Download Evaluation of a 5-Marker Blood Test for Colorectal Cancer Early

Published OnlineFirst November 11, 2015; DOI: 10.1158/1078-0432.CCR-15-1268
Clinical
Cancer
Research
Biology of Human Tumors
Evaluation of a 5-Marker Blood Test for Colorectal
Cancer Early Detection in a Colorectal Cancer
Screening Setting
Simone Werner1, Friedemann Krause2, Vinzent Rolny2, Matthias Strobl2,
David Morgenstern3, Christian Datz4, Hongda Chen1, and Hermann Brenner1,5
Abstract
Purpose: In initial studies that included colorectal cancer
patients undergoing diagnostic colonoscopy, we had identified
a serum marker combination able to detect colorectal cancer with
similar diagnostic performance as fecal immunochemical test
(FIT). In this study, we aimed to validate the results in participants
of a large colorectal cancer screening study conducted in the
average-risk, asymptomatic screening population.
Experimental Design: We tested serum samples from 1,200
controls, 420 advanced adenoma patients, 4 carcinoma in situ
patients, and 36 colorectal cancer patients with a 5-marker blood
test [carcinoembryonic antigen (CEA)þanti-p53þosteopontinþ
sepraseþferritin]. The diagnostic performance of individual markers and marker combinations was assessed and compared with
stool test results.
Results: AUCs for the detection of colorectal cancer and
advanced adenomas with the 5-marker blood test were 0.78
[95% confidence interval (CI), 0.68–0.87] and 0.56 (95% CI,
0.53–0.59), respectively, which now is comparable with guaiacbased fecal occult blood test (gFOBT) but inferior to FIT. With
cutoffs yielding specificities of 80%, 90%, and 95%, the sensitivities for the detection of colorectal cancer were 64%, 50%, and
42%, and early-stage cancers were detected as well as late-stage
cancers. For osteopontin, seprase, and ferritin, the diagnostic
performance in the screening setting was reduced compared with
previous studies in diagnostic settings while CEA and anti-p53
showed similar diagnostic performance in both settings.
Conclusions: Performance of the 5-marker blood test under
screening conditions is inferior to FIT even though it is
still comparable with the performance of gFOBT. CEA and
anti-p53 could contribute to the development of a multiple
marker blood-based test for early detection of colorectal
cancer. Clin Cancer Res; 1–9. 2015 AACR.
Introduction
vasive stool tests are an attractive alternative for colorectal cancer
screening due to their low cost and suitability for home use or in a
primary care setting. However, the aversion to stool handling for
gFOBT and FITs may limit the acceptance of stool tests (7). Bloodbased tests may be more readily accepted by patients for many
medical conditions, and surveys conducted in individuals eligible
for colorectal cancer screening showed a strong preference for
collection of blood over feces (8). While the development of a
simple and convenient blood test with diagnostic performance
comparable with FIT could contribute to improvement in the
acceptance of colorectal cancer screening, the diagnostic performance of currently existing blood tests is insufficient for their
routine use (9).
The identification of a blood-based multimarker panel for
detection of colorectal cancer in a marker identification program
(MIP) study has been previously described (10). While most
controls were recruited in a screening setting, colorectal cancer
cases recruited from surgery centers were used to enrich the
number of cancer cases due to the low prevalence of colorectal
cancer in the screening population (10, 11). A multivariate
analysis of 60 candidate biomarkers identified a 6-marker panel
consisting of CEA, ferritin, seprase, osteopontin, anti-p53 autoantibody, and CYFRA-21-1. With sensitivity and specificity of
69.6% and 95.0%, respectively, for the detection of colorectal
cancer, the diagnostic performance of the panel was comparable
with the performance of FIT (10).
To validate this result in the target population of screening, we
combined 5 of the 6 candidate biomarkers to a 5-marker blood
With over 1.3 million new cancer cases worldwide and almost
700,000 deaths each year, colorectal cancer is one of the most
common cancers (1). Because of the slow progression from
detectable and curable precancerous lesions to colorectal cancer
and the strong dependence of prognosis on stage at diagnosis,
early detection of colorectal cancer has great potential to reduce
the burden of this disease (2–4). Colonoscopy is the gold standard for detecting colorectal cancer and its precursors, but its
application as primary screening test is impaired by high costs,
limited capacities, and typically lower adherence (5, 6). Nonin-
1
Division of Clinical Epidemiology and Aging Research, German Cancer Research Center (DKFZ), Heidelberg, Germany. 2Roche Diagnostics GmbH, Penzberg, Germany. 3Roche Diagnostic Operations, Inc.,
Indianapolis, Indiana. 4Department of Internal Medicine, KH Oberndorf, Teaching Hospital of the Paracelsus Private Medical University of
Salzburg, Oberndorf, Austria. 5German Cancer Consortium (DKTK),
German Cancer Research Center (DKFZ), Heidelberg, Germany.
Note: Supplementary data for this article are available at Clinical Cancer
Research Online (http://clincancerres.aacrjournals.org/).
Corresponding Author: Hermann Brenner, Division of Clinical Epidemiology and
Aging Research (C070), German Cancer Research Center, Im Neuenheimer Feld
581, Heidelberg D-69120, Germany. Phone: þ49-6221-421300; Fax: þ49-6221421302; E-mail: [email protected]
doi: 10.1158/1078-0432.CCR-15-1268
2015 American Association for Cancer Research.
www.aacrjournals.org
Downloaded from clincancerres.aacrjournals.org on August 12, 2017. © 2015 American Association for Cancer
Research.
OF1
Published OnlineFirst November 11, 2015; DOI: 10.1158/1078-0432.CCR-15-1268
Werner et al.
Translational Relevance
Blood tests for early detection of colorectal cancer would be
a highly attractive alternative to endoscopic examinations and
stool tests. In the last decade, many candidate biomarkers were
identified in studies that included symptomatic colorectal
cancer cases recruited in clinics. However, the majority of
these biomarkers failed in subsequent studies under screening
conditions which stresses the importance of the study setting
for biomarker discovery and validation. Here, we tested 1,660
blood samples from participants of screening colonoscopy
with a 5-marker blood test [carcinoembryonic antigen (CEA)
þ anti-p53 þ osteopontin þ seprase þ ferritin]. The diagnostic
performance of the 5-marker test was comparable with guaiacbased fecal occult blood test (gFOBT) but inferior to fecal
immunochemical test (FIT). Of note, a combination of antip53 and CEA was sufficient to reach the same diagnostic
performance under screening conditions as the whole 5-marker panel suggesting preference for these two markers for future
multimarker panel development.
test and retrospectively analyzed 1,660 prospectively collected
blood samples from all available colorectal cancer cases, most
advanced adenoma cases, and a selection of controls of the BliTz
study collective. The BliTz study is an ideal setting for such a
validation, as both cases and controls were recruited prior to
diagnosis at screening colonoscopy (12–14). We aimed to determine the diagnostic performance of the 5-marker blood test for
detection of colorectal cancer and advanced adenomas in the
screening setting. Furthermore, we intended to compare the
results with previous results from the diagnostic setting and with
results of a commercial FIT and a commercial gFOBT.
Materials and Methods
Study design and sample selection
Samples were drawn from the BliTz study collective. BliTz
(Begleitende Evaluierung innovativer Testverfahren zur Darmkrebs-Fr€
uherkennung) is an ongoing prospective screening study
conducted in cooperation with more than 20 gastrointestinal
practices in southwestern Germany. Detailed information on
the Blitz study can be found elsewhere (12–14). Briefly, since
the end of 2005 precolonoscopy stool and blood samples from
more than 7,000 participants of screening colonoscopy were
collected. After colonoscopy, basic demographic and clinical
data were extracted from colonoscopy and histology reports in
a standardized manner by trained research assistants who were
blinded to blood and stool test results. Cancer stages were
classified according to the UICC (Union for International
Cancer Control) classification and advanced adenomas were
defined as adenomas with at least one of the following features:
1 cm in size, tubulovillous or villous components, and highgrade dysplasia. Further patient data were collected with a short
patient questionnaire. Personal and clinical data were stored
separately in access databases for reasons of privacy protection.
The study was approved by the Ethics Committee of the
University of Heidelberg (Heidelberg, Germany) and informed
consent was obtained from all participants. Auditing was conducted by the coordinating study center at the German Cancer
OF2 Clin Cancer Res; 2016
Research Center and by the Coordination Centre for Clinical
Trials (KKS) Heidelberg (Heidelberg, Germany).
For the current analysis, participants with colorectal cancer
(N ¼ 36), participants with carcinoma in situ (CIS, N ¼ 4), and
participants with advanced adenoma (N ¼ 420) were compared
with controls without colorectal neoplasms (N ¼ 1,200). While
the colorectal cancer and the CIS group were comprised of all
available colorectal cancer/CIS patients from the BliTz study
recruited until the February 20, 2013, for the advanced adenoma
and the control group, representative samples were selected from
the available BliTz participants. BliTz participants whose blood
was taken after colonoscopy or for whom the date of blood
withdrawal was unknown were excluded. Furthermore, study
participants with a history of colorectal cancer or inflammatory
bowel disease, participants with a previous colonoscopy within
the last 5 years, and persons ages below 50 or over 79 years were
excluded, because these participants would typically not be considered to be the target population for colorectal cancer screening.
In the control group also participants with insufficient bowel
preparation before colonoscopy or incomplete colonoscopy were
excluded to avoid false-negative colonoscopy results. Because this
study was conducted in a true screening population in which
colorectal cancer and adenoma patients are expected to be, on
average, slightly older and to include a larger proportion of men,
we did not match for these factors as this might lead to biased
estimates of specificity in such a setting (15).
Handling of blood samples
After blood withdrawal in the gastrointestinal practices, serum
samples were incubated at room temperature for 30 to 60 minutes
to allow blood clotting and centrifuged at 2,000 to 2,500 g for
10 minutes. Then they were transported to the DKFZ laboratory in
a cold chain (medium transport time: 1 day), centrifuged again,
aliquoted, and stored at 80 C. For testing, the serum samples
were randomized and shipped on dry ice to Roche Diagnostics
GmbH. There was not more than one freeze–thaw cycle for each
sample. The laboratory staff was blind to any information regarding the study population.
Immunoassays
The five biomarkers CEA, ferritin, seprase, osteopontin, and
anti-p53 antibody were measured quantitatively on a cobas e601
platform. The assays for each marker are designed as sandwich
assays based on the streptavidin–biotin technology. The capture
antibodies are biotinylated and bind to streptavidin-coated
microparticles. The secondary antibodies, covalently linked to
Ruthenium complexes, are used for electrochemoluminescent
detection (16).
For CEA and ferritin, the commercial tests Elecsys CEA (Roche
Diagnostics GmbH, catalog number: 11731629 322) and Elecsys
Ferritin (Roche Diagnostics GmbH, catalog number: 04491785
190) were used according to the manufacturer's instructions.
Calibration was performed with the CEA CalSet (Roche Diagnostics GmbH, catalog number 11731645322) and the Ferritin
CalSet (Roche Diagnostics GmbH, catalog number 03737586
190) in accordance with the package inserts.
Reagents for the quantitative analysis of seprase, osteopontin,
and anti-p53 antibody for the cobas e platform were available as
prototypes at Roche Diagnostics. Performance characteristics for
prototype assays are expected to be similar to those seen for
Clinical Cancer Research
Downloaded from clincancerres.aacrjournals.org on August 12, 2017. © 2015 American Association for Cancer
Research.
Published OnlineFirst November 11, 2015; DOI: 10.1158/1078-0432.CCR-15-1268
A 5-Marker Blood Test for CRC Early Detection
commercial Elecsys assays, that is, repeatability CV 5%–6%,
intermediate precision/total imprecision CV 7%–8%, dilution linearity within 10%. For the prototype calibrators
M-Cal-Seprase, M-Cal-Osteopontin, and M-Cal-anti-p53, a full
calibration was performed.
gFOBTs and FITs
In the context of the BliTz study, different stool tests for the early
detection of colorectal cancer were evaluated. For all except two
participants included into this study, FIT results were available
and for most study participants (HemOccult, Beckman Coulter
GmbH) gFOBT results were also available. While participants
recruited before February 2009 were tested by the quantitative FIT
RIDASCREEN Hemoglobin (R-Biopharm AG) as described elsewhere (17), for participants recruited since February 2009 the
quantitative FIT FOB Gold (Sentinel Diagnostics) was used. Until
January 2012, BliTz participants collected native stool samples
that were immediately frozen and thawed once before conducting
the FIT according to the manufacturers' instructions in a central
laboratory. Since the end of January 2012, participants directly
used the buffer-filled stool collection tubes from Sentinel Diagnostics for sample collection and mailed them to the central
laboratory for analysis according to the manufacturer's instructions. For all three FIT conditions (frozen stool þ RIDASCREEN,
frozen stool þ FOB Gold, fresh stool in buffer-filled sentinel
tubes þ FOB Gold), we calculated cutoffs for test positivity based
on all available BliTz controls with this FIT condition. At 96%
specificity, cutoffs were 9.6 mg hemoglobin/g stool for the
RIDASCEEN test, 42.5 mg hemoglobin/g stool for the FOB Gold
test with frozen stool samples, and 15.3 mg hemoglobin/g stool
for the FOB Gold test with fresh stool samples collected in bufferfilled Sentinel tubes.
Data processing and statistical analysis
For processing of the data obtained from the immunoassays,
the evaluation software OASE was used. For further data analyses,
statistical software [R version 3.1.0 (18) and SAS version 9.2 (SAS
Institute)] was used.
Basic demographic characteristics in the study population
(sex, age, and UICC stage) were summarized. The results of the
five individual tests were combined into one single diagnostic
result (the "score") at the biostatistics department of Roche
Diagnostics using a defined algorithm with a predefined cutoff.
This algorithm was selected by penalized LASSO regression on
data of the MIP study (10) and reoptimized in a second large
panel with screening controls and enriched cases (CT study).
For the reoptimized algorithm, the MIP study marker CYFRA21 was dispensable. That is why this marker was not tested in
the BliTz study anymore. An algorithm for the marker combination CEA þ anti-p53 was derived from a logistic regression
model trained on data from the CT study. Sociodemographic
characteristics of this study collective can be found in Supplementary Table S1.
Univariate marker results and results for marker combinations
were compared between participants with colorectal cancer and
controls and between participants with advanced adenomas and
controls. Clinical performance [sensitivity and specificity with
exact 95% confidence intervals (CI)] and ROC curves for detection
of colorectal cancer and advanced adenoma were determined. In
addition to analyses in the whole study population, we performed
stage-specific analyses. AUCs were compared by the DeLong
www.aacrjournals.org
method with the R package "pROC" (19). For patients with gFOBT
and FIT results, we also evaluated agreement between the
5-marker blood test and the stool tests.
Results
Characterization of the study population
For validation of the 5-marker blood test, 1,660 participants
(36 patients with invasive colorectal cancer, 4 CIS patients, 420
advanced adenoma patients, and 1,200 participants free of
neoplasms) were selected from eligible participants of the
colorectal cancer screening study BliTz as described in the
STAndards for the Reporting of Diagnostic accuracy studies
(STARD) diagram (see Fig. 1). Sociodemographic characteristics of all participants with valid measurement results (n ¼
1,656) are summarized in Table 1. As expected, for a true
colorectal cancer screening setting, the average age among
colorectal cancer, CIS, and advanced adenoma cases is slightly
higher than among controls (mean SD: 66.0 6.2, 63.0 5.3, and 63.6 6.7 vs. 62.0 6.1 years). Also, the proportion of
men is higher among cases than among controls (colorectal
cancer, 72.2% men; CIS, 75.0% men; advanced adenomas,
64.7% men; controls, 45.5% men). Among the group of colorectal cancer patients, early-stage cancers (UICC stage I/II) were
equally common as late-stage cancers (UICC stage III/IV).
Diagnostic performance of the original marker panel
For blood samples from all except four subjects (one advanced
adenoma patient and three controls) valid measurements for all
five Elecsys Assays (CEA, ferritin, seprase, osteopontin, and antip53) could be obtained. We used a predefined algorithm obtained
in the MIP study and validated in the CT study to combine the
results from the five Elecsys Assays into a single prediction score.
ROC curve analysis revealed an AUC of 0.78 (95% CI, 0.68–0.87)
for the discrimination of colorectal cancer patients (stage I–IV)
and controls and an AUC of 0.56 (95% CI, 0.53–0.59) for the
discrimination of advanced adenoma patients and controls
(see Fig. 2). In a sensitivity analysis including 4 CIS patients, the
diagnostic performance for the detection of colorectal cancer
(stage 0–IV) was slightly worse with an AUC of 0.76 (95% CI,
0.67–0.85).
When using a cutoff at 90% specificity in the CT study, the
5-marker combination yielded sensitivities of 44 (95% CI,
28–62) and 12 (95% CI, 9–15) for colorectal cancer and
advanced adenomas at a specificity of 94 (95% CI, 92–95).
When adjusting the cutoffs to yield specificities of 80%, 90%,
and 95% in the BliTz collective, sensitivities for the detection of
colorectal cancer were 64%, 50%, and 42%. These values are
lower than the sensitivities observed in the MIP study (10) and
the CT study (see Supplementary Table S2). Sensitivities for
advanced adenomas were below 30% even at cutoffs yielding
80% specificity (see Table 2).
With AUCs of 0.80 (95% CI, 0.67–0.93) and 0.75 (95% CI,
0.62–0.88), the ability of the 5-marker blood test to detect earlyand late-stage cancer was similar (P, 0.57) (see Table 2).
Diagnostic performance of single markers
To further evaluate the loss of performance of the 5-marker
panel in the BliTz study collective, we compared the univariate
results for CEA, ferritin, seprase, osteopontin, and anti-p53 in the
BliTz and in the CT study. Interestingly, two of the five markers
Clin Cancer Res; 2016
Downloaded from clincancerres.aacrjournals.org on August 12, 2017. © 2015 American Association for Cancer
Research.
OF3
Published OnlineFirst November 11, 2015; DOI: 10.1158/1078-0432.CCR-15-1268
Werner et al.
Parcipants with signed consent, colonoscopy results, and quesonnaires parcipang unl February 20, 2013
(N = 5,781)
Exclusion of parcipants without adequate blood samples:
• No serum sample available (N = 332)
• Blood withdrawal aer colonoscopy (N = 38) or unknown me of blood withdrawal (N = 158)
Exclusion of parcipants who do not represent the target populaon of screening:
• History of CRC or inflammatory bowel disease (N = 41)
• Colonoscopy history in the last 5 years or unknown colonoscopy history (N = 282 + 70)
• Age <50 years, age ≥80 years, or age unknown (N = 89 + 58 + 1)
Exclusion of parcipants with potenally false negave results (only in parcipants free of
neoplasms):
•
Inadequate bowel preparaon before colonoscopy (N = 307)
•
Incomplete colonoscopy (N = 62)
Parcipants eligible for sample selecon (N = 4,567)
CRC and
CIS
(N = 42)
Not further
defined polyp
(N = 36)
Advanced
adenoma
(N = 483)
Non-advanced
adenoma
(N = 916)
Free of colorectal
neoplasms
(N = 3,090)
Exclusion of parcipants with
crical informaon missing at
me of sample selecon (N = 2)
Available CRC and CIS
paents (N = 40)
Representave sample of adenoma
paents1,2 (N = 420)
Representave sample of controls
free of neoplasms2 (N = 1,200)
Sample set for measurements with
the 5-marker blood test (N = 1,660)
No measurement
results (N = 4)
Samples with valid measurement results
(N = 1,656)
Figure 1.
1
STARD diagram of participants of the BliTz study (December 2005–February 2013). , 420 is the number of available advanced adenoma patients in BliTz
2
that was anticipated at the time of study design. , Preferentially, participants were selected from the subgroup with valid FIT results (FIT result available and stool
sampling before colonoscopy) which comprises >90% of all BliTz participants. CRC, colorectal cancer.
(CEA and anti-p53) showed very similar diagnostic performance
in both studies [AUC in BliTz, 0.84 (95% CI, 0.78–0.90) and 0.57
(95% CI, 0.51–0.63); AUC in CT study, 0.77 (95% CI, 0.73–0.81)
and 0.59 (95% CI, 0.56–0.62)] while the other markers did not
(see Fig. 3A–E). The largest decrease of diagnostic performance
was seen for seprase with an AUC of 0.78 (95% CI, 0.74–0.82) in
the CT study and an AUC of 0.60 (95% CI, 0.49–0.70) in the BliTz
study.
OF4 Clin Cancer Res; 2016
Diagnostic performance of a 2-marker combination
Because of the poor univariate results for seprase, osteopontin,
and ferritin in the BliTz study collective, we decided to perform an
exploratory evaluation of the marker combination CEA þ antip53. For training of the algorithm, data from the CT study were
used. In the BliTz study, the 2-marker combination reached an
AUC of 0.85 (95% CI, 0.78–0.91) for the detection of colorectal
cancer (see Fig. 3F) and an AUC of 0.56 (95% CI, 53–59) for the
Clinical Cancer Research
Downloaded from clincancerres.aacrjournals.org on August 12, 2017. © 2015 American Association for Cancer
Research.
Published OnlineFirst November 11, 2015; DOI: 10.1158/1078-0432.CCR-15-1268
A 5-Marker Blood Test for CRC Early Detection
Table 1. Study population characteristics
Group
Characteristics
Number of participants
Measurements available
Age [mean SD, years]
Sex
Male (n, %)
Female (n, %)
Stage
UICC 0
UICC I
UICC II
UICC III
UICC IV
Control
1,200
1,197
62.0 6.1
Advanced adenoma
420
419
63.6 6.7
CIS
4
4
63.0 5.3
CRC
36
36
66.0 6.2
545 (45.5%)
652 (54.5%)
271 (64.7%)
148 (35.3%)
3 (75.0%)
1 (25.0%)
26 (72.2%)
10 (27.8%)
—
—
—
—
—
—
—
—
—
—
4
—
—
—
—
—
13
5
16
2
Abbreviations: CRC, colorectal cancer; n, number.
detection of advanced adenomas. When adjusting the specificity
to 80%, 90%, and 95%, the sensitivities for colorectal cancer
detection were 67%, 58%, and 47% for the 2-marker combination. For the same specificities, the sensitivities of CEA alone were
62%, 50%, and 40%.
Comparison of the 5-marker blood test with stool tests
For 23 colorectal cancer cases, 301 advanced adenoma cases,
and 899 controls, results of the 5-marker blood test and both stool
tests (gFOBT and FIT) were available. Table 3 shows the diagnostic
performance of all tests in this subpopulation. For reasons of
comparison, the specificities of the FIT and the 5-marker blood
test were adjusted to the specificity of the gFOBT (96%). At this
specificity, both the 5-marker blood test and the gFOBT could
identify 39% of all colorectal cancer cases. Of note, the gFOBT and
the 5-marker blood test primarily detected different colorectal
cancer patients which led to an increased sensitivity of 65% (at
minimal loss of specificity) when combining gFOBT and blood
test results. For advanced adenomas, the sensitivities of the
5-marker blood test, the gFOBT, and the combination of both
tests were 8%, 4%, and 12%. With sensitivities of 78% for
colorectal cancer and 28% for advanced adenomas, the FIT was
superior to the other tests and its combination with the 5-marker
panel could only slightly further increase sensitivities (83% for
colorectal cancer, 35% for advanced adenomas) at the cost of a
lower specificity (92%).
Discussion
In this study, we evaluated a 5-marker blood test for colorectal
cancer early detection in a large screening study. The AUCs for the
detection of colorectal cancer and advanced adenomas were
0.78 (95% CI, 0.68–0.87) and 0.56 (95% CI, 0.53–0.59), respectively. At specificities of 80%, 90%, and 95%, the sensitivities for
the detection of colorectal cancer reached 64%, 50%, and 42%,
respectively. Early-stage cancers were detected at least as well as
late-stage cancers. Of note, two of the five markers (CEA and antip53) were sufficient to achieve a similar or even better diagnostic
Figure 2.
Diagnostic performance of the 5-marker blood test for the detection of colorectal cancer (A; CRC) and advanced adenomas (B).
www.aacrjournals.org
Clin Cancer Res; 2016
Downloaded from clincancerres.aacrjournals.org on August 12, 2017. © 2015 American Association for Cancer
Research.
OF5
Published OnlineFirst November 11, 2015; DOI: 10.1158/1078-0432.CCR-15-1268
Werner et al.
Table 2. Sensitivities and specificities of the 5-marker blood test at different cutoffs
Cutoff obtained
Cutoff adjusted
Group
n
in CT study
at 95% specificity
Sensitivity of the 5-marker blood test in % (95% CI)
Colorectal cancer
36
44 (28–62)
42 (26–59)
UICC I–II
18
50 (26–74)
44 (22–69)
UICC III–IV
18
39 (17–64)
39 (17–64)
CIS
4
0 (0–60)
0 (0–60)
Advanced adenoma
419
12 (9–15)
9 (6–12)
Specificity of the 5-marker blood test in % (95% CI)
Controls
1,197
94 (92–95)
95 (94–96)
performance for colorectal cancer (AUC, 0.85; 95% CI, 0.78–
0.91). In a subsample of participants with gFOBT, FIT, and blood
test results, we directly compared the performance of these tests.
With a sensitivity of 39% at a specificity of 96%, gFOBT and the
5-marker blood test performed equally well and sensitivity
increased to 65% when both tests were combined. Nevertheless,
both tests and their combination were outperformed by FIT.
In our previous MIP study, a 6-marker blood test (5-marker
panel with the addition of CYFRA21–1) was able to detect
colorectal cancer patients with a sensitivity of 69.6% at a specificity of 95%, which was similar to the diagnostic performance
of FIT in the MIP study (10). In the CT study that was used to
refine the algorithm for the 5-marker blood test, the observed
sensitivity of the 5-marker blood test was 68.1% at 95% specificity. The obvious drop of diagnostic performance of the
5-marker blood test in participants selected from the BliTz study
collective may be explained by differences in the study populations. While BliTz is a real screening study in which all cases were
recruited before diagnosis at screening colonoscopy, colorectal
cancer cases in the MIP and CT study had to be enriched with
patients recruited at surgery units. For clinically recruited colorectal cancer cases, there is the possibility that blood marker
levels are altered by diagnostic or therapeutic interventions or
lifestyle changes in response to the colorectal cancer diagnosis.
Furthermore, in contrast to screening settings, in clinical settings
it cannot be guaranteed that blood withdrawal and blood
storage conditions at recruitment site are exactly the same for
cases and controls. Last but not the least, colorectal cancer
patients recruited in clinical settings are often in a more
advanced stage and more often present symptoms than colorectal cancer patients diagnosed at screening colonoscopy (20).
Our univariate analyses suggest that not all markers of the
5-marker blood test are as prone to study setting issues as others.
With an AUC of 0.84, the diagnostic performance of CEA was even
better than in the CT study (AUC, 0.77) and for anti-p53, the
AUCs in both studies were quite similar (AUC in BliTz, 0.57; AUC
in CT study, 0.59). These results suggest that the observed
diagnostic performance for CEA and anti-p53 indeed represents
true cancer-specific differences between cases and controls. For
ferritin, osteopontin, and seprase, other factors might have
contributed to the previously observed good diagnostic performance for colorectal cancer detection. Serum ferritin levels, for
instance, can be influenced by age, sex, body mass index, acute
or chronic inflammation, and aspirin use in addition to cancer
(21). For colorectal cancer, the situation is especially complex,
because possible positive correlations between cancer-specific
processes and ferritin, as found for other cancers (21), might be
antagonized by iron deficiency anemia caused by chronic
gastrointestinal bleedings.
OF6 Clin Cancer Res; 2016
Cutoff adjusted
at 90% specificity
Cutoff adjusted
at 80% specificity
50 (33–67)
50 (26–74)
50 (26–74)
25 (1–81)
16 (12–19)
64 (46–79)
72 (47–90)
56 (31–78)
25 (1–81)
25 (21–30)
90 (88–92)
80 (78–82)
The sensitivity of anti-p53 is limited due to the fact that not
all colorectal cancer patients have p53 mutations and not all
patients with p53 mutations produce antibodies against this
tumor suppressor protein (22). Nevertheless, the remarkably
high specificity of anti-p53 for cancer, which can also be seen in
the very steep slope of the left part of its ROC curve (see Fig.
3B), makes it possible to increase the sensitivity of conventional tumor markers without reducing specificity to a relevant
extent when anti-p53 is added in a marker combination (23).
The combined use of CEA and anti-p53 for colorectal cancer
detection has been evaluated in earlier studies, and sensitivities
between 33% and 73% have been reported (22–26). However,
none of these studies was performed in a screening setting
and only Kojima and colleagues reported specificities for this
marker combination. In their newer and larger study from
2011, a sensitivity of 48% at 93% specificity for the detection
of colorectal cancer was reached (25), which is very similar to
our findings in participants of screening colonoscopy. It should
be stated that both CEA and anti-p53 are not cancer type–
specific and have been found in patients with other cancers like
lung cancer (27, 28). Thus, there is a possibility that some of the
controls with false-positive test results actually are persons with
an undiscovered other malignancy.
With 47% and 6% sensitivity at 95% specificity, the capability
of the marker combination CEA þ anti-p53 to detect colorectal
cancer and advanced adenomas was comparable or even superior
to Epi proColon, which, to our knowledge, is the only prospectively evaluated blood test for early detection of colorectal cancer
so far (29). Although the 2-marker combination CEA þ anti-p53
alone cannot compete with the FIT, it appears plausible that
a combination with further blood biomarkers, such as DNA
methylation markers (30), miRNA markers (31), autoantibody
markers (24), or protein markers (32, 33) might increase the
diagnostic performance sufficiently for an application in mass
screening.
Blood tests seem to be better accepted in public than stool
tests. For instance, in a study by Adler and colleagues, over 100
persons that refused to participate in screening colonoscopy
were offered a choice of a blood-based or a stool-based colorectal cancer early detection test and while 83% of the participants picked the blood test, only 15% picked the stool test
(34). One major advantage of FIT over current blood tests, in
addition to the better diagnostic performance for detection of
colorectal cancer, is its ability to detect a relevant proportion of
advanced adenomas. Stool tests might have a larger potential to
capture localized tumor effects in general such as excretion of
blood or components of tumor cells with stool which would be
hard, if not impossible, to detect by blood tests, in particular, at
early tumor stage and for precursors of colorectal cancer. In our
Clinical Cancer Research
Downloaded from clincancerres.aacrjournals.org on August 12, 2017. © 2015 American Association for Cancer
Research.
Published OnlineFirst November 11, 2015; DOI: 10.1158/1078-0432.CCR-15-1268
A 5-Marker Blood Test for CRC Early Detection
Figure 3.
A–E, Univariate analysis: Performance
of CEA, anti-p53, osteopontin, ferritin
and seprase in the BliTz and in the CT
study. F, Diagnostic performance of
the 2-marker combination anti-p53 þ
CEA in the BliTz study.
analyses, the diagnostic performance for the detection of
advanced adenomas was poor and the diagnostic performance
for the detection of non-advanced adenomas is expected to be
even worse. Non-advanced adenomas were deliberately not
included in our analyses as their transition rates to more
www.aacrjournals.org
advanced neoplasms are low (35) and there is an ongoing
debate whether they should be considered as one of the target
lesions for colorectal cancer screening or not (36). For the
development of future blood tests for early detection of colorectal cancer, it would be beneficial to identify markers that also
Clin Cancer Res; 2016
Downloaded from clincancerres.aacrjournals.org on August 12, 2017. © 2015 American Association for Cancer
Research.
OF7
Published OnlineFirst November 11, 2015; DOI: 10.1158/1078-0432.CCR-15-1268
Werner et al.
Table 3. Comparison of 5-marker blood test, gFOBT, and FIT in participants with stool test results
5-Marker
5-Marker blood
blood test
gFOBT
test þ gFOBTa
Sensitivity CRCb in % (95% CI)
39 (20–61)
39 (20–61)
65 (43–84)
8 (5–12)
4 (2–7)
12 (9–16)
Sensitivity advanced adenomac in % (95% CI)
96 (94–97)
96 (94–97)
92 (90–93)
Specificityd in % (95% CI)
FIT
78 (56–93)
28 (23–34)
96 (94–97)
5-Marker blood
test þ FITa
83 (61–95)
35 (30–41)
92 (90–94)
The combination 5-marker blood test þ gFOBT/FIT was considered positive if either the 5-marker blood test, the gFOBT/FIT, or both tests were positive.
n ¼ 23.
c
n ¼ 301.
d
n ¼ 899.
a
b
detect advanced adenomas. In the meantime, efforts to increase
public's adherence for stool test should be enhanced.
To our knowledge, our study is the first to test a 5-marker blood
test, including CEA and anti-p53, in subjects from a true screening
setting. There are specific strengths and limitations that have to be
considered. One strength is that cases and controls were selected
from participants of screening colonoscopy that represent the
target population for colorectal cancer screening. With over 1,600
study participants, including 1,200 controls and 400 advanced
adenoma patients, our sample size was very large. So estimates for
specificity and sensitivity for the detection of advanced adenomas
could be determined very precisely and 95% CIs are small.
Furthermore, we used a predefined algorithm trained on data of
our previous studies to avoid overfitting, a serious problem seen
in many multimarker studies (37). A limitation is the relatively
small number of colorectal cancer cases included in this study
which is due to the low prevalence of colorectal cancer in participants of screening colonoscopy. This limited our options to
perform subgroup-specific analyses. In addition, we did not
evaluate the value of the 5-marker blood test or individual
markers for prognosis or monitoring colorectal cancer patients.
In conclusion, the validation of a 5-marker blood test for
colorectal cancer early detection in participants selected from a
screening study collective uncovered decreased diagnostic performance for the markers ferritin, osteopontin, and seprase, compared with previous evaluations in studies conducted among
cases recruited in clinical settings. Thus, the overall diagnostic
performance estimates for the 5-marker blood test dropped from
values comparable with FIT in the clinical setting to values
comparable with gFOBT in our study. Our results furthermore
underline the potential of CEA and anti-p53 to discriminate
cancer patients and controls under screening conditions suggesting their potential to contribute to the development of a multimarker blood-based test for early detection of colorectal cancer.
Disclosure of Potential Conflicts of Interest
S. Werner reports receiving commercial research grants from Roche
Diagnostics. H. Brenner reports receiving commercial research grants from
Roche Diagnostics, and other commercial research support from Applied
Proteomics. No potential conflicts of interest were disclosed by the other
authors.
Authors' Contributions
Conception and design: S. Werner, M. Strobl, D. Morgenstern, H. Brenner
Development of methodology: D. Morgenstern
Acquisition of data (provided animals, acquired and managed patients,
provided facilities, etc.): S. Werner, M. Strobl, C. Datz, H. Brenner
Analysis and interpretation of data (e.g., statistical analysis, biostatistics,
computational analysis): S. Werner, F. Krause, V. Rolny, H. Brenner
Writing, review, and/or revision of the manuscript: S. Werner, F. Krause,
V. Rolny, M. Strobl, D. Morgenstern, C. Datz, H. Chen, H. Brenner
Study supervision: M. Strobl, H. Brenner
Acknowledgments
The authors acknowledge the excellent cooperation of gastroenterology
practices in patient recruitment and of Labor Limbach in sample collection.
The authors also thank Dr. Katja Butterbach and Ulrike Schlesselmann for their
excellent work in laboratory preparation of blood samples and Isabel Lerch,
Susanne K€
ohler, Utz Benscheid, Jason Hochhaus, and Maria Kuschel for their
contribution in data collection, monitoring, and documentation.
Grant Support
This study was financed by Roche Diagnostics GmbH, Penzberg, Germany.
The costs of publication of this article were defrayed in part by the payment of
page charges. This article must therefore be hereby marked advertisement in
accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Received May 29, 2015; revised October 28, 2015; accepted October 31, 2015;
published OnlineFirst November 11, 2015.
References
1. Ferlay J, Soerjomataram I, Dikshit R, Eser S, Mathers C, Rebelo
M, et al. Cancer incidence and mortality worldwide: sources,
methods and major patterns in GLOBOCAN 2012. Int J Cancer
2015;136:E359–86.
2. Leslie A, Carey FA, Pratt NR, Steele RJ. The colorectal adenoma-carcinoma
sequence. Br J Surg 2002;89:845–60.
3. Brenner H, Altenhofen L, Stock C, Hoffmeister M. Natural history of
colorectal adenomas: birth cohort analysis among 3.6 million participants
of screening colonoscopy. Cancer Epidemiol Biomarkers Prev 2013;
22:1043–51.
4. Brenner H, Stock C, Hoffmeister M. Effect of screening sigmoidoscopy and
screening colonoscopy on colorectal cancer incidence and mortality: systematic review and meta-analysis of randomised controlled trials and
observational studies. BMJ 2014;348:g2467.
OF8 Clin Cancer Res; 2016
5. Hassan C, Giorgi Rossi P, Camilloni L, Rex DK, Jimenez-Cendales B, Ferroni
E, et al. Meta-analysis: adherence to colorectal cancer screening and the
detection rate for advanced neoplasia, according to the type of screening
test. Aliment Pharmacol Ther 2012;36:929–40.
6. Lansdorp-Vogelaar I, Knudsen AB, Brenner H. Cost-effectiveness of colorectal cancer screening. Epidemiol Rev 2011;33:88–100.
7. Carroll MR, Seaman HE, Halloran SP. Tests and investigations for colorectal cancer screening. Clin Biochem 2014;47:921–39.
8. Osborne J, Wilson C, Moore V, Gregory T, Flight I, Young G. Sample
preference for colorectal cancer screening tests: blood or stool? Open J Prev
Med 2012;2:326–31.
9. Fung KY, Nice E, Priebe I, Belobrajdic D, Phatak A, Purins L, et al. Colorectal
cancer biomarkers: to be or not to be? Cautionary tales from a road well
travelled. World J Gastroenterol 2014;20:888–98.
Clinical Cancer Research
Downloaded from clincancerres.aacrjournals.org on August 12, 2017. © 2015 American Association for Cancer
Research.
Published OnlineFirst November 11, 2015; DOI: 10.1158/1078-0432.CCR-15-1268
A 5-Marker Blood Test for CRC Early Detection
10. Wild N, Andres H, Rollinger W, Krause F, Dilba P, Tacke M, et al. A
combination of serum markers for the early detection of colorectal cancer.
Clin Cancer Res 2010;16:6111–21.
11. Karl J, Wild N, Tacke M, Andres H, Garczarek U, Rollinger W, 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.
12. Hundt S, Haug U, Brenner H. Comparative evaluation of immunochemical
fecal occult blood tests for colorectal adenoma detection. Ann Intern Med
2009;150:162–9.
13. Haug U, Hundt S, Brenner H. Quantitative immunochemical fecal occult
blood testing for colorectal adenoma detection: evaluation in the target
population of screening and comparison with qualitative tests. Am J
Gastroenterol 2010;105:682–90.
14. Brenner H, Tao S, Haug U. Low-dose aspirin use and performance of
immunochemical fecal occult blood tests. JAMA 2010;304:2513–20.
15. Brenner H, Altenhofen L, Tao S. Matching of controls may lead to biased
estimates of specificity in the evaluation of cancer screening tests. J Clin
Epidemiol 2013;66:202–8.
16. Roche Diagnostics Ltd. Elecsys Technology brochure (2011) [accessed
2014 Oct 24 ]. Available from: http://www.cobas.com/home/product/
clinical-and-immunochemistry-testing/technology-elecsys-ecl.html.
17. Brenner H, Haug U, Hundt S. Sex differences in performance of fecal occult
blood testing. Am J Gastroenterol 2010;105:2457–64.
18. R Core Team (2014). R: A language and environment for statistical
computing. R Foundation for Statistical Computing, Vienna, Austria.
Available from: http://www.R-project.org/.
19. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under
two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44:837–45.
20. Tao S, Hundt S, Haug U, Brenner H. Sensitivity estimates of blood-based
tests for colorectal cancer detection: impact of overrepresentation of
advanced stage disease. Am J Gastroenterol 2011;106:242–53.
21. Alkhateeb AA, Connor JR. The significance of ferritin in cancer: antioxidation, inflammation and tumorigenesis. Biochim Biophys Acta
2013;1836:245–54.
22. Suppiah A, Greenman J. Clinical utility of anti-p53 auto-antibody: systematic review and focus on colorectal cancer. World J Gastroenterol
2013;19:4651–70.
23. Muller M, Meyer M, Schilling T, Ulsperger E, Lehnert T, Zentgraf H, et al.
Testing for anti-p53 antibodies increases the diagnostic sensitivity of
conventional tumor markers. Int J Oncol 2006;29:973–80.
24. Chen H, Werner S, Tao S, Zornig I, Brenner H. Blood autoantibodies against
tumor-associated antigens as biomarkers in early detection of colorectal
cancer. Cancer Lett 2014;346:178–87.
www.aacrjournals.org
25. Kojima T, Matsui T, Fujimitsu Y, Kojima H, Uno H, Hiramatsu K, et al.
Titration of serum CEA, p53 antibodies and CEA-IgM complexes in 142
patients with colorectal cancer and 150 healthy blood donors. Ann Cancer
Res Ther 2011;19:15–19.
26. Kojima T, Yoshikawa K, Matsui T, Kodera Y, Kojima H. Titration of serum
CEA, p53 antibodies and CEA-IgM complexes in patients with colorectal
cancer. Mol Med Rep 2009;2:477–80.
27. Grunnet M, Sorensen JB. Carcinoembryonic antigen (CEA) as tumor
marker in lung cancer. Lung Cancer 2012;76:138–43.
28. Lei QQ, Liu JW, Zheng H. Potential role of anti-p53 antibody in diagnosis
of lung cancer: evidence from a bivariate meta-analysis. Eur Rev Med
Pharmacol Sci 2013;17:3012–8.
29. Church TR, Wandell M, Lofton-Day C, Mongin SJ, Burger M,
Payne SR, et al. Prospective evaluation of methylated SEPT9 in
plasma for detection of asymptomatic colorectal cancer. Gut 2014;
63:317–25.
30. Summers T, Langan RC, Nissan A, Brucher BL, Bilchik AJ, Protic M,
et al. Serum-based DNA methylation biomarkers in colorectal
cancer: potential for screening and early detection. J Cancer 2013;4:
210–6.
31. Luo X, Burwinkel B, Tao S, Brenner H. MicroRNA signatures: novel
biomarker for colorectal cancer? Cancer Epidemiol Biomarkers Prev
2011;20:1272–86.
32. Alvarez-Chaver P, Otero-Estevez O, Paez de la Cadena M, RodriguezBerrocal FJ, Martinez-Zorzano VS. Proteomics for discovery of candidate colorectal cancer biomarkers. World J Gastroenterol 2014;20:
3804–24.
33. Chen H, Zucknick M, Werner S, Knebel P, Brenner H. Head-to-head
comparison and evaluation of 92 plasma protein biomarkers for early
detection of colorectal cancer in a true screening setting. Clin Cancer Res
2015;21:3318–26.
34. Adler A, Geiger S, Keil A, Bias H, Schatz P, deVos T, et al. Improving
compliance to colorectal cancer screening using blood and stool based tests
in patients refusing screening colonoscopy in Germany. BMC Gastroenterol 2014;14:183.
35. Brenner H, Altenhofen L, Stock C, Hoffmeister M. Expected long-term
impact of the German screening colonoscopy programme on colorectal
cancer prevention: analyses based on 4,407,971 screening colonoscopies.
Eur J Cancer 2015;51:1346–53.
36. Garcia M, Mila N, Binefa G, Borras JM, Espinas JA, Moreno V. False-positive
results from colorectal cancer screening in Catalonia (Spain), 2000–2010.
J Med Screen 2012;19:77–82.
37. Vickers AJ, Jang K, Sargent D, Lilja H, Kattan MW. Systematic review of
statistical methods used in molecular marker studies in cancer. Cancer
2008;112:1862–8.
Clin Cancer Res; 2016
Downloaded from clincancerres.aacrjournals.org on August 12, 2017. © 2015 American Association for Cancer
Research.
OF9
Published OnlineFirst November 11, 2015; DOI: 10.1158/1078-0432.CCR-15-1268
Evaluation of a 5-Marker Blood Test for Colorectal Cancer
Early Detection in a Colorectal Cancer Screening Setting
Simone Werner, Friedemann Krause, Vinzent Rolny, et al.
Clin Cancer Res Published OnlineFirst November 11, 2015.
Updated version
Supplementary
Material
E-mail alerts
Reprints and
Subscriptions
Permissions
Access the most recent version of this article at:
doi:10.1158/1078-0432.CCR-15-1268
Access the most recent supplemental material at:
http://clincancerres.aacrjournals.org/content/suppl/2015/11/11/1078-0432.CCR-15-1268.DC1
Sign up to receive free email-alerts related to this article or journal.
To order reprints of this article or to subscribe to the journal, contact the AACR Publications
Department at [email protected].
To request permission to re-use all or part of this article, contact the AACR Publications
Department at [email protected].
Downloaded from clincancerres.aacrjournals.org on August 12, 2017. © 2015 American Association for Cancer
Research.