SEARCH

SEARCH BY CITATION

Keywords:

  • EGFR;
  • KRAS;
  • lung cancer;
  • cytology;
  • fine-needle aspiration;
  • FNA;
  • endobronchial ultrasound-guided fine-needle aspiration;
  • EBUS

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. FUNDING SOURCES
  7. REFERENCES

BACKGROUND:

The aim of this study was to validate clinical utilization of routinely prepared cytology specimens for molecular testing to detect EGFR or KRAS mutations in lung cancer.

METHODS:

From September 2009 to April 2010, the authors collected 209 cytology specimens from patients with lung cancer at the MD Anderson Cancer Center Department of Pathology. The specimens included 99 cases of endobronchial ultrasound-guided (EBUS) fine-needle aspiration (FNA), 67 cases of computed tomography (CT)-guided FNA, 27 cases of body fluid, 10 cases of ultrasound-guided of superficial FNA, and 6 cases of other cytology specimens. DNA sequencing for EGFR exons 18-21 and KRAS codons 12, 13, and 61 was performed.

RESULTS:

The overall specimen insufficiency rate was low (6.2%). EBUS (4%) and body-fluid cases (3.7%) showed lower insufficiency rates than the other cases. Similar insufficiency rates were observed in smears (6.1%) and cell-block sections (6.4%). EGFR mutations were detected in 19.4% (34 of 175) of nonsmall cell lung carcinoma (NSCLC) with a significantly higher frequency in adenocarcinoma (29%, 29 of 100) than in nonadenocarcinoma (7%, 5 of 75, P = .002). KRAS mutations were detected in 23.6% (41 of 174) of NSCLCs with no statistical differences between adenocarcinoma (26%, 26 of 102) and nonadenocarcinoma (21%, 17 of 72, P = .86). Higher frequencies of EGFR mutations in exons 19 and 21 (65%) than in exons 18 and 20 were detected.

CONCLUSIONS:

Our findings support clinical utilization of routinely prepared cytology specimens, including EBUS, CT/US. FNAs and body fluid specimens, as a reliable source for molecular testing to detect EGFR or KRAS mutations in patients with NSCLC. Cancer (Cancer Cytopathol) 2011;. © 2011 American Cancer Society.

Lung cancer is the leading cause of cancer deaths in both men and women in the United States.1 Patients with nonsmall cell lung cancer (NSCLC), which accounts for approximately 85% of all lung cancers and is most frequently diagnosed in the advanced clinical stages, have a poor prognosis.2 Recent advances in targeted therapy have produced promising treatment responses in patients with advanced NSCLC.3, 4

Targeted therapy for lung cancer is currently applied to patients with NSCLC and mutations in epidermal growth factor receptor (EGFR). EGFR (c-erbB-1, HER-1) is a transmembrane receptor protein with an extracellular binding domain and a cytoplasmic tyrosine kinase domain. Ligand binding with transforming growth factor α (TGF-α) and epidermal growth factor (EGF) results in autophosphorylation of key tyrosine residues in the receptor cytoplasmic domain and further activation of downstream signaling events that trigger antiapoptosis, cell proliferation, angiogenesis, tumor invasion, and metastasis. Cancer cells with EGFR mutations not only possess a growth advantage over those with wild-type EGFR but also show increased sensitivity to the anti-EGFR tyrosine kinase inhibitors (TKIs).5 The TKIs, such as gefitinib and erlotinib, that can block ligand binding have produced favorable treatment responses in patients who have advanced NSCLC with mutated EGFR.6, 7 Furthermore, the presence of EGFR mutations can be used as a prognostic factor.8 EGFR mutations are most common in lung adenocarcinoma. The mutation frequencies vary among ethnic groups, being seen more often among Asians (20%-40%) than among Caucasians (5%-20%).9 Mutations in EGFR exons 18-21 occur in the majority, and these are usually deletions in exon 19 and point mutations in exon 21.9

Another marker frequently tested in patients with NSCLC is Kristen-Rous sarcoma virus (KRAS) gene. KRAS mutation may be a predictor for resistance to TKIs therapy in patients with NSCLC10, 11 and a prognostic marker,5 although it is still controversial whether KRAS can predict resistance to TKIs.5, 11, 12 A major challenge in evaluating TKI treatment efficacy in patients with advanced NSCLC is the complexity of EGFR and KRAS mutations, which vary by ethnicity, sex, and smoking history.3, 4 Resistance to TKI therapy is also associated with new mutations.13-15 Therefore, assessment of EGFR and KRAS status in patients with NSCLC is crucial to determine the benefits of targeted therapies using TKI.

For molecular testing in patients with NSCLC, histopathologic samples obtained by core-needle biopsy or surgical resection are most commonly used to ensure adequate tissue. Routinely acquired cytology specimens are not commonly used for molecular testing on the assumption that cytology material is most likely inadequate for such testing. Therefore, the use of cytology specimens for molecular testing in patients with NSCLC is largely absent from most clinical trials and not frequently used in daily practice. At MD Anderson, we have increasingly used cytology specimens for molecular testing when cytology material was the only specimen source available or the core-needle biopsy sample was insufficient for testing. With the increased use of endobronchial ultrasound-guided (EBUS) biopsy for staging and diagnosis of lung cancer, EBUS FNA cytology specimens may be the only tissue source for diagnosis and molecular testing.

Although attempts have been made to use routinely collected cytology specimens including fine-needle aspiration (FNA) biopsy and body fluids for molecular testing, most published studies have used cytology materials specifically collected and stored for molecular testing.16-19 Published studies describing the use of routinely collected cytology specimens, such as smears or cell blocks, for molecular testing in patients with NSCLC are few and usually consist of small cohorts.19-22 If routine cytology specimens could be used, then more invasive procedures performed to obtain sufficient tumor tissue for molecular testing could be reduced or avoided. To evaluate the clinical validity of using cytology specimens for molecular testing, we conducted a study using cytology specimens, mainly smears and cell-block material, in lung cancer patients for molecular testing to detect EGFR and KRAS mutations.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. FUNDING SOURCES
  7. REFERENCES

The internal review board at MD Anderson Cancer Center approved the study.

Cytology Specimens

From September 2009 to April 2010, we identified 209 consecutive patients with lung cancer for whom cytology specimens were collected and submitted for molecular testing at the MD Anderson Cancer Center Department of Pathology. There were 114 women and 95 men. The patient's ages ranged from 29 to 91 years with a mean age of 63 years. Ethnically, 179 patients were Caucasian, 15 were black, 8 were Asian, and 7 were Hispanic. Of the 209 specimens, 79% (166 of 209) were collected at MD Anderson Cancer Center, and 21% (43 of 209) were collected at outside institutions from those patients who were referred to MD Anderson Cancer Center for treatment.

The cytology specimens consisted of 99 EBUS FNAs, 67 CT-guided FNAs, 27 pleural/pericardial effusions, 10 ultrasound-guided or superficial FNAs, 1 bronchoalveolar lavage, and 1 bronchial washing. The sources of 4 specimens were uncertain. The CT-FNA biopsies included 48 cases of primary lung carcinoma and 19 cases of metastatic lung carcinoma in bone (8 cases), lymph node (4 cases), adrenal gland (3 cases), soft tissue (3 cases), and liver (1 case).

Cytology Specimen Preparation

Direct smears were prepared routinely as follows: slides used for direct smears were fixed in Carnoy solution and stained with Papanicolaou (Pap) stain. The regressive Pap staining method was performed by hand as follows: 95% alcohol (10 dips), tap water (10 dips), Harris hematoxylin (up to 1 minute), tap water (10 dips), 1% hydrochloric acid (1 quick dip), and tap water to rinse off the acid. After checking staining quality of cellularity microscopically, the slides were rinsed in 95% (10 dips, ×3), 100% alcohol (10 dips, ×2), xylene (10 dips, ×2), and then cover slipped.

For cases requiring immediate assessment, a modified quick-staining technique (progressive method) was used. Fixed glass slides were stained as follows: reagent water (10 dips), Harris hematoxylin (8 seconds), reagent water (10 dips), 95% alcohol (10 dips, ×2), EA-50 (2-3 minutes), 95% alcohol (10 dips), 100% alcohol (10 dips), and xylene (10 dips or until cleared).

For air-dried smears, the Diff-Quik stain (Stat Lab Medical Products, Lewisville, Texas) was used following the manufacturer's manual.

Cell-block sections were prepared routinely as follows: the rinse (RPMI 1640, Gibco brand; Invitrogen by Life Technologies, Carlsbad, California) from FNA or body fluid was centrifuged at 1500 rpm for 5 minutes. The cell pellet was mixed with an equal amount of 10% formalin and 95% ethanol and then centrifuged at 1500 rpm for 10 minutes. The button of concentrated material was folded in shark-skin filter paper and placed within a tissue cassette. The tissue cassette was placed in a specimen bucket with 10% formalin and processed in the histology lab. The sections from the paraffin-embedded block were stained with hematoxylin and eosin.

Specimen Selection for Molecular Testing

Cytology diagnoses of lung carcinoma were rendered by 13 cytopathologists at the MD Anderson Cancer Center Department of Pathology. Classification of lung carcinoma was based on morphological interpretation, immunoperoxidase studies (when needed), and patient history. Immunostains for thyroid transcription factor 1 (TTF-1) were performed in 64 cases. TTF-1 was positive in 30 cases of adenocarcinoma, 3 cases of nonsmall cell carcinoma not otherwise specified, and 7 cases of small cell carcinoma. Positivity for TTF-1 was a criterion contributing to adenocarcinoma classification.

Cytology specimens were reviewed by pathologists (J.S. and G.S.) to determine the quality and quantity of tumor tissue for molecular testing. Cytology specimens that were qualified for molecular testing usually contained at least 40% tumor cells, a threshold strongly recommended by our molecular diagnostics laboratory based on a test sensitivity of 5% to 10% tumor cells. The percentage of tumor cells in the smears and cell-block sections was estimated in samples with relatively abundant tumor. In samples with less abundant tumor (higher levels of benign cells), the percentage of tumor was determined by counting cells in representative high-power fields. Some equivocal cases were reviewed by 2 pathologists (J.S. and G.S.) for consensus. In exceptional cases, we have submitted samples with tumor in the 20% to 40% range but only after a very careful accounting of cells. This threshold for tumor concentration was maintained to minimize the risk of a false-negative result. Tumor groups on smears (99 cases) and cell-block sections (110 cases) were marked by pathologists (J.S. and G.S.) and sent to the MD Anderson Cancer Center Molecular Diagnostic Lab. Only cell-block samples were used for molecular testing on outside pathology cases from referred patients, except in unusual circumstances and with the consent of the referring pathologist. To evaluate the association between specimen cellularity and specimen adequacy, a semiquantitative estimation of cellularity was recorded as follows: sparse cellularity (<300 tumor cells), low cellularity (300-1000 tumor cells), normal cellularity (>1000 tumor cells) in specimens. The tumor cellularity was compared with specimen insufficiency.

Molecular Testing

Tumor tissue in the marked area was scraped from the glass slides under direct observation or under a dissecting microscope. DNA was extracted using a QIAmp DNA Mini kit (Qiagen, Valencia, California) or a PicoPure DNA extraction kit (Molecular Devices, Sunnyvale, California) according to the manufacturer's instructions.

EGFR and KRAS mutations were detected by using DNA sequencing. The sequencing of EGFR exons 18 to 21 was performed by using a BigDye Terminator v3.1 Cycle sequencing kit (Applied Biosystems) after a nested polymerase chain reaction (PCR) procedure as previously described.23 The sequencing of KRAS codons 12, 13, and 61 was performed by using Pyrosequencing PSQ96 HS System (Biotage AB, Uppsala, Sweden) after a conventional PCR as previously described.24

EGFR and KRAS results were reported by the molecular diagnostics laboratory as positive, negative, or inadequate for complete evaluation. Inadequate samples include samples with insufficient quantity of DNA for testing, poor quality of DNA for PCR amplification, and poor quality DNA sequences for analysis. Testing of inadequate samples was repeated at least once before they were classified as inadequate.

For the purposes of this study, an insufficient molecular testing result on a cytologic sample is defined as follows: 1) failure of PCR amplification in at least one of the EGFR exons (absent a mutation in any of the other EGFR exons) with negative or no KRAS testing; or 2) failure of PCR amplification in at least one of the KRAS codons (absent a mutation in any of the other KRAS codons) with negative or no EGFR testing. The frequencies of mutations of EGFR or KRAS were compared with cytologic diagnoses. The distribution of EGFR or KRAS mutations from patients with adenocarcinoma, nonsmall cell carcinoma (not otherwise specified), and squamous/ poorly differentiated carcinoma was recorded. Mutation frequencies in EGFR exons 18-21 and KRAS codons 12, 13, and 61 were also compared with cytologic diagnoses.

Statistical Analysis

Descriptive statistics were calculated; χ2 or Fisher exact tests were used to assess the association between categorical variables. P values (2-sided test) <.05 were considered significant. All statistical analyses were carried out using SAS 8.0 software (SAS Institute, Cary, North Carolina).

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. FUNDING SOURCES
  7. REFERENCES

Specimen Insufficiency Rate for Molecular Testing

There were 26 reported PCR failures (13 EGFR and 13 KRAS). According to our definition, 13 cases were classified as “specimen insufficiency” that cannot be used to direct clinical decisions. The cytological specimen insufficiency rates for EGFR and KRAS mutation are summarized in Table 1. The combined specimen insufficiency rate was 6.2% (13 of 209). EBUS specimens (4%, 4 of 99) and body-fluid specimens (3.7%, 1 of 27) showed low specimen insufficiency rates for molecular testing, followed by specimens collected using CT-guided FNA (7.5%, 5 of 67), and ultrasound-guided FNA/superficial FNA performed by cytopathologists (10%, 1 of 10). The specimens collected as bronchial washing, or bronchioloalveolar lavage, or uncertain origin showed a high insufficiency rate (33%, 2 of 6).

Table 1. Specimen Insufficiency Rates for EGFR or KRAS Molecular Testing by Cytology Specimen Type
SpecimensCase No.Insufficiency (%)
  1. Abbreviations: CT-FNA, computed tomography-guided fine-needle aspiration; EBUS, endobronchial ultrasound-guided; US, ultrasound.

EBUS994 (4)
CT-FNA675 (7.5)
Body Fluid271 (3.7)
US/Superficial FNA101 (10)
Others62 (33)
Total20913 (6.2)

The specimen insufficiency rates were 6.1% (6 of 99) in smears and 6.4% (7 of 110) in cell-block sections. A higher specimen insufficiency rate (16%, 7 of 43) was observed in specimens sent by referring institutions than in specimens acquired at MD Anderson (3.6%, 6 of 166; Table 2).

Table 2. Specimen Insufficiency Rates for EGFR or KRAS Molecular Testing by Cytology Preparation
SpecimensCase No.Insufficiency (%)
  1. Abbreviation: MDACC, MD Anderson Cancer Center.

Smears996 (6.1)
Cell block1107 (6.4)
Inside MDACC cases1666 (3.6)
Outside referral cases437 (16.0)

PCR Failure Rates and Tumor Cellularity

The estimation of tumor cellularity was conducted in 195 specimens with 22 failed PCRs for either EGRR or KRAS. Samples with normal cellularity (>1000 tumor cells) had a significantly lower insufficiency rate than samples with sparse cellularity (<300 tumor cells, P = .03; Table 3). The differences between sparse and low cellularity samples (P = .12) and between low and normal cellularity (P = .58) were not significant.

Table 3. Polymerase Chain Reaction Failure Rates by Tumor Cellularity
 Tumor Cellularity 
PCRSparse (%)Low (%)Normal (%)Total
  1. Abbreviation: PCR, polymerase chain reaction.

  2. Sparse is defined as <300 tumor cells; Low, 300-1000 tumor cells; and Normal, >1000 tumor cells.

Failed5 (29)6 (12)11 (8.7)22
Succeed12 (71)46 (88)115 (91.3)173
Total1752126195

Distribution of EGFR and KRAS Mutations

The cytology diagnoses were classified in the following categories: adenocarcinoma (107 cases); nonsmall cell carcinoma (NSC), not otherwise specified (43 cases); squamous cell carcinoma (SCC) / poorly differentiated carcinoma (PD) (36 cases including 5 cases of poorly differentiated carcinoma); and small cell carcinoma (23 cases). No mutations were detected in any of the 23 cases of small cell carcinoma. The distribution of EGFR or KRAS mutations in combined NSCLC is given in Table 4. Mutations in the EGFR genes were detected in 19.4% (34 of 175) cases of combined NSCLC. The highest frequency of EGFR mutations was observed in adenocarcinoma (29%, 29 of 100), followed by NSC (9.8%, 4 of 41), and SCC of PD (2.9%, 1 of 34); this difference was significant (P = .002).

Table 4. Distribution of EGFR or KRAS Mutations in Fine-Needle Aspiration Biopsies from Patients With Nonsmall Cell Lung Carcinoma
DiagnosesEGFRKRAS
 Mutation/Case No. (%)Mutation/Case No. (%)
  1. Abbreviations: NSC, nonsmall cell lung carcinoma, not otherwise specified; PD, poorly differentiated carcinoma; SCC, squamous cell carcinoma.

  2. P-value was calculated to compare the mutations between adenocarcinoma and nonadenocarcinoma.

Adenocarcinoma29/100 (29)26/102 (26)
NSC4/41 (9.8)16/40 (40)
SCC/PD1/34 (2.9)1/32 (3.1)
Total34/175 (19.4)41/174 (23.6)
P.002.86

Mutations in the KRAS genes were detected in 23.6% (41 of 174) combined NSCLC cases (Table 4). The highest KRAS mutation rate was observed in NSC (40%, 16 of 40), followed by adenocarcinoma (26%, 26 of 102), and then SCC of PD (3.1%, 1 of 32). The difference in the KRAS mutation rate between adenocarcinoma and nonadenocarcinoma was not significant (P = .86) (Table 4).

Distribution of Mutations in EGFR Exons and KRAS Codons in Lung NSCLC

The distribution of EGFR mutations in the 4 exons and KRAS mutations in the 3 codons are illustrated in Table 5. EGFR mutations occurred predominantly in single EGFR exons, but they appeared in 2 exons in 5 cases. The EGFR mutations were most frequently detected in exon 19 (6.1%, 11 of 180) and least frequently in exon 18 (2.2%, 4 of 181). Among cases of adenocarcinoma, EGFR mutations were most frequent in exon 21, followed by exons 19, 20, and 18 (Table 5). The frequencies of EGFR mutations in exons 19 or 21 were significantly different between adenocarcinomas and combined nonadenocarcinomas. No statistical differences were observed in EGFR exons 18 and 20 between the 2 groups (Table 5).

Table 5. Frequencies of EGFR or KRAS Mutation in Nonsmall Cell Lung Carcinoma
GenesAdenocarcinomaNSCSCC/PDTotalP
 No. (%)No. (%)No. (%)No. (%) 
  1. Abbreviations: NSC, nonsmall cell carcinoma, not otherwise specified; SCC, squamous cell carcinoma; PD, poorly differentiated carcinoma.

  2. EGFR mutation x 1 (27); mutations x 2 (5).

  3. KRAS mutation x 1 (41).

EGFR 183 (1.7)1 (0.5)04/181 (2.2).62
EGFR 1910 (5.6)1 (0.5)011/180 (6.1).01
EGFR 207 (3.9)1 (0.6)1 (0.6)9/177 (5.1).18
EGFR 2113 (7.3)0013/179 (5.1)<.001
KRAS 1227 (15.3)9 (5.1)1 (0.6)37/176 (21.0).01
KRAS 131 (0.6)2 (0.9)03/176 (1.7)1.0
KRAS 611 (0.6)001/175 (0.6)1.0

KRAS mutations mainly occurred in codon 12 (21%, 37 of 176), followed by codon 13 (1.7%, 3 of 176). Only 1 KRAS mutation in codon 61 was observed in the adenocarcinoma cases (0.6%, 1 of 175). All KRAS mutations occurred in only 1 codon. No mutations were observed simultaneously in both EGFR and KRAS genes. The frequency of KRAS mutations in codon 12 was significantly different between adenocarcinomas and nonadenocarcinomas (Table 5). No statistical differences in KRAS codons 13 or 61 mutations were observed between adenocarcinomas and combined nonadenocarcinomas (Table 5).

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. FUNDING SOURCES
  7. REFERENCES

In this study, we observed an overall high efficiency using cytology specimens for molecular testing to detect EGFR or KRAS mutation in lung cancer patients. Our data validate the use of routinely processed cytology specimens (including EBUS, CT/US-guided FNAs, and body fluids) for molecular testing. Specimen adequacy for smears and cell-block preparations were comparable. In addition, we found a good correlation between cytology diagnoses and EGFR or KRAS mutations in lung carcinoma, further validating the use of cytology specimens for molecular testing in lung cancer patients. Sufficient material with satisfactory testing results for EGFR or KRAS has been reported in cytology specimens collected in special fixatives or saline or frozen body fluids.16-18 Our data indicate that with careful review, routinely prepared cytologic samples can also be highly informative.

At MD Anderson, cytology specimen selection and collection are similar to those for core biopsy specimens in that we mark slides and scrape tumor tissue for DNA extraction. Laser dissection is not required to obtain sufficient material for the testing. Our study shows that the 2 major cytology preparations, smears and cell-block sections, are equally good for molecular testing Our data also indicate that samples with sparse cellularity (<300 tumor cells) have a higher PCR failure rate.

We found that EBUS and body-fluid samples produced higher sufficiency rates than CT-guided FNA. We speculate that specimen selection may be a factor for higher specimen insufficiency in CT-guided FNA specimens because most FNA material was preserved as a second-line tissue source for molecular testing in case core biopsy material was inadequate. At MD Anderson, concurrently collected FNA samples are frequently used for a preliminary diagnosis of NSCLC to guide the confirmatory core-needle biopsy. Therefore, a few passes of FNA with relatively limited material may result in suboptimal material for molecular testing. In contrast, EBUS FNA is often used to collect material for a final diagnosis, with the need for molecular testing understood by the bronchoscopist. Another concern using cytology specimens for molecular testing for EGFR and KRAS is the cellular heterogeneity of the specimen. To achieve the optimal specimen purity with tumor cells, we performed tissue mapping on the slides to define the location of tumor fragments or areas of relative tumor enrichment. Considerable care was taken to ensure sufficient tumor purity in the mapped areas. In addition, PCR assay was used to amplify the specific targeted sequences encompassing the mutated area in EGFR or KRAS genes to ensure specificity of the subsequent sequencing. To address this issue, we are currently studying the mutation rates for EGFR and KRAS in cytology specimens versus core biopsy and excisional specimens processed at MD Anderson for molecular testing. Our preliminary data indicate no significant difference in the mutation rates between samples obtained by cytology versus histology (data not shown). We observed good correlations between cytology classifications and the frequencies of EGFR and KRAS mutations. The highest and lowest mutation rates of EGFR were observed in adenocarcinoma and squamous carcinoma, respectively. Our data on further characterization of EGFR or KRAS mutation showed mutation patterns comparable to those published.25, 26 EGFR exons 19 and 21 were most frequently detected in our study, accounting for 65% of mutations in the tested EGFR exons. KRAS mutations in exon 1 (codons 21 and 13) were predominant in our study. In addition, EGFR and KRAS mutations in our cohort were mutually exclusive. These results are consistent with the literature.5, 27

In our study cohort, we observed a higher EGFR mutation rate in adenocarcinoma (29%) compared with those in published data. It is unclear whether this is the result of our methodology or reflects a bias due to the diverse patient population seen at MD Anderson or reflects a bias introduced by selective testing directed by the clinical staff.

In summary, our findings support clinical utilization of routine cytology specimens, including EBUS, CT/US FNA, and body fluids, as a reliable source for molecular testing to detect EGFR or KRAS mutations in lung cancer patients.

FUNDING SOURCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. FUNDING SOURCES
  7. REFERENCES

No specific funding was disclosed.

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. FUNDING SOURCES
  7. REFERENCES