CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2008;6:1122–1128
`
`Improved Diagnosis of Colorectal Cancer Using a Combination of Fecal
`Occult Blood and Novel Fecal Protein Markers
`
`JOHANN KARL,* NORBERT WILD,* MICHAEL TACKE,* HERBERT ANDRES,* URSULA GARCZAREK,‡
`WOLFGANG ROLLINGER,* and WERNER ZOLG*
`
`*Department of New Technologies, and the ‡Department of Biostatistics, Professional Diagnostics, Roche Diagnostics GmbH, Penzberg, Germany
`
`Background & Aims: Annual testing for fecal occult
`blood is recommended as first-line screening for the detec-
`tion of colorectal cancer (CRC), but is affected by limited
`sensitivity. We initiated a proteomics-based search for
`novel biomarkers to improve the sensitivity of detection of
`CRC in stool samples. Methods: Six markers, including
`immunologic fecal occult blood test (iFOBT), were evalu-
`ated in a collective of 551 samples (186 CRC, 113 advanced
`adenoma, and 252 control patients) to establish the diag-
`nostic performance
`of
`each marker
`and marker
`combinations. Results: We tested the known stool mark-
`ers hemoglobin (iFOBT), hemoglobin-haptoglobin, calpro-
`tectin, carcinoembryogenic antigen, and the novel fecal
`markers tissue inhibitor of metalloproteinase-1 (TIMP-1)
`and S100A12. The best diagnostic performance was found
`for S100A12 with an area under the curve of 0.95, followed
`by TIMP-1 (0.92), hemoglobin-haptoglobin (0.92), hemo-
`globin (0.91), calprotectin (0.90), and carcinoembryogenic
`antigen (0.66). By using Bayes logistic regression as a math-
`ematic model, the highest sensitivity (88%) for the detection
`of CRC at 95% specificity was obtained with the marker pair
`S100A12 and hemoglobin-haptoglobin. Increasing the spec-
`ificity to 98%, the combination of S100A12, hemoglobin-
`haptoglobin, and TIMP-1 resulted in a sensitivity of 82%,
`with the highest increase of sensitivity found in early tumor
`stages (international union against cancer stage I: 74% sen-
`sitivity vs 57% of the best single marker). Conclusions:
`Depending on the specificity selected, a marker pair, S100A12
`and hemoglobin-haptoglobin, or a triple combination includ-
`ing TIMP-1, allowed the detection of CRC at significantly
`higher rates than can be obtained with iFOBT alone.
`
`C olorectal cancer (CRC) is one of the most prevalent can-
`
`cers worldwide and the lifetime risk is almost 6%.1 Early
`detection is clearly a key factor in reducing mortality from
`CRC.2 Several screening regimens for CRC are recommended,
`including colonoscopy, fecal occult blood testing (FOBT), and
`fecal DNA analysis. Although colonoscopy remains the gold
`standard for the detection of colon lesions, compliance is low
`owing to uncomfortable and unpleasant preparation proce-
`dures. Other limitations of colonoscopy for primary screening
`are the risk of complications, costs, and access. In contrast,
`stool-based testing is well accepted, despite limitations such as
`low sensitivity and dietary influences. The commonly used
`guaiac-based FOBT is an effective screening tool when used
`programmatically, reducing the incidence3 and the mortality.4,5
`Superior performance can be attributed to immunochemical
`
`FOBT (iFOBT) assays, which are specific for human hemoglo-
`bin, and eliminate the need for dietary restrictions and have a
`similar or better sensitivity. A recent study including 21,805
`asymptomatic Japanese patients reported a sensitivity for
`iFOBT of 65.8% for detecting cancer (specificity, 95%).6 Not-
`withstanding improved sensitivity compared with guaiac-based
`FOBT, about one third of invasive cancer remained undetected
`in this study. New tools improving the sensitivity of CRC
`screening are therefore needed. Recently, a second-generation
`fecal DNA test with improved performance has been reported.7
`Fecal DNA testing and virtual colonoscopy now are included in
`the joint guidelines for CRC screening.8
`Applying proteomics approaches to identify new screening
`markers we analyzed the protein expression in colon tissue and
`found strongly increased expression of S100A12 in CRC.9 The
`objective of the present study was to examine the clinical per-
`formance of fecal S100A12 and other selected biomarkers for
`early detection of CRC in stool samples in comparison with
`iFOBT and to evaluate if marker combinations can improve the
`sensitivity further.
`
`Materials and Methods
`Study Design
`Stool samples were collected prospectively in 2 Euro-
`pean multicenter studies. The first study recruited patients at
`gastroenterology units in an average-risk screening population
`that underwent a preventive check by colonoscopy. Patients
`with symptoms of gastrointestinal events such as rectal bleed-
`ing, recent change in bowel habits, or lower abdominal pain,
`and if FOBT testing was performed before colonoscopy, were
`excluded. From each participant a colonoscopy was performed
`at the participating centers with preparation and sedation used
`at each site. The stool samples had to be collected before
`colonoscopy. In control patients collection also was permitted
`if performed more than 3 days after colonoscopy. The size and
`location of each lesion were recorded. A pathologist examined
`each surgical resection specimen on site to determine the diag-
`
`Abbreviations used in this paper: BLR, Bayes Logistic Regression;
`CEA, carcinoembryogenic antigen; CRC, colorectal cancer; FOBT, fecal
`occult blood test; iFOBT, immunologic fecal occult blood test; TIMP-1,
`tissue inhibitor of metalloproteinase-1; UICC,
`international union
`against cancer.
`
`© 2008 by the AGA Institute
`1542-3565/08/$34.00
`doi:10.1016/j.cgh.2008.04.021
`
`Geneoscopy Exhibit 1045, Page 1
`
`

`

`October 2008
`
`COMBINATIONS OF FECAL PROTEIN MARKERS IN CRC 1123
`
`Table 1. Basic Characteristics of Patient Collectives and Measured Marker Concentrations
`
`Patients, (n)
`
`Mean age (⫾SD), (y)
`
`Female
`
`Male
`
`Sex
`
`Controls
`Healthya
`Hemorrhoids
`Diverticulosis
`Hyperplastic polyps
`Other bowel diseases
`Advanced adenoma
`CRC collective I (all stages)b
`UICC 0/I
`UICC II
`UICC III
`UICC IV
`Without staging
`CRC collective II (all stages)c
`
`252
`132
`28
`73
`14
`5
`113
`101
`1/22
`27
`12
`23
`16
`85
`
`63.0 ⫾ 8.0
`62.3 ⫾ 6.8
`60.1 ⫾ 7.1
`64.7 ⫾ 9.5
`67.9 ⫾ 9.9
`59.2 ⫾ 6.7
`66.8 ⫾ 8.5
`68.4 ⫾ 11.5
`65.0 ⫾ 10.0
`73.1 ⫾ 10.9
`70.9 ⫾ 12.3
`69.6 ⫾ 10.3
`61.9 ⫾ 12.6
`64.0 ⫾ 11.8
`
`151
`81
`13
`46
`8
`3
`48
`48
`8
`14
`8
`12
`6
`44
`
`101
`51
`15
`27
`6
`2
`65
`53
`15
`13
`4
`11
`10
`41
`
`NOTE. Median values shown.
`aNo evidence of bowel disease.
`bNo CRC patient underwent FOBT or had visible blood in his/her stool before colonoscopy.
`cCRC patients underwent a colonoscopy because of a positive FOBT or because of visible blood in their stool; in 11 patients the reason for
`colonoscopy was unknown.
`
`nosis and the respective staging. Because of the low incidence of
`approximately 0.4% CRC patients within the preventive screen-
`ing population, cancer patients were additionally recruited in a
`second prospective study at different surgery units without any
`restrictions regarding symptoms or FOBT testing. In this study
`the stool samples were collected before surgery. The diagnosis
`of CRC was confirmed by pathologic staging of each patient by
`pathologists on site. The research protocols for both studies
`were reviewed and approved by the appropriate ethics commit-
`tees and all participants gave written informed consent.
`Clinical samples from both multicenter studies were com-
`piled for the evaluation. Group A comprised the control cohort
`with 252 patients from study I. All patients with adenoma or
`inflammatory bowel diseases were excluded. Group B com-
`prised the advanced adenoma cohort containing 113 patients
`from study I and study II with any lesion containing high-grade
`dysplasia, villous or tubovillous architecture, or tubular ade-
`noma with a diameter of at least 1 cm. Group C comprised the
`CRC cohort with 186 CRC patients from study I and study II
`(Table 1 and Figure 1). To avoid a positive bias for the iFOBT
`assay, the cancer patients were divided into collective I (no
`FOBT testing or visible blood in stool), and collective II (no
`restrictions applied). Only 4 CRC patients from collective I had
`an additional inflammatory bowel disease. To assess the influ-
`ence of tumor localization on the diagnostic result, collective I
`was subdivided into right-sided CRC (cecum to colon transver-
`sum) and left-sided CRC (flexura lienalis to rectum).
`
`Stool Sample Collection
`Each participant provided 2 different portions of ap-
`proximately 1 g of feces from one bowel movement using a
`stool
`collection tube
`(identification number 80.623.022;
`Sarstedt, Nümbrecht, Germany). Subjects were given detailed
`instructions for stool collection and no dietary or medication
`modifications were required. The stool samples were frozen at
`
`⫺20°C within 24 hours of collection and transferred within 2
`weeks to a ⫺70°C freezer. Stool samples were transported on
`dry ice to our laboratory and stored at ⫺70°C.
`
`Immunoassays
`Six
`immunoassays were measured: Hemoglobin
`(RIDASCREEN Hemoglobin, R-Biopharm, Darmstadt, Ger-
`many), Hemoglobin-haptoglobin (RIDASCREEN Hemoglobin-
`Haptoglobin), calprotectin (calprotectin test; Nova Tec Immun-
`diagnostica GmbH, Dietzenbach, Germany), tissue inhibitor of
`metalloproteinase-1 (TIMP-1)
`(Quantikine human TIMP-1;
`R&D Systems, Minneapolis, MN), and carcinoembryonic anti-
`
`Figure 1. Flow chart of the patients selected for the study.
`
`Geneoscopy Exhibit 1045, Page 2
`
`

`

`1124 KARL ET AL
`
`CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No. 10
`
`Table 1. Continued
`
`S100A12 (ng/g)
`
`Hemoglobin (␮g/g)
`
`Hemoglobin-haptoglobin (␮g/g)
`
`TIMP-1 (ng/g)
`
`Calprotectin (␮g/g)
`
`CEA (␮g/g)
`
`32.1
`
`32.2
`18.7
`54.7
`25.6
`22.5
`55.2
`2153.4
`
`393.8
`2813.3
`3746.9
`2543.6
`2314.7
`2144.5
`
`<0.2
`
`⬍0.2
`⬍0.2
`⬍0.2
`⬍0.2
`⬍0.2
`⬍0.2
`1.4
`
`0.2
`0.5
`3.6
`3.4
`3.9
`4.9
`
`<0.2
`
`⬍0.2
`⬍0.2
`⬍0.2
`⬍0.2
`⬍0.2
`⬍0.2
`3.2
`
`0.2
`2.1
`5.8
`6.2
`7.0
`6.2
`
`0.9
`
`0.6
`0.0
`1.6
`2.2
`2.0
`2.3
`92.9
`
`22.2
`126.1
`107.6
`136.6
`123.3
`93.5
`
`24.4
`
`22.4
`21.5
`26.9
`23.0
`26.2
`27.2
`420.5
`
`179.2
`550.2
`542.5
`312.8
`677.7
`350.3
`
`26.3
`
`27.9
`21.7
`27.8
`19.0
`22.9
`24.2
`51.7
`
`51.7
`40.0
`49.7
`54.5
`56.8
`42.9
`
`gen (CEA) (Roche Diagnostics GmbH, Mannheim, Germany).
`The first 3 assays were used as recommended by the manufac-
`turer; however, an optimized stool extraction procedure was
`used. The extraction ratio (1:50) and the extract dilution factor
`(1:10) of these 3 assays was maintained. TIMP-1 and CEA were
`adapted to stool using 1:6 and 1:400 extract dilutions, respec-
`tively. An enzyme-linked immunosorbent assay was developed
`in-house for the detection of S100A12. Rabbits were immu-
`nized with recombinant full-length S100A12 expressed in Esch-
`erichia coli. The immunoglobulin G fraction was biotinylated or
`digoxigenylated to build a sandwich enzyme-linked immu-
`nosorbent assay using streptavidin-coated plates. Stool extracts
`were diluted 1:25 in sample dilution buffer and 50 ␮L of the
`diluted sample (or standard) was transferred to each well. Sub-
`sequently, 50 ␮L of antibody mix was added containing 0.5
`␮g/mL biotinylated polyclonal antibody “S100A12” and 0.5
`␮g/mL digoxigenylated polyclonal antibody “S100A12” in assay
`buffer. The plates were incubated for 60 minutes, washed 3
`times with 350 ␮L of washing buffer, and 100 ␮L of 25 mU/mL
`digoxigenin-POD conjugate was added and incubated again for
`60 minutes. Plates were washed 3 times with 350 ␮L of washing
`buffer, 100 ␮L ABTS solution was added, and then incubated
`
`for 60 minutes. The absorbance was measured at 405/620 nm.
`Recombinant full-length S100A12 was used for calibration.
`
`Fecal Analysis
`All stool samples were processed in a single laboratory
`with a modification of a recent procedure10 using a freshly
`prepared extraction buffer (Tris 0.1 mol/L, pH 8.0, citric acid
`0.1 mol/L, urea 1.0 mol/L, CaCl2 0.01 mol/L, bovine serum
`albumin 0.5%), adding a protease inhibitor cocktail (Mini Com-
`plete EDTA–free, Roche Diagnostics GmbH). The stool samples
`were thawed and 50 to 100 mg of each sample were transferred
`to a fecal sample preparation kit (Roche Diagnostics GmbH,
`Mannheim). A 50-fold excess of extraction buffer was added by
`weight, the samples were mixed vigorously for 30 minutes,
`transferred to a 10-mL tube, and centrifuged at 1200 ⫻ g for 10
`minutes. Supernatants were filtered using a 5-␮m cut-off filter
`(Ultrafree-CL, Millipore, Schwalbach, Germany), aliquoted, and
`stored at ⫺70°C. The stool extracts were randomized and all
`biomarkers were measured in our laboratory with the exception
`of calprotectin, which was determined at an external site (A¨rzte
`für Labormedizin Limbach und Kollegen, Heidelberg, Ger-
`many). Each marker was assessed independently from the same
`
`Table 2. Marker Stability in Stool Extracts
`
`Marker
`
`Positive samples
`(n)
`
`Mean recovery (⫾SD)
`after 1 day at RT (%)
`
`Minimum recovery
`after 1 day at RT
`(%)
`
`Mean recovery (⫾SD)
`after 3 days at RT
`(%)
`
`S100A12
`Hemoglobin
`Hemoglobin-haptoglobin
`TIMP-1
`Calprotectin
`CEA
`
`20
`18
`15
`7
`20
`20
`
`RT, room temperature.
`aOne sample at the lowest detection limit of the assay.
`
`87 ⫾ 15
`79 ⫾ 23
`78 ⫾ 33
`97 ⫾ 32a
`96 ⫾ 10
`99 ⫾ 6
`
`52
`45
`33
`18a
`74
`95
`
`73 ⫾ 20
`59 ⫾ 30
`72 ⫾ 30
`100 ⫾ 44a
`94 ⫾ 19
`100 ⫾ 5
`
`Geneoscopy Exhibit 1045, Page 3
`
`

`

`October 2008
`
`COMBINATIONS OF FECAL PROTEIN MARKERS IN CRC 1125
`
`Table 3. Univariate Results
`
`Sensitivity at 95% specificity, %
`
`Marker
`
`Median AUCa
`(p5–p95)
`
`CRC
`collective
`Ib
`
`CRC
`collective
`IIc
`
`Advanced
`adenomac
`
`S100A12
`Hemoglobin
`Hemoglobin-
`haptoglobin
`TIMP-1
`Calprotectin
`CEA
`
`0.95 (0.90–0.98)
`0.91 (0.85–0.95)
`0.92 (0.88–0.97)
`
`0.92 (0.87–0.96)
`0.90 (0.84–0.95)
`0.66 (0.57–0.73)
`
`82
`82
`82
`
`73
`62
`21
`
`84
`87
`85
`
`72
`56
`20
`
`13
`20
`20
`
`10
`12
`8
`
`AUC, area under the curve of a receiver operating characteristics
`graph.
`aOne hundred–fold Monte Carlo cross-validation, median plus ⬎0.05
`and 0.95 quantile.
`bOne hundred–fold Monte Carlo cross-validation, median.
`cSensitivities were estimated in predictions (rule generated with CRC
`collective I vs controls).
`
`stool extract in duplicate. Laboratory personnel were unaware
`of clinical data. The maximum biomarker concentration of the
`respective sample pairs was used for further analysis.
`
`Statistical Analysis
`The diagnostic potential of the biomarkers was eval-
`uated by receiver operator characteristics curves11 and by
`determining the sensitivity at a preset specificity of 95% or
`98%, respectively. Bayes Logistic Regression (BLR) was used
`as a mathematic model for marker combinations12 as imple-
`mented in the Bayesian binary regression software. Results of
`the BLR were evaluated by 100 runs in a Monte-Carlo cross-
`validation design13 applied on CRC collective I and controls.
`Within each run two thirds of all cases and controls, respec-
`tively, were selected randomly as a training set. BLR was
`
`Table 4. Sensitivities of Marker Combinations
`
`Figure 2. Receiver-operating characteristic curves of S100A12 and
`marker combinations in stool. The markers have been determined in
`CRC-collective I (101 patients) and controls (252 patients).
`
`applied to the training set to generate a diagnostic rule. A
`threshold on the estimated posterior case probabilities was
`determined on the controls of the training set to achieve an
`apparent specificity of 95% or 98% for the multivariate diag-
`nostic rule. This rule then was applied to the remaining third
`of the data to estimate sensitivity and specificity at the given
`threshold. All multivariate results on the CRC collective I are
`therefore reported as median sensitivities from cross-valida-
`tion. BLR then was applied to all samples in collective I to
`learn a final diagnostic rule and again its thresholds were
`determined. This rule with these thresholds then was applied
`to subgroups of the CRC collective I (UICC stages, left and
`right colon) and the apparent sensitivities are reported to see
`trends in stages and location of the cancer. Note that these
`
`Collective
`
`Patients
`
`Hemoglobin Hemoglobin-haptoglobin S100A12
`
`S100A12 ⫹
`hemoglobin-haptoglobin
`
`S100A12 ⫹
`hemoglobin-haptoglobin ⫹
`TIMP-1
`
`Sensitivity (%) at a specificity of 95%
`
`Median sensitivities from
`cross-validation
`CRC collective I
`Apparent sensitivities when
`applying the final
`multivariate rule to
`subcollectives
`UICC 0/I
`UICC II
`UICC III
`UICC IV
`Left-sided CRC
`Right-sided CRC
`CRC collective II
`Advanced adenomas
`
`101a
`
`82
`
`82
`
`82
`
`88
`
`23
`27
`12
`23
`70
`31
`85
`113
`
`74
`85
`92
`83
`83
`81
`87
`20
`
`78
`81
`83
`83
`81
`81
`85
`20
`
`57
`96
`92
`87
`84
`81
`84
`13
`
`78
`93
`92
`96
`90
`84
`88
`22
`
`88
`
`78
`96
`83
`96
`89
`87
`88
`20
`
`aSixteen CRC patients with unknown UICC staging were included; hence patients with known stages will not sum up to 101.
`
`Geneoscopy Exhibit 1045, Page 4
`
`

`

`1126 KARL ET AL
`
`CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No. 10
`
`estimates by construction can be overoptimistic, but the
`trends still give important information. The final rule also
`was used to predict test results for the CRC collective II and
`the adenomas. The apparent sensitivities in these groups
`were estimated without any overfit caused by learning.
`
`Results
`Marker Candidates
`Six marker candidates were evaluated alone or in com-
`bination for the detection of CRC in stool samples: hemoglobin
`(iFOBT), hemoglobin-haptoglobin, calprotectin, CEA, TIMP-1,
`and S100A12.
`
`Analyte Stability in Stool Extracts
`The analyte stability was determined in stool extracts
`after storage at room temperature for 1 or 3 days, respectively.
`CEA (99% and 100%), calprotectin (96% and 94%), and TIMP-1
`(97% and 100%) were very stable, followed by S100A12 (87% and
`73%). Hemoglobin and hemoglobin-haptoglobin appeared to be
`less stable (Table 2). The interpatient variability of analyte
`recovery for hemoglobin and hemoglobin-haptoglobin was
`higher than for the other biomarkers.
`
`Analyte Concentrations
`All biomarkers were measured in the patient collec-
`tives described in Table 1. The median level of S100A12 was
`comparable in both CRC collectives (2153 vs 2145 ng/g),
`being much lower in the control group (32.1 ng/g), but
`without striking differences between the control subgroups
`(Table 1). The levels in the advanced adenoma collective were
`comparable with the concentrations found in the diverticu-
`losis group (55.2 vs 54.7 ng/g), and only twice as high as in
`other controls. Similar results were found for TIMP-1 and
`calprotectin. With hemoglobin and hemoglobin-haptoglobin
`
`the median level in the control group was below the mea-
`suring range of the assays. In both assays the median was
`higher in CRC collective II than in collective I owing to the
`different inclusion criteria.
`
`Univariate Analysis
`The diagnostic performance for distinguishing CRC
`from controls was determined by receiver operator characteris-
`tic curve analysis. S100A12 showed the best discrimination
`followed by TIMP-1, hemoglobin-haptoglobin, hemoglobin,
`and calprotectin, whereas CEA did not reach a diagnostic rele-
`vant discriminatory power (Table 3). More important for a
`screening marker is the sensitivity at a high specificity level,
`which was arbitrarily set to 95%. S100A12, hemoglobin, and
`hemoglobin-haptoglobin all achieved a high sensitivity of 82%
`in CRC collective I (Table 3). The detection rate of advanced
`adenoma was low for all markers.
`
`Multivariate Analysis
`To test if marker combinations can improve the clinical
`performance we combined the markers using BLR. As shown in
`Figure 2, the area under the curve of S100A12 could be increased
`further by marker combinations (from 0.95 to 0.96). More impor-
`tant is the improvement of the sensitivity with the best single
`markers from 82% to 88% by combining S100A12 and hemoglo-
`bin-haptoglobin at 95% specificity (Table 4). A combination of
`more than 2 markers did not increase the sensitivity further.
`Although achieving a high sensitivity is of prime importance for
`CRC patients, a high specificity also is crucial for a screening
`marker to avoid distress by false-positive results. Hence, we deter-
`mined the sensitivity of marker combinations at the even more
`restrictive specificity of 98%. The best sensitivity could be achieved
`using a combination of hemoglobin-haptoglobin, S100A12, and
`TIMP-1 (Table 4). Especially in early cancer stages UICC I and II,
`a strong increase was seen from 57% to 74% and from 74% to 93%,
`respectively, comparing univariate and multivariate analysis. Left-
`
`Table 4. Continued
`
`Sensitivity (%) at a specificity of 98%
`
`Hemoglobin
`
`Hemoglobin-haptoglobin
`
`S100A12
`
`S100A12 ⫹
`hemoglobin-haptoglobin
`
`S100A12 ⫹
`hemoglobin-haptoglobin ⫹
`TIMP-1
`
`70
`
`52
`70
`75
`78
`73
`65
`79
`12
`
`73
`
`57
`74
`75
`78
`71
`74
`78
`12
`
`67
`
`30
`81
`75
`70
`67
`61
`66
`4
`
`79
`
`52
`81
`83
`87
`76
`77
`82
`9
`
`82
`
`74
`93
`83
`91
`86
`84
`86
`12
`
`Geneoscopy Exhibit 1045, Page 5
`
`

`

`October 2008
`
`COMBINATIONS OF FECAL PROTEIN MARKERS IN CRC 1127
`
`sided or right-sided cancer was detected with a comparable sensi-
`tivity irrespective of the location.
`
`Discussion
`The aim of this study was to improve the early detection
`of CRC in stool samples. We evaluated the clinical performance of
`fecal S100A12, a novel marker identified in our former proteomics
`study,9 either alone or in combination with other selected biomar-
`kers in stool samples. Besides hemoglobin (iFOBT)6,14,15 as refer-
`ence we also included hemoglobin-haptoglobin complex,16 CEA,17
`and calprotectin,18 –20 as well as TIMP-1, which has been described
`in plasma from CRC patients.21,22
`Few data have so far been published for S100A12 as a cancer
`marker. In contrast to a recent study23 we found increased levels
`of S100A12 in tumor tissue from CRC patients as well as high
`concentrations in serum of CRC patients versus healthy con-
`trols, but a loss of specificity in inflammatory diseases.9 Speci-
`ficity problems in serum are not surprising because increased
`levels of S100A12 have been shown in the context of inflam-
`mation.24 –27 In contrast, a fecal assay will be specific for gastro-
`intestinal diseases, for which gastrointestinal inflammation still
`would be detected, but interference from nongastrointestinal
`inflammation would be less of a problem.
`We found high concentrations of S100A12 in stool samples
`from CRC patients versus control patients giving the best overall
`diagnostic performance with an area under the curve of 0.95. At
`95% specificity S100A12 reached the same sensitivity (82%) as
`hemoglobin and hemoglobin-haptoglobin in CRC collective I.
`Fecal TIMP-1 achieved 73%. Inflammation might lead to an over-
`estimated sensitivity for S100A12. However, only 4 patients in
`CRC collective I had an additional inflammatory disease. Looking
`specifically at these inflammatory CRC patients, S100A12, iFOBT,
`and TIMP-1 gave an identical clinical classification (3 positive and
`1 negative). From this limited number of inflammatory patients a
`bias for S100A12 in our study was not detectable, but should be
`considered in follow-up studies.
`Importantly, calprotectin was less sensitive than S100A12,
`although both markers are expressed in granulocytes and have
`similar biological properties. Our result of 62% sensitivity cor-
`relates well with the rate of 63% found in a Norwegian screening
`study.20 In our hands, CEA showed a limited diagnostic perfor-
`mance that did not support the findings of Kim et al.17
`In CRC collective II including FOBT-positive patients or
`patients with rectal bleeding in initial stool samples, a slightly
`higher sensitivity was detected for hemoglobin and hemoglo-
`bin-haptoglobin, whereas the remaining markers were indepen-
`dent of this preselection. However, 100% sensitivity in collective
`II was not achieved with these 2 assays because fecal blood–
`based assays rely on intermittent and localized bleeding. In our
`study the sensitivity of iFOBT was much higher than in a large
`study from Japan that reported a sensitivity of 65.8% for CRC
`(specificity at 95%) when screening 21,805 asymptomatic pa-
`tients.6 Because both studies applied similar collection schemes,
`sampling stool from a single bowel movement, 2 factors might
`account for this discrepancy. Our extraction procedure ap-
`peared to release hemoglobin more efficiently than the method
`recommended by the manufacturer of the hemoglobin assay
`used in our study (data not shown). In addition, our study was
`not conducted in a true screening situation in which the sen-
`sitivity might be lower because of a higher percentage of earlier-
`stage samples. However, the sensitivity for the detection of
`
`adenomas with iFOBT at 95% specificity was comparable with
`the 20% found in our study versus the 20% found for adenomas
`greater than 10 mm in the study described previously.6
`Our conclusions from the univariate analysis are as follows:
`(1) S100A12 is a novel stool marker, resulting from proteomics
`approaches, with comparable performance with iFOBT; (2)
`both variants of iFOBT (hemoglobin and hemoglobin-hapto-
`globin) are comparable; and (3) TIMP-1 is an additional fecal
`biomarker with high clinical utility.
`The introduction of iFOBT has brought an improvement over
`the guaiac-based FOBT.28 However, an unacceptably high number
`of cancers were not detected in time to start effective therapy and
`none of the single markers solved this problem. Therefore, we
`investigated marker combinations. At the more restrictive require-
`ment of 98% specificity, BLR selected the combination S100A12,
`hemoglobin-haptoglobin, and TIMP-1, increasing the sensitivity in
`CRC collective I from 73% (hemoglobin-haptoglobin) to 82% (Ta-
`ble 4). However, hemoglobin-haptoglobin also might be inter-
`changeable with hemoglobin because of a very similar perfor-
`mance. By using CRC collective II we could validate our results
`with an independent patient cohort, obtaining a sensitivity of 86%.
`When we analyzed our data from CRC collective I by UICC stages
`we found a trend towards a better discrimination of CRC patients
`in early stages using the triple combination. At 98% specificity the
`combination improved the sensitivity in stage 0/I by 17% to 74%
`and in stage II by 19% to 93%, compared with the best single
`marker (Table 4). A validation of these stage-specific results will
`have to be addressed in future studies with a sufficiently large
`number of independent and unbiased patients. Hence, we did not
`re-analyze CRC collective II by stages.
`Because patients in a screening situation are likely to collect
`stool samples at home, the sampling procedure and analyte sta-
`bility are crucial. Novel sampling devices, that are prefilled with
`extraction buffer and can be mailed to the laboratory at ambient
`temperature, have been introduced. They are easy to use and
`reduce discomfort for the patient, which in turn improves patient
`compliance. TIMP-1, CEA, and calprotectin were very stable,
`whereas S100A12 showed sufficient stability, but hemoglobin and
`the hemoglobin-haptoglobin complex seemed to be more critical.
`Two limitations are apparent from our study. First, inflam-
`matory processes of the bowel lead to high levels of S100A12 in
`stool as described for Crohn’s disease29 and inflammatory bowel
`disease.30 Yet, we believe that this concurrent appearance of
`S100A12 in CRC and inflammatory bowel diseases will not
`compromise the decision to trigger colonoscopy because in
`these cases the diagnostic procedures will be adjusted to the
`medical history of the patient. Still, any false-positive result
`caused by significant gastrointestinal
`inflammation of un-
`known cause would justify a follow-up colonoscopy. Second,
`the CRC patients in our study were not recruited within an
`average-risk screening population.
`Our study describes a significant improvement in the early
`diagnosis of CRC by combining markers that aim to trigger fol-
`low-up colonoscopy. Two new fecal markers for the detection of
`CRC, S100A12 and TIMP-1, have been evaluated in this study.
`They show comparable diagnostic performance with the estab-
`lished iFOBT. The combination of S100A12, hemoglobin-hapto-
`globin, and TIMP-1 reached a sensitivity greater than 80% at a high
`specificity (98%), giving noninvasive CRC screening in stool a new
`perspective. We are planning the evaluation of this diagnostic
`
`Geneoscopy Exhibit 1045, Page 6
`
`

`

`1128 KARL ET AL
`
`CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 6, No. 10
`
`algorithm to perform colonoscopies in a multicenter screening
`study.
`
`References
`
`1. Jemal A, Murray T, Ward E, et al. Cancer statistics, 2005. CA
`Cancer J Clin 2005;55:10 –30.
`2. Etzioni R, Urban N, Ramsey S, et al. The case for early detection.
`Nat Rev Cancer 2003;3:243–252.
`3. Mandel JS, Church TR, Bond JH, et al. The effect of fecal occult-
`blood screening on the incidence of colorectal cancer. N Engl
`J Med 2000;343:1603–1607.
`4. Kronborg O, Fenger C, Olsen J, et al. Randomised study of
`screening for colorectal cancer with faecal-occult-blood test. Lan-
`cet 1996;348:1467–1471.
`5. Hardcastle JD, Chamberlain JO, Robinson MHE, et al. Random-
`ised controlled trial of faecal-occult-blood screening for colorectal
`cancer. Lancet 1996;348:1472–1477.
`6. Morikawa T, Kato J, Yamaji Y, et al. A comparison of the immuno-
`chemical fecal occult blood test and total colonoscopy in the asymp-
`tomatic population. Gastroenterology 2005;129:422– 428.
`7. Itzkowitz SH, Jandorf L, Brand R, et al. Improved fecal DNA test
`for colorectal cancer screening. Clin Gastroenterol Hepatol 2007;
`5:111–117.
`8. Levin B, Lieberman DA, McFarland B, et al. Screening and sur-
`veillance for the early detection of colorectal cancer and adeno-
`matous polyps, 2008: a joint guideline from the American Cancer
`Society, the US Multi-Society Task Force on Colorectal Cancer,
`and the American College of Radiology. CA Cancer J Clin 2008.
`Available
`from: http://caonline.amcancersoc.org/cgi/content/
`full/CA.2007.0018v1. Accessed on: March, 27 2008.
`9. Thierolf M, Hagmann ML, Pfeffer M, et al. Towards a comprehensive
`proteome of normal and malignant human colon tissue by 2-D-LC-
`ESI-MS and 2-DE proteomics and identification of S100 A12 as
`potential cancer biomarker. Proteomics Clin Appl 2008;2:11–22.
`10. Tøn H, Brandsnes Ø, Dale S, et al. Improved assay for fecal
`calprotectin. Clin Chim Acta 2000;292:41–54.
`11. Zweig MH, Campell G. Receiver-operating characteristics plots: a
`fundamental evaluation tool
`in clinical medicine. Clin Chem
`1993;93:561–577.
`12. Genkin A, Lewis DD, Madigan D. Large-scale Bayesian logistic re-
`gression for text categorization. Technometrics 2007;49:291–304.
`13. Dudoit S, van der Laan MJ. Asymptotics of cross-validated risk
`estimation in estimator selection and performance assessment.
`Stat Methodol 2005;2:131–154.
`14. Allison JE, Tekawa IS, Ransom LJ, et al. A comparison of fecal
`occult-blood tests for colorectal-cancer screening. N Engl J Med
`1996;334:155–159.
`15. Levi Z, Rozen P, Hazazi R, et al. A quantitative immunochemical
`fecal occult blood test for colorectal neoplasia. Ann Intern Med
`2007;146:244 –255.
`16. Sieg A, Thoms C, Lüthgens K, et al. Detection of colorectal
`neoplasms by the highly sensitive hemoglobin-haptoglobin com-
`plex in feces. Int J Colorectal Dis 1999;14:267–271.
`17. Kim Y, Lee S, Park S, et al. Gastrointestinal tract cancer screen-
`ing using fecal carcinoembryonic antigen. Ann Clin Lab Sci 2003;
`33:32–38.
`18. Tibble J, Sigthorsson G, Foster R, et al. Faecal calprotectin and
`faecal occult blood tests in the diagnosis of colorectal carcinoma
`and adenoma. Gut 2001;49:402– 408.
`19. Von Roon AC, Karamountzos L, Purkayastha S, et al. Diagnos-
`tic precision of fecal calprotectin for inflammatory bowel dis-
`ease and colorectal malignancy. Am J Gastroenterol 2007;
`102:803– 813.
`20. Hoff G, Grotmol T, Thiis-Evensen E, et al. Testing for fecal cal-
`protectin (PhiCal) in the Norwegian Colorectal Prevention trial on
`flexible sigmoidoscopy screening: comparison with an immuno-
`
`chemical test for occult blood (FlexSure OBT). Gut 2004;53:
`1329 –1333.
`21. Holten-Andersen MN, Christensen IJ, Nielsen HJ, et al. Total
`levels of tissue inhibitor of metalloproteinases 1 in plasma yield
`high diagnostic sensitivity and specificity in patients with colon
`cancer. Clin Cancer Res 2002;8:156 –164.
`22. Holten-Andersen MN, Nielsen HJ, Sorensen S, et al. Tissue in-
`hibitor of metalloproteinases-1 in the postoperative monitoring of
`colorectal cancer. Eur J Cancer 2006;42:1889 –1896.
`23. Hsieh HL, Schäfer BW, Sasaki N, et al. Expression analysis of
`S100 proteins and RAGE in human tumors using tissue microar-
`rays. Biochem Biophys Res Commun 2003;307:375–381.
`24. Foell D, Kane D, Bresnihan B, et al. Expression of pro-inflamma-
`tory protein S100A12(EN-RAGE) in rheumatoid and psoriatic ar-
`thritis. Rheumatology 2003;42:1383–1389.
`25. Foell D, Seeliger S, Vogl T, et al. Expression of S100A12
`(EN-RAGE) in cystic fibrosis. Thorax 2003;58:613– 617.
`26. Foell D, Kucharzik T, Kraft M, et al. Neutrophil derived human
`S100A12 (En-RAGE) is strongly expressed during chronic active
`inflammatory bowel disease. Gut 2003;52:847– 853.
`27. Foell D, Fukiko I, Vogl T, et al. S100A12 (EN-RAGE) in monitoring
`Kawasaki disease. Lancet 2003;361:1270 –1272.
`28. Guittet L, Bouvier V, Mariotte N, et al. Comparison of a guaiac
`based and an immunochem