Molecular Monitoring in Endometrial Cancer—Ready for Prime Time?

In this issue of Clinical Cancer Research, Ashley and colleagues expand the current knowledge of cell-free DNA (cfDNA) as a diagnostic tool, prognostic indicator, and surveillance measure in endometrial cancer (1). This builds on a rapidly growing evidence base for the use of liquid biopsies in cancer and is timely for endometrial cancer given the rapid use of molecular therapeutics in this disease. In contrast to other disease sites, the diagnosis, recurrence, and progression of endometrial cancer is not accurately predicted by the trend of a biochemical marker (1, 2). One such example of this is the use of CA-125 as a blood-based tumor marker in a subset of patients with ovarian cancer, which is less reliable in the endometrial cancer population (3). Disease monitoring in endometrial cancer therefore currently relies upon serial radiographic imaging and clinical assessments. This highlights an important role for the liquid biopsy as a potentially sensitive and specific clinical instrument for earlier detection of endometrial cancer at diagnosis, recurrence, and progression (3). Here, we review where cfDNA may supplement current gaps in endometrial cancer management and identify challenges that remain in delivering molecular monitoring to the clinic.

Endometrial cancer is the sixth most common female malignancy worldwide and the most common gynecologic malignancy in high-income countries (4). Over the past four decades, the incidence has been increasing and endometrial cancer resulted in nearly 100,000 deaths globally in 2020 (4). While most low-risk patients diagnosed with early-stage disease can be cured with surgery alone and have good prognoses, the 5-year relative survival for patients diagnosed with advanced disease of stage II or higher is poor and has not significantly improved since 1975 (1, 5). Targeted therapeutics have emerged as an area of research interest, particularly with the increasing availability of high-throughput DNA sequencing technologies and better understanding of the mechanisms of cancer progression (6). With the limitations of existing cancer screening and surveillance tools, minimally invasive strategies such as cfDNA for disease detection and monitoring remains an important research focus.

In the current study, investigators correlated the mutational landscape of primary endometrial tumors with circulating tumor DNA (ctDNA) levels—the tumor-derived fraction of cfDNA—and longitudinally monitored patients for a median time of 33 months (1). They first aimed to determine if ctDNA levels were associated with histologic subtype or disease stage in patients with a new endometrial cancer diagnosis. Tumor tissues from 42 patients were sent for MSK-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT) sequencing of 468 cancer-related genes. Of these patients, 36 had a pre- and/or postoperative blood draw with the required plasma cfDNA input. Tumor-derived mutations were found in cfDNA of 8 patients and 6 of these patients were identified as having stage III or IV disease. The results demonstrated significant correlation between the quantity of ctDNA in plasma and surgical stage. There was also correlation, although not significant, between increased ctDNA detection and endometrial cancers with more aggressive histologic types. In accordance with these data, Bolivar and colleagues previously correlated the presence of plasma ctDNA mutations with stage at the time of debulking surgery in 48 patients with endometrioid endometrial cancer (2). Fifteen patients (33%) had at least one matched mutation detected in both the tumor and plasma ctDNA which included a panel of CTNNB1, KRAS, PTEN, and PIK3CA genes. Presence of a plasma mutation in these 15 patients was significantly associated with advance stage at hysterectomy, lymphovascular space invasion (LVSI), which is an indicator of poor prognosis, and primary tumor size (2). This study did not follow the postoperative course of patients to analyze time course of mutational clearance from blood or mutation re-emergence as a prediction of recurrence. Chicchillitti and colleagues previously reported elevated cfDNA levels in patients with endometrial cancer with grades 2 and 3 histology when compared with those with grade 1 histology (7). This also aligns with results from Vizza and colleagues who reported significantly higher levels of cfDNA in high-grade endometrial cancer histology (6). They also reported increased cfDNA integrity (an index of the ratio between long and short cfDNA fragments) in patients with LVSI. These studies support the increasing body of evidence that cfDNA may function in risk stratification and treatment de-escalation for patients with early-stage endometrial cancer when the role of adjuvant therapy is not clearly defined.

Beyond its use as a diagnostic tool and prognostic indicator, Ashley and colleagues profiled longitudinal cfDNA samples to identify if serial monitoring can detect minimal residual disease (MRD) for disease monitoring and if ctDNA detection is associated with disease recurrence or progression. Plasma samples were collected every 6 months and samples from the 25 patients with plasma ctDNA detected at recurrence and/or post-surgery underwent targeted sequencing. The identified mutations and change in mutant allele fraction closely followed disease progression in this population.

Many unanswered questions exist in the management of endometrial cancer and these results postulate that ctDNA measurements may one day supplement such clinical conundrums. One example lies in the management of patients with isolated tumor cells (ITC). Interestingly, in the referenced study, ctDNA was detected in a stage 1A high-grade adenosarcoma and stage 1B high-grade mixed serous and endometrioid carcinoma with ITCs and LVSI. No other low-grade stage 1 patients with endometrioid endometrial cancer had detectable preoperative ctDNA. Management strategies for ITCs currently remains controversial. A large multi-institutional retrospective study published in 2021 assessed the role of adjuvant treatment in patients with endometrial cancer with low volume metastasis in sentinel lymph nodes (8). In 247 analyzed patients, 132 (53.4%) had ITCs and 115 (46.6%) had micrometastasis (MM). Among all patients, 38 (17 in the ITC group and 21 in the MM group) recurred within 4 years post-surgery. This prompts us to ask whether cfDNA sequencing may help risk stratify these patients. It is also established that localized recurrence of endometrial cancer can be cured with salvage therapy in select cases. In a 2019 study, a cohort of 2,691 patients with endometrial cancer were analyzed, 91% with endometrioid histology. Of these patients, 61% were stage IA and 57% had grade 1 morphology. Within the median follow-up time of 6.1 years, 194 patients (7.2%) experienced recurrence and 99 patients (3.7%) had local regional recurrence (LRR). Patients with LRR had a significantly longer median overall survival (OS; 14.0 years) than those with pelvic and distant recurrences (1.2 and 1.0 years, respectively; ref. 9). All patients with LRR who received salvage radiotherapy, surgery, and chemotherapy had significantly improved OS. After initial LRR and salvage therapy, 40.4% of patients did not have a second failure during the 5.9-year post-recurrence follow-up period. Early detection of LRR with more sensitive MRD detection using cfDNA analysis may offer opportunities for salvage therapy before distant failure occurs.

Several completed and ongoing studies across other disease sites are investigating correlations between cfDNA/ctDNA levels, early detection, and disease monitoring (10). Studies of lung cancer, for example, are diving deeper into molecular monitoring of treatment response and drug resistance. Tumors contain heterogeneity both within primary tumors and between primary tumors and sites of metastasis, which generates susceptibility for adaptation and development of drug resistance (11). The BENEFIT trial (NCT02282267) showed high rates of de novo EGFR mutations in serial plasma cfDNA samples of patients with lung cancer that were not present in somatic tissue, suggesting a molecular-based method of measuring acquired resistance that is less invasive and perhaps more informative than re-biopsy (12). This may be applied to treatment planning and guiding drug development to ensure patients receive individualized therapy that is selected to target specific pathways driving tumorigenesis (11).

There remain significant challenges in the technical aspect of bringing the liquid biopsy to clinical practice. Cells within the tumor and tumor microenvironment release DNA that harbors pathogenic driver mutations and noncancerous DNA, which are both incorporated within a given cfDNA sample. False positivity can thus be influenced by tumor-related mutations released from normal cells (10). Different timepoints within the course of a single disease may also be characterized by varying fractions of tumor DNA shedding. In addition to this, patients with early-stage cancer or low volume disease shed less tumor DNA into plasma, which makes detection of molecular markers challenging and affects the sensitivity of sequencing methods (13, 14). A 2022 review proposed three stages which are needed prior to regulated use of molecular biomarkers: (i) validation to ensure assays are reproducible and reliable, (ii) a prospective feasibility trial including a small sample size of the intended-use population to assess sensitivity, specificity, and positive and negative predictive values, and (iii) a prospective pivotal study in the intended-use population with a large sample size to assess efficacy, safety, and cost–benefit analysis (14). It is important to acknowledge and address the potential for sensitive early detection methods to result in lead-time bias if used to evaluate outcomes from time of molecular recurrence. This can occur when testing increases the perceived survival time without affecting the course of disease or OS (Fig 1; refs. 13, 15). Clinical trials have begun to observe mortality from the date of randomization rather than survival to eliminate differential lead time, however this cannot reasonably be translated into real-world studies (15).

The ability to detect disease at a molecular level opens new opportunities to redefine treatment success following surgery, radiation, systemic, or immunotherapy, as well as the development of new strategies to strive towards cure integrating and sequencing therapies with clarity of measured and quantitative effect. While targeted approaches have been shown to be highly effective for monitoring active or high-burden cancers, the limit of detection is bounded by the number of mutant DNA molecules available for detection from a blood sample. Therefore, reliance on relatively few targets with potential mutations may not deliver the sensitivity necessary to detect MRD or establish true disease clearance. To this end, genome- and epigenome-wide approaches that integrate mutation, copy number, and/or methylation analysis of thousands of sites have shown sensitivities orders of magnitude lower than gene sequencing alone (16–18). However, it remains to be shown in endometrial cancer whether this level of sensitivity is necessary to establish a clinically meaningful measurement of MRD. Other biospecimens such as urine may also yield informative levels of ctDNA and be more readily detected using cfDNA sequencing techniques (19). Studies such as Ashley and colleagues’ demonstrate that ctDNA levels in blood closely mirror the clinical course in endometrial cancer. The question remains—How low do we need to go?

Moving forward, if we can reduce bias and develop a validation tool to use cfDNA in the appropriate clinical context, only then will there be promise for its implementation in clinical care. The Cancer Genome Atlas has defined four molecular endometrial cancer subtypes: POLE ultra-mutated, microsatellite instability high, copy number low, and copy number high (20). Future studies that include and stratify larger patient populations within each subtype may help define a clearer path forward, especially as we move toward individualized targeted treatment in cancer care. If cfDNA evolves into a validated, sensitive, and specific molecular marker, it may also forge a path for more accurate risk-stratification and precise molecular monitoring in high-risk populations such as patients with early-stage disease or atypical endometrial hyperplasia receiving hormonal therapy for fertility preservation. Although it remains to be seen if implementing cfDNA analysis to routine clinical follow-up will have practice changing implications, this proof-of-principle study takes important strides in the effort to define the role of cfDNA in endometrial cancer. We hope this paper will encourage investigators of prior first-line and recurrent endometrial cancer studies to publish cfDNA data where available to put previously analyzed data into greater context as we determine how it may best be used in practice.