American Journal of Gastroenterology
`C(cid:2) 2008 by Am. Coll. of Gastroenterology
`Published by Blackwell Publishing
`
`ISSN 0002-9270
`doi: 10.1111/j.1572-0241.2008.02088.x
`
`A Simplified, Noninvasive Stool DNA Test
`for Colorectal Cancer Detection
`Steven Itzkowitz, M.D.,1 Randall Brand, M.D.,2 Lina Jandorf, M.A.,1 Kris Durkee, Ph.D.,9
`John Millholland, Ph.D.,9 Linda Rabeneck, M.D., M.P.H.,3 Paul C. Schroy III, M.D., M.P.H.,4
`Stephen Sontag, M.D.,5 David Johnson, M.D.,6 Sanford Markowitz, M.D.,7 Lawrence Paszat, M.D., M.Sc.,8
`and Barry M. Berger, M.D.9
`1Department of Medicine and Oncological Sciences, Mount Sinai School of Medicine, New York, New York;
`2GI Division, Evanston Northwestern Healthcare, Evanston, Illinois; 3Sunnybrook Regional Cancer Centre,
`University of Toronto, Toronto, Canada; 4GI Division, Boston University School of Medicine, Boston,
`Massachusetts; 5GI Section, Hines Veterans Affairs Hospital, Hines, Illinois; 6Gastroenterology Division,
`Eastern VA Medical School, Norfolk, Virginia; 7Howard Hughes Medical Institute and Department of Medicine,
`Case Western Reserve University, Cleveland, Ohio; 8Sunnybrook Health Sciences Centre, Toronto, Ontario; and
`9Exact Sciences Corporation, Marlborough, Massachusetts
`
`BACKGROUND: As a noninvasive colorectal cancer (CRC) screening test, a multi-marker first generation stool DNA
`(sDNA V 1.0) test is superior to guaiac-based fecal occult blood tests. An improved sDNA assay
`(version 2), utilizing only two markers, hypermethylated vimentin gene (hV) and a two site DNA
`integrity assay (DY), demonstrated in a training set (phase 1a) an even higher sensitivity (88%) for
`CRC with a specificity of 82%.
`
`AIM:
`
`To validate in an independent set of patients (phase 1b) the sensitivity and specificity of sDNA
`version 2 for CRC.
`
`RESULTS:
`
`METHODS:
`
`Forty-two patients with CRC and 241 subjects with normal colonoscopy (NC) provided stool samples,
`to which they immediately added DNA stabilizing buffer, and mailed their specimen to the
`laboratory. DNA was purified using gel-based capture, and analyzed for hV and DY using methods
`identical to those previously published.
`Using the same cutpoints as the 1a training set (N = 162; 40 CRCs, 122 normals), hV demonstrated
`a higher and DY a slightly lower sensitivity, for a combined sensitivity of hV + DY of 86%. Optimal
`cutpoints based on the combined phase 1a + 1b dataset (N = 445; 82 CRCs, 363 normals) yielded
`a CRC sensitivity of 83%. The vast majority of cancers were detected regardless of tumor stage,
`tumor location, or patient age. Assay specificity in the phase 1b dataset for hV, DY, and hV + DY was
`82%, 85%, and 73%, respectively, using the phase 1a cutpoints. Optimal cutpoints based on the
`combined phase 1a + 1b dataset yield a specificity of 82%.
`CONCLUSIONS: This study provides validation of a simplified, improved sDNA test that incorporates only two
`markers and that demonstrates high sensitivity (83%) and specificity (82%) for CRC. Test
`performance is highly reproducible in a large set of patients. The use of only two markers will make
`the test easier to perform, reduce the cost, and facilitate distribution to local laboratories.
`
`(Am J Gastroenterol 2008;103:2862–2870)
`
`INTRODUCTION
`
`Screening for colorectal cancer (CRC) is a highly effective
`intervention that substantially reduces cancer-specific mor-
`tality by detecting early stage CRC and premalignant lesions.
`Despite the recommendations of all major medical societies,
`fewer than 60% of eligible individuals over age 50 have un-
`dergone CRC screening (1–3). Although several CRC screen-
`ing methods are available, colonoscopy is being increasingly
`
`used as the primary screening tool because of its excel-
`lent diagnostic accuracy and ability to remove precancerous
`and early cancerous lesions. However, the invasive nature of
`the procedure itself, as well as the many physician, patient,
`and organizational barriers, limit its effectiveness. This has
`spawned efforts to develop an accurate noninvasive screen-
`ing test that would increase adherence with CRC screening
`guidelines by individuals who are reluctant to undergo inva-
`sive tests, or in situations where colonoscopy screening is not
`
`2862
`
`Downloaded from http://journals.lww.com/ajg by GqmaMshhiPwrY8JZZlbQeyBNj0CtyHw3cdfO4RhDK417XVuRs981F
`
`zLCaaUwJda3Odak2w7SKRw1wwxEppfMPmZztfrtBsZJDIBVzyQPhfbqMsmrymCyJ3yKLYpzGa/6 on 10/23/2023
`
`Geneoscopy Exhibit 1058, Page 1
`
`

`

`Simplified Stool DNA Test for Colorectal Cancer Detection
`
`2863
`
`sDNA from CRC patients and those with NC. The first set
`(phase 1a) would be the training set; the second set (phase
`1b) would be the validation set. The results of the training set
`(phase 1a) demonstrated a sensitivity of 88% for CRC, and
`a specificity of 82% (12). The purpose of the present study
`was to validate in an independent set of patients the perfor-
`mance of the version 2 assay with respect to the sensitivity
`and specificity for CRC.
`
`METHODS
`
`Source of Clinical Material
`The same seven centers that participated in the phase 1a study
`also participated in the present phase 1b study. These centers
`represent a spectrum of academic medical settings (com-
`munity based to tertiary care). Each center obtained local
`Institutional Review Board approval prior to beginning the
`study. Phase 1b was originally planned to include 125 new
`patients with CRC and 200 new subjects with NC. The quota
`of normal subjects was readily achieved, but CRC enrollment
`was slower than expected. Therefore, upon the advice of the
`statisticians from an independent contract research organiza-
`tion (Battelle CRO, Needham, MA), phase 1b was terminated
`with the enrollment of 50 CRC patients.
`Between June 2005 and February 2006, subjects 50–
`80 yr of age were eligible for the study if they were found at the
`time of colonoscopy to have either CRC, or a NC. The latter
`group consisted of individuals in whom the bowel prepara-
`tion (prep) was classified as very good to excellent (deemed
`adequate to exclude polyps >5mm), the colonoscopy was
`complete to the cecum, and the mucosa was free of any type
`of mucosal lesion or polyps. We offered entry to consecu-
`tive presenting patients with CRC at each site. Individuals
`were excluded if any of the following conditions applied:
`any contraindication to colonoscopy or conscious sedation;
`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 polypo-
`sis, hereditary nonpolyposis CRC, inflammatory bowel dis-
`ease, and strong family history of CRC (two or more first
`degree relatives with CRC, or one or more first-degree rel-
`atives with CRC younger than age 50), personal history of
`colorectal adenomas or CRC, prior colorectal resection for
`any reason, current pregnancy or lactation. Gastrointestinal
`symptoms were not an exclusion criterion and were reported
`by 66/82 (80.5%) subjects with CRC, but only 40/363 (11%)
`subjects with NC. The preparation for, and performance of,
`colonoscopy was done according to standard operating pro-
`cedures at each site. The histologic diagnosis of CRC was
`verified by a board-certified pathologist. Cancers were staged
`according to the TNM (Tumor-Nodes-Metastasis) classifica-
`tion. Left-sided cancers were defined as those arising at, or
`distal to, the splenic flexure.
`
`feasible. Indeed, the recently updated guidelines endorsed by
`the American Cancer Society, the U.S. Multi-Society Task
`Force on Colorectal Cancer, and the American College of
`Radiology encourage all average risk individuals over age
`50 to undergo CRC screening with either a structural or a
`noninvasive screening test (4).
`Several studies have demonstrated the feasibility of ex-
`tracting and detecting human DNA from stool (reviewed
`in Ref. 5). The DNA markers in these studies comprise
`mutations of genes involved in the predominant chromo-
`somal instability pathway (such as APC, p53, and K-ras)
`and DNA alterations reflecting the microsatellite instability
`pathway (Bat-26) and abnormal apoptosis. Initial studies of
`stool DNA (sDNA) using stool samples from patients al-
`ready determined by colonoscopy to have colon cancer, ade-
`nomas, or a normal colon, reported sensitivities of 62–91%
`for CRC, 27–82% for advanced adenomas, and specificities
`of 93–96% (5, 6). These encouraging data prompted a large,
`prospective, multi-center screening study of more than 4,000
`average-risk, asymptomatic individuals over age 50. The re-
`sults demonstrated fourfold greater sensitivity for detecting
`CRC with the sDNA test compared to Hemoccult II fecal oc-
`cult blood test (51.6% vs 12.9%, P = 0.003), with comparable
`specificity (94.4% vs 95.2%) (7). Despite its superiority over
`Hemoccult II, the prototype sDNA test (version 1.0) exhib-
`ited lower than expected sensitivity, due to an unexpectedly
`low rate of positivity for the DNA integrity assay (DIA) com-
`ponent of the assay. Despite precautions such as immediate
`chilling of samples and rapid delivery to the laboratory by
`express courier, subsequent research demonstrated that the
`cause of the suboptimal performance of DIA was a result
`of DNA degradation during the transit of specimens to the
`laboratory (8).
`Several technical and conceptual improvements have now
`been incorporated into a newer assay version (version 2).
`First, the addition of a DNA-stabilizing buffer to the stool im-
`mediately upon defecation prevents DNA degradation during
`transport for several days and enhances the performance of
`DIA (8). Second, a gel-based DNA capture approach, rather
`than the original bead-based technology, permits better ex-
`traction of DNA from stool (9). Third, a new marker, hyper-
`methylated vimentin gene, has been included in the sDNA as-
`say. The use of this new marker is based on the evidence that
`the epigenetic phenomenon of promoter methylation is a key
`pathway by which colon cancers develop (10). Vimentin pro-
`tein is not normally expressed by colonic epithelial cells but
`is typically expressed by mesenchymal cells. The vimentin
`gene is minimally methylated in normal colonic epithelial
`cells, but was found to be highly methylated in colon can-
`cer cell lines and in 53–83% of colon cancer tissues (11).
`In a recent study, hypermethylated vimentin was detected in
`the sDNA of 43/94 (46%) patients with CRC versus 20/198
`(10%) with a normal colonoscopy (NC) (11).
`To test the performance of the version 2 assay for CRC,
`we conducted a two-phase study. The study was designed to
`use the version 2 assay to analyze two sets of patients with
`
`Downloaded from http://journals.lww.com/ajg by GqmaMshhiPwrY8JZZlbQeyBNj0CtyHw3cdfO4RhDK417XVuRs981F
`
`zLCaaUwJda3Odak2w7SKRw1wwxEppfMPmZztfrtBsZJDIBVzyQPhfbqMsmrymCyJ3yKLYpzGa/6 on 10/23/2023
`
`Geneoscopy Exhibit 1058, Page 2
`
`

`

`had been completed and scored. Descriptive statistics were
`used to characterize the data. Positivity for each marker (hV
`and DY) is reported separately. The sDNA test, which com-
`bines both markers, is considered positive if either hV or DY
`is positive. Sensitivity and specificity with 95% confidence
`intervals (CI) were computed for all markers. The t-tests and
`χ 2 tests, comparing the CRC to the NC group, were used
`to examine associations between patient characteristics (e.g.,
`gender, age, time since colonoscopy) or markers. P values
`less than 0.05 were considered significant. SPSS (version
`14) (SPSS, Inc., Chicago, IL) was used for all analyses.
`
`RESULTS
`
`Patient Population
`Initially, 51 patients with CRC and 248 subjects in the NC
`group were enrolled. Of the patients in the CRC group, nine
`were excluded: three because of a positive family history or
`personal history of CRC, and six because the cancer was
`interpreted as high-grade dysplasia (HGD) and not invasive
`cancer. Of the subjects in the NC group, seven were excluded
`because no collection date was indicated (N = 1), there was
`inadequate stool weight (N = 2), or because of a previous
`history of cancer (one CRC, one breast, one leukemia, one
`larynx). Five subjects with CRC between ages 42 and 50 were
`included because they fulfilled all other eligibility criteria.
`Thus, the phase 1b subject set consisted of 283 patients (42
`CRC, 241 NC).
`Table 1 lists the demographic characteristics of the sub-
`jects studied. There were no significant differences between
`the two groups in terms of gender or collection interval. The
`number of days between colonoscopy and stool sample col-
`lection was longer in the CRC group, perhaps owing to factors
`related to patients adjusting to their new diagnosis. The NC
`group was younger than the CRC group. Among those with
`CRC, there was no difference in mean age according to can-
`cer stage. Almost two-thirds of all cancers were early stage
`(I and II), and two-thirds of CRCs were located distal to the
`splenic flexure.
`
`Assay Sensitivity
`The same markers used in the phase 1a study were analyzed
`using the identical sample collection kit, DNA stabilization
`buffer, and gel-based DNA purification. Table 2 shows the
`sensitivity and specificity of the markers in the phase 1a study,
`phase 1b study, and the combined dataset. Using cutpoint val-
`ues derived from phase 1a, the sensitivity of hV as a single
`marker in phase 1b was 81% (95% CI = 66.7–90.0%), higher
`than the value found in the phase 1a study. The sensitivity of
`DY was 60% (95% CI = 44.5–73.0%), slightly lower than
`phase 1a results. Combining both markers yielded a sensitiv-
`ity of 86% (95% CI = 72.2–93.3%), a value almost identical
`to the phase 1a result of 88%. Optimal cutpoints based on
`the combined phase 1a + 1b dataset yielded a sensitivity for
`hV + DY of 83% (95% CI = 73.4–89.5%). The positive
`likelihood ratio for the combined dataset was 4.49.
`
`2864
`
`Itzkowitz et al.
`
`Sample Collection
`To avoid any possible effect of the colonoscopic bowel prep
`or any biopsies done at the time of exam on test results, each
`subject was asked to provide a single stool sample approxi-
`mately 6–14 days after colonoscopy. For patients with CRC,
`the sample was provided prior to beginning the presurgical
`bowel prep. Subjects were given a special stool collection
`kit that mounts on the toilet bowl along with detailed in-
`structions. Immediately following defecation, subjects added
`250 mL of a DNA stabilizing buffer (8) to the passed stool.
`Acceptable specimens were at least 35 g, with no upper limit
`of quantity. The specimen was shipped overnight at room
`temperature using a coded identifier to keep the laboratory
`blinded to the clinical source. The collection interval was de-
`fined as the number of hours from the time of defecation until
`the specimen arrived in the laboratory. Stool samples were
`processed and analyzed without knowledge of clinical infor-
`mation. The details of sample processing and human DNA
`purification have been described previously (12).
`
`DNA Integrity Assay
`The DIA assay was performed using real-time PCR as de-
`scribed previously (12). The assay has been converted to a
`multiplex format where four primer/probe pairs simultane-
`ously interrogate the presence and quantity of 200, 1300,
`1800, and 2400 bp human DNA fragments at two loci: 5p21
`(Locus D); LOC91199 (Locus Y). For the four different-sized
`fragments at each of these two loci, a numerical cutpoint was
`determined based on the genome equivalents amplified for
`each fragment, above which the assay was considered positive
`for that fragment size. Any sample that returned values above
`the established cutoffs for at least three of the eight fragments
`was considered a positive (abnormal) DIA-DY test.
`
`Vimentin Methylation Assay
`Bisulfite conversion of DNA was performed as previously de-
`scribed (13). Methylation specific-PCR (MS-PCR) reactions
`were performed using 0.5 µM armed primers for vimentin
`(IDT, Coralville, IA), 1X HotStar buffer, 1.25 U HotStar poly-
`merase (Qiagen, Alameda, CA), 200 µM dNTP (Promega,
`Madison, WI), and 10 µL (capture stool) DNA in a final vol-
`◦
`ume of 50 µL. Cycling conditions were 95
`C for 14.5 min
`◦
`◦
`followed by 40 cycles of 94
`C for 30 s, 68
`C (vimentin methy-
`◦
`◦
`lated) or 62
`C (vimentin unmethylated) for 1 min, 72
`C for
`◦
`1 min with final 72
`C for 5 min. Samples were visualized
`on 4% NuSieve 3:1 agarose (FMC, Rockland, ME) gels us-
`ing a Stratagene EagleEye II (Stratagene, La Jolla, CA) still
`image system. Samples were scored positive (hypermethy-
`lated vimentin present) if MS-PCR band intensity exceeded a
`previously determined level. Positive samples were repeated
`in duplicate to confirm methylation status. Primer sequences
`are available on request.
`
`Data Analysis
`To ensure adequate blinding of data, Battelle CRO maintained
`all data and did not release this information until all assays
`
`Downloaded from http://journals.lww.com/ajg by GqmaMshhiPwrY8JZZlbQeyBNj0CtyHw3cdfO4RhDK417XVuRs981F
`
`zLCaaUwJda3Odak2w7SKRw1wwxEppfMPmZztfrtBsZJDIBVzyQPhfbqMsmrymCyJ3yKLYpzGa/6 on 10/23/2023
`
`Geneoscopy Exhibit 1058, Page 3
`
`

`

`Simplified Stool DNA Test for Colorectal Cancer Detection
`
`2865
`
`Normal Colonoscopy
`(N = 241)
`90 (37.3%)
`28.18 ±7.38
`10.68 ±2.92
`56.90 ±6.33
`N/A
`N/A
`N/A
`N/A
`
`P
`value
`
`0.841
`0.091
`0.003
`0.0
`
`Table 1. Demographic Characteristics of the Study Population (Phase 1b)
`Colorectal Cancer
`(N = 42)
`15 (35.7%)
`30.31 ±7.43
`16.07 ±10.95
`67.44 ±11.21
`69.12 ±10.91
`68.11 ±12.66
`65.33 ±11.16
`67.93 ±9.00
`
`Male, N (%)
`Collection interval, h (mean ± SD)
`Time since c’scopy, days (mean ± SD)
`Age, yr (mean ± SD)
`Stage I
`Stage II
`Stage III
`Stage IV
`Stage of cancer, n (%)
`Stage I
`Stage II
`Stage III
`Stage IV
`Size of cancer, cm (mean ± SD)
`Stage I
`Stage II
`Stage III
`Stage IV
`Location of cancer, n (%)
`Distal
`Proximal
`
`11 (26.2%)
`14 (33.3%)
`14 (33.3%)
`3 (7.1%)
`3.40 ±1.23
`3.81 ± 2.11
`3.77 ± 1.76
`7.33 ± 5.80
`
`28 (66.7%)
`14 (33.3%)
`
`N/A
`N/A
`N/A
`N/A
`
`N/A
`N/A
`N/A
`N/A
`
`N/A
`N/A
`
`Six (14%) of the 42 CRC cases were not associated with
`detectable hV or DY. There were no apparent distinguishable
`clinicopathological features of these few tumors. The mean
`age of this group was 70.7 yr; four were stage III, two were
`stage II, and four were in the proximal colon.
`
`Assay Specificity
`Among the stool specimens from 241 subjects with nor-
`mal colonoscopies in phase 1b, 18% contained hV and 15%
`demonstrated abnormal apoptosis by the DY assay, giving
`single marker specificities of 82% and 85%, respectively
`(Table 2). Combining both markers gave a specificity of 73%
`(95% CI = 67.1–78.2%). When the phase 1a and 1b datasets
`were combined, optimal cutpoints based on this more robust
`dataset (363 NC’s) yielded specificities of 83%, 96%, and
`82% for hV, DY, and hV + DY, respectively. The 95% CI
`of the specificity yielded by the phase 1a cutpoints applied
`to the 1a dataset, the 1a cutpoints applied to the 1b dataset
`and the cutpoints newly determined from the combined 1a +
`1b dataset all overlap. The latter cutpoints were identical to
`the 1a cutpoint for hV but were adjusted with respect to DY
`
`Table 2. Sensitivity and Specificity of Version 2 Assay
`
`to capture more fully the biologic variability represented by
`the larger combined NC data. The negative likelihood ratio
`for the combined dataset was 0.21.
`The relative ability of cutpoints to separate the means of
`Gaussian distributions of a population of subjects with CRC
`from that of a population of subjects with normal colonos-
`copies is illustrated in Figure 1. The phase 1a training set
`cutpoint on the 1a dataset and the 1a + 1b cutpoint on the
`1a + 1b dataset both cluster near 2 sigma of separation, with
`the 1a training set cutpoint on the 1b dataset somewhat below.
`The latter performance reflects the lowered specificity seen
`with the DY marker in the larger population of the 1b dataset
`(85%) compared to that of the smaller 1a dataset (93%).
`The 95% CI of the sensitivity and specificity generated by
`each of these three cutpoints on the three datasets overlap
`(Table 2).
`
`Influence of Tumor Stage on Marker Expression
`In phase 1b, hV was associated with the vast majority of
`cancers regardless of tumor stage (Table 3). A positive DY
`assay, however, was associated more frequently with earlier
`
`Phase 1b: (1a cutpoints)
`Sensitivity (N = 42 CRC)
`Specificity (N = 241 NL)
`Phase 1a: (Ref. 12)
`Sensitivity (N = 40 CRC)
`Specificity (N = 122 NL)
`Combined 1a and 1b dataset
`Sensitivity (N = 82 CRC)
`Specificity (N = 363 NL)
`
`hV
`
`DY
`
`81% [66.7–90.0%]
`82% [76.4–86.1%]
`
`73% [57.2–83.9%]
`87% [79.8–91.8%]
`
`77% [66.6–84.6%]
`83% [79.3–86.9%]
`
`60% [44.5–73.0%]
`85% [80.0–89.0%]
`
`65% [49.5–77.9%]
`93% [86.6–96.1%]
`
`48% [37.1–58.2%]
`96% [93.0–97.3%]
`
`hV + DY [95% CI]
`
`86% [72.2–93.3%]
`73% [67.1–78.2%]
`
`88% [73.9–94.5%]
`82% [74.2–87.8%]
`
`83% [73.4–89.5%]
`82% [77.2–85.2%]
`
`Downloaded from http://journals.lww.com/ajg by GqmaMshhiPwrY8JZZlbQeyBNj0CtyHw3cdfO4RhDK417XVuRs981F
`
`zLCaaUwJda3Odak2w7SKRw1wwxEppfMPmZztfrtBsZJDIBVzyQPhfbqMsmrymCyJ3yKLYpzGa/6 on 10/23/2023
`
`Geneoscopy Exhibit 1058, Page 4
`
`

`

`2866
`
`Itzkowitz et al.
`
`Receiver Operator Characteristics
`
`3 Sigma
`
`2.5 Sigma
`
`2.0 Sigma
`
`1.5 Sigma
`
` 0 Sigma
`
`0.10
`
`0.20
`
`0.30
`
`0.40
`
`0.50
`
`0.60
`
`0.70
`
`0.80
`
`0.90
`
`1.00
`
`1-Sensitivity (%)
`
`1.00
`
`0.90
`
`0.80
`
`0.70
`
`0.60
`
`0.50
`
`0.40
`
`0.30
`
`0.20
`
`0.10
`
`0.00
`0.00
`
`Specificity (%)
`
`Figure 1. The receiver operator plot shows five curves of iso-performance ranging from 3 standard deviations of separation of the means of
`normal and affected populations (3 sigma) to no separation (0 sigma). These curves assume that the underlying distribution consists of two
`Gaussians, each having the same standard deviation, and a mean separated by the amount of discriminating power as measured in units of
`standard deviation. The 1a training set cutpoint on the 1a dataset (•) and the 1a + 1b cutpoint on the 1a + 1b dataset ((cid:1)) cluster near 2 sigma
`of separation, the 1a training set cutpoint on the 1b dataset ((cid:2)) somewhat below. The diagonal represents the line of no discrimination.
`
`stage cancers than with late stage disease. Taken together, DY
`and hV detected all stage I and stage IV cancers, and almost
`all stage II and III cancers. The 83% CRC sensitivity seen in
`the 82 cancers of the combined phase 1a + 1b dataset was
`independent of stage.
`
`marker hV positivity remained associated with older age, but
`only among NCs. DY alone demonstrated no association with
`patient age. Combining hV and DY in both datasets revealed
`an association with older age among NCs but not patients
`with CRC.
`
`Influence of Tumor Location on Marker Expression
`In the present study, distal cancers were more likely than
`proximal cancers to be positive for both the hV and DY
`markers (Table 4). However, this was statistically signifi-
`cant only for DY, so that the combination of the two mark-
`ers detected cancers regardless of location. A similar finding
`was observed in phase 1a (12). When the results from phase
`1a and 1b were combined, there was a slight trend toward
`detection of left-sided cancers that just reached statistical
`significance.
`
`Influence of Patient Age on Sensitivity and Specificity
`In phase 1a, we previously observed that hV positivity was
`associated with older age in both healthy controls and can-
`cer patients (Table 5) (12). In the present study, there was a
`weaker (not statistically significant) association between hV
`and older age among NCs but not among cancer patients
`(Table 5). When the two datasets were combined, single
`
`Influence of Family History or Symptoms on Test Results
`The possible effect of family history or symptoms was an-
`alyzed in the combined dataset (phase 1a + 1b), using the
`phase 1a cutpoints. Among the 82 subjects with CRC, none
`reported a family history of colon cancer or polyps, so this did
`not influence test results. Among the 363 subjects with NC,
`only 24 reported a family history of CRC or polyps and four
`(16.6%) had a positive sDNA test, which is not significantly
`different from those with a negative family history.
`Among the 82 subjects with CRC, the frequency of a
`positive sDNA test was 57/66 (86.4%) among those with
`symptoms, compared to 14/16 (87.3%) among those without
`symptoms. Among the 363 subjects with NC, the frequency
`of a positive sDNA test was 12/40 (30%) among those with
`symptoms, compared to 73/323 (22.6%) among those without
`symptoms. In both cases, these differences were not statisti-
`cally significant (P > 0.9 for subjects with CRC, and P > 0.3
`for those with NC).
`
`Downloaded from http://journals.lww.com/ajg by GqmaMshhiPwrY8JZZlbQeyBNj0CtyHw3cdfO4RhDK417XVuRs981F
`
`zLCaaUwJda3Odak2w7SKRw1wwxEppfMPmZztfrtBsZJDIBVzyQPhfbqMsmrymCyJ3yKLYpzGa/6 on 10/23/2023
`
`Geneoscopy Exhibit 1058, Page 5
`
`

`

`Simplified Stool DNA Test for Colorectal Cancer Detection
`
`2867
`
`Table 3. Association Between Marker Detection and Cancer Stage
`
`hV
`
`DY
`
`Phase 1b: (1a cutpoints)
`Sensitivity (N = 42 CRC)
`Stage I (N = 11)
`Stage II (N = 14)
`Stage III (N = 14)
`Stage IV (N = 3)
`Phase 1a: (Ref. 12)
`Sensitivity (N = 40 CRC)
`Stage I (N = 8)
`Stage II (N = 10)
`Stage III (N = 17)
`Stage IV (N = 5)
`Combined dataset: (1a + 1b cutpoints)
`Sensitivity (N = 82 CRC)
`Stage I (N = 19)
`Stage II (N = 24)
`Stage III (N = 31)
`Stage IV (N = 8)
`
`81% [66.7–90.0%]
`91% [62.3–98.4%]
`86% [60.1–96.0%]
`64% [38.3–83.7%]
`100% [43.8–100%]
`
`72.5% [57.2–83.9%]
`75% [40.9–92.8%]
`80% [49.0–94.3%]
`76% [52.7–90.4%]
`40% [11.8–76.9%]
`
`77% [66.6–84.6%]
`84% [62.4–94.5%]
`83% [64.1–93.3%]
`71% [53.4–83.9%]
`63% [30.4–86.3%]
`
`60% [44.5–73.0%]
`73% [43.4–90.3%]
`71% [45.4–88.3%]
`43% [21.4–67.4%]
`33% [6.1–79.2%]
`
`65.0% [49.5–77.9%]
`62.5% [30.6–86.3%]
`70% [39.7–89.2%]
`59% [36.0–78.4%]
`80% [37.6–96.4%]
`
`48% [37.1–58.2%]
`53% [31.7–72.7%]
`54% [35.1–72.1%]
`39% [23.7–56.2%]
`50% [21.5–78.5%]
`
`hV + DY [95% CI]
`
`86% [72.2–93.3%]
`100% [74.1–100%]
`86% [60.1–96%]
`71% [45.4–88.3%]
`100% [43.8–100%]
`
`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%]
`
`83% [73.4–89.5%]
`84% [62.4–94.5%]
`88% [69.0–95.7%]
`81% [63.7–90.8%]
`75% [40.9–92.9%]
`
`Performance of the Assay in Patients
`With Adenomas with HGD
`Although advanced adenomas with HGD were not considered
`qualifying lesions, there were six patients in phase 1b and one
`patient in phase 1a who were thought to have CRC (so their
`sDNA was analyzed) but found on final pathological review to
`have noninvasive neoplastic changes, so these patients were
`excluded from the main analysis. Table 6 demonstrates that
`six of seven individuals with HGD were associated with hV,
`and three of seven demonstrated positive DY, regardless of
`adenoma location or patient age.
`
`DISCUSSION
`
`Because the prevalence of colorectal polyps is common in
`persons over age 50, and colonoscopy is both diagnostic and
`therapeutic, colonoscopy has been endorsed as the preferred
`CRC screening approach by several authorities. Indeed, the
`American College of Gastroenterology (14) and the New
`York City Department of Health and Mental Hygiene (NYC
`
`Table 4. Association Between Marker Detection and Cancer
`Location
`hV + DY
`
`hV
`
`DY
`
`∗
`
`69.2%
`89.3%
`0.175
`
`Phase 1b: (1a cutpoints)
`Right-sided (N = 13)
`Left-sided (N = 28)
`P value
`†
`Phase 1a
`Right-sided (N = 14)
`64.3%
`Left-sided (N = 26)
`76.9%
`0.414
`P value
`Combined dataset: (1a + 1b cutpoints)
`Right-sided (N = 27)
`66.7%
`Left-sided (N = 54)
`83.3%
`0.119
`P value
`
`23.1%
`78.6%
`0.0006
`
`28.6%
`84.6%
`0.0006
`
`14.8%
`64.8%
`0.00001
`
`69.2%
`96.4%
`0.058
`
`78.6%
`92.3%
`0.273
`
`70.4%
`90.7%
`0.044
`
`∗
`†
`
`Tumor location not known for one case.
`Tumor location misclassified for three cases in previous report (Ref. 12); conclusions
`unchanged.
`
`DOHMH) (15) have both endorsed the procedure, and recent
`surveys indicate that among New York City adults over age
`50, screening colonoscopy rates have increased from 47% to
`60% in the last 4 yr (16, 17).
`While the ultimate goal is to decrease mortality and mor-
`bidity from CRC, there remain formidable physician, patient,
`and organizational barriers to screening colonoscopy. In one
`recent study, even when medically insured patients were re-
`ferred by their physician for screening colonoscopy, and pa-
`tients were personally navigated through the system to facili-
`tate the completion of colonoscopy, fully one-third of patients
`still did not complete the procedure (18). Moreover, in a large
`safety net health care system, a disappointing 42% of patients
`did not attend their colonoscopy appointment (19). Test dis-
`comfort, invasiveness, embarrassment, and self-efficacy have
`been identified as important barriers to more effective screen-
`ing (20, 21). Patient inconvenience with colonoscopy has re-
`cently been quantified, revealing that patients spend an av-
`erage of 21 h involved in bowel preparation, travel, waiting,
`colonoscopy procedure, and recovery, and another 20 h af-
`ter the procedure before they are completely back to normal
`(22). Inadequate cleansing of the colon can limit the detec-
`tion of neoplasia (23, 24) and even result in the scheduling
`of a repeat colonoscopy. Some large studies report that 23%
`of colonoscopies were associated with an inadequate prep
`(24). Superimposed on these barriers is the increased volume
`in the endoscopy suite itself, and the demands for screen-
`ing colonoscopy to compete with the need to perform the
`procedure for diagnostic and surveillance purposes (25–27).
`All of these issues have prompted some to advocate a mixed
`strategy for screening, whereby colonoscopy is used as a
`diagnostic follow-up tool for a less invasive but positive
`test (28).
`The availability of a noninvasive screening test that is con-
`venient, safe, and easy to perform at home without bowel
`preparation or dietary restriction has the potential to signifi-
`cantly increase participation in CRC screening. The ability to
`
`Downloaded from http://journals.lww.com/ajg by GqmaMshhiPwrY8JZZlbQeyBNj0CtyHw3cdfO4RhDK417XVuRs981F
`
`zLCaaUwJda3Odak2w7SKRw1wwxEppfMPmZztfrtBsZJDIBVzyQPhfbqMsmrymCyJ3yKLYpzGa/6 on 10/23/2023
`
`Geneoscopy Exhibit 1058, Page 6
`
`

`

`2868
`
`Itzkowitz et al.
`
`Table 5. Association Between Markers and Patient Age
`
`∗
`
`Phase 1b (using phase 1a cutpoints)
`Normal c’scopy
`Colon cancer
`Phase 1a (Ref. 12)
`Normal c’scopy
`Colon cancer
`Phase 1a + 1b
`Normal c’scopy
`Colon cancer
`
`hV
`
`Neg
`
`56.5
`68.6
`
`57.5
`60.3
`
`56.7
`63.8
`
`Pos
`
`58.7
`67.2
`
`∗
`
`64.3
`67.7
`
`60.2
`67.4
`
`P
`
`0.084
`0.741
`
`0.006
`0.066
`
`0.003
`0.229
`
`Pos
`
`56.6
`66.2
`
`58.4
`64.2
`
`57.3
`65.7
`
`DY
`
`Neg
`
`57.0
`69.3
`
`58.4
`68.3
`
`56.2
`67.3
`
`P
`
`0.670
`0.361
`
`0.974
`0.225
`
`0.462
`0.516
`
`hV + DY
`Neg
`
`56.6
`70.7
`
`57.4
`69.0
`
`56.6
`67.6
`
`P
`
`0.289
`0.349
`
`0.011
`0.287
`
`0.003
`1.0
`
`Pos
`
`57.7
`66.9
`
`62.7
`65.2
`
`59.9
`66.3
`
`∗
`
`Values represent mean age of patients in each category.
`
`mail sDNA test specimens to the laboratory avoids the need
`for a formal health care visit, prevents loss of time from work,
`and removes the disadvantage of geographic distance. Previ-
`ous studies reveal that the patients perceive sDNA testing
`to have distinct advantages over existing screening strate-
`gies, including fecal occult blood test (FOBT) (12, 29, 30).
`One study found that although colonoscopy was the preferred
`CRC screening test overall, among patients who preferred a
`noninvasive test, sDNA was preferred over FOBT because of
`perceptions of superior accuracy and a less demanding prepa-
`ration (30). In addition, among patients who performed the
`sDNA test, 87% responded that the test was very easy/easy to
`perform, 91% indicated that they would be very likely/likely
`to repeat the test, and 52% replied that they had never been
`screened for CRC (31).
`In a screening study design, a prototype (version 1.0) multi-
`target sDNA assay demonstrated 52% sensitivity and 94%
`specificity for CRC in asymptomatic average-risk individu-
`als, representing a fourfold greater sensitivity over guaiac-
`based FOBT (7). Our research group has been investigat-
`ing a second generation sDNA assay that has both technical
`and biomarker enhancements. Given the formidable time and
`resource constraints involved in performing a large-scale,
`prospective screening study, and not knowing how a new
`sDNA assay would perform, we designed a two-phase study
`that included patients found by colonoscopy to have either
`CRC or a normal colon. As such, our results do not reflect
`a true screening population. In the first phase, we reported
`in a test set of subjects that the simplified sDNA test with
`only two markers (hV and DY) had a sensitivity for CRC of
`87.5%, with a specificity of 82% (12). We herein describe
`
`Table 6. Performance of Assay in Adenomas With HGD
`
`Age
`
`Location
`
`Diameter (mm)
`
`Transverse
`Sigmoid
`Sigmoid
`Rectum
`Rectum
`Rectum
`Cecum
`
`∗
`
`15
`45
`47
`32
`−
`27
`30
`
`86
`63
`71
`70
`80
`52
`62
`∗
`
`From phase 1a study.
`
`hV
`+
`+
`+
`+
`+
`+
`−
`
`DY
`−
`−
`−
`+
`+
`+
`−
`
`hV + DY
`+
`+
`+
`+
`+
`+
`−
`
`very similar results in an independent set of CRC and normal
`subjects. In the present validation set, sensitivity for CRC
`was 86% (95% CI = 72.2–93.3%) using the same cutpoints
`as the previous study. The high sensitivity occurred regard-
`le