Baseline neutrophil-to-lymphocyte ratio as a predictive and prognostic biomarker in patients with metastatic castration-resistant prostate cancer treated with cabazitaxel versus abiraterone or enzalutamide in the CARD study

Background There is growing evidence that a high neutrophil-to-lymphocyte ratio (NLR) is associated with poor overall survival (OS) for patients with metastatic castration-resistant prostate cancer (mCRPC). In the CARD study (NCT02485691), cabazitaxel significantly improved radiographic progression-free survival (rPFS) and OS versus abiraterone or enzalutamide in patients with mCRPC previously treated with docetaxel and the alternative androgen-receptor-targeted agent (ARTA). Here, we investigated NLR as a biomarker. Patients and methods CARD was a multicenter, open-label study that randomized patients with mCRPC to receive cabazitaxel (25 mg/m2 every 3 weeks) versus abiraterone (1000 mg/day) or enzalutamide (160 mg/day). The relationships between baseline NLR [< versus ≥ median (3.38)] and rPFS, OS, time to prostate-specific antigen progression, and prostate-specific antigen response to cabazitaxel versus ARTA were evaluated using Kaplan–Meier estimates. Multivariable Cox regression with stepwise selection of covariates was used to investigate the prognostic association between baseline NLR and OS. Results The rPFS benefit with cabazitaxel versus ARTA was particularly marked in patients with high NLR {8.5 versus 2.8 months, respectively; hazard ratio (HR) 0.43 [95% confidence interval (CI) 0.27-0.67]; P < 0.0001}, compared with low NLR [7.5 versus 5.1 months, respectively; HR 0.69 (95% CI 0.45-1.06); P = 0.0860]. Higher NLR (continuous covariate, per 1 unit increase) independently associated with poor OS [HR 1.05 (95% CI 1.02-1.08); P = 0.0003]. For cabazitaxel, there was no OS difference between patients with high versus low NLR (15.3 versus 12.9 months, respectively; P = 0.7465). Patients receiving an ARTA with high NLR, however, had a worse OS versus those with low NLR (9.5 versus 13.3 months, respectively; P = 0.0608). Conclusions High baseline NLR predicts poor outcomes with an ARTA in patients with mCRPC previously treated with docetaxel and the alternative ARTA. Conversely, the activity of cabazitaxel is retained irrespective of NLR.


INTRODUCTION
Life-extending therapeutic options for patients with metastatic castration-resistant prostate cancer (mCRPC) have expanded in recent years from docetaxel chemotherapy to include cabazitaxel, androgen-signaling-targeted inhibitors (abiraterone and enzalutamide), radium-223, and sipuleucel-T. More recently, poly (ADP-ribose) polymerase inhibitors (olaparib and rucaparib) have become available for patients with DNA damage repair abnormalities. [1][2][3] Docetaxel and androgen-signaling-targeted inhibitors also significantly prolong overall survival (OS) in metastatic hormone-sensitive prostate cancer, in combination with androgen deprivation therapy (ADT). 1,3 Apalutamide, enzalutamide, and darolutamide are also approved for non-mCRPC. 1,3 Although there have been advancements in determining optimal sequences of the many available agents for advanced prostate cancer, some uncertainties remain. It is now clear that patients who progress on abiraterone respond poorly to enzalutamide, and vice versa. This is likely because androgen-signaling-targeted inhibitors target the same pathway and share common resistance mechanisms. 2,[4][5][6] Results of the CARD study also indicate that patients who have received docetaxel and progressed within 12 months on abiraterone or enzalutamide should receive cabazitaxel over a second alternative androgensignaling-targeted inhibitor. 6 Cabazitaxel was associated with improved radiographic progression-free survival (rPFS), OS, prostate-specific antigen (PSA) response, and tumor response versus abiraterone or enzalutamide. 6 A deeper understanding of the patient and tumor characteristics impacting treatment response is necessary, however, to inform optimal use and timing of available therapies. 7 Much research has focused on the development of accurate prognostic biomarkers beyond conventional indicators such as stage, PSA level, Gleason score, and metabolic factors. 8 Tumor-promoting inflammation is an enabling characteristic that underlies several hallmarks of cancer. 9 Inflammation plays a crucial role in tumor development and metastatic progression in many cancers, including prostate cancer. 10,11 Inflammation can also affect immune surveillance and responses to therapy. 10,11 The neutrophil-to-lymphocyte ratio (NLR), a measure of the absolute neutrophil count (ANC) divided by the absolute lymphocyte count (ALC), is a biomarker of this systemic inflammation that is easy and inexpensive to measure in daily practice. NLR is a biomarker that integrates assessment of systemic inflammation as well as lymphocyte counts that are associated with the adaptive immune response to disorders ranging from infections to cancer. 12,13 High NLR is associated with poor OS in many non-metastatic and metastatic solid tumors. 14 In addition, high NLR has been associated with an increased risk of many nonneoplastic disorders, and has been shown to be predictive of the risk of developing cardiovascular disease. 15 There is also growing evidence that high NLR is associated with poor OS in mCRPC. [16][17][18][19][20][21][22][23][24][25][26] Although NLR is a continuous variable and an optimal cut-off for high NLR is not yet defined, retrospective analyses of phase III randomized studies and several meta-analyses have identified elevated baseline NLR as a strong indicator of prognosis in mCRPC. 16,18,21,[24][25][26] A recent meta-analysis involving 32 studies and 21 949 participants with prostate cancer found that high baseline NLR, neutrophil, and monocyte counts predict worse OS. 24 The significant association between NLR and survival was maintained in a subgroup analysis including patients with mCRPC only, with the authors concluding that NLR may be a predominant prognostic factor in patients with mCRPC. 24 Prognostic models for OS for patients with mCRPC have been developed with applications in clinical practice and clinical trial design. [27][28][29] These models include parameters such as performance status, lactate dehydrogenase (LDH), PSA, albumin, hemoglobin, alkaline phosphatase, and presence of bone metastases. [27][28][29] In addition to these traditional variables and treatment, NLR (<2.5 versus !2.5) was identified as a novel independent prognostic factor in the minimally symptomatic mCRPC setting by a validated model developed using the phase III PREVAIL database. 29 As the PREVAIL study evaluated the efficacy and safety of enzalutamide versus placebo in chemotherapy-naive patients with mCRPC, this model reflects current clinical practice and points toward a role for NLR in risk stratification at diagnosis. 29 The aim of this analysis was to investigate NLR as a predictive and prognostic biomarker in patients enrolled in the CARD study. Analyses investigating the association between baseline NLR and rPFS were prespecified; however, the CARD study was not powered on these analyses.

Study design
CARD was a multicenter, multinational, open-label study designed to compare cabazitaxel with abiraterone or enzalutamide in patients with mCRPC previously treated with docetaxel who progressed on the alternative androgensignaling-targeted inhibitor (abiraterone or enzalutamide). Overall, 255 patients were randomized 1 : 1 to receive cabazitaxel (25 mg/m 2 every 3 weeks plus prednisone daily and granulocyte colony-stimulating factor) versus abiraterone (1000 mg plus prednisone daily) or enzalutamide (160 mg daily) until radiographic progression, unacceptable toxicity, start of subsequent treatment, or patient request to discontinue treatment. The full study design, eligibility criteria, and efficacy and safety results have been described previously. 6 The study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines and all patients provided written informed consent. The study is registered with clinicaltrials.gov, number NCT02485691.

Procedures and outcomes
Hematology testing, which included ANC, ALC, and the calculation of NLR (ANC divided by ALC), was part of routine monitoring during the CARD study. The primary outcome of the CARD study was imaging-based PFS (also known as rPFS), defined as the time from randomization to radiologic tumor progression (assessed using RECIST 1.1) or progression of bone lesions [assessed using the Prostate Cancer Clinical Trials Working Group (PCWG2) criteria], or death from any cause. Secondary outcomes included OS, PSA response, and time to PSA progression. OS was defined as the interval between the date of randomization and the date of death from any cause. PSA response was defined as a decline in serum PSA of !50% from baseline confirmed by a second measurement at least 3 weeks later, in patients with a baseline PSA of !2 ng/ml. Time to PSA progression was calculated as the time from randomization to first documented PSA progression, which was defined in patients with a decline in PSA from baseline as an increase of !25% (at least 2 ng/ml) from the lowest value, confirmed at least 3 weeks later. In patients without a decline in PSA from baseline, PSA progression was defined as an increase of !25% (at least 2 ng/ml) from the baseline value following 12 weeks of treatment, confirmed at least 3 weeks later.

Statistical analysis
The CARD study was designed to have 80% power to detect a hazard ratio (HR) of 0.67 (cabazitaxel versus an androgensignaling-targeted inhibitor) in the analysis of the primary endpoint (rPFS) with the use of a stratified log-rank test at a two-sided alpha level of 5%. Approximately 234 patients needed to be randomized to collect data on 196 events (achieved at the cut-off date of 27 March 2019). All analyses reported here were carried out using data obtained at this cut-off date. Analyses were carried out in the intention-totreat population. Baseline NLR was a prespecified prognostic factor in the CARD study; however, the analyses reported here were not powered on this parameter. Analyses investigating the association between baseline NLR (< versus ! median) and rPFS were planned in the protocol (see Supplementary Figure S1, Tables S1-S3, available at https://doi.org/10.1016/j.esmoop.2021.100241). All other analyses were post hoc.
The treatment effects of cabazitaxel versus abiraterone or enzalutamide on OS, rPFS, PSA response, and time to PSA progression by NLR subgroup were investigated using KaplaneMeier survival estimates. The prognostic association between baseline NLR (as a continuous variable) and OS was investigated using stratified multivariate Cox regression with stepwise selection of covariates, including adjustment for treatment arm. In total, 14 pre-planned prognostic factors were evaluated. Categorical covariates were M1 disease, visceral metastases, prior therapy with curative intent (radical prostatectomy or radiation therapy for localized disease), type of progression, and Gleason score 8-10 at diagnosis, and continuous covariates were hemoglobin, duration of first ADT, NLR, neutrophil counts, age, testosterone, and log 10 -transformed LDH, alkaline phosphatase, and PSA. The significance thresholds for entry into and removal from the model were 0.10 and 0.05, respectively.
Exploratory analyses were conducted using Receiver Operating Characteristic (ROC) curves to estimate the optimal NLR threshold for the prediction of OS. As there was no strong evidence from the ROC analyses suggesting an optimal threshold, the median NLR (3.38) was selected as the cut-off threshold.
Pre-specified analyses were conducted using SAS version 9.2 and all post hoc analyses were conducted using SAS version 9.4.

Patient population
A total of 255 patients with mCRPC were randomized to receive cabazitaxel (n ¼ 129) or a second androgensignaling-targeted inhibitor (n ¼ 126) (Supplementary Figure S1, available at https://doi.org/10.1016/j.esmoop. 2021.100241). At the cut-off date, median follow-up was 9.2 months and median duration of treatment was 22.0 weeks (range 13.1-30.4 weeks) with cabazitaxel compared with 12.5 weeks (range 9.9-23.4 weeks) with abiraterone or enzalutamide. Patient demographics and clinical characteristics have been reported previously. 6 Briefly, 31% of patients were aged !75 years, 18% had visceral metastases, and most patients had pain progression at randomization (69%; Table 1). Median baseline neutrophil and lymphocyte counts were 4.48 Â 10 9 /l and 1.3 Â 10 9 /l across treatment arms, respectively, and the median baseline NLR was 3.38. Overall, patients with high NLR at baseline had higher PSA and LDH values compared with patients with low NLR at baseline (Table 1).

Treatment exposure by NLR subgroup
The treatment durations and frequency of dose modifications associated with cabazitaxel and abiraterone/enzalutamide for each NLR subgroup are shown in Supplementary

DISCUSSION
Although the prognostic value of baseline NLR has been the subject of debate across several malignancies, it is now accepted that baseline NLR is prognostic in mCRPC. 11 A greater understanding of the predictive value of NLR, however, is necessary to guide decision making in the context of the many available treatment options. The CARD study established the superiority of cabazitaxel over abiraterone or enzalutamide in patients with mCRPC who had previously received docetaxel and an alternative androgen-signaling-targeted inhibitor. 6 Evaluating the prognostic and predictive value of NLR in such a study was thus of great interest. The present findings may be summarized as follows: first, high NLR was prognostic when all arms were combined, but analysis by treatment arm showed that NLR was prognostic only in patients treated with abiraterone or enzalutamide, and not in those treated with cabazitaxel. Patients with high NLR at baseline who received abiraterone or enzalutamide had a 4-month lower median OS compared with patients with low NLR (9.5 versus 13.3 months), whereas patients treated with cabazitaxel with a high NLR had a numerically longer median OS than those with a low NLR (15.3 versus 12.9 months). Second, high NLR, compared with low NLR, was associated with reduced antitumor activity of abiraterone or enzalutamide in terms of rPFS and PSA response. Conversely, although cabazitaxel was superior to abiraterone or enzalutamide regardless of low or high NLR, the clinical antitumor activity of cabazitaxel appeared numerically greater in patients with high NLR.
This association between high NLR and poorer responses to androgen-signaling-targeted inhibitors seen in the CARD population is consistent with several retrospective studies. 17,19,20,22,23 Leibowitz-Amit et al. 17 first reported that NLR >5 was associated with a poor PSA response and shorter OS in two independent cohorts of patients treated with abiraterone at the Princess Margaret and Royal Marsden Cancer Centers. Other investigators have reported similar findings, either with abiraterone or enzalutamide. 19,20,22,23 In a cohort of 193 patients with mCRPC treated with enzalutamide after docetaxel, patients with baseline NLR >3 had a median OS of 10.4 months, compared with 16.9 months in patients with baseline NLR 3. 19 A post hoc analysis of the pivotal phase III COU-AA-302 study conducted in 1088 asymptomatic or minimally symptomatic chemotherapy-naive patients with mCRPC also identified high baseline NLR (>2.5) as a predictor of shorter OS, rPFS, and PSA-PFS in patients receiving abiraterone. 23 In contrast, patients enrolled in CARD who received cabazitaxel consistently had improved clinical outcomes compared with patients who received an androgensignaling-targeted inhibitor, regardless of baseline NLR. Patients with high baseline NLR had higher baseline ANC values compared with those with low baseline NLR; however, this was not associated with an increase in cabazitaxel dose intensity. To our knowledge, this is the first analysis to demonstrate that treatment with cabazitaxel may counteract the poor prognosis associated with high baseline NLR in patients with advanced prostate cancer.
Given that patients with mCRPC and high NLR have poor outcomes when treated with abiraterone or enzalutamide, this biomarker can be used to identify patients in whom treatment with a second androgen-signaling-targeted inhibitor should be avoided. Furthermore, cabazitaxel may counteract the adverse prognosis associated with high NLR, comparable with how trastuzumab reverses the poor prognosis associated with HER2 positivity. 30 As NLR can be measured during a simple blood test, NLR status could be an easy and accessible way of determining the optimal next therapy in patients with mCRPC who progress on docetaxel and an initial androgen-signaling-targeted inhibitor.
In cancer patients, baseline NLR is likely a biomarker of systemic inflammation, which has emerged as a key mechanism for tumor progression. 11 Although NLR is measured in the peripheral blood, it is the levels of leukocytes in the tumor microenvironment that are the crucial factor. However, it is possible that elevated baseline NLR reflects either a decreased lymphocyte count or an increased myeloid cell count, or both. There is increasing     ESMO Open evidence that tumor-associated myeloid cells drive tumor progression through multiple mechanisms. 11,31,32 Neutrophils can suppress CD8þ T-cell activation to facilitate metastasis. 33 In cancer, neutrophils are also often associated with myeloid-derived suppressor cells (MDSCs), which also contribute to an immunosuppressive microenvironment. 32 It has been shown that androgen deprivation leads to subsequent MDSC infiltration and interleukin-23 production, fostering mCRPC development. 34 The findings that, in solid tumors and hematologic malignancies, grade !3 neutropenia or leucopenia during chemotherapy is consistently associated with improved OS, may also reflect the role of neutrophils in promoting cancer progression, although the effects of increased cytotoxic drug exposure should also be taken into consideration. 35 In a randomized study comparing docetaxel with mitoxantrone in 228 Chinese patients with mCRPC, docetaxel was associated with a much greater OS benefit (an increase of 8.2 months) than observed in the randomized TAX327 study, conducted in 1006 Caucasian patients (an increase of 2.4 months). 36,37 The increased OS benefit seen in the Chinese study was associated with an increased rate of grade !3 neutropenia (58% versus 32%). 36 38 There are several important limitations of the present analysis. Firstly, prior administration of corticosteroids that induce the release of neutrophils and suppress lymphocyte counts, thereby altering NLR, may represent a confounding factor. 21 Secondly, the CARD study was powered for rPFS, but was not powered for the secondary endpoints and post hoc analyses reported here, due to the relatively small number of patients enrolled in CARD (n ¼ 255). Third, patients in CARD progressed with prior abiraterone or enzalutamide within 12 months and results may therefore not be generalizable to patients progressing with androgensignaling-targeted inhibitors beyond 12 months, although median time to progression with abiraterone or enzalutamide did not exceed 1 year in randomized studies conducted in chemotherapy-naive mCRPC. 4,5 The results reported here also may not be generalizable to patients receiving an androgen-signaling-targeted inhibitor in non-mCRPC. Therefore, although the results suggest that patients with high NLR may have a greater benefit with cabazitaxel over a second androgen-signaling-targeted inhibitor, further validation of this is required. Finally, the median NLR value used in the present study may not represent the optimal cut-off value for defining those patients with poor outcomes. In this subanalysis, the median NLR (3.38) was used as the cut-off, as ROC did not identify an alternative optimal cut-off. As discussed, there is no clear consensus regarding what the optimal cut-off is, with other studies using cut-offs ranging from 2.5 to 5. 17,19,20,22,23 Rationales for these cut-offs vary per study, although the reason is frequently based on the median value of the population or the median of an earlier study. The variability in cut-offs highlights a need for validation and identification of an optimal cut-off that can be used consistently in future studies. More research into risk stratification using this continuous variable as a biomarker is needed.

Conclusions
In conclusion, the findings of this analysis show that high NLR at baseline associates with poor outcomes with abiraterone or enzalutamide, but does not appear to impact the antitumor activity of cabazitaxel. We hypothesize that high NLR reflects an immunosuppressive environment in which tumor-associated neutrophils and MDSCs promote tumor growth. Such an environment may induce resistance to abiraterone or enzalutamide, while cabazitaxel, by targeting the microtubule network of these neutrophils and myeloid cells, contributes to improving OS. Combined with the previously published survival and health-related quality of life benefits, 6,39 the finding that cabazitaxel is superior to a second androgen-signaling-targeted inhibitor regardless of baseline NLR strengthens the support for the use of cabazitaxel over abiraterone or enzalutamide in this setting. Further studies are necessary to fully elucidate the underlying mechanisms, which should contribute to a better understanding of the role of the tumor microenvironment in prostate cancer.

ROLE OF THE FUNDING SOURCE
This work was supported by Sanofi Genzyme, which funded the CARD study, provided the study treatments, and collaborated with the co-authors on study design, data analysis, data interpretation, and writing of the report. The manuscript was written with editorial support from medical writers, funded by the sponsor. The corresponding author had full access to the data and was responsible for the decision to submit for publication.