Assessment of Circulating Tumor Cells as a Predictive Biomarker of Histology in Women With Suspected Ovarian Cancer

Abstract

Background

The clinical assessment of circulating tumor cells (CTCs) as a blood-based biomarker is FDA-approved for use in breast, colorectal, and prostate cancers. The objective of this prospective clinical study was to determine whether pretreatment CTCs are a useful diagnostic biomarker in women with complex pelvic masses.

Methods

Whole blood was collected from 49 women with newly diagnosed pelvic masses. The presence of CTCs was compared between women with and without ovarian cancer histopathologic diagnosis after surgery using a Chi-squared test.

Results

CTCs were absent in those with benign disease (0/14), present in 17% (5/29) of patients with a histologic diagnosis of ovarian carcinoma, and present in 80% (4/5) of patients with ovarian metastases from other cancers (P = 0.001). All 5 women with ovarian cancer who had CTCs present presented stage III or IV of the disease (P = 0.13).

Conclusions

CTCs were more prevalent in patients with metastases to the ovary than in primary ovarian carcinomas.

Ovarian neoplasms are common, but only a minority are malignant. It is difficult to differentiate malignant ovarian masses from benign neoplasms as there are currently no diagnostic tests sensitive enough to differentiate between the 2 without resulting in a high number of false positive results.¹ CA-125 is the most widely cited and only clinically available biomarker for ovarian cancer.^2-6 While CA-125 can be used as a serum marker of ovarian cancer recurrence, it has a relatively low sensitivity and specificity for the screening and diagnosis of ovarian cancer.⁷ Only 50% of patients with early-stage ovarian cancers have elevated serum CA-125,⁸ and a number of non-ovarian cancer conditions have falsely elevated CA-125. There is a strong need to identify testable biomarkers that can be used at the time of diagnosis of a pelvic mass to triage patients to the appropriate provider and treatment.

Circulating tumor cells (CTCs) separate from a malignant tumor and circulate in the bloodstream, leading to dissemination of disease. They not only colonize, but also stimulate tumor growth, in addition to recruiting stromal cells and enhancing the tumor’s invasive potential.⁹ Clinical assessment of CTCs as a blood-based biomarker is FDA-approved for use in breast, prostate, and colorectal cancers. Higher concentrations of CTCs were associated with worse outcomes and prognoses in these cancers.^10-13 Few studies, however, have examined CTCs in ovarian cancer patients, and the studies that have been done have only looked at the correlation of disease response to investigational drugs.^14-17 Improved methods of identifying and isolating CTCs have made it possible to quantify CTC levels in the blood at the time of diagnosis, and could be used as a predictive or prognostic biomarker if clinically validated.

We hypothesized that the assessment of CTCs in women with newly diagnosed ovarian masses could be used as a serum-based predictive biomarker of epithelial ovarian malignancies. The objective of this prospective pilot study was to evaluate whether pretreatment CTCs could be used as a diagnostic biomarker of primary ovarian malignancy in women with pelvic masses.

Materials and Methods

This study was approved by the Institutional Review Board (IRB) at the University of Minnesota (Study #1312M46201). Women with newly diagnosed complex pelvic masses evaluated in the gynecologic oncology clinic at the University of Minnesota Gynecologic Oncology Clinic from April 1, 2014 to June 30, 2016 were recruited for this prospective clinical study. Written informed consent was obtained from all participants.

To perform CTC enumeration, 7.5 cc of whole blood was collected in a CellSearch collection tube prior to surgery. More specifically, for those patients who underwent surgery as their initial form of treatment (eg, no neoadjuvant therapy), the blood samples were collected on an average of 10.3 days prior to surgery (median = 8.0 days; range 0 to 31 days). Patients who received neoadjuvant chemotherapy treatment prior to surgery, had blood collected on an average of 9.6 days prior to start of this chemotherapy course (median = 8 days; range 2 to 17 days).

Carcinoma cells were positively selected using magnetic beads conjugated to an anti-EpCAM antibody. CTC enumeration was performed using photomicroscopic image analysis after staining cells with DAPI, anti-CD45, and an anti-cytokeratin cocktail comprised of CK8, CK18, and CK19. CTCs were identified as positive if they were EpCAM-positive, CK-positive, DAPI-positive and CD45-negative and had the morphology of a single intact carcinoma cell (eg, no cell clusters identified).

Patient demographic and clinical data were extracted from the electronic medical record, including age at diagnosis, race, pelvic mass origin and histology, tumor grade and stage, tumor size, and lymph node status. Serum levels of CA-125 were also assessed per standard of care.

Statistical Analysis

The goal of the analysis was to determine whether the presence of CTCs (0, > 0) at the time of diagnosis correlated with a cancer diagnosis and ovarian cancer diagnosis. Demographics and clinical characteristics were summarized using descriptive statistics. The proportions of women with and without CTCs were compared by cancer and ovarian cancer diagnoses using Fisher’s Exact tests. Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPP), and their corresponding 95% confidence intervals were calculated for use of CTCs to detect any cancer diagnosis. Among those with ovarian cancer, measures of aggressiveness including histologic type (high-grade epithelial vs low-grade epithelial vs borderline epithelial, or stromal tumors including granulosa cell tumors), stage (I;II vs III;IV), grade, tumor size and lymph node involvement were also compared between women with and without CTCs. All analyses were conducted using SAS version 9.4 (Cary, NC) and P-values < 0.05 were considered statistically significant.

Results

We prospectively enrolled 49 women with newly diagnosed complex pelvic masses (Table 1). The mean age of participants was 60.0 ± 11.8 years; 90.7% were white (non-Hispanic). Most (93.8%) of these women underwent surgical staging, and 14.6% received neoadjuvant chemotherapy prior to surgery. The median baseline CA-125 was 129.5 units/ml (range: 5 to 2,782 units/ml). The normal range is <35 units/ml.

Table 1.

Patient Characteristics

Variable	n	%
Age, years, mean (SD)	49	60.0 (11.8)
Race
White, Non-Hispanic	39	90.7
Other	4	9.3
Unknown	6
Surgery
No	3	6.3
Yes	45	93.8
No follow-up	1
Neoadjuvant Chemotherapy
No	41	85.4
Yes	7	14.6
No follow-up	1
Baseline CA-125, median (min-max)	46	129.5 (5-2782)

n = 49

Of the 49 enrolled patients, 35 (71.4%) had a histologic diagnosis of malignancy. CTCs were discovered in 9/49 (18.4%) of the patients; the median CTCs among those in whom they were detected was 2 (range 1 to 21). CTCs were detected in 9/35 (25.7%) of patients with cancer as compared to 0/14 (0.0%) not diagnosed with cancer (P = 0.045) (Table 2). The sensitivity of the CTC test for predicting a malignant histology of pelvic masses was 25.7% (95% CI: 12.5%, 43.3%). The specificity of CTCs for identifying malignant histology was 100.0% (95% CI: 76.8%, 100.0%). The PPV for CTCs was 100.0% and the NPV of CTCs was 35.0%.

Table 2.

Cancer Diagnoses by CTC Group

	CTCs = 0		CTCs > 0
	n	%	n	%	P value*
Cancer Diagnosis					0.045
No	14	100.0	0	0.0
Yes	26	72.2	9	25.7
Ovarian Cancer					1.00
No Cancer or Other Cancer	15	79.0	4	21.1
Yes	24	82.8	5	17.2
Unknown Cancer Origin	1
Cancer Type					0.001
None (Benign)	14	100.0	0	0.0
Ovarian Cancer	24	82.8	5	17.2
Other Cancer	1	20.0	4	80.0
Unknown Cancer Origin	1

*Fisher’s Exact Test

The percent of patients with detectable CTCs was higher in patients with cancers of non-ovarian origin that metastasized to the ovaries as compared to patients with primary ovarian cancer. More specifically, CTCs were detected in 4 out of 5 patients (80.0%) with malignant ovarian masses not of ovarian origin. These tumors included two Krukenberg tumors (gastrointestinal origin), one metastatic endometrial cancer, and one case of abdominal soft-tissue sarcoma with peritoneal carcinomatosis. CTCs were detected in only 5 out of 29 patients (17.2%) with proven primary ovarian cancer.

We then investigated any potential associations of CTCs with ovarian cancer histologic subtype, tumor grade, tumor stage and size, and lymph node status among those with ovarian cancer (Table 3). Of these, the only notable correlation of CTC presence was with higher stage (FIGO stage III or IV) malignant ovarian tumors. All 5 patients with detectable CTCs (5/18) had advanced stage disease (III or IV), whereas none had stage I;II disease (0/11; P = 0.13). Further, while not statistically significant, none of the patients with epithelial borderline or stromal ovarian tumors had CTCs present (P = 0.61).

Table 3.

Characteristics by CTC Presence Among Those With Ovarian Cancer

	CTCs = 0		CTCs > 0
	n	%	n	%	P value*
Ovarian Cancer Type					0.61
High-grade epithelial	13	81.3	3	18.8
Low-grade epithelial	4	66.7	2	33.3
Borderline	5	100.0	0	0.0
Stromal	2	100.0	0	0.0
Tumor Grade					1.00
G1	5	83.3	1	16.7
G2	3	100.0	0	0.0
G3	10	83.3	2	16.7
Missing	6		2
Stage					0.13
I-II	11	100.0	0	0.0
III-IV	13	72.2	5	27.8
Tumor Size (T)					0.35
1	8	100.0	0	0.0
2	2	100.0	0	0.0
3	12	75.0	5	25.0
Missing	2		1
Lymph Nodes (N)					0.64
0	9	90.0	1	10.0
1	5	71.4	2	28.6
X	8	88.9	1	11.1
Missing	2		1
Debulking					1.00
Suboptimal	3	75.0	1	25.0
Optimal	12	85.7	2	14.3
Missing	9		2

*Fisher’s Exact test

Discussion

The primary focus of this investigation was the potential utility of CTCs as a diagnostic biomarker of ovarian cancers. Our study found that the PPV of CTCs for malignancy was 100%, which is consistent with at least one other study.¹⁸ Thus, in the clinical setting, if a patient was found to have CTCs at presentation, evaluating and triaging would need to be prioritized and completed in a timely manner. Although the detection rate of primary ovarian malignancies was low, enumeration of CTCs was prominent in women with Krukenberg tumors, (ie, metastatic cancers that were of non-ovarian origin but which upon initial clinical presentation were suspicious for ovarian cancer).

As a prognostic biomarker, CTCs have been studied most extensively in prostate, breast, and pancreatic cancers, where it was found that higher concentrations of CTCs correlate with a worse outcome.^10-13,19 Several published studies have correlated the presence of CTCs with shorter, disease-free overall survival (OS) in ovarian cancer.²⁰ Notably, a meta-analysis by Zhou et al reported CTC detection as a prognostic factor associated with worse survival, and this correlation did vary based on subgroup and method used to enumerate CTCs. The subgroup that had CTCs detected using the CellSearch method specifically showed no difference in OS, in contrast to CTC detection using reverse transcriptase-polymerase chain reaction (RT-PCR).²⁰ We speculate that the overall lower yield of CTCs using CellSearch was the primary reason that this method failed to associate significantly with prognosis. Also of note, as this was a meta-analysis comparing several studies, is that cutoffs were variable across studies, further confounding the interpretation of this finding.²⁰ In general, the yield of CTCs across several of these cancers is rather limited, and in our study on ovarian cancer, only 17% of patients with proven histopathologic diagnosis of primary ovarian carcinoma had detectable CTCs. Therefore, the identification of CTCs is not likely to be a useful diagnostic or screening biomarker for ovarian cancer. This finding of relatively low CTCs at baseline is consistent with other studies in ovarian cancer that used alternate approaches, including microarray analysis.²¹ However, our results also stand in sharp contrast to another study that used the same system (CellSearch) but reported identification of CTCs in > 80% of 78 women with ovarian cancer.²² That specific study examined peripheral blood specimens from new patients, as well as those with recurrent disease. The threshold for positive CTCs in that study was ≥2. A potential limitation of that report was that 23% of patients (18/78 total, newly diagnosed and recurrent) had 1 CTC detected, but it was not verified whether CTCs at that quantity were truly malignant or represented benign circulating epithelial cells. It is unclear whether the unusually high yield of positive CTC values could also be due to presence of non-malignant cells.²² Our study was initiated at a time when platforms such as CellSearch were available for use at our center and others, and thus, enumeration of CTCs was readily feasible. However, in the relatively short timespan of several years, new technologies have emerged that have greatly improved the sensitivity of CTC capture, and even further permitted pharmacogenomic and other forms of profiling to glean more information from CTCs. As one example, Lee et al recently reported a more cutting-edge approach, using an electronically conductive chip as part of a microfluidic platform to capture CTCs. They determined that 98% of ovarian cancer patients were found to have CTCs using that platform. Of these, 24 had newly diagnosed disease and 30 were recurrent with a greater number of CTCs found in the newly diagnosed group, albeit at relatively low levels with a median of 3 cells.²³ This, and other similar novel platforms that incorporate size-dependent separation of CTCs,²⁴ may ultimately provide better approaches to enumeration and analysis in the years to come.

Krukenberg tumors are secondary forms of ovarian malignancy, most often originating from primary carcinomas of the breast, endometrium, stomach, appendix, colon and rectum.^25-28 Because the initial evaluation and work-up of women with newly diagnosed pelvic and ovarian masses raises suspicion for primary ovarian cancer, there is a strong risk that this patient population can be misdiagnosed, and consequently, not receive appropriate therapies geared toward treating the true cancer of origin.²⁵ While the crux of making the correct diagnosis differentiating these metastatic tumors to the ovary from primary ovarian cancers is based on careful histopathologic analysis and clinical history and staging, some cases present diagnostic dilemmas. Thus, biomarkers that could further differentiate these 2 populations of cancer could be of clinical use in these specific cases. Admittedly, however, the CTC method in our study only enumerates the carcinoma cells, and cannot subtype the epithelial cell of origin. While the intended primary purpose of our study was not to detect differences in these 2 populations, the fact is that such a high disparity in CTC enumeration between primary ovarian cancers (17.2%) as compared to cases of Krukenberg tumors (80.0%), raises the question of whether or not CTC enumeration can fill that clinically relevant need.

The main limitations of our study include the low yield of CTCs as compared to the overall number of patients with cancer, and the overall limited number of patients evaluated. Although CTC enumeration is FDA-approved and available in CLIA-approved laboratories at major medical centers, there are logistical challenges to obtaining accurate CTC concentrations. The reasons for this may be severalfold. The CellSearch method we used only targets EpCam and various cytokeratins, and does not detect EpCAM negative cells. This fact may explain, at least in part, any variation in CTC detection across cancer types, as cell surface markers vary.²¹ This point is crucial, as it has become clear that relying on EpCam as the cell surface marker best recognized by CellSearch leads to low sensitivity of the assay. Likely compounding this issue is the fact that EpCam expression is downregulated as malignant cells undergo epithelial-to-mesenchymal transition (EMT), a process that is not only associated with altered cell morphology, but also with underlying molecular expression profiles, increased rate of metastasis, and potential drug resistance.^29-32 Another potential solution to increase yield is to standardize pre-enrichment of blood samples prior to enumeration. Pearl et al reported the use of an initial enrichment step that significantly increased sensitivity and specificity of invasive CTCs in patients with stage I or II epithelial ovarian cancers.¹⁸

We expect that the CTC research community will continue to discuss this issue in depth and come to a consensus of how to best optimize targeting CTC cell surface markers to accurately identify CTCs. CTC clusters have been implicated as mediators of cancer metastasis,³³ and specifically in ovarian cancer as well.^21,24 As more studies are aimed at refining the technique of CTC identification using novel and more sensitive methodologies such as microfluidics, it is likely that identification of CTC clusters, in addition to single cells, may increase the yield of detectable CTCs from patient samples.²³ In turn, better detection of both single cell and cell clusters together may prove more useful in predicting clinically meaningful outcomes of women with ovarian cancer. Adapting available alternate methods of cell separation, such as flow cytometry, has been used to detect CTCs, and to provide novel approaches to overcome this problem as well. In one such study, Kim et al used a MUC-1 antibody conjugated to a rhodamine isothiocyanate magnetic nanoparticle to improve binding capacity of CTCs, and were able to capture as many as 100 malignant ovarian cells for each 50 µl sample of whole blood.³⁴ This method is an example of a potentially significant step in improving cell capture and sensitivity in a minimal amount of blood. Furthermore, detection of other circulating biomarkers, including plasma-borne microRNAs, is also being explored and may provide additional and more reliable screening approaches.³⁵

In conclusion, CTCs are unlikely to be a useful diagnostic biomarker for ovarian cancer when detected and enumerated using the CellSearch method alone. Despite a low prevalence of CTCs overall using this method, the PPV of this test for indication of malignancy is impressive, and if discovered, further management should not be delayed regardless of the primary source of the malignancy. However, in women presenting with complex adnexal masses, CTCs were more prevalent in patients with cancer of non-ovarian origin that metastasized to the ovaries than in patients with primary ovarian cancers. We postulate that CTC enumeration may correlate with tumor burden in non-ovarian malignancies, such as gastrointestinal cancers, that manifest with ovarian metastasis at the time of diagnosis. Ongoing and future studies representing the evolution of CTC testing that includes pharmacogenomic profiling may provide more sensitive and clinically useful approaches to assessment of CTCs and their role in ovarian cancer.

Funding

This research was supported by a NIH Clinical and Translational Science KL2 Scholar Award 8UL1TR000114 (to E.L.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Additional funding support came from the Litman Family Fund for Cancer Research; the University of Minnesota Deborah E. Powell Center for Women’s Health Interdisciplinary Seed Grant support (Grant #PCWH-2013-002) (E.L.); Minnesota Masonic Charities; the Masonic Cancer Center (Grant # P30 CA77598) and Department of Medicine, Division of Hematology, Oncology and Transplantation, University of Minnesota.

Abbreviations:

CTCs

circulating tumor cells

IRB

Institutional Review Board

PPV

positive predictive value

NPP

negative predictive value

RT-PCR

reverse transcriptase-polymerase chain reaction

OS

overall survival

EMT

epithelial-to-mesenchymal transmition