First-Line Therapy for Metastatic Castration-sensitive Prostate Cancer: a Network Meta-analysis

Hong Kong J Radiol 2022 Mac;25(1):6-15

 

 

¹ Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong

² Department of Clinical Oncology, The University of Hong Kong, Hong Kong

 

  Full Article in PDF

Prostate cancer (PC) has the second highest incidence and is the third leading cause of cancer death among men in the world. In 2018, more than 1.2 million new cases were diagnosed and 359,000 deaths were reported worldwide.[1] Approximately 89% of newly diagnosed cases are locoregional, with a 5-year survival rate of nearly 100%; 6% are diagnosed at the metastatic stage, which has a significantly worse 5-year relative survival of 30.2%.[2]

 

Androgen deprivation therapy (ADT) has been the standard of care for patients with metastatic PC since Huggins and Hodges discovered the hormone dependency of PC in the 1940s.[3] ADT was proven to be effective in producing improved radiological and biochemical profiles and prolonging overall survival (OS).[4] Multiple randomised controlled trials (RCTs) have been performed to evaluate the benefits of other therapies added to ADT for metastatic castration-sensitive PC (mCSPC). Following the STAMPEDE Arm C,[5] [6] CHAARTED,[7] [8] and GETUG-AFU 15 trials,[9] [10] docetaxel (Doce) plus ADT was recommended as first-line treatment (especially for high-volume disease) for mCSPC in 2015.[11] Treatment protocols have further evolved over the past 5 years. Positive results from the LATITUDE,[12] [13] STAMPEDE Arm G,[14] [15] TITAN,[16] ARCHES[17] and ENZAMET[18] studies have shown that addition of abiraterone acetate plus prednisolone (AAP), apalutamide (APA), or enzalutamide (ENZA) to ADT showed superior results compared to ADT alone in mCSPC and are now considered standard treatment protocols.[19] However, no head-to-head RCT has been conducted to compare the survival benefits of these regimens. We therefore conducted a network meta-analysis to guide the selection of the best first-line combination therapy for mCSPC.

 

This study was conducted in accordance with the PRISMA (Preferred Reporting Items for Systematic Review and Meta-analysis) guidelines.[20] We included RCTs only if they included treatment of mCSPC. We specifically included trials of treatment regimens containing ADT plus AAP, APA, Doce, or ENZA to assess the efficacy in systemic treatments. Included clinical trials needed to contain a control arm with ADT alone or with placebo for carrying out indirect comparison by network meta-analysis.

 

We performed a comprehensive systematic literature search for full-length journal publications, which included mCSPC ADT. The PubMed/MEDLINE Ovid, Embase, Cochrane Library, CINAHL Databases, trial registries, and other sources were employed and searched from 1990 to the present day, with the most recent search carried out on 30 September 2020. Key words and medical sub-heading (MeSH) terms included: ‘prostate cancer’, ‘metastatic’, ‘castration-sensitive’, and ‘treatment’. Key words used for searching included: “prostate” OR “prostatic”; “neoplasm” OR “cancer” OR “cancers” OR “cancerous” “tumor” OR “tumour”; “metastatic” OR “metastasis”; “treatment” OR “therapy”; and “castration-sensitive” OR “castration”. We aimed at identifying any journal articles or abstract proceedings that published the efficacy of first-line therapy in mCSPC. The key words and MeSH terms within each concept were then separated by the Boolean operator ‘AND’. Only articles written in English were included. The details and results of the literature search are provided in Figure 1 and Table 1.

 

Figure 1. PRISMA flowchart showing the results of systematic review identified from PubMed/MEDLINE, Ovid, Embase, Cochrane Library, CINAHL Databases, trial registries and other sources.

 

Table 1. Characteristics of included trials.

 

The results of the findings were first screened by title and abstract for appropriateness by two independent reviewers (KYCZ and AKHF), while the third reviewer (THS) was consulted if there were any disputes on selection of relevant literature. For relevant abstracts, the full papers would be reviewed for inclusion. More updated versions of the publications were adopted for meta-analysis when more than one version of the same studies were found. Two authors (KYCZ and AKHF) were responsible for independent data extraction based on the same criteria.

 

The primary endpoint of our study was OS, defined as the time from the date of randomisation until the date of death from any cause. The secondary endpoint was progression-free survival (PFS), the time from the date of randomisation to the date of first disease progression (locoregional or distant) or death from any cause, whichever occurred earlier.

 

The assessment of study quality for all RCTs was carried out independently by two authors (KYCZ and AKHF) using the Cochrane Collaboration tool.[21] Domains for the risk of bias assessment included: (1) the randomisation process; (2) deviations from the intended interventions; (3) missing data; (4) measurements of the outcomes (OS, PFS); and (5) selection of reported results. Risk of bias was judged as ‘low risk’, ‘some concerns’, or ‘high risk’ for individual domains based on the information provided by the authors of the included trials Table 2.

 

Table 2. Risk of bias summary

 

Our primary analysis in comparing the efficacy of different therapies on mCSPC was performed with both random- and fixed-effects models under a frequentist framework. Reported hazard ratios (HRs) for OS and PFS in eight included trials were incorporated into the mathematical model. Since PFS was displayed in various forms among different studies, we adopted radiological PFS as our primary interest. For the studies with no radiological PFS provided, clinical or biochemical PFS was adopted as a surrogate in our analysis.

 

Relative effects of each treatment were evaluated in indirect comparison by network meta-analysis. The I² and Q statistics were used to quantify the heterogeneity among different trials; an I² value of >50% and/or significant Q statistic at p < 0.1 was regarded as significant heterogeneity. P-scores were further employed to evaluate the mean extent of probability that one treatment outperforms the others.[22] In other words, the higher the P-score, the higher the certainty that a treatment is better than the others. Efficacy of treatments on OS and PFS was ranked by the relative P-score in our analysis.

 

Publication bias could not be formally assessed in this network meta-analysis because of the small number of trials included. Despite the real potential for this bias, given the relatively small number of trials, we judged the certainty in the evidence was unlikely to be decreased by this concern.

 

All statistical analyses were performed using opensource software, (R version 4.0.2, R Foundation for Statistical Computing, Vienna, Austria, 2018). A p value of <0.05 was considered statistically significant.

 

The network meta-analysis consisted of eight trials and 7790 patients, including CHAARTED, GETUG, LATITUDE, TITAN, ARCHES, and ENZAMET. Two studies from STAMPEDE (with Doce+ADT and AAP+ADT as their experimental regimens, respectively).[5] [6] [7] [8] [9] [10] [12] [13] [14] [15] [16] [17] [18] We also included the most updated results of STAMPEDE, from the 2020 European Society for Medical Oncology Congress.[15]

 

ADT was the reference category. Treatments were grouped into four categories: Doce+ADT, AAP+ADT, APA+ADT and ENZA+ADT. The network is displayed in Figure 2.

 

Figure 2. Schematic of the Network of Evidence Used in Network Meta-analysis for metastatic castration-sensitive prostate cancer treatments. The size of nodes of the diagram represents the total study population adopting the treatment and the thickness of the line is proportional to the number of comparisons among the studies.

 

AAP+ADT had the longest OS, followed by APA+ADT, ENZA+ADT and Doce+ADT with respective P-scores of 86%, 70%, 61% and 33% where a higher score means a higher probability of being the best treatment (Figure 3). Their corresponding HRs (95% confidence interval; 95% CI) compared to ADT alone were 0.63 (0.56-0.71), 0.67 (0.51-0.89), 0.70 (0.57-0.87), and 0.79 (0.71-0.89). There was no significant heterogeneity observed (I² = 0%; p = 0.610 for the Q statistic).

 

Figure 3. Forest plot for OS showing results comparing mCSPC treatments against ADT alone (upper) and Doce+ADT (lower) from network meta-analysis. HR<1 is in favour of ADT alone (upper) / docetaxel (lower).

 

Results of our study in PFS did not match with our analysis in OS. Figure 4 shows that ENZA+ADT had the highest P-score of 98%, implying a high certainty that ENZA+ADT was the best treatment in terms of PFS (HR=0.40, 95% CI=0.34-0.46). APA+ADT, AAP+ADT and Doce+ADT had P-scores of 75%, 51% and 26%, respectively. Their corresponding HRs (95% CI) were 0.48 (0.39-0.60), 0.58 (0.52-0.65) and 0.67 (0.60-0.74).

 

Figure 4. Forest plot for PFS showing results comparing mCSPC treatments against ADT alone (upper) and Doce+ADT (lower) from network meta-analysis. HR<1 is in favour of ADT alone (upper) / docetaxel (lower).

 

The same results were obtained in fixed- and randomeffects models in both OS and PFS analyses.

 

Figure 3 shows our explorative analysis in OS comparing different treatment combinations against Doce+ADT. Our study revealed that only AAP+ADT showed a significant benefit when compared with Doce+ADT (HR=0.79, 95% CI=0.67-0.94). A P-score of 86% was recorded when compared with Doce+ADT. Other modalities, including APA+ADT and ENZA+ADT, had no significant OS benefit compared to Doce+ADT.

 

In the explorative analysis against Doce+ADT, ENZA+ADT (HR=0.59, 95% CI=0.49-0.72) and APA+ADT (HR=0.72, 95% CI=0.57-0.92) had significant benefit in PFS. Although all four treatment options showed superiority over ADT alone (HR=1.50, 95% CI=1.35-1.67), AAP+ADT did not show a significant difference in improving the PFS of mCSPC patients, as compared to Doce+ADT (HR=0.87, 95% CI=0.74-1.02).

 

Similarly, the results of exploratory analysis were consistent in fixed- and random-effects models.

 

With updated evidence in 2020, the primary and secondary endpoints were reached and revealed that the addition of AAP, ENZA, APA, or Doce to ADT prolonged the OS and PFS of mCSPC patients. Among the four treatment modalities, AAP+ADT provided the best effect in prolonging OS. This result differs from that of another network meta-analysis published in early 2020, which suggested that ENZA+ADT is the most effective in prolonging OS.[22] This discrepancy may be because the ARCHES trial, which compared ENZA+ADT versus ADT alone in mCSPC (median OS HR = 0.81, 95% CI = 0.53-1.25, p = 0.3361) was not included in their analysis. Furthermore, the most updated OS data in 2020 from the STAMPEDE trial for AAP+ADT versus ADT alone is included in our analysis.[15] One must also notice that some RCTs had a rather short median follow-up and the OS can be considered premature; for instance, the median followups for ARCHES and TITAN were 14.4 months and 22.7 months, respectively.[16] [17]

 

In contrast to the OS results, our network meta-analysis showed that ENZA+ADT is the most favourable in terms of prolonging PFS. Despite the differences in PFS definition in each trial — be it radiographic, biochemical, or clinical—the PFS HRs were rather consistent across the trials concerning the same drug. We therefore postulate that the prevention of all forms of progression may be best achieved by ENZA+ADT as compared to other treatment modalities.

 

With the difference in OS and PFS results in mind, it is however important to note that there is no established evidence towards PFS being a surrogate for OS, while the latter is considered a golden endpoint. Moreover, OS is also affected by subsequent salvage treatment beyond progression and therefore the sequence of drugs may contribute to the difference in PFS and OS results. Cross-resistance exists between androgen receptor-targeted therapies, for instance, AAP+ADT and ENZA+ADT.[23] The sequence of agent initiation may affect the OS. There is some evidence suggesting that the use of AAP+ADT before ENZA+ADT gives rise to better outcomes. A phase 2 randomised crossover trial conducted by Khalaf et al24 has compared the prostate-specific antigen (PSA) decline >50% (PSA50) on second-line therapy and the median time to second PSA progression for AAP+ADT followed by ENZA+ADT (arm A) versus ENZA+ADT followed by AAP+ADT (arm B). It was shown that the sequence of AAP+ADT followed by ENZA+ADT (arm A) achieved superior PSA50 and time to second PSA progression (HR=0.75, 95% CI=0.53-1.06). Although this trial was designed for metastatic castration-resistant PC, it is still a direct comparison of the biochemical effectiveness between the two sequences on PC. We believe this might explain the different results in PFS and OS.

 

In addition, the follow-up period of ARCHES (14.4 mo) and ENZAMET (34 mo) for ENZA+ADT were relatively short compared with that of LATITUDE (51.8 mo) and STAMPEDE (73.2 mo) for AAP+ADT. It remains unclear whether extended follow-up of studies of ENZA+ADT would lead to a better OS outcome. Thus, further analysis may be required in the future.

 

The above results on OS, PFS and time to PSA progression do not provide a clear answer to which agent is more preferable. Several other factors including the potential toxicity profile and quality of life (QOL) should be taken into account. Apart from the general adverse effects shared among androgen receptor targeting agents, there are some adverse events unique to each specific agent that may influence the treatment choice. For example, ENZA is commonly associated with fatigue (24.1%) and musculoskeletal events (26.4%) including falls and fractures[17]; abiraterone acetate exerts excessive mineralocorticoid activity; thus, it must be given together with prednisolone, leading to some steroid-specific adverse events which are not common in other agents[13]; APA causes particularly high incidence of rash (27.1%) as compared to ADT alone (8.5%).[16] Patients’ QOL with different agents is a considerable factor. A QOL sub-study extended from the STAMPEDE trial suggested a significantly higher global QOL score in the first 2 years of treatment with AAP+ADT as compared with Doce+ADT.[25] The health-related QOL, Patient Health Questionnaire-9 depression score, and Montreal Cognitive Assessment score of AAP+ADT and ENZA+ADT were assessed in another trial.[26] Minimal difference was seen in terms of Montreal Cognitive Assessment score; while AAP+ADT was shown to be associated with better health-related QOL and Patient Health Questionnaire-9 scores over time, with greater significance in the elderly group.[26]

 

The choice of treatment should always include careful discussion and shared decision making among physicians, the patient, and caregivers. It is observed that AAP+ADT is superior to Doce+ADT, ENZA+ADT, and APA+ADT in terms of OS, QOL, and being the first agent initiated in mCSPC patients. However, the abovementioned evidence on OS, PFS, sequence of treatments, adverse events, and QOL may only serve as a guide; no single factor should be used to pursue one treatment over another.

 

Our results are in line with a recently published study by Wang et al[27] suggesting that AAP+ADT treatment were more effective in prolonging OS, while ENZA+ADT might provide better PFS in mCSPC patients. The strength of our study is that we have included the most up-to-date data from all these major trials, including the STAMPEDE with AAP+ADT as the experimental regimen.[15] Also, the comprehensive inclusion of all the most updated studies in this network meta-analysis covering more than 7500 patients provides a clearer comparison of survival data among different treatment modalities, in addition to our previously published preliminary results.[28]

 

However, several limitations exist in our comparison and the findings should be interpreted within this context. Similar to the other network meta-analyses in the field, this analysis only provided indirect comparisons and rankings among the named systematic treatments by comparing the efficacy in the RCTs, which cannot replace head-to-head clinical trials for comparison. As shown in Table 1, the study population and inclusion criteria varied among the different RCTs. The variations among different studies may contribute to heterogeneity in treatment effect, which limits our power in comparing the efficacy among the systematic treatments. Furthermore, our study did not account for other outcomes, such as QOL and cost-effectiveness. Given that the relevant data in either prospective or retrospective clinical trials are not adequate among our study populations, current clinical trials provide limited insights on the effects on QOL and cost-effectiveness among different patients.

 

Another possible weakness of our study is that we have not performed subgroup analysis according to low/high risk (according to LATITUDE study) and low/high volume (according to CHAARTED study). However, recent studies have disproven the importance of risk and volume classification in selecting treatment regimens.[29] Also, our analysis only included drug regimens and did not include radiotherapy to the primary site as a comparison arm.

 

We found that AAP+ADT is the most effective first-line treatment for mCSPC in terms of OS, while ENZA+AA may provide better PFS. Clinicians should take these findings into consideration when planning treatment of mCSPC.