Liquid Biopsy for Cancer Detection: Clinical and Epidemiologic Considerations

In this issue of Clinical Cancer Research, Jeannot and colleagues present a study evaluating circulating DNA (ctDNA) assays for posttreatment monitoring of cervical cancer, suggesting that detection of HPV E7 ctDNA can predict relapse (1). Detection of circulating nucleic acids in blood, also referred to as “liquid biopsy,” has been heralded for some time as a promising approach for a wide range of clinical applications including early cancer detection, predicting cancer outcomes, and posttreatment monitoring (2). Because only few cancer sites can be directly sampled, blood-based detection of tumor markers is intriguing; however, few liquid biopsy applications have become clinical reality so far. This study reports promising results for a potential clinical use of ctDNA for recurrence monitoring and provides an opportunity to highlight key considerations and challenges. To illustrate these points, we contrast the study by Jeannot and colleagues to the application of liquid biopsies for cancer early detection (Fig. 1).

A necessary requirement for a successful liquid biopsy test is a specific tumor signal, like a somatic mutation, methylation, mRNA, or miRNA that is present in a tumor, but not, or to a much lesser extent, in normal tissue, and is released into the bloodstream. Most cancers have different molecular subtypes; achieving high sensitivity for early detection requires coverage of several pathways by an assay. In contrast, recurrence monitoring can focus on molecular changes detected in tumor tissue. However, an important challenge for posttreatment monitoring is the impact of treatment on clonal evolution. If treatment targets a specific pathway driving the primary cancer, clones with different molecular features may be selected. Thus, both for early detection and posttreatment monitoring, the complexity of molecular pathways underlying the targeted tumors needs to be considered. Exceptions include hereditary cancers with a specific germline mutation and cancers initiated by viral oncogenes, which typically have a more homogeneous molecular tumor spectrum. Because almost all cervical cancers are caused by persistent infections with carcinogenic HPV, driving both the initial tumor and recurrences, detecting HPV appears to be an ideal liquid biopsy target.

The performance characteristics of liquid biopsy assays determine their potential clinical application. Sensitivity of an assay is related to its limit of detection, and the amount of nucleic acid released into the bloodstream. Specificity is related to the presence of the marker in other tissues or blood components; because of their abundance, even low levels in normal tissues can limit discrimination from specific tumor signals. In the context of early detection, liquid biopsy approaches may be more likely to detect advanced cancers than early changes (3). Detection at advanced stages, or when a cancer is already clinically manifest, may provide little or no benefit for cancer outcomes. Preinvasive tumors are an ideal target for cancer prevention but typically do not release specific nucleic acids into circulation. A highly sensitive assay may detect lower stage tumors before they become clinically manifest. To have meaningful clinical utility, liquid biopsy markers need high positive predictive value (PPV) for cancer. The PPV is directly related to the specificity of the assay and disease prevalence in the target population (4). Several studies have reported driver mutations in blood from healthy individuals, which may pose a challenge for the development of highly sensitive and specific early detection assays (5).

The situation is different for post-treatment monitoring. Targeting specific molecular changes derived from patient tumors can enhance specificity, and assays may have less stringent analytic sensitivity requirements if tumor load is higher and cells are released into circulation early. An additional benefit of liquid biopsy assays in the setting of posttreatment monitoring is that they can detect recurrences independent of the site (local, regional, distant) while clinical approaches such as imaging are typically more targeted. In the specific example, while HPV DNA is commonly present in transient epithelial infections that have very low risk of cancer, these infections do not lead to release of virus into the bloodstream, guaranteeing high specificity of HPV ctDNA for invasive cancers.

The specific clinical application directly informs the population(s) to be tested. In the general population, cancers are rare and false positive results may have significant negative consequences (4). Focusing on a subset of the population that is at increased risk of cancer (e.g., based on family history or other risk assessment strategies) would increase disease prevalence and specificity. In contrast, the population for post-treatment monitoring is much smaller and disease risk is higher, thus false positive results may be better tolerated. Tumor characteristics such as stage, grade, histology, and molecular profiles, as well as treatment modality, could be used to identify patient populations that would most benefit from liquid biopsy approaches. For any clinical application, it is important to evaluate liquid biopsy assays in the respective target populations, covering the relevant age range, disease spectrum, and demographic features.

The final consideration involves understanding the current clinical approaches and interventions to compare liquid biopsy performance to standard of care. For early detection, this involves a formal comparison of test performance to current practice (considering screening approaches, clinical symptoms, etc.) for cancer detection to estimate the benefit in terms of improved outcomes or survival. Other important considerations include the timing and frequency of testing (i.e., how often and at what interval), as well as the benefits and harms associated with diagnostic workup procedures. These considerations also apply in the context of post-treatment monitoring. For example, as suggested by Jeanott and colleagues, a trial comparing the performance of HPV ctDNA testing to standard imaging approaches is necessary to fully evaluate the benefits of liquid biopsy for detection of cervical cancer recurrence. Another key consideration for posttreatment monitoring is the availability of beneficial second-line therapies, for example, such as salvage surgery for patients with residual disease as noted by the study authors.

In summary, while liquid biopsies for cancer detection are intriguing and deserve our attention, various challenges need to be overcome before these tests become a clinical reality. The considerations depend on the specific application of the assay. As with all clinical interventions, it is important to weigh the benefits and harms of new approaches. Evaluation of an early detection assay that targets the general population needs to be particularly concerned about the consequences of false positive results and needs to demonstrate a clinical benefit of cancer detection. Because the population is much smaller and disease risk is higher in a posttreatment setting, the benefit–risk ratio may be more favorable for new tests. The study by Jeannot and colleagues demonstrates a potential clinical benefit of detecting HPV ctDNA to monitor recurrences. As outlined above, due to its high specificity for the underlying tumor, HPV ctDNA represents an ideal target for detecting cervical cancer recurrences. Yet, even in this ideal setting, only 63% of patients had detectable HPV ctDNA before treatment, suggesting that only a subset of patients could be monitored with HPV ctDNA. We concur with the authors that clinical studies are needed comparing the liquid biopsy approach to currently established standards to fully understand the clinical potential of their elegant approach. Importantly, we focused only on two liquid biopsy applications. Other approaches for molecular profiling, tumor prognosis, and monitoring clonal evaluation for adaptive treatments, require consideration of other opportunities and challenges (2).

No disclosures were reported.

This work was supported by the Intramural Research Program of the NCI (ZIA CP010152 00030).