Tumor Marker Kinetics as Prognosticators in Patients with Unresectable Gallbladder Adenocarcinoma Undergoing Palliative Chemotherapy.

BACKGROUND/AIMS: To determine the prognostic value of carcinoembryonic antigen (CEA) and carbohydrate antigen (CA) 19-9 in gallbladder cancer (GBC) during palliative chemotherapy.
METHODS: One hundred and twenty-three patients with pathologically confirmed unresectable GBC were included. Differences in serum CEA and CA 19-9 levels before and after chemotherapy were measured. Receiver operating characteristic curve analysis, Kaplan-Meier analyses of CEA, CA 19-9, and combined changes were performed to assess the optimal cutoff values and survival rates.
RESULTS: Patients with decreased tumor markers had significantly better progression-free survival (PFS) and overall survival (OS) than patients with increased tumor markers. The pre- and postchemotherapy CA 19-9 ratio had the highest area-under-the-curve values for predicting 3-month PFS and 1-year OS. In the multivariate analysis, increases in serum CA 19-9 during palliative chemotherapy in patients with unresectable GBC was an independent prognosticator of poor PFS and OS, with hazard ratios of 2.20 (p=0.001) and 1.67 (p=0.020), respectively. Patients with increases ＞10-fold were considered to have progressive disease, whereas individuals with increases ＞3-fold were likely to benefit from early imaging follow-up.
CONCLUSIONS: CA 19-9 kinetics was a reliable prognosticator of PFS and OS in patients with unresectable GBC who underwent palliative chemotherapy.

INTRODUCTION

Gallbladder cancer (GBC) is the most common malignant tumor of the biliary tract. GBC is associated with a poor prognosis, with a 5-year overall survival (OS) rate of 18.5%, and only 20% of patients are eligible for resection at the time of diagnosis.1 Palliative chemotherapy has shown survival benefits,2 but the response rates are relatively low (17.1% to 36.6%).3 The median survival time of patients with GBC is 4.6 to 11.7 months.2,3

In biliary tract cancer, serum carbohydrate antigen (CA) 19-9 and carcinoembryonic antigen (CEA) have been associated with prognosis.4 Changes in these tumor markers during treatment were reported as prognosticators in pancreatic cancer5–8 and cholangiocarcinoma.9 For GBC, serum CEA and CA 19-9 are useful diagnostic10–13 and prognostic12,14–17 markers. However, previous studies employed spot measurements of CEA or CA 19-9, and studies evaluating the prognostic role of tumor marker kinetics in GBC have not been conducted.

The aim of the present study was to assess the prognostic values of serum CEA and CA 19-9 and their kinetics during palliative chemotherapy in patients with unresectable GBC.

MATERIALS AND METHODS

1. Patients and study design

A single-center retrospective study was conducted in patients with unresectable GBC, which was pathologically confirmed as adenocarcinoma between January 2005 and April 2015 at Seoul National University Hospital. Patients who underwent at least four cycles of palliative chemotherapy and who had baseline and postchemotherapy tumor marker records were included. Patients with other malignant tumors diagnosed previously, a history of systemic chemotherapy, or normal pre- and postchemotherapy tumor marker values were excluded (Fig. 1). The final analysis set included 123 patients.

Fig. 1

Patient selection criteria.

GB, gallbladder; CA 19-9, carbohydrate antigen 19-9; CEA, carcinoembryonic antigen.

All patients were followed up until 30 November 2015, and observations were censored at the time of death or loss to follow-up.

2. Data collection

Patient characteristics such as age, sex, symptoms at admission, and comorbid disease status (Charlson comorbidity index score)18 were obtained. Variables in tumor characteristics including tumor location (fundus, body, and neck) and distant metastasis were collected.

Serum CEA and CA 19-9 levels at baseline and postchemotherapy were evaluated. Baseline levels were measured within 3 days before chemotherapy initiation (CEApre and CA 19-9pre). Postchemotherapy levels were measured within 3 days after the end of the second cycle of chemotherapy (CEApost and CA 19-9post). Serum CEA and CA 19-9 were measured using a commercially available immunoradiometric assays (CA 19-9: IZO TOP®, Institute of Isotopes Co., Ltd., Budapest, Hungary; CEA: RIAKEY®, Shinjin Medics Inc., Goyang, Korea). The initial total serum bilirubin levels and prothrombin times were also evaluated.

Data on progression-free survival (PFS) and OS were collected. Disease progression (PD) was assessed using abdominal computed tomography every two to four cycles of chemotherapy. PFS data was censored according to the date of loss to follow-up. The date of death was sourced from the records of the Korean Central Cancer Registry.

3. Statistical analyses

Tumor marker kinetics were defined as and . Combined tumor marker kinetics were defined as COMBchange=CEAchange×CA 19-9change. The median, and first and third quadrant values, of the kinetic parameters were calculated.

Receiver operating characteristic (ROC) curve analysis of CEAchange, CA 19-9change, and COMBchange was performed to assess 3-month PFS and 1-year OS rates. The cutoff value for the highest sum of sensitivity and specificity was used for further analyses. Kaplan-Meier analyses were performed for survival evaluation. The log-rank test was used to assess the relationships between tumor marker kinetic parameters and PFS or OS.

Univariate analysis with the log-rank test was conducted to compare survival using the cutoff values of the tumor marker kinetic parameters. Factors associated with survival in the univariate analysis with a p-value <0.10 were used in multivariate analysis. Cox regression analysis was performed to identify independent prognosticators.

Hazard ratios and 95% confidence intervals were calculated for each predictive factor. Two-sided p-values of <0.05 were considered statistically significant. All statistical analyses were performed with IBM SPSS Statistics version 22.0 (IBM Corp., Armonk, NY, USA).

RESULTS

1. Patient demographics

Baseline patient characteristics are listed in Table 1. The median age was 64 years (range, 25 to 85 years). Thirty patients had stage IIIb disease, all of which had unresectable disease due to extensive liver invasion or regional lymph node metastasis. The majority of patients underwent gemcitabine plus cisplatin chemotherapy (61.0%) with a mean duration of 4.8 (±3.0) cycles. Other chemotherapy regimens included TS-1 plus cisplatin (15.5%) with a mean duration of 5.2 (±3.1) cycles, gemcitabine plus oxaliplatin (13.8%) with a mean duration of 5.9 (±3.7) cycles, gemcitabine plus TS-1 (4.1%), infusional 5-fluorouracil, doxorubicin, and mitomycin-C (1.6%), capecitabine alone (1.6%), and TS-1 alone (2.4%). The median Charlson comorbidity index was 8 (range, 3–13). The median serum CEA and CA 19-9 levels were 3.9 mg/L (range, 0.5 to 1,350 mg/L) and 413 U/mL (range, 1 to 145,000 U/mL), respectively. The median interval between diagnosis and chemotherapy was 16 days (range, 0 to 123 days). The median PFS and OS were 3.9 and 8.1 months, respectively.

Table 1

Baseline Characteristics of Eligible Patients

Characteristic	Value
Sex
Male	62 (50.4)
Female	61 (49.6)
Age, yr
≥65	57 (46.3)
<65	66 (53.7)
Drinking history
Yes	31 (25.2)
No	92 (74.8)
Smoking history
Yes	27 (22.0)
No	96 (78.0)
Symptoms
Yes	113 (91.9)
No	10 (8.1)
CCI (cancer score subtracted)
≥4.0	46 (37.4)
<4.0	77 (62.6)
Total bilirubin, mg/dL
≥3.0	32 (26.0)
<3.0	91 (74.0)
CEA, ng/mL
≥4.0	66 (46.3)
<4.0	57 (53.7)
CA 19-9, U/mL
≥400	68 (55.3)
<400	55 (44.7)
Performance status (ECOG)
0 or 1	103 (82.1)
2	22 (17.9)
Location
Fundus and body	95 (77.2)
Neck	28 (22.8)
Metastasis
Yes	78 (63.4)
No	45 (36.6)
Stage
IIIb	30 (24.4)
IVa	15 (12.2)
IVb	78 (63.4)
Biliary drainage
Yes	51 (41.5)
No	7 (58.5)
Chemotherapy
Gemcitabine-based	99 (80.5)
Others	24 (19.5)
Concurrent radiotherapy	5 (4.1)
CEAchange (n=68)	1.41 (0.06–10.0)
CA19-9change (n=106)	1.00 (0.01–46.9)
COMBchange (n=51)	1.19 (0.01–186)

Data are presented as number (%) or median (range).

CCI, Charlson comorbidity index; CEA, carcinoembryonic antigen; CA 19-9, carbohydrate antigen 19-9; ECOG, Eastern Cooperative Oncology Group; COMB, combination.

2. Prognostic value of CEAchange, CA 19-9change, and COMBchange

In the ROC analysis, the areas-under-the-curve of the CEAchange, CA 19-9change, and COMBchange for predicting 3-month PFS were 0.727, 0.750, and 0.734, whereas those for 1-year OS were 0.623, 0.742, and 0.720, respectively (Fig. 2).

Fig. 2

Receiver operating characteristic (ROC) curve analysis of tumor marker kinetics as predictors of survival. (A) Progression-free survival. (B) Overall survival.

CEA, carcinoembryonic antigen; CA 19-9, carbohydrate antigen 19-9.

The relationships between tumor marker kinetics and survival are shown in Table 2. CA 19-9change was significantly correlated with PFS and OS. However, CEAchange was significantly correlated with PFS and OS only at the cutoff value of 1.0. The relationship between COMBchange and PFS and OS was variable but significant at most values with the exception of a cutoff value of 2.0 for OS.

Table 2

Sensitivity and Specificity for Predicting PFS and OS According to Serum Tumor Marker Changes after Chemotherapy

	PFS			OS
	CEAchange	CA 19-9change	COMBchange	CEAchange	CA 19-9change	COMBchange
Cutoff 0.5
Sensitivity, %	91.7*	84.2*	82.8*	82.8†	82.8†	82.8†
Specificity, %	31.2*	51.0*	50.0*	50.0†	50.0†	50.0†
<0.5	5.93±1.50‡	6.00±1.11‡	5.93±1.19‡	10.20±2.65‡	11.40±1.19‡	11.40±1.49‡
≥0.5	2.57±0.63‡	2.47±0.29‡	2.10±0.41‡	6.60±0.39‡	6.80±0.23‡	6.00±0.61‡
p-value	0.067	<0.001	0.009	0.258	0.001	0.011
Cutoff 1.0
Sensitivity, %	83.3*	68.4*	72.4*	72.4†	72.4†	72.4†
Specificity, %	59.4*	69.4*	68.2*	68.2†	68.2†	68.2†
<1.0	5.87±0.36‡	5.60±0.79‡	5.30±0.78‡	11.40±0.80‡	9.60±1.12‡	8.93±2.16‡
≥1.0	2.31±0.36‡	2.17±0.27‡	2.10±0.23‡	6.17±0.50‡	6.63±0.43‡	6.17±0.52‡
p-value	0.002	<0.001	0.002	0.036	0.008	0.011
Cutoff 2.0
Sensitivity, %	27.8*	45.6*	55.2*	55.2†	55.2†	55.2†
Specificity, %	85.4*	85.7*	72.7*	72.7†	72.7†	72.7†
<2.0	4.27±0.97‡	4.40±0.71‡	4.37±0.54‡	7.47±0.80‡	8.80±0.99‡	7.87±1.11‡
≥2.0	2.47±0.53‡	2.10±0.40‡	2.17±0.49‡	6.57±0.25‡	6.53±0.54‡	6.53±0.47‡
p-value	0.397	0.001	0.039	0.692	0.021	0.173

Data are presented as mean±SD.

PFS, progression-free survival; OS, overall survival; CEA, carcinoembryonic antigen; CA 19-9, carbohydrate antigen 19-9; COMB, combination.

Sensitivity plus specificity for 3-month PFS;

Sensitivity plus specificity for 1-year OS;

Survival in months.

Survival curves according to tumor kinetic parameters using a cutoff value of 1.0 are shown in Fig. 3. The median PFS was 5.9 and 2.3 months in patients with a CEAchange of <1 and ≥1, respectively (p=0.002). The median OS was 11.4 and 6.2 months in those with a CEAchange of <1 and ≥1, respectively (p=0.036). The median PFS was 5.6 and 2.2 months in patients with a CA 19-9change of <1 and ≥1, respectively (p<0.001). The median OS was 9.6 and 6.6 months in patients with a CA 19-9change of <1 and ≥1, respectively (p=0.008). The median PFS was 5.3 and 2.1 months in those with a COMBchange of <1 and ≥1 (p=0.002). The median OS was 8.9 and 6.2 months in those with a COMBchange of <1 and ≥1, respectively (p=0.011).

Fig. 3

(A–F) Kaplan-Meier plot of progression-free and overall survival according to serum tumor marker changes after chemotherapy using a cutoff value of 1.0.

CEA, carcinoembryonic antigen; CA 19-9, carbohydrate antigen 19-9; COMB, combination.

CA 19-9change was the most valuable prognostic marker. Kaplan-Meier analyses according to CA 19-9change cutoff value of 0.4, 1.0, and 2.0 (which represent the first quadrant, median, and third quadrant) are shown in Fig. 4. Patients with a CA 19-9change <0.4 had significantly better survival compared to those with greater changes. The linearity of the PFS and OS in were statistically significant (both p<0.001). Because CA 19-9change was the most valuable prognosticator, we used CA 19-9change as a marker in further analyses.

Fig. 4

Progression-free (A) and overall survival (B) according to changes in carbohydrate antigen (CA) 19-9.

3. CA 19-9change as a predictor of response to chemotherapy

Responses after four-cycle (rather than 3-month due to the variety of chemotherapy regimen) of chemotherapy were assessed according to the modified Response Evaluation Criteria in Solid Tumors (version 1.1). Sixteen patients (13.0%) had a partial response, 41 (33.3%) had stable disease, and 66 (53.7%) showed PD. A high CA 19-9change was correlated with PD (p=0.001), but a low CA 19-9change was not associated with partial response (p=0.500). Three patients with a CA 19-9change >10.0 showed PD after four-cycle of chemotherapy, and 20 patients with CA 19-9change >3.0, 17 (85%) of whom underwent four-cycle of chemotherapy showed PD.

4. Prognostic value of CA 19-9change associated with PFS

The correlations of CA 19-9change with other baseline variables were analyzed, but there were no significant correlations (Table 3). Univariate analysis revealed that male sex (p=0.095), a primary mass located in the gallbladder neck (p=0.076), positive distant metastasis (p=0.018), CEA ≥4.0 ng/mL at diagnosis (p=0.045), and CA 19-9change ≥1.0 (p<0.001) had a p-value of <0.10 for PFS. However, CA 19-9change ≥1.0 alone was an independent prognosticator of PFS (p=0.001) (Table 4).

Table 3

Correlations between Changes in CA 19-9 and Other Variables

	CA 19-9		p-value
	Decrease	Increase
Sex
Male	26	30
Female	27	23	0.560
Age, yr
≥65	27	23
<65	26	30	0.560
Drinking
Yes	16	12
No	37	41	0.509
Smoking
Yes	11	13
No	42	40	0.817
CCI (cancer subtracted score)
≥4.0	21	20
<4.0	32	33	1.000
Symptoms
Yes	47	50
No	6	3	0.488
Location
Fundus and body	35	45
Neck	18	8	0.041
Distant metastasis
Yes	31	39
No	22	14	0.151
Biliary drainage
Yes	25	19
No	28	34	0.324
CEA, ng/mL
≥4.0	25	27
<4.0	28	26	0.846
CA 19-9, U/mL
≥400	37	27
<400	16	26	0.073

CA 19-9, carbohydrate antigen 19-9; CCI, Charlson comorbidity index; CEA, carcinoembryonic antigen.

Table 4

Prognosticators of PFS in Patients with Unresectable Gallbladder Cancer

	Univariate			Multivariate
	No.	PFS	p-value	HR	95% CI	p-value
Sex
Male	62	2.7
Female	61	4.8	0.095	0.81	0.53–1.23	0.317
Age, yr
≥65	57	3.9
<65	66	3.9	0.918
Drinking
Yes	31	4.0
No	92	3.7	0.315
Smoking
Yes	27	3.9
No	96	4.0	0.839
Symptom
Yes	113	3.9
No	10	2.6	0.826
Performance status (ECOG)
0 or 1	101	4.0
2	22	3.2	0.505
Location
Fundus and body	95	3.4
Neck	28	4.5	0.076	0.79	0.46–1.38	0.410
Distant metastasis
Yes	78	2.7
No	45	5.7	0.018	1.45	0.92–2.29	0.115
Biliary drainage
Yes	51	3.4
No	72	4.8	0.171
Total bilirubin, mg/dL
≥3.0	32	3.4
<3.0	91	4.3	0.336
CEA, ng/mL
≥4.0	66	3.2
<4.0	57	4.4	0.066	1.44	0.91–2.29	0.112
CA 19-9, U/mL
≥400	68	3.9
<400	55	3.9	0.328
Chemotherapy
GP	75	4.0
Others	48	3.4	0.851
Interval between diagnosis and treatment, day
≤15	59	4.0
>15	64	3.9	0.433
CA 19-9change
≥1.0	53	5.6
<1.0	53	2.2	<0.001	2.20	1.39–3.47	0.001

PFS, progression-free survival; HR, hazard ratio; CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; CEA, carcinoembryonic antigen; CA 19-9, carbohydrate antigen 19-9; GP, gemcitabine plus cisplatin.

5. Prognostic value of CA 19-9change associated with OS

Univariate analyses revealed that male sex (p=0.058), an Eastern Cooperative Oncology Group performance score of 2 (p=0.050), distant metastasis (p=0.072), CEA ≥4.0 ng/mL (p=0.001), CA 19-9 ≥400 U/mL (p=0.007), and CA 19-9change ≥1.0 (p=0.008) had a p-value of <0.10. Multivariate Cox regression analysis showed that baseline CEA ≥4.0 ng/mL, baseline CA 19-9 ≥400 U/mL, and CA 19-9change ≥1.0 were independent prognosticators of OS (p=0.018, p=0.022, and p=0.020, respectively) (Table 5).

Table 5

Prognosticators of OS in Patients with Unresectable Gallbladder Cancer

	Univariate			Multivariate
	No.	OS	p-value	HR	95% CI	p-value
Sex
Male	62	7.0
Female	61	8.8	0.058	0.81	0.53–1.23	0.320
Age, yr
≥65	57	6.9
<65	66	9.5	0.130
Drinking
Yes	31	7.5
No	92	8.1	0.473
Smoking
Yes	27	8.3
No	96	7.8	0.419
Symptoms
Yes	113	7.9
No	10	8.9	0.274
Performance status (ECOG)
0 or 1	101	8.3
2	22	6.6	0.050	1.15	0.69–1.93	0.592
Location
Fundus and body	95	7.8
Neck	28	9.5	0.104
Distant metastasis
Yes	78	6.8
No	45	9.6	0.072	1.35	0.87–2.11	0.186
Biliary drainage
Yes	51	7.0
No	72	8.3	0.669
Total bilirubin, mg/dL
≥3.0	32	7.0
<3.0	91	8.1	0.763
CEA, ng/mL
≥4.0	66	6.8
<4.0	57	9.6	0.001	1.76	1.10–2.80	0.018
CA 19-9, U/mL
≥400	68	7.0
<400	55	9.6	0.007	1.74	1.08–2.80	0.023
Chemotherapy
GP	75	7.8
Others	48	8.3	0.250
CA 19-9change
≥1.0	53	6.6
<1.0	53	9.6	0.008	1.67	1.08–2.58	0.020

OS, overall survival; HR, hazard ratio; CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; CEA, carcinoembryonic antigen; CA 19-9, carbohydrate antigen 19-9; GP, gemcitabine plus cisplatin.

6. Effect of total serum bilirubin on CA 19-9 levels

The correlation between serum CA 19-9 and serum total bilirubin level, and that between CA 19-9change and serum total bilirubin were not significant (p=0.155 and 0.845, respectively). We did a subgroup analysis which divided the subjects according to serum total bilirubin levels, and 29 out of 106 patients had a serum total bilirubin level >3.0 mg/dL (Table 6). The median PFS was 4.0 and 2.7 months in patients with a CA 19-9change of <1 and ≥1, respectively (p=0.296). The median OS was 7.1 and 6.7 months in patients with a CA 19-9change of <1 and ≥1, respectively (p=0.955).

Table 6

Survival Analysis of High- and Low-Serum Bilirubin Subgroups

	Bilirubin ≥3 mg/dL			Bilirubin <3 mg/dL
	CA 19-9change <1 (n=18)	CA 19-9change ≥1 (n=11)	p-value	CA 19-9change <1 (n=35)	CA 19-9change ≥1 (n=42)	p-value
PFS, median, mo	4	2.7	0.296	5.9	2.2	<0.001
OS, median, mo	7.1	6.7	0.955	10.2	6.6	<0.001

CA 19-9, carbohydrate antigen 19-9; PFS, progression-free survival; OS, overall survival.

DISCUSSION

In this retrospective study, we aimed to assess the prognostic value of serum tumor marker kinetics after chemotherapy in patients with unresectable GBC. The results revealed that tumor marker changes after first two cycles of chemotherapy were independent prognosticators of survival. Serum CEA, serum CA 19-9, or a combination of the two were valuable prognosticators; however, among them, CA 19-9 kinetics was the most valuable prognosticator of survival. In addition, the serum CA 19-9 level after two cycles of chemotherapy was a valuable predictor of PD after four cycles of chemotherapy. Therefore, we suggest that patients with >10-fold increase in serum CA 19-9 after two cycles of chemotherapy should be considered as having PD, and patients with >3-fold increase in serum CA 19-9 should be considered for early imaging studies.

Multivariate analyses in previous studies of patients with GBC that utilized specific cutoff values of spot serum CA 19-9 did not identify serum CA 19-9 as an independent prognosticator.19–21 This could be explained by the wide variation in CA 19-9 secretion levels between GBC cases. Although the serum CA 19-9 levels vary between patients and may not represent tumor burden, the relative value of CA 19-9 in a single patient may be reflective of the tumor burden, regardless of the absolute CA 19-9 level. The present study showed that the relative change in CA 19-9 did not correlated with spot serum CA 19-9 measurements, and that the relative change was an independent prognosticator, similar to previous findings in pancreatic cancer5–8 and cholangiocarcinoma.9

Yu et al.14 focused on the change of serum CA 19-9 and CEA after resection of GBC. They showed that increased CA 19-9 and CEA after GBC resection were independent prognosticators of poor survival. Combined with postoperative pathology, which is the most important prognosticator, they classified postsurgical patients into different prognostic groups, which aided further decision making. Unlike patients with resectable GBC, patients with unresectable GBC undergo palliative chemotherapy and exhibit variable tumor responses, a valid assessment of such responses is critical in further decision making. The present study focused on the relationship between chemotherapy response and tumor marker changes, and identified criteria for early imaging follow-up or chemotherapy discontinuation based on these changes.

We excluded subjects with normal pre- and postchemotherapy tumor markers in this study and this may create some degree of selection bias. However, excluding such subjects does not facilitate any favorable tendency in the conclusion of this study. On the other hand, we could not conclude some issues regarding chemotherapy response due to the insufficient number of subjects. Despite tumor marker kinetics significantly correlated with progressive disease, we could not draw a conclusion correlating tumor marker kinetics with partial or complete response. Only 16 patients showed partial response to chemotherapy and there were no patients with complete response. Studies with a larger number of subjects are warranted to reach a conclusion.

High baseline serum total bilirubin is correlated with high baseline serum CA 19-9 level, and could therefore influence CA 19-9change.22 There is no universal agreement on how to adjust the serum CA 19-9 level. Kim et al.23 suggested a formula correcting the effect of serum total bilirubin on serum CA 19-9 level, which divides serum CA 19-9 level with serum total bilirubin level. However, the serum total bilirubin threshold 2.0 mg/dL makes this formula inadequate to this study due to the measurable nature of tumor marker changes. Instead, we did a subgroup analysis, and 29 of 106 patients had a serum total bilirubin level >3.0 mg/dL (Table 6). High baseline serum bilirubin (>3.0 mg/dL) tended to be associated with improved survival in patients with a lower CA 19-9change, although the association did not reach significance, most likely because of the small number of patients. Therefore, the influence of combined total bilirubin and serum CA 19-9 changes should be evaluated in further studies with a larger number of patients.

In conclusion, CA 19-9 kinetics is a valuable prognosticator of patient survival and treatment response during chemotherapy for unresectable GBC.