J Cancer Res Clin Oncol (2013) 139:681689

DOI 10.1007/s00432-012-1371-3

ORIGINAL PAPER

Prognostic value of CA 19-9, CEA, CRP, LDH and bilirubin levels

in locally advanced and metastatic pancreatic cancer: results
from a multicenter, pooled analysis of patients receiving palliative
chemotherapy
Michael Haas Volker Heinemann Frank Kullmann Rudiger P. Laubender
Christina Klose Christiane J. Bruns Stefan Holdenrieder Dominik P. Modest
Christoph Schulz Stefan Boeck

Received: 24 November 2012 / Accepted: 29 December 2012 / Published online: 13 January 2013
Springer-Verlag Berlin Heidelberg 2013

Abstract
Purpose CA 19-9 is the only established tumor marker in
pancreatic cancer (PC); the prognostic role of other serum
markers like CEA, CRP, LDH or bilirubin has not yet been
defined.
Methods We pooled pre-treatment data on CA 19-9,
CEA, CRP, LDH and bilirubin levels from two German
multicenter randomized phase II trials together with prospective patient data from one high-volume German Cancer Center. Marker levels were assessed locally before the
start of palliative first-line therapy for advanced PC and
serially during treatment (for CA 19-9 only). Clinical and
M. Haas
V. Heinemann
D. P. Modest
C. Schulz

S. Boeck (&)
Department of Internal Medicine III and Comprehensive Cancer
Center, Klinikum Grosshadern, Ludwig-Maximilians-University
of Munich, Marchioninistr. 15, 81377 Munich, Germany
e-mail: stefan.boeck@med.uni-muenchen.de
F. Kullmann
Department of Medicine I, Gastroenterology/Oncology,
Klinikum Weiden, Weiden, Germany
R. P. Laubender
Institute of Medical Informatics, Biometry and Epidemiology,
Ludwig-Maximilians-University of Munich, Munich, Germany
C. Klose
Institute for Medical Biometry and Informatics,
University of Heidelberg, Heidelberg, Germany
C. J. Bruns
Department of Surgery and Comprehensive Cancer Center,
Klinikum Grosshadern, Ludwig-Maximilians-University of
Munich, Munich, Germany
S. Holdenrieder
Institute of Clinical Chemistry and Clinical Pharmacology,
University Hospital Bonn, Bonn, Germany

biomarker data (overall 12 variables) were correlated with

the efficacy endpoints time-to-progression (TTP) and
overall survival (OS) by using uni- and multivariate Cox
models.
Results Data from 291 patients were included in this
pooled analysis; 253 patients (87 %) received treatment
within prospective clinical trials. Median TTP in the study
cohort was 5.1 months and median OS 9.0 months. In
univariate analysis, pre-treatment CA 19-9 (HR 1.55),
LDH (HR 2.04) and CEA (HR 1.89) levels were significantly associated with TTP. Regarding OS, baseline CA
19-9 (HR 1.46), LDH (HR 2.07), CRP (HR 1.69) and bilirubin (HR 1.62) were significant prognostic factors.
Within multivariate analyses, pre-treatment log [CA 19-9]
(as continuous variable for TTP) and log [bilirubin] as well
as log [CRP] (for OS) had an independent prognostic value.
A CA 19-9 decline of C25 % during the first two chemotherapy cycles was predictive for TTP and OS, independent
of the applied CA 19-9 assay.
Conclusion Baseline CA 19-9 and CA 19-9 kinetics
during first-line chemotherapy are prognostic in advanced
PC. Besides that finding other serum markers like CRP,
LDH and bilirubin can also provide prognostic information
on TTP and OS.
Keywords
factor

Gemcitabine
Pancreatic cancer
Prognostic

Introduction
The diagnosis of pancreatic cancer (PC) is still associated
with a dismal prognosis: only 1015 % of patients initially
present with resectable disease and can undergo curativeintent surgery. However, with a 5-year survival rate of only

123

682

5 %, most of these patients suffer from relapse and are

treated with systemic chemotherapy or chemoradiotherapy
in the course of their disease (Vincent et al. 2011). Since
1997, gemcitabine has been the standard chemotherapy
agent for advanced PC (Heinemann et al. 2012). The
addition of the tyrosine kinase inhibitor erlotinib has shown
a clinically moderate but statistically significant prolongation of survival for metastatic disease (Moore et al.
2007). The combination of 5-fluorouracil, folinic acid,
oxaliplatin and irinotecan (FOLFIRINOX) has recently
been described to almost double the survival time, but this
rather intensive combination scheme can only be applied to
relatively young and medically fit patients without relevant
co-morbidities and normal bilirubin values, thus appealing
to only a minority of patients (Conroy et al. 2011). To date,
however, there is no assured biomarker which might predict the efficacy of one of these specific treatment
regimens.
Several serum tumor markers have been investigated
in PC with CA 19-9 being the one with the highest
evidence to date (Boeck et al. 2006; Duffy et al. 2010).
CA 19-9 is also the most commonly used PC tumor
marker in clinical routine, as many studies in patients
with resectable and advanced disease proved CA 19-9 to
be a useful tool for the evaluation of treatment response
and for the prediction of prognosis (Hess et al. 2008;
Reni et al. 2009; Boeck et al. 2010; Humphris et al.
2012). Two important research issues, however, still
remain unclear: first, if a CA 19-9 decline during chemotherapy is predictive for efficacy endpoints (and specifically which cutoff to use for a definition of a
clinically relevant CA 19-9 decline) and if the application
of unique CA 19-9 assay is required to generate valid
and comparable data (Boeck et al. 2006; Duffy et al.
2010). Early evidence suggests that also other serum
markers like CEA and LDH levels might have a prognostic relevance in patients with advanced PC receiving
palliative treatment (Stocken et al. 2008; van Cutsem
et al. 2009; Duffy et al. 2010). Furthermore, there is
increasing evidence that an acute phase response plays an
important role in the pathobiology and prognosis of PC
and that the serum C-reactive protein (CRP) concentration thus might also serve as a potential biomarker in this
disease (van Cutsem et al. 2009; Moses et al. 2009; Pine
et al. 2009).
The primary aim of this large retrospective pooled
analysis was to define the prognostic role of several clinical
and pre-treatment laboratory factors (like CA 19-9, CEA,
CRP, LDH and bilirubin) with regard to time-to-progression (TTP) and overall survival (OS) in a homogeneous
patient population with advanced PC receiving first-line
palliative chemotherapy (mainly within a prospective

123

J Cancer Res Clin Oncol (2013) 139:681689

clinical trial). We furthermore investigated the impact of

CA 19-9 kinetics during treatment on outcome, separately
analyzed for the applied CA 19-9 assay (standardized CA
19-9 Elecsys assay vs. any assay).

Patients and methods

Patient population and treatment
Patients included in this pooled analysis were recruited
from two previously published German randomized phase
II trials (Boeck et al. 2008; Kullmann et al. 2009), and
additionally, also, patients with advanced PC treated
between July 2002 and June 2008 at the outpatient clinic of
the Pancreas Center at the Department of Internal
Medicine III, Ludwig-Maximilians-University of Munich
(LMU) were considered eligible. All patients had histologically or cytologically proven exocrine PC and were
treated with palliative chemotherapy (or chemoradiotherapy) for locally advanced or metastatic disease. Previous
treatment with chemo- or chemoradiotherapy was not
allowed except adjuvant treatment after curative-intent
surgical tumor resection. Patients with endocrine pancreatic tumors or secondary malignancies were excluded. As
reported previously, patients treated within the Ro96 trial
received gemcitabine/capecitabine, gemcitabine/oxaliplatin
or capecitabine/oxaliplatin, respectively (Boeck et al.
2008). Patients included into the GEMOXCET study were
treated with either gemcitabine/oxaliplatin or gemcitabine/
oxaliplatin plus cetuximab (Kullmann et al. 2009). Patients
treated outside clinical trials at the Pancreas Center at
the Department of Internal Medicine III receivedbased
on the decision of the treating medical oncologist SB and
VHstandard gemcitabine or gemcitabine-based chemotherapy until disease progression, unacceptable toxicity or
patient refusal.
Measurement of laboratory parameters
The baseline concentrations of CA 19-9, CEA, LDH,
hemoglobin and bilirubin were measured routinely in each
included patient before treatment initiation. As described
previously, a subgroup of patients whose CA 19-9 values
had been measured by a unique method (Elecsys, Roche
Diagnostics, Germany) was generated based on patients
treated at the LMU and within the GEMOXCET trial
(Kullmann et al. 2009; Boeck et al. 2010). Data on pretreatment CRP levels were only available from LMU
patients. Serial measurements were conducted for CA 19-9
every three to 4 weeks during treatment. All serum markers
were measured prospectively in each participating patient.

682

5 %, most of these patients suffer from relapse and are

treated with systemic chemotherapy or chemoradiotherapy
in the course of their disease (Vincent et al. 2011). Since
1997, gemcitabine has been the standard chemotherapy
agent for advanced PC (Heinemann et al. 2012). The
addition of the tyrosine kinase inhibitor erlotinib has shown
a clinically moderate but statistically significant prolongation of survival for metastatic disease (Moore et al.
2007). The combination of 5-fluorouracil, folinic acid,
oxaliplatin and irinotecan (FOLFIRINOX) has recently
been described to almost double the survival time, but this
rather intensive combination scheme can only be applied to
relatively young and medically fit patients without relevant
co-morbidities and normal bilirubin values, thus appealing
to only a minority of patients (Conroy et al. 2011). To date,
however, there is no assured biomarker which might predict the efficacy of one of these specific treatment
regimens.
Several serum tumor markers have been investigated
in PC with CA 19-9 being the one with the highest
evidence to date (Boeck et al. 2006; Duffy et al. 2010).
CA 19-9 is also the most commonly used PC tumor
marker in clinical routine, as many studies in patients
with resectable and advanced disease proved CA 19-9 to
be a useful tool for the evaluation of treatment response
and for the prediction of prognosis (Hess et al. 2008;
Reni et al. 2009; Boeck et al. 2010; Humphris et al.
2012). Two important research issues, however, still
remain unclear: first, if a CA 19-9 decline during chemotherapy is predictive for efficacy endpoints (and specifically which cutoff to use for a definition of a
clinically relevant CA 19-9 decline) and if the application
of unique CA 19-9 assay is required to generate valid
and comparable data (Boeck et al. 2006; Duffy et al.
2010). Early evidence suggests that also other serum
markers like CEA and LDH levels might have a prognostic relevance in patients with advanced PC receiving
palliative treatment (Stocken et al. 2008; van Cutsem
et al. 2009; Duffy et al. 2010). Furthermore, there is
increasing evidence that an acute phase response plays an
important role in the pathobiology and prognosis of PC
and that the serum C-reactive protein (CRP) concentration thus might also serve as a potential biomarker in this
disease (van Cutsem et al. 2009; Moses et al. 2009; Pine
et al. 2009).
The primary aim of this large retrospective pooled
analysis was to define the prognostic role of several clinical
and pre-treatment laboratory factors (like CA 19-9, CEA,
CRP, LDH and bilirubin) with regard to time-to-progression (TTP) and overall survival (OS) in a homogeneous
patient population with advanced PC receiving first-line
palliative chemotherapy (mainly within a prospective

123

J Cancer Res Clin Oncol (2013) 139:681689

clinical trial). We furthermore investigated the impact of

CA 19-9 kinetics during treatment on outcome, separately
analyzed for the applied CA 19-9 assay (standardized CA
19-9 Elecsys assay vs. any assay).

Patients and methods

Patient population and treatment
Patients included in this pooled analysis were recruited
from two previously published German randomized phase
II trials (Boeck et al. 2008; Kullmann et al. 2009), and
additionally, also, patients with advanced PC treated
between July 2002 and June 2008 at the outpatient clinic of
the Pancreas Center at the Department of Internal
Medicine III, Ludwig-Maximilians-University of Munich
(LMU) were considered eligible. All patients had histologically or cytologically proven exocrine PC and were
treated with palliative chemotherapy (or chemoradiotherapy) for locally advanced or metastatic disease. Previous
treatment with chemo- or chemoradiotherapy was not
allowed except adjuvant treatment after curative-intent
surgical tumor resection. Patients with endocrine pancreatic tumors or secondary malignancies were excluded. As
reported previously, patients treated within the Ro96 trial
received gemcitabine/capecitabine, gemcitabine/oxaliplatin
or capecitabine/oxaliplatin, respectively (Boeck et al.
2008). Patients included into the GEMOXCET study were
treated with either gemcitabine/oxaliplatin or gemcitabine/
oxaliplatin plus cetuximab (Kullmann et al. 2009). Patients
treated outside clinical trials at the Pancreas Center at
the Department of Internal Medicine III receivedbased
on the decision of the treating medical oncologist SB and
VHstandard gemcitabine or gemcitabine-based chemotherapy until disease progression, unacceptable toxicity or
patient refusal.
Measurement of laboratory parameters
The baseline concentrations of CA 19-9, CEA, LDH,
hemoglobin and bilirubin were measured routinely in each
included patient before treatment initiation. As described
previously, a subgroup of patients whose CA 19-9 values
had been measured by a unique method (Elecsys, Roche
Diagnostics, Germany) was generated based on patients
treated at the LMU and within the GEMOXCET trial
(Kullmann et al. 2009; Boeck et al. 2010). Data on pretreatment CRP levels were only available from LMU
patients. Serial measurements were conducted for CA 19-9
every three to 4 weeks during treatment. All serum markers
were measured prospectively in each participating patient.

J Cancer Res Clin Oncol (2013) 139:681689

Study design and statistical analyses

This tumor marker study was designed, conducted and
analyzed according to the 2005 REMARK guidelines
(REporting recommendations for tumor MARKer prognostic studies) as appropriate (McShane et al. 2005). The
pre-defined primary endpoint for the current retrospective
pooled analysis was to define the prognostic value of 12
clinical and laboratory parameters with regard to TTP and
OS in patients with advanced PC undergoing palliative
chemotherapy (or chemoradiotherapy). TTP was defined as
the time interval between the initiation of first-line therapy
until disease progression ascertained by imaging; OS was
calculated from the time point of initiation of first-line
therapy until death from any cause. The observations were
censored if disease progression was apparent through
clinical deterioration without imaging for TTP, and for OS
for patients being alive at a pre-defined time point (October
15, 2008). The applied cutoffs (in order to generate
dichotomized variables) for the assessed laboratory values
were derived either from previously published reports,
from the reference level of the validated assay or based on
the median values of the respective subgroups. Additionally, we also analyzed the assessed laboratory values as
continuous variables by transforming them into the natural
logarithm (log [variable]). Survival times were estimated
by using the KaplanMeier method; differences were calculated by using the log-rank test on a significance level of
0.05 (two sided). Parameters that reached statistical significance in univariate analysis were included in multivariate Cox regression models using backward elimination.
In order to evaluate the prognostic significance of serial
CA 19-9 measurements during treatment, the lowest CA
19-9 measurement (=nadir) in addition to the baseline
measurement within the first two cycles of chemotherapy
(between day 28 8 and day 56 8) was taken into
account. The 16 patients who underwent chemoradiotherapy were excluded from this analysis due to the heterogeneity of the treatment modality. For the definition of a
biochemical response to therapy, two cutoffs of a 25 and
50 % decline during treatment as proposed by other
authors were chosen (Boeck et al. 2006). In order to
evaluate the impact of the respective method of CA 19-9
measurement on the outcome results, an additional subgroup analysis including only values assessed with the
unique assay (Elecsys, Roche Diagnostics, Germany) was
performed.
Several potential confounders affecting the accuracy of
CA 19-9 kinetics have been described (Boeck et al. 2006).
In order to minimize their influence, the CA 19-9 kinetic
analyses were repeated including only patients without
signs of cholestasis at baseline (defined as total bilirubin
B1.5 9 upper limit of normal, ULN) and patients with a

683

CA 19-9 level at baseline of above the upper limit of the

reference range ([37 U/ml). Furthermore, the landmark
method was used at day 56 8 to avoid the effect of a
guarantee time bias. A landmark is a time point where all
patients still have to be on the treatment protocol and up to
which the evaluation of tumor marker response must have
taken place. Thus, the allocation to a certain prognostic
group depending on definitions of CA 19-9 decline (25 or
50 %) must have taken place before that time point and
survival times are considered only after the landmark.

Results
Patient characteristics
Data from 291 PC patients were included into this pooled
multicenter analysis: 116 patients from the LMU Pancreas Center, 122 patients from the Ro96 trial and 53
patients from the GEMOXCET study. Baseline patient
characteristics are summarized within Table 1. At treatment initiation, 243 patients (84 %) had metastatic disease
and 48 patients (16 %) had locally advanced tumors. The
organ most commonly affected by metastases was the liver
(69 %), followed by lung (18 %) and peritoneum (5 %).
The majority of patients treated at the LMU received
gemcitabine-based chemotherapy (61 %) or a capecitabine-based regimen (23 %); only 16 (of the 24 patients
with locally advanced disease) received upfront chemoradiotherapy. Overall, 253 of all patients (87 %) received
treatment within a prospective clinical trial. Median TTP
from treatment initiation in the study cohort was estimated
with 5.1 months; median OS was 9.0 months (Table 1). At
the time of final data analysis, 237 of the 291 PC patients
(81 %) had died.
Univariate analysis of pre-treatment prognostic factors
Univariate analyses of 12 different clinical characteristics
and laboratory parameters were conducted regarding TTP
and OS, respectively (for details see Table 2). Regarding
TTP, the following clinical parameters revealed a statistical
significance: KPS C90 %, tumor grading with good or
median differentiated tumors (G1 ? G2) and age above
64 years (Table 2). Regarding the endpoint OS, locally
advanced tumors, a KPS C90 % and good or median differentiated tumors (G1 ? G2) had an improved survival.
Among laboratory parameters, normal LDH values
(defined as B250 U/l) had a prognostic significance for
both TTP (5.4 vs. 3.4 months, p \ 0.001) and OS (9.9 vs.
5.9 months, p \ 0.001). When LDH was analyzed as
continuous variableafter transformation into the natural
logarithm (log [LDH])and included into a Cox

123

684
Table 1 Patient characteristics

J Cancer Res Clin Oncol (2013) 139:681689

Characteristic

LMU
patients

Ro96

GEMOXCET

Overall

116

122

53

291

Median

63

63

63

63

Range

3978

4075

3175

3178

Age (years)

Gender
Male

73 (63 %)

82 (67 %)

37 (70 %)

192 (66 %)

Female

43 (37 %)

40 (33 %)

16 (30 %)

99 (34 %)

Stage of disease (at study entry)

Locally advanced

24 (21 %)

24 (20 %)

48 (16 %)

Metastatic

92 (79 %)

98 (80 %)

53 (100 %)

243 (84 %)

29 (25 %)

21 (17 %)

8 (15 %)

58 (20 %)

83 (72 %)

55 (45 %)

31 (59 %)

169 (58 %)

Previous surgery
Karnofsky status (KPS)
90100 %
6080 %

32 (28 %)

66 (54 %)

10 (19 %)

108 (37 %)

Missing

1 (1 %)

1 (1 %)

12 (23 %)

14 (5 %)

G1 ? G2

46 (40 %)

49 (40 %)

22 (42 %)

117 (40 %)

G3 ? G4

59 (51 %)

45 (37 %)

21 (40 %)

125 (43 %)

Missing

11 (9 %)

28 (23 %)

10 (19 %)

49 (17 %)

Tumor grading

Localization of primary
Pancreatic head

72 (62 %)

67 (55 %)

26 (49 %)

165 (57 %)

Pancreatic body or tail

44 (38 %)

28 (23 %)

19 (36 %)

91 (31 %)

Missing

27 (22 %)

8 (15 %)

35 (12 %)

Histology
Adenocarcinoma

112 (97 %)

122 (100 %)

53 (100 %)

287 (99 %)

4 (3 %)

4 (1 %)

Median

749

1,114

3,461

1,137

Range

881,626

6100,000

8100,000

6100,000

Patients treated within a prospective

clinical trial

78 (67 %)

122 (100 %)

53 (100 %)

253 (87 %)

Patients deceased

100 (86 %)

97 (80 %)

40 (76 %)

237 (81 %)

Median TTP (months)

4.1

5.5

4.9

5.1

Median OS* (months)

10.1

9.7

7.3

9.0

Acinar cell carcinoma

Pre-treatment CA 19-9 [U/ml]

LMU Ludwig-MaximiliansUniversity, OS overall survival,

TTP time-to-progression
(* p = 0.019)

proportional hazard regression model, the level of significance remained with a HR of 1.65 (95 % CI 1.182.31,
p = 0.004) for TTP and a HR of 1.93 (95 % CI 1.402.65,
p \ 0.001) for OS (Table 2). Measurements of serum CRP
levels at baseline also had a prognostic impact for TTP and
OS: while our pre-defined cutoff of 1.0 mg/dl did not reach
a level of statistical significance for TTP, this could be
observed after transformation into the natural logarithm log
[CRP] (HR 1.24; 95 % CI 1.031.49, p = 0.026).
Regarding OS, both the cutoff of 1.0 mg/dl (11.4 vs.
6.8 months, p = 0.009) and the transformed values log
[CRP] (HR 1.40, 95 % CI 1.161.69, p = 0.001) had a

123

highly significant impact on OS. Bilirubin values below the

ULN (1.0 mg/dl) did not influence TTP (4.9 vs.
4.4 months, p = 0.202) nor did log [bilirubin]. However,
patients with normal bilirubin levels at chemotherapy initiation lived significantly longer (9.7 vs. 7.2 months,
p = 0.003), and also, the continuous variable log [bilirubin] significantly influenced OS (HR 1.43; 95 % CI
1.201.70, p \ 0.001).
Among the assessed serum tumor markers, CEA values
below or above a cutoff of 4.5 ng/ml showed a significant
correlation with TTP (6.1 vs. 3.5 months, p \ 0.001), and
comparable results were seen when log [CEA] was

J Cancer Res Clin Oncol (2013) 139:681689

685

Table 2 Univariate analysis of pre-treatment prognostic factors

Table 2 continued

Variable

Variable

Median
TTP
(months)

Hazard
ratio
(95 % CI)

A: Time-to-progression (TTP)

[4.5 ng/ml

Stage of disease
Locally advanced
Metastatic

log [CEA]
48
243

4.4
5.1

0.868

1
0.97 (0.681.39)

Karnofsky performance status (KPS)

90100 %

169

5.5

6080 %

108

3.6

G1 ? G2

117

6.1

G3 ? G4

125

4.4

T1 ? T2

48

5.2

T3 ? T4

209

4.7

T (continuous)

256

1
0.004

1.51 (1.142.01)

Tumor grading

101

3.5

<0.001

1.89 (1.352.63)

0.004

1.11 (1.031.19)

0.964

0.99 (0.661.50)

0.351

2.17 (0.4311.12)

0.005

1.63 (1.162.32)

0.037

1.33 (1.021.74)

0.013

1.42 (1.081.88)

203

Hemoglobin
C12 g/dl
\12 g/dl

115
47

log [Hemoglobin]

148

4.1
4.4

B: Overall survival (OS)

0.008

Locally advanced

48

12.9

1.50 (1.102.03)

Metastatic

243

8.3

151

4.2

C64 years

140

5.8

192

4.5

90100 %

169

9.9

0.412

0.85 (0.581.25)

6080 %

108

8.0

0.832

1.02 (0.841.25)

G1 ? G2

117

10.8

G3 ? G4

125

8.3

T1 ? T2

48

8.3

T3 ? T4

209

9.2

99

5.3

1
0.015

0.72 (0.540.94)
1

0.915

1.02 (0.771.34)

Pre-treatment CA 19-9

T (continuous)
256
Age (at study entry)

20

3.8

B63 years

151

9.1

128

6.1

0.76 (0.451.27)

C64 years

140

8.8

[1,000 U/ml

140

4.0

1.11 (0.661.84)

log [age]

291

Pre-treatment CA 19-9
128

6.1

[1,000 U/ml

140

4.0

log [CA 19-9]

268

Male

192

8.6

0.002

1.55 (1.162.06)

Female

99

9.7

<0.001

1.15 (1.081.22)

Pre-treatment CA 19-9 (Elecsys)

B1,000 U/ml

62

6.6

[1,000 U/ml

61

3.5

log [CA 19-9]

123

log [LDH]

247

Not detectable*

20

8.9

B1,000 U/ml

128

10.5

0.002

2.00 (1.293.11)

[1,000 U/ml

140

8.0

<0.001

1.19 (1.081.31)

5.4
3.4

<0.001

1
2.04 (1.472.84)

0.004

1.65 (1.182.31)

CRP
59

6.1

[1.0 mg/dl

56

3.8

log [CRP]

110

0.062

1.52 (0.972.36)

0.026

1.24 (1.031.49)

Bilirubin
B1.0 mg/dl

224

4.9

[1.0 mg/dl

58

4.4

log [Bilirubin]

282

CEA
102

0.82 (0.581.17)

0.290

0.91 (0.771.08)
1

0.554

0.93 (0.721.20)

0.204

0.58 (0.261.34)

6.1

1
0.202

1.26 (0.881.79)

0.228

1.14 (0.921.41)

1
0.196

0.84 (0.641.01)

Pre-treatment CA 19-9

LDH
168
79

1
0.271

Gender

B1,000 U/ml

B250 U/l
[250 U/l

T stage

B1,000 U/ml

0.003

Tumor grading

Not detectable*

B4.5 ng/ml

Karnofsky performance status (KPS)

B63 years

B1.0 mg/dl

Hazard
ratio
(95 % CI)

Stage of disease

Age (at study entry)

Female

Median
TTP
(months)

T stage

Gender
Male

1
0.76 (0.451.27)
0.020

1.11 (0.661.84)

0.006

1.46 (1.111.90)

<0.001

1.13 (1.061.19)

0.048

1.49 (1.002.22)

0.002

1.15 (1.061.26)

<0.001

2.07 (1.552.78)

<0.001

1.93 (1.402.65)

Pre-treatment CA 19-9
B1,000 U/ml

128

10.5

[1,000 U/ml

140

8.0

log [CA 19-9]

268
Pre-treatment CA 19-9 (Elecsys)
B1,000

62

10.3

[1,000

61

8.1

log [CA 19-9]

123

LDH
B250 U/l

168

9.9

[250 U/l

79

5.9

log [LDH]

CRP
B1.0 mg/dl

59

11.4

[1.0 mg/dl

56

6.8

1
0.009

1.69 (1.142.52)

123

686

J Cancer Res Clin Oncol (2013) 139:681689

Table 2 continued
Variable

log [CRP]

Median
TTP
(months)

115

Hazard
ratio
(95 % CI)
0.001

1.40 (1.161.69)

Bilirubin
B1.0 mg/dl

224

9.7

[1.0 mg/dl
log [Bilirubin]

58

7.2

1
0.003
<0.001

1.62 (1.182.24)
1.43 (1.201.70)

CEA
B4.5 ng/ml

102

10.3

[4.5 ng/ml

101

9.0

log [CEA]
C12 g/dl

115

9.0

47

7.1

log [Hemoglobin]

1.35 (0.991.83)

0.053

1.07 (1.001.15)

0.463

1.15 (0.791.67)

0.547

0.64 (0.152.73)

Bold values indicate statistical significance (p \ 0.05)

* Defined as CA 19-9 \ 2.6 U/ml

analyzed as continuous variable (HR 1.11, 95 % CI

1.031.19, p = 0.004). However, the correlation of pretherapeutic CEA values with OS did not reach a level of
statistical significance. The distribution of median pretreatment CA 19-9 levels in the different study populations
is shown in Table 1. The median value of CA 19-9 among
all patients was 1,137 U/ml. LMU patients had the lowest
median value of 749 U/ml, whereas the participants of the
GEMOXCET trial had the highest median count of
3,641 U/ml. Using the pre-defined CA 19-9 cut-off of
1,000 U/ml, a significant correlation with TTP was
observed (6.1 vs. 4.0 months, p = 0.002). These results
were also seen when only values measured with the Elecsys assay (n = 123) were taken into account (median
TTP: 6.6 vs. 3.5 months, p = 0.002). Concerning OS, CA
19-9 values below 1,000 U/ml were linked with an
improved OS, regardless whether all CA 19-9 values (10.5
vs. 8.0 months, p = 0.006) or only those CA 19-9 levels
determined with the Elecsys assay (10.3 vs. 8.1 months,
p = 0.048) were taken into account. After transformation
of the CA 19-9 values into the natural logarithm, a highly
significant correlation of baseline CA 19-9 as continuous
variable with both TTP and OS was observed through all
subgroups (see Table 2).
Multivariate analysis of pre-treatment prognostic
factors
In a multivariate Cox model for TTP where the variables
age, tumor grading, KPS, log [LDH], log [CEA] and log
[CA 19-9] were included, only age and log [CA 19-9]
showed independent statistical significance (Table 3).

123

Prognostic role of serial CA 19-9 measurements

during treatment

1
0.054

Hemoglobin
\12 g/dl

A multivariate model for the endpoint OS contained the

parameters tumor stage, tumor grading, KPS, log [CA
19-9], log [LDH] and log [bilirubin]. Log [bilirubin] was
the only laboratory parameter that kept statistical significance besides tumor stage and KPS. A subgroup analysis
including a smaller number of patients furthermore
revealed an independent prognostic value for log [CRP]
(HR 1.32; 1.061.63, p = 0.011) (see Table 3).

Overall, at least one follow-up measurement of CA 19-9

during the first two cycles of chemotherapy was available
for 186 patients (64 %), among them 83 patients whose
values were measured by the Elecsys method. Two cutoffs of a CA 19-9 response during treatment were applied
(25 and 50 %, respectively) in order to define a biochemical treatment response. As summarized within Table 4, a
strong correlation was observed between a CA 19-9 decline
of at least 25 % with OS (11.9 vs. 8.2 months, p = 0.003)
for all values and for the subgroup determined by the
Elecsys assay (13.4 vs. 8.6 months, p = 0.004). Adopting
the 50 % cutoff in the univariate analysis, no significant
correlation with TTP or OS could be observedneither for
all 186 patients nor for the subgroup of 83 patients measured with the Elecsys assay (Table 4). In a second step,
more stringent inclusion criteria (serum bilirubin
B1.5 9 ULN and baseline CA 19-9 [ 37 U/ml) were
applied for the landmark analysis of the serial CA 19-9
kinetics (see validation analysis in Table 4): this resulted in 138 eligible patients, 64 of them were assessed only
by the Elecsys assay. Again, a robust correlation with OS
was seen for the 25 % cutoff (11.3 vs. 6.2 months,
p \ 0.001; Fig. 1) for all patients and also for the Elecsys
subgroup (11.8 vs. 7.5 months, p = 0.001).
The baseline parameters log [CA 19-9], tumor stage,
tumor grading and KPS together with the variable CA
19-9 decline of at least 25 % were subsequently included
into a multivariate Cox model: besides tumor grading and
KPS, also the 25 % cutoff for the CA 19-9 decline during
treatment showed an independent prognostic significance
for the endpoint OS (HR 2.40, 95 % CI 1.543.74,
p \ 0.001).

Discussion
Within a pooled analysis based on data from two German
randomized phase II trials in conjunction with data from
one high-volume German Cancer Center, we were able to
investigate the prognostic role of 12 clinical and biochemical variables in patients with advanced PC. As

J Cancer Res Clin Oncol (2013) 139:681689

687

Table 3 Multivariate analysis of pre-treatment prognostic factors

Variable

HR (95 % CI)

A: Time-to-progression (TTP), n = 133

Age (B63 years vs. older)

0.66 (0.440.98)

0.040

Tumor grading (G1 ? G2 vs.

G3 ? G4)

1.07 (0.691.65)

0.766

KPS (90100 % vs. 6080 %)

1.30 (0.871.95)

0.208

log [CA 19-9]

1.18 (1.091.28)

<0.001

log [LDH]

1.13 (0.642.00)

0.678

log [CEA]
log [CA 19-9] Elecsys (n = 84)

1.07 (0.981.18)
1.08 (0.941.24)

0.115
0.257

log [CRP] (n = 84)

1.03 (0.821.30)

0.794

Stage of disease (locally advanced vs.

metastatic)

1.56 (1.032.35)

0.036

Tumor grading (G1 ? G2 vs.

G3 ? G4)

1.32 (0.941.85)

0.107

KPS (90100 % vs. 6080 %)

1.91 (1.372.67)

<0.001

log [CA 19-9]

1.02 (0.951.10)

0.581

log [LDH]

1.21 (0.831.75)

0.321

log [Bilirubin]

1.83 (1.432.36)

<0.001

log [CA 19-9] Elecsys (n = 102)

1.04 (0.941.50)

0.429

log [CRP] (n = 93)

1.32 (1.061.63)

0.011

B: Overall survival (OS), n = 183

Bold values indicate statistical significance (p \ 0.05)

expected, stage of disease and KPS were significant prognostic factors for TTP and OS also in our study population
(Heinemann et al. 2012). Interestingly, an age of C64 years
at study entry was associated with a prolonged TTP (5.8 vs.
4.2 months), without having an impact on OS (8.8 vs.
9.1 months), respectively. Thus, one might conclude that
age per se should be no contraindication for the use of
palliative chemotherapy in PC. We furthermore hypothesize that the prolongation of TTP in older patients might
argue for a slower tumor growth, which however was not
transferred into a survival benefit (possibly due to agedependent co-morbidities).
LDH is a known parameter indicating a high turnover of
cells with subsequent release of the intracellular enzyme.
The upper limit of normal (250 U/l) was chosen as a cutoff
for this study, and significant differences for TTP and OS
were observed in univariate (but not in multivariate)
analyses. The transformation into the natural logarithm,
potentially a more adequate method for the evaluation of a
continuous variable with a broad range, also revealed a
correlation with clinical outcome. Up to now, there are
only few reports describing a potential prognostic role of
LDH in patients with PC (Bramhall et al. 2001; Stocken
et al. 2008). Remarkably, baseline CRP levels also indicated a highly significant correlation with OS when using a
cutoff of 1.0 mg/dl in the univariate analysis (Falconer

et al. 1995). The transformation of CRP into the natural

logarithm was similarly associated with TTP and OS. On
multivariate analysis, log [CRP] retained independent significance for OS. These results are consistent with other
reports on CRP in the perioperative or palliative setting for
PC (Moses et al. 2009; Pine et al. 2009). Data from
resectable colorectal cancer demonstrated that a low CD4?
T-lymphocyte infiltration in the tumor was linked with
elevated CRP serum levels, both reflecting a poor cancerspecific survival. Thus, tumor progression might not
merely be determined by tumor biology only, but also by
the immunologic response of the host (Canna et al. 2005).
Both parameters, LDH and CRP, were also shown to have
a significant impact on OS in an univariate subgroup
analysis (published online only) from the large phase III
AViTA study (n = 607): LDH levels at baseline of [ULN
resulted in a HR for death of 0.59 (95 % CI 0.430.82) as
did pre-treatment CRP levels of [1.4 mg/dl (HR 0.65,
95 % CI 0.510.84) (van Cutsem et al. 2009).
Baseline CA 19-9 levels were significant predictors of
TTP in uni- and multivariate analyses in our patient cohort
(Tables 2 and 3). This observation was detected when we
applied a cutoff of 1,000 U/ml for dichotomizing the CA
19-9 variable as well as for the analysis of log [CA 19-9] as
continuous variable. Of note, this association seems not to
be assay dependent, as results were the same when analyzing the Elecsys only subgroup in univariate analysis.
Regarding OS, pre-treatment CA 19-9 also had a highly
significant impact on survival, irrespective of the applied
assay (Table 2). One important scientific (and clinically
relevant) finding from this study is summarized in Table 4:
a CA 19-9 decline of C25 % during treatment seems
appropriate to define subgroups with a different outcome
for TTP and OS. In contrast, the 50 % cutoff seems inappropriate as a marker for a biochemical treatment response.
Interestingly, this observation again was independent of the
applied CA 19-9 assay. These data are consistent with the
result reported from Hess and colleagues who found that an
early decrease in CA 19-9 concentration of at least 50 %
after two cycles of chemotherapy was not associated with a
prolonged OS in their biomarker analysis from the large
randomized SAKK 44/00 trial (Hess et al. 2008). Thus, one
might conclude that an early CA 19-9 decline of 25 %
during treatment is adequate in order to define a prognostically favorable subgroup of PC patients and that is
important to use the same assay in the follow-up of a single
patient (intraindividual approach); however, the use of a
unique assay for all patients (e.g., within the setting of a
multicenter study) does not necessarily seem to be required
for generating valid tumor marker data (Boeck et al. 2006).
The strength of this pooled retrospective, multicenter
study arises from the fact that we were able to analyze a
large patient number [according to the REMARK

123

688
Table 4 Prognostic value of
CA 19-9 kinetics during the first
two treatment cycles (day
2064)

J Cancer Res Clin Oncol (2013) 139:681689

Overall survival (OS)

CA 19-9
decline

OS
(months)

Time-to-progression (TTP)
HR (95 % CI)

TTP
(months)

HR (95 % CI)

0.003

1.62 (1.172.25)

5.8

0.018

1.49 (1.012.09)

0.076

1.34 (0.971.87)

0.372

1.16 (0.841.61)

7.0
2.6

0.003

2.18 (1.283.72)

5.8

0.785

1.08 (0.631.83)

0.011

1.69 (1.122.55)

0.508

1.14 (0.781.66)

0.016

2.11 (1.133.95)

0.748

1.11 (0.602.04)

All patients (n = 186)

C25 %

110

11.9

\25 %

76

8.2

C50 %

77

11.9

\50 %

109

8.8

4.4
5.5
5.4

Elecsys patients only (n = 83)

C25 %
\25 %

46
37

13.4
8.6

0.004

2.03 (1.243.33)

C50 %

27

12.4

0.259

1.36 (0.792.34)

\50 %

56

9.5

6.1

Validation* all patients (n = 138)

Bold values indicate statistical

significance (p \ 0.05)
* Patients selected for
validation: landmark at day
56 8; serum bilirubin
B1.5 mg/dl; CA
19-9 C 37 U/ml

C25 %

89

11.3

\25 %

49

6.2

C50 %

59

11.3

\50 %

79

6.8

<0.001

2.12 (1.443.14)

0.032

1.51 (1.032.22)

3.8
2.8
3.6
3.5

Validation* Elecsys patients only (n = 64)

C25 %

39

11.8

\25 %

25

7.5

C50 %

22

11.8

\50 %

42

7.6

0.001

2.52 (1.404.52)

5.9
2.8

0.146

1.57 (0.852.88)

4.7
4.7

Fig. 1 KaplanMeier curve for

OS based on a CA 19-9 decline
of C25 % during the first two
chemotherapy cycles: subgroup
CA 19-9 validationall
patients (n = 138; for details
see text and Table 4); median
OS: 11.3 versus 6.2 months
(HR 2.12, p \ 0.001)

guidelines (McShane et al. 2005)] and that most of the

patients included (87 %) were treated within prospective
clinical trials. All marker data included in the profound
statistical analyses (e. g., including a landmark method)
were collected prospectively. Furthermore, we were able to
generate differentiated data on the value of the applied CA
19-9 assay; an approach that isat least to our

123

knowledgeunique up to now. Limitations are based on

the different numbers that were available for each single
variable (e. g., CRP data were available from LMU patients
only) and on a significant proportion of missing data for the
analysis of the CA 19-9 kinetics (n = 186/291)a problem that often is observed when subgroup data from multicenter PC trials are reported.

J Cancer Res Clin Oncol (2013) 139:681689

In conclusion, this study confirms the important role of

CA 19-9 as tumor marker in advanced PC and also highlights the potential role of other serum markers like CRP,
LDH and bilirubin to serve as prognostically relevant
factors in this disease. These findings may play an important role for future stratification procedures in clinical trials
and also for the selection of good risk and poor risk
PC patients for different treatment strategies (e. g., singeagent gemcitabine vs. intensive FOLFIRINOX chemotherapy; Conroy et al. 2011; Heinemann et al. 2012).
However, a prospective validation of these novel hypothesis-generating data is recommended.
Acknowledgments The authors would like to thank all patients and
their families, nurses, study coordinators and investigators that
actively participated in the Ro96 and the GEMOXCET study, thereby
enabling this pooled analysis. This work is part of the doctoral thesis
of Michael Haas.
Conflict of interest All authors declare that they have no conflict of
interest.

References
Boeck S, Stieber P, Holdenrieder S, Wilkowski R, Heinemann V
(2006) Prognostic and therapeutic significance of carbohydrate
antigen 199 as tumor marker in patients with pancreatic cancer.
Oncology 70(4):255264
Boeck S, Hoehler T, Seipelt G, Mahlberg R, Wein A, Hochhaus A
et al (2008) Capecitabine plus oxaliplatin (CapOx) versus
capecitabine plus gemcitabine (CapGem) versus gemcitabine
plus oxaliplatin (mGemOx): final results of a multicenter
randomized phase II trial in advanced pancreatic cancer. Ann
Oncol 19(2):340347
Boeck S, Haas M, Laubender RP, Kullmann F, Klose C, Bruns CJ
et al (2010) Application of a time-varying covariate model to the
analysis of CA 199 as serum biomarker in patients with
advanced pancreatic cancer. Clin Cancer Res 16(3):986994
Bramhall SR, Rosemurgy A, Brown PD, Bowry C, Buckels JA (2001)
Marimastat as first-line therapy for patients with unresectable
pancreatic cancer: a randomized trial. J Clin Oncol
19(15):34473455
Canna K, McArdle PA, McMillan DC, McNicol AM, Smith GW,
McKee RF et al (2005) The relationship between tumour
T-lymphocyte infiltration, the systemic inflammatory response
and survival in patients undergoing curative resection for
colorectal cancer. Br J Cancer 92(4):651654

689
Conroy T, Desseigne F, Ychou M, Bouche O, Guimbaud R, Becouarn
Y et al (2011) FOLFIRINOX versus gemcitabine for metastatic
pancreatic cancer. N Engl J Med 364(19):18171825
Duffy MJ, Sturgeon C, Lamerz R, Haglund C, Holubec VL, Klapdor
R et al (2010) Tumor markers in pancreatic cancer: a European
Group on Tumor Markers (EGTM) status report. Ann Oncol
21(3):441447
Falconer JS, Fearon KC, Ross JA, Elton R, Wigmore SJ, Garden OJ
et al (1995) Acute-phase protein response and survival duration
of patients with pancreatic cancer. Cancer 75(8):20772082
Heinemann V, Haas M, Boeck S (2012) Systemic treatment of
advanced pancreatic cancer. Cancer Treat Rev 38(7):843853
Hess V, Glimelius B, Grawe P, Dietrich D, Bodoky G, Ruhstaller T
et al (2008) CA 199 tumour-marker response to chemotherapy
in patients with advanced pancreatic cancer enrolled in a
randomised controlled trial. Lancet Oncol 9(2):132138
Humphris JL, Chang DK, Johns AL, Scarlett CJ, Pajic M, Jones MD
et al (2012) The prognostic and predictive value of serum
CA19.9 in pancreatic cancer. Ann Oncol 23(7):17131722
Kullmann F, Hollerbach S, Dollinger MM, Harder J, Fuchs M,
Messmann H et al (2009) Cetuximab plus gemcitabine/oxaliplatin (GEMOXCET) in first-line metastatic pancreatic cancer: a
multicentre phase II study. Br J Cancer 100(7):10321036
McShane LM, Altman DG, Sauerbrei W, Taube SE, Gion M, Clark
GM, Statistics Subcommittee of the NCI-EORTC Working
Group on Cancer Diagnostics (2005) REporting recommendations for tumour MARKer prognostic studies (REMARK). Br J
Cancer 93(4):387391
Moore MJ, Goldstein D, Hamm J, Figer A, Hecht JR, Gallinger S et al
(2007) Erlotinib plus gemcitabine compared with gemcitabine
alone in patients with advanced pancreatic cancer. A phase III
trial of the National Cancer Institute of Canada Clinical trials
group. J Clin Oncol 25(15):19601966
Moses AG, Maingay J, Sangster K, Fearon KC, Ross JA (2009) Proinflammatory cytokine release by peripheral blood mononuclear
cells from patients with advanced pancreatic cancer: relationship
to acute phase response and survival. Oncol Rep 21(4):10911095
Pine JK, Fusai KG, Young R, Sharma D, Davidson BR, Menon KV
et al (2009) Serum C-reactive protein concentration and the
prognosis of ductal adenocarcinoma of the head of pancreas. Eur
J Surg Oncol 35(6):605610
Reni M, Cereda S, Balzano G, Passoni P, Rognone A, Fugazza C et al
(2009) Carbohydrate antigen 199 change during chemotherapy for
advanced pancreatic adenocarcinoma. Cancer 115(12):26302639
Stocken DD, Hassan AB, Altman DG, Billingham LJ, Bramhall SR,
Johnson PJ et al (2008) Modelling prognostic factors in
advanced pancreatic cancer. Br J Cancer 99(6):883893
Van Cutsem E, Vervenne WL, Bennouna J, Humblet Y, Gill S, Van
Laethem JL et al (2009) Phase III trial of bevacizumab in
combination with gemcitabine and erlotinib in patients with
metastatic pancreatic cancer. J Clin Oncol 27(13):22312237
Vincent A, Herman J, Schulick R, Hruban RH, Goggins M (2011)
Pancreatic cancer. Lancet 378(9791):607620

123