Detection and Prognostic Significance of Circulating Tumor Cells in Patients With Metastatic Thyroid Cancer

Abstract

Context:

Individual patient prognostication for advanced thyroid cancer (TC) is challenging. Circulating tumor cells (CTCs) have been shown to be a valuable prognostic marker for other solid cancers.

Objective:

We hypothesized that CTCs are present in the blood of patients with advanced TC and their number can predict overall survival (OS).

Setting:

This is a prospective study at a tertiary cancer hospital.

Patients, Interventions, and Main Outcome Measures:

Initial studies were performed with TC cell lines to determine the feasibility of detection using the Veridex CellSearch. CTC enumeration was performed in blood samples from 18 patients with distantly metastatic medullary TC (metMTC), 14 with distantly metastatic differentiated TC (metDTC), and 10 controls with a history of TC but no evidence of disease. The prognostic value of CTC levels to predict OS in metMTC patients was assessed.

Results:

CellSearch detected cells from MTC and DTC but not anaplastic TC cell lines. Six metMTC patients but no metDTC or control patients had more than or equal to 5 CTCs detected by the CellSearch assay. Median survival in metMTC patients with more than or equal to 5 CTCs was 13 months vs 51.5 months for those with less than 5 CTCs (P = .0116). The hazard ratio for mortality of patients with more than or equal to 5 CTCs compared with those with less than 5 CTCs was 3.95 (1.20–13.0, P = .0245).

Conclusions:

The presence of more than or equal to 5 CTCs in patients with metMTC is associated with worse OS. Larger cohorts are required to validate the prognostic value of CTC enumeration.

Circulating tumor cells were detected in blood from patients with metastatic medullary thyroid carcinoma, and were prognostic of increased risk for mortality.

The prognosis for most patients diagnosed with either differentiated thyroid carcinoma (DTC) or medullary TC (MTC) is excellent and can be readily predicted by clinicopathologic staging characteristics determined at initial presentation (1). Identification of distant metastases may signify patients at higher risk for disease-related mortality, but considerable heterogeneity exists; patients with a significant burden of metastatic disease may remain stable for many years, even in the absence of systemic therapy. Escalating concentrations of serum biochemical tumor markers, thyroglobulin for DTC and calcitonin (Ctn) and carcinoembryonic antigen (CEA) for MTC, can predict disease progression and mortality, especially when doubling times are shorter than 6 months (2, 3), but serial measurement over several years is required for optimal assessment. Further, advanced TCs, including anaplastic TC (ATC), metastatic DTC (metDTC), and metastatic MTC (metMTC) may not produce these markers, reflecting the poor differentiation of the malignancy. Additionally, studies of novel targeted therapies for metastatic disease generally do not identify changes in these biomarkers as strongly predictive of cancer response to treatment, particularly when drug therapy directly alters transcription of the biomarker's gene (4). Thus, there remains a critical need to develop new prognostic and predictive biomarkers to inform clinical management.

The presence of tumor cells in the circulation has long been presupposed to be essential to development of distant metastases (5–7). Detection and enumeration of circulating tumor cells (CTCs) have been shown to be prognostic and predictive biomarkers in multiple solid cancers (8–10). Additionally, methods to isolate and molecularly analyze CTCs can permit genetic characterization of cancer cells, ie, a “liquid biopsy,” in patients with solid tumors. Of various methods that have been introduced, the CellSearch System (Veridex LLC) for enumeration of CTCs, an assay based on the expression of epithelial-cell adhesion molecule (EpCAM), has been rigorously validated and is cleared by the Food and Drug Administration as a prognostic biomarker for patients with metastatic breast, prostate, and colorectal carcinomas (8–11). We hypothesized that CellSearch can identify CTCs from patients with advanced TC and that the detection of CTCs could lead to development of new prognostic and predictive biomarkers for these malignancies.

Patients and Methods

Immunohistochemical staining of EpCAM

EpCAM expression was previously reported on DTC but not on ATC tumor tissues (12, 13). However, no report in MTC was found. Therefore the expression of EpCAM on MTC tumor was examined. Slides were obtained from paraffin blocks of MTC pathologic samples. Immunohistochemical staining was performed using standard protocol. Briefly, slides were treated with 3% hydrogen peroxide, and incubated with 1:200 primary antibody to EpCAM (VU-1D9; Santa Cruz Biotechnology) at 4°C overnight. Slides were then incubated with peroxidase-conjugated secondary antibody (Santa Cruz Biotechnology). Daiminobenzidine chromogen (Santa Cruz Biotechnology) was used for visualization.

Recovery of TC cell lines experiments

To ascertain whether CellSearch can identify TC cells, recovery experiments were performed using the ATC cell line HT-101, the papillary TC (PTC) cell line TPC-1, and the MTC cell lines TT and MZ-CRC-1. Ten cultured cells were microscopically drawn into a micropipette and added to 7.5-mL tubes of blood obtained from a healthy male. The spiked blood samples were then individually subjected to the CellSearch analysis.

Patients

From the Multidisciplinary Endocrine Center at MD Anderson Cancer Center, we recruited 18 subjects with metMTC, 14 with metDTC (10 PTC, 3 Hürthle cell TC, and 1 follicular TC), and 10 with history of TC (9 DTC and 1 MTC) and no biochemical or radiographic evidence of disease (NED) for more than or equal to 5 years. We excluded patients with other active malignancies. All metMTC and metDTC patients had extracervical distant metastases and a radiographically measurable tumor burden. All patients gave their written informed consent for the study. The protocol and consent were approved by the Institutional Review Board.

Baseline clinical data at the time of sample collection were collected and included age, gender, age at initial diagnosis, systemic therapy, number of metastatic organs level of tumor markers (thyroglobulin for DTC, Ctn, and CEA for MTC), Ctn and CEA doubling times based on at least 4 serial levels, and information regarding somatic and germline RET mutation status for metMTC patients. Because CTCs were only identified in the metMTC group, these patients were prospectively monitored with date of last follow-up or death recorded.

Enumeration of CTCs by CellSearch

From each patient, 2 7.5-mL CellSave tubes (Veridex LLC) of blood were collected, maintained at room temperature, and processed promptly. The CellSearch system was used to isolate CTCs according to the manufacturer's instructions. Briefly, epithelial cells were immunomagnetically enriched with EpCAM antibodies, followed by immunohistochemical characterization and analysis using the CellSpotter Analyzer, a 4-color semiautomated fluorescence microscope. Cells were defined as CTCs if fulfilling the criteria: round to oval morphology, a visible nucleus positively staining for 4,2-diamidino-2-phenylindole dihydrochloride, positive staining for epithelial marker cytokeratin, and negative for the lymphocyte marker CD45. The technical details including accuracy, precision, linearity, and reproducibility have been previously described (7). Patient CTC number was based on the mean obtained from the 2 tubes.

Statistical analysis

All data were expressed as mean ± SD and median (range). Statistical analysis was performed using software JMP version 11 (SAS Institute). Variables that were positively skewed were log transformed before analysis. Comparison of continuous data was done using an unpaired t test or nonparametric k-sample test, and χ² or Fisher's exact tests were used for categorical data. The sensitivity and specificity of CTC cutoff were determined with receiver operator curve analysis. Correlation analysis was done using Pearson or Spearman analysis. Overall survival (OS) was defined as the elapsed time from blood collection to death. Patients were censored at last follow-up if death had not occurred. Kaplan-Meier survival curves were used for survival analysis, and log-rank and Wilcoxon tests were employed to detect differences between groups. Proportional hazards test was performed for multivariate regression. P < .05 was considered significant.

Results

TC cell recovery study

Previous studies had identified membranous and cytoplasmic staining of EpCAM on DTC cells, and we extended this observation in primary MTC tumor specimens (Figure 1). To determine whether EpCAM-based detection with CellSearch could enumerate TC cells, we tested the assay by spiking normal blood with established TC cell lines. These experiments established a 100% recovery rate using the CellSearch assay for blood spiked with the PTC cell line TPC-1 and the MTC cell lines TT and MZ-CRC-1 (data not shown). No tumor cells were identified in the blood spiked with the anaplastic carcinoma cell line HT-101, or in control specimens without cells spiked. For this reason, ATC patients were excluded from subsequent study.

Figure 1.

Representative finding of immunohistochemical staining of epithelial-cell adhesion molecule in medullary thyroid carcinoma (MTC) tumor tissues. Membranous and cytoplasmic staining of MTC tumor cells.

Patient characteristics

Clinical characteristics of enrolled subjects are summarized in Table 1. The mean age at diagnosis and at time of sampling were 42.5 ± 14.1 and 51.9 ± 11.2 years for metMTC patients, and 51.4 ± 9.6 and 59.4 ± 10.4 years for metDTC patients, respectively. All sites of distant metastases for metMTC and metDTC patients included lung, bone, liver, and kidney. All of the metDTC patients and 13/18 metMTC patients had experienced progression of cancer within the year before study as determined by retrospective conventional imaging assessment. Eight of the metMTC patients and 9 of the metDTC patients were on tyrosine kinase inhibitors at the time of sampling, and 6 of metMTC and 2 of metPTC patients had previously received systemic therapy. None of patients had systemic therapy stopped within one month before sampling. All metDTC patients had underwent previous radioactive iodine treatment more than 1 year before sampling.

Table 1.

Demographic and Clinical Characteristics of Patients With metMTC, metDTC or NED

	metMTC	metDTC	NED
n (M/F)	18 (9/9)	14 (7/7)	10 (9/1)
Age at diagnosis (y)	42.5 ± 14.1	51.4 ± 9.6	41.1 ± 15.7
Age at study entry (y)	51.9 ± 11.2	59.4 ± 10.4	52.2 ± 15.4
Current systemic therapy n (%)	8 (44.4)	9 (64.3)	—
Previous systemic therapy n (%)	6 (33.3)	2 (14.3)	—
Metastasis n (cervical/extracervical)	12/18	3/14	—

metMTC, distantly metMTC; metDTC, distantly metDTC; NED, NED patients for more than or equal to 5 years.

Detection of CTCs in metMTC and metDTC patients

Data duplicates were available for 38/42 subjects. The correlation between the CTC counts obtained from both tubes showed a correlation coefficient R² = 0.964, suggesting good intraassay agreement. Using the conventional cutoff of more than or equal to 5 CTCs/7.5 mL that has been used to define elevated levels for other solid malignancies (11, 14, 15), sensitivity and specificity to distinguish metastatic disease patients from NED controls were 20% and 100%, respectively. However, by receiver operator curve analysis, the optimal cutoff to distinguish metastatic disease patients from NED controls was more than or equal to 1 CTC/7.5 mL, with sensitivity and specificity of 41% and 90%, respectively (area under curve 0.69, P = .0154). Thirteen of 18 metMTC patients had detectable CTCs (range 1–49 CTCs/7.5 mL), of whom 6/18 had CTCs more than or equal to 5 and 5/18 had 0 CTCs. One metDTC patient, who had an aggressive tall cell variant PTC and lung metastases, had 3 CTCs/7.5 mL. All of the other metDTC patients and NED controls had less than 1 CTCs/7.5mL, except one from each group had 1 CTCs/7.5 mL (Figure 2). Clinical data and CTC levels of metDTC patients were tabulated in Supplemental Table 1. Because only 2 metDTC patients had more than or equal to 1 CTCs, no further analysis was done of outcomes for DTC patients as a function of CTC status.

Figure 2.

CTCs levels in metMTC and metDTC patients and NED controls.

Baseline characteristics and CTC levels of metMTC patients

The CTC numbers and baseline clinical and known mutational information at the time of blood collection for the 18 metMTC patients is provided in Table 2. The mean Ctn and CEA doubling times, calculated at the time of blood CTC sampling, were 24.3 ± 48.3 and 24.4 ± 53.7 months, respectively. The mean number of metastatic sites was 3 ± 1. Results from germline RET gene analysis was available in all patients, 6 of whom were found to carry a germline RET mutation. Nine patients had undergone tumor somatic mutation testing for clinical purposes to identify RET and RAS “hotspot” mutations, with a RET mutation identified in 3, a KRAS mutation reported in 2, and no mutations found in the remaining 4 patients.

Table 2.

Baseline Characteristics and CTCs Levels of metMTC Patients

Patient Number	Age at Diagnosis (y)	Ctn DT (m)	CEA DT (m)	Number of Metastatic Sites	Current Therapy	Germline RET Mutation	Somatic Mutation	CTC Number per 7.5 mL
1	20.4	2.7	38.69	4	N	C634R	N/A	8.0
2	39.3	154.9	−18.65	2	Lenvatinib	WT	WT	4.0
3	19.3	145.8	100.24	4	N	C634Y	N/A	7.5
4	27.1	25.6	55.6	2	N	WT	N/A	0.5
5	36.0	7.9	14.27	3	Lenvatinib	WT	N/A	1.0
6	31.1	8.1	30.04	3	N	C634R	WT	7.5
7	35.0	3.1	2.82	3	Lenvatinib	WT	RET del629-31	11.5
8	36.0	12.0	40.81	4	N	C618G	N/A	0
9	52.1	28.4	−11.09	4	N	WT	WT	0
10	53.0	−20.2	−44.58	4	Sunitinib/valproic acid	WT	KRAS G12V	0.5
11	55.2	6.9	8.7	1	N	WT	N/A	1.5
12	60.0	2.6	4.11	5	N	WT	RET M918T	38.0
13	66.5	4.8	7.52	3	N	WT	RET M918T	0
14	42.3	1.7	2.75	4	Lenvatinib	WT	N/A	0
15	63.5	−20.1	11.08	2	Sunitinib	WT	KRAS A146V	1.0
16	40.3	41.7	198.7	3	Sunitinib	V804M	N/A	0
17	35.9	21.3	−5.45	2	Sunitinib	C618S	WT	1.0
18	52.8	9.5	3.89	4	N	WT	N/A	49.0

Ctn DT, Ctn doubling time; CEA DT, CEA doubling time.

The log-transformed mean number of CTCs/7.5 mL in metMTC patients was 7.3 ± 13.8. No significant differences were observed for baseline clinic parameters between the groups with and without elevated CTCs, including age at diagnosis, Ctn and CEA doubling times, number of metastatic sites, current or history of systemic therapy, and presence of a germline RET mutation (Table 3).

Table 3.

Comparison of Baseline Parameters Between Patients With <5 CTCs/7.5 mL and ≥5 CTCs/7.5 mL

	<5 CTCs/7.5 mL	≥5 CTCs/7.5 mL	P Value
N	12	6
Gender (M/F)	7/5	2/4	.62
Age at diagnosis, years	45.6 ± 12.2	36.4 ± 16.8	.20
Positive germline RET mutation, %	25	50	.34
Age at sampling, years	54.2 ± 11.9	47.5 ± 9.0	.25
Log-mean Ctn, pg/mL	2182	12 633	.18
Ctn doubling time, months	22.1 ± 45.6	28.6 ± 57.5	.80
CEA doubling time, months	21.6 ± 61.5	30.0 ± 37.7	.77
Number of metastatic sites	2.8 ± 1.0	3.8 ± 0.8	.052
Tyrosine kinase inhibitor, %
Concurrent	58%	17%	.15
At any time	75%	83%	.99

CTCs level, biomarker doubling times, and OS in metMTC patients

To evaluate the prognostic significance of elevated CTCs in metMTC patients, the duration of survival following blood sampling was determined. At the time of final analysis, 14 of 18 patients were deceased, with a median follow-up of 36.3 months (1–78.6 months). All 6 patients with more than or equal to 5 CTCs/7.5 mL at the time of sampling died during follow-up, compared with 8 of 12 (67%) with less than 5 CTCs/7.5 mL (P = .245). The median survival from time of sampling was 13 months for those with more than or equal to 5 CTCs, and 51.5 months for those with less than 5 CTCs. Kaplan-Meier analysis demonstrated significantly shorter survival from sampling for those with more than or equal to 5 CTCs (log-rank P = .0116 and Wilcoxon P = .0137) (Figure 3). In contrast, using the cutoff of more than or equal to 1 CTCs that had optimally distinguished metastatic disease from NED controls, there was no significant difference in survival between those metMTC patients with more than or equal to 1 and those with less than 1 CTCs/7.5 mL. Ctn doubling time at time of CTC sampling was 15.9 ± 40.6 months in patients who died compared with 53.5 ± 67.9 months in those who were still alive at last follow-up (P = .082). By comparison, CEA doubling times were 15.9 ± 40.6 and 26.2 ± 58.7 months, respectively (P = .798). The hazard ratio (HR) for mortality of more than or equal to 5 CTCs compared with less than 5 CTCs was 3.95 (95% confidence interval [CI] 1.20–13.0, P = .0245), whereas for Ctn and CEA doubling time less than 6 months compared with more than or equal to 6 months, the HRs were only 1.68 (95% CI 0.55–5.29, P = .358) and 1.30 (95% CI 0.42–3.77, P = .6327), respectively. In multivariate analysis, the presence of more than or equal to 5 CTCs/7.5 mL significantly predicted shortened OS after adjusting for age at diagnosis, Ctn and CEA doubling times at sampling, number of metastatic sites, presence of a germline RET mutation, and current/previous systemic therapy.

Figure 3.

OS according to CTC levels. HRs for mortality: CTCs/7.5 mL more than or equal to 5 vs less than 5: 3.95 (1.20–13.0).

Discussion

Identification of biomarkers that provide accurate prognostic information can critically aid the clinical management of patients with cancer. In this study, we have demonstrated that CTCs can be detected in the blood of a subset of patients with metMTC, using the EpCAM-based CellSearch methodology that has previously been shown to be an accurate and reproducible assay for CTCs for several cancers of epithelial origin (7). Further, an elevated level of CTCs enumerated by this test appears to be prognostic of an increased risk for mortality, potentially to a greater degree than the current biomarkers currently used, Ctn and CEA doubling times.

The frequency with which CellSearch detected CTCs in metMTC and the range of absolute levels in each patient were consistent with reports in patients with breast, prostate, and colon carcinomas. The good correlation of CTC numbers between duplicated samples demonstrated the reproducibility of enumeration of CTCs, even at low numbers, and would suggest that further studies can be performed using only a single blood sample. Therefore, the CellSearch method provides an accurate and reproducible assay for CTC enumeration in metMTC patients.

The prognostic value of CTCs measured by CellSearch has been well recognized and proven by prospective trials in other solid tumors. A cutoff of at least 5 CTCs per 7.5 mL was the strongest predictor of OS on multivariate analysis in nonsmall-cell lung (15), breast (14), and prostate cancer (11). Moreover, CTC number predicted the response to chemotherapy in nonsmall-cell lung cancer (15) and metastatic breast cancer (16). Despite our study's small sample size, we were able to observe that CTC number of at least 5 was independently prognostic of a significantly worse OS in metMTC patients than a lower CTC number. Indeed, CTC level outperformed established prognostic factors for MTC, Ctn and CEA doubling times, potentially replacing the current need to rely upon multiple measurements over months or years to calculate doubling times with a single point-in-time assay. Of note, the metMTC patients we recruited had advanced disease with distant metastases. The presence and roles of CTC enumeration in previously untreated patients or those with low risk of recurrence or metastasis need to be investigated. A major limitation of our study is the small number of subjects, and clearly these results need to be validated in a larger, potentially more diverse cohort of patients, including identification of cutoffs with optimal predictive values. Further, nearly half of the patients were receiving various types of systemic therapy that may potentially confound CTC measurements. For example, the use of antiangiogenic therapies that alter vascular structures might contribute variability to the ability of tumor cells to access the circulation. Thus further studies should separate treatment naive patients from those already on such therapies for metastatic disease. In addition, studies to evaluate the changes of CTC levels after initiating systemic therapy are necessary elucidate the potential role of CTC enumeration in prediction of response to treatment.

EpCAM expression, the primary target of CTC capture by CellSearch, differs among various carcinomas and may be lost in the evolution of the disease, thus contributing to variation in the ability to detect CTCs in patient samples. EpCAM expression has been previously reported in DTC (13). In the recovery experiments reported here, CellSearch was able to detect cells from PTC lines. Nonetheless, only 1 of 14 metDTC patients had detectable CTCs using CellSearch, and none had levels above the conventional cutoff. Decreased membrane expression and increased nuclear accumulation of EpCAM has been previously reported in aggressive PTC, which may have contributed to the failure to detect CTCs in metDTC patients in our study (17). Consistent with this observation, circulating epithelial cells have been observed in patients with DTC when cells are counted by methods not requiring membrane-bound EpCAM for detection (18). Thus, other approaches to detecting and enumerating CTCs in patients with metDTC will need to be explored.

In conclusion, this is the first study to evaluate CTC number in TC using CellSearch technology. CTCs were mainly identified in metMTC patients, and in this population, the presence of more than or equal to 5 CTCs per tube was associated with worse OS. Further, elevated levels of CTCs was more predictive than Ctn and CEA doubling time, suggesting that a single measurement could be of clinical value compared with the serial assessment required for the conventional biomarkers. Larger cohorts are required to validate these observations. Subsequent studies are needed to determine whether CTCs could identify treatment naive metMTC patients who might benefit from systemic therapy, or whether alterations in CTCs could predict response to treatment and improved outcomes.