Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins, Johns Hopkins University, Baltimore, MD, USA
Correspondence: Emmanuel S. Antonarakis, Prostate Cancer Research Program, Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University, 1650 Orleans Street, CRB1-1M45, Baltimore, MD 21231, USA.
To determine whether the pretreatment neutrophil-to-lymphocyte ratio (NLR), a measure of systemic inflammatory response, is associated with overall survival (OS) in men receiving chemotherapy with docetaxel for metastatic castration-resistant prostate cancer (mCRPC).
Patients and Methods
Records from 238 consecutive patients who were treated with first-line docetaxel-containing chemotherapy for mCRPC at a single high-volume centre from 1998 to 2010 (and who had adequate information to enable calculation of NLR) were reviewed.
Univariable and multivariable Cox regression models were used to predict OS after chemotherapy initiation.
In univariable analyses, the NLR as a discrete variable (optimal threshold 3.0) was significantly associated with OS (P = 0.001).
In multivariable analyses, a lower NLR (≤3.0) was associated with lower risk of all-cause mortality (P = 0.002).
In Kaplan–Meier analysis, the median OS was higher (18.3 vs 14.4 months) in patients that did not have an elevated NLR than in those with an elevated NLR (log-rank; P < 0.001).
Men who were treated with first-line docetaxel for mCRPC who had a low pretreatment NLR (≤3.0) had significantly longer OS.
NLR may be a potentially useful clinical marker of systemic inflammatory response in predicting OS in men with mCRPC who receive docetaxel and may be helpful to stratify patients for clinical trials.
These findings derived from a retrospective analysis need to be validated in larger populations in prospective studies, and in the context of different therapies.
Chemotherapy with docetaxel remains the standard initial systemic therapy for men with metastatic castration-resistant prostate cancer (mCRPC) [1, 2]. As docetaxel is a cytotoxic agent with potential adverse effects and its efficacy varies between patients, several pretreatment factors have been evaluated for their ability to predict outcome after therapy . This will become even more important in the coming years with an increasing number of possible treatment strategies, as novel therapies including cabazitaxel, abiraterone, enzalutamide and sipuleucel-T have been approved for therapy of these men and several other drugs including novel androgen-modulating approaches, therapeutic cancer vaccines, angiogenesis inhibitors or epigenetic therapies are being developed [4-13]. Several variables have been reported to be prognostic factors in men with mCRPC who will receive chemotherapy with docetaxel [14-16]. Gleason score, Eastern Cooperative Oncology Group (ECOG) performance status, presence of visceral metastases, presence of pain, PSA levels, haemoglobin, albumin, lactate dehydrogenase (LDH) and alkaline phosphatase levels have been reported to be associated with overall survival (OS). Furthermore, models have been developed that integrate these factors to predict individual survival probabilities and may be helpful to stratify patients with mCRPC in randomised phase III trials .
The neutrophil-to-lymphocyte ratio (NLR) combines circulating neutrophil and lymphocyte counts and is one of the most common indicators of inflammatory status in patients with cancer. The NLR has been reported to be significantly associated with clinical outcomes in several types of cancer [17-21]. In the last few years, there has been increasing evidence that disease progression in cancer is dependent on a complex interaction of the characteristics of both tumour and host factors including weight loss, performance status and a systemic inflammatory response [22, 23]. These host factors seem to be linked to each other, e.g. patients with elevated C-reactive protein levels were found to show increased weight loss and poorer performance status suggesting systemic inflammation being an important aetiological factor in the nutritional and functional decline of patients with advanced cancer [22, 23].
The purpose of the present study was to examine the association between NLR and OS in men with mCRPC receiving first-line chemotherapy with docetaxel. To do this, we examined the chemotherapy database at a single high-volume academic cancer centre.
Patients and Methods
For the present Institutional Review Board-approved study, data from 238 consecutive men who were treated for mCRPC at the Johns Hopkins Sidney Kimmel Comprehensive Cancer Center between 1 January 1998 and 31 December 2010 were reviewed. All patients received first-line docetaxel-containing chemotherapy, either according to standard-of-care at the USA Food and Drug Administration (FDA)-approved dose and schedule, or as participants in clinical study protocols incorporating first-line docetaxel (at the same dose and schedule) given with or without additional experimental drugs. All patients had pathologically confirmed adenocarcinoma of the prostate, all showed evidence of progressive mCRPC (by PSA or clinical/radiographic criteria) and all had not received prior chemotherapy for mCRPC.
The primary objective of the present study was to examine the association between the pretreatment NLR (before initiation of docetaxel-containing chemotherapy) and OS of men with mCRPC. Data were retrospectively collected including demographic and clinicopathological information, pertinent laboratory values, prior treatment details, and follow-up data specifically pertaining to survival status. Docetaxel chemotherapy cycles were defined as 21-day treatment periods with docetaxel (75 mg/m2 i.v.) administered on day 1 of each treatment cycle, together with prednisone 5 mg twice daily. Disease burden and response to chemotherapy was evaluated following the criteria of the Prostate Cancer Clinical Trial Working Group 2 (PCWG2) . OS was defined as the time from the first chemotherapy treatment with docetaxel to the date of death from any cause (or patients were censored at the date of last follow-up).
Univariable and multivariable Cox regression models were used to adjust for potential confounders in predicting OS. Covariates with significant P values (<0.05) in univariate analysis were included in the multivariable analysis. A stepwise forward-selection method was used to construct multivariable regression model by setting P value for entry at 0.15. Only those covariates that had P values <0.20 in multivariable analysis are reported in the final model. A regression tree approach was used to find the best threshold values of confounding risk factors as previously described . The NLR was calculated by dividing the peripheral-blood absolute neutrophil count by the absolute lymphocyte count, and was treated as a continuous variable. The optimal threshold to dichotomize NLR was determined to be 3.0 (P = 0.003) using the maximally selected log-rank test. Multivariable analyses were performed for NLR as a continuous and dichotomous variable. The Kaplan–Meier product-limit method was used to estimate event-time distributions for OS, and P values were computed using the log-rank test .
For all statistical analyses, tests were two-sided and a P < 0.05 was considered to indicate statistical significance. All statistical analyses were performed using the R software package.
In all, 247 consecutive chemotherapy-naïve men with mCRPC were treated with docetaxel-containing chemotherapy from 1998 to 2010, of which nine men were excluded without available data on pretreatment NLR. The NLR was treated as a discrete variable using the optimal threshold, which was determined to be 3.0 using the maximally selected log-rank test. In all, 168 of the 238 patients (70%) had a pretreatment NLR of >3. The baseline clinicopathological characteristics of the entire cohort of 238 individuals are shown in Table 1. Most men were Caucasian (75.8%). The median (range) age was 68.3 (44.6–84.5) years. Almost 80% of all patients received initial local therapy including surgery (20.9%), external beam radiotherapy (37.3%) or both (20.4%). The median (range) Gleason sum was 8 (5–10). Docetaxel chemotherapy was administered for a median (range) of 6 (1–19) cycles. Bisphosphonates were applied in 32.6% of all cases. Of all patients, 90.3% had bone metastases, of which most patients (85%) had >10 metastatic lesions. Lymph node metastases were found in 51.3% of patients, 77.7% of these with more than five metastatic lymph nodes. Metastases to the liver and lungs were found in 9.3% and 7.2% of all patients, respectively; in many cases those patients had more than five metastases in the liver (63.6%) or the lung (76.5%). At the time of docetaxel initiation, 40.3% of all men (96/238) had measurable soft-tissue disease according to the PCWG2 criteria . Among patients with measurable disease (i.e. soft tissue disease), tumour response rates (defined by Response Evaluation Criteria In Solid Tumors [RECIST] 1.1 criteria) to docetaxel chemotherapy were 42.9%.
Table 1. Baseline characteristics
WBC, white blood cell; ALT, alanine aminotransferase; AST, aspartate aminotransferase.
Number of patients
Race (white), n (%)
Median (range) age, years
Primary treatment, n (%)
Median (range) number of chemotherapy cycles
Gleason score, n (%)
Median (range) number of previous hormonal therapies
Bisphosphonate use, n (%)
ECOG Performance Score, n (%)
Median (range) pain score (0–10)
Alkaline phosphatase, U/L
Platelet count x 109/L
WBC count x 109/L
5410 (1700–13 180)
7150 (2360–20 870)
Serum creatinine, mg/dL
Presence of bone metastases, n (%)
No. of metastases (n), n (%)
Presence of lymph node metastases, n (%)
Presence of lung metastases, n (%)
Presence of liver metastases, n (%)
Presence of measurable disease, n (%)
Presence of objective tumour response, n (%)
baseline PSA level at trial entry, ng/mL
best PSA response, % reduction
Association of NLR to OS
The median (range) follow-up for the present cohort was 15.0 (1.5–90.2) months. Patient age, race, type of primary treatment, Gleason score, or the presence of metastases to bone or lymph node were not associated with OS in univariable analyses (Table 2). In contrast, OS was associated with ECOG performance score (P = 0.001), number of chemotherapy cycles administered (P = 0.002), as well as certain baseline laboratory parameters including haemoglobin (P < 0.001), albumin (P = 0.015), alkaline phosphatase (P < 0.001), aspartate aminotransferase (AST, P < 0.001), LDH (P < 0.001), and baseline serum PSA level (P = 0.026). The pretreatment NLR was significantly associated with OS (P = 0.002). In Kaplan–Meier analysis (Fig. 1), patients with an elevated NLR (>3.0) had a shorter OS (14.4 vs 18.3 months, P < 0.001) than those without an elevated NLR (≤3.0).
Table 2. Univariable analysis of OS
HR for death (95% CI of HR)
Please refer to Table 1 for percentages of patients in each subgroup. WBC, white blood cell.
1.012 (0.996, 1.028)
None vs surgery
1.536 (0.420, 1.011)
None vs radiation
None vs both
Number of chemotherapy cycles
>5 vs ≤5
0.613 (0.452, 0.833)
7 vs ≤6
1.370 (0.715, 2.620)
8–10 vs ≤6
1.600 (0.855, 3.010)
1.082 (0.947, 1.236)
Yes vs no
0.834 (0.610, 1.140)
ECOG Performance Score
≥1 vs 0
1.610 (1.100, 2.354)
Pain score (0–10)
1.030 (0.957, 1.108)
<12 vs ≥12.1
2.090 (1.570, 2.770)
≤3.7 vs >3.7
0.394 (0.284, 0.548)
Alkaline Phosphatase (U/L)
<98 vs ≥99
1.840 (1.370, 2.480)
1.001 (1.000, 1.001)
1.002 (1.001, 1.003)
WBC count (x109/L)
1.000 (1.998, 1.003)
1.000 (0.999, 1.000)
1.000 (1.000, 1.000)
1.022 (1.008, 1.036)
<3 vs ≥3
1.675 (1.219, 2.302)
1.021 (1.012, 1.030)
Presence of bone metastases
Yes vs no
1.125 (0.674, 1.878)
Number of bone metastases
>10 vs <10
1.362 (1.066, 1.738)
Presence of lymph node metastases
Yes vs no
1.030 (0.777, 1.365)
Number of lymph node metastases
>5 vs ≤5
1.354 (0.830, 2.209)
Presence of lung metastases
Yes vs no
2.925 (1.732, 4.939)
Number of lung metastases
>5 vs ≤5
3.199 (0.712, 14.370)
Presence of liver metastases
Yes vs no
1.593 (1.001, 2.535)
Number of liver metastases
>5 vs ≤5
1.390 (0.525, 3.683)
Presence of measurable disease
Yes vs no
0.921 (0.693, 1.225)
Baseline (log2) PSA level
1.249 (1.028, 1.518)
Multivariable Cox proportional hazard regression analysis was performed after treating NLR as a dichotomous and continuous variable (Table 3). A forward-stepwise selection method was used to construct multivariable Cox proportional hazard models using significant covariates in univariable analysis by setting the threshold P value at 0.15. The presence of liver and lung metastasis was significant in univariable analysis, but these parameters were highly associated with elevated NLR as shown in Table 1, and thus they were excluded from multivariable analysis. In multivariable analysis, certain parameters were dichotomized into low and high levels, based on optimal thresholds determined using the maximally selected log-rank test. LDH was also significant in univariable analysis, but was excluded from multivariable analysis due to missing data on 137 patients. In this multivariable model (Table 3), the number of chemotherapy cycles, haemoglobin, albumin and AST levels, and the baseline serum PSA level were all significantly associated with OS. Patients with a NLR of >3.0 were at significantly higher risk of death than patients with a NLR of ≤3.0 (hazard ratio [HR] 1.883; 95% CI 1.248–2.842; P = 0.002). NLR was an independent predictor of survival in the multivariable model as both a continuous (P = 0.045) and a dichotomized variable (P = 0.002). The multivariable model with and without NLR was constructed to assess improvement of model discrimination upon addition of NLR. The concordance index (c-index) with and without NLR was 0.746 and 0.727, respectively.
Table 3. Multivariable model predicting OS
NLR as continuous variable
Dichotomized NLR (>3 vs ≤3)
95% CI of HR
95% CI of HR
Number of chemotherapy cycles: >5 vs ≤5
Haemoglobin: >12 vs ≤12 g/dL
Albumin: >3.7 vs ≤3.7 U/L
AST: >30 vs ≤30 U/L
NLR: continuous and dichotomized
PSA (log2) at baseline
Alkaline phosphatase: <98 vs ≥99 U/L
Several measures of systemic inflammatory response have been described in patients with cancer. In addition to NLR, these include elevated C-reactive protein concentrations, and increased white cell, neutrophil or platelet counts . Moreover, decreased concentrations of albumin are also recognised as a marker of inflammation . Several reports have previously shown an association between the inflammatory status and clinical outcome of patients with cancer [18-20]. This has led to the development of simple prognostic scores combining two or more markers of the inflammatory response to predict outcome. The NLR and the platelet-to-lymphocyte ratio, as well as the Glasgow Prognostic Score (GPS, that combines C-reactive protein and albumin concentrations) have been shown to be of clinical use for risk stratification before surgery or chemotherapy treatment [22, 27]. In addition to its role as an inflammation marker, NLR is also associated with immune function. It has been described that low absolute lymphocyte counts are associated with a generalised state of immunosuppression in several types of cancer, which seems to be associated with impaired survival in these patients .
In the present study, we sought to analyse the association between the pretreatment NLR and OS in men who were treated with first-line docetaxel-containing chemotherapy for mCRPC. We have shown that the NLR is significantly associated with OS in these men. Using an optimal threshold of 3.0, patients with elevated NLR above this level were at significantly higher risk (HR 1.833, P = 0.002) of death than patients with NLRs below this level. Furthermore, pretreatment NLR was associated with best PSA response.
The prognostic value of pretreatment NLR and other measures of systemic inflammatory response in patients with cancer receiving chemotherapy has been reported by other groups in different disease contexts. Keizman et al.  recently showed in 109 patients with metastatic RCC who were treated with sunitinib that a NLR of ≤3 was associated with superior progression-free survival (HR 0.285; P < 0.001) and OS (HR 0.3; P = 0.043). Similarly, in a study of 349 patients with unresectable metastatic colorectal cancer from two cohorts who received first-line palliative chemotherapy, a NLR of >5 was confirmed to independently predict OS in the training cohort (P = 0.002), as well as in the validation cohort (P < 0.001) . A more recent study in patients with breast cancer showed that an NLR of >3.3 before initiation of chemotherapy was an independent predictor of higher mortality (HR 4.09, P = 0.002) . In patients who were treated with everolimus for metastatic RCC, the pre-treatment NLR was an independent prognostic factor for progression-free survival and OS .
The NLR has also been shown to be useful in patients with cancer undergoing surgery for the primary tumour or metastases. Sarraf et al.  reported on 177 patients that underwent complete resection for primary non-small cell lung cancer. The authors found that increasing preoperative NLRs were associated with higher tumour stage (P = 0.019) and remained an independent predictor of survival. Halazun et al.  reported an increased risk of both recurrence (HR 4.521, P < 0.001) and death (HR 2.261, P < 0.001) among 400 patients who underwent hepatic resection for colorectal liver metastases and who had an NLR of >5. In a cohort of 281 patients with localised non-clear cell RCC that underwent surgery with curative intent, the NLR was an independent prognostic factor for disease-free survival . In another study, NLR was associated with OS and cancer-specific survival in patients undergoing radical cystectomy for muscle-invasive urothelial cancer of the bladder .
Similarly, in 230 patients that were diagnosed with colorectal cancer, NLRs of >5 at time of diagnosis were correlated with inferior OS and cancer-specific survival in multivariable analyses . Finally, in the Glasgow Inflammation Outcome Study, 897 men who were diagnosed with prostate cancer were followed over time . Individuals that showed an elevated NLR of ≥5 at ≤2 years after initial diagnosis had a higher risk of all-cause mortality (HR 2.38, P < 0.001) .
Patients with advanced cancer often show neutrophilia and lymphopenia simultaneously. Although absolute neutrophil and lymphocyte counts have been significantly associated with OS in some cancers , the NLR seems to be more stable and less variable, as absolute counts can alter from individual to individual by various physiological, pathological, and physical factors .
The prognostic impact of inflammatory markers on OS has also been observed for C-reactive protein in patients treated with docetaxel for mCRPC. To this end, in a group of 80 consecutive patients, Ito et al.  found C-reactive protein as a continuous variable as well as a discrete variable (<5 vs ≥5) to be an independent prognostic factor for OS (HR 1.95, P < 0.001).
Inflammation-based markers, such as the NLR, the platelet-to-lymphocyte ratio or the GPS (C-reactive protein, albumin) are widely available and inexpensive measurements that are easy to integrate into pretreatment evaluation. Used in a routine manner, these markers might help to better predict patient outcome and help patient stratification into different treatment groups within the context of clinical trial design. Proctor et al.  showed recently that a derived NLR (dNLR) based on white cell and neutrophil counts (dNLR = neutrophil count to [white cell count – neutrophil count] ratio) predicts survival in patients with cancer. This dNLR might be even more widely available than NLR.
Similar to findings from previous studies, we have confirmed several prognostic factors that independently influence OS in men who receive docetaxel chemotherapy for mCRPC [15, 36, 37]. These factors included haemoglobin (>12 vs ≤12 g/dL; HR 0.588; P = 0.009), albumin (>3.7 vs ≤3.7 U/L; HR 0.454; P < 0.001) and baseline serum PSA (HR 1.079; P = 0.027) levels measured at baseline.
The present study has several limitations. Data on all patients were reviewed retrospectively from a single institution and the number of patients is relatively small. Patients in the present study were treated upon their clinician's practices or according to a variety of study protocols. In addition, there was no control over additional or further therapies that men included in the present study could receive once chemotherapy with docetaxel was completed. Also, neutrophil and lymphocyte counts might have been influenced by concurrent infections, comorbidities or additional drugs. Our database did not capture information about thyroid disease, haematological disorders, acute and chronic inflammatory diseases or corticosteroid use. Also, our database did not include information on pretreatment C-reactive protein levels. In the present study, we focused on the prognostic role of pretreatment NLR in men with mCRPC receiving docetaxel, but we did not assess NLR changes during treatment. Pond et al.  reported a significant association between neutropenia on day 8 and improved OS in men who received docetaxel for mCRPC. For this reason, NLR during or after docetaxel treatment may also function as a prognostic marker in these patients.
In conclusion, we have investigated the prognostic impact of the NLR on OS in men receiving first-line chemotherapy with docetaxel for mCRPC. We show that the NLR is significantly associated with clinical outcome in these patients. Men with an elevated pretreatment NLR of >3.0 were found to be at higher risk of death after adjusting for other prognostic variables. The pretreatment NLR may be used to stratify patients during enrollment of clinical studies. These findings require validation in further prospective studies, and in the context of additional therapies for advanced prostate cancer.
Nuhn received a fellowship of the German Research Foundation (DFG; Nu 277/1-1).
Antonarakis is partially funded by grant P30 CA006973.
Conflict of Interest
The authors declare that there are no conflicts of interest.