Promises and Pitfalls of Using Liquid Biopsy for Precision Medicine

In the era of precision medicine, liquid biopsies are increasingly being studied as a tool that can capture tumor evolution in real time and thus guide systemic treatment. In this article, we will refer to the analysis of circulating tumor DNA (ctDNA) and circulating tumor cells (CTC) only and we will not cover other liquid biopsy biomarkers such as circulating RNAs, proteins, metabolites, and exosomes.

Sampling a patient's blood may give information about the genomic profile of a given cancer (1–3) and provide an assessment of tumor burden (4), without the need of invasive procedures. Over the last few years, several studies have been published supporting the analytical and clinical validity of CTC (5) and ctDNA assays (6) in cancer.

However, according to a recent ASCO review there is still insufficient evidence of clinical utility for the majority of ctDNA assays in advanced cancer and no evidence of clinical utility in early-stage disease or cancer screening (7). Moreover, no CTC assay is currently being used in the clinic.

Recent reviews have addressed the use of liquid biopsies (8–10), focusing especially on the technologies and their analytical and clinical validity (8). This review aims to focus more on the remaining challenges that currently prevent the use of liquid biopsies clinically (demonstration of clinical utility). To that end, we will provide an update (2) on ongoing/completed key clinical studies using CTCs and ctDNA for clinical decision (Table 1). Additionally, the main clinical applications of CTCs and ctDNA that are currently being explored are summarized in Fig. 1A.

For CTC and ctDNA assays, there is a need to standardize preanalytical variables and for cross-platform comparison studies. To address these challenges, initiatives are underway both in Europe (Cancer-ID; ref. 11) and United States (BloodPAC; ref. 12) aiming to standardize preanalytical issues and compare the performance of different liquid biopsy assays in the same patient samples.

In particular, the low concentration of CTCs and ctDNA limits the use of liquid biopsies in early-stage cancer and drawing a large volume of blood is not always clinically feasible. The addition of extra markers for CTCs, the use of implanted devices containing materials that bind ctDNA (13) or assays that can simultaneously test for multiple mutations in the same reaction (14), appear to be promising in increasing the amount of CTCs/ctDNA detected. Alternative approaches that integrate ctDNA mutations with multiple other blood-based analytes (such as exomes, CTCs, ctDNA epigenetics, metabolites) may also be required.

In the metastatic setting, the only ctDNA assay that is currently used in solid tumors is the cobas EGFR Mutation Test v2. This assay is used for patients with metastatic non–small cell lung cancer (NSCLC) and can detect 42 mutations in exons 18, 19, 20, and 21 of the EGFR gene encompassing the T790M-resistant mutation. A positive finding—of an actionable mutation in plasma—using this assay can inform treatment selection relating to erlotinib and osimertinib. However, a negative result—such as the absence of the T790M mutation in a patient with clinical or radiological progression—should be considered inconclusive and DNA from a tumor biopsy should also be assessed (15).

The assessment of PIK3CA mutations in plasma cell-free DNA may be the first liquid biopsy to be used in the clinic for metastatic breast cancer (MBC). This is based on the analysis of PIK3CA mutations in ctDNA among 549 hormone receptor (HR)+/HER2− MBC patients enrolled in the SOLAR-1 trial (16). Progression-free survival (PFS) was significantly prolonged when the PI3KCA selective inhibitor alpelisib was added to fulvestrant in patients with detectable PIK3CA mutations in ctDNA. If the FDA approves alpelisib in this setting, PIK3CA mutation ctDNA testing will likely become a companion diagnostic test. Moreover, the presence of PIK3CA mutations in plasma ctDNA in patients with endocrine-resistant estrogen receptor (HR)+/HER2− advanced breast cancer, identified patients that could benefit from the combination of the pan-PI3K inhibitor buparlisib with fulvestrant (BELLE-2 trial; ref. 17). However, the clinical development of buparlisib has been discontinued for toxicity issues. The value of PIK3CA ctDNA genotyping has also been evaluated in other studies. In the BOLERO-2 study, the addition of everolimus to exemestane prolonged median PFS irrespective of PIK3CA genotype (18). In the PALOMA-3 study, PIK3CA ctDNA levels after 15 days treatment appear to be predictive of PFS on palbociclib and fulvestrant (19).

In patients with MBC, the use of digital PCR assays (19, 20) showed that estrogen receptor 1 (ESR1) mutations in ctDNA are frequently subclonal, and occur later during metastatic aromatase inhibitor (AI) therapy. Results from retrospective ctDNA analysis in the SoFEA trial showed that ESR1 mutation analysis in plasma might be useful to direct the choice of further endocrine-based therapy; because patients with plasma ESR1 mutations (63 of 161) have a shorter PFS on subsequent AI-based therapy (18 patients) compared with fulvestrant (45 patients; ref. 21). The results of the above study have been recently extended in 383 patients in a combined meta-analysis of the Sofea and Efect studies (22). However, the benefit of adding CDK4/6 inhibitors to endocrine treatment was largely irrespective of the presence of ESR1 mutations (21). As the combination of endocrine treatment with CDK4/6 inhibitors has become the new standard of care, the clinical value of ESR1 mutation detection in patients progressing on AIs is limited to patients that will subsequently receive a second line of endocrine monotherapy without CDK4/6 inhibitors. Moreover, analyses of plasma baseline ESR1 mutations in the BOLERO-2 trial, demonstrated that the presence of ESR1 mutations was a marker of poor prognosis (23). In a subgroup analysis, it was suggested that patients with the Y537S ESR1 mutation did not derive benefit from the addition of everolimus to exemestane, although these data need independent validation in larger series. Currently, there is no clear role for plasma ESR1 mutations as a tool to guide treatment in patients with MBC.

ctDNA assays have been incorporated in trials for various solid tumors, with lung (24) and prostate cancer (25) being lead indications.

Several studies are also evaluating the role of CTC enumeration and characterization in guiding treatment decision, especially in metastatic breast and prostate cancer. The French STIC CTC study (NCT01710605; refs. 26–28), demonstrated the “noninferiority” of a CTC-based treatment decision vs. a clinician-based treatment decision, in the choice of first-line treatment (single-agent hormone therapy vs. chemotherapy) of HR+ MBC and confirmed the adverse prognostic value of baseline CTC count (27). In the CTC-driven arm, for patients that received chemotherapy because of their high CTC count (≥5 CTCs) PFS was significantly longer than in the clinically-driven arm (HR 0.67; 95% CI, 0.49–0.92; P = 0.01; ref. 28). However, interpretation of the results of this study is challenging due to the change in the standard of care for these patients, which is currently the combination of hormone therapy with CDK4/6 inhibitors.

Beyond CTC enumeration, there is preliminary evidence that the molecular analysis of CTCs in patients with MBC with bone (29) or brain (30) metastases might help direct the choice of treatment. Finally, preliminary evidence suggests that the use of highly sensitive assays for analyzing prostate CTC-derived transcripts (31, 32) may help to guide therapies, especially in advanced prostate cancer with bone metastases that cannot be easily biopsied. Patients with prostate cancer with nuclear-localized androgen receptor splice variant 7 (AR-V7)-positive CTCs exhibited resistance to abiraterone and enzalutamide (31) and had a superior overall survival (OS) when treated with taxanes (32). In a recent consensus statement on circulating biomarkers for advanced prostate cancer (33), 31% of the experts preferred the EPIC AR-V7 CTC protein assay—laboratory-developed and validated (32, 34)—as an AR-V7 test to use in clinical practice, whereas 7.67% of the scientists chose the Hopkins/Qiagen AR-V7 RT-PCR mRNA-based assay using the AdnaTest platform (35).

Plasma ctDNA analyses using NGS cancer gene panels (either pan-cancer or disease-specific) have been shown to be an attractive alternative to tumor tissue biopsy NGS analysis in molecular screening programs; and have the potential to increase patient access to clinical trials with new compounds (36, 37), as in the case of the basket trial SUMMIT (38).

Alongside treatment selection, another potential application of liquid biopsies is in monitoring treatment response. This might allow for the earlier withdrawal of expensive and potentially toxic drugs when a liquid biopsy test suggests that there will be no benefit. CTC elimination after short-term drug exposure has been demonstrated to be an early response endpoint in metastatic castration-resistant prostate cancer, more reliable than PSA levels, which are affected by modulations in androgen receptor signaling (39). In the SWOG S0500 clinical trial, patients with MBC and persistently increased CTC levels after one cycle of first-line chemotherapy were shown to have a poorer outcome, although an early switch to an alternative chemotherapy regimen did not improve outcome (40). In this setting, it might be interesting to test whether a switch to a therapy targeting specific genomic aberrations, detected using a liquid biopsy test, might be a better approach instead of switching to a different chemotherapy regimen. Beyond CTC enumeration, monitoring of the CTC genomic and epigenomic profile might also better inform treatment selection. Indeed, the presence of ESR1 methylation in CTCs from serial blood samples was shown to be associated with lack of response to everolimus/exemestane (41). Finally, serial ctDNA evaluation can be used for treatment monitoring (4, 19).

Immune checkpoint inhibitors, such as inhibitors of programmed-death receptor 1 (PD1) or programmed death receptor ligand (PDL1) have become a standard therapy in several cancer types. Preliminary evidence suggests that longitudinal on-treatment monitoring of CTCs (42), ctDNA (43, 44), or exomal PD-L1 (45) dynamics might be used as a marker of response to identify the patients more likely to benefit from immunotherapy, thus sparing nonresponding patients from such treatment. However, definitive studies to change clinical practice are needed. Blood-based monitoring may also help decide whether to continue immunotherapy or not in cases of pseudoprogression (46).

The molecular basis of the acquired resistance to targeted therapies represents one of the main challenges in cancer research, with large implications in the clinical setting. Serial ctDNA analysis has emerged as a promising tool to identify and track the mechanisms of drug resistance (47–49). ctDNA and CTCs have been useful tools to detect multiple genomic alterations (in genes such as ESR1, genes of the mitogen-activated protein kinase (MAPK) pathway and the RB1, T790M, KRAS, and BRAF genes) that emerge following treatment with various therapies (AI, CDK4/6 inhibitors; osimertinib; cetuximab; BRAF inhibitors) in breast cancer (50–54), lung cancer (55), colon cancer (49), and melanoma (56, 57), respectively. CTC and ctDNA analysis may, therefore, have a complementary utility to detect mechanisms of resistance (58, 59). Results from the above studies suggest that different mechanisms of resistance can often coexist within a given patient (55), with frequent subclonal mutations (50), so targeting only one pathway might result in a transient benefit for the patient. Combination therapy might provide a better strategy to delay drug resistance in this setting.

A pooled analysis of individual patient data from more than 1,500 patients from 21 studies provided evidence that baseline CTC counts, as determined by Cellsearch, is an independent poor prognostic factor in patients with breast cancer treated with neoadjuvant chemotherapy (60).

The purpose of administering adjuvant treatment after surgery is to eradicate tumor cells undetectable by conventional imaging called minimal residual disease (MRD). Proof-of-principle studies demonstrated that persistent detection of ctDNA after local therapy [surgery or radiotherapy (RT)] is associated with a higher risk of recurrence (14, 61). In this setting, monitoring multiple mutations per patient has been shown to improve the sensitivity for ctDNA detection (14). Moreover, it has been suggested that CTC detection before or after adjuvant chemotherapy is associated with a worse clinical outcome (62–64). Furthermore, the detection of CTCs 5 years or more after diagnosis was recently found to be associated with a higher risk of late recurrence in patients with HR+HER2− localized breast cancer (65). The correlation between ctDNA/CTCs levels and the lead time to clinical relapse is shown in Fig. 1B.

Preliminary data support the design of clinical trials for selection of adjuvant therapy based on MRD detection and characterization, in addition to characterization of the primary tumor (66). The first international trial to test this new model of drug development has been the Treat CTC trial (67). In the Treat CTC trial, we have explored the question of whether six cycles of trastuzumab can eliminate chemotherapy-resistant CTCs following standard (neo)adjuvant chemotherapy and surgery in women with HER2-non-amplified breast cancer. The trial was terminated by an independent data monitoring committee after 63 patients had been randomized because it demonstrated that trastuzumab could not eliminate CTCs in this setting. In line with the Treat CTC trial results, the NSABP B47 phase III trial including more than 3,000 patients demonstrated that 1 year of adjuvant trastuzumab did not improve invasive disease-free survival when added to standard chemotherapy in HER2-negative breast cancer (68). More examples such as the Treat CTC/NSABP B47 example are needed to provide evidence that CTCs or ctDNA elimination after short drug exposure can provide relevant information that can be used in addition to data from the activity of the drug in the metastatic and neoadjuvant setting in order to make more informed decisions before testing this drug in a large phase III adjuvant trial. Moreover, CTCs and ctDNA can be used in the future for designing trials aiming to de-escalate (in CTC/ctDNA-negative patients) or escalate (in CTCs/ctDNA-positive patients) systemic or locoregional treatment.

In early-stage cancers, the identification of patients at higher risk of recurrence who will benefit from adjuvant RT remains challenging. Promising preliminary results were reported in the recent analysis of patients with stage pT1-T2 and pN0-N1 breast cancer from the National Cancer Database and the multicenter phase III SUCCESS clinical trial (69). Adjuvant RT after breast-conserving surgery was associated with longer OS among patients of both cohorts with detectable CTCs before adjuvant therapy and was also associated with longer locoregional-free survival and disease-free survival in patients with CTCs from the SUCCESS cohort. This benefit was not observed in patients without CTCs and prospective validation of these findings is required.

There are efforts to explore whether ctDNA testing can be used as a tool for cancer diagnosis. For instance, the detection of circulating, cancer-derived (Epstein Barr Virus) EBV DNA in plasma has proven to be a useful screen for nasopharyngeal carcinoma in asymptomatic subjects, with high sensitivity and specificity (70). The use of ctDNA sequencing seems similarly promising in the identification of patients with somatic mutations associated with increased risks of hematologic cancer. However, caution is warranted because clonal hematopoiesis with somatic mutations was detected in 10% of elderly patients, whereas the risk of developing a hematologic malignancy was modest (1% per year; ref. 71). Moreover, new blood tests (72, 73) seem to be useful in discriminating cancer patients from healthy controls and allow for the detection of early cancers and also localize to the organ of origin such as CancerSEEK (72) and one further method, based on a genome-wide bisulfite sequencing of plasma DNA, that can identify the specific methylation profiles of each tissue (73). Another approach to identify the tissue of origin has been mRNA sequencing of tumor-educated platelets (74). There are several ethical issues related to the clinical testing of a cancer diagnostic test. Such a test needs to be sensitive enough to detect surgically resectable tumors but also highly specific, since false-positive results will cause unnecessary imaging workup and severely affect quality of life. Prospective clinical trials to demonstrate the clinical utility of a liquid biopsy cancer screening test in addition to, or instead of, standard screening programs are required (75).

There are increasing data on the analytical and clinical validity of commercially available liquid biopsy assays; however, demonstration of clinical utility is needed. To this aim, there is a need of clinical trials with inclusion criteria based on ctDNA/CTCs detection and characteristics in order to make decisions of treatment escalation or de-escalation. The detection of PIK3CA mutations in plasma ctDNA to guide treatment with the PIK3CA inhibitor alpelisib might provide the first example of a ctDNA assay with clinical utility in MBC. In the near future multigene ctDNA assays are expected to be used for guiding treatment selection in the metastatic cancer setting. Moreover, efforts should be made to validate the use of highly sensitive liquid biopsy assays (76) in the early disease setting to identify patients with MRD, who will benefit from adjuvant treatment, or those without MRD, who can be spared the toxicity of adjuvant treatment. Therefore, the development of targeted drugs designed to eliminate dormant tumor cells (77) or maintain them in a quiescent status might be an attractive perspective in order to delay/prevent the progression from MRD to overt metastases. The results from the ongoing trials are expected to improve patient outcomes and change the way we treat cancer patients.

M. Ignatiadis is a consultant/advisory board member of Novartis, Pfizer, Celgene, Tesaro, and Seattle Genetics. No conflicts of interest was disclosed by the other author.

M. Ignatiadis received funding from “Les Amis de l'Institut Bordet” (no grant number applicable) and from the “Fondation Contre le Cancer” (Grant No. 2016-124).