CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2007;5:111–117
`
`Improved Fecal DNA Test for Colorectal Cancer Screening
`
`STEVEN H. ITZKOWITZ,* LINA JANDORF,* RANDALL BRAND,‡ LINDA RABENECK,§ PAUL C. SCHROY III,㛳
`STEPHEN SONTAG,¶ DAVID JOHNSON,# JOEL SKOLETSKY,** KRIS DURKEE,** SANFORD MARKOWITZ,‡‡
`AND ANTHONY SHUBER**
`
`*Department of Medicine and Oncological Sciences, Mount Sinai School of Medicine, New York, New York; ‡Gastrointestinal Division, Evanston Northwestern
`Healthcare, Evanston, Illinois; §Toronto Sunnybrook Regional Cancer Centre, University of Toronto, Toronto, Ontario, Canada; 㛳Gastrointestinal Division, Boston
`University School of Medicine, Boston, Massachusetts; ¶Gastrointestinal Section, Hines Veterans Affairs Hospital, Hines, Illinois; #Gastroenterology Division, Eastern
`VA Medical School, Norfolk, Virginia; **Exact Sciences Corporation, Marlborough, Massachusetts; ‡‡Howard Hughes Medical Institute and Department of Medicine,
`Case Western Reserve University, Cleveland, Ohio
`
`Background & Aims: Fecal DNA testing has shown
`greater sensitivity than guaiac-based occult blood tests
`for noninvasive colorectal cancer (CRC) screening. The
`prototype assay (version 1), which analyzed 22 gene mu-
`tations and DNA integrity assay (DIA), showed a sensitivity
`of 52% for CRC detection and a specificity of 94% in aver-
`age-risk individuals. The present study was conducted to
`determine the sensitivity and specificity of a second-gener-
`ation assay (version 2) that uses improved DNA stabiliza-
`tion/isolation techniques and a new promoter methylation
`marker. Methods: Forty patients with CRC and 122 sub-
`jects with normal colonoscopy provided stool samples to
`which DNA preservation buffer was added immediately.
`DNA was purified using gel-based capture, and analyzed for
`the original panel of 22 mutations, DIA, and 2 new pro-
`moter methylation markers. Results: By using DNA that
`was optimally preserved and purified from stool, the sensi-
`tivity of the prototype version 1 assay increased to 72.5%
`because of enhanced performance of DIA. Vimentin gene
`methylation alone provided sensitivity and specificity of
`72.5% and 86.9%, respectively. The optimal combination of
`vimentin methylation plus DIA resulted in 87.5% sensitivity
`and 82% specificity; cancers were detected regardless of stage
`or location. False-positive vimentin methylation was associ-
`ated with older age. Conclusions: An improved fecal DNA
`test that incorporates only 2 markers shows much higher
`sensitivity for CRC. The new assay is easier to perform and
`should be less costly, thereby facilitating its use for nonin-
`vasive CRC screening.
`
`S creening for colorectal cancer (CRC) is arguably the most
`
`effective intervention for preventing any cancer. Unfortu-
`nately, despite the recommendations of all major medical soci-
`eties, fewer than half of eligible individuals older than age 50
`have undergone CRC screening.1–3 In the United States,
`colonoscopy is being used increasingly as a primary screening
`tool because of its excellent diagnostic accuracy and ability to
`remove precancerous and early cancerous lesions. However, the
`availability of an accurate, noninvasive screening test might
`increase compliance with CRC screening guidelines by individ-
`uals who are reluctant to undergo more invasive tests, or situ-
`ations in which colonoscopy screening is not feasible or readily
`available.
`Several studies have shown the feasibility of detecting colon
`tumor–specific products in the stool.4 The markers in these
`studies represent alterations of genes involved in the predomi-
`nant chromosomal instability pathway (such as APC, p53, and
`
`K-ras), the microsatellite instability pathway (Bat-26), and mark-
`ers of abnormal apoptosis. Studies using stool samples from
`patients already known to have colon cancer, adenomas, or a
`normal colon report sensitivities of 62%–91% for CRC, 27%–
`82% for advanced adenomas, and specificities of 93%–96% in
`individuals with a normal colonoscopy.4,5 These encouraging
`data prompted a large, prospective, multicenter study in more
`than 4000 average-risk, asymptomatic individuals older than
`age 50. The results showed a higher sensitivity for detecting
`cancer with the fecal DNA test compared with Hemoccult II
`(Beckman Coulter, Fullerton, CA) (51.6% vs 12.9%, P ⫽ .003),
`with comparable specificity (94.4% vs 95.2%, respectively).6 De-
`spite superior sensitivity over Hemoccult II, the prototype fecal
`DNA test (version 1) revealed lower than expected sensitivity,
`which was owing to an unexpectedly low rate of positivity for
`the DNA integrity assay (DIA) component. In retrospect, it was
`learned that the suboptimal performance of DIA was a result of
`DNA degradation during transit of specimens to the laboratory,
`despite precautions such as immediate chilling of samples and
`rapid delivery by express courier.
`Since that time, pilot studies have shown that several tech-
`nical and conceptual advances could improve fecal DNA test-
`ing. First, adding a DNA-stabilizing buffer to the stool imme-
`diately on defecation was shown to prevent DNA degradation
`for several days and enhance the performance of DIA.7 Second,
`a gel-based DNA capture approach, rather than the original
`bead-based technology, allowed for enhanced extraction of
`DNA from stool.8 Finally, promoter methylation has become
`recognized as a key pathway by which colon cancers develop.9
`This epigenetic alteration is not detected by approaches that
`analyze for gene mutations. Vimentin, a gene that typically is
`considered a product of mesenchymal cells, is not methylated in
`normal colonic epithelial cells, but becomes highly methylated
`in colon cancer cell lines and in 53%– 83% of colon cancer
`tissues.10 Vimentin methylation also has been detected in the
`stool from 43 of 94 (46%; 95% confidence interval [CI], 36%–
`56%) patients with CRC vs 20 of 198 (specificity, 90%; 95%
`confidence interval [CI], 85%–94%) with a normal colonos-
`copy,10 suggesting that methylation markers might contribute
`to a fecal DNA assay panel.
`
`Abbreviations used in this paper: CI, confidence interval; CRC, colo-
`rectal cancer; DIA, DNA integrity assay; DY, locus D (5p21) and locus
`Y (LOC91199); HLTF, Helicase-like Transcription Factor; MSP, methyl-
`ation-specific polymerase chain reaction; NC, normal colonoscopy;
`PCR, polymerase chain reaction.
`© 2007 by the AGA Institute
`1542-3565/07/$32.00
`doi:10.1016/j.cgh.2006.10.006
`
`Geneoscopy Exhibit 1006, Page 1
`
`

`

`112 ITZKOWITZ ET AL
`
`CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 1
`
`These improvements of better DNA stabilization, enhanced
`DNA extraction, and use of gene-specific methylation have been
`incorporated into a second-generation fecal DNA test (version 2).
`The purpose of the present study was to determine the sensi-
`tivity and specificity of the newer version 2 assay for detection
`of CRC.
`
`Methods
`Study Design
`This study was designed in 2 phases. Phase 1 involved
`analyzing stool samples from approximately 50 patients with
`CRC and 200 patients with normal colonoscopy (NC) to define
`suitable DIA cut-off values and to determine optimal markers
`for the new assay. Phase 2, which is ongoing, was designed as a
`validation set in which an additional 125 patients with CRC
`and 200 patients with NC will be analyzed using the optimal
`marker panel from phase 1. Without knowing the performance
`of the new assay, we decided to analyze specimens from phase
`1 after 45 CRC and 150 NC patients were enrolled, which had
`a negligible effect on the initial estimations for setting cut-off
`points for the DIA assay. The findings presented herein repre-
`sent the results of phase 1.
`
`Source of Clinical Material
`Seven centers participated in this study, representing a
`spectrum of academic medical settings (community based to
`tertiary care). Each center obtained local institutional review
`board approval before beginning the study. The number of
`patients contributed by each site varied depending on when
`institutional review board approval was obtained, with a mean
`number of 24 stool samples per site (range, 8 – 42). Between
`January and September 2005, subjects who were 50 – 80 years of
`age were eligible for the study if they were found at the time of
`colonoscopy to have either CRC or NC. The latter group con-
`sisted of individuals in whom the bowel preparation was clas-
`sified as very good to excellent, the colonoscopy was complete
`to the cecum, and the mucosa was free of any type of mucosal
`lesion or polyps. Although they were younger than age 50, 4
`subjects (3 CRC, 1 NC) between the ages of 44 and 50 were
`included because they fulfilled all other eligibility criteria. In-
`dividuals were excluded if any of the following conditions
`applied: any contraindication to colonoscopy or conscious se-
`dation; personal history of, or coexistent, cancer except basal
`and squamous cell carcinomas of the skin; active therapy with
`chemotherapy or radiation therapy for a concurrent cancer;
`high-risk conditions such as familial adenomatous polyposis,
`hereditary nonpolyposis colorectal cancer, inflammatory bowel
`disease, and strong family history of CRC (2 or more first-
`degree relatives with CRC, or 1 or more first-degree relatives
`with CRC younger than age 50), personal history of colorectal
`polyps, prior colorectal resection for any reason, current preg-
`nancy, or lactation. The presence of gastrointestinal symptoms
`was not an exclusion criterion, although patients with NC were
`almost all asymptomatic and presented for routine screening.
`The preparation for, and performance of, colonoscopy was
`performed according to standard operating procedures at each
`site. The histologic diagnosis of CRC was verified by a board-
`certified pathologist. Cancers were staged according to the
`TNM classification. Left-sided cancers were defined as those
`arising at, or distal to, the splenic flexure.
`
`Sample Collection
`To avoid any possible effect of the colonoscopic bowel
`preparation on test results, each subject provided a single stool
`sample approximately 6 –14 days after colonoscopy. In the case
`of patients with CRC, the sample was provided before begin-
`ning the presurgical bowel preparation. Subjects were given
`detailed instructions and a special stool collection kit that is
`mounted on the toilet bowl. Immediately after defecation, sub-
`jects added 250 mL of a DNA-stabilizing buffer7 to a stool
`specimen of at least 50 g. Only 10 patients provided less than
`50 g of stool, and, of these, 3 subsequently provided an ade-
`quate second specimen. The specimen was shipped at room
`temperature overnight using a coded identifier provided by an
`external clinical research organization (Carestat Inc., Newton,
`MA) to keep the laboratory blinded to the clinical source. The
`clinical research organization was responsible for maintaining
`all of the clinical data files. The collection interval was defined
`as the number of hours from the time of defecation until the
`specimen arrived in the laboratory. Stool samples were pro-
`cessed and analyzed without knowledge of clinical information.
`The details of sample processing and human DNA purification
`have been described previously.7
`
`Version 1 Assay
`Samples were processed for 22 specific mutations ac-
`cording to Whitney et al8 using a gel-based DNA capture
`approach (Effipure; Exact Sciences Corporation, Marlborough,
`MA) with the following modifications: (1) DNA amplifications
`were increased to 60 cycles; (2) single base extension reactions
`included internal controls, that is, 0.5-umol/L internal control
`primers and 25 ng (mutant reactions) or 5 ng (wild-type reac-
`tions); (3) acyclopol enzyme was increased to 0.027 U/reaction;
`and (4) extension reactions were treated with 0.1 uL of shrimp
`alkaline phosphatase (SAP; Promega, Madison, WI) at 37°C
`for 30 minutes before analysis by capillary electrophoresis
`(Applied Biosystems 3100 instrument; Applied Biosystems,
`Foster City, CA).
`
`DNA Integrity Assay
`The DIA was performed using real-time polymerase
`chain reaction (PCR) as described previously.8 The assay was
`converted to a multiplex format in which 4 primer/probe pairs
`simultaneously interrogated the presence and quantity of 200-,
`1300-, 1800-, and 2400-bp human DNA fragments at 4 loci:
`5p21 (locus D), 17p13 (locus E), HRMT1L1 (locus X), and
`LOC91199 (locus Y).
`
`Methylation Assay
`Stool samples were processed for vimentin and Helicase-
`like Transcription Factor (HLTF) analysis according to Whiney
`et al8 by using the following capture sequences: vimentin
`(Vimcp50a: 5=- GGCCAGCGAGAAGTCCACCGAGTCCTGCAG-
`GAGCCGC -3=; Vimcp29b: 5=- GAGCGAGAGTGGCAGAGGACT-
`GGACCCCGCCGAGG -3=), and HLTF (methylation-specific poly-
`merase chain reaction [MSP]5cp: 5=-CAAATGAACCTGACC-
`TTCCCGGCGTTCCTCTGCGTTC-3=). Bisulfite conversion of
`DNA was performed as previously described.11,12 MSP PCR reac-
`tions were performed using 0.5-umol/L armed primers for either
`HLTF MSP-5 or vimentin MSP-29 (IDT, Coralville, IA). HLTF
`MSP-5 primer sequences have been reported previously.13 Modi-
`
`Geneoscopy Exhibit 1006, Page 2
`
`

`

`January 2007
`
`IMPROVED FECAL DNA SCREENING FOR COLON CANCER 113
`
`Table 1. Comparison of CRC Detection for Stool Samples Obtained Precolonoscopy and Postcolonoscopy (Version 1 Assay)
`
`Precolonoscopy6
`
`Postcolonoscopy
`
`No. positive/total
`
`% (95% CI)
`
`No. positive/total
`
`% (95% CI)
`
`Total
`Stage I
`Stage II
`Stage III
`Stage IV
`Unknown
`
`16/31
`8/15
`5/8
`3/8
`0
`—
`
`51.6 (34.8–68.0)
`53.3 (30.1–75.2)
`62.5 (30.6–86.3)
`37.5 (13.7–69.4)
`—
`—
`
`20/47
`7/16
`2/10
`7/11
`3/7
`1/3
`
`42.6 (29.5–56.7)
`43.8 (23.1–66.8)
`20.0 (5.7–51.0)
`63.6 (35.4–84.8)
`42.9 (15.8–75.0)
`33.3 (6.2–79.2)
`
`fied HLTF MSP-5 methylation-specific forward primers 5=-
`GACGTCTAACTAAACTCGCGA-3= and reverse primers 5=-TTT-
`TAGGTCGTTAGATCGAGC-3= were extended by a 5=
`tag
`sequence 5=-GCGGTCCCAATAGGGTCAGT-3=, which is not de-
`rived from the HLTF sequence, but which allows for more robust
`sequence-specific
`template amplification. Vimentin MSP-29
`primer sequences have been reported previously.10 Primers were
`combined with 1⫻ HotStar buffer, 1.25 U HotStar polymerase
`(Qiagen, Alameda, CA), 200 ␮mol/L deoxynucleoside triphosphate
`(Promega), and 10 ␮L (capture stool) DNA in a final volume of 50
`␮L. Cycling conditions were 95°C for 14.5 minutes followed by 40
`cycles of 94°C for 30 seconds, 57°C (HLTF), 68°C (vimentin
`methylated) or 62°C (vimentin unmethylated) for 1 minute, 72°C
`for 1 minute, with final 72°C for 5 minutes. Samples were visu-
`alized on 4% NuSieve 3:1 agarose (FMC, Rockland, ME) gels using
`a Stratagene EagleEye II (Stratagene, La Jolla, CA) still-image sys-
`tem. Samples were scored as positive if the PCR band intensity
`exceeded a previously determined level. Positive samples were re-
`peated in duplicate to confirm methylation status.
`
`Patient Satisfaction Questionnaire
`Individuals submitting a stool sample were asked to
`complete a brief, 6-item satisfaction questionnaire designed by
`the authors to assess satisfaction with the new stool collection
`kit. The questionnaire was made available to all co-investiga-
`tors, who then distributed it to participants without tracking.
`Completed questionnaires were mailed back anonymously
`without identifiers, so it was not possible to determine the
`response rate.
`
`Data Analysis
`Descriptive statistics were used to characterize the data.
`The sensitivities and specificities with 95% confidence intervals
`were computed for all markers. t tests and ␹2 tests comparing
`the CRC with the NC group were used to examine associations
`between patient characteristics (eg, sex, age, time since colonos-
`copy) or markers. P values less than .05 were considered signif-
`icant. SPSS (version 14; SPSS, Chicago, IL) was used for all
`analyses.
`
`Results
`Rationale for Postcolonoscopy Stool Samples
`In the previous multicenter study using version 1, pre-
`colonoscopy stool samples were collected from an asymptom-
`atic average-risk population undergoing screening colonos-
`copy.6 That study required 4404 patients to identify 31 cancers.
`During that study period, several of the same clinical centers
`
`also participated in a parallel study to collect postcolonoscopy
`stool samples from patients with CRC. Both sets of samples
`were processed in a blinded fashion alongside 1423 CRC-neg-
`ative samples using the version 1 assay. At that time, stools were
`collected without DNA stabilization buffer and DNA was ex-
`tracted by a bead-capture technique. As shown in Table 1, CRC
`detection for the precolonoscopy and postcolonoscopy stool
`groups was 51.6% (95% CI, 34.8%– 68.0%) and 42.6% (95% CI,
`29.5%–56.7%), respectively (not statistically significant). There
`was also no significant difference in detection frequency accord-
`ing to tumor stage, although the number of patients was small.
`If anything, postcolonoscopy stool samples showed a lower
`sensitivity, suggesting that such samples may underestimate the
`assay’s sensitivity. Thus, tumor manipulation or other potential
`factors at the time of colonoscopy do not appear to bias in favor
`of molecular CRC detection, suggesting that analysis of post-
`colonoscopy stool samples is adequate for estimating the per-
`formance of a fecal DNA test.
`
`Patient Population
`Initially, 45 patients in the CRC group and 150 subjects
`in the NC group were enrolled. Of the patients in the CRC
`group, 5 were excluded because of age younger than 40 years
`(n ⫽ 2), carcinoma in situ (n ⫽ 1), history of colitis (n ⫽ 1), and
`the finding of an adenoma instead of cancer (n ⫽ 1). Of the
`
`Table 2. Demographic Characteristics of the
`Study Population
`
`Colon cancer
`(N ⫽ 40)
`
`NC
`(N ⫽ 122)
`
`24 (60)
`38.9 (21.6)
`12.4 (6.2)
`
`65.6 (10.3)
`71.0 (7.4)
`67.1 (10.6)
`64.9 (10.0)
`56.8 (11.3)
`
`8 (20)
`10 (25)
`17 (43)
`5 (12)
`
`29 (73)
`
`62 (51)
`27.4 (5.2)
`10.2 (5.5)
`
`58.5 (7.2)a
`N/A
`N/A
`N/A
`N/A
`
`N/A
`N/A
`N/A
`N/A
`
`N/A
`
`Male, N (%)
`Mean collection interval, h (⫾SD)
`Days since colonoscopy
`(mean ⫾ SD)
`Age, y (mean ⫾ SD)
`Stage I
`Stage II
`Stage III
`Stage IV
`Stage of cancer, N (%)
`I
`II
`III
`IV
`Location of cancer, N (%)
`Left-sided
`
`aP ⫽ .03.
`
`Geneoscopy Exhibit 1006, Page 3
`
`

`

`114 ITZKOWITZ ET AL
`
`CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 1
`
`Table 3. Sensitivity and Specificity of Version 1 Assay
`
`Imperiale et al6
`
`Present studya
`
`No. positive/total
`
`% (95% CI)
`
`No. positive/total
`
`% (95% CI)
`
`Sensitivity
`All version 1 markers
`MuMu22
`DIA
`Specificity
`All version 1 markers
`MuMu22
`DIA
`
`16/31
`16/31
`1/31
`
`79/1423
`65/1423
`18/1423
`
`51.6 (34.8–68.0)
`51.6 (34.8–68.0)
`3.2 (0.6–16.2)
`
`94.4 (93.1–95.5)
`95.4 (94.2–96.4)
`98.7 (98.0–99.9)
`
`29/40
`17/40
`26/40
`
`13/122
`5/122
`10/122
`
`72.5 (57.2–83.9)
`42.5 (28.5–57.8)
`65.0 (49.5–77.9)b
`
`89.3 (82.6–93.7)
`95.9 (90.8–98.2)
`91.8 (85.6–95.5)
`
`aUsing stabilization buffer ⫹ gel capture (postcolonoscopy stool samples).
`bP ⬍ .0001.
`
`subjects in the NC group, 28 were excluded because of inade-
`quate colonoscopy preparation (n ⫽ 13), insufficient stool sam-
`ple (n ⫽ 7), strong family history of CRC (n ⫽ 2), personal
`history of polyps (n ⫽ 3), polyp found on colonoscopy (n ⫽ 2),
`and patient withdrawal (n ⫽ 1).
`Table 2 lists the demographic characteristics of the 40 CRC
`and 122 NC subjects studied. There were no significant differ-
`ences between the 2 groups in terms of sex, collection interval,
`or number of days after colonoscopy that the stool sample was
`collected. The NC group was younger than the CRC group (P ⫽
`.03). Among those with CRC, there was no difference in mean
`age according to cancer stage. Almost half of all cancers were
`early stage (I and II), and approximately three quarters of the
`CRCs were left-sided.
`
`Version 1 Marker Performance Using
`Optimized Sample Collection and
`Purification Techniques
`
`The same version 1 markers used in the previous mul-
`ticenter study6 were analyzed after the stool was collected using
`
`Table 4. Sensitivity and Specificity of DIA Combinations
`
`improved sample collection with DNA stabilization buffer7 and
`a gel-based DNA purification method.8 As shown in Table 3,
`the sensitivity of all version 1 markers was increased using the
`newer collection and purification methods from 51.6% (95% CI,
`34.8%– 68.0%) to 72.5% (95% CI, 57.2%– 83.9%; not statistically
`significant). The sensitivity of DIA when using buffer with
`gel-based purification was increased markedly from 3.2% (95%
`CI, 0.6%–16.2%) to 65% (95% CI, 49.5%–77.9%) (P ⬍ .0001).
`Specificity was not affected significantly by the newer collection
`and purification methods.
`
`Performance of the DNA Integrity Assay
`The DIA was analyzed according to a previously deter-
`mined requirement that 4 of 12 PCR fragments, excluding the
`200-bp fragments, must be greater than the individual frag-
`ment thresholds for a sample to be positive. Additional analyses
`involved all combinations of fragments, including the 200-bp
`fragment, to determine the optimal combination for maximum
`sensitivity and specificity. All combinations of DIA markers
`were similar with regard to sensitivity (60%– 65%), with speci-
`
`Sensitivity (N ⫽ 40)
`
`Specificity (N ⫽ 122)
`
`No. positive
`
`% (95% CI)
`
`No. positive
`
`% (95% CI)
`
`DEXY without 200 bp
`DEXY with 200 bp
`DEX
`DEY
`DXY
`EXY
`DE
`DX
`DY
`EX
`EY
`XY
`D
`E
`X
`Y
`
`26
`25
`24
`25
`25
`26
`25
`26
`26
`24
`24
`26
`25
`24
`25
`24
`
`65.0 (49.5–77.9)
`62.5 (47.0–75.8)
`60.0 (44.6–73.7)
`62.5 (47.0–75.8)
`62.5 (47.0–75.8)
`65.0 (49.5–77.9)
`62.5 (47.0–75.8)
`65.0 (49.5–77.9)
`65.0 (49.5–77.9)
`60.0 (44.6–73.7)
`60.0 (44.6–73.7)
`65.0 (49.5–77.9)
`62.5 (47.0–75.8)
`60.0 (44.6–73.7)
`62.5 (47.0–75.8)
`60.0 (44.6–73.7)
`
`10
`6
`13
`12
`10
`15
`11
`14
`9
`11
`10
`10
`12
`8
`11
`8
`
`91.8 (85.6–95.5)
`95.1 (89.7–97.8)
`89.3 (82.6–93.7)
`90.2 (83.6–94.3)
`91.8 (85.6–95.5)
`87.7 (80.7–92.4)
`91.0 (84.6–94.9)
`88.5 (81.7–93.0)
`92.6 (86.6–96.1)
`91.0 (84.6–94.9)
`91.8 (85.6–95.5)
`91.8 (85.6–95.5)
`90.2 (83.6–94.3)
`93.4 (87.6–96.6)
`91.0 (84.6–94.9)
`93.4 (87.6–96.6)
`
`NOTE. DIA loci are D (5p21), E (17p13), X (HRMT1L1), and Y (LOC91199). DIA analysis was performed with and without the 200-bp fragment
`for DEXY. Because this did not change the results, remaining analyses were performed with the inclusion of the 200-bp fragment. The DY
`combination gave maximum sensitivity and specificity (shown in bold).
`
`Geneoscopy Exhibit 1006, Page 4
`
`

`

`January 2007
`
`IMPROVED FECAL DNA SCREENING FOR COLON CANCER 115
`
`Table 5. Sensitivity and Specificity of Different Marker Combinations
`
`Sensitivity (n ⫽ 40)
`
`Specificity (n ⫽ 122)
`
`No. positive/total
`
`% (95% CI)
`
`No. positive/total
`
`% (95% CI)
`
`Methylation (total):
`HLTF
`Vimentin
`DIA-DY (from Table 4)
`Vimentin ⫹ DYa
`Stage I
`Stage II
`Stage III
`Stage IV
`
`31
`15
`29
`26
`35/40
`6/8
`9/10
`16/17
`4/5
`
`77.5 (62.5–87.7)
`37.5 (24.2–53.0)
`72.5 (57.2–83.9)
`65.0 (49.5–77.9)
`87.5 (73.9–94.5)
`75.0 (40.9–92.8)
`90.0 (59.6–98.2)
`94.1 (73.0–99.0)
`80.0 (37.6–96.4)
`
`19
`9
`16
`9
`22/122
`—
`—
`—
`—
`
`84.4 (77.0–89.8)
`92.6 (86.6–96.1)
`86.9 (79.8–91.8)
`92.6 (86.6–96.1)
`82.0 (74.2–87.8)
`—
`—
`—
`—
`
`aVimentin ⫹ DY was the optimal marker combination.
`
`ficities ranging from 87.7% to 95.1% (Table 4). The DY combi-
`nation had the highest overall sensitivity (65%; 95% CI, 49.5%–
`77.9%) and specificity (92.6%; 95% CI, 86.6%–96.1%).
`
`Optimal Marker Combinations for Maximal
`Sensitivity and Specificity
`Of the methylation markers, HLTF was 2-fold less sen-
`sitive than vimentin, and did not significantly improve overall
`methylation marker sensitivity and specificity when combined
`with vimentin (Table 5). Vimentin alone gave a sensitivity of
`72.5% (95% CI, 57.2%– 83.9%) and a specificity of 86.9% (95% CI,
`79.8%–91.8%)—values that are nearly identical to the composite
`version 1 panel of MuMu22 plus DIA (see Table 3). Examples of
`vimentin methylation in normal and cancer specimens are
`shown in Figure 1. We explored marker combinations to deter-
`mine which formulation would provide maximum sensitivity
`and specificity. As noted earlier, DIA-DY alone gave a sensitivity
`and specificity of 65% and 92.6%, respectively. The least complex
`assay consisted of hypermethylation of vimentin and DIA-DY
`(vim ⫹ DY), yielding a maximum sensitivity of 87.5% (95% CI,
`73.9%–94.5%), with a specificity of 82.0% (95% CI, 74.2%– 87.8%)
`(Table 5). Importantly, among the 40 cancers, vim⫹DY de-
`tected cancers regardless of stage.
`
`Influence of Tumor Location and Patient Age
`on Marker Expression
`There was a significant predilection for left-sided can-
`cers to be DY positive (Table 6). However, vimentin detected
`cancers regardless of the location (Table 6). Thus, the combi-
`nation of vim⫹DY detected cancers regardless of the location.
`We observed that among individuals with NC, those with
`a positive methylation marker (either vimentin or HLTF)
`were significantly older than those with a negative test (Table
`7). A similar trend was noted among the CRC patients. There
`were 22 subjects (18%) with false-positive vimentin and/or
`DY in the NC group (Table 5). Among these, 14 were positive
`for vimentin alone (specificity, 88.6%), 6 were positive for
`DIA alone (specificity, 95.1%), and 2 were positive for DIA
`and vimentin (specificity, 98.4%). The mean age of subjects in
`these groups was 65.1, 58.5, and 59 years, respectively, indi-
`cating that false-positives for vimentin methylation were
`associated with older age.
`
`Patient Satisfaction
`Forty-one percent of the respondents to the satisfaction
`questionnaire were men, and 40% were older than age 60. The
`percentage that found it easy or very easy to (1) perform the test,
`(2) open the preservative bottle, or (3) add the preservative to the
`specimen was 97%, 96%, and 100%, respectively, and 93% of re-
`spondents felt very comfortable performing the stool test. Impor-
`tantly, 84% would repeat the test if recommended by their doctor.
`
`Discussion
`Screening is a cost-effective yet underused strategy for
`reducing CRC incidence and mortality. Concerns about test dis-
`comfort, invasiveness, embarrassment, and self-efficacy have been
`identified as important barriers to more effective screening.14,15
`The availability of a noninvasive screening test that is convenient,
`safe, and easy to perform at home without bowel preparation or
`dietary restriction has the potential to significantly increase par-
`ticipation. Prior studies clearly have shown that fecal DNA testing
`fulfills these criteria and has distinct advantages over existing
`screening strategies, including the fecal occult blood test.16 In
`addition, because fecal DNA testing involves mailing of specimens,
`geographic access becomes less of a barrier, there is no loss of time
`from work, and no formal health care visit. Improved performance
`
`Figure 1. Detection of vimentin methylation in fecal DNA analyzed by
`VIM-29 PCR primers. Amplification of fecal DNA from 9 normal (N) and
`5 colon cancer (C) patients. Upper panel (VIM-29M) shows vimentin
`gene methylation and the lower panel (VIM-29U) shows control wild-
`type amplification of unmethylated vimentin sequences derived from
`normal cells in all samples. Positive (⫹) vimentin methylation is noted
`next to the sample identification number. Assay controls include un-
`methylated (U) and methylated (M) DNA samples, and negative water
`blank (deionized water).
`
`Geneoscopy Exhibit 1006, Page 5
`
`

`

`116 ITZKOWITZ ET AL
`
`CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 1
`
`Table 6. Association Between Marker Detection and Cancer Location
`
`Right (N ⫽ 11)
`Left (N ⫽ 29)
`P value
`
`Version 1
`
`54.5%
`79.3%
`NS
`
`DIA-DY
`
`36.4%
`75.9%
`.03
`
`HLTF
`
`36.4%
`37.9%
`NS
`
`Vimentin
`
`HLTF ⫹ Vimentin
`
`Vimentin ⫹ DIA-DY
`
`72.7%
`72.4%
`NS
`
`81.8%
`75.9%
`NS
`
`90.9%
`86.2%
`NS
`
`data would further validate the role of fecal DNA testing as an
`alternative screening option.
`We herein report the results of a second-generation fecal DNA
`test that takes advantage of technical enhancements and new
`knowledge about molecular pathways of colon carcinogenesis.
`Our findings build on the promising results of the prototype
`(version 1) multitarget DNA panel that showed 52% sensitivity
`and 94% specificity for CRC in asymptomatic average-risk indi-
`viduals.6 By using improved methods for stabilizing DNA dur-
`ing specimen transport, and better extraction of DNA from
`stool, we now find that the sensitivity for the version 1 marker
`panel is increased from 52% to 72.5%. This improvement was a
`direct result of an increase in DIA sensitivity from 3% to 65%,
`confirming the observations of Olson et al7 that the low sensi-
`tivity of DIA in the previous multicenter study was as result of
`the DNA degradation. This increase in sensitivity for the full
`version 1 panel was achieved at the expense of a relatively minor
`decrease in specificity from 94.4% to 89.3%. Thus, even with the
`original markers, improved DNA preservation and extraction
`from stool resulted in a better screening test.
`This advance notwithstanding, the MuMu22⫹DIA panel is
`complex given its many target genes and the need to conduct
`numerous PCR reactions. It is noteworthy, therefore, that the
`72.5% sensitivity of the full version 1 mutation panel achieved
`in the present study by better preservation and purification
`methods also was achieved by using either the DIA-DY or
`vimentin methylation markers alone. Indeed, because the sen-
`sitivity of 72.5% (95% CI, 57.2%– 83.9%) for vimentin methyl-
`ation alone in the present study is higher than our previous
`finding of 46% (95% CI, 35%–56%) with this marker,10 this
`further suggests that better DNA preservation and extraction
`improves the results of methylation marker assays in stool. The
`combination of vimentin plus DIA-DY resulted in an optimal
`sensitivity of 87.5% (95% CI, 73.9%–94.5%). This optimal com-
`bination detected most cancers, including early (stage I) lesions,
`regardless of their location in the colon. These results support
`the biological significance of abnormal apoptosis and aberrant
`gene methylation in colorectal carcinogenesis.
`
`It is not known how the new version 2 assay will perform in
`a screening setting where most cancers are detected at an earlier
`stage.6 Although only half the CRC cases in the present study
`were stage I or II lesions, the fact that the new assay detected
`75%–90% of these early stage lesions (compared with 53%– 63%
`in the multicenter screening study), and that precolonoscopy
`stool samples may have even higher sensitivity than post-
`colonoscopy samples (Table 1), encourages us to think that the
`new assay will perform well in a screening setting. In addition,
`historically, the best results with the version 1 assay were ob-
`tained when stools were frozen within a few hours of collection,
`thereby optimizing the quality of the DNA.17 The fact that the
`DNA stabilization buffer now appears to mimic this condition
`suggests that DNA degradation will be less of a problem when
`the assay is applied in a screening setting. Others recently have
`reported that when fecal DNA was better preserved with buffer,
`the sensitivity of a different long DNA assay was 44% in a
`limited number of patients.18
`The improvement in sensitivity to 88% for vim ⫹ DY was at the
`expense of a lower specificity of 82%, representing 22 false-positive
`results. Further analysis revealed that older age may have contrib-
`uted to a false-positive test result. Because many genes become
`methylated as a function of age,9 we postulate that this may be
`caused by age-related methylation changes in the vimentin pro-
`moter. Alternatively, vimentin methylation frequently is present in
`aberrant crypt foci,10 suggesting that some false-positive test re-
`sults for vimentin methylation may arise in individuals with a
`particularly high burden of such aberrant crypt foci, who may
`accordingly be at increased risk for future development of colonic
`neoplasia. Although we cannot completely exclude the possibility
`that some patients with a positive vimentin⫹DY test and a NC
`might have a missed adenoma or cancer, we took care to include
`only patients in whom colonoscopy was complete to the cecum
`and the bowel preparation was considered to have adequately
`visualized the entire colonic mucosa.19
`Inherent in the success of any screening test is patient ac-
`ceptance. Our preliminary experience with the new fecal DNA
`test in which patients add stabilization buffer to stool confirms
`
`Table 7. Association Between Markers and Patient Age
`
`Version 1
`
`DIA-DY
`
`HLTF
`
`Vimentin
`
`HLTF ⫹ Vimentin
`
`Vimentin ⫹ DIA-DY
`
`Normal colonoscopy
`Negative
`Positive
`P value
`Colon cancer
`Negative
`Positive
`P value
`
`58.3
`58.9
`NS
`
`67.5
`65.0
`NS
`
`58.4
`58.4
`NS
`
`68.3
`64.2
`NS
`
`58.0
`63.0
`.04
`
`65.1
`66.5
`NS
`
`57.5
`64.3
`.00
`
`60.3
`67.7
`.04
`
`57.6
`62.6
`.03
`
`59.8
`67.4
`.05
`
`57.4
`62.7
`.01
`
`69.0
`65.2
`NS
`
`NOTE. Values represent mean age (in years).
`
`Geneoscopy Exhibit 1006, Page 6
`
`

`

`January 2007
`
`IMPROVED FECAL DNA SCREENING FOR COLON CANCER 117
`
`a very high degree of patient satisfaction. These findings sup-
`port results from the previous multicenter study in which
`subjects preferred fecal DNA testing to both the fecal occult
`blood test and colonoscopy.16
`Although screening colonoscopy is becoming more readily
`used, there are manpower and resource issues that have to be
`overcome