Is the Mirror Image Method Really Useful in Tumor Tissue Bank Quality Control?

Abstract

Oncology research projects are highly dependent on the quality of tumor samples stored in the biobank. Microscopic control is important to ensure the quality of the frozen sample (Does the sample correspond to tumor tissue? Does the sample contain a sufficient number of tumor cells for molecular analysis?). The aim of this study was to evaluate the value of the mirror image method in quality control of colonic adenocarcinoma samples stored in a tumor bank. Microscopic concordance for the differentiation grade, malignant and normal cell percentages, necrosis, mucinous component, and ulceration was assessed on 82 colon adenocarcinoma banked samples and their paired, formalin-fixed, paraffin-embedded mirror controls. Molecular concordance for KRAS status was evaluated in 76 of these 82 cases. Morphological correspondence between frozen and mirror samples was good for the mucinous component (intraclass correlation coefficient [ICC] = 0.81), moderate for differentiation (Cohen's kappa coefficient [k] = 0.67), fair for malignant cells (ICC = 0.44), and poor for ulceration (k = 0.08), normal tissue (ICC = 0.36), and necrosis (ICC = 0.13) percentages. Molecular correspondence for KRAS status was almost perfect (95% correspondence, k = 0.88) between frozen and mirror samples. In conclusion, the mirror sample method is not a good alternative for microscopic and molecular control of frozen colonic adenocarcinoma samples.

Introduction

Colorectal cancer (CRC) is the third most common cancer diagnosed worldwide in men and the second in women. Despite advances in screening, diagnosis, and management of the disease, it remains the fourth cancer in terms of mortality.^1–3 Union for International Cancer Control (UICC) staging is the only prognostic classification used in clinical practice to select patients for adjuvant chemotherapy.⁴ However, histoclinical parameters only poorly predict the heterogeneity of disease and are insufficient for recurrence and prognostic prediction in an individual patient. Currently, CRC has relatively few established biomarkers to predict patient outcome. Among the molecular markers that have been extensively investigated for CRC characterization and prognosis, microsatellite instability is the only marker showing a significant prognostic factor in CRC.^1,2,5–7 RAS mutation status is used to decide the therapeutic protocol in metastatic CRC patients.^8–10 Molecular research on CRC aims to find markers involved in tumor progression, leading to a better understanding of the carcinogenesis process and thus allowing new prognostic markers and novel therapeutic targets to be identified. These types of studies require high-quality biospecimens. Thus, fresh frozen tissue can be useful because it produces higher yield and higher quality nucleic acids compared with formalin-fixed paraffin-embedded (FFPE) tissue, in which nucleic acids are fragmented.^11,12 However, even if nucleic acids from FFPE specimens had a poorer quality than nucleic acids from fresh frozen samples, FFPE biospecimens are still perfectly adequate for most tests, including next-generation sequencing.¹³

Biobanked fresh frozen tissues are increasingly used in medical research projects. Biobanks can provide researchers with a reliable and organized source of human tissue for RNA-based analysis. RNA integrity of biobanked specimens is important when applications involve gene expression studies such as Reverse Transcriptase-PCR (RT-PCR) and cDNA microarray. Using degraded RNA can lead to incorrect and unreproducible results, both in microarray experiments and real-time PCR. Lack of reproducibility in various studies raises concerns about variation in quality of tissues used for such studies.^14,15 Therefore, histological control of stored tissue is a crucial step before embarking on time-consuming, labor-intensive, and costly projects. Microscopic control is crucial to ensure that biospecimens are properly sampled. Indeed, up to 10% of biospecimens can be unusable due to sampling inadequacy (e.g., with no or insufficient percentage of malignant cells).^16–18 Moreover, histological evaluation of biospecimens enables an estimate of the malignant cell percentage to be performed. Reported studies on CRC have used tissue samples where more than 60% of the cells are tumor cells.¹⁹ When controlled, 27% to 90% of CRC samples fulfilled this criterion and are therefore suitable for molecular analysis.¹⁸

Morphological control of a biospecimen can be performed using one of three different methods: a frozen section, a cytology imprint, or a mirrored sample.¹⁷ Cutting a frozen section from the biospecimen is the reference method because it allows visualization of the stored biospecimen itself. However, two main drawbacks are attributable to this method. The first one is the need to at least partially thaw the sample to make the frozen section, thus potentially leading to nucleic acid and protein degradation. The second one is consumption of the sample, decreasing the quantity of available tissue. This can be important when samples are small. An additional disadvantage with this method is the risk of contaminating the extracted nucleic acid or protein with optimal cutting temperature compound, which can be a particular problem in proteomics (ion suppression in mass spectrometry).²⁰

Cytology imprint is a rapid, economic, and conservative method, but Boudou-Rouquette et al. show the poor efficiency of this method when assessing tumor cell percentage.¹⁷ The mirror image method involves taking a sample close to the frozen tissue specimen, then performing a conventional FFPE histological analysis. This technique allows very good morphological control of the tissue to be performed as well as immunohistochemical analysis without compromising the frozen sample. However, the mirror sample is tissue next to the frozen sample, not the frozen tissue itself. The microscopic correspondence between frozen and mirror samples has been reported to be excellent in prostate cancer.^21,22 In colonic adenocarcinoma, neither the morphological nor the molecular correspondence between frozen and mirror samples has been studied yet. Knowledge of these two parameters (microscopic and molecular correspondence) is required to establish quality control protocols for CRC biobanking.

The aim of this study was to assess the value of the mirror image method in quality control of colonic adenocarcinoma stored in a tumor bank.

Materials and Methods

Population studied

The study included frozen and paired mirror tissues sampled from 82 colectomy specimens received at the Pathology Department of Reims University Hospital (France) for histological diagnosis from 2006 to 2016 and stored in the Tumorothèque de Champagne Ardenne Biobank. The patients had not been treated with chemotherapy or radiotherapy before surgery. The study was performed in accordance with the ethical standards laid down in the Declaration of Helsinki. Written patient consent for biobanking and biospecimen use was obtained in all cases. Approval had been obtained from the Tissue Bank Management Board (authorization number DC-2008-374) for the study.

Tissue banking process

The frozen and mirror samples were collected from colectomy specimens as follows: a tumor sample was excised with a scalpel and then divided into two parts of the same size (mean 5 mm²). The cut sides of the two samples were not inked. Frozen samples were placed in labeled cryovials, snap-frozen in liquid nitrogen (−196°C), and then transferred to an ultralow-temperature freezer (−80°C) for storage. The mirror samples were fixed in 10% neutral buffered formalin for 6–24 hours and embedded in paraffin and then they underwent routine histopathological processing.

Morphological correspondence

Morphological assessment of fresh frozen samples was performed on 5-μm-thick frozen (−20°C) sections stained with hematoxylin–eosin. For mirror FFPE blocks, hematoxylin–eosin–saffron staining was performed on 3-μm-thick sections. No staining efficiency comparison was performed with respect to frozen and FFPE sections as both protocols were standard protocols. Neither the frozen nor the FFPE blocks had sections cut from them before this study, but a few sections had to be discarded before a section of sufficient quality for morphological analysis was cut. The following morphologic criteria were blindly quantitatively evaluated by two pathologists (A.M. and O.D.): percentage of tumor cells (TT), nontumor cells (NT), necrosis area, and mucinous component. Tumor cell percentage was calculated using the ratio of the number of tumor cell nuclei to the number of total cell nuclei. Presence or absence of ulceration and tumor differentiation grade according to World Health Organization (WHO) guidelines was also established.¹ In case of a discrepancy, a consensus diagnosis was reached after collegial discussion.

Molecular correspondence

DNA extraction and purification were performed on frozen and FFPE mirror samples using commercial kits (QIAamp DNA Micro Kit; Qiagen, Hilden, Germany and Maxwell16 FFPE Plus Tissue LEV DNA Purification kit, Promega, Madison, WI, respectively), according to the manufacturer's protocols. For each sample, an in-house pyrosequencing assay for KRAS exon 2 was realized: first, DNA was amplified by PCR using the PyroMark PCR kit (Qiagen) in a 25 μL final volume with the following primer set: 5′-AGAGAGGCCTGCTGAAAATGAC-3′ and 5′-Biotin-TAGCTGTATCGTCAAGGCACTCT-3′. The PCR program was as follows: a first step at 95°C for 15 minutes, then 38 cycles of 30 seconds at 95°C, 30 seconds at 58°C, 60 seconds at 72°C, a final step at 72°C for 10 minutes, and a hold at 8°C. The PCR generated an 86-bp amplicon comprising codons 12 and 13 of KRAS. Binding of biotinylated PCR products to Sepharose beads was performed with PyroMark Binding Buffer (Qiagen) and annealing of the sequencing primer (5′-CTTGTGGTAGTTGGAGCT-3′) to the DNA template with PyroMark Annealing Buffer (Qiagen). Then, the pyrosequencing reaction was performed using PyroMark reagents on a PSQ96MA analyzer (Qiagen). For each discordant frozen and mirror pair of samples, a real-time allele-specific PCR assay was performed using the KRAS mutation analysis kit (EntroGen, Woodland Hills, CA), according to the manufacturer's instructions, on a LightCycler 480 instrument (Roche, Rotkreuz, Switzerland). This assay is able to detect 1% mutation in a background of wild-type DNA and has been used for detection of the following nucleotide changes: c.34G>A, c.34G>C, c.34G>T, c.35G>A, c.35G>C, c.35G>T, and c.38G>A.

Statistical analyses

Quantitative variables are described as mean ± standard deviation and qualitative variables as number and percentage. Microscopic variable concordance between frozen and mirror samples was evaluated by using contingency tables and Cohen's kappa coefficient (k) or intraclass correlation coefficient (ICC), as appropriate. Interobserver reliability was assessed by using Cohen's kappa coefficient or ICC, as appropriate. Agreements evaluated with the kappa coefficient were considered slight for coefficients between 0.01 and 0.20, fair between 0.21 and 0.40, moderate between 0.41 and 0.60, substantial between 0.61 and 0.80, and almost perfect between 0.81 and 1.00. Agreements evaluated with ICC were considered poor when <0.40; fair to moderate between 0.40 and 0.59, good between 0.60 and 0.74, and excellent between 0.75 and 1.²³

RNA quality and quantity were compared by paired t-tests. Statistical analyses were performed with SAS, version 9.4 (SAS Institute, Inc., Cary, NC). For all tests, p < 0.05 was considered to be statistically significant.

Results

Microscopic correlation

Microscopic control of frozen samples revealed that five samples (6.1%) were not adenocarcinoma samples, but either adenoma remnants (n = 2) or normal colonic mucosa (n = 3). Mirror images of these five samples were concordant for the two adenoma cases (Fig. 1A, B), but were discrepant for the three other cases. In these three cases, mirror samples were adenocarcinoma samples with 20%–30% of malignant cells and the paired frozen sample did not contain tumor cells (Fig. 1C–F). Inversely, three mirror samples contained less than 5% tumor cells and the paired frozen sample contained 30%–50% tumor cells. In total, a major discrepancy between frozen and paired mirror samples was found in 6/82 cases (7.3%). A good morphological correlation, defined by similarities of all histological parameters between frozen and mirror samples, was found in 25/82 cases (30.5%), including the two adenoma cases (Fig. 1G, H).

FIG. 1.

Representative examples of microscopic comparisons between paired mirror and fresh frozen samples. (A, B) Adenoma remnant with perfect morphological correspondence between the paired mirror image and frozen section. (A) Mirror image (Hematein–Eosin–Saffron, × 2 magnification). (B) Frozen section (Hematein–Eosin, × 2 magnification). (C, D) A case of major morphological discrepancy between the paired mirror image and frozen section. (C) Well-differentiated adenocarcinoma on mirror sample (Hematein–Eosin–Saffron, × 2 and square × 20 magnification). (D) Normal colonic mucosa on frozen section (Hematein–Eosin, × 2 and square × 20 magnification). (E, F) Another case of major morphological discrepancy between the paired mirror image and frozen section. (E) Moderately differentiated adenocarcinoma on mirror sample (Hematein–Eosin–Saffron, × 2 and square × 20 magnification). (F) Necrosis of frozen section (Hematein–Eosin, × 2 and square × 20 magnification). (G, H) Adenocarcinoma with perfect morphological correspondence between the paired mirror image and frozen section. (G) Mirror image (Hematein–Eosin–Saffron, × 2 magnification). (H) Frozen section (Hematein–Eosin, × 2 magnification).

The mean percentage of tumor cells was 40.6% ± 20.4% in frozen samples and 38.4% ± 19.5% in mirror samples. The mean percentage of nontumor cells was 12.9% ± 24.4% in frozen samples and 16.7% ± 25.4% in mirror samples. The mean percentage of necrosis was 0.7% ± 2.5% in frozen samples and 2.5% ± 8.7% in mirror samples. The mean percentage of the mucinous component was 6.0% ± 18.1% in frozen samples and 8.3% ± 21.5% in mirror samples. Only 19 (23.2%) frozen samples and mirror samples had >60% tumor cells and <20% necrosis.

Morphological correlation between frozen and mirror samples assessed by ICC was excellent for mucinous component (0.81), fair for tumor cell (0.44), and poor for normal tissue (0.36) and necrosis (0.13) percentages. Figure 2 summarizes the correlation of tumor cell, mucinous component, and necrosis percentages between the paired frozen and FFPE samples (Fig. 2). Morphological resemblance between frozen and mirror samples assessed by k was substantial for differentiation grade (0.67) and poor for ulceration (0.08) (Table 1).

FIG. 2.

Scatter plot and regression line between malignant cell, necrosis, and mucinous component percentages in paired frozen and FFPE samples. (A) Scatter plot graphical representation and regression line of malignant cell percentage of paired frozen and FFPE colon adenocarcinoma samples. (B) Scatter plot graphical representation and regression line of necrosis percentage of paired frozen and FFPE colon adenocarcinoma samples. (C) Scatter plot graphical representation and regression line of mucinous component percentage of paired frozen and FFPE colon adenocarcinoma samples. FFPE, formalin-fixed paraffin-embedded.

Table 1.

Morphological Correspondence Between Frozen and Mirror Samples

	Frozen sample	Mirror sample	Correlation coefficient
Percentage of malignant cells, mean ± SD	40.6 ± 20.4	38.4 ± 19.4	0.44^b
Percentage of nontumoral tissue, mean ± SD	12.9 ± 24.4	16.7 ± 25.4	0.36^b
Percentage of necrosis, mean ± SD	0.7 ± 2.5	2.5 ± 8.7	0.13^b
Mucinous component, mean ± SD	6.0 ± 18.1	8.3 ± 21.5	0.81^b
Ulceration, present/absent	1/81	18/64	0.08^a
Differentiation, well/moderate/poor	16/49/8	12/50/11	0.67^a

Interpretation of Cohen's kappa: level of agreement:

0–0.20: slight, 0.21–0.40: fair, 0.41–0.60: moderate, 0.61–0.80: substantial, and 0.81–1.00: almost perfect.

Cohen's kappa.

ICC.

ICC, intraclass correlation coefficient; SD, standard deviation.

Molecular correspondence

KRAS mutational status was successfully determined in 76 adenocarcinoma cases for both frozen and mirror samples. Initial pyrosequencing analysis revealed 10 discrepant paired frozen and FFPE cases. Discrepant results were confirmed in four cases and quantitative PCR (qPCR) analysis enabled correction of false-positive pyrosequencing results in the other six specimens (in all six cases, the error was in the mirrored FFPE sample). Finally, a total correlation for KRAS status between frozen and mirror samples was found in 72/76 cases (95%) (Table 2). KRAS mutations were found in 23/76 cases in both frozen and FFPE mirror samples. The same mutations were found in these 23 paired frozen and FFPE samples: c.35G>A in nine cases (39%), c.35G>T in six cases (26%), c.38G>A in five cases (22%), c.35G>C in two cases (9%), and c.34G>A in one case (4%). The molecular correlation between frozen and mirror samples assessed by k was almost perfect (0.88). In four discrepant cases, KRAS mutations were found in frozen samples only and thus with two different analysis methods. The mutations only found in frozen samples were c.35 G > C in two cases and c.35 G > A in the two other cases. In these four cases, the percentage of tumor cells was ≥20% and concordant in both mirror and frozen samples.

Table 2.

Molecular Correspondence Between Frozen and Mirror Samples for KRAS Mutation

	KRAS wt mirror sample	KRAS mutated mirror sample	Total
KRAS wt frozen sample	49	0	49
KRAS mutated frozen sample	4	23	27
Total	53	23	76

Discussion

Molecular research on CRC is highly dependent on the quality of tumor samples stored in the biobank. Microscopic control is important to ensure that the stored CRC tissues are really tumor tissues and contain a sufficient number of tumor cells for molecular analysis. The value of the mirror sample method as a surrogate of the frozen sample for microscopic and molecular control was unknown. In this study, we evaluated the value of the mirror sample method for quality control of colonic adenocarcinoma samples stored in a tumor bank. We demonstrate the importance of performing microscopic control of CRC samples before molecular analyses. Indeed, 6% of frozen and mirror tumor samples were not found to be malignant. Thus, the use of uncontrolled biospecimens can lead to false and irreproducible results.

In our study, the morphological correspondence between paired frozen and mirror samples was satisfactory only for mucinous component percentage assessment (ICC = 0.81). The morphological correlation between paired frozen and mirror samples was insufficient for other microscopic criteria, particularly tumor cell and necrosis percentages. Moreover, in six cases (7.3%), a major discrepancy (nontumoral sample vs. tumoral sample) was found between paired frozen and mirror samples. Tumor cell and necrosis percentage criteria are important parameters to consider when deciding whether a particular sample should be included or excluded from a study. For example, The Cancer Genome Atlas requires samples to have more than 60% tumor cells and less than 20% necrosis.¹⁹ Up to 20% of tumor cells are required for mutational analyses by direct sequencing techniques.^24,25 Thus, the mirror sample method is not a good surrogate of a frozen section to assess tumor cell percentage and necrosis in CRC. The use of a mirror sample may result in inappropriate sample selection and therefore false results. The mirror sample method might be more efficient to assess the tumor cell percentage in tumors other than CRC. Li et al. showed that solid organ cancers such as renal cancer had higher tumor cell content than hollow organs.²⁶ In this type of tumor, the morphological agreement between frozen and mirror samples could be higher than in CRC samples—further investigations are needed. We acknowledge, however, in our study, the cut surfaces of each block that were closest to each other were not inked, meaning that there is good potential for mounting the frozen block in the wrong orientation and cutting and comparing surfaces that are more spatially distant and therefore potentially less concordant.

Our study also shows that frozen and mirror samples have discrepancies at the molecular level. Four of the 76 cases of our cohort had discordant KRAS status. In these four cases, KRAS mutations were found in frozen samples only, which could not be explained by tumor cell percentage. Formalin fixation could partly explain the discrepancy. Indeed, a lower frequency of KRAS mutation in FFPE samples compared with paired frozen samples has previously been reported in 6% of CRC cases with the high-resolution melting technique and 18% with Sanger sequencing.²⁷ DNA extracted from FFPE tissue harbors potential sequence artifacts such as nucleotide deamination that causes C:G>T:A changes^28,29 and can be associated with higher false-negative and false-positive mutation rates than frozen tissue.²⁷ Intratumoral heterogeneity could also partly explain discrepant KRAS results. Intratumoral heterogeneity for RAS mutation was found in 10% to 40% of primary or metastatic colorectal adenocarcinomas.^30,31 Coexistence of different RAS mutations has been described in 28% of metastatic colorectal carcinomas.³⁰ In our cohort, the 23 mutated cases had identical mutations with the use of frozen and paired FFPE samples. Intratumoral heterogeneity raises the question of which sample should be analyzed. Some authors recommend analyzing a DNA cocktail extracted from different tumor samples of the same patient.³⁰ This mixing technique could better reflect the overall mutational status of the patient, but it does not reflect the mutational status of the biobanked sample itself.

In conclusion, our study shows that performing microscopic control before molecular analyses of CRC tumor samples is a step that is crucial to avoid sampling inadequacy. The microscopic control should be performed using the biobanked sample that will be used for molecular analyses. The mirror sample is not a good enough surrogate of the biobanked sample that will be used for molecular analysis.

Footnotes

Acknowledgments

The authors thank the Regional Platform of Innovative Biology (PRBI) for their excellent technical assistance. This study was funded by grants from the Centre Hospitalier Universitaire de Reims, Appel d'Offre Local 2015. Grant Numbers AOL 2015-11 (CBR). This study sponsor had no role in the study design and collection, analysis, and interpretation of data.

Author Disclosure Statement

No conflicting financial interests exist.

References

Bosman

, Carneiro

, Hruban

, Theise

. WHO Classification of Tumours of the Digestive System. Lyon: IARC Press; 2010.

Stewart

, Wild

. World Cancer Report Lyon, France: International Agency for Research on Cancer. World Health Organization; Lyon, France: iARC Press; 2014.

Ronnekleiv-Kelly

, Burkhart

, Pawlik

. Molecular markers of prognosis and therapeutic targets in metastatic colorectal cancer. Surg Oncol, 2016; 25:190–199.

Sobin

, Gospodarowicz

, Wittekind

. Colon and Rectum. TNM Classification of Malignant Tumours, 7th edition. West Sussex: Blackwell Publishing Ltd.; 2010; 100–105.

Popat

, Hubner

, Houlston

. Systematic review of microsatellite instability and colorectal cancer prognosis. J Clin Oncol, 2005; 23:609–618.

Bertagnolli

, Redston

, Compton

, et al. Microsatellite instability and loss of heterozygosity at chromosomal location 18q: Prospective evaluation of biomarkers for stages II and III colon cancer—A study of CALGB 9581 and 89803. J Clin Oncol, 2011; 29:3153–3162.

Gray

, Quirke

, Handley

, et al. Validation study of a quantitative multigene reverse transcriptase-polymerase chain reaction assay for assessment of recurrence risk in patients with stage II colon cancer. J Clin Oncol, 2011; 29:4611–4619.

Lech

, Słotwiński

, Słodkowski

, Krasnodębski

. Colorectal cancer tumour markers and biomarkers: Recent therapeutic advances. World J Gastroenterol, 2016; 22:1745–1755.

Gonzalez-Pons

, Cruz-Correa

. Colorectal cancer biomarkers: Where are we now?. Biomed Res Int, 2015; 2015:149014.

10.

Hutchins

, Southward

, Handley

, et al. Value of mismatch repair, KRAS, and BRAF mutations in predicting recurrence and benefits from chemotherapy in colorectal cancer. J Clin Oncol, 2011; 29:1261–1270.

11.

Shabihkhani

, Lucey

, Wei

, et al. The procurement, storage, and quality assurance of frozen blood and tissue biospecimens in pathology, biorepository, and biobank settings. Clin Biochem, 2014; 47:258–266.

12.

Hedegaard

, Thorsen

, Lund

, et al. Next-generation sequencing of RNA and DNA isolated from paired fresh-frozen and formalin-fixed paraffin-embedded samples of human cancer and normal tissue. PLoS One, 2014; 9:e98187.

13.

Einaga

, Yoshida

, Noda

, et al. Assessment of the quality of DNA from various formalin-fixed paraffin-embedded (FFPE) tissues and the use of this DNA for next-generation sequencing (NGS) with no artifactual mutation. PLoS One, 2017; 12:e0176280.

14.

Simeon-Dubach

, Burt

, Hall

. Quality really matters: The need to improve specimen quality in biomedical research. J Pathol, 2012; 228:431–433.

15.

Rudloff

, Bhanot

, Gerald

, et al. Biobanking of human pancreas cancer tissue: Impact of ex-vivo procurement times on RNA quality. Ann Surg Oncol, 2010; 17:2229–2236.

16.

Qualman

, France

, Grizzle

, et al. Establishing a tumour bank: Banking, informatics and ethics. Br J Cancer, 2004; 90:1115–1119.

17.

Boudou-Rouquette

, Touibi

, Boëlle

P-Y

, Tiret

, et al. Imprint cytology in tumor tissue bank quality control: An efficient method to evaluate tumor necrosis and to detect samples without tumor cells. Virchows Arch, 2010; 456:443–447.

18.

Galissier

, Schneider

, Nasri

, et al. Biobanking of fresh-Frozen human adenocarcinomatous and normal colon tissues: Which parameters influence RNA quality?. PLoS One, 2016; 11:e0154326.

19.

The Cancer Genome Atlas Network. Comprehensive molecular characterization of human colon and rectal cancer. Nature, 2012; 487:330–337.

20.

Shah

, Zhang

, Choi

, et al. Tissue proteomics using chemical immobilization and mass spectrometry. Anal Biochem, 2014; 469:27–33.

21.

Sircar

, Gaboury

, Ouadi

, et al. Isolation of human prostatic epithelial plasma membranes for proteomics using mirror image tissue banking of radical prostatectomy specimens. Clin Cancer Res, 2006; 12:4178–4184.

22.

Brimo

, Sircar

, Chevalier

, et al. Banking of fresh-frozen prostate tissue using the alternate mirror image protocol: Methods, validation, and impact on the pathological prognostic parameters in radical prostatectomy. Cell Tissue Bank, 2012; 13:631–638.

23.

Cicchetti

. Guidelines, criteria, and rules of thumb for evaluating normed and standardized assessment instrument in psychology. Psychol Assess, 1994; 6:284–290.

24.

Thunnissen

, Kerr

, Herth

, et al. The challenge of NSCLC diagnosis and predictive analysis on small samples. Practical approach of a working group. Lung Cancer, 2012; 76:1–18.

25.

Park

, Holmes-Tisch

, Cho

, et al. Discordance of molecular biomarkers associated with epidermal growth factor receptor pathway between primary tumors and lymph node metastasis in non-small cell lung cancer. J Thorac Oncol, 2009; 4:809–815.

26.

, Guo

, Sun

, Chen

. Histological assessment of tumor tissue samples via the mirror image method. Biopreserv Biobank, 2015; 13:25–30.

27.

Solassol

, Ramos

, Crapez

, et al. KRAS mutation detection in paired frozen and Formalin-Fixed Paraffin-Embedded (FFPE) colorectal cancer tissues. Int J Mol Sci, 2011; 12:3191–3204.

28.

, Dobravic

. Sequence artifacts in DNA from formalin-fixed tissues: Causes and strategies for minimization. Clin Chem, 2015; 61:64–71.

29.

Kim

, Park

, Ji

, et al. Deamination effects in formalin-fixed, paraffin-embedded tissue samples in the era of precision medicine. J Mol Diagn, 2017; 19:137–146.

30.

Richman

, Chambers

, Seymour

, et al. Intra-tumoral heterogeneity of KRAS and BRAF mutation status in patients with advanced colorectal cancer (aCRC) and cost-effectiveness of multiple sample testing. Anal Cell Pathol (Amst), 2011; 34:61–66.

31.

Jeantet

, Tougeron

, Tachon

, et al. High intra- and inter-tumoral heterogeneity of RAS mutations in colorectal cancer. Int J Mol Sci, 2016; 17:E2015.