Circulating Tumor DNA Biomarkers for Response Assessment in Prostate Cancer

In this issue of Clinical Cancer Research, Tolmeijer and colleagues (1) have published an important new finding regarding circulating tumor DNA (ctDNA) fraction and response in metastatic castration-resistant prostate cancer (mCRPC). Why is this important and what do we now understand better as a consequence of these data?

Metastatic prostate cancer is notorious for being a bone dominant disease and bone metastatic disease is especially difficult to assess for response using conventional imaging. In the typical scenario, bone lesions are assessed by bone scans (bone scintigraphy) for progression but not response. Lymph nodes are involved with prostate cancer a substantial proportion of time but typically these nodes are either not measurable or barely measurable. Thus, measuring responses in mCRPC by conventional radiographic criteria is frequently problematic (and typically slow).

The advent of PSA testing in the 1990s was met with great enthusiasm. There was a perception that declines in PSA could be used for reliable assessment of both response and progression. Today, using guidelines from the Prostate Cancer Working Group (2) rates of various PSA declines are categorized in virtually all advanced prostate cancer trials. Though PSA declines are important, by no means do they entirely reflect the course of the disease. Therapies such as sipuleucel-T and radium-223 dichloride demonstrate improvements in overall survival (OS) in large randomized trials but have little effect on PSA declines (3, 4). Further, studies of PSA decline and progression are imperfectly correlated with survival in numerous large mCRPC trials. A recent example oof PSA imperfection was demonstrated in the randomized TheraP trial (5) where treatments with lutetium-177-PSMA-617 (PSMA-Lu-177) and cabazitaxel were compared. In terms of PSA declines, the PSMA-Lu-177 performed superiorly to cabazitaxel, but in terms of OS, they were nearly identical.

Taken together it is important to assess new parameters for therapeutic responses (and resistance) in mCRPC. Ideally, these parameters would have relatively rapid changes so that patient's exposure to ineffective therapies would be limited. Clearly, there is no benefit when patients are treated with ineffective interventions.

In most patients without cancer, circulating free DNA (cfDNA) is predominantly is derived from white blood cells (6). In patients with cancer, some percentage of the cfDNA may (or may not) be derived from cancerous cells, this is termed ctDNA. The proportion of ctDNA/cfDNA can be used to calculate the percentage of ctDNA.

The manuscript by Tolmeijer and colleagues (1) describes 81 patients with mCRPC treated with abiraterone (N = 51) or enzalutamide (N = 30). These patients were part of two prospective observational studies conducted in the Netherlands. In each instance these patients had no prior exposure to abiraterone or enzalutamide or docetaxel. Plasma was collected at baseline and 4 weeks after initiation of therapy. The ctDNA fraction was calculated using a combination of somatic mutations determined by targeted sequencing and genome copy number profiles. The ctDNA was initially classified into detectable and undetectable fractions. Clinical outcomes were assessed by both progression-free survival (PFS) and OS. The median follow-up was 27 months and the majority of patients (64%) enrolled in the studies had died. Thus, follow-up was reasonably mature (an important aspect of this study). Too often immature analyses are used in these types of investigation.

At baseline the majority of patients (59%) had detectable baseline ctDNA. Four weeks after treatment, not surprisingly the typical ctDNA fraction was lower and the number of patients with detectable ctDNA was lower (36%). The relationship between ctDNA and PFS and OS was striking. For those with detectable versus undetectable ctDNA at baseline, statistically significant findings in PFS and OS were reported; the median PFS was 5.8 versus 20.2 months and the median survival was 22.7 months versus not reached. On the basis of these data, there is clear evidence that baseline ctDNA serves as a prognostic marker in these abiraterone/enzalutamide-treated patients with mCRPC.

Certain men treated with abiraterone or enzalutamide had a detectable ctDNA prior to therapy and an undetectable ctDNA 4 weeks later. For these men the median PFS was 15.6 months compared with a PFS of 4.8 months for men whose ctDNA was detectable both at baseline and 4 weeks after starting treatment. OS in these two groups was also strikingly distinct, 27.7 versus 16.0 months. These findings are important. When assessing the ctDNA changes in the context of other variables of interest, a multivariate model demonstrated the statistical significance of the ctDNA findings. Additional analyses looked at durable versus nondurable responses to abiraterone or enzalutamide. Compared with PSA responses, the ctDNA changes 4 weeks after treatment started were superior in predicting duration of response.

These data indicate the importance of ctDNA both in determining prognosis (at baseline), and for changes in ctDNA 4 weeks posttreatment. Therewas clear data to associate these changes with duration of PFS and OS. Exploratory multivariate analyses emphasized that ctDNA were independent of other blood-based biomarkers. Taken together, these studies demonstrate the importance of ctDNA in terms of rapidly predicting PFS and OS after hormonal treatments in patients with advanced mCRPC.

Additional studies suggest that ctDNA can be used both for mCRPC prognosis and response after either hormonal or chemotherapy treatments (7, 8). Thus, broadly, ctDNA findings appear to be independent of treatment modality. More studies will be necessary to substantiate this broad claim. Regardless, ctDNA provides a rapid assessment of patients undergoing mCRPC therapy and provides clinicians with a pragmatic strategy to improve the care of patients.

The Tolmeijer and colleagues data did not delve deeply into specific ctDNA findings but it is known that not all genetic alterations are created equal. More data are required to evaluate the functional importance of individual gene changes in ctDNA but studies from Jayaram and colleagues (8) suggest that patients with persistent TP53, RB1, or PTEN gene alterations posttreatment have a significantly shorter survival as compared with those patients with no such alteration. Further work is need to calibrate early changes in specific gene alterations and outcomes. It is clear that therapies may selectively alter some but not all ctDNA changes and the consequences of eliminating certain “drivers” from a heterogenous tumor are likely important.

Other novel strategies are being developed for the rapid ascertainment of prognosis and response to treatment. Measuring treatment response by using ctDNA methylation markers may provide additional insights over current ctDNA assays which are gene-centric (9). Assessments of circulating tumor cells (CTC), their contents, and CTC morphology can potentially provide rapid and important clinical information (10, 11). Extracellular vesicles (EV) and their cargoes contain a multiplicity of information relevant to prostate cancer prognosis and treatment (12). In addition, molecular imaging techniques including both PET and single-photon emission computed tomography are now providing rapid and quantitative prognostic and predictive information after PSMA Lu-177 treatments (13, 14). A new method of assessing DNA damage using PET imaging of PARP expression suggests the ability of monitoring the effects of DNA damaging therapy in near real time (15). Undoubtedly, more work will follow.

In addition to providing important information on prognosis and response/resistance. Biomarkers can also provide important value regarding selection of therapies. In these case biomarkers can be predictive not just prognostic. Clear cases in point include the use of BRCA2 mutations in selection of patients for PARP inhibition (16) or the PSMA PET selection of patients for PSMA Lu-177 therapy (17). Perhaps combinations of serum proteins, ctDNAs, CTCs, EVs, and novel molecular imaging will be even better than any of these assessments alone (Table 1).

As clinicians better and more rapidly understand response and resistance to anticancer therapies, superior management strategies are likely to emerge. With earlier detection of resistance, exposures to ineffective treatments can be minimized. Thus, these studies provide hope that by detecting drug resistance earlier, clinicians can take advantage of this knowledge by introducing alternative therapeutic strategies sooner. Whether or not earlier interventions will provide better patient outcomes is a question best resolved by clinical trialists. This question is worthy of an answer.

O. Sartor reports grants and personal fees from Advanced Accelerator Applications, Amgen, AstraZeneca, Bayer, Janssen, and Sanofi; personal fees from ART BioScience, Astellas Pharma, Clarity Pharmaceuticals, EMD Serono, Fusion Pharmaceuticals, Isotopen Technologien, MacroGenics, Novartis, Pfizer, Point Biopharma, Ratio, Telix Pharmaceuticals, and TeneBio; and grants from InVitae, Merch, and Lantheus outside the submitted work. No other disclosures were reported.

Dr. Sartor is supported in part by the NCI (180068).