Highlights
- •Three next generation antiandrogens significantly prolong overall survival in nm-CRPC
- •Mature survival data for these treatments are now available for meta-analysis
- •Simulated patient data-level may help to highlight differences among trials
- •Different study designs and follow-up may affect data interpretation
Abstract
There was a high medical need for patients with non-metastatic castration-resistant prostate cancer (nmCRPC) when several next-generation anti-androgens (apalutamide, enzalutamide, and darolutamide) demonstrated clinically relevant delays in metastasis onset. However, to date, few publications have assessed the pooled effect of these treatments on overall survival (OS). We performed a systematic review and meta-analysis of all randomized, placebo-controlled studies investigating a systemic treatment in nmCRPC. Publications were identified by searching several databases on April 7, 2021. The primary objective of this analysis was to determine the OS benefit. Secondary outcomes included the relative risk (RR) of adverse events (AEs) and grade 3-4 AEs. A sensitivity analysis with simulated data was also conducted to examine the influence of the study designs on the results. Three randomized controlled studies (SPARTAN, PROSPER, ARAMIS) met our inclusion criteria. Pooled meta-analyses showed a significant benefit in OS with the active agents versus placebo (hazard ratio [HR] 0.74, 95% confidence interval [CI] 0.65–0.83), as well as increased risk of any grade (RR 1.09, 95% CI 1.01–1.17) and grade 3-4 AEs (RR 1.50, 95% CI 1.23–1.83). The sensitivity analysis with SPARTAN-like simulated populations demonstrated that when using ARAMIS statistical design, OS would be statistically significant in 98.1% of the cases, at a shorter follow-up and with lower number of events. First-line treatment of nmCRPC patients with anti-androgens increased OS with an acceptable safety profile. In light of the different study designs and follow-up, results should be interpreted separately.
Keywords
Introduction
Prostate cancer is the most frequent genitourinary tumor worldwide, with almost 1,500,000 new cases diagnosed annually.
1
Since the 1940s, androgen deprivation therapy (ADT) has been the backbone treatment for advanced prostate cancer.- Sung H
- Ferlay J
- Siegel RL
- et al.
Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries.
CA Cancer J Clin. 2021; 71 (Epub 2021 Feb 4. PMID: 33538338): 209-249https://doi.org/10.3322/caac.21660
2
Nevertheless, many patients on continuous ADT experience disease progression despite castrate levels of testosterone, transitioning to a disease stage known as “castration-resistant”.Non-metastatic castration-resistant prostate cancer (nmCRPC) refers to patients with rising prostate-specific antigen (PSA) levels despite castrate levels of testosterone and no evidence of distant metastasis.
3
These patients are mostly asymptomatic, with a similar health-related quality of life (QoL) to people of the same age in the general population.4
The development of metastasis in this population is the point of no return marking the final stage of the disease, associated with pain, functionality and QoL deterioration, and increased mortality.5
Metastasis-free survival (MFS) is acknowledged as a relevant endpoint in clinical studies for patients with nmCRPC,
6
and apalutamide, darolutamide, and enzalutamide were approved based on studies using MFS as a primary endpoint.7
, 8
, 9
These three agents are now recommended in international clinical guidelines for patients with high-risk nmCRPC.10
, 11
, 12
They have also demonstrated a significant benefit in overall survival (OS),13
, 14
, 15
and are the first agents to do so in nmCRPC.We conducted a systematic review and meta-analysis of clinical studies using these three agents with the aim of (1) estimating the pooled OS benefit of these treatments and (2) evaluating the relative risk of adverse events (AEs) and grade 3-4 AEs of these agents in patients with nmCRPC. We also examined if different statistical approaches were comparable using simulated patient-level data.
Materials and methods
This systematic review and meta-analysis is reported according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement.
16
, 17
The study protocol is available upon request, in lieu of a formal ethics committee review; the principles of the Helsinki Declaration were followed.Search Strategy
We performed a systematic search of the PubMed, Scopus, Cochrane Library, and Google Scholar databases for clinical studies in patients with nmCRPC published from database inception to April 7, 2021. The following key search terms were used: [(“castration resistant” AND “prostate cancer”) OR “CRPC”] AND (“non-metastatic” OR “nonmetastatic” OR “M0”) AND “overall survival”. The publication language was limited to English. When multiple publications on the same study were found, the most recent data were used.
Eligibility Criteria and Study Selection
We included prospective, randomized, placebo-controlled phase III studies of systemic therapy in surgically or medically castrated patients with nmCRPC and testosterone levels ≤50 ng/dL. Only studies with a significant benefit in OS among their outcomes were selected.
Two authors (OAC and AQG) independently performed the searches and screened identified publications for inclusion. Any discrepancies were resolved by a third author (JRMT).
Quality Assessment
Two authors (AQG and OAC) independently evaluated the risk of bias in each study using the Cochrane risk of bias tool (RoB 2).
18
Discrepancies were resolved by consensus. Randomization, blinding, control-group, crossover, follow-up time, and number of events for OS of each study were also used to assess the risk of bias.Data Extraction
Publication year, first author, title, journal, and identification number (PMID or similar) from each publication were extracted to a text file for screening purposes, then descriptive, efficacy and safety variables from each study were extracted by OAC and AQG from the full text of each article using a standardized form.
Statistical Analysis
A standard pairwise meta-analysis was performed using R statistical software (Version 1.1.419 – 2009–2018 RStudio, Inc.) with R package meta. Aggregated data from efficacy and safety variables were analyzed in the whole intention-to-treat (ITT) population. The primary outcome was OS, and safety data were considered as secondary outcomes. All measurements were processed as hazard ratios (HR) or risk ratios (RR) with 95% confidence intervals (CIs) when available. Chi-squared test of Q was employed to test heterogeneity among different studies, and I2 was used to quantify the inconsistency. Tau-squared was used to analyze the magnitude of heterogeneity. A P value of <.05 in chi-squared test and/or an I2 >75% was considered indicative of heterogeneity.
The meta-analysis was performed using a fixed-effect model (Mantel-Haenszel method) if P > .1 and I2 <50%; otherwise, a random-effect model (DerSimonian-Laird method) was used.
19
, 20
A P value of <.05 was considered statistically significant.Simulated Patient Data
Considering the similarities and differences in design and follow-up of the clinical studies, we hypothesized that small differences in statistical design between studies could affect the magnitude of the results. Hence, we conducted an analysis using simulated patient-level data to determine if the different statistical approaches used in the included studies were comparable as described elsewhere.
21
SPARTAN data was used as a model due to the longer follow-up. Individual patient data were obtained from published Kaplan-Meier curves using mathematical algorithms.22
Simulated patient data were obtained by resampling the estimated survival functions obtained from the SPARTAN study
9
to evaluate if there would have been different statistical outcomes in OS if the SPARTAN statistical plan (eg, follow-up, alpha-spending function) had been different. Simulated populations were created solving an indeterminate compact system of equations that arose from the imposition of the OS results in the study subgroups. A multidisciplinary group of clinicians specialized in the treatment of advanced prostate cancer analyzed the simulated data and searched for clinical inconsistencies based on their expertise.Results
Search Results and Risk of Bias
The initial literature search yielded 378 potentially relevant publications. After removal of duplicates, reviews, opinion articles, and retrospective studies (Supplementary Figure S1), three publications were included (Table 1) for the SPARTAN,
14
PROSPER,15
and ARAMIS studies.13
Table 1Summary and baseline characteristics of the included studies.
| Study nameAuthors, year | Population | N | Experimental arm (n) Control arm (n) | Patient crossover n (%) | PSA blinding | Assigned treatment Discontinuations n (%) | Discontinuations other than AEs or PD n (%) | Median follow-up, months | Deaths n (%) |
|---|---|---|---|---|---|---|---|---|---|
| SPARTAN Smith et al 2021 14 | nmCRPC PSA-DT ≤10 months Local disease permitted when LN below aortic bifurcation <2 cm | 1207 | Apalutamide (803) Placebo (401) b | 76 (19.1) | Y | 566 (70.5) 398 (100) | 90 (11.2) 62 (15.5) | 48.0 | 427 (35.6) |
| PROSPER Sternberg et al 2020 15 | nmCRPC PSA-DT ≤10 months Local disease permitted when LN below aortic bifurcation <1.5 cm | 1401 | Enzalutamide (933) Placebo (468) c | 87 (18.6) | Y | 552 (59.2) 465 (100) | 116 (12.4) 70 (15.0) | 52.0 | 466 (33.3) |
| ARAMIS Fizazi et al 2020 13 | nmCRPC PSA-DT ≤10 months Local disease permitted when LN below aortic bifurcation <2 cm | 1509 | Darolutamide (955) Placebo (554) | 170 (30.7) | N | 488 (51.1) 554 (100.0) | 137 (14.3) 184 (33.2) | 29.0 | 254 (16.8) |
Abbreviations: AEs = adverse events; LN = lymph nodes; N = no; n = number; PD = progressive disease; PSA = prostate-specific antigen; PSA-DT = prostate-specific antigen doubling time; Y = yes.
a Reasons for discontinuation include: patient withdrawal and other reasons (SPARTAN and PROSPER) and personal reasons and investigators judgement (ARAMIS). bcThree randomized patients did not receive placebo.
Although all studies had a similar study design (Table 1) and positive results for OS (Table 2), they had some design and data differences.
15
, 14
, 13
The three studies showed low risk of bias.Table 2Summary of efficacy in the included studies.
| Study | SPARTAN 14 | PROSPER 15 | ARAMIS 13 |
|---|---|---|---|
| Comparators | Apalutamide vs. placebo | Enzalutamide vs. placebo | Darolutamide vs. placebo |
| MFS | |||
| Median, months | 40.5 vs. 16.2 | 36.6 vs. 14.7 | 40.4 vs. 18.4 |
| HR (95% CI) | 0.28 (0.23–0.35) | 0.29 (0.24–0.35) | 0.41 (0.34–0.50) |
| Net benefit, months | 24.3 | 21.9 | 22.0 |
| OS | |||
| Median, months (95% CI) | 73.9 (61.2–NR) vs. 59.9 (52.8–NR) | 67.0 (64.0–NR) vs. 56.3 (54.4–63.0) | NR vs. NR |
| HR (95% CI) | 0.78 (0.64–0.96); P = .016 | 0.73 (0.61–0.89); P = .001 | 0.69 (0.53–0.88); P = .003 |
| Net benefit, months | 14 | 10.7 | – |
| Protocol defined events at Final analysis | 427 | 596 | 240 |
| Final analysis | 428 | 466 | 254 |
| Sensitivity analysis, OS HR | 0.69 | – | 0.59–0.68 23 |
| Net benefit | 21.1 | – | – |
| Treatment duration, months | 32.9 vs. 11.5 | 33.9 vs. 14.2 | 25.8 vs. 11.6 |
| In patients who underwent crossover | 26.1 | 14.5 | 11.0 |
| Subsequent treatment, n (%) | 386 (48.1) vs. 285 (71.6) | 310 (33.3) vs. 303 (65.2) | 141 (14.8) vs. 307 (55.4) |
| % active treatments after discontinuation/total discontinuations | 68.2 vs. 88.5 | 56.1 vs. 65.2 | 29.0 vs. 55.4 |
Abbreviations: CI = confidence interval; HR = hazard ratio; MFS = metastasis-free survival; NA/NR = not available/not reported; NR = not reached; OS = overall survival.
a medians refer to time on active treatment after crossover from placebo arm
b including 170 patients which crossed from placebo to receive darolutamide
Overall Survival
All studies had OS among secondary endpoints and demonstrated a significant benefit (Table 2). Different interim analyses and alpha-spending functions were used for each study. In PROSPER, three interim analyses, and a final analysis of OS were planned after 135 (9.6%), 285 (20.3%), 440 (31.4%), and 596 (42.5%) deaths, respectively. Only the first
8
and third15
interim analyses were published, the last one being considered final because of the statistically significant improvement in OS observed. Two interim analyses and one final analysis were planned for the SPARTAN study when 104 (8.6%), 285 (23.6%), and 426 (35.3%), respectively, deaths had occurred.9
, 14
, 24
For the ARAMIS study, one interim and one final analysis for OS were planned when 140 (9.3%) and 240 (15.9%) deaths had occurred.7
, 13
PROSPER and SPARTAN used the O'Brien-Fleming alpha-spending function to set the significance threshold in all analyses,14
, 15
while ARAMIS used an exponential function with Rho10.13
Pooled meta-analysis showed a significant overall benefit regarding OS in patients treated with the systemic treatment versus placebo (HR 0.74, 95% CI 0.65–0.83; Figure 1). The I2 test showed low inconsistency among the studies (I2=0%; P = .75).
Figure 1Forest plot of overall survival in the SPARTAN, PROSPER, and ARAMIS studies. ADT = androgen deprivation therapy; CI = confidence interval; HR = hazard ratio; seTE = standard error treatment effect; TE = treatment effect.
The treatment effect on OS was consistent in subgroup analysis by PSA-DT (≤6 vs. >6 months), use of bone sparing agents (Yes vs. No), and presence of locoregional nodal disease (No vs. Yes) (Supplementary Table S2 and Figures S2, S3 and S4).
Safety Analysis
All three studies evaluated grade -3-5 AEs in the ITT populations; PROSPER and SPARTAN assessed AEs using the Common Terminology Criteria for Adverse Events (CTCAE) version 4.0,
14
, 15
and ARAMIS used CTCAE version 4.03.13
PROSPER and SPARTAN reported safety profiles after a similar follow-up (48 and 52 months, respectively), with median treatment durations of 33.9 and 32.9 months, respectively.
14
, 15
ARAMIS initially reported the safety analysis only for the double-blind treatment period, with a median duration of exposure of 18.5 months in the final OS analysis.13
Subsequently, updated safety data from ARAMIS were reported for the combined double-blind and open-label treatment periods and with a median treatment exposure of 25.8 months (Table 3).25
The proportion of patients with any AE, grade 3-4 AEs and serious AEs were different between the studies in placebo arms. AEs leading to discontinuation and grade 5 AEs were low in all three studies.Table 3Summary of safety in the included studies.
| SPARTAN 14 | PROSPER 15 | ARAMIS 13 | ||
|---|---|---|---|---|
| Double-blind study | Open-label extension | |||
| Comparators | Apalutamide vs. placebo | Enzalutamide vs. placebo | Darolutamide vs. placebo | |
| AEs, n (%) | ||||
| Any AE | 1154 (96.1) | 1256 (89.7) | 1257 (83.3) | 1296 (85.9) |
| By comparator | 781 (97.3) vs. 373 (93.7) | 876 (94.2) vs. 380 (81.7) | 818 (85.7) vs. 439 (79.2) | 857(89.8) vs. 439 (79.2) |
| AEs leading to death | 26 (2.2) | 54 (3.9) | 57 (3.8) | – |
| By comparator | 24 (3.0) vs. 2 (0.5) | 51 (5.5) vs. 3 (0.7) | 38 (4.0) vs. 19 (3.4) | – |
| Grade 3–4 AEs | 594 (49.5) | 572 (40.8) | 371 (24.6) | 423 (28.1) |
| By comparator | 449 (55.9) vs. 145 (36.4) | 446 (48.0) vs. 126 (27.1) | 251 (26.3) vs. 120 (21.7) | 303 (31.8) vs. 120 (21.7) |
| Serious AEs | 389 (32.4) | 472 (33.8) | 370 (24.5) | 427 (28.3) |
| By comparator | 290 (36.1) vs. 99 (24.9) | 372 (40.0) vs. 100 (21.5) | 249 (26.1) vs. 121 (21.8) | 306 (32.1) vs. 121 (21.8) |
| AEs leading to discontinuation | ||||
| By comparator | 120 (14.9) vs. 29 (7.3) | 158 (17.0) vs. 41 (8.8) | 85 (8.9) vs. 48 (8.7) | 100 (10.5) vs. 48 (8.7) |
AE = adverse event.
a PROSPER reported grade ≥3 AEs. –, not reported
Pooled meta-analysis showed a significant risk for any AEs (RR 1.09, 95% CI 1.01–1.17; Figure 2A) and grade 3/4 AEs (RR 1.50, 95% CI 1.23–1.83; Figure 2B) with active treatment versus placebo. I2 showed a high inconsistency among the study results for any grade AEs (I2=89%, P < .01) and grade 3/4 AEs (I2=77%, P = .01).
Figure 2Forest plot of (A) any grade and (B) grade 3/4 adverse events in the SPARTAN, PROSPER, and ARAMIS studies. ADT = androgen deprivation therapy; CI = confidence interval; RR = risk ratio; ADT = androgen deprivation therapy.
However, when the updated safety data from ARAMIS was used in the pooled meta-analysis for grade 3/4 AEs, the inconsistency among the studies significantly decreased (I2=25%; P = .27; Figures 3A-B).
Figure 3Forest plot of (A) any grade and (B) grade 3/4 adverse events in the SPARTAN, PROSPER, and ARAMIS studies, including the long-term follow-up data in ARAMIS. ADT = androgen deprivation therapy; CI = confidence interval; RR = risk ratio.
Given the differences in follow-up and duration of exposure across studies, we also assessed relative risks by comparing individual AE risk in the active arms with the placebo arms (Table 4 and Figures. 4-5). This analysis showed that enzalutamide was associated with higher RR for fatigue, hypertension, falls, and fractures. Apalutamide was associated with a higher RR for rash and diarrhea. Including the updated safety data and 7.3 additional months of exposure, darolutamide was associated with a higher incidence of all individual AEs, including any-grade rash.
Table 4Relative risks of individual AEs in the included studies.
| SPARTAN 14 | PROSPER 15 | ARAMIS 13 | ||
|---|---|---|---|---|
| Double-blind study | Open-label extension | |||
| Comparators | Apalutamide vs. placebo | Enzalutamide vs. placebo | Darolutamide vs. placebo | |
| Median duration of treatment, months | 32.9 vs. 11.5 | 33.9 vs. 14.2 | 18.5 vs. 11.6 | 25.8 vs. 11.6 |
| Any grade AEs, % patients (RR) | ||||
| Fatigue | 33 vs. 21 (1.6) | 46 vs. 22 (2.1) | 13.2 vs. 8.3 (1.6) | 14.3 vs. 8.3 (1.7) |
| Hypertension | 28 vs. 21 (1.3) | 18 vs. 6 (3.0) | 7.8 vs. 6.5 (1.2) | 9.0 vs. 6.5 (1.4) |
| Falls | 22 vs. 9.5 (2.3) | 18 vs. 5 (3.6) | 5.2 vs. 4.9 (1.1) | 6.9 vs. 4.9 (1.4) |
| Fractures | 18 vs. 7.5 (2.4) | 18 vs. 6 (3.0) | 5.5 vs. 3.6 (1.5) | 8.3 vs. 3.6 (2.3) |
| Rash | 26 vs. 6.3 (4.1) | 4 vs. 3 (1.3) | 3.1 vs. 1.1 (2.8) | 3.8 vs. 1.1 (3.5) |
| Diarrhea | 23 vs. 15 (1.5) | 12 vs. 10 (1.2) | – | – |
| Musculoskeletal events | – | 34 vs. 23 (1.5) | – | – |
| Cardiac arrhythmia | – | – | 7.3 vs. 4.3 (1.7) | – |
| Hot flush | 15 vs. 8.5 (1.7) | 14 vs. 8 (1.8) | 6 vs. 4.5 (1.3) | – |
| Grade 3–4 AEs, % patients (RR) | ||||
| Fatigue | 0.9 vs. 0.3 (3.0) | 4 vs. 1 (4.0) | 0.4 vs. 0.9 (0.4) | – |
| Hypertension | 16 vs. 12 (1.3) | 6 vs. 2 (3.0) | 3.5 vs. 2.3 (1.5) | – |
| Falls | 2.7 vs. 0.8 (3.3) | 2 vs. 1 (2.0) | 0.9 vs. 0.7 (1.3) | – |
| Fractures | 4.9 vs. 1 (4.9) | – |
| – |
| Rash | 5.2 vs 0.3 (17.3) | – | 0.2 vs. 0.2 (1.0) | – |
| Diarrhea | 1.5 vs. 0.5 (3.0) | 1 vs. <1 (1.3) | – | – |
| Cardiac arrhythmia | – | – | 1.8 vs. 0.7 (2.6) | – |
| Hot flush | 0.0 vs. 0.0 | – | 0.0 vs. 0.0 | – |
AE = adverse event; RR = relative risk.
a PROSPER reported grade ≥3 AEs. –, not reported
Figure 4Spider plot of studies for any grade adverse events.
Figure 5Spider plot of studies for grade 3/4 adverse events.
Simulated Patient Data
To check whether the differences in follow-up, number of planned interim analysis, and alpha-spending functions could affect the punctual estimation of OS in all studies (Supplementary Table S1), we analyzed the SPARTAN data (longer follow-up), under the ARAMIS study design conditions (shortest follow-up). The use of two and one interim analysis in ARAMIS and PROSPER, together with the use of the O'Brien-Fleming and Rho10 exponential functions, respectively, directly affected the alpha boundary of both studies. In the absence of individual patient data, a set of different simulated populations were generated using the same median OS times and HR as in the SPARTAN final analysis in each patient subgroup (based on age, Eastern Cooperative Oncology Group performance status, PSA-DT, and presenceor absence of loco-regional disease).
14
The simulated populations were then validated by expert clinicians in nmCRPC patients.Simulated populations can be analyzed at any point of planned events and simulate censored patients. Thus, the distribution of significant results in these SPARTAN simulated populations with two interim analyses and a final analysis using the O'Brien-Fleming function were more likely to be positive at the second interim analysis. If one interim analysis and one final analysis with an exponential Rho10 alpha-spending function are used, the method maximizes the probability of positive result at the final analysis (Supplementary Figure S5).
The analysis of SPARTAN-like simulated populations using one interim analysis and one final analysis (ARAMIS design) with potential function Rho10 achieved a significant advantage in OS in 98.1% of the cases with 28% total events (67% of planned events) at the second interim analysis (Figure 5A). This strategy maximized the highest P value boundary at the second stop despite less follow-up time. This is favored by an elevated Rho value in potential alpha-spending function, while small Rho values would perform similar to the O'Brien-Fleming function (Supplementary Figure S6).
Figure 5ACumulative positive results with SPARTAN simulated populations using SPARTAN (blue) and ARAMIS (orange) statistical designs.
Discussion
We conducted a systematic review and meta-analysis of systemic treatments with OS benefit in nmCRPC and found that apalutamide, enzalutamide, and darolutamide showed clear advantages in OS.
13
, 14
, 15
Previous meta-analyses have also demonstrated a statistically significant benefit in OS with next generation anti-androgens in nmCRPC,26
but these were performed with immature data from interim analyses with non-significant OS, whereas ours used the most up-to-date results. Our data show that in patients with nmCRPC receiving treatment with these molecules, risk of death would be reduced by an 26%. Across trials, reported HR ranged from 0.69 (ARAMIS) to 0.78 (SPARTAN). When looking at the net benefit of the active treatment compared to the control arm, enzalutamide and apalutamide would provide additional 11 and 14 months of OS, respectively, while ARAMIS has not yet achieved survival medians preventing this calculation. This 11-14 months increment on survival represents a considerable magnitude of benefit, especially when compared with survival seen in later stages of the disease.27
, 28
- Ryan CJ
- Smith MR
- Fizazi K
- et al.
Abiraterone acetate plus prednisone versus placebo plus prednisone in chemotherapy-naive men with metastatic castration-resistant prostate cancer (COU-AA-302): final overall survival analysis of a randomised, double-blind, placebo-controlled phase 3 study.
Lancet Oncol. 2015; 16: 152-160
However, clear differences among the studies regarding follow-up, number of patients, and statistical design were apparent.
13
, 14
, 15
First, the follow-up period was shorter in the ARAMIS study (29 months),13
where median OS was not reached, compared with follow-up in PROSPER (48 months)15
and SPARTAN (52 months),14
which both had more mature OS data.PROSPER, SPARTAN and ARAMIS recruited 1401,1207 and 1509 patients respectively, the latter having the highest statistical power because it recruited 25% more patients than SPARTAN. All studies considered the use of alpha-spending functions to determine OS stopping boundaries in the interim analysis. While SPARTAN and PROSPER used the O'Brien-Fleming alpha-spending function,
14
, 15
ARAMIS used a Rho-family spending function with parameter Rho10.13
Both methods are valid when interim analyses are planned and alpha spending needs to be distributed among them. However, the distribution of alpha spending differed between the O'Brien-Fleming and exponential methods when a Rho10 parameter was set. The higher Rho parameter, the greater the chance of positive results in the final analysis when compared with the O'Brien-Fleming method.13
, 14
, 15
Due to variability in the number of patients, follow-up time, and statistical design for OS, we performed a sensitivity analysis with a simulated SPARTAN population, according to accepted practices.
21
, 29
By changing the alpha-spending function, we demonstrated that SPARTAN could have achieved a positive result earlier. Also, punctual estimation of OS is usually less precise with a short follow-up period.Patients in the placebo arm of SPARTAN had the highest incidence of any AE (93.7%) compared with PROSPER (81.7%) and ARAMIS (79.2%), and a higher occurrence of grade 3/4 and individual AEs. The visit schedules were different between studies. In SPARTAN, AEs were initially evaluated every 4 weeks, although a protocol amendment towards the end of the study allowed for a decrease in the frequency after the first year of treatment.
14
In contrast, PROSPER and ARAMIS evaluated AEs each 16 weeks of treatment.13
, 15
These differences may have introduced bias, especially for AEs reported by patients during study visits and without an impact on clinical history (falls, hot flushes, fatigue, diarrhea), suggesting that frequency of assessment affected AE incidence.30
Additionally, the reporting period for safety results were different among studies. In PROSPER and SPARTAN, the long-term safety profile in the experimental arm was reported after more than 33.9 and 32.9 months of treatment, respectively, whereas ARAMIS safety data were limited to the double-blind period and reported after 18.5 months in the experimental arm, 11.6 months for the placebo arm, and 11 months in the crossover group.
13
Updated data from the double-blind and open-label treatment periods of ARAMIS showed increases in the incidence of any-grade, serious, and grade 3/4 AEs with longer follow-up and treatment exposure.25
Therefore, the safety profile of darolutamide in ARAMIS should be taken in the context of the shorter follow-up duration and exposure to active drug compared with SPARTAN and PROSPER.30
Conclusions
This meta-analysis showed a significant improvement in terms of OS, and an increase in any-grade and grade 3/4 AEs. Exhaustive examination of apparently similar studies revealed that differences in design, follow-up, number of patients, and the statistical plan exist, potentially playing a role in the interpretation of results. Using simulated patient-level data, results can vary using similar, but not equivalent, statistical approaches, highlighting the critical role that the statistical design plays in the results of each study. Studies with less mature data due to shorter follow-up, although valid in terms of statistical significance, are more uncertain in terms of punctual estimation of OS.
Author contributions
All authors contributed to the development of the manuscript draft, the development of inclusion criteria, the risk of bias assessment, the data extraction criteria, and performing the search strategy, and provided statistical expertise and expertise on nmCRPC. All authors have read and provided feedback on the manuscript draft, and approved the final draft of the manuscript for submission.
Disclosure
Óscar Amor-Carro, Jacobo Muñoz del Toro, Alberto Arribas and Alba Quesada are employees of Janssen. Dr. Eduardo Useros Rodríguez is a former employee of Janssen. Dr. Andrés Rodríguez has received consulting and speaker fees from Janssen and speaker fees from Astellas. Dr. Fernando López-Campos has received consulting and speaker fees from Janssen, Astellas, AstraZeneca and Merck Sharp & Dohme (MSD), and is receiving research funding from Astellas. Dr Alejo Rodriguez-Vida reports serving in an advisory role for MSD, Pfizer, Bristol Myers Squibb (BMS), Astellas, Janssen, Bayer, Clovis, Telix, and Roche; is receiving honoraria or travel expenses from Pfizer, MSD, Astellas, BMS, Janssen, AstraZeneca, Roche, Bayer, and Sanofi Aventis; and is receiving research funding from Takeda, Pfizer, and Merck.
Acknowledgments
We would like to thank Simone Tait of Springer Healthcare Communications who edited the manuscript before submission. This medical writing assistance was funded by Janssen. This work was supported by Janssen. The funding supported expenses related to the publication of the manuscript, any software requirements, and access to the full versions of all papers needed for the review. The authors did not receive any funding in terms of participation in this publication.
Appendix. Supplementary materials
References
- Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries.CA Cancer J Clin. 2021; 71 (Epub 2021 Feb 4. PMID: 33538338): 209-249https://doi.org/10.3322/caac.21660
- Timing of androgen-deprivation therapy in patients with prostate cancer with a rising PSA (TROG 03.06 and VCOG PR 01-03 [TOAD]): a randomised, multicentre, non-blinded, phase 3 trial.Lancet Oncol. 2016; 17: 727-737
- Managing nonmetastatic castration-resistant prostate cancer.Eur Urol. 2019; 75: 285-293
- Effect of apalutamide on health-related quality of life in patients with non-metastatic castration-resistant prostate cancer: an analysis of the SPARTAN randomised, placebo-controlled, phase 3 trial.Lancet Oncol. 2018; 19: 1404-1416
- Metastases-free survival rate in non-metastatic castration-resistant prostate cancer: now living on its own! [Article in Spanish].Actas Urol Esp. 2019; 43: 167-168
- An FDA review of drug development in nonmetastatic castration-resistant prostate cancer.Clin Cancer Res. 2020; 26: 4717-4722
- Darolutamide in nonmetastatic, castration-resistant prostate cancer.N Engl J Med. 2019; 380: 1235-1246
- Enzalutamide in men with nonmetastatic, castration-resistant prostate cancer.N Engl J Med. 2018; 378: 2465-2474
- Apalutamide treatment and metastasis-free survival in prostate cancer.N Engl J Med. 2018; 378: 1408-1418
- Prostate cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up.Ann Oncol. 2020; 31: 1119-1134
- EAU-EANM-ESTRO-ESUR-SIOG Guidelines on Prostate Cancer.European Association of Urology. 2021;
- NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines): Prostate Cancer. Version 2.2021.National Comprehensive Cancer Network. 2021;
- Nonmetastatic, castration-resistant prostate cancer and survival with darolutamide.N Engl J Med. 2020; 383: 1040-1049
- Apalutamide and overall survival in prostate cancer.Eur Urol. 2021; 79: 150-158
- Enzalutamide and survival in nonmetastatic, castration-resistant prostate cancer.N Engl J Med. 2020; 382: 2197-2206
- Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015.statement. Syst Rev. 2015; 4: 1
- Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement.BMJ. 2009; 339: b2535
- RoB 2: a revised tool for assessing risk of bias in randomised trials.BMJ. 2019; 366: l4898
- Meta-analysis in clinical trials.Control Clin Trials. 1986; 7: 177-188
- Meta-analysis: fixed-effect model.Am J Orthod Dentofacial Orthop. 2020; 157: 134-137
- Using simulation studies to evaluate statistical methods.Stat Med. 2019; 38: 2074-2102
- Enhanced secondary analysis of survival data: reconstructing the data from published Kaplan-Meier survival curves.BMC Med Res Methodol. 2012; 12: 9
- Analysis of the effect of crossover from placebo (PBO) to darolutamide (DARO) on overall survival (OS) benefit in the ARAMIS Trial.J Clin Oncol. 2021; 39: 240
- Apalutamide and overall survival in non-metastatic castration-resistant prostate cancer.Ann Oncol. 2019; 30: 1813-1820
- Safety of darolutamide (DARO) for nonmetastatic castration-resistant prostate cancer (nmCRPC) from extended follow-up in the phase III ARAMIS trial.J Clin Oncol. 2021; 39: 239
- Apalutamide, enzalutamide, and darolutamide for non-metastatic castration-resistant prostate cancer: a systematic review and network meta-analysis.Int J Clin Oncol. 2020; 25: 1892-1900
- Enzalutamide in men with chemotherapy-naive metastatic castration-resistant prostate cancer: extended analysis of the phase 3 PREVAIL study.Eur Urol. 2017; 71: 151-154
- Abiraterone acetate plus prednisone versus placebo plus prednisone in chemotherapy-naive men with metastatic castration-resistant prostate cancer (COU-AA-302): final overall survival analysis of a randomised, double-blind, placebo-controlled phase 3 study.Lancet Oncol. 2015; 16: 152-160
- An evaluation of homogeneity tests in meta-analyses in pain using simulations of individual patient data.Pain. 2000; 85: 415-424
- Inferences about drug safety in phase 3 trials in oncology: examples from advanced prostate cancer.J Natl Cancer Inst. 2020; : djaa134
Article Info
Publication History
Published online: November 21, 2021
Accepted:
November 13,
2021
Received in revised form:
November 12,
2021
Received:
September 2,
2021
Identification
Copyright
© 2021 The Authors. Published by Elsevier Inc.
User License
Creative Commons Attribution (CC BY 4.0) | How you can reuse 
Elsevier's open access license policy
Creative Commons Attribution (CC BY 4.0)
Permitted
- Read, print & download
- Redistribute or republish the final article
- Text & data mine
- Translate the article
- Reuse portions or extracts from the article in other works
- Sell or re-use for commercial purposes
Elsevier's open access license policy
00219-6&id=gr1.jpg){kind=link}
00219-6&id=gr2.jpg){kind=link}
00219-6&id=gr3.jpg){kind=link}
00219-6&id=gr4.jpg){kind=link}
00219-6&id=gr5.jpg){kind=link}
00219-6&id=gr6.jpg){kind=link}