Circulating tumor cells as a surrogate marker for determining response to chemotherapy in patients with advanced gastric cancer

Abstract

The purpose of this study was to quantify circulating tumor cells (CTCs) in advanced gastric cancer (AGC) patients, and to demonstrate the role of CTCs in cancer therapy. This study investigates the hypothesis that CTCs can predict clinical outcomes in patients with AGC. From November 2007 to June 2009, 52 patients with AGC were enrolled into a prospective study. The chemotherapy regimen was an S‐1‐based regimen (S‐1 with or without cisplatin) or paclitaxel. CTCs of whole blood at baseline, 2 weeks, and 4 weeks after initiation of chemotherapy, were isolated and enumerated using immunomagnetics. Patients with ≥4 CTCs at 2‐week points and 4‐week points had a shorter median progression‐free survival (PFS) (1.4, 1.4 months, respectively) than those with the median PFS of <4 CTCs (4.9, 5.0 months, respectively) (log‐rank test; P < 0.001, P < 0.001, respectively). Patients with ≥4 CTCs at 2‐week points and 4‐week points had shorter median overall survival (OS) (3.5, 4.0 months, respectively) than those with the median PFS of <4 CTCs (11.7, 11.4 months, respectively) (log‐rank test; P < 0.001, P = 0.001, respectively). In conclusion, this study demonstrates that CTC measurement may be useful as a surrogate marker for determining response to S‐1‐based or paclitaxel regimens in AGC.

(Cancer Sci 2010; 101: 1067–1071)

Gastric cancer is more prevalent in Asia, Eastern Europe, and Central and South America than in other areas. In Japan, this cancer is one of the most common causes of cancer‐related mortality, despite dramatic advances in diagnosis and treatment. Outcomes are extremely poor in patients with unresectable gastric cancer, with the median survival ranging from 3 to 5 months with the best supportive care.^{( 1} , ² , ^{3 )} The ability to identify patients with the worst prognoses or those destined to progress quickly could have broad clinical applications.

Circulating tumor cells (CTCs) or disseminated tumor cells (DTCs) in bone marrow and peripheral blood from patients with cancers have been documented.^{( 4} , ⁵ , ^{6 )} Braun et al. ^{( 7} , ^{8 )} reported that ∼30% of women with primary breast cancer have DTCs in bone marrow, and a 10‐year follow‐up of these patients revealed a significantly decreased disease‐free survival and overall survival (OS) when compared with patients without DTCs. However, aspiration of bone marrow is time consuming and, in many cases, uncomfortable for the patients precluding multiple samplings for therapy monitoring studies. Therefore, recent efforts have concentrated on the detection of CTCs in the peripheral blood of cancer patients. Cristofanilli et al. ^{( 9} , ^{10 )} showed in a prospective study that CTC detection provided significant prognostic information for patients with metastatic breast cancer. Cohen et al. ^{( 11 )} showed that the number of CTCs before and during treatment was an independent predictor of PFS and OS in patients with metastatic colorectal cancer. It is not clear whether CTC detection using this system provides prognostic information for patients with advanced gastric cancer. We initiated this study to evaluate whether CTCs could serve as a prognostic and/or predictive marker in patients with AGC.

Materials and Methods

Patients.  All patients were enrolled using institutional review board‐approved protocols at the Cancer Institute Hospital at the Japanese Foundation for Cancer Research and provided informed consent. The study population consisted of patients aged 18 years or older with histologically proven AGC. Other inclusion criteria were Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 2; adequate organ function; and S‐1‐based (S‐1 with or without cisplatin) or paclitaxel chemotherapy regimen. The subjects were five patients treated with S‐1 (40 mg/m², twice daily, days 1–28, repeated every 6 weeks), 26 patients treated with S‐1 plus CDDP (S‐1 40 mg/m², twice daily, days 1–21, CDDP 60 mg/m², day 8, repeated every 5 weeks), and 21 patients treated with paclitaxel (80 mg/m², weekly).

Sample preparation for isolation of CTCs from blood.  Blood was drawn from advanced gastric cancer patients into 10 mL of evacuated blood for CTC in a Cell Save Preservative Tube (Veridex, Raritan, NJ, USA). Blood was always drawn from cancer patients before treatment initiation (baseline), 2 weeks, and 4 weeks after the administration of an S‐1‐based or paclitaxel regimen. The CellSearch system (Veridex) consists of the CellPrep system, the CellSearch Epithelial Cell Kit (for the measurement of CTC), and the CellSpotter Analyzer. The CellPrep system is a semi‐automated sample preparation system, and the CellSearch Epithelial Cell Kit consists of ferrofluids coated with epithelial cell‐specific EpCAM antibodies to immunomagnetically enrich epithelial cells; a mixture of two phycoerythrin‐conjugated antibodies that bind to cytokeratin 8, 18, and 19; an antibody to CD45 conjugated to allophycocyanin; nuclear dye 4′,6‐ diamidino‐2‐phenylindole (DAPI) to fluorescently label the cell; and buffers to wash, permeabilize, and resuspend the cells. Sample processing and evaluation were done as described by Allard et al. Briefly, 7.5 mL of blood for CTCs were mixed with 6 mL of buffer, centrifuged at 800 g for 10 min, and then placed on the CellPrep system. After aspiration of the plasma and buffer layer by instrument, ferrofluids were added. After incubation and subsequent magnetic separation, unbound cells and remaining plasma were aspirated. The staining reagents were then added in conjunction with a permeabilization buffer to fluorescently label the immunomagnetically labeled cells. After incubation in the system, the magnetic separation was repeated, and excess staining reagents were aspirated. In the final processing step, the cells were resuspended in the MagNest Cell Presentation Device (Veridex). This device consists of a chamber and two magnets that orient the immunomagnetically labeled cells for analysis using the CellSpotter Analyzer.

Sample analysis.  The MagNest was placed on the CellSpotter Analyzer, a four‐color semi‐automated fluorescence microscope. Image frames covering the entire surface of the cartridge for each of the four fluorescent filter cubes were captured. The captured images containing objects that met predetermined criteria were automatically presented in a web‐enabled browser from which final selection of cells was made by the operator. The criteria for an object to be defined as a CTC include round to oval morphology, a visible nucleus (DAPI positive), positive staining for cytokeratin, and negative staining for CD45. Results of cell enumeration are always expressed as the number of cells per 7.5 mL of blood for CTCs.

Statistical analysis.  Progression‐free survival (PFS) was defined as the elapsed time from blood collection to progression. Kaplan–Meier survival plots were generated based on CTC levels each time blood was collected, and the curves were compared using a log‐rank testing. A P‐value <0.05 was considered significant. Cox proportional hazards regression was used to determine univariate and multivariate hazard ratios for selected potential predictors of PFS and OS. The distribution of patients above and below the CTC threshold and clinical response was compared using Fisher’s exact test.

Results

Patient characteristics.  A total of 52 patients were enrolled. Patients’ characteristics at baseline are summarized in Table 1. Patients’ characteristics were as follows: median age, 62 years (range, 24–78 years); PS 0/1/2, 39/12/1; primary tumor +/−, 33/19; and regimen S‐1/S‐1 with cisplatin/paclitaxel, 5/26/21. Thirty‐five patients had diffuse‐type histology (67.3%). Seventeen patients (32.7%) had intestinal type. Among 52 patients, the best response rates were 28.8% (complete response [CR]/partial response [PR]/stable disease [SD]/progressive disease [PD]: 0/15/19/18). Of 31 patients treated with the S‐1‐based regimen (S‐1 alone or S‐1/cisplatin [CDDP]) assessable for response, we observed 14 PR (45.2%), 11 patients (35.5%) with SD, and six patients (19.4%) with PD during treatment. The overall response rate was 45.2%. On the other hand, of 21 patients treated with the weekly paclitaxel regimen assessable for response, we observed one PR (4.8%), eight patients (38.1%) with SD, and 12 patients (57.1%) with progression of disease during treatment, for an overall response rate (RR) of 4.8% (Table 2).

Table 1.

 Patient demographics

Demographic	Number or median (range)
Median age (range)	62 (24–78)
Male/female	44/8
PS: 0/1/2	39/12/1
S1‐based/PAC regimen	31/21
Line: 1st/2nd	34/18
Histopathology: diffuse/intestinal type	35/17
Primary tumor: +/−	33/19
Sites of metastasis: +/−
Liver	24/28
Lung	3/49
Bone	1/51
Peritoneum	22/30
Lymph node	37/15

Table 2.

 Objective response

	S1‐based regimen (31) S1 alone (5), S1/CDDP (26) 1st line (31)	PAC (21) Weekly PAC (21) 1st line (3), 2nd line (18)
CR	0	0
PR	14	1
SD	11	8
PD	6	12

CDDP, cisplatin; CR, complete response; PAC, paclitaxel; PD, progressive disease; PR, partial response; SD, stable disease.

Stratification according to CTC levels.  To select a level of circulating tumor cells that most clearly distinguished patients with a response of chemotherapy, thresholds of 1 to 88 cells for 2‐week point were systematically correlated with PFS for 26 of the 30 patients in the training set. The median PFS among patients with levels above or below each threshold differed at the level of one circulating tumor cell per 7.5 mL of blood, and reached a plateau at approximately four cells per 7.5 mL of blood. At the latter level, the Cox proportional hazards ratio signifying the difference between slow and rapid progression of disease also reached a plateau. Thus, a cut‐off of four circulating tumor cells per 7.5 mL of blood was chosen to distinguish patients.^{( 12 )} The Kaplan–Meier circulating tumor‐cell counts were available at a 2‐week point for 26 of the thirty patients in the training set and for 21 of the 22 patients in the validation set. Neither PFS nor OS was significantly different in the two sets (data not shown). Because the two sets of data were nearly identical, they were combined for the estimation of PFS and OS for the entire population.

CTCs and imaging to assess response to therapy.  Thirty‐four (65.4%) of 52 patients were classified as having non‐progressive disease (non‐PD), with 24 of these patients (46.2%) having <4 CTCs and 10 patients (19.2%) having ≥4 CTCs before the initiation of therapy. Ten (19.2%) of 52 patients were classified as having PD, with 11 of these patients (21.2%) having <4 CTCs and seven patients (13.4%) having ≥4 CTCs before the initiation of therapy. The difference between the clinical responses and CTC levels were not significant. In contrast, 33 (64.7%) of 51 patients were classified as having non‐PD, with 33 of these patients (64.7%) having <4 CTCs and no patients (0%) having ≥4 CTCs at 2 weeks. Eighteen (35.3%) of 51 patients were classified as having PD, with 11 of these patients (21.6%) having <4 CTCs and seven patients (13.7%) having ≥4 CTCs at 2 weeks. The difference between the clinical responses and CTC levels was highly significant. (P = 0.001, Fisher’s exact test). Thirty‐two (64%) of 48 patients were classified as having non‐PD, with 31 of these patients (64.6%) having <4 CTCs and one patient (2.0%) having ≥4 CTCs at 4 weeks. Sixteen (33.3%) of 48 patients were classified as having PD, with eight of these patients (16.7%) having <4 CTCs and eight patients (16.7%) having ≥4 CTCs at 4 weeks. The difference between the clinical responses and CTC levels were highly significant (P < 0.001, Fisher’s exact test) (Table 3).

Table 3.

 CTCs and correlation with response assessment by imaging

	Non‐PD			PD			Fisher’s exact P‐values
	No. of patients	CTCs <4 (%)	CTCs ≥4 (%)	No. of patients	CTCs <4 (%)	CTCs ≥4 (%)
Baseline	34	24 (46.2)	10 (19.2)	18	11 (21.2)	7 (13.4)	0.544
2 week	33	33 (64.7)	0 (0)	18	11 (21.6)	7 (13.7)	0.001
4 week	32	31 (64.6)	1 (2.0)	16	8 (16.7)	8 (16.7)	<0.001

CTCs, circulating tumor cells; PD, progressive disease.

Analysis of PFS according to CTC level.  Figure 1 shows the Kaplan–Meier plots for prediction of PFS using the baseline CTC counts (Fig. 1a), at 2 weeks (Fig. 1b), and at 4 weeks (Fig. 1c). Seventeen of the patients (32.7%) had ≥4 CTCs per 7.5 mL of blood at baseline. These patients had no significantly different PFS compared with that of patients with <4 CTCs per 7.5 mL of blood at baseline. Patients with ≥4 CTCs at the 2‐week point had a shorter median PFS (1.4 months; 95% confidence interval [CI], 1.2–1.6) than the median PFS of <4 CTCs at 2 weeks (4.9 months; 95% CI, 4.0–5.8) (P < 0.001) (Fig. 1b). Patients with ≥4 CTCs at the 4‐week point had a shorter median PFS (1.4 months; 95% CI, 1.3–1.5) than the median PFS of <4 CTCs at 4 weeks (5.0 months; 95% CI, 3.9–6.1) (P < 0.001) (Fig. 1c). With the S‐1‐based regimen, 10 patients had ≥4 CTCs per 7.5 mL of blood at baseline. These patients had no significantly different PFS compared with 21 patients with <4 CTCs per 7.5 mL of blood at baseline. Patients with ≥4 CTCs at the 2‐week point had a shorter median PFS (1.2 months) than the median PFS of <4 CTCs at 2 weeks (6.0 months; 95% CI, 4.3–7.7) (P < 0.001). Patients with ≥4 CTCs at the 4‐week point had a shorter median PFS (2.3 months; 95% CI, 0.7–3.9) than the median PFS of <4 CTCs at 4 weeks (6.3 months; 95% CI, 3.0–9.7) (P < 0.001). With the paclitaxel regimen, seven patients had ≥4 CTCs per 7.5 mL of blood at baseline. These patients had no significantly different PFS compared with 14 patients with <4 CTCs per 7.5 mL of blood at baseline. Patients with ≥4 CTCs at the 2‐week point had a shorter median PFS (1.4 months; 95% CI, 1.4–1.5) than the median PFS of <4 CTCs at 2 weeks (4.3 months; 95% CI, 3.5–5.2) (P < 0.001). Patients with ≥4 CTCs at the 4‐week point had a shorter median PFS (1.4 months; 95% CI, 1.0–1.8) than the median PFS of <4 CTCs at 4 weeks (4.4 months; 95% CI, 3.6–5.3) (P < 0.001).

Figure 1.

 Kaplan–Meier plots of progression‐free survival (PFS) in advanced gastric cancer patients with less than four circulating tumor cells (CTCs) or ≥4 CTCs at baseline (a), 2 weeks (b), and 4 weeks (c).

Analysis of OS according to CTC level.  Figure 2 shows the Kaplan–Meier plots for prediction of OS using baseline CTC counts (Fig. 2a), at 2 weeks (Fig. 2b), and at 4 weeks (Fig. 2c). Seventeen of the patients (32.7%) with ≥4 CTCs per 7.5 mL of blood at baseline had no significant different OS compared with patients with <4 CTCs per 7.5 mL of blood at baseline. Patients with ≥4 CTCs at the 2‐week point had a shorter median OS (3.5 months; 95% CI, 3.0–4.0) than the median OS of <4 CTCs at 2 weeks (11.7 months; 95% CI, 5.3–18.2) (P < 0.001) (Fig. 2b). Patients with ≥4 CTCs at the 4‐week point had a shorter median OS (4.0 months; 95% CI, 2.6–5.5) than the median OS of <4 CTCs at 4 weeks (11.4 months; 95% CI, 3.3–19.5) (P = 0.001) (Fig. 2c). With the S‐1 based regimen, 10 patients had ≥4 CTCs per 7.5 mL of blood at baseline. These patients had no significant different OS compared with 21 patients with <4 CTCs per 7.5 mL of blood at baseline. Patients with ≥4 CTCs at the 2‐week point had a shorter median OS (1.3 months) than the median OS of <4 CTCs at 2 weeks (13.8 months; 95% CI, 9.4–18.2) (P < 0.001). Patients with ≥4 CTCs at the 4‐week point had a shorter median OS (4.0 months; 95% CI, 2.3–5.7) than the median OS of <4 CTCs at 4 weeks (>11.7 months) (P = 0.031). With the paclitaxel regimen, seven patients had ≥4 CTCs per 7.5 mL of blood at baseline. These patients had no significant different OS compared with 14 patients with <4 CTCs per 7.5 mL of blood at baseline. Patients with ≥4 CTCs at the 2‐week point had a shorter median OS (3.5 months; 95% CI, 3.1–4.0) than the median OS of <4 CTCs at 2 weeks (6.5 months; 95% CI, 5.9–7.2) (P < 0.001). Patients with ≥4 CTCs at the 4‐week point had a shorter median OS (3.5 months; 95% CI, 2.3–4.7) than the median OS of <4 CTCs at 4 weeks (6.5 months; 95% CI, 5.5–7.5) (P = 0.013).

Figure 2.

 Kaplan–Meier plots of overall survival (OS) in advanced gastric cancer patients with less than four circulating tumor cells (CTCs) or ≥4 CTCs at baseline (a), 2 weeks (b), and 4 weeks (c).

Univariate and multivariate analysis of predictors of PFS and OS.  Univariate and multivariate Cox proportional hazards regression was performed to assess the association between factors of interest and PFS or OS. In univariate analysis, PS, treatment regimen, line of chemotherapy, and CTC levels (cut‐off, 4) at 2 and 4 weeks predicted PFS and OS (Table 4). In order to evaluate the independent predictive effect of chemotherapy, multivariate Cox regression analysis was carried out (Table 5). CTC levels at 2 and 4 weeks were the strongest predictors.

Table 4.

 Univariate Cox regression analysis of independent parameters for prediction of PFS and OS

Parameter	No. of patients	PFS				OS
		HR	95% CI	P‐values	χ2	HR	95% CI	P‐values	χ2
ECOG, 2 vs 1 vs 0	52	1.817	1.010–3.268	0.046	0.042	2.795	1.416–5.516	0.003	0.002
Treatment regimen	52	0.422	0.225–0.792	0.007	0.006	0.239	0.106–0.538	0.001	<0.001
Line of therapy	52	3.155	1.577–6.311	0.001	0.001	4.527	2.031–10.088	<0.001	<0.001
CTCs at the 2nd week	51	22.633	6.214–82.429	<0.001	<0.001	42.796	8.382–218.515	<0.001	<0.001
CTCs at the 4th week	48	15.947	5.380–47.271	<0.001	<0.001	4.699	1.751–12.609	0.002	0.001

CI, confidence interval; CTCs, circulating tumor cells; ECOG, Eastern Cooperative Oncology Group; HR, hazard ratio; OS, overall survival; PFS, progression‐free survival.

Table 5.

 Multivariate Cox regression analysis for prediction of PFS and OS

Parameter	PFS				OS
	No. of patients	HR	95% CI	P‐values	No. of patients	HR	95% CI	P‐values
No. of patients	51				51
Line of therapy, 1st vs 2nd		0.463	0.219–0.977	0.043		0.307	0.129–0.731	0.008
Lymph node metastasis		0.458	0.214–0.980	0.044
CTCs at the 2nd week		0.049	0.012–0.199	<0.001		0.037	0.007–0.191	<0.001
Model χ2			<0.001				<0.001
No. of patients	48				48
Line of therapy, 1st vs 2nd		0.412	0.192–0.880	0.022		0.217	0.089–0.504	<0.001
CTCs at the 4th week		0.082	0.027–0.224	<0.001		0.216	0.077–0.607	0.004
Model χ2			<0.001				<0.001

CI, confidence interval; CTCs, circulating tumor cells; HR, hazard ratio; OS, overall survival; PFS, progression‐free survival.

Discussion

The CellSearch system is designed to enrich and enumerate CTCs from peripheral blood. Furthermore, it is the first system to validate the clinical use of CTCs in patients with advanced gastric cancer. Our results show that the system is a suitable tool for assessment of CTCs in these patients, enabling reliable detection of CTCs in whole blood.

Approaches for isolation of CTCs in a research setting range from enrichment of tumor cells using density‐gradient centrifugation^{( 13} , ¹⁴ , ^{15 )} and flow cytometry.^{( 16} , ^{17 )} CTC number as quantified by the CellSearch methodology^{( 18} , ¹⁹ , ²⁰ , ^{21 )} has been shown to have prognostic significance, and post‐therapy decreases and increases in CTC number are associated with a superior and inferior survival, respectively, in patients with breast cancer, prostate cancer, and colorectal cancer. In this study, a finding of <4 CTCs in 7.5 mL of peripheral blood at 2 and 4 weeks after initiation of chemotherapy was associated with significantly longer PFS and OS as compared with these patients with ≥4 CTCs in 7.5 mL of peripheral blood. The results of this analysis demonstrated that the presence of four or more CTCs in 7.5 mL of blood before initiation of chemotherapy is not associated with PFS and OS. But the levels of CTCs at 2 and 4 weeks after initiation of chemotherapy are predictive of treatment efficacy, PFS, and OS. The presence of at least four CTCs at 2 and 4 weeks is a strong independent prognostic factor for inferior PFS and OS. These data demonstrate that CTC measurement may be a useful biomarker for monitoring response to therapy in AGC.

Outcomes are extremely poor in patients with ≥4 CTCs at 2 and 4 weeks, with the median OS ranging from 2 to 5 months. These data suggest the value of this technology in the identification of chemotherapy‐resistant patients who could benefit from early treatment change and/or more investigational. Further study should prospectively address whether a change of treatment based on ≥4 CTCs at 2 or 4 weeks after initiation of chemotherapy early in the course of treatment will result in improvement in OS. CTC levels drawn at 2 and 4 weeks, before typical imaging intervals, may have the potential to suggest treatment choices and spare unnecessary toxicity by suggesting that an early change in therapy is warranted. Because the CellSearch system has not been approved in Japan, the price of one sample costs about ¥80 000 as in the case of the extra laboratory in the clinical trial. Several prospective trials led to the FDA approval of CTC counts for monitoring of patients with breast, colorectal, and prostate cancer. We expect CTC counts for monitoring of patients with gastric, breast, colorectal, and prostate cancer to be approved in Japan.

In conclusion, this study demonstrates the independent predictive value of CTCs for patients initiating chemotherapy for AGC. The data obtained in this clinical trial of the CellSearch system were for enumeration of CTCs in AGC. Our study was not designed to assess whether a change in therapy based on ≥4 CTCs is beneficial. However, clinical trials to explore this hypothesis are warranted.

Acknowledgments

This work was supported by Taiho Pharmaceutical Co., Ltd. The excellent technical assistance of Dr Yoshimasa Kawazoe, Dr Koichi Takagi, Sayuri Minowa, Harumi Shibata, and Mariko Kimura is greatly appreciated.

Disclosure Statement

The authors have no conflict of interest.