a combination of serum markers for the early detection of...

Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Copyright © 2010 American Association for Cancer Research

A combination of serum markers for the early detection of

colorectal cancer

Running title: Serum markers for early detection of colorectal cancer

Norbert Wild,1 Herbert Andres,1 Wolfgang Rollinger,1 Friedemann Krause,2 Peter

Dilba,2 Michael Tacke,3 and Johann Karl1

1Department of New Technologies, 2Department of Biostatistics, and 3Department of

Antibody Development, Professional Diagnostics, Roche Diagnostics GmbH, Penzberg,

Germany

Grant support: none

Funding: The study was funded by Roche Diagnostics GmbH, Germany.

Corresponding author:

Dr. N. Wild, Department of New Technologies, Professional Diagnostics, Roche Diagnostics

GmbH, Nonnenwald 2, 82377 Penzberg, Germany

phone: 0049-8856-604547

fax: 0049-8856-604513

Published OnlineFirst on August 26, 2010 as 10.1158/1078-0432.CCR-10-0119

Research. on May 14, 2018. © 2010 American Association for Cancerclincancerres.aacrjournals.org Downloaded from

Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on August 26, 2010; DOI: 10.1158/1078-0432.CCR-10-0119

http://clincancerres.aacrjournals.org/


e-mail: [email protected]

Conflict of interest disclosure:

The study was funded by Roche Diagnostics GmbH, Germany. The authors are employees of

Roche Diagnostics GmbH, Germany. Patents by the authors on the results of this study,

including individual markers as well as marker combinations, have been filed.

Keywords: colorectal cancer, serum markers, FOBT, screening, marker combination

Abbreviations: FOBT, fecal occult blood test; CRC, colorectal cancer; CEA,

carcinoembryogenic antigen; FIT, fecal immunochemical test; OPN, osteopontin, Hb-hpt,

hemoglobin-haptoglobin complex; UICC, international union against cancer. CYFRA 21-1

and ELECSYS are trademarks of Roche





Translational relevance

The current practice of colorectal cancer screening is based on stool assays and imaging

methods e.g. colonoscopy. To pre-select patients, fecal occult blood testing or fecal

immunochemical testing is routinely used prior to colonoscopy and only patients with a

positive test result are referred to a specialist. While these assays are useful screening tools,

patient compliance with stool based assays tends to be low. Serum based assays for the early

detection of colorectal cancer are highly attractive as they could be integrated into any regular

health check-up without the need for additional stool sampling, thereby increasing acceptance

upon patients. The combination of serum markers presented in this study intends to trigger a

follow-up colonoscopy for a final diagnosis. It could be an alternative approach for the early

detection of colorectal cancer.





Purpose: Fecal occult blood testing (FOBT) is recommended as first line screening to

detect colorectal cancer (CRC). We evaluated markers and marker combinations in serum as

an alternative to improve the detection of CRC.

Experimental Design: Using penalized logistic regression, six markers were selected for

evaluation in 1027 samples (301 CRC, 143 patients with adenoma, 266 controls, 141 disease

controls, 176 other cancer). The diagnostic performance of each marker and of marker

combinations was assessed.

Results: To detect CRC from serum samples we tested 22 biomarkers. Six markers were

selected for a marker combination including the known tumor markers CEA and CYFRA 21-

1 as well as novel markers or markers that are less routinely used for detection of CRC:

ferritin, osteopontin, anti-p53 and seprase. CEA showed the best sensitivity at 95% specificity

with 43.9%, followed by seprase (42.4%), CYFRA 21-1 (35.5%), osteopontin (30.2%),

ferritin (23.9%) and anti-p53 (20.0%). A combination of these markers gave 69.6% sensitivity

at 95% specificity and 58.7% at 98% specificity. Focusing on UICC stages 0–III reduced the

sensitivity slightly to 68.0% and 53.3%, respectively. In a sub-collective, where matched stool

samples were available (75 CRC cases and 234 controls), the sensitivity of the marker

combination was comparable to fecal immunochemical testing (FIT) with 82.4% and 68.9%

vs. 81.8% and 72.7% at 95% and 98% specificity, respectively.

Conclusions: The performance of the serum marker combination is comparable to FIT.

This provides a novel tool for CRC screening to trigger a follow-up colonoscopy for a final

diagnosis.





The early detection of colorectal cancer (CRC) significantly improves the prognosis of

patients and is a key factor to reduce the mortality from CRC (1). Recently, the American

Cancer Society, the US Multi-Society Task Force on Colorectal Cancer, and the American

College of Radiology have issued joint guidelines for CRC screening to include well known

procedures as guaiac based fecal occult blood testing (FOBT) and fecal immunochemical

testing (FIT), colonoscopy, sigmoidoscopy and double contrast barium enema, but also two

more recent methods, computer tomography colonography (CTC) and fecal DNA testing (2).

The guidelines of the US Preventive Services Task Force have also been up-dated, but do not

include the latter two methods (3).

Serum-based, minimally invasive markers would be highly attractive for CRC screening

as they could easily be integrated in any health check-up without the need of additional stool

sampling. Though numerous biomarkers are under evaluation for the detection of CRC from

serum, none of them has sufficient sensitivity and specificity to be considered in the current

guidelines (4). CEA and carbohydrate antigens e.g. CA19-9 have been assessed more

intensely but with varying results depending on the study design and the study population (4).

Recently, nucleic acid based markers, including RNA, DNA and epigenetic modifications,

have come into focus (5-12). While some of the markers e.g. a panel of 5 RNA markers (7) or

hypomethylated SEPT-9 DNA (11), showed good sensitivities of 88 % and 68 %, the

respective specificities of 69 % and 89 % will burden a high number of patients with a false

positive result if used for screening. In a more recent study the sensitivity and specificity of

SEPT-9 was confirmed with 73% and 91%, respectively (12). GCC RNA (5) and L6 RNA (6)

showed high sensitivity with 74% and 79% at 95% and 100% specificity but have not yet

been extensively validated.





Using an ELISA format we evaluated new tumor markers together with established

markers. The clinical performance of these markers was assessed as stand alone markers or in

marker combinations with the intention not to replace but to initiate a follow-up colonoscopy

for a definite diagnosis.





Materials and Methods

Study Design and Sample Collection

Blood and stool samples were collected prospectively in two European multi-center

studies that have previously been described (13). The respective research protocols were

reviewed and approved by the appropriate ethics committees and written informed consent

was obtained from patients prior to sample collection.

All blood samples were collected before surgery or colonoscopy according to standard

operating procedures in standardized collection tubes that were provided to the participating

centers. Whole blood was collected in S-Monovette tubes without gel separator (Sarstedt).

The blood samples were centrifuged, serum was collected, dispensed into cryotubes and

stored at -70°C. After arrival in our laboratory the samples were thawed, aliquoted and stored

again at -70 °C.

Clinical samples from both multi-center studies were compiled for the evaluation (Table

1 and Figure 1). A: Control cohort with 266 patients. Patients with adenoma or inflammatory

bowel diseases were excluded. GI-healthy controls were only pre-selected to be free of any

bowel disease. No other exclusion criteria, e.g. chronic diseases, were applied to controls. B:

Advanced adenoma cohort containing 143 patients with any lesion containing high-grade

dysplasia, villous or tubovillous architecture, or tubular adenoma with at least 1 cm, and C:

CRC cohort with 301 colorectal cancer patients from both studies. Samples from 141 patients

with benign gastrointestinal conditions were collected in both studies (Table 1). Samples from

patients with other cancer diagnosis were collected in surgical units. Patients with co-

morbidities commonly found in a screening population were not excluded from the study. The

co-morbidities of the control and cancer cohort are given in Supplement S1.





[Insert Table 1 here]

[Insert Fig. 1 here]

Immunoassays

The ELISA tests were measured blinded in duplicate in serum at Roche Diagnostics

GmbH, Penzberg, Germany. Laboratory personnel were unaware of clinical data. To exclude

lot to lot variance a single assay lot was used for all samples wherever feasible and controls

were run. For each assay a sample aliquot was thawed and residual volume was discarded

after testing. The tumor markers of the Elecsys test menu from Roche Diagnostics, Mannheim,

were measured: CEA, CA15-3, CA125, CA19-9, CA72-4, CYFRA 21-1, ferritin, NSE, AFP

and HCG+β. Commercial ELISA kits were obtained for: TIMP-1 and HGF (R&D Systems),

interleukin-6 (Roche Diagnostics GmbH), interleukin-8 (BD Biosciences).

For all other biomarkers prototype ELISA kits were developed using standard assay

procedures as described previously for Nicotinamide N-methyltransferase (NNMT) (14),

PSME3 (15) and S100A12 (16). The technical performance of the prototype ELISA kits was

validated to meet the following specifications (not shown): intra-assay CV and inter-assay CV

(< 15 %), linearity of dilution (+/- 10 % of theoretical value) and analyte recovery in spiked

samples (100 +/- 20 %). While the intra-assay CV was specified to be < 15 %, it was found to

be routinely < 10 % for double determinations of patient samples. The sensitivity of the





assays was adjusted to detect base levels in healthy subjects, with the exception of anti-p53,

since healthy subjects are expected to be autoantibody negative.

For the seprase ELISA (Swiss-Prot accession: Q12884) serum samples were diluted ad

0.05% in an antibody mix in incubation buffer (40 mM phosphate buffer pH7.4, 10 mM

EDTA, 0.1% Tween 20, 2 mg/ml BSA, 1 mg/ml bovine IgG) containing monoclonal anti-

seprase antibodies D8 and D28 0.375 µg/ml each, labeled with digoxigenin and biotin,

respectively (17). After 2 hours at room temperature aliquots were transferred onto a

strepavidin-coated microtiter plate and incubated for one hour. After washing (TBS, pH7.4,

0.05% Tween 20), 100 µl of anti-digoxigenin peroxidase conjugate (30 mU/ml) were added

for one hour, washed and then followed by tetramethylbenzidine (TMB) substrate for color

development. The reaction was stopped after one hour with 2 N sulfuric acid and the

absorbance was measured at 450 / 620 nm. The seprase concentrations were calculated in

arbitrary units using a serial dilution of a human serum containing a high seprase

concentration.

Osteopontin (OPN) was measured using an ELISA based on rabbit polyclonal antibodies

against peptides covering amino acids 211–228 and 288–304 of human osteopontin (Swiss-

Prot accession: P10451). Biotinylated anti-osteopontin (288–304) antibodies were used to

capture and digoxygenylated anti-osteopontin (211–228) antibodies to detect the analyte.

Samples were diluted 1:100 into an antibody mix in incubation buffer (s. seprase assay)

containing 2.2 µg/ml of each of the anti-osteopontin antibodies. After 1 hour 100 µl aliquots

were transferred to a streptavidin-coated microtiter plate, incubated for another hour and

finally processed as described above. The assay was calibrated in pmol/l using a serial





dilution in equine serum of an osteopontin fragment, amino acids 211–314, expressed in E.

coli.

ASC (Swiss-Prot accession: Q9ULZ3) and Serpin B5 (Swiss-Prot accession: P36952)

were measured with ELISA assays using rabbit polyclonal antibodies raised against full

length recombinant protein from E.coli. Additionally an antibody against amino acids 331-

347 of Serpin B5 was raised and used as capture antibody. Capture antibodies were

biotinylated, while the detection antibodies were digoxigenylated. The respective sandwich

forming antibodies were diluted in incubation buffer (s. seprase assay) at a concentration of 1

µg/ml each, 100 µl of this mix were added to 20 µl standard or serum and incubated in a

streptavidin-coated microtiter plate. The plates were washed and processed as above. The

assays were calibrated against a serial dilution of recombinant protein from E. coli in equine

serum.

For the detection of anti-p53 auto-antibodies two biotinylated peptides representing

amino acids 11–35 and 41-60 of human p53(Swiss-Prot accession: P04637) in PBS, 0.05 %

Tween 20 at 50 ng/ml were immobilized on a streptavidin-coated microtiter plate with 100

µl/well. After washing 100 µl of serum samples diluted 1:50 in PBS / 0.05 % Tween 20 were

added and incubated for 120 min at 30° C. After washing bound IgG was labeled with 25

mU/ml anti-human-IgG-peroxidaxe-conjugate. Subsequent to washing, bound conjugate was

detected as in the seprase assay. The assay was calibrated using a highly positive serum with

1000 U/ml correspondings to an optical density of 3.0 at 450 nm.

Fecal assays, hemoglobin-haptoglobin complex and S100A12, were performed as

reported (13).





Statistical Analysis

A rule for serum biomarkers was determined on a training-test-design based on Monte

Carlo Cross Validation (MCCV). A special type of penalized logistic regression, Lasso-

Regression, was used, both as classification algorithm and as selection procedure for a subset

of relevant biomarkers (18,19).

Two different biomarker rules were established for different subsets of the study

population, one including all CRC patients and controls (algorithm I), and a second rule for

the same collective but without CRC stage IV (algorithm II) as classified according to the

staging of the “International union against cancer” (UICC). The software “R” was used for

rule generation.

The results were assessed by receiver operating characteristics (ROC) analysis and by

determining the sensitivity at a preset specificity of 95% or 98%, respectively (20).

Data were first divided randomly into training (67%) and test set (33%) and stratified by

age, gender, survey panel and site. Then a MCCV with 100 runs was executed on the training

set. The medians of the sensitivities and specificities over all 100 runs of MCCV are reported

as training results. They were used to determine an optimized set of parameters for the Lasso-

Regression. The optimized Lasso-Regression was applied to the training-set for generating a

final diagnostic rule and determining the final thresholds at the specificities of 95% and 98%.

This rule was then applied to the independent test-set. By application of the final rule to the

test-set, the test results were obtained (ROC-curves, sensitivities, specificities) for the

assessment of the diagnostic potential of the biomarkers. The final rule was also applied to the

adenoma panel and to the individual CRC stages. Due to sample numbers of the individual

cancer stages the analysis by stages included samples from training and test set.





Results

Marker candidates

Marker candidates were identified using different methods covering proteomics methods,

Affymetrix studies, literature search and through collaborations (not shown). 60 biomarkers,

including 10 established tumor markers from the Elecsys system (Roche Diagnostics,

Mannheim), were considered as potential biomarkers. Of these 52 were evaluated in 50 CRC

patients and 50 apparently healthy controls, that were not used later on in the enlarged sample

panel for the multivariate analysis (not shown). Finally 22 biomarkers were assessed in an

extensive patient panel that included screening relevant controls (Table 1).


Univariate analysis

When the biomarkers were measured in the patient collectives described in Table 1, the

median concentrations of the markers differed by a factor of < 3. While most markers gave

increased concentrations in CRC patients vs. controls, the concentrations of ferritin and

surprisingly of seprase were decreased (Table 2). When we pre-validated seprase with western

blots of tumor and adjacent healthy tissue from CRC patients, we detected increased

expression in cancer tissue (not shown) as has been described by others (21). However, when

we measured seprase in serum, we found decreased concentrations in cancer patients with a

median of 26.1 vs. 37.6 U/ml.

In a clinical setting the sensitivity of a biomarker is most relevant. Therefore, the cut-off

was set to achieve a specificity of 95% in controls that reflected the screening situation. The





control cohort was age matched with a median age of 62.0 years to the CRC patients and

included relevant gastrointestinal diseases. As no exclusion criteria besides inflammatory

bowel diseases or adenoma were applied to the control collective, patients with co-morbidities

commonly found in people aged over 50 e.g. hypertension or diabetes, were included

(Supplement S1). When the specificity was set to 95 %, none of the markers was sufficiently

sensitive to be a screening marker on its own. The established tumor marker CEA gave the

highest sensitivity with 43.9%, followed by the novel marker seprase (42.4%), CYFRA 21-1

(35.5%) and OPN (30.2%). Ferritin and anti-p53 were less sensitive with 23.9% and 20.0%,

respectively. The tumor markers AFP, CA15-3 and HCG+β reached sensitivities below 10 %

and were excluded from further evaluation. None of the markers selected for a marker

combination (s. multivariate analysis) detected advanced adenomas with sufficient sensitivity

(Table 3).


When analyzed by tumor stages, seprase was at least 10% more sensitive in early stages

than CEA or CYFRA 21-1 (Table 3), in stage I even 22.8% compared with these two markers.

In the screening relevant stages 0–III seprase reached a sensitivity of 45.4% and a very similar

sensitivity of 42.4% including all stages. CEA and CYFRA 21-1 gave a comparable

sensitivity only when all stages were considered as these markers have an excellent sensitivity

in stage IV patients. However, they were less sensitive in stages 0–III with 30.9% and 22.3%,

respectively.





Multivariate analysis

To improve the clinical performance, we evaluated marker combinations by application

of Lasso-Regression including either all stages (algorithm I) or only the more screening

relevant stages 0–III (algorithm II). Focusing on stages 0–III (Table 4), the following markers

were selected with Lasso Regression: CEA, ferritin and seprase in each of 100 runs of the

MCCV in the training-set. OPN was selected in 99, CYFRA 21-1 and anti-p53 in 68 runs,

respectively. Though additional markers were selected by the algorithm, a marked drop in the

selection frequency from 68 to 26 runs or less beyond these six markers was evident

(Supplement S2). To estimate the stability of the algorithm, the marker combination selected

in the training-set was validated in an independent test-set (Table 4). The combination of the 6

top markers significantly improved the sensitivity of CRC detection. For the evaluation of

cancer patients with stages 0-III the sensitivity at 95% specificity increased from 45.4% to

68.0 % in the test-set as compared to the best single marker seprase (Tables 3 and 4). When

all tumor stages were included, the sensitivity improved to 69.6% vs. 43.9% for CEA. A

detailed analysis by tumor stages revealed a remarkable increase of the sensitivity at 95%

specificity. The sensitivities for the stages I-III improved from 36.0%, 49.3% and 51.3%,

respectively, with the best single marker, to 44.0%, 75.4% and 71.4% with the marker

combination. Due to sample numbers, patients from the training set were included in this

analysis, which might lead to an increased sensitivity in the individual stages. In patients with

adenomas the sensitivity increased to 22.7%, but was still insufficient.






For a screening marker, that will trigger a follow-up colonoscopy for the final diagnosis,

a higher specificity would reduce the number of unwarranted colonoscopies, thereby reducing

the burden of false positive results for the patients. Hence, we investigated the sensitivity at

98% specificity. Though algorithm I lost 9.6% and algorithm II 8.4% in the training set, the

sensitivity was still high with 61.2% (all stages) and 56.2% (stages 0–III), that translated in

the test set to a sensitivity of 58.7% and 53.3%, respectively. For the individual cancer stages

the sensitivity of the serum marker combination was at 98% specificity comparable to or

higher than the sensitivities found for the individual markers in the univariate analysis at 95%

specificity (Table 3).

We applied algorithm II (stages 0-III) to CRC patients from the screening situation,

(study I, Figure 1). Based on n=12, where values for all 6 markers were available, the

sensitivity was 66.7% (34.9 - 90.1, 90% CI) and 41,7% (15.2 - 72.3, 90% CI) at 95 % and

98 % specificity, respectively. However, these results are preliminary and have to be taken

with care due to the low sample number. Clearly, the sensitivity of the marker combination

needs to be tested in an independent screening study to fully assess its potential.

The specificity of the CRC marker combination vs. disease controls, including benign

gastrointestinal diseases (Table 5), was assessed by applying the cut-off derived from controls

at 98% specificity. The specificity of the combination was 60.3% against all diseases with

variations against specific subgroups. When the specificity was assessed in patients suffering

from other cancer types, applying again the cut-off derived from controls at 98% specificity,





the specificity was 75.0% (Table 5). It ranged from excellent 97.0 % in prostate cancer to

46.7 % in lung cancer.


We compared the serum marker combination to fecal hemoglobin-haptoglobin and to a

combination of hemoglobin-haptoglobin with S100A12 (13). In a subset including 75 CRC

patients and 234 controls from the total study collective, where matched serum and stool

samples were available, the respective assays were run in parallel. In these patients the

sensitivity of the serum marker combination was higher than in the complete patient collective

(Table 4). When compared to the fecal methods, the serum marker combination appeared to

be as sensitive as hemoglobin-haptoglobin with 82.4% vs. 81.8% and 68.9% vs. 72.7%, but it

was less sensitive than the fecal marker combination that gave a sensitivity of 90.9% and

79.2% at 95% and 98% specificity, respectively.





Discussion

Biomarker based CRC screening currently applies the detection of occult fecal blood

using either FOBT or FIT though these methods have limitations in sensitivity and / or

specificity. We had recently evaluated the potential of fecal marker combinations to improve

the sensitivity of CRC screening (13). A combination of Hb-Hpt, S100A12 and, depending on

the specificity level, TIMP-1 resulted in an increased sensitivity. While such a combination of

fecal markers is useful, serum markers could be integrated in a health check-up more easily

having a significant impact on CRC screening due to improved patient compliance. In

contrast to stool samples, microbial contamination or interference by an ill-defined matrix is

not an issue with serum, implicating a more reproducible recovery of analytes. Unfortunately,

current serum markers lack sufficient sensitivity or specificity to be used as standalone

screening markers (4). In consequence marker combinations have been explored to solve this

dilemma. e.g. for CRC two studies showed an increase of the AUC from 0.75 with the best

single marker to 0.80 and 0.81 with combinations of CEA + CA72-4 or CEA + CA72-4 +

CA19-9 (22, 23), respectively.

When we applied the principles from our recent study to serum markers and reduced the

number of potential marker candidates in a stepwise procedure, we identified a combination

of six markers that detects CRC in serum with high sensitivity: Seprase, osteopontin, anti-p53

autoantibodies, ferritin, CEA and CYFRA 21-1.

An increased expression of seprase in cancer including colorectal cancer has been shown

and our western blot results from tissue confirmed this (21, 24-26). However, most

surprisingly the concentrations in serum appeared to be reduced instead, when we used an in-

house ELISA to discriminate CRC from controls. As seprase is a membrane bound protein,





this might be due to processing that could differentially lead to soluble seprase or to a loss of

an epitope of the specific antibodies used in our assay. Alternatively, isoforms might be

considered. However, only an additional, intracellular isoform has so far been reported, which

does not explain the discrepant results in tissue and serum (27). The mechanism behind the

serum concentrations is therefore currently unclear. Although the concentrations found in

serum are in opposition to the expression in tissue, we found a higher sensitivity in early CRC

stages as has been demonstrated by Henry et al using semiquantitative immunohistochemistry

(25).

Osteopontin has been linked to various cancer types e.g. breast, colon, prostate or lung

cancer (28-31). A special focus has been its contribution to tumor cell invasion and metastasis

making it a promising tumor marker, although serum concentrations reported by Fedarko et al

were not elevated in colon cancer (31). The concentrations found in our study were only

slightly increased but still contributed to the marker combination.

This also applies to ferritin, where a lower serum concentration in CRC is correlated with

tumor size and blood loss in the bowel (32, 33). However, it appears that such small

differences of CRC vs. controls give an insufficient sensitivity for a standalone screening

marker (34).

Anti-p53 autoantibodies in cancer have been investigated in depth (35, 36). They lack

sensitivity but contribute excellent specificity. Our results confirm findings of Müller et al

that a combination of serum anti-p53 with conventional tumor markers increases the

sensitivity of cancer detection (37). For CRC it increased by 13.5% from 59.7% to 73.2%,

when anti-p53 positive status was added to CEA. The high sensitivity in this study might be

attributed to the patient collective lacking early stages almost entirely. In contrast, we





included a significant number of patients with early stage disease as we wanted to assess our

marker combination in samples closely representing a screening situation.

The markers selected by Lasso-regression are insufficient diagnostic markers when

considered on their own but showed an additive value when combined. CEA is most

commonly used in the follow-up of CRC patients after surgery: CEA is recommended for

follow-up but not for diagnosis due to its low sensitivity in early CRC stages in the guidelines

of the American Society of Clinical Oncology (38). In our study both established tumor

markers CEA and CYRA 21-1 showed a good sensitivity in stage IV but only an intermediate

performance in early stages. When all six markers were combined to analyze stages I-IV, the

sensitivity increased notably as compared to the best single marker CEA. The increase was

also evident when only stages 0–III were considered. Here especially seprase contributed to

the marker combination as it had the best performance of all markers in early stages.

The current benchmark for marker based CRC screening is FIT and any serum marker

should at least be as sensitive. The sensitivity of our marker combination in stages 0–III

(68.0%) is comparable to a Japanese screening study that reported 65.8% sensitivity, both at

95% specificity (39). In a subset of our patients we could confirm this in a side by side

comparison. The sensitivity was slightly higher in these patients than in the total study

collective suggesting that the subset was not strictly representative for all patients. However,

the combination of serum markers and FIT, represented in our study by fecal hemoglobin-

haptoglobin, achieved an almost identical performance.

The benefit of an algorithm based marker combination is an improved sensitivity at high

specificity. Simple addition of a positive status for several assays will immediately introduce

a loss of specificity. The combination of CEA with anti-p53 as done by Müller et al (37) using





a simple “AND”-rule for the addition of true positives without adding false positives is only

feasible due to the excellent specificity of anti-p53. The more markers are combined with

such a procedure, the more false positives will be included, as the specificity of these markers

is usually lower than 100%. Since different markers will generate false positive results in

different patients, it will significantly reduce the overall specificity.

The specificity of the serum combination was assessed in disease controls that included

benign gastrointestinal conditions and in patients suffering from other cancer types. A

specificity of 60.3% in disease controls might be less problematic as the patients are

symptomatic and can easily by identified by a physician in an otherwise non-symptomatic

screening population. Such patients might be referred for colonoscopy without the need of an

initial screening assay. On the other hand, a specificity of 75.0% against other cancer types

will lead to positive results in colonoscopy negative patients where no symptoms might be

obvious. While a positive rate of 25% in these patients appears high at first glance, it is also a

chance to identify undetected cancer when the follow-up is adjusted to the specific patient

situation, e.g. lung cancer in smokers or breast cancer in women. In patients where no cancer

is detected despite of a positive test result and negative follow-up procedures, monitoring is

an option, to control for changes of marker concentrations over time. This could either reveal

a false positive result, or trigger another follow-up if a significant increase is observed.

However, only longitudinal studies with high patient numbers can address this issue.

Considering the sensitivity of the novel test procedure, the specificity in normal screening

participants and the anticipated increased acceptance upon patients as compared to stool

testing, we believe that the advantages of a serum screening assay outweigh the risks

associated with it.





The primary focus of our study was to evaluate the feasibility of the detection of CRC

from serum. Though we included a significant number of patients with early CRC stages, the

study population was collected only partially in a screening setting (study I, Figure 1). The

validation of the results in a true screening population with increased patient numbers is

therefore mandatory.

In conclusion, we describe a combination of six serum markers for the early detection of

CRC that will initiate a follow-up colonoscopy for a definite diagnosis. The performance of

the marker combination is comparable to FIT and might improve the acceptance of CRC

screening in the general population as it can easily be integrated in any health check-up

without the need for stool sampling. We are currently validating the algorithm in a multicenter

screening study.





References

1. Etzioni R, Urban N, Ramsey S, et al. The case for early detection. Nat Rev Cancer

2003;3:243–52.

2. Levin B, Lieberman DA, McFarland B, et al. Screening and surveillance for the early

detection of colorectal cancer and adenomatous 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;58:130–60.

3. U.S. Preventive Services Task Force, collaborators: Calonge N, Petitti DB, Curry S, et al.

Screening for colorectal cancer: U.S. Preventive Services Task Force recommendation

statement. Ann Intern Med 2008;149:627–37.

4. Hundt S, Haug U, Brenner H. Blood markers for early detection of colorectal cancer: A

systematic review. Cancer Epidemiol Biomarkers Prev 2007;16:1935–53.

5. Bustin SA, Gyselman VG, Williams NS, Dorudi S. Detection of cytokeratins 19/20 and

guanylyl cyclase C in peripheral blood of colorectal cancer patients. Br J Cancer

1999;79:1813-20.

6. Schiedeck TH, Wellm C, Roblick UJ, Broll R, Bruch HP. Diagnosis and monitoring of

colorectal cancer by L6 blood serum polymerase chain reaction is superior to

carcinoembryonic antigen-enzyme-linked immunosorbent assay. Dis Colon Rectum

2003;46:818-25.

7. Han M, Liew CT, Zhang HW, et al. Novel blood-based, five-gene biomarker set for the

detection of colorectal cancer. Clin Cancer Res 2008;14:455-60.





8. Ng EKO, Chong WWS, Jin H, et al. Differential expression of microRNAs in plasma of

colorectal cancer patients: A potential marker for colorectal cancer screening. Gut

2009;58:1375-81.

9. Flamini E, Mercatali L, Nanni O, et al. Free DNA and carcinoembryonic antigen serum

levels: An important combination for diagnosis of colorectal cancer. Clin Cancer Res

2006;12:6985-8.

10. Lofton-Day C, Model F, deVos T, et al. DNA methylation biomarkers for blood-based

colorectal cancer screening. Clin Chem 2008;54:414-23.

11. DeVos T, Tetzner R, Model F, et al. Circulating methylated SEPT9 DNA in plasma is a

biomarker for colorectal cancer. Clin Chem 2009;55:1337-46.

12. Tänzer M, Balluff B, Distler J, et al. Performance of epigenetic markers SEPT9 and

ALX4 in plasma for detection of colorectal precancerous lesions. PloS one 2010;5:e9061

13. Karl J, Wild N, Tacke M, et al. Improved diagnosis of colorectal cancer using a

combination of fecal occult blood and novel fecal protein markers. Clin Gastroenterol

Hepatol 2008;6:1122–8.

14. Roessler M, Rollinger W, Palme S, et al. Identification of nicotinamide N-

methyltransferase as a novel serum tumor marker for colorectal cancer. Clin Cancer Res

2005;11:6550–7.

15. Roessler M, Rollinger W, Mantovani-Endl L, et al. Identification of PSME3 as a novel

serum tumor marker for colorectal cancer by combining two-dimensional polyacrylamide

gel electrophoresis with a strictly mass spectrometry-based approach for data analysis.

Mol Cell Proteomics 2006;5:2092–101.





16. Thierolf M, Hagmann M-L, 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.

17. Piñeiro-Sánchez ML, Goldstein LA, Dodt J, Howard L, Yeh Y, Chen W-T. Identification

of the 170-kDa Melanoma membrane-bound gelatinase (Seprase) as a serine integral

membrane protease. J Biol Chem 1997;272:7595–601.

18. Tibshirani B. Regression shrinkage and selection via the Lasso. J R Statist Soc B

1996;58:267–88.

19. Park MY, Hastie T. L1 regularization path algorithm for generalized linear models. J R

Statist Soc B 2007;69:659-77.

20. Zweig MH, Campell G. Receiver-operating characteristics plots: A fundamental

evaluation tool in clinical medicine. Clin Chem 1993;93:561–77.

21. Iwasa S, Okoka K, Chen W-T, et al. Increased expression of seprase, a membrane-type

serine protease, is associated with lymph node metastasis in human colorectal cancer.

Cancer Letters 2005;227:229–36.

22. Carpelan-Homström M, Louhimo J, Stenman U-H, Alfthan H, Haglungd C. CEA, CA

19-9 and CA 72-4 improve the diagnostic accuracy in gastrointestinal cancers. Anticancer

Res 2002;22:2311-6.

23. Carpelan-Holmström M, Louhimo J, Stenman U-H, Alfthan H, Järvinen H, Haglungd C.

Estimating the probability of cancer with several tumor markers in patients with

colorectal disease. Oncology 2004;66:296-302.





24. Kennedy A, Dong H, Chen D, Chen W-T. Elevation of seprase expression and promotion

of an invasive phenotype by collagenous matrices in ovarian tumor cells. Int J Cancer

2009;124:27-35.

25. Henry LR, Lee H-O, Lee JS, et al. Clinical implications of fibroblast activation protein in

patients with colon cancer. Clin Cancer Res 2007;13:1736-41.

26. Chen D, Kennedy A, Wang J-Y, et al. Activation of EDTA-resistant gelatinases in

malignant human tumors. Cancer Res 2006;66:9977-85.

27. Goldstein LA, Chen W-T. Identification of an alternatively spliced seprase mRNA that

encodes a novel intracellular isoform. J Biol Chem 2000;275:2554–9.

28. Rohde F, Rimkus C, Friederichs J, et al. Expression of osteopontin, a target gene of de-

regulated Wnt signaling, predicts survival in colon cancer. Int J Cancer 2007;121:1717-

23.

29. Irby RB, McCarthy SM, Yeatman TJ. Osteopontin regulates multiple functions

contributing to human colon cancer development and progression. Clin Exp Metastasis

2004;21:515-23.

30. McAllister SS, Gifford AM, Greiner AL, et al. Systemic endocrine instigation of indolent

tumor growth requires osteopontin. Cell 2008;133:994-1005.

31. Fedarko NS, Jain A, Karadag A, van Eman MR, Fisher LW. Elevated serum bone

sialoprotein and osteopontin in colon, breast, prostate, and lung cancer. Clin Cancer Res

2001;7:4060-6.

32. Kishida T, Sato J, Fujimori S, et al. Clinical significance of serum iron and ferritin in

patients with colorectal cancer. J Gastroenterol 1994;29:19-23.





33. Li F, Kishida T, Kobayashi M. Serum iron and ferritin levels in patients with colorectal

cancer in relation to the size, site, and disease stage of cancer. J Gastroenterol

1999;34:195-9.

34. Scholefield JH, Robinson MHE, Bostock K, Brown NS. Serum ferritin - screening test

for colorectal cancer? Dis Colon Rectum 1998;41:1029-32.

35. Soussi T. p53 antibodies in the sera of patients with various types of cancer: A review.

Cancer Res 2000;60:1777-88.

36. Russo Am Bazan V, Iacopetta B, Kerr D, Soussi T, Gebbia N for the TP53-CRC

Collaborative Study Group. The TP53 colorectal cancer international collaborative study

on the prognostic and predictive significance of p53 mutation: Influence of tumor site,

type of mutation, and adjuvant treatment. J Clin Oncol 2005;23:7518-28.

37. Müller M, Meyer M, Schilling T, et al. Testing for anti-p53 antibodies increases the

diagnostic sensitivity of conventional tumor markers. Int J Oncology 2006;29:973-80.

38. Locker GY, Hamilton S, Harris J, et al. ASCO 2006 Update of Recommendations for the

Use of Tumor Markers in Gastrointestinal Cancer. J Clin Oncology 2006;24:5313-27.

39. Morikawa T, Kato J, Yamaji Y, Wada R, Mitsushima T, Shiratori S. A comparison of the

immunochemical fecal occult blood test and total colonoscopy in the asymptomatic

population. Gastroenterology 2005;129:422–8.





Acknowledgements

We are indebted to S. Ahmeti, H. Bodenmueller, H. Duefel, A. Engel, S. Freiburghaus, M.

Grünefeld, M.L. Hagmann, E. Hoess, D. Knodel, T. Kott, B. Kanne, U. Kunert, M. Roeßler, P.

Stegmueller, M. Thierolf, W. Uhl and R. Vogel for their contributions to this study.

Participating Clinicians

The following clinical investigators contributed to the prospective sample collection: H.

Bertram, GPR Klinikum Rüsselsheim, Germany; R. Dahmen, Universitätsklinikum

Heidelberg, Germany; A. Eickhoff, Klinikum Ludwigshafen, Germany; H. Friess,

Universitätsklinikum Heidelberg, Germany; F. van de Mierop, AZ Sint Augustinus Wilrijk,

Belgium; H.D. Janisch, Gastroenterology Specialist, Erlangen, Germany; J. Janssen, Sint-

Elisabeth ZH, Turmhout, Belgium; K.W. Jauch, Klinikum Großhadern, Germany; H. Kalthoff,

Conbio GmbH, Hamburg, Germany; M. Khoury, Gastroenterology Specialist, Hof, Germany;

W. Kerzel, Gastroenterology Specialist, Forchheim, Germany; M. Prilutskaya, ProteoGenex,

Los Angeles, US; J.F. Riemann, Klinikum Ludwigshafen, Germany; W. Rösch, Krankenhaus

Nordwest, Frankfurt am Main, Germany; G. Rohr, Hochtaunus Kliniken GmbH, Bad

Homburg v.d.H., Germany; S. Rossol, Krankenhaus Nordwest Frankfurt am Main, Germany;

D. Scheck, Gastroenterology Specialist, St. Ingbert, Germany; W. Tacke, Gastroenterology

Specialist, Königstein, Germany; H. Thiel, Gastroenterology Specialist, Homburg/Saar,

Germany; W. Weber-Guskar, Gastroenterology Specialist, Feldafing, Germany.





Table 1. Basic characteristics of patient collectives.

Patients, n

Age, mean, median, range, y

Sex female male

CRC collective 301 66.6 / 67 (35-96) 144 157 - UICC 0 / I 6/53 65.9 / 66 (44-83) 25 34 - UICC II 68 71.2 / 72 (45-93) 34 34 - UICC III 76 63.6 / 62 (35-91) 34 42 - UICC IV 68 66.6 / 66 (42-89) 34 34 - w/o staging† 30 65.0 / 64 (43-96) 17 13 Control Collective 266 62.1 / 62 (18-89) 162 104 - GI-healthy* 135 61.6 / 62 (18-79) 83 52 - Hemorrhoids 29 55.9 / 57 (27-74) 15 14 - Diverticulosis 76 64.9 / 65 (37-89) 47 29 - Hyperplastic polyps 14 67.9 / 67 (54-86) 8 6 - Other bowel diseases 12 58.7 / 59 (37-75) 9 3 Advanced adenoma 143 66.6 / 66 (41-85) 59 84 Disease controls 141 59.1 / 62 (23-91) 76 65 - Diverticulitis 30 69.1 / 69 (51-91) 13 17 - Colitis 29 61.9 / 64 (25-90) 15 14 - Morbus Crohn 18 41.3 / 39 (23-67) 13 5 - Colitis ulcerosa 14 56.1 / 55 (24-83) 9 5 - Infection related diarrhea 12 56.2 / 61 (29-79) 8 4 - Inflammatory GI-disease 22 56.6 / 60 (32-87) 9 13 - Ulcer 12 66.7 / 69 (40-89) 8 4 - Other GI-disease 4 55.0 / 61 (31-68) 1 3 Other cancer 176 63.2 / 64 (34-88) 101 75 - Bladder 8 64.4/ 68 (53-75) 3 5 - Breast 44 62.8 / 65 (34-88) 44 - - Endometrium 15 66.3 / 67 (47-84) 15 - - Kidney 23 62.7 / 62 (49-83) 9 14 - Lung 30 65.1 / 65 (47-84) 7 23 - Ovary 23 59.9 / 61 (34-86) 23 - - Prostate 33 62.7 / 62 (51-74) - 33

* no evidence of bowel disease. † staging as given by a pathologist I-III or II-III.





Table 2. Univariate analysis of marker performance ranked by AUC of CRC vs. controls.

Marker Units Concentration, median, range Cut-off at 95 %

specificity

Sensitivity (%) at 95 % specificity

AUC CRC vs. controls

Controls (n = 266)

Advanced adenoma (n = 143)

Cancer (n = 301)

Seprase U/ml 37.6 (6.99 - 125.0) 33.7 (5.25 - 70.8) 26.1 (1.93 - 81.8) 24.3 42.4 0.79 CYFRA 21-1 ng/ml 0.83 (0.14 - 8.66) 0.99 (0.25 - 6.20) 1.55 (0.17 - 433.5) 2.1 35.5 0.78 CEA ng/ml 1.49 (0.20 - 9.03) 1.80 (0.23 - 10.0) 3.86 (0.20 - 5187) 4.8 43.9 0.77 OPN pmol/ml 405.7 (7.06 - 1264) 480.2.1 (167.4 - 1513) 627.0 (124.3 - 6396) 814.2 30.2 0.77 Il-8 pg/ml 18.6 (5.55 - 1175) 20.1 (7.66 – 243.7) 35.2 (7.84 - 2656) 71.4 27.9 0.76 Il-6 pg/ml 2.52 (0.32 - 164.5) 2.54 (1.07 - 77.5) 4.36 (0.90 - 304.1) 6.7 30.6 0.74 NNMT pg/ml 307.1 (42.7 - 2600) 372.5 (98.1 - 2035) 582.9 ( 39.3 - 2633) 958.8 30.6 0.72 HGF pg/mL 451.3 (2930 - 4207) 490.5 (253.2 - 10438) 633.9 (104.4 - 11856) 812.8 30.4 0.70 ASC pg/mL 312.8 (121.5 - 3052) 322.7 (160.4 - 1981) 434.0 (124.9 - 3163) 828.2 18.3 0.69 S100 A12 ng/mL 14.9 (2.03 - 133.0) 13.7 (0.71 - 133.6) 25.7 (5.08 - 1718) 61.9 20.3 0.69 CA72-4 U/mL 0.94 (0.36 - 24.7) 0.90 (0.37 - 106.6) 1.59 (0.48 - 853.6) 8.0 20.6 0.66 TIMP-1 ng/mL 96.6 (55.6 - 275.2) 95.84 (58.3 - 713.1) 115.9 (51.1 - 1769) 156.7 26.8 0.66 FERR ng/mL 148.6 (7.56 - 2000) 143.2 (7.86 - 1435) 85.5 (2.39 - 1921) 35.2 23.9 0.65 CA19-9 U/mL 6.60 (0.60 - 116.0) 8.14 (0.60 - 100.3) 11.6 (0.60 - 18316) 34.0 22.6 0.64 PSME-3 pg/mL 53.0 (0.00 - 1402) 59.3 (0.00 – 935.9) 80.3 (0.00 - 5263) 189.9 19.7 0.64 Serpin B5 mU/mL 149.4 (0.00 - 872.9) 193.5 (0.00 - 2069) 209.6 (0.00 - 7303) 409.0 23.9 0.62 anti-p53 U/ml 15.4 (0.00 - 327.0) 15.3 (0.00 - 538.4) 15.7 (0.00 - 42144) 43.5 20.0 0.57 NSE ng/mL 9.70 (1.19 - 23.7) 8.61 (0.54 - 30.5) 8.73 (2.11 - 315.5) 16.1 14.0 0.57 AFP IU/mL 2.00 (0.50 - 19.7) 2.23 (0.50 - 10.3) 1.76 (0.5 - 19.2) 5.8 6.6 0.56 CA125 U/mL 11.7 (1.34 - 108.3) 11.8 (2.58 - 134.6) 12.4 (0.60 - 716.3) 28.3 17.6 0.56 CA15-3 U/mL 17.2 (5.37 - 73.1) 16.5 (4.51 - 92.8) 16.3 (1.42 - 48.1) 33.3 4.0 0.53 HCG+ß mIU/mL 0.31 (0.10 - 8.24) 0.1 (0.1 - 10.9) 0.27 (0.10 - 847.7 5.3 3.7 0.51

Research.

on May 14, 2018. ©

2010 Am

erican Association for C

ancerclincancerres.aacrjournals.org

Dow

nloaded from

Author m

anuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Author M

anuscript Published O

nlineFirst on A

ugust 26, 2010; DO

I: 10.1158/1078-0432.CC

R-10-0119



Table 3. Sensitivities (%) of marker candidates by tumor stages at 95 % specificity.

N / N* CEA Seprase CYFRA 21-1 OPN Ferritin Anti-p53 Colorectal cancer

Stage 0 6 / 6 16.7 16.7 16.7 16.7 16.7 33.3 Stage I 53 / 50 13.2 36.0 13.2 13.2 13.2 16.0

Stage II 68 / 67 36.8 49.3 26.5 22.1 30.9 17.9 Stage III 76 / 76 34.2 51.3 23.7 27.6 34.2 15.8 Stage IV 68 / 66 88.2 31.8 80.9 54.4 13.2 24.2

Stage 0 – III† 233 / 229 30.9 45.4 22.3 23.2 27.0 18.8 All stages 301 / 295 43.9 42.4 35.5 30.2 23.9 20.0

Adenoma Advanced adenoma 143 / 141 7.7 11.3 8.4 9.1 9.8 5.6

* Reduced sample number for seprase and anti-p53 due to sample volume. † 30 additional samples included with staging as given by a pathologist I-III or II-III.

Research.

on May 14, 2018. ©

2010 Am



Dow

nloaded from

Author m


Author M


nlineFirst on A

ugust 26, 2010; DO

I: 10.1158/1078-0432.CC

R-10-0119



Table 4. Sensitivities of the serum marker combination. Comparison of the serum combination vs. fecal markers in a sub-colletive (75 CRC, 234 controls) with matched serum and stool samples.

N Sensitivity test-set (%) at 95% specificity Sensitivity test set (%) at 98% specificity

Training-set* Test-set† Training-set* Test-set† Algorithm I (all stages)

- Stage 0‡ 6 n.a. 50.0 (15.3 - 84.7) n.a. 33.3 (6.3 - 72.9) - Stage I‡ 50 n.a. 44.0 (32.0 - 56.6) n.a. 32.0 (21.2 - 44.5) - Stage II‡ 65 n.a. 75.4 (65.0 - 83.9) n.a. 64.6 (53.7 - 74.5) - Stage III‡ 70 n.a. 71.4 (61.2 - 80.2) n.a. 64.3 (53.8 - 73.8) - Stage IV‡ 65 n.a. 93.8 (86.5 - 97.9) n.a. 87.7 (78.9 - 93.7) - all stages§ 285 70.8 (60.0 - 79.7) 69.6 (60.7-77.4) 61.2 (47.6 - 73.9) 58.7 (49.6-67.4)

Algorithm II (stages 0-III) - Stages 0-III§ 220 64.6(53.2 - 72) 68.0 (58.0 - 76.9) 56.2 (40.8 - 66.7) 53.3 (43.2 - 63.2)

Algorithm II (adenoma) - Advanced adenoma 141 n.a. 22.7 (17.0 - 29.3) n.a. 12.1 (7.8 - 17.5)

Algorithm I: Serum combination vs. fecal markers in 75 CRC and 234 controls with matched serum and stool samples׀׀ - Serum marker combination 75 82.4 (73.5 - 89.3) 68.9 (58.9 - 77.7) - Fecal Hb-Hpt 75 81.8 (73.0 – 88.7) 72.7 (63.2 – 80.9) - Fecal marker combination:

Hb-Hpt + S100A12 75 90.9 (83.6 – 95.7) 79.2 (70.2 – 86.5)

*One hundred fold Monte Carlo Cross Validation, median plus 0.05 and 0.95 quantile. †Apparent sensitivities when the cut-off from the training set was applied, 90 % CI. ‡The cut-off from the training set was applied to all CRC samples with the respective stage, including samples from training and test set. §29 additional samples included with staging as given by a pathologist I-III or II-III. .Samples from the training and test set where matched stool samples were available. Algorithm from the training set was applied׀׀

Research.

on May 14, 2018. ©

2010 Am



Dow

nloaded from

Author m


Author M


nlineFirst on A

ugust 26, 2010; DO

I: 10.1158/1078-0432.CC

R-10-0119



N Total number of cancer patients where results were available for all six markers of the combination. Cancer patients and the control collective (n = 266) were split 2/3 (training set) and 1/3 (test set) to optimize and test algorithms I and II. n.a. Not applicable. The cut-off from the training set on CRC and controls was applied. Hb-Hpt: Hemoglobin-haptoglobin complex.

Research.

on May 14, 2018. ©

2010 Am



Dow

nloaded from

Author m


Author M


nlineFirst on A

ugust 26, 2010; DO

I: 10.1158/1078-0432.CC

R-10-0119



Table 5. Specificity of the marker combination against disease controls and other cancer

N Specificity (%)* Disease Controls 141 60.3 (53.0 - 67.2) - Chronic bowel diseases† 44 47.7 (34.6 - 61.1) - Colitis 29 62.1 (45.1 - 77.1) - Diverticulitis 30 63.3 (46.7 - 77.9) - Inflammatory GI-disease‡ 22 68.2 (48.5 – 84.0) - Other§ 4 100.0 (47.3 – 100)- Ulcer 12 66.7 (39.1 - 87.7) Other Cancer 176 75.0 (69.1 - 80.3) - Bladder cancer 8 87.5 (52.9 - 99.4) - Breast cancer 44 77.3 (64.5 - 87.1) - Endometrium 15 86.7 (63.7 - 97.6) - Kidney cancer 23 69.6 (50.4 - 84.8) - Lung cancer 30 46.7 (30.8 – 63.0) - Ovary cancer 23 69.6 (50.4 - 84.8) - Prostate cancer 33 97.0 (86.4 - 99.8)

*Apparent specificities when the algorithm was applied using a cut-off derived from the control cohort at 98 % specificity, 90 % CI. †Colitis ulcerativa, infection related diarrhea, Morbus Crohn. ‡Appendicitis, cholangitis, mesenteritis, pancreatitis, proctitis. §Carcinoid, lipoma.





Figure Legends Fig. 1. Flow chart of the patients selected for the study.




Published OnlineFirst August 26, 2010.Clin Cancer Res Norbert Wild, Herbert Andres, Wolfgang Rollinger, et al. colorectal cancerA combination of serum markers for the early detection of

Updated version

10.1158/1078-0432.CCR-10-0119doi:

Access the most recent version of this article at:

Material

Supplementary

http://clincancerres.aacrjournals.org/content/suppl/2011/01/02/1078-0432.CCR-10-0119.DC1

Access the most recent supplemental material at:

Manuscript

Authoredited. Author manuscripts have been peer reviewed and accepted for publication but have not yet been

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://clincancerres.aacrjournals.org/content/early/2010/08/26/1078-0432.CCR-10-0119To request permission to re-use all or part of this article, use this link



http://clincancerres.aacrjournals.org/lookup/doi/10.1158/1078-0432.CCR-10-0119

http://clincancerres.aacrjournals.org/content/suppl/2011/01/02/1078-0432.CCR-10-0119.DC1

http://clincancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://clincancerres.aacrjournals.org/content/early/2010/08/26/1078-0432.CCR-10-0119


a combination of serum markers for the early detection of...

Documents