THE ADDITION OF CHEMOTHERAPY IN THE DEFINITIVE MANAGEMENT OF HIGH RISK PROSTATE CANCER

Abstract

In attempt to improve long-term disease control outcomes for high-risk prostate cancer, numerous clinical trials have tested the addition of chemotherapy (CTX)—either adjuvant or neoadjuvant—to definitive local therapy, either radical prostatectomy (RP) or radiation therapy (RT).

Neoadjuvant trials generally confirm safety, feasibility, and pre-RP PSA reduction, but rates of pathologic complete response are rare, and no indications for neoadjuvant CTX have been firmly established. Adjuvant regimens have included CTX alone or in combination with androgen deprivation therapy (ADT).

Here we provide a review of the relevant literature, and also quantify utilization of CTX in the definitive management of localized high-risk prostate cancer by querying the National Cancer Data Base (NCDB). Between 2004 and 2013, 177 patients (of 29,659 total) treated with definitive RT, and 995 (of 367,570 total) treated with RP had CTX incorporated into their treatment regimens. Low numbers of RT + CTX patients precluded further analysis of this population, but we investigated the impact of CTX on overall survival (OS) for patients treated with RP +/− CTX. Disease-free survival or biochemical-recurrence-free survival are not available through the NCDB. Propensity-score matching (PSM) was conducted as patients treated with CTX were a higher-risk group. For non-matched groups, OS at 5-years was 89.6% for the CTX group versus 95.6%, for the no-CTX group (p < 0.01). The difference in OS between CTX and no-CTX groups did not persist after PSM, with 5-year OS 89.6% versus 90.9%, respectively (HR 0.99; P = 0.88).

In summary, CTX was not shown to improve OS in this retrospective study. Multimodal regimens—such as RP followed by ADT, RT, and CTX; or RT in conjunction with ADT followed by CTX—have shown promise, but long-term follow-up of randomized data is required.

INTRODUCTION

In attempt to improve disease control outcomes for high-risk prostate cancer, numerous clinical trials have tested the addition of chemotherapy (CTX)—either adjuvant or neoadjuvant—to definitive local therapy, either radical prostatectomy (RP) or radiation therapy (RT).

Neoadjuvant regimens supplemented to local therapy have included estramustine and etoposide,¹ docetaxel alone,^2–6 or docetaxel in combination with mitoxantrone,^7–10 estramustine,^11,12 capecitabine,¹³ nab-paclitaxel,¹⁴ gefitinib,¹⁵ bevacizumab,¹⁶ and/or androgen deprivation therapy (ADT).^17–23 These neoadjuvant trials generally confirm feasibility, safety, and PSA reduction prior to RP, but pathologic complete response is rare and no indications for neoadjuvant CTX have been firmly established.^7,17

Adjuvant regimens following RP have included CTX alone^24–28 or in combination with ADT and RT.^29–32 The recently published results of NRG Oncology/RTOG Study 0621—a phase 2 trial of adjuvant RT, ADT, and docetaxel for high-risk post-RP patients—are encouraging and the 3-year progression-free survival of 73% demonstrates a significant improvement from historical controls, but the authors acknowledge that randomized studies are needed.³¹ Following RT and ADT, the addition of adjuvant CTX also seems promising as 4-year results from the randomized phase 3 trial RTOG 0521 suggest a 10% improvement in disease-free survival, and a 4% improvement to overall survival (OS).³³

Incorporation of CTX into definitive treatment regimens remains outside the routine standard of care, especially since the relative trials are generally small in number of patients included, limited in terms of long-term follow-up, mostly single arm, and heterogeneous in terms of inclusion criteria and treatment paradigms. Due to the rise of active surveillance, prostate cancer incidence seems likely to move away from low-risk patients and towards high-risk patients. Additionally, with increasing public awareness of active surveillance, some patients may be reluctant for treatment intensification, even in warranted scenarios. RT-based regimens now incorporate use of more effective dose-escalated modalities in addition to long-term ADT. With these changes to management of high-risk prostate cancer, the question of the exact scenarios for which CTX has been tested and may provide a supplemental benefit gains importance.

NEOADJUVANT CHEMOTHERAPY

In the early 1990s several reports emerged describing the effectiveness of CTX for metastatic castration-resistant prostate cancer, which was at that time a relatively novel concept.^34–36 Subsequently, interest arose in moving CTX to the definitive setting. Table 1 demonstrates a comparison of the numerous trials that will be discussed below in terms of regimens, eligibility criteria, and outcomes.

Table 1.

Trial comparisons.

Study	Regimen	Inclusion criteria; Any of the below	Local therapy	Number of patients who completed CTX and underwent local therapy	Median follow- up time (months)	Percent with recurrence*
Neoadjuvant CTX
Clark et al, 2001; Cleveland Clinic1	Estramustine + etoposide × 3 cycles (28 day cycles)	T2b-T3; GS ≥8; PSA ≥15	RP	18	14	12
Hussain et al, 2003; Karmanos and Michigan12	Docetaxel + estramustine q21days for 3–6 cycles	T2b-T3; GS ≥8; PSA ≥15	RP or RT	28	130	64
Ryan et al, 2004; Memorial Sloan Kettering39	Vinblastine (6 weeks on, 2 weeks off) + estramustine × 2 neoadjuvant cycles, followed by concurrent vinblastine and estramustine with RT	GS ≥8 and PSA ≥10; GS ≥7 and PSA ≥20; T3 and PSA ≥20; T4; N1	RT	23	60	65
Febbo et al, 2005; Dana Farber3	Weekly docetaxel × 6 months	T3; GS ≥8; PSA ≥20	RP	19	26.5	63.2
Vuky et al, 2009; Virginia Mason Medical Center15	Docetaxel (3 weeks on 1 week off) and daily gefitinib × two months	T2b-T3; GS ≥8; PSA ≥20	RP	22	28	34
Ross et al, 2012; Prostate Cancer Clinical Trials Consortium16	Docetaxel × 6 (q21 days) with bevacizumab (q21 days) given with the first 5 cycles	T3; GS ≥8; PSA ≥20; PSA velocity >2 ng/mL/y	RP	37	N/A	49
Zhao et al, 2015; Cleveland Clinic6	Weekly docetaxel × 6 weeks	T2b-T3; GS≥8; PSA ≥15	RP	28	49.5	57
Bergstrom et al, 2017; Oregon/VA Portland/Washington8	Docetaxel (weekly) + mitoxantrone (3 out of 4 weeks) × 4 months	T2c-T3a; GS ≥4+3; PSA ≥15	RP	54	120	63
Neoadjuvant CTX + ADT
Pettaway et al, 2000; MD Anderson20	LHRH agonist and antiandrogen + alternating cycles (× 12 weeks) of ketoconazole + doxorubicin or vinblastine + estramustine	T3; GS 7 with PSA ≥10; T1-2 with GS ≥8	RP	33	13	31
Konety et al, 2004; Memorial Sloan Kettering18	LHRH agonist + 4–6 cycles of carboplatin, paclitaxel, and estramustine	≥T3; GS ≥8; PSA ≥20	RP	35	29	55
Prayer-Galetti et al, 2007; Italy21	LHRH agonist + docetaxel (q21 days) and estramustine × 4 cycles	≥T3; GS ≥8; PSA ≥15	RP	18	53	58
Kelly et al, 2008; CALGB 9981146	LHRH agonist + paclitaxel weekly, carboplatin monthly, and estramustine × 4 cycles	≥T3b; GS ≥7 and PSA >20	RT	27	38	70
Chi et al, 2008; Canadian multicenter17	LHRH agonist and antiandrogen + docetaxel (6 weeks with 3 or on 2 weeks off) for 3 cycles	≥T3; GS ≥8; PSA ≥20; GS 7 with 3 or more positive cores; PSA ≥10 with 3 or more positive cores	RP	64	42.7	30
Sella et al, 2008; Israel22	LHRH agonist and antiandrogen + docetaxel (q21days) and estramustine × 4 cycles	≥T2c; GS ≥8; PSA ≥20	RP	22	23.6	45.4
Mellado et al, 2009; Spain45	LHRH agonist and antiandrogen + docetaxel (3 weeks on, 1 week off) × 3 cycles	T3; T1c-T2 with GS ≥4+3 or PSA >20	RP	51	35	41.2
Narita et al, 2012; Akita University, Japan19	LHRH agonist and antiandrogen + docetaxel (weekly) and estramustine × 6 weeks	≥T3; GS ≥9; PSA ≥15	RP	18	18	16.7
Thalgott et al, 2014; Germany23	LHRH agonist and antiandrogen	>40% 5-yr biochemical recurrence risk62	RP	29	48.6	55.2
Fizazi, et al. 2015; GETUG 1247	LHRH agonist alone LHRH agonist + docetaxel (q3 weeks) and estramustine × 4 cycles	≥T3; GS ≥8; vs. PSA >20; N1	RP or RT	206 vs. 207	105.6	54 vs. 43
Adjuvant CTX
Schmidt et al, 2006; National Prostatic Cancer Project—RP Protocol49	Cyclophosphamide q3weeks × 2 years vs. estramustine × 2 years vs. observation	T2c – T3b; N1	RP	184	120	56 vs. 46 vs. 46
Schmidt et al, 2006; National Prostatic Cancer Project—RT Protocol49	Cyclophosphamide q3weeks × 2 years vs. estramustine × 2 years vs. observation	T2c – T3b; N1	RT	253	120	77 vs. 49 vs. 63
Kibel et al, 2007; Multicenter25	Docetaxel (3 weeks on, 1 week off) × 6 cycles	>50% 3-yr biochemical recurrence	RP	76	29.2	60.5
Cetnar et al, 2008; University of Pennsylvania24	Paclitaxel weekly (3 weeks on, 1 week off) and estramustine × 4 cycles	≥50% 2-year PSA failure63	RP	17	24	30
Ahlgren et al, 2016 SPCG 1253	Docetaxel q3weeks × 6 cycles vs. survellance	pT2 with positive margin and GS ≥4+3; pT3b and GS >3+4; N1 and GS >3+4	RP	459	56.8	47.9 vs. 38.9
Adjuvant CTX + RT + ADT
Hussain et al, 2012; University of Maryland55	LHRH agonist + paclitaxel weekly concurrent with adjuvant RT	pT3N0N+ disease or rising PSA ≥ 0.05	RP	30	74.9	37
Hurwitz et al, 2017; RTOG 062131	LHRH agonist and antiandrogen + docetaxel (q3weeks) × 6 cycles	post-RP PSA nadir > 0.2 ng/mL and GS ≥7; post-RP PSA nadir of <0.2 but ≥pT3 and GS ≥8	RT	74	52.8	35.1

^*Different regimen types and lengths allow only rough comparisons due to the possibility immortal time bias. Biochemical or clinical recurrence.

Neoadjuvant chemotherapy without androgen deprivation therapy

In 1998 at Cleveland Clinic, a trial was initiated that treated high-risk patients with three cycles of estramustine and etoposide prior to RP and bilateral pelvic lymphadenectomy.¹ The regimen was delivered to 18 patients, and while there was a higher than expected rate of organ-confined disease on surgical pathology, histologically, there was no evidence for antitumor effect beyond what would have been expected with ADT alone.¹ This apparent lack of a markedly improved antitumor effect compared to ADT, in conjunction with a relatively high rate (17%) of thromboembolic adverse events (a known association with the synthetic steroidal estrogen, estramustine), led the authors to determine that while the regimen was feasible, other potentially more efficacious regimens should be considered.¹

In 1999 at the Karmanos Cancer Institute and University of Michigan, 21 men with were treated with estramustine and docetaxel for a maximum of six cycles, this time prior to either RP or RT.¹² Again, histological activity was demonstrated, but relative activity of CTX compared to ADT remained in question.¹² None of the 10 patients who underwent RP had a pathologic complete response.¹² To offset the thrombotic effects of estramustine (and after 3 patients developed deep venous thrombosis), low-dose warfarin was eventually instituted, which effectively stopped thromboembolism in the remaining patients.¹² PSA was lowered to a degree consistent with the above Cleveland Clinic trial.¹ Namely, median PSA nadir was 0.5 ng/mL, with 76% of patients achieving a 90% decline and the remainder achieving a decline between 50 – 90%.¹² Comparatively, in the Cleveland clinic trial, half of the patients achieved an undetectable PSA prior to RP (less than 0.2 ng/mL), and the remaining patients had PSAs between 0.2 and 0.7 ng/mL (median 98% reduction).¹

Subsequently in 2001, the Cleveland Clinic group began a new Phase II trial of weekly docetaxel alone for six weeks prior to RP.^2,6 Again, reductions in serum PSA following the CTX regimen were demonstrated in the majority of the 29 patients (79%); though, only 24.1% of patients had ≥ 50% reduction in PSA,^2,6 which was somewhat less than the rates demonstrated in the Karmanos/Michigan study of estramustine and docetaxel.¹² Again, the neoadjuvant CTX regimen was feasible and reasonably well tolerated, with no significant increase to RP morbidities.⁶ None of these patients achieved a pathologic complete response.⁶ The most recent report of these patients describes 36% alive and recurrence-free.⁶ Notably, PSA response to neoadjuvant therapy—an endpoint in several of these neoadjuvant chemotherapy trials—was not found to be a predictor of long-term outcomes.⁶ In a later attempt by this group to see if another agent might more effectively elicit tumor response, 18 patients were treated with nab-paclitaxel prior to RP.¹⁴ Again, no pathologic complete responses were generated and only 16% had PSA reductions > 50%.¹⁴ It should be noted that there is evidence from the metastatic setting that suggests PSA reductions with CTX may be enhanced by the addition of prednisone, though it is unclear if this benefit might be translated to the neoadjuvant setting.³⁷

In 2001 a study was open through Dana Farber that again tested neoadjuvant docetaxel, but this time docetaxel was given for 6 months instead of 6 weeks.³ Endorectal MRI was also incorporated to monitor tumor response to CTX.³ In the 19 patients, median maximum tumor volume decreased by 1.5 cm³ (48.3%) and median prostate size decreased by 3.6 cm³ (28.9%) after the six months of docetaxel.³ PSA decrease of ≥50% was noted in the majority of patients (58%) and mean PSA reduction was 64%.³ No pathologic complete responses were demonstrated.³

Also in 2001, a multicenter Phase 1/2 study opened in the Northwest; neoadjuvant treatment was with docetaxel and mitoxantrone for 16 weeks prior to RP.^7–10 The regimen was safely tolerated, and of the 54 patients included in the most recent report, recurrence-free survival was 29%.⁸ Lymph node status, PSA density, and increased prostate VEGF expression were found to be predictive of recurrence.⁸

VEGF expression had been previously hypothesized to be involved in pathogenesis of prostate metastases,³⁸ and in further investigation of VEGF as a driver of prostate cancer progression, the Prostate Cancer Clinical Trials Consortium performed a phase 2 multicenter trial, with patients treated 2006 – 2008, of docetaxel and the VEGF inhibitor bevacizumab for six cycles (21 day cycles).¹⁶ Of the 41 patients included in analysis, 22% achieved >50% decline in PSA with neoadjuvant therapy, and 29% achieved >50% decline in tumor volume on endorectal MRI.¹⁶ Again, no patients experienced a pathologic complete response.¹⁶

Other failures in obtaining pathologic complete response were demonstrated in Friedman et al (3–6 months of docetaxel and capecitabine; 40% achieving ≥50% PSA reduction),¹³ and Vuky et al (2 months of neoadjuvant docetaxel and gefitinib).¹⁵

Collectively, determining the effect of neoadjuvant CTX on biochemical control and pathologic disease characteristics such as surgical margin status was not possible via the above phase II trials. Testosterone monitoring conducted in the above trials generally suggests that CTX on its own does appear to act in a mechanism independent from ADT to lower PSA; the extent to which there may be additive or synergistic effects between CTX and ADT remains unclear.

Finally, with respect to neoadjuvant CTX prior to RT, this has been investigated in single-institutional Phase 2 trial, Ryan et al, which enrolled a very high-risk (relative to many of the above trials) cohort of patients to a protocol of neoadjuvant vinblastine and estramustine followed by the same regimen concurrent with RT (75.6 Gy).^39,40 PSA nadirs of 0 were achieved in 70% of the 23 patients, and this seemed to be a strong predictor of outcomes, as not achieving a PSA nadir of 0 was associated with a fivefold increase in risk of developing metastatic disease.^39,40

Neoadjuvant chemotherapy with androgen deprivation therapy

In 2006 the Cancer and Leukemia Group B (CALGB) 90203 randomized phase 3 trial was initiated testing RP with or without the addition of neoadjuvant ADT and docetaxel.¹¹ Adjuvant RT is allowed at the discretion of the treating physician. Men eligible for this study must have an estimated ≤60% probability of freedom from disease recurrence at 5 years following RP.¹¹ CALGB 90203 has met its accrual goal, though results will not be available for several years.⁴⁴

Several phase II trials investigating a similar paradigm have been conducted, and confirmed the safety and feasibility of their various regimens.^18,20,21 As expected, PSA values decreased dramatically with the addition of ADT, compared to the above trials of neoadjuvant CTX alone. Interestingly, an Italian study of 19 men included one patient with a pathologic complete response and another six patients (31%) with only small foci of tumor remaining comprising <10% of the prostate volume.²¹ Disease-free survival was found to be associated with pathologic response to neoadjuvant therapy.²¹ In a Canadian multicenter study of 64 men, two achieved a pathologic complete response and an additional 25% had ≤5% tumor remaining;¹⁷ in a Spanish study of 51 men, three achieved a pathologic complete response;⁴⁵ and in a Japanese study of 18 men, two achieved a pathologic complete response.¹⁹ Conversely, in an Israeli study of 22 men,²² and in a German study of 30 men,²³ no pathologic complete responses were demonstrated. It does seem that pathologic complete responses, though rare, are possible with the addition of ADT to neoadjuvant CTX.

Prior to RT, neoadjuvant CTX and ADT has been investigated in a multicenter phase 2 study, Cancer and Leukemia Group B (CALGB) 99811.⁴⁶ In addition to ADT, patients received neoadjuvant carboplatin, paclitaxel, and estramustine.⁴⁶ The regimen was found to be tolerable and feasible in the multicenter setting.⁴⁶

Data from the randomized Phase 3 GETUG 12 trial—testing the addition of adjuvant docetaxel and estramustine versus ADT alone prior to definitive local therapy via RP or RT—continues to mature.⁴⁷ After staging lymphadenectomy, local therapy was decided upon at multidisciplinary conference; patients with node-negative disease could undergo RP or RT, while patients with node-positive disease could undergo RT or no local therapy.⁴⁷ Early results show a benefit for CTX in terms of 8-year relapse-free survival (62% vs. 50%) but not yet metastasis-free survival or OS.⁴⁷

ADJUVANT CHEMOTHERAPY

Adjuvant therapy affords patients with more favorable pathologic characteristics to perhaps be spared treatment regimen intensification via RT or CTX. The obvious cost of moving treatment escalation to the adjuvant setting is that there is no chance of tumor response/downstaging—which can influence surgical outcome, prognosis, and need for adjuvant RT.

Adjuvant chemotherapy without androgen deprivation therapy

Perhaps the earliest attempt to evaluate adjuvant CTX was the National Prostate Cancer Project, that enrolled 1978 – 1985, and randomized patients, via two protocols (post-RP and post-RT, staging pelvic lymphadenectomy required for both arms), to either cyclophosphamide, estramustine, or observation.⁴⁹ When interpreting the results (Table 1), it should be noted that lymph node involvement was considerably lower in the RP arm (29%) compared to the RT arm (63%).⁴⁹ The authors ultimately concluded that adjuvant estramustine was of clearest benefit in RT patients with extensive (>20%) pelvic nodal involvement.⁴⁹

At the University of Pennsylvania between 2001 and 2004, 17 patients with at least 50% probability of 2-year PSA failure were treated with adjuvant paclitaxel and estramustine following RP.²⁴ The actual median risk of PSA failure in this group of men was 70%, and, interestingly, only 30% developed PSA failure (P = 0.001).²⁴

Between 2002 – 2004, patients at >50% recurrence risk after RP accrued on a relatively large phase II multicenter trial (Kibel et al) investigating adjuvant docetaxel for six cycles.²⁵ Using a nomogram comprised of historical controls, the median progression-free survival of 15.7 months was deemed to be better than the predicted 10 months, though 30% of patients experienced Grade ≥3 toxicity.²⁵

Initiated in 2006, Veteran’s Affairs Cooperative Studies Program (CSP) 553 was a randomized trial testing RP with or without the addition of adjuvant docetaxel and prednisone.^50,51 VA CSP 553 did not meet its accrual goal of 300 men and was closed early; so far it has been reported in abstract form only, with the underpowered results demonstrating a non-statistically significant trend for benefit in terms of progression free survival for the overall intention to treat population, but a statistically significant benefit for African American patients and ≥T3b tumors, on the pre-specified subgroup analyses.⁵²

In 2016, results of SPCG12 were reported in abstract form; from 2005 – 2010, 459 Scandinavian men were randomized to docetaxel versus surveillance with results showing no benefit for docetaxel in terms of biochemical disease free survival.⁵³ In fact, numerically the docetaxel arm did worse by approximately 10%, though this was not statistically significant, and the authors even suggest that a certain subgroup of patients seems to progress biochemically more rapidly with docetaxel monotherapy.⁵³

Adjuvant chemotherapy with androgen deprivation therapy

Extrapolating from efficacy of ADT in other settings, it seems possible or even likely that the combination of CTX and ADT could be superior to CTX alone. RTOG 99-02 was a phase III randomized trial testing the addition of four cycles of paclitaxel, estramustine, and etoposide to ADT and RT.⁵⁴ Due to excess thromboembolic toxicity, despite the eventual addition of warfarin, the trial was stopped early of its intended sample size of 1,440, after a total of 397 patients were accrued.⁵⁴ Toxicities were considerably more common in the experimental arm—gastrointestinal, renal/genitourinary, and especially, hematologic (40% experienced Grade 3–4)—though late toxicities at 2 or 3 years were not different between arms.⁵⁴

Early results of RTOG 0521—a randomized phase III trial of adjuvant ADT and six cycles of docetaxel following definitive RT in men with any of 1) Gleason ≥9; 2) ≥T2, Gleason 8, PSA <20; or 3) Gleason ≥7, PSA ≥20—have been reported in abstract form and demonstrated a statistically significant benefit for the addition of docetaxel in terms of 4-year disease-free survival (65% vs. 55%) and OS (93% vs. 89%).³³ Node positive men and men with PSA >150 were excluded. Longer follow-up is planned and will reveal if the degree of these survival differences magnify over time.

Trimodality therapy—surgery, adjuvant radiation therapy, and chemotherapy

University of Maryland initiated a protocol in 1999 testing the addition of adjuvant ADT and concurrent chemoradiation (paclitaxel) following RP.⁵⁵ Of the 30 patients enrolled at a median follow-up time of 74.9 months, 37% experienced biochemical progression.⁵⁵

The recently published results of RTOG 0621 demonstrated favorable 3-year freedom from progression of 73%.³¹ This regimen was deemed well-tolerated except for Grade 3 and 4 neutropenia, and this treatment paradigm is the most uniform among the adjuvant CTX trials given that it pre-specifies use of ADT and adjuvant RT. Longer-term follow-up is needed before an analysis of OS will be performed, and a follow-up phase 3 study is being planned—NRG-GU002—which will randomize men treated with RP with high-risk post-operative features to receive ADT and adjuvant RT with or without six cycles of adjuvant docetaxel.⁵⁶

In summary of the above, results from adjuvant CTX trials are encouraging, though will require confirmation in large phase 3 trials.

CHEMOTHERAPY UTILIZATION IN THE UNITED STATES

In attempt to quantify the extent to which CTX has been incorporated into definitive treatment of high-risk prostate cancer thus far in the United States, we sought to utilize the National Cancer Data Base (NCDB) to investigate utilization patterns and compare OS between regimens +/− CTX.

Methods and materials

The NCDB is a joint project of the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society, and is the largest clinical registry in the world, incorporating approximately 70% of new cancer cases from the United States.⁵⁷ Given no patient identifiers are available through the NCDB, no institutional review board approval was required to conduct this investigation.

Patients that were included had histologically-proven invasive prostate adenocarcinoma and met at least one of the following criteria: American Joint Committee on Cancer (AJCC) stage ≥T2c or greater, Gleason score ≥8, or PSA ≥20. Exclusion criteria consisted of: patients with rare histologies including sarcomas, and neuroendocrine/small-cell cancers, metastatic patients, cases with missing outcomes, and cases with unknown CTX status. Node positive patients were not excluded.

Cases that were treated with CTX more than 8 months following local therapy were excluded in order to maintain a population of non-metastatic patients who received either neoadjuvant or adjuvant CTX only. Eight months was selected as a reasonable cutoff beyond which CTX delivery would be more likely to be for metastatic and not adjuvant treatment—while still allowing an adequate interval for the completion of adjuvant RT following RP, if delivered.

Incorporation of CTX into the treatment regimen remains investigational and a non-standard of care approach, and so low numbers of cases were to be expected. Patients treated with either local therapy—RP or RT were both initially included. It became apparent that the number of patients treated with definitive RT and CTX was too small to pursue further statistical analysis of OS. The number of patients treated with RP and CTX was considerably higher, so we proceeded with analysis of OS. Statistical analysis was conducted using SAS Version 9.4 with software macros designed by the Biostatistics and Bioinformatics Shared Resource of the Winship Cancer Institute at Emory University.⁵⁸ The significance level was set at 0.05. Descriptive statistics were generated to summarize patient, disease, and treatment characteristics. Patients receiving neoadjuvant and adjuvant CTX were grouped together to preserve sample size. The univariate associations between covariates and study cohorts (CTX vs. no CTX) were assessed using the Chi-square test for categorical covariates and ANOVA for numerical covariates. Univariate analysis (UVA) between each covariate including study cohorts and study outcome were assessed using Cox proportional hazards models and log-rank tests, and logistic regression for binary cohorts (CTX). Start date for OS was established as the date of onset of definitive treatment, which was taken as RP unless neoadjuvant CTX was utilized, in which case the start date was established as the initiation of CTX, because, as described above, clinical trials testing the addition of neoadjuvant CTX have demonstrated high degree of tumor down-staging, and progression during neoadjuvant CTX is very rare. In multivariable analysis (MVA), the Cox proportional hazard model was applied for OS, and the model was built by a backward variable selection method applying an alpha = .20 removal criteria. Kaplan Meier plots were produced to compare the survival curves by cohorts.

Propensity score matching (PSM) method was also implemented to reduce treatment selection bias. A logistic regression model predicting CTX was carried out to estimate the propensity score by covariates that predict OS in multivariable model and known confounders. Patients from the CTX group were matched to the no CTX group at a ratio of 1:5 based on the propensity score using a greedy algorithm.⁵⁹ After matching, the balance of covariates between the two cohorts was evaluated by the standardized differences and a value of < 0.1 was considered as negligible imbalance.⁶⁰ The effects were estimated in the matched sample by a Cox model with a robust variance estimator for OS.⁶¹

Results

High-risk prostate cancer patients diagnosed between 2004 and 2013 were included in the analysis. During this timeframe, 29,659 patients who met the above criteria were treated with definitive RT, and only 177 of these had CTX incorporated into the treatment regimen; 367,570 patients who met the above criteria underwent RP, and 995 of these patients received CTX. Given the small numbers of patients treated with RT and CTX, only analysis of RP patients treated with CTX was continued. For Consolidated Standards of Reporting Trials (CONSORT) diagram outlining change in RP patient numbers by exclusion criteria, see Figure 1.

Figure 1.

Consolidated Standards of Reporting Trials (CONSORT) diagram.

NCDB = National Cancer Data Base; PSA = Prostate-specific antigen; RP = Radical prostatectomy; CTX = Chemotherapy

For UVA of patient and tumor characteristics by cohort, please see Table 2. All of following were associated (all P < 0.001, unless otherwise specified) with use of CTX: younger age, white race (P = 0.007), academic/research facility type, earlier year of diagnosis, increasing clinical and pathologic T stage, positive clinical and pathologic N stage, positive surgical margins, increasing Gleason score, increasing PSA, positive surgical margins, use of RT, and use of ADT. Distance to facility was not significantly different between CTX and no CTX groups (P = 0.512). The CTX group had a longer median time to RP after diagnosis compared to the no CTX group (111 versus 76.5 days, respectively; P < 0.001). However, when considering median time from diagnosis to start of any definitive treatment regimen (either RP or start date of CTX in the case of neoadjuvant CTX treatment), the CTX group had a shorter median time to onset of therapy compared to the no CTX group (51 versus 65 days, respectively; P < 0.001). The CTX group had a longer median duration to the start of ADT compared to the no CTX group (81 versus 66 days, respectively; P < 0.001).

Table 2.

Patient and Clinical Characteristics and Treatment Details by Cohort

Variable	No CTX (n = 366,575)	CTX (n = 995)	P*
Age at diagnosis, median, y	61	60	< 0.001
Race, No. (%)
White	304,387 (83.0)	862 (86.6)	0.007
Black	45,367 (12.4)	92 (9.3)
Others/Unknown	16,821 (4.6)	41 (4.1)
Charlson-Deyo Score, No. (%)
0	307,154 (83.8)	852 (85.6)	0.113
1	52,957 (14.4)	122 (12.3)
2+	6,464 (1.8)	21 (2.1)
Facility Type, No. (%)
Non-Academic/Research Program	210,830 (57.6)	334 (33.6)	< 0.001
Academic/Research Program	155,393 (42.4)	659 (66.3)
Year of Diagnosis, No. (%)
≥ 2004 – ≤ 2007	129,192 (35.2)	501 (50.4)	< 0.001
> 2007 – ≤ 2009	83,964 (22.9)	220 (22.1)
> 2009 – ≤ 2011	83,900 (22.9)	162 (16.3)
> 2011 – ≤ 2013	69,519 (19.0)	112 (11.3)
AJCC clinical T stage, No. (%)
1	213,891 (58.4)	328 (33.0)	< 0.001
2	89,747 (24.5)	319 (32.1)
3	10,243 (2.8)	164 (16.5)
4	358 (0.1)	15 (1.5)
Unknown	52,336 (14.3)	169 (17.0)
AJCC clinical N stage, No. (%)
0	282,538 (77.1)	663 (66.6)	< 0.001
1	1,067 (0.3)	47 (4.7)
Unknown	82,970 (22.6)	285 (28.6)
AJCC pathologic T stage, No. (%)
1–2	26,1170 (71.2)	242 (24.3)	< 0.001
3	93,604 (25.5)	630 (63.3)
4	2,113 (0.6)	38 (3.8)
Unknown	9,688 (2.6)	85 (8.5)
AJCC pathologic N stage, No. (%)
0	263,117 (71.8)	623 (62.6)	< 0.001
1	9,745 (2.7)	253 (25.4)
Unknown	93,713 (25.6)	119 (12.0)
Gleason Score, No. (%)
2–7	315,957 (86.2)	348 (35.0)	< 0.001
8–10	45,214 (12.3)	585 (58.8)
Unknown	5,404 (1.5)	62 (6.2)
PSA, No. (%)
< 10	34,719 (9.5)	187 (18.8)	< 0.001
≤ 10- < 20	264,260 (72.1)	530 (53.3)
≥ 20	24,624 (6.7)	198 (19.9)
Unknown	42,972 (11.7)	80 (8.0)
Surgical Margin, No. (%)
Negative	275,028 (75.0)	529 (53.2)	< 0.001
Positive	87,590 (23.9)	414 (41.6)
Unknown	3,957 (1.1)	52 (5.2)
RT, No. (%)
No RT	344,656 (94.0)	727 (73.1)	< 0.001
Adjuvant RT	15,781 (4.3)	196 (19.7)
RT–adjuvant criteria not met	6,138 (1.7)	72 (7.2)
ADT, No. (%)
No	338,908 (92.4)	346 (34.8)	< 0.001
Yes	18,193 (5.0)	634 (63.7)
Unknown	9,474 (2.6)	15 (1.5)

Abbreviations: CTX, Chemotherapy; AJCC, American Joint Committee on Cancer; PSA, Prostate-specific antigen; RT, Radiation therapy; ADT, Androgen deprivation therapy

^*Bolded P Values are significant

In the MVA (Table 3), use of CTX was associated with inferior OS with a hazard ratio (HR) of 1.19 (95% confidence interval [CI] 1.01 – 1.39; P = 0.039). As expected, age at diagnosis, race, Charlson-Deyo score, facility type, AJCC clinical T and N stage, AJCC pathologic T and N stage, Gleason score, PSA, and surgical margin status were all significantly predictive for OS.

Table 3.

Multivariable Subgroup Analysis of Overall Survival for the No CTX and CTX Cohorts

Variable	Hazard Ratio (95% CI)	P*
Chemotherapy use
Yes	1.19 (1.01 – 1.39)	0.039
No	―	―
Age at diagnosis, median	1.06 (1.06 – 1.06)	< 0.001
Race, No. (%)
White	―	―
Black	1.41 (1.35 – 1.47)	< 0.001
Others/Unknown	0.78 (0.72 – 0.84)	< 0.001
Charlson-Deyo Score, No. (%)
0	―	―
1	1.59 (1.54 – 1.65)	< 0.001
2+	2.59 (2.41 – 2.77)	< 0.001
Facility Type, No. (%)
Non-Academic/Research Program	―	―
Academic/Research Program	0.87 (0.84 – 0.90)	< 0.001
Year of Diagnosis, No. (%)
≥ 2004 – ≤ 2007	―	―
> 2007 – ≤ 2009	1.05 (1.01 – 1.09)	0.023
> 2009 – ≤ 2011	1.05 (0.99 – 1.10)	0.082
> 2011 – ≤ 2013	1.08 (1.00 – 1.16)	0.050
AJCC clinical T stage, No. (%)
1	―	―
2	1.11 (1.08 – 1.15)	< 0.001
3	1.30 (1.22 – 1.39)	< 0.001
4	1.52 (1.19 – 1.93)	< 0.001
Unknown	1.12 (1.07 – 1.18)	< 0.001
AJCC clinical N stage, No. (%)
0	―	―
1	1.22 (1.03 – 1.42)	0.018
Unknown	0.99 (0.95 – 1.04)	0.739
AJCC pathologic T stage, No. (%)
1–2	―	―
3	1.35 (1.31 – 1.40)	< 0.001
4	2.07 (1.86 – 2.31)	< 0.001
Unknown	1.16 (1.07 – 1.25)	< 0.001
AJCC pathologic N stage, No. (%)
0	―	―
1	1.55 (1.45 – 1.66)	< 0.001
Unknown	0.97 (0.93 – 1.01)	0.101
Gleason Score, No. (%)
2–7	―	―
8–10	1.65 (1.60 – 1.71)	< 0.001
Unknown	1.23 (1.12 – 1.35)	< 0.001
PSA, No. (%)
< 10	0.82 (0.78 – 0.85)	< 0.001
≤ 10– < 20	―	―
≥ 20	0.97 (0.92 – 1.03)	0.381
Unknown	0.94 (0.89 – 0.99)	0.028
Surgical Margin, No. (%)
Negative	―	―
Positive	1.15 (1.12 – 1.19)	< 0.001
Unknown	0.93 (0.82 – 1.06)	0.277

Abbreviations: CTX, Chemotherapy; CI, Confidence interval; AJCC, American Joint Committee on Cancer; PSA, Prostate-specific antigen; ADT, Androgen deprivation therapy

^*Bolded P Values are significant

Kaplan-Meier analysis demonstrated 89.6% versus 96.5% 10-year OS for CTX and no CTX groups, respectively (P < 0.01). See Figure 2 for Kaplan Meier curves demonstrating OS for unmatched cohorts.

Figure 2.

Kaplan-Meier curves demonstrating overall survival for unmatched cohorts.

With PSM by ratio of 1:5, a total of 5,870 patients were matched—4882 patients in the no CTX group and 988 patients in the CTX group. Table 4 demonstrates balance check for PSM groups; groups were well-matched. After PSM, 10-year OS was not significantly different between groups—89.6% versus 90.9% for the CTX and the no CTX groups, respectively (Figure 3; HR 0.99; 95% CI 0.82 – 1.19; P = 0.88).

Table 4.

Patient and Clinical Characteristics and Treatment Details for Propensity-Matched Cohorts

Variable	No CTX	CTX	P	SD
Age at diagnosis, median, y (SD)	59.79 (7.39)	59.45 (7.46)	0.332	0.034
Race, No. (%)
White	4,204 (86.11)	857 (86.74)	0.871	0.018
Black	464 (9.50)	90 (9.11		0.014
Others/Unknown	214 (4.38)	41 (4.15)		0.012
Charlson-Deyo Score, No. (%)
0	4,143 (84.86)	845 (85.53)	0.813	0.019
1	639 (13.09)	122 (12.35)		0.022
2+	100 (2.05)	21 (2.13)		0.005
Facility Type, No. (%)
Non-Academic/Research Program	1,752 (35.89)	334 (33.81)	0.213	0.044
Academic/Research Program	3,130 (64.11)	654 (66.19)		0.044
Year of Diagnosis, No. (%)
≥ 2004 – ≤ 2007	2,415 (49.47)	494 (50.00)	0.966	0.011
> 2007 – ≤ 2009	1,075 (22.02)	220 (22.27)		0.006
> 2009 – ≤ 2011	830 (17.00)	162 (16.40)		0.016
> 2011 – ≤ 2013	562 (11.51)	112 (11.34)		0.006
AJCC clinical T stage, No. (%)
1	1,636 (33.51)	328 (33.20)	0.887	0.007
2	1,613 (33.04)	319 (32.29)		0.016
3	733 (15.01)	159 (16.09)		0.030
4	63 (1.29)	15 (1.52)		0.019
Unknown	837 (17.14)	167 (16.90)		0.006
AJCC clinical N stage, No. (%)
0	3,316 (67.92)	659 (66.7)	0.642	0.026
1	201 (4.12)	46 (4.66)		0.026
Unknown	1,365 (27.96)	283 (28.64)		0.015
AJCC pathologic T stage, No. (%)
1–2	1,110 (22.74)	242 (24.49)	0.646	0.041
3	3,197 (65.49)	628 (63.56)		0.040
4	178 (3.65)	38 (3.85)		0.011
Unknown	397 (8.13)	80 (8.1)		0.001
AJCC pathologic N stage, No. (%)
0	3,114 (63.79)	621 (62.85)	0.786	0.019
1	1,175 (24.07)	248 (25.1)		0.024
Unknown	593 (12.15)	119 (12.04)		0.003
Gleason Score, No. (%)
2–7	1,659 (33.98)	347 (35.12)	0.736	0.024
8–10	2,951 (60.45)	584 (59.11)		0.027
Unknown	272 (5.57)	57 (5.77)		0.009
PSA, No. (%)
< 10	977 (20.01)	186 (18.83)	0.845	0.030
≤ 10– < 20	2,556 (52.36)	529 (53.54)		0.024
≥ 20	983 (20.14)	198 (20.04)		0.002
Unknown	366 (7.50)	75 (7.59)		0.004
Surgical Margin, No. (%)
Negative	2,596 (53.17)	529 (53.54)	0.901	0.007
Positive	2,066 (42.32)	412 (41.70)		0.013
Unknown	220 (4.51)	47 (4.76)		0.012
ADT, No. (%)
No	1,704 (34.90)	345 (34.92)	0.963	0.000
Yes	3,098 (63.46)	628 (63.56)		0.002
Unknown	80 (1.64)	15 (1.52)		0.010
Time from diagnosis to treatment onset (RP or CTX), median, days (SD)	61.24 (39.52)	59.75 (49.99)	0.301	0.035

Abbreviations: CTX, Chemotherapy; SD, Standardized difference; AJCC, American Joint Committee on Cancer; PSA, Prostate-specific antigen; ADT, Androgen deprivation therapy

Figure 3.

Kaplan-Meier curves demonstrating overall survival for propensity-matched cohorts.

Discussion

Low numbers of high-risk prostate cancer patients have been treated in the United States with regimens including CTX, and these patients are an especially high-risk group within the high-risk criteria. This was to be expected, as incorporation of CTX into a definitive treatment of prostate cancer is still, at this point, non-standard, and so a patient treated with CTX would have been most likely to be either 1) very high risk and judged by his physicians to potentially benefit from addition of CTX to improve outcomes or 2) enrolled on a clinical trial specifically testing addition of CTX. The numbers of patients for whom CTX was incorporated into the treatment regimen was significantly higher than the sum of all patients enrolled on known relevant clinical trials, meaning that at least some patients are being treated with this technique off-trial.

With PSM, the inferior OS for the CTX group dissipated, suggesting that in a group of more closely matched patients, at the very least CTX does not appear to worsen long-term OS. Notably, however, the NCDB is unable to account for potential lasting morbidities caused by CTX—such as neuropathy—which can impact quality of life in long-term survivors of treatment. Even though results from our PSM groups do not show a benefit from CTX in terms of OS, it is possible that a benefit might be seen in terms of freedom from biochemical recurrence, disease-free survival, or distant metastasis-free survival, all of which are not available through the NCDB. Additionally, PSM using the NCDB cannot account for many prognostic factors for RP patients—including pre-treatment PSA velocity, post-RP PSA nadir, and perineural invasion. While summed Gleason score is available, more than half of the patients in the database have unknown values of the primary and secondary Gleason patterns, and so this is another factor that cannot be taken into account via PSM. The NCDB also does not distinguish between adjuvant and early-salvage RT, meaning our adjuvant RT group is likely a mix between these different scenarios, with early-salvage RT representing a more unfavorable population. Identification of specific CTX agents is also not available in the NCDB. Due to evolving CTX paradigms (towards basis around docetaxel), many of the patients included in the NCDB during the timeframe analyzed were likely treated with regimens that would no longer be favored. Finally, the NCDB is subject to the inherent limitations of retrospective data; for example, it is possible that a small portion of our patient population may have been treated without definitive intent due to errors in coding, despite our careful selection criteria. It should be noted that the NCDB database used for this analysis ends at 2014, and the incidence of CTX use may rise in the future with ongoing publications of positive clinical trials.

CONCLUSIONS

Incorporation of CTX into the treatment regimen for high-risk prostate cancer patients treated with RP remains uncommon and investigational, as demonstrated by the available literature, and this analysis of the NCDB. Multimodal regimens such as RP followed by ADT, RT, and CTX, or RT in conjunction with ADT followed by CTX, have shown promise, but will require long-term follow-up of randomized data—treatment with these regimens should still be explicitly reserved for the clinical trial setting. Further studies may elucidate the particular clinical and pathologic scenarios for which a benefit from CTX might be maximized.

Abbreviation List

CTX

Chemotherapy

RP

Radical prostatectomy

ADT

Androgen deprivation therapy

RT

Radiation therapy

NCDB

National Cancer Data Base

OS

Overall survival

CoC

Commission on Cancer

AJCC

American Joint Committee on Cancer

PSA

Prostate-specific antigen

UVA

Univariate analysis

MVA

Multivariable analysis

PSM

Propensity score matching

HR

Hazard ratio

CI

Confidence interval