 arrays, including the aforementioned 86-reference arrays used in the refRMA normalization step. All arrays used RNA extracted from RNAL NSCLC specimens. Thus, a percent present value outside the QC limits of 61.18% to 74.78% disqualified a patient result from being reported. Poly-A controls were spiked in at the beginning of target preparation and used to assess the overall success of the amplification and array processing. The poly-A controls lys, phe, thr, and dap were required to be called present and express at an increasing intensity. Two positive controls, a high positive control and a low positive control, were included in the assay QC system to cover the operational aspects. Both the high and low positive controls used RNA extracted from RNAL-preserved NSCLC patient tumor specimens. Total RNA from each control was run alongside commercial and validation specimens starting at cDNA synthesis. Each control had to generate a risk score within two SD of the validated expected average. The averages were determined through data collected across 11 different days and four operators, using two lots of the Ovation Pico WTA System V2 kit and three lots of the Encore Biotin Module, for a total of 41 replicates per control type. The current acceptable QC ranges for the high and low positive control risk scores are 0.841 to 2.537 and −1.538 to −0.019, respectively. A no-template control (NTC) was selected to be used as a negative control. Potential contamination from assay equipment and reagents can cause nonspecific amplification. This control verifies that patient specimens do not generate false results that could arise from such contamination. The NTC was included in each run of an amplification plate. Nuclease-free water was added in place of template RNA in the NTC well. The NTC subsequently was treated with the same reverse transcription and amplification reagents as other controls (high and low positive controls) and patients on the plate. The NTC was carried through the entire amplification process, with the resulting cDNA yield assessed after amplification. NTC samples had to measure <46.94 ng/μL of cDNA to pass QC. This NTC control QC limit was determined based on 12 NTC assays run over a period of 4 months across three operators using two lots of the Ovation Pico WTA System V2 kit. A failed NTC control would disqualify results from all patient samples in the same run from being released, and amplifications from these patient samples would be repeated after the laboratory was decontaminated. The strength of agreement between the assay results obtained from matched FF and RNAL tissues was estimated by the Pearson correlation coefficient, the percentage of concordance, sensitivity, specificity, and the McNemar's test.13CLSICLSI document EP12–A2. User Protocol for Evaluation of Qualitative Test Performance; Approved Guideline-Second Edition. Clinical and Laboratory Standards Institute, Wayne, PA2008Google Scholar All confidence intervals for proportions throughout the article are reported as Wilson score intervals, because the Wilson score interval has a coverage probability closest to the nominal level when compared to the exact and Wald intervals.14Agresti A. Coull B. Approximate is better than “exact” for interval estimation of binomial proportion.Am Stat. 1998; 52: 119-126Google Scholar For both analyses, the two replicates within each tissue format were averaged. The strength of agreement between the replicates of FF tissues within each specimen were compared using the same statistical methods, but based on single replicate values. Reportable range was calculated as “the range of test result values over which the laboratory can establish or verify the accuracy of the instrument, kit, or test system measurement response.”7CLSICLSI Document MM22. Microarrays for Diagnosis and Monitoring Infectious Diseases; Approved Guideline. Clinical and Laboratory Standards Institute, Wayne, PA2014Google Scholar,p.10 To obtain a satisfactory estimate of the range of test result values, the targeted population must be assayed sufficiently well. Because the latter was outside the scope of this study, the range of risk scores theoretically possible under the signature is reported.15Plamadeala V. Huang S. McCreary S.M. Reitze N.J. Ewing A.L. Gabrin M.J. Bennett A.E. Mulligan J.M. Wilson C.L. Wang D. Analytical performance of a formalin-fixed paraffin-embedded tissue-based 634-probe prognostic assay for predicting outcome of patients with stage II colon cancer.Appl Immunohistochem Mol Morphol. 2014; 22: 308-316Crossref PubMed Scopus (2) Google Scholar In the computational algorithm of the assay, the expressions for the 15 genes in the signature were obtained by the refRMA normalization approach, with the reference set composed of 86 assays on NSCLC patient specimens in the RNAL tissue format. Maximum likelihood was implemented to fit the parametric distribution for each gene by using the fitdistr function from the package MASS in R software version 3.2.0 (http://cran.r-project.org). Based on the Shapiro-Wilks test, the histograms for all genes were symmetric; although not all of the 15 underlying distributions appeared Gaussian. If the Shapiro-Wilks test P value for the expression of a particular gene was <0.10, a t-distribution was used; otherwise, a Gaussian distribution was fitted to the data. As a final check for deviation from the obtained fit, the Kolmogorov-Smirnov test was applied as a goodness-of-fit test. The final fitted distributions and the Kolmogorov-Smirnov test for goodness-of-fit P value for each gene are reported in Table 1. After fitting univariate distributions to the expressions of each gene, the Cholesky's decomposition was applied to estimate the variance–covariance matrix. The expressions for the 15 genes were then simulated based on the multivariate model. The simulation was iterated 1,000,000 times, and the assay risk score was computed for all iterations. The reportable range of the assay was set at the 0.05 and 99.95 percentiles of this simulated risk score distribution. In accordance with the Clinical Laboratory Standards Institute guidelines EP09-A2 and GP29-A2, Deming regression was used to detect bias between the clinical and the reference laboratories.8CLSICLSI document EP09-A2-IR. Method Comparison and Bias Estimation Using Patient Samples; Approved Guideline-Second Edition (Interim Revision). Clinical and Laboratory Standards Institute, Wayne, PA2010Google Scholar, 9CLSICLSI document GP29–A2. Assessment of Laboratory Tests When Proficiency Testing is Not Available; Approved Guideline-Second Edition. Clinical and Laboratory Standards Institute, Wayne, PA2008Google Scholar The slope (for detecting proportional bias) and intercept (for detecting constant bias) of this regression were reported. In addition, Pearson correlation and percent concordance were used to measure the strength of agreement between the risk scores obtained in the two laboratories. Because the assay was executed in a single laboratory, the variability-causing factors examined in the precision study are limited to different operators, days, and production lots of two critical reagents. The precision study consisted of two substudies. The first substudy determined the variability due to days and operators using three high- and three low-risk patient specimens. To evaluate the variability that was due to technical error, ie, the repeatability of the assay, one of the three specimens in each risk group was run in triplicate by each of three operators, across three days. The remaining two specimens had only one replicate per day per operator. This substudy was processed entirely with one lot of the Ovation Pico WTA System V2 and one lot of the Encore Biotin Module. The second substudy assessed the variability introduced by different lots of the two critical reagents. One high-risk patient and one low-risk patient were selected to be profiled for this study. Each patient had nine replicates amplified using a single lot of Ovation Pico WTA System V2, and nine additional replicates amplified using a second lot of Ovation Pico WTA System V2. Of the nine replicates amplified per Ovation Pico WTA System V2 lot, three replicates were fragmented and labeled using a single lot of Encore Biotin Module, across a total of three lots. This substudy was performed on a different day from the first substudy. The same design was used for low-risk and high-risk specimens in each substudy. A sample size of 63 assays was obtained in each risk group. The precision study results were analyzed in terms of the categorical assay calls. The overall interrun (reproducibility) concordance percent was computed as the average pairwise replicate concordance across different runs within a specimen, defined in Frampton et al16Frampton G.M. Fichtenholtz A. Otto G.A. Wang K. Downing S.R. He J. et al.Development and validation of a clinical cancer genomic profiling test based on massively parallel DNA sequencing.Nat Biotechnol. 2013; 31: 1023-1031Crossref PubMed Scopus (1449) Google Scholar as follows:2×∑ici∑ij(rij1+rij2)×100%,(3) where ci is the number of concordant pairs of correct calls for the ith specimen, rij1 = {0,1} is the number of correct calls detected in replicate 1 for the ith specimen, jth pair, and rij2 = {0,1} is the number of correct calls detected in replicate 2 for the ith specimen, jth pair. In the denominator, all possible pairs within a specimen are considered. Similarly, the overall intrarun (repeatability) concordance percent was computed as the average pairwise replicate concordance across different replicates, within a specimen and within a run:2×∑ijcij∑ijk(rijk1+rijk2)×100%,(4) where cij is the number of concordant pairs of correct calls for the ith specimen, jth run, rijk1 = {0,1} is the number of correct calls detected in replicate 1 for the ith specimen, jth run, kth pair, and rijk2 = {0,1} is the number of correct calls detected in replicate 2 for the ith specimen, jth run, kth pair. The inter- and intrarun concordance estimates were computed by risk level. Following the National Committee on Clinical Laboratory Standards EP-17 guideline, the assay's analytical sensitivity can be defined by the lower limit of quantification (LLOQ) of the assay analyte.11NCCLSNCCLS document EP17-A. Protocols for Determination of Limits of Detection and Limits of Quantification; Approved Guideline. NCCLS, Wayne, PA2004Google Scholar Treating the total extracted RNA as the analyte, the assay requires the target RNA concentration of 12.5 ng/μL (50 ng total) be used as the input for the amplification process. The purpose of this study was to show that the assay does not yield different risk scores at concentrations higher than the requirement of 12.5 ng/μL. In other words, the objective was to show that the LLOQ for the assay target RNA concentration is at most 12.5 ng/μL. In particular, three different concentrations were studied: 10.5 ng/μL, 12.5 ng/μL, and 14.5 ng/μL, with each level assessed in five replicates. The effect of different RNA concentrations on the risk scores was tested through a one-way analysis of variance model. The total RNA input for each concentration level was 42 ng, 50 ng, and 58 ng, respectively. Three gDNA concentrations were studied with five microarray replicate assays generated at each of three spike-in concentrations, for two specimens. A full, two-way analysis of variance of the risk scores versus specimen, gDNA concentration, and an interaction term was fit to the study results. The hypothesis about an effect of gDNA concentration on the risk scores was tested. The percent concordance in risk category between the matched FF and RNAL tissue assays was 84% (Table 2) (95% Wilson CI, 70%–92%). The sensitivity of the assay using the RNAL platform with respect to the FF platform was 90% (95% Wilson CI, 0.69–0.97), and the specificity was 79% (95% Wilson CI, 0.60–0.91) with a Pearson correlation of the risk scores of 0.74 (Figure 1) (95% CI, 0.63–0.85). This level of agreement is comparable to the inherent reproducibility of the assay observed within biological replicates of FF tissue, which had a concordance of 79% (Table 3) (95% Wilson CI, 65%–89%) and Pearson correlation of 0.83 (Figure 2) (95% CI, 0.74–0.92). Although the observed specificity of the RNAL platform with respect to the FF platform is lower than the observed sensitivity, suggesting the assay is sensitive to the selection of platform in the low-risk category, the exact McNemar's test for a difference in risk classification had a P value of 0.5 in the case of the FF versus RNAL tissue comparison, and 0.2 in the case of the FF replicate comparison. There is no evidence of a difference in risk classification in either case. This indicates that the assay results are in a high level of agreement between the two platforms and within the FF platform; however, further evaluation is warranted as more data are accumulated.Table 2Concordance in Risk Category between Matched FF and RNAL TissuesRNALHigh riskLow riskFF High risk172 Low risk519n = 43.FF, fresh frozen; RNAL, RNAlater. Open table in a new tab Table 3Concordance in Risk Category between Replicates of FF TissueFF replicate 2High riskLow riskFF replicate 1 High risk132 Low risk721n = 43.FF, fresh frozen. Open table in a new tab Figure 2Risk scores from biological replicates of fresh frozen (FF) tissue from non–small-cell lung carcinoma patients (n = 43). Each open circle represents the risk score of each FF tissue within each of its biological replicates. The dotted lines represent the clinically validated threshold between low (≤−0.1) and high (>−0.1) risk categories. The solid line is the 45° line.View Large Image Figure ViewerDownload Hi-res image Download (PPT) n = 43. FF, fresh frozen; RNAL, RNAlater. n = 43. FF, fresh frozen. The reportable range of the assay was determined to be −4.295 to 4.210 risk score units. An assay with a risk score outside this range would not be reported. Thirty-four pairs of assays were produced for the accuracy study: within each pair one assay was run in the clinical laboratory (Helomics Corporation, Pittsburgh, PA) and one in the reference laboratory (Almac Diagnostics, Craigavon, Northern Ireland, UK). Each of the 34 arrays satisfied the QC criteria in both laboratories. The assay risk scores are displayed in Figure 3 and Table 4. The assays run in the clinical laboratory were compared to those in the reference laboratory via Deming regression. Because the regression slope was not statistically different from 1 and the intercept was not statistically different from 0, there was no evidence of either a proportional or a constant bias between the two laboratories. Furthermore, the Pearson correlation between the risk scores in the two laboratories was 0.88 (95% CI, 0.77–0.94), and the percent concordant categorical assay results was 94% (95% Wilson CI, 81%–98%).Table 4Concordance in Calculated Risk Category between the Reference and Productions LaboratoriesReference laboratoryHigh riskLow riskClinical laboratory High risk200 Low risk212n = 34. Open table in a new tab n = 34. The interrun concordance for the high-risk group was 100%, and for the low-risk group, it was 99.3%. The intrarun concordance for the high-risk group was 100%, and for the low-risk group, it was 97.6%. Near-perfect agreement among multiple replicates within a run and between runs renders the assay highly repeatable and reproducible. No evidence was found that the mean risk scores differ across the studied concentration levels (F-test P value = 0.6). Thus, the assay LLOQ for the total RNA is at most 12.5 ng/μL. No evidence was found that gDNA has an effect on the risk score values at the tested concentration levels (F-test P value = 0.5), concluding that gDNA is not an interference material in the assay. The 15-gene prognostic signature is a complex multianalyte molecular assay that classifies early-stage, completely resected, and untreated NSCLC patients as having either good or poor prognosis. After confirming its prognostic value in several studies,5Zhu C.Q. Ding K. Strumpf D. Weir B.A. Meyerson M. Pennell N. Thomas R.K. Naoki K. Ladd-Acosta C. Liu N. Pintilie M. Der S. Seymour L. Jurisica I. Shepherd F.A. Tsao M.S. Prognostic and predictive gene signature for adjuvant chemotherapy in resected non-small-cell lung cancer.J Clin Oncol. 2010; 28: 4417-4424Crossref PubMed Scopus (357) Google Scholar, 6Der S.D. Sykes J. Pintilie M. Zhu C.Q. Strumpf D. Liu N. Jurisica I. Shepherd F.A. Tsao M.S. Validation of a histology-independent prognostic gene signature for early-stage, non-small-cell lung cancer including stage IA patients.J Thorac Oncol. 2014; 9: 59-64Abstract Full Text Full Text PDF PubMed Scopus (195) Google Scholar the assay was successfully adapted to employ an RNA-stabilized tissue format in lieu of the FF format. In the clinical setting, the FF tissue format may impart operational challenges, primarily as a result of difficulties in the FF tissue preparation and transportation. Using tissue preserved in an RNA stabilization reagent will circumvent these challenges and facilitate its availability to patients. To ensure robust performance in the clinical setting, the assay was evaluated in a series of analytical validation studies. To our knowledge, among the existing prognostic signatures for NSCLC reported in the literature,17Katz J.R. He J. Van Den Eeden S.K. Zhu Z.H. Gao W. Pham P.T. Mulvihill M.S. Ziaei F. Zhang H. Su B. Zhi X. Quesenberry C.P. Habel L.A. Deng Q. Wang Z. Zhou J. Li H. Huang M.C. Yeh C.C. Segal M.R. Ray M.R. Jones K.D. Raz D.J. Xu Z. Jahan T.M. Berryman D. He B. Mann M.J. Jablons D.M. A practical molecular assay to predict survival in resected non-squamous, non-small-cell lung cancer: development and international validation studies.Lancet. 2012; 379: 823-832Abstract Full Text Full Text PDF PubMed Scopus (261) Google Scholar, 18Tang H. Xiao G. Behrens C. Schiller J. Allen J. Chow C.W. Suraokar M. Corvalan A. Mao J. White M.A. Wistuba I.I. Minna J.D. Xie Y. A 12-gene set predicts survival benefits from adjuvant chemotherapy in non-small-cell lung cancer patients.Clin Cancer Res. 2013; 19: 1577-1586Crossref PubMed Scopus (195) Google Scholar, 19Wistuba I.I. Behrens C. Lombardi F. Wagner S. Fujimoto J. Raso M.G. Spaggiari L. Galetta D. Riley R. Hughes E. Reid J. Sangale Z. Swisher S.G. Kalhor N. Moran C.A. Gutin A. Lanchbury J.S. Barberis M. Kim E.S. Validation of a proliferation-based expression signature as prognostic marker in early stage lung adenocarcinoma.Clin Cancer Res. 2013; 19: 6261-6271Crossref PubMed Scopus (80) Google Scholar, 20Xie Y.1 Minna J.D. A lung cancer molecular prognostic test ready for prime time.Lancet. 2012; 379: 785-787Abstract Full Text Full Text PDF PubMed Scopus (12) Google Scholar the 15-gene assay is the only gene expression-based assay with validations of both its prognostic ability and analytical performance. The multivariate nature of the assay analyte, the binary specification of the assay result, and the novelty of the assay collectively posed some challenges in the exact application of standard analytical validation prescriptions. Nevertheless, the results obtained in the conducted analytical performance studies present a well-controlled process. Typically, the reproducibility and repeatability of a quantitative assay are expressed in terms of the CV percentage. Despite the one-to-one correspondence between the 15-gene assay risk scores and the binary assay calls, which would justify a precision study in terms of risk scores, the CV percent is not a suitable measure for quantifying this assay's repeatability and reproducibility, because the denominator in the CV percent can be very close to zero in the low-risk group. Instead, the categorical methods used in Frampton et al16Frampton G.M. Fichtenholtz A. Otto G.A. Wang K. Downing S.R. He J. et al.Development and validation of a clinical cancer genomic profiling test based on massively parallel DNA sequencing.Nat Biotechnol. 2013; 31: 1023-1031Crossref PubMed Scopus (1449) Google Scholar were used to describe the repeatability and reproducibility of the assay in terms of the binary calls. The precision study used a large sample size and yielded nearly perfect concordance estimates between the replicates. The drawback of these concordance estimators is the lack of appropriate confidence interval methodology. In the absence of a reference assay method or a reference laboratory offering the same assay, the analytical accuracy of the 15-gene assay was benchmarked against a laboratory that is well-recognized, Clinical Laboratory Improvement Amendment-registered, and accredited by the College of American Pathologists for performing microarray assays. No evidence of bias was found between the two laboratories, further supporting that the assay is performed in a controlled environment. The definition of the LLOQ and the evaluation methodology based on serial dilution is well understood for one-analyte assays. For a multiplex assay, such as the 15-gene signature, it is impractical to assess individual features using the typical dilution studies. However, there is a minimum total RNA concentration used as input for the amplification process that allows accurate measurements of both individual features and final assay results. Because the assay protocol calls for a target RNA concentration of 12.5 ng/μL (50 ng total), the concentration used to validate the signature clinically and analytically, the sensitivity of the assay was assessed by establishing 12.5 ng/μL as an upper bound on the LLOQ for the total RNA concentration. This is a nonstandard, yet reasonable way to evaluate the sensitivity of the assay. In the analytical specificity study, the assay was subjected to the presence of a maximal amount of residual gDNA contamination. During RNA extraction, some amount of gDNA may be coprecipitated. Because gDNA is a potential interference material, a subsequent DNase digestion step was performed. The specificity study found that although residual gDNA may be present as a result of protocol deviations during the DNase treatment step, it does not affect the assay results even in its maximal amount. Despite the complexities of the assay process, the 15-gene signature was shown to maintain consistent results when subject to different variability and bias factors, and is suitable for the intended use." @default.
• W1575098853 created "2016-06-24" @default.
• W1575098853 creator A5001678193 @default.
• W1575098853 creator A5008484301 @default.
• W1575098853 creator A5015224570 @default.
• W1575098853 creator A5022668934 @default.
• W1575098853 creator A5023495969 @default.
• W1575098853 creator A5026267250 @default.
• W1575098853 creator A5026902904 @default.
• W1575098853 creator A5027837366 @default.
• W1575098853 creator A5029960857 @default.
• W1575098853 creator A5031214249 @default.
• W1575098853 creator A5041136752 @default.
• W1575098853 creator A5056014309 @default.
• W1575098853 creator A5059189045 @default.
• W1575098853 creator A5060615213 @default.
• W1575098853 creator A5067127926 @default.
• W1575098853 creator A5073362806 @default.
• W1575098853 creator A5075347427 @default.
• W1575098853 date "2015-07-01" @default.
• W1575098853 modified "2023-09-26" @default.
• W1575098853 title "Analytical Performance of a 15-Gene Prognostic Assay for Early-Stage Non–Small-Cell Lung Carcinoma Using RNA-Stabilized Tissue" @default.
• W1575098853 cites W1987638453 @default.
• W1575098853 cites W2038934354 @default.
• W1575098853 cites W2081709843 @default.
• W1575098853 cites W2091596054 @default.
• W1575098853 cites W2111501574 @default.
• W1575098853 cites W2115157373 @default.
• W1575098853 cites W2158086631 @default.
• W1575098853 cites W2160981405 @default.
• W1575098853 cites W2167278653 @default.
• W1575098853 cites W2168266830 @default.
• W1575098853 cites W2316854742 @default.
• W1575098853 cites W4234236664 @default.
• W1575098853 doi "https://doi.org/10.1016/j.jmoldx.2015.03.005" @default.
• W1575098853 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/25960256" @default.
• W1575098853 hasPublicationYear "2015" @default.
• W1575098853 type Work @default.
• W1575098853 sameAs 1575098853 @default.
• W1575098853 citedByCount "8" @default.
• W1575098853 countsByYear W15750988532015 @default.
• W1575098853 countsByYear W15750988532017 @default.
• W1575098853 countsByYear W15750988532018 @default.
• W1575098853 countsByYear W15750988532020 @default.
• W1575098853 crossrefType "journal-article" @default.
• W1575098853 hasAuthorship W1575098853A5001678193 @default.
• W1575098853 hasAuthorship W1575098853A5008484301 @default.
• W1575098853 hasAuthorship W1575098853A5015224570 @default.
• W1575098853 hasAuthorship W1575098853A5022668934 @default.
• W1575098853 hasAuthorship W1575098853A5023495969 @default.
• W1575098853 hasAuthorship W1575098853A5026267250 @default.
• W1575098853 hasAuthorship W1575098853A5026902904 @default.
• W1575098853 hasAuthorship W1575098853A5027837366 @default.
• W1575098853 hasAuthorship W1575098853A5029960857 @default.
• W1575098853 hasAuthorship W1575098853A5031214249 @default.
• W1575098853 hasAuthorship W1575098853A5041136752 @default.
• W1575098853 hasAuthorship W1575098853A5056014309 @default.
• W1575098853 hasAuthorship W1575098853A5059189045 @default.
• W1575098853 hasAuthorship W1575098853A5060615213 @default.
• W1575098853 hasAuthorship W1575098853A5067127926 @default.
• W1575098853 hasAuthorship W1575098853A5073362806 @default.
• W1575098853 hasAuthorship W1575098853A5075347427 @default.
• W1575098853 hasBestOaLocation W15750988531 @default.
• W1575098853 hasConcept C104317684 @default.
• W1575098853 hasConcept C126322002 @default.
• W1575098853 hasConcept C142724271 @default.
• W1575098853 hasConcept C143998085 @default.
• W1575098853 hasConcept C146357865 @default.
• W1575098853 hasConcept C151730666 @default.
• W1575098853 hasConcept C2777546739 @default.
• W1575098853 hasConcept C2777714996 @default.
• W1575098853 hasConcept C502942594 @default.
• W1575098853 hasConcept C54355233 @default.
• W1575098853 hasConcept C67705224 @default.
• W1575098853 hasConcept C71924100 @default.
• W1575098853 hasConcept C86803240 @default.
• W1575098853 hasConceptScore W1575098853C104317684 @default.
• W1575098853 hasConceptScore W1575098853C126322002 @default.
• W1575098853 hasConceptScore W1575098853C142724271 @default.
• W1575098853 hasConceptScore W1575098853C143998085 @default.
• W1575098853 hasConceptScore W1575098853C146357865 @default.
• W1575098853 hasConceptScore W1575098853C151730666 @default.
• W1575098853 hasConceptScore W1575098853C2777546739 @default.
• W1575098853 hasConceptScore W1575098853C2777714996 @default.
• W1575098853 hasConceptScore W1575098853C502942594 @default.
• W1575098853 hasConceptScore W1575098853C54355233 @default.
• W1575098853 hasConceptScore W1575098853C67705224 @default.
• W1575098853 hasConceptScore W1575098853C71924100 @default.
• W1575098853 hasConceptScore W1575098853C86803240 @default.
• W1575098853 hasIssue "4" @default.
• W1575098853 hasLocation W15750988531 @default.
• W1575098853 hasLocation W15750988532 @default.
• W1575098853 hasOpenAccess W1575098853 @default.
• W1575098853 hasPrimaryLocation W15750988531 @default.
• W1575098853 hasRelatedWork W2009966535 @default.
• W1575098853 hasRelatedWork W2019988726 @default.
• W1575098853 hasRelatedWork W2029605690 @default.
• W1575098853 hasRelatedWork W2054378706 @default.

• next