Annals of Internal Medicine
`Stool DNA and Occult Blood Testing for Screen Detection of
`Colorectal Neoplasia
`
`David A. Ahlquist, MD; Daniel J. Sargent, PhD; Charles L. Loprinzi, MD; Theodore R. Levin, MD; Douglas K. Rex, MD;
`Dennis J. Ahnen, MD; Kandice Knigge, MD; M. Peter Lance, MD; Lawrence J. Burgart, MD; Stanley R. Hamilton, MD;
`James E. Allison, MD; Michael J. Lawson, MD; Mary E. Devens; Jonathan J. Harrington; and Shauna L. Hillman, MS
`
`Article
`
`Background: Stool DNA testing is a new approach to colorectal
`cancer detection. Few data are available from the screening setting.
`
`Objective: To compare stool DNA and fecal blood testing for
`detection of screen-relevant neoplasia (curable-stage cancer, high-
`grade dysplasia, or adenomas ⬎1 cm).
`
`Design: Blinded, multicenter, cross-sectional study.
`
`Setting: Communities surrounding 22 participating academic and
`regional health care systems in the United States.
`
`Participants: 4482 average-risk adults.
`
`Measurements: Fecal blood and DNA markers. Participants col-
`lected 3 stools, smeared fecal blood test cards and used same-day
`shipment to a central facility. Fecal blood cards (Hemoccult and
`HemoccultSensa, Beckman Coulter, Fullerton, California) were
`tested on 3 stools and DNA assays on 1 stool per patient. Stool
`DNA test 1 (SDT-1) was a precommercial 23-marker assay, and a
`novel test (SDT-2) targeted 3 broadly informative markers. The
`criterion standard was colonoscopy.
`
`Results: Sensitivity for screen-relevant neoplasms was 20% by
`SDT-1, 11% by Hemoccult (P ⫽ 0.020), 21% by HemoccultSensa
`(P ⫽ 0.80); sensitivity for cancer plus high-grade dysplasia did not
`
`differ among tests. Specificity was 96% by SDT-1, compared with
`98% by Hemoccult (P ⬍ 0.001) and 97% by HemoccultSensa (P ⫽
`0.20). Stool DNA test 2 detected 46% of screen-relevant neo-
`plasms, compared with 16% by Hemoccult (P ⬍ 0.001) and 24%
`by HemoccultSensa (P ⬍ 0.001). Stool DNA test 2 detected 46%
`of adenomas 1 cm or larger, compared with 10% by Hemoccult
`(P ⬍ 0.001) and 17% by HemoccultSensa (P ⬍ 0.001). Among
`colonoscopically normal patients, the positivity rate was 16%
`with SDT-2, compared with 4% with Hemoccult (P ⫽ 0.010)
`and 5% with HemoccultSensa (P ⫽ 0.030).
`
`Limitations: Stool DNA test 2 was not performed on all subsets of
`patients without screen-relevant neoplasms. Stools were collected
`without preservative, which reduced detection of some DNA mark-
`ers.
`
`Conclusion: Stool DNA test 1 provides no improvement over
`HemoccultSensa for detection of screen-relevant neoplasms. Stool
`DNA test 2 detects significantly more neoplasms than does Hemoc-
`cult or HemoccultSensa, but with more positive results in colono-
`scopically normal patients. Higher sensitivity of SDT-2 was particu-
`larly apparent for adenomas.
`
`Ann Intern Med. 2008;149:441-450.
`For author affiliations, see end of text.
`
`www.annals.org
`
`Colorectal cancer remains the second most common
`
`cause of death among the types of cancer (1). Al-
`though screening reduces colorectal cancer mortality (2–
`6), observed reductions have been modest (6, 7) and more
`than one half of adults in the United States have not re-
`ceived screening (8). More accurate, user-friendly, and
`widely distributable tools have the potential to improve
`screening effectiveness, acceptability, and access.
`Several molecular approaches to screening stool for
`colorectal cancer have been studied and reviewed (9, 10),
`and stool DNA testing has been jointly endorsed by the
`American Cancer Society, the U.S. Multi-Society Task
`Force on Colorectal Cancer, and the American College of
`Radiology (11). The advantages of stool DNA testing in-
`clude noninvasiveness, absence of bowel preparation or di-
`etary restrictions, and ease of access via mail courier. How-
`ever, the reported accuracy of stool DNA tests for the
`detection of colorectal neoplasia varies. In clinical studies
`that used different assays and selected groups (12–20), sen-
`sitivities ranged from 62% to 100% for colorectal cancer
`and 27% to 82% for advanced adenomas, with specificities
`ranging from 82% to 100%. In the only reported multi-
`center study on asymptomatic average-risk patients (21), a
`precommercial multitarget DNA assay (SDT-1, a proto-
`type of PreGenPlus, EXACT Sciences, Marlborough, Mas-
`
`sachusetts) detected 52% of cases of colorectal cancer,
`compared with 13% by Hemoccult (P ⫽ 0.003), at speci-
`ficities of 94.4% and 95.2%, respectively.
`The accuracy of stool DNA testing is influenced by
`both biological and technical factors. A panel of markers
`must be used to accommodate the molecular heterogeneity
`of colorectal neoplasia, and marker selection critically af-
`fects discrimination (9). Unlike occult bleeding, which is
`intermittent (22), DNA markers seem to be shed continu-
`ously by exfoliation (23). Thus, the multiple stool sam-
`pling practiced with fecal occult blood tests may not be
`necessary with stool DNA tests. However, recovery of the
`
`See also:
`
`Print
`Editors’ Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442
`Editorial comment. . . . . . . . . . . . . . . . . . . . . . . . . . 509
`Summary for Patients. . . . . . . . . . . . . . . . . . . . . . . I-20
`
`Web-Only
`Appendix
`Conversion of graphics into slides
`Audio Summary
`
`© 2008 American College of Physicians 441
`
`Geneoscopy Exhibit 1044, Page 1
`
`

`

`Article Stool DNA versus Occult Blood Testing
`
`Context
`Because the colonic mucosa constantly sheds cells, testing
`stool for cancer-related genes could be better for colorec-
`tal cancer screening than testing for occult bleeding, which
`is intermittent.
`
`Content
`A total of 3764 healthy adults had screening colonoscopy,
`fecal occult blood testing with Hemoccult and Hemoccult-
`Sensa, and both a first- and a second-generation stool DNA
`test (SDT-1 and SDT-2, respectively) for a battery of cancer
`genes. The sensitivity of SDT-1 and HemoccultSensa was
`very similar for screen-relevant neoplasms (20% and 21%,
`respectively), whereas the sensitivity of SDT-2 was 40%.
`
`Caution
`The authors could not measure the specificity of SDT-2.
`
`Implication
`A second-generation stool test for cancer genes is sub-
`stantially more sensitive than fecal occult blood testing.
`
`—The Editors
`
`minute quantities of human DNA and assay of tumor-
`specific DNA alterations from stool present technical chal-
`lenges and require exquisite laboratory sensitivity to
`achieve optimal detection rates.
`Our primary aim was to compare the precommercial
`stool DNA test (SDT-1), which was studied by Imperiale
`and colleagues (21), with widely used fecal occult blood
`tests for the detection of screen-relevant neoplasia, defined
`as curable-stage colorectal cancer (no distant metastases),
`high-grade dysplasia, or adenomas larger than 1 cm. A
`secondary aim was to explore neoplasm detection by an-
`other stool DNA test 2 (SDT-2), which uses a more
`broadly informative marker panel.
`
`METHODS
`Table 1 lists the genes used in our test panels and
`defines several key terms.
`Design
`We conducted this multicenter, prospective, triple-
`blinded trial, targeting average-risk persons, from 2001 to
`2007. A group of national experts on colorectal cancer
`screening advised on study design, and institutional review
`boards at each site approved the study. Because we did not
`know the effect of diet and medications on DNA assays,
`patients were randomly assigned at entry to group A (re-
`striction of red meat and therapeutic doses of nonsteroidal
`anti-inflammatory drugs for 3 days before and during stool
`collections) or group B (no such restrictions). All patients
`were asked not to ingest vitamin C for the 3 days before
`and during stool collections. For the companion test, we
`chose Hemoccult (Beckman Coulter, Fullerton, Califor-
`
`442 7 October 2008 Annals of Internal Medicine Volume 149 (cid:127) Number 7
`
`nia), the most widely used fecal occult blood test, which
`was used in the trials that established the benefit of screen-
`ing for fecal occult blood (2– 4). As a second companion
`test, we chose the next-generation guaiac test Hemoccult-
`Sensa (Beckman Coulter). We compared fecal blood re-
`sults from 3 stools per patient with stool DNA on 1 stool.
`Experienced technicians performed stool DNA and occult
`blood testing in separate central
`laboratories without
`knowledge of clinical findings or the results of other tests.
`All patients who completed stool collections also had
`colonoscopy, which served as the criterion standard. We
`did not have access to data until after they had been ana-
`lyzed by statisticians and released by a data monitoring
`board.
`Participants
`We recruited asymptomatic persons age 50 to 80 years
`who were at average risk for colorectal cancer from com-
`munities surrounding 22 participating academic and re-
`gional health care systems through direct mail and multi-
`media advertisements. The exclusion criteria were structural
`colorectal evaluation (endoscopic or radiographic) within 10
`years; fecal blood testing within 1 year; overt rectal bleeding
`within 1 month; previous colorectal resection; aerodiges-
`tive cancer within 5 years; inability to stop therapeutic
`doses of nonsteroidal anti-inflammatory drugs or anti-
`coagulants; coagulopathy; contraindications to colonos-
`copy; chemotherapy within 3 months; high-risk conditions
`for colorectal cancer, such as familial adenomatous polyp-
`osis, the Lynch syndrome, or other cancer syndromes; pre-
`vious colorectal cancer or adenoma; inflammatory bowel
`disease; or more than 2 first-degree relatives with colorectal
`neoplasia. Study assistants at each site registered partici-
`pants and randomly assigned them by using a Web-based
`management system; distributed fecal blood test cards,
`stool collection containers, and colonoscopy preparation
`materials; and provided instructions.
`Stool Collection and Processing
`Patients collected 3 stools by using plastic buckets
`mounted to the toilet seat. Promptly after each individual
`collection, patients smeared stool onto both windows of
`their Hemoccult and HemoccultSensa cards and then ex-
`press-shipped smeared cards and the whole stool (sealed in
`
`Table 1. Definitions
`
`Gene targets in stool DNA test panels:
`Test 1: point mutations on K-ras, APC, and p53; microsatellite marker
`BAT-26; long DNA
`Test 2: point mutations on K-ras, scanned mutator cluster region of APC,
`vimentin methylation
`Screen-relevant neoplasia: colorectal cancer, high-grade dysplasia, adenomas
`ⱖ1 cm
`Sensitivity: rate of test positivity for those with screen-relevant neoplasia
`Specificity: rate of test negativity for those without screen-relevant neoplasia
`Test positivity: rate of positive stool test results for individual colonoscopic
`findings or groups of findings
`
`www.annals.org
`
`Geneoscopy Exhibit 1044, Page 2
`
`

`

`a bucket in an insulated container cooled with ice packs) to
`the Mayo Clinic in Rochester, Minnesota. We froze the
`first stool from each participant whole at ⫺80 °C on re-
`ceipt and sent it in batches on dry ice to EXACT Sciences
`(Marlborough, Massachusetts) for DNA assay; each of the
`subsequent 2 stools were archived in aliquots at ⫺80 °C. If
`the first stool weighed less than 30 g or was received more
`than 48 hours after defecation, it was rejected for DNA
`analysis and the second or third stool (if it met inclusion
`criteria) was sent for DNA assay.
`Stool Assays
`DNA Testing
`All assays were polymerase chain reaction– based and
`were run at EXACT Sciences. Stool DNA test 1 was per-
`formed as described in Imperiale and colleagues’ study
`(21). The marker panel for SDT-1 included 21 tumor-
`specific point mutations (3 on the K-ras gene, 10 on the
`APC gene, and 8 on the p53 gene); the microsatellite-
`instability marker BAT-26; and long DNA, a marker for
`delayed apoptosis, which is characteristic of exfoliated neo-
`plastic colonocytes (12). For SDT-2, sequence-specific
`DNA markers were detected by acrylamide gel electro-
`phoresis, as described by Whitney and colleagues (24); the
`panel consisted of 3 tumor-specific markers broadly infor-
`mative for both colorectal cancer and adenomas (25): K-ras
`mutations, scanning of APC mutator cluster regions, and
`methylation of the vimentin gene. We used methods de-
`scribed elsewhere to detect mutant K-ras (12), APC scan-
`ning (25), and vimentin gene methylation (20) assays. We
`defined any positive component marker result according to
`the manufacturer’s preestablished criteria as a positive test
`result.
`
`Occult Blood Testing
`The manufacturer that developed the Hemoccult and
`HemoccultSensa cards, without rehydration, trained tech-
`nicians on-site at the Mayo Clinic. As recommended by
`the manufacturer, the technicians added the catalyst solu-
`tion to cards stored at ambient temperature within 48 to
`72 hours of collection. We defined a spreading (enlarging)
`blue color in 60 seconds in any window of the cards as a
`positive result and any other result as negative.
`Colonoscopy
`After cathartic preparation, experienced endoscopists
`performed colonoscopy in all patients. If the examination
`did not reach the cecum or inspected less than 90% of the
`mucosa, the patient was disqualified. Endophotographs
`documented cecal intubation, and the size and location of
`all lesions were recorded. Costs not covered by third parties
`were reimbursed by study funding.
`Pathologic Examination
`Local pathologists examined all endoscopically or sur-
`gically sampled lesions. A gastrointestinal pathologist at the
`coordinating site reexamined all lesions to confirm diagno-
`
`www.annals.org
`
`Stool DNA versus Occult Blood Testing
`
`Article
`
`screen-relevant neo-
`sis. Classification discrepancies of
`plasms were adjudicated by a second expert pathologist.
`We categorized patients with multiple neoplasms accord-
`ing to the most advanced lesion. For assay of markers
`in screen-relevant neoplasms, DNA was extracted from
`microdissected tissue.
`
`Statistical Analysis
`We calculated sample size to ensure adequate power to
`detect differences in sensitivity comparisons. We powered
`the study to ensure an adequate number of cases of cur-
`able-stage colorectal cancer and high-grade dysplasia and
`assumed their combined prevalence to be at least 1.5%. A
`sample size of 2900 would yield an expected 43 curable-
`stage cancer or high-grade dysplasia cases, which would
`provide 90% power to detect a 35% improvement in sen-
`sitivity of SDT-1 over the Hemoccult test by using a
`2-sided McNemar test with ␣⫽ 0.05 (assuming Hemoc-
`cult sensitivity of 25%). The protocol specified interim
`analyses at one half and three quarters of full enrollment to
`see whether it was necessary to stop the study early if test
`sensitivities differed significantly or to revise sample size
`requirements on the basis of observed prevalence of the
`target lesion. At the first interim analysis, lesion prevalence
`was lower than expected, and we readjusted the sample size
`to 4434 patients. However, before we completed enroll-
`ment, the manufacturer altered the SDT-1 assay, which
`prompted an unplanned interim analysis after 2497 pa-
`tients. On the basis of these interim results, we stopped
`SDT-1 testing and began doing the SDT-2 test.
`To accomplish a secondary aim of this trial (to see
`whether restricting diet and medication affects the specific-
`ity of the SDT test), we randomly assigned persons to
`pretest restrictions or no restrictions. The sample size cal-
`culated for the sensitivity comparison provided 85% power
`to detect a 4% difference in specificity between random-
`ization groups. Because SDT specificity was the same in
`both groups, we pooled the results for all analyses.
`We included all patients tested with SDT-1. We com-
`pared stool test sensitivities and specificities by using the
`McNemar test. We used a chi-square test or the Fisher
`exact test to compare baseline characteristics between co-
`horts and assay performance in subsets of patients. All P
`values are 2-sided.
`Per agreement with EXACT Sciences, we did the
`SDT-2 test on all patients with cancer, high-grade dyspla-
`sia, and adenomas larger than 2 cm from the full enroll-
`ment period as well as on a random sample of 50 patients
`with 1- to 2-cm adenomas and 75 with normal colonos-
`copy results. To estimate the population-level sensitivity
`for the SDT-2 test, we used all case patients tested with
`SDT-2 and reweighted the calculation to be proportional
`to the observed prevalence of each screen-relevant neopla-
`sia category in the entire population with screen-relevant
`neoplasias. Because we did not do the SDT-2 test on all
`
`7 October 2008 Annals of Internal Medicine Volume 149 (cid:127) Number 7 443
`
`Geneoscopy Exhibit 1044, Page 3
`
`

`

`Article Stool DNA versus Occult Blood Testing
`
`Figure 1. Study flow diagram.
`
`Cancellations, protocol violations,
`or ineligibility (n = 477)
`
`Colonoscopy did not reach cecum
`or view >90% of colorectum
`(n = 171)
`
`Stools not collected within 120 d
`(n = 68)
`
`Non–curable-stage colorectal
`cancer (n = 2)
`
`Enrolled (n = 4482)
`
`Retained (n = 4005)
`
`Retained (n = 3834)
`
`Retained (n = 3766)
`
`Evaluable (n = 3764)
`
`Stool testing
`
`Hemoccult and HemoccultSensa
`(n = 3764)
`
`Stool DNA test 1 before
`stoppage at interim
`analysis (n = 2497)
`
`Stool DNA test 2 on all patients with
`colorectal cancer (n = 19), high-grade
`dysplasia (n = 20), and adenoma
`≥2 cm (n = 53); 50 randomly selected
`patients with 1- to 2-cm adenoma;
`and 75 patients with normal
`colonoscopy results (n = 217)
`
`subsets without screen-relevant neoplasia, we could not cal-
`culate specificity for screen-relevant neoplasia. To compare
`test positivity rates in patient subsets, we used the McNe-
`mar test.
`Role of the Funding Source
`The National Cancer Institute funded this study and
`monitored conduct. EXACT Sciences performed DNA as-
`says at no cost, and Beckman Coulter provided Hemoccult
`and HemoccultSensa cards at no cost. EXACT Sciences
`limited SDT-2 coverage to screen-relevant neoplasms and a
`subset of normal control participants. Neither company
`influenced study oversight, data analysis, or reporting.
`
`RESULTS
`Patients
`Of the 4482 persons enrolled, 3764 (84%) were evalu-
`able. We excluded 545 patients because of cancellations,
`protocol violations, or ineligibility; 171 because of incom-
`plete colonoscopies; and 2 because of distant metastases
`
`444 7 October 2008 Annals of Internal Medicine Volume 149 (cid:127) Number 7
`
`(Figure 1). Table 2 shows demographic and colorectal le-
`sion characteristics. We found screen-relevant neoplasms in
`290 (7.7%) patients; 39 had nonmetastatic cancer or high-
`grade dysplasia and 251 had adenomas that were 1 cm or
`larger. Major complications from colonoscopy occurred in
`4 patients; no procedure-related deaths were reported.
`
`Occult Blood Testing: Hemoccult versus HemoccultSensa
`Detection sensitivities for the 290 screen-relevant neo-
`plasms found among all 3764 evaluable participants were
`10% (95% CI, 7% to 13%) with Hemoccult and 18%
`(CI, 13% to 22%) with HemoccultSensa (P ⬍ 0.001).
`Based on all 3474 participants without screen-relevant neo-
`plasia, the Hemoccult specificity of 98% (CI, 98% to
`99%) was slightly higher than that of HemoccultSensa
`(97% [CI, 96% to 97%]) (P ⬍ 0.001).
`Hemoccult and HemoccultSensa positivity rates for
`the 39 patients with colorectal cancer or high-grade dys-
`plasia were 33% (CI, 19% to 48%) and 44% (CI, 28% to
`59%), respectively (P ⫽ 0.100). For the 251 patients with
`www.annals.org
`
`Geneoscopy Exhibit 1044, Page 4
`
`

`

`Stool DNA versus Occult Blood Testing
`
`Article
`
`Table 2. Baseline Characteristics and Colorectal Findings
`
`Characteristic
`
`All Patients
`(n ⴝ 4482)
`
`Evaluable Patients*
`(n ⴝ 3764)
`
`Patients Tested with SDT-1†
`(n ⴝ 2497)
`
`Patients Tested with SDT-2‡
`(n ⴝ 217)
`
`Age, y
`Mean (SD)
`Median (range)
`Women, n (%)
`White, n (%)
`Colorectal findings, n (%)
`Screen-relevant neoplasia
`Cancer
`Stage I
`Stages II and III
`Cancer ⫹ high-grade dysplasia
`Adenoma ⱖ1 cm
`Adenoma ⬎2 cm
`Adenoma ⬍1 cm
`Hyperplastic polyps
`Other
`Normal
`
`63.8 (8.29)
`65 (50–81)
`2341 (52.2)
`4184 (93.4)
`
`–
`
`–
`–
`–
`–
`–
`–
`–
`–
`–
`
`63.7 (8.25)
`65 (50–80)
`1964 (52.2)
`3522 (93.6)
`
`290 (7.7)
`
`11 (0.3)
`8 (0.2)
`39 (1.0)
`251 (6.7)
`53 (1.4)
`785 (20.9)
`492 (13.1)
`86 (2.3)
`2111 (56.1)
`
`60.4 (7.86)
`59 (50–80)
`1348 (54.0)
`2314 (92.7)
`
`157 (6.3)
`
`6 (0.2)
`6 (0.2)
`22 (0.9)
`135 (5.4)
`21 (0.8)
`469 (18.8)
`341 (13.7)
`57 (2.3)
`1473 (59.0)
`
`66.4 (7.17)
`67 (51–80)
`108 (49.8)
`201 (92.6)
`
`142 (65.4)
`
`11 (5.1)
`8 (3.7)
`39 (18.0)
`103 (47.5)
`53 (24.4)
`Not tested
`Not tested
`Not tested
`75 (34.6)
`
`SDT ⫽ stool DNA test.
`* Patients who met all inclusion criteria. Both Hemoccult and HemoccultSensa (Beckman Coulter, Fullerton, California) were performed on all evaluable participants.
`† On the basis of results from an interim analysis, SDT-1 was terminated.
`‡ All participants with cancer, high-grade dysplasia, and adenomas ⱖ2 cm from the full enrollment period are included, as are random samples from 50 patients with 1- or
`2-cm adenomas and 75 with normal colonoscopy results.
`
`adenomas 1 cm or larger, the positivity rates were 6% (CI,
`3% to 9%) versus 14% (CI, 9% to 18%) (P ⫽ 0.001).
`Stool DNA versus Occult Blood Testing
`SDT-1 versus Occult Blood Testing
`Based on the first 2497 evaluable participants (Table
`3), the sensitivity of SDT-1 for screen-relevant neoplasia
`was higher than that of Hemoccult (20% [CI, 14% to
`26%] vs. 11% [CI, 6% to 16%]; P ⫽ 0.020) but not that
`of HemoccultSensa (21% [CI, 15% to 27%]; P ⫽ 0.80).
`For all target lesion groupings, specificities were slightly
`but significantly lower for SDT-1 than for Hemoccult
`but not HemoccultSensa, and the positive likelihood
`ratios for SDT-1 were lower than for either Hemoccult or
`
`HemoccultSensa for the more advanced groupings of
`screen-relevant neoplasms (Table 3).
`Based on test positivity data in subsets of screen-rele-
`vant neoplasms (Table 4), SDT-1 had higher detection
`rates than Hemoccult for 1- to 2-cm adenomas but not for
`any other subset. Stool DNA test 1 had a lower positivity
`rate for detecting invasive cancer than did HemoccultSensa
`(25% [CI, 5% to 57%] vs. 75% [51% to 100%]; P ⫽
`0.010).
`
`SDT-2 versus Occult Blood Testing
`Table 3 shows the sensitivity of SDT-2, Hemoccult,
`and HemoccultSensa for screen-related neoplasia. The
`
`Table 3. Summary of Test Performance
`
`Index Test
`
`Screen-Relevant
`Neoplasia, n*
`
`Positive Test
`Result, n
`
`Sensitivity
`(95% CI)
`
`No Screen-Relevant
`Neoplasia, n
`
`Negative Test
`Result, n
`
`Specificity
`(95% CI)
`
`Positive
`Likelihood
`Ratio (95% CI)
`
`Negative
`Likelihood
`Ratio (95% CI)
`
`Hemoccult (n ⫽ 2497)
`HemoccultSensa
`(n ⫽ 2497)
`SDT-1 (n ⫽ 2497)
`SDT-2 (n ⫽ 217)
`
`157
`157
`
`157
`142
`
`17
`33
`
`31
`66
`
`11 (6–16)†
`21 (15–27)§
`
`20 (14–26)
`40 (32–49)¶
`
`2340
`2340
`
`2340
`75
`
`2297
`2258
`
`2246
`NA**
`
`98 (98–99)‡
`97 (96–97)㛳
`
`5.9 (3–10)
`6.0 (4–9)
`
`96 (95–97)
`NA
`
`4.9 (3–7)
`NA
`
`0.9 (0.9–1.0)
`0.8 (0.8–0.9)
`
`0.8 (0.8–0.9)
`NA
`
`NA ⫽ not available; SDT ⫽ stool DNA test.
`* Includes curable-stage cancer, high-grade dysplasia, and adenomas ⱖ1 cm.
`† P ⫽ 0.02 for STD-1 vs. Hemoccult.
`‡ P ⬍ 0.001 for STD-1 vs. Hemoccult.
`§ P ⫽ 0.80 for STD-1 vs. HemoccultSensa.
`㛳 P ⫽ 0.40 for STD-1 vs. HemoccultSensa.
`¶ We calculated the weighted sensitvity for SDT-2 with the following equation: reweighted sensitivity ⫽ (% [colorectal cancer ⫹ high-grade dysplasia] ⫻ PR) ⫹ (%
`adenomas ⱖ2 cm ⫻ PR) ⫹ (% adenomas 1–2 cm ⫻ PR) ⫽ (0.13 ⫻ 0.49) ⫹ (0.18 ⫻ 0.57) ⫹ (0.68 ⫻ 0.34). PR ⫽ proportion of participants for that category of
`screen-relevant neoplasia in the entire population with screen-relevant neoplasia. See “Comparison of Stool DNA Tests” for statistical comparisons of SDT-1 and SDT-2 in
`participants who had both DNA tests performed.
`** We did not calculate formal specificity because SDT-2 was not performed on all subsets without screen-relevant neoplasia.
`
`www.annals.org
`
`7 October 2008 Annals of Internal Medicine Volume 149 (cid:127) Number 7 445
`
`Geneoscopy Exhibit 1044, Page 5
`
`

`

`Article Stool DNA versus Occult Blood Testing
`
`Table 4. Test Positivity Rates
`
`Stool DNA Test and Patient Subset
`
`SDT-1
`Screen-relevant neoplasms‡ (n ⫽ 157)
`Site
`Proximal§ (n ⫽ 56)
`Distal§ (n ⫽ 101)
`Cancer (n ⫽ 12)
`Cancer ⫹ high-grade dysplasia (n ⫽ 22)
`Adenoma ⱖ1 cm (n ⫽ 135)
`Adenoma ⬎2 cm (n ⫽ 21)
`Adenoma 1–2 cm (n ⫽ 114)
`Adenoma ⬍1 cm (n ⫽ 469)
`Hyperplastic polyps (n ⫽ 341)
`Other (n ⫽ 57)
`Normal (n ⫽ 1473)
`Age ⬍65 y (n ⫽ 1040)
`Age ⬎65 y (n ⫽ 433)
`
`SDT-2
`Screen-relevant neoplasms‡ (n ⫽ 142)
`Site
`Proximal§ (n ⫽ 56)
`Distal§ (n ⫽ 86)
`Cancer (n ⫽ 19)
`Cancer ⫹ high-grade dysplasia (n ⫽ 39)
`Adenoma ⱖ 1 cm (n ⫽ 103)
`Adenoma ⬎ 2 cm (n ⫽ 53)
`Adenoma 1–2 cm (n ⫽ 50)
`Normal (n ⫽ 75)
`Age ⬍65 y (n ⫽ 36)
`Age ⬎65 y (n ⫽ 39)
`
`Positive SDT Result
`(95% CI), %
`
`Positive Hemoccult
`Result (95% CI),%
`
`P Value*
`
`Positive HemoccultSensa
`Result (95% CI), %
`
`P Value†
`
`20 (14–26)
`
`20 (9–30)
`20 (12–28)
`25 (5–57)
`36 (16–56)
`17 (11–23)
`19 (5–42)
`17 (10–24)
`4 (2–6)
`5 (2–7)
`5 (1–15)
`4 (3–5)
`3 (2–4)
`5 (3–7)
`
`46 (38–55)
`
`45 (32–58)
`48 (37–58)
`58 (36–80)
`49 (33–64)
`46 (35–54)
`57 (43–70)
`34 (21–47)
`16 (8–24)
`6 (4–29)㛳
`26 (12–39)
`
`11 (6–16)
`
`7 (2–17)
`13 (6–19)
`50 (22–78)
`41 (20–61)
`6 (2–10)
`14 (3–36)㛳
`4 (1–8)㛳
`3 (1–5)
`1 (0.2–3)㛳
`2 (0.04–9)㛳
`2 (1–2)
`2 (1–2)
`2 (1–3)
`
`16 (10–22)
`
`9 (1–16)
`21 (12–30)
`47 (25–70)
`33 (19–48)
`10 (4–15)
`13 (4–22)
`6 (1–17)㛳
`4 (1–11)㛳
`3 (0–15)㛳
`5 (1–17)㛳
`
`0.020
`
`21 (15–27)
`
`0.070
`0.100
`0.30
`0.80
`0.004
`0.60
`0.004
`0.30
`0.003
`0.30
`⬍0.001
`0.009
`0.020
`
`⬍0.001
`
`⬍0.001
`⬍0.001
`0.40
`0.100
`⬍0.001
`⬍0.001
`0.002
`0.010
`0.56
`0.010
`
`11 (3–19)
`27 (18–35)
`75 (51–100)
`55 (34–75)
`16 (9–22)
`33 (13–54)
`12 (6–18)
`5 (3–7)
`4 (2–6)
`5 (1–15)‡
`3 (2–4)
`3 (2–4)
`3 (2–5)
`
`24 (17–31)
`
`13 (4–21)
`31 (22–41)
`63 (41–85)
`44 (28–59)
`17 (9–24)
`25 (13–36)
`8 (2–19)㛳
`5 (1–13)㛳
`6 (1–19)㛳
`5 (1–17)㛳
`
`0.80
`
`0.20
`0.20
`0.010
`0.20
`0.74
`0.30
`0.40
`0.70
`0.70
`1.00
`0.20
`0.40
`0.30
`
`⬍0.001
`
`⬍0.001
`0.020
`0.70
`0.60
`⬍0.001
`⬍0.001
`0.005
`0.030
`1.00
`0.010
`
`SDT ⫽ stool DNA test.
`* For SDT versus Hemoccult.
`† For SDT versus HemoccultSensa.
`‡ Patients were classified by the most advanced lesion found (in descending order of severity: cancer, high-grade dysplasia, or adenoma). The distribution of screen-relevant
`neoplasms differs from that in Table 3. See Table 2 for details of patients tested with SDT-2.
`§ Proximal and distal sites are relative to the splenic flexure. “Proximal” includes the cecum and ascending and transverse colon; “distal” includes the splenic flexure,
`descending colon, sigmoid, and rectum.
`㛳 The exact CI was used when the numerator was ⬍5.
`
`weighted estimate of SDT-2 sensitivity for screen-relevant
`neoplasms was 40% (CI, 32% to 49%), compared with the
`observed Hemoccult and HemoccultSensa sensitivities of
`11% (CI, 6% to 16%) and 21% (CI, 15% to 27%), re-
`spectively (Table 3). Table 4 shows the test positivity rates
`for subgroups of colorectal lesions.
`For adenomas 1 cm or larger, the positivity rate was
`46% (CI, 35% to 54%) by SDT-2 versus 10% (CI, 4%
`to 15%) by Hemoccult (P ⬍ 0.001) and 17% (CI, 9%
`to 24%) by HemoccultSensa (P ⬍ 0.001). Neoplasm
`site did not affect detection by SDT-2; however, detec-
`tion rates were lower for lesions proximal to the splenic
`flexure than for distal lesions with both Hemoccult (9%
`[CI, 1% to 16%] vs. 21% [CI, 12% to 30%]; P ⫽
`0.060) and HemoccultSensa (13% [CI, 4% to 21%] vs.
`31% [CI, 22% to 41%]; P ⫽ 0.010). Study compari-
`sons are based on a single stool per patient for SDT-2
`and 3 stools per patient for fecal blood tests; differences
`in test performance are larger if fecal blood test results
`are analyzed on fewer than 3 stools per patient (Figure
`2). For patients with normal colonoscopy results, the
`
`446 7 October 2008 Annals of Internal Medicine Volume 149 (cid:127) Number 7
`
`positivity rate was 16% (CI, 8% to 24%) by SDT-2,
`compared with 4% (CI, 1% to 11%) by Hemoccult
`(P ⫽ 0.010) and 5% (CI, 1% to 13%) by Hemoccult-
`Sensa (P ⫽ 0.030).
`Frequencies of DNA Markers in Neoplastic Tissue
`Nearly all of the tissues analyzed from screen-relevant
`neoplasms contained at least 1 marker from the SDT-2
`panel;
`fewer than two thirds of the tissues contained
`SDT-1 markers (Table 5).
`
`Negative Stool DNA Tests in Patients with
`Screen-Relevant Neoplasia
`Incomplete marker recovery from stools, instability of
`long DNA, and lesion size influenced test results. Recovery
`of individual markers in stool when they were present in
`tumor tissue from the same patient was 40% (17 of 42) for
`SDT-1 and 39% (51 of 130) for SDT-2. The sensitivity of
`long DNA decreased as the time from defecation to freez-
`ing the stool increased; sensitivity for screen-relevant neo-
`plasms decreased from 11% on stools received and frozen
`12 to 24 hours after defecation to 1.8% at 24 to 35 hours
`
`www.annals.org
`
`Geneoscopy Exhibit 1044, Page 6
`
`

`

`Stool DNA versus Occult Blood Testing
`
`Article
`
`Comparison of Stool DNA Tests
`We performed the 2 stool DNA tests on 69 patients
`with screen-relevant neoplasia
`and 54 with normal
`colonoscopy results. Positivity rates were higher with
`SDT-2 than with SDT-1 among the 69 patients with
`screen-relevant neoplasms (43% [CI, 32% to 55%] vs.
`20% [11% to 30%]; P ⫽ 0.001), including the subset of
`12 patients with colorectal cancer (58% [CI, 28% to 85%]
`vs. 25% [6% to 57%]; P ⫽ 0.050). Positivity rates were
`higher when comparing results for all 47 patients with
`adenomas 1 cm or larger (45% [CI, 31% to 59%] vs. 13%
`[CI, 3% to 22%]; P ⫽ 0.003), the 21 patients with ade-
`nomas larger than 2 cm (62% [CI, 38% to 82%] vs. 19%
`[CI, 5% to 42%]; P ⫽ 0.003), and the 26 patients with
`adenomas between 1 and 2 cm (31% [CI, 14% to 52%] vs.
`8% [CI, 1% to 25%]; P ⫽ 0.030). Test positivity in those
`with normal colonoscopy results was significantly higher
`with SDT-2 than with SDT-1 (13% [CI, 5% to 25%] vs.
`2% [CI, 1% to 10%]; P ⫽ 0.030).
`
`DISCUSSION
`In this multicenter study, we found that the precom-
`mercial stool DNA test (SDT-1) provides no meaningful
`improvements over the widely used fecal occult blood tests
`Hemoccult and HemoccultSensa for detection of screen-
`relevant colorectal neoplasms. Neoplasm detection rates by
`SDT-1 were no better than HemoccultSensa for any subset
`of neoplasms; in fact, HemoccultSensa detected signifi-
`cantly more cases of cancer than did SDT-1. In addition,
`the estimated positive likelihood ratios for cancer or high-
`grade dysplasia tended to be larger and the negative likeli-
`hood ratios smaller with both fecal blood tests than with
`SDT-1.
`Because of the poor performance of SDT-1, we exam-
`ined SDT-2 (which targets a potentially more informative
`marker panel) in a more limited manner. Stool DNA test 2
`had significantly better neoplasm detection rates than did
`the fecal blood tests or SDT-1. We accounted for the
`higher detection rates of SDT-2 by better detection of ad-
`vanced adenomas, a finding of particular importance for
`cancer prevention. Whereas fecal blood testing detected
`proportionately fewer proximal than distal colorectal neo-
`plasms, corroborating previous findings (26, 27), site did
`not affect detection by stool DNA testing.
`Detection of adenomas is essential for screening to
`prevent colorectal cancer. In our blinded comparison, us-
`ing colonoscopy results as the criterion standard, SDT-2
`detected 3 times more adenomas that were 1 cm or larger
`than did SDT-1, 4 times more than did Hemoccult and 3
`times more than did HemoccultSensa. Compared with
`SDT-1, all markers in the SDT-2 panel occur early in the
`adenoma-to-cancer progression (25, 28, 29) and were col-
`lectively more informative for adenomas on our tissue anal-
`yses; this accounted for the superior detection rates of
`SDT-2 over SDT-1. Our finding that guaiac testing is
`
`7 October 2008 Annals of Internal Medicine Volume 149 (cid:127) Number 7 447
`
`and to 0% at more than 35 hours after defecation (P ⫽
`0.010). The sensitivity of mutation markers did not change
`over this time range. The median size of screen-relevant
`neoplasms missed by SDT-2 was 15 mm, compared with
`20 mm for detected lesions (P ⫽ 0.003). Neither demo-
`graphic variables nor randomization group affected stool
`DNA test positivity.
`
`Positive Stool DNA Tests in Patients without
`Screen-Relevant Neoplasia
`Diet and use of nonsteroidal anti-inflammatory drugs
`affected HemoccultSensa results but not stool DNA test or
`Hemoccult results. For HemoccultSensa, positivity rates in
`colonoscopically normal patients randomly assigned to
`group A (restricted diet and medications) were 2% (CI,
`1% to 2%), compared with 4% (CI, 3% to 5%) for those
`in group B (unrestricted) (P ⫽ 0.030). In contrast, posi-
`tivity rates for the 2 stool DNA tests were unaffected
`(SDT-1: 4% for group A vs. 3% for group B [P ⫽ 0.31];
`SDT-2: 19% for group A vs. 13% for group B [P ⫽
`0.50]), as were those for Hemoccult (1% for group A vs.
`2% for group B; P ⫽ 0.170). Age influenced SDT-2 re-
`sults in patients with normal colonoscopy results, as posi-
`tivity rates increased from 6% (CI, 4% to 29%) for pa-
`tients younger than age 65 years to 26% (CI, 12% to 39%)
`for those 65 years of age or older (P ⫽ 0.020) (Table 4).
`Positivity rates for the component mar