Background
In patients with metastatic castration resistant prostate cancer (mCRPC) first-line cytotoxic therapy with docetaxel is standard of care. About 45 % of patients are primary non-responders and tumor progression occurs after a median of 6–8 months [
1‐
4]. Thus, early prediction of treatment efficacy is relevant for optimized and individualized treatment strategies, especially since recently newer agents like abiraterone, enzalutamide and radium-223 were established which may enable sequential treatment strategies and are applicable either prior or secondary to docetaxel [
5‐
7].
During therapy, objective treatment response (TR) is routinely monitored with computed tomography (CT) and by additional bone scintigraphy with T-99 m labeled diphosphonates every four to six cycles of chemotherapy. Thus, the first response evaluation is performed 3–4 months after therapy initiation [
6,
8]. Also, the prostate-specific antigen (PSA) value as a tumor marker requires a treatment interval of about 3 months to reach prognostic significance and is unreliable to reveal treatment response, as is the case with other serum derived markers such as lactate dehydrogenase and alkaline phosphatase [
5,
9‐
12].
Consequently, there is a demand for a reliable predictive surrogate marker as an early indicator of treatment efficacy. A potential blood derived biomarker is the quantitative detection of circulating tumor cells (CTCs) that are highly investigated [
13‐
18]. A US federal Food and Drug Administration (FDA) approved device for CTC-quantification and treatment monitoring in metastatic PC is the CellSearch™-System. Using this device, a threshold of ≥5 CTCs per 7.5 ml blood demonstrated prognostic significance for overall survival (OS) in metastatic PC [
9‐
11].
There is evidence that CTC-counts are an early predictive surrogate marker for objective TR in breast and colorectal cancer [
19‐
22]. Similarly in mCRPC patients elevated pre-treatment CTC-counts are associated with reduced radiologic response rates [
23]. Therefore, we performed a prospective study in mCRPC patients, assessing longitudinal categorical and continuous CTC-count status during therapy in association with response assessments by radiographic RECIST- (Response Evaluation Criteria In Solid Tumors) and clinical criteria. The main objective was to compare the predictive and prognostic value of early categorical and continuous CTC-count status, after one cycle of chemotherapy for objective therapy response (TR), progression free (PFS) and overall survival (OS).
Discussion
In this prospective clinical trial we examined the predictive and prognostic value of early CTC-count status for monitoring treatment efficacy in mCRPC patients. To the best of our knowledge this is the first study that comparatively investigates the predictive and prognostic value of early continuous vs. categorical CTC-count status for therapy response in a homogenous cohort of mCRPC patients, assessing several defined longitudinal time points during chemotherapy with docetaxel. According to our data, categorical CTC-count status relative to a threshold of 5 CTCs is an early predictor of treatment response, already at the end of the first cycle of chemotherapy, 9–12 weeks before the first radiologic objective response assessment. In contrast early continuous CTC-values displayed no early predictive value for therapy response by morphologic RECIST or clinical criteria. The predictive value of early categorical post-treatment CTC-count status for therapy response was similarly observed in metastatic breast and colorectal cancer [
19‐
21,
30]. In our cohort pretreatment CTC-counts were not associated with therapy response. In contrast earlier results in mCRPC-patients revealed an association of elevated baseline CTC-counts (≥5 CTCs) with reduced radiologic and biochemical response while the post-treatment CTC-dynamic revealed an association with PSA-response [
23].
Focusing on the predictive value of the continuous CTC-count status for therapy response, early continuous CTC-values at q1 were not predictive for objective response, although the comparison of median absolute CTC-counts significantly differed in patients with PD vs. non-PD at all time points. The missing predictive value of continuous CTC-counts might be reflected by the CTC-kinetics. After one cycle of chemotherapy early CTC-kinetics revealed a decrease of median CTC-values in non-PD and PD patients alike. According to rTR in non-PD patients the initial median CTC decrease was 80 % (p = 0.002) and in PD patients 60.4 % (p = 0.1) and for cTR 77.7 % (p = 0.002) and 50 % (p = 0.086). However, at later stages during therapy an increase of median CTC-counts was measured only in patients with PD. Therefore, continuous CTC-counts seem to require an interval of up to 12 weeks to reflect therapy response in imaging devices. To our knowledge we are the first to present analyses of continuous CTC-count status in relation to radiologic response, during a defined course of docetaxel in mCRPC patients. Earlier studies investigated the association and predictive value of directional or delta changes of log10 transformed CTCs with clinical response in overall interval-adjusted analyses [
28,
29]. In CRPC-patients mean changes of CTC-counts were not predictive for clinical PD, neither by unidirectional nor by delta changes despite a trend for increasing CTC-counts in patients with PD and relatively unchanged values in non-PD. In this cohort mean changes in CTC-counts did not differ significantly between patients with PD vs. non-PD [
28]. In contrast, investigating a mixed cohort of castration sensitive (CSPC) and mCRPC patients, an average CTC-decrease in non-PD and a CTC-increase in PD patients was observed, with a significant individual predictive value of CTC-value changes for the risk of clinical progression. Concordance analyses of directional CTC-changes with clinical outcome revealed a sensitivity of 79 % and a specificity of 75 % [
29]. Similarly Goodman et al. demonstrated in analyses conducted on log10 transformation of CTC-counts that elevated CTC numbers during treatment were associated with a higher risk for transition from HSPC to CRPC. In this study only baseline CTC-counts were independently prognostic for the risk of CRPC [
31].
Thus, there is an inconsistent body of evidence concerning whether continuous CTC-counts might have predictive value for therapy response. Our results indicate that continuous CTC-counts as a predictive marker require sampling at multiple time points over a period of up to 12 weeks. This is in accordance with a larger study that included different types of cancers, and which demonstrated that about 3 months are needed to monitor the treatment effect on continuously assessed CTCs [
32]. Comparing different concepts of % change in CTCs with a defined % reduction confidence or fold changes with absolute and proportional reduction cutoffs vs. categorical CTC-enumeration demonstrated that a static CTC cutoff is the best method to determine whether a therapy is effective [
32]. Similarly a fold change in CTC-counts was only moderately associated with survival time and proportional changes of continuous CTC-counts with ≥30 % or ≥50 % revealed conflictive results with respect to survival [
11,
23,
27].
Focusing on PFS we could demonstrate that early categorical posttreatment CTC-counts were highly prognostic, as was demonstrated earlier for other carcinomas [
19,
20,
30,
33]. In addition, we demonstrated categorical CTC-counts, after the first cycle of docetaxel, as an independent prognostic marker for PFS. Interestingly, in our study baseline CTC-values did not display prognostic value for PFS. This contrasts earlier results in CSPC patients, where unfavorable categorical baseline and posttreatment CTC-counts were associated with reduced time to castration-refractory stages, but is in accordance with results in metastatic colorectal cancer [
20,
31,
34]. Investigating the continuous CTC-count status as a prognostic tool for PFS, in our study post-treatment CTC-values displayed a significant prognostic value, despite a distinct lower hazard ratio when compared to categorical CTC-counts. Consequently early continuous CTC-counts after one cycle of docetaxel were not independently prognostic for PFS. Therefore early categorical posttreatment CTC-counts seem to be a superior prognostic marker for PFS, when compared to the continuous CTC-value status.
In our study response was primary defined by the objective radiologic response (rTR) performed by RECIST 1.1 criteria and secondary by clinical response (cTR). Both modalities were rated blinded to CTC-assessments. With regard to the selected RECIST 1.1 criteria, we handled bone metastases as non-measurable lesions as all of the bone metastases being present in the patients examined by our study show no measurable soft-tissue component. Patients with non-measurable disease only at baseline were allowed, with a defined PD in case of an increased tumor burden or a substantial worsening as defined by RECIST 1.1 [
26]. This approach is in accordance with earlier studies using RECIST criteria but might be limited due to a potential misinterpretation of a radiographic flare response with sclerosing of osseous metastases as PD. It is well known that under various therapies in bone metastases an increased activity of osteoblasts can occur resulting in an higher density in CT which could be misinterpreted as progressive disease [
11,
31,
35]. However, it has to be taken into account that the interpretation as PD defined by RECIST 1.1 requires an uniequivocal progression with a total increase in tumor burden. An only modest increase is not regarded as enough for PD in non-target lesions. This approach as conducted in our study limits, however not completely omits the incorrect classification of patients with stable disease as PD. In our trial the additionally evaluated clinical response (cTR) revealed similar results when compared to rTR. Similar to earlier studies cTR was assessed mainly based on imaging findings with RECIST evaluations of CT-scans and concomitant bone scans where new lesions (≥2) indicated progression [
26‐
28]. Supplementary, medical assessments including symptomatic progression and performance status were taken into account, and to a lesser extent laboratory studies with increasing PSA, LDH and AP levels indicating progression [
11,
28]. Beside the objective rTR evaluation by RECIST criteria, the cTR assessment in our study may incorporate a subjective bias and represents a limitation. Therefore further studies should apply defined weight bearing percentages for the individual clinical criteria as demonstrated by Gonzales et al. or should consider the recommendations of the prostate cancer clinical trials working group (PCWG2 guidelines) for response assessment [
26,
27,
29].
Focusing on OS, categorical CTC-count status displayed a significant association with OS for each time point during therapy including baseline. Early categorical CTC-counts at q1 were confirmed as an independent prognostic surrogate for OS. Thus, our data complement other studies in which the prognostic value of a defined CTC-threshold for OS has been demonstrated for baseline and 2–5 weeks after therapy induction or during later intervals [
10,
11,
23,
34,
36]. Although the delineated threshold for survival varies across studies from 3 to 5 CTCs [
9‐
11,
24,
31,
34,
36‐
38] in our study the FDA-approved threshold for therapy monitoring of <5 vs. ≥5 CTCs was applicable as a predictor of treatment efficacy in our cohort. Regarding continuous CTC-values our study revealed also the early posttreatment continuous CTC status as an independent prognostic marker for OS at q1 despite a lower hazard ratio when compared to categorical CTC-counts. These results supplement earlier data by Scher et al. demonstrating that elevated continuous CTC-values were associated with a higher risk of death and decreased survival [
27].
As presented in additional files (Additional files
1,
2,
3 and
4) we performed further exploratory analyses investigating early CTC-dynamics from q0 to q1. We could demonstrate the conversion of CTC-counts below the established threshold of 5 CTCs relevant for PFS and OS, as was demonstrated earlier for OS [
10,
11,
23]. Similarly, a 50 % decrease algorithm as a potential measure of continuous CTC-value changes revealed prognostic impact on OS and PFS. Assessing the 50 % decrease, 44 % of the patients reaching the 50 % decline in CTC-counts simultaneously demonstrated a CTC-count decrease below the threshold of 5 CTCs. Our results complement initial trials demonstrating that not only the conversion of CTC-counts to a favorable level of <5 CTCs, but also a percent decrease shows a prognostic value for survival in mCRPC patients as it was demonstrated initially for OS by using a proportional fall of ≥30 % of continuously assessed CTC-counts [
11]. In contrast a recent study demonstrated no significant association of a 50 % decrease with OS [
23]. Thus there is inconsistent evidence regarding the prognostic value of continuous CTC-values for survival whereas the prognostic value of categorical CTC-counts was confirmed in a broad range of trials with respect to OS. Taken together categorical CTC-counts seem to be a superior surrogate marker when compared to continuous CTC-values presenting in our study a higher hazard ratio for survival and a clinically easily applicable threshold with an early predictive value for objective treatment response. In addition categorical CTC-counts displayed an early independent value for PFS and OS, applicable already after one cycle of chemotherapy, up to 3.7 months before the first imaging staging procedures. Therefore, the aim of therapies should be the conversion of CTC-counts to favorable values, as has been suggested also by others [
32]. Limiting our results might be attributed to the small patient cohort and need to be confirmed in a large cohort especially with regard to randomized CTC-guided treatment strategies. In addition further studies need to address the molecular characterization of CTCs as a liquid biopsy [
39‐
42].
Competing interests
The authors have no actual or potential conflict of interest in relation to this article to declare. The study was supported by Sanofi-Aventis, Frankfurt, Germany and the Siegfried Gruber-Foundation, Munich, Germany.
Authors’ contributions
MT and RN analyzed data and wrote the manuscript, VK performed statistical analyses and counseling, MR contributed to concept design, JG and MR directed the study and contributed to data interpretation, MT, MMH and CK performed the research and contributed to data collection, ME, MS and BJK evaluated imaging datasets, MMH edited the manuscript, BR and UA performed the CTC-detection as trained operators. All authors read and approved the final manuscript.