Reassessment of the Efficacy of Carboplatin for Metastatic Urothelial Carcinoma in the Era of Immunotherapy: A Systematic Review and Meta-analysis

Keiichiro Mori , Victor M. Schuettfort , Takafumi Yanagisawa , Satoshi Katayama , Benjamin Pradere , Ekaterina Laukhtina , Pawel Rajwa , Hadi Mostafaei , Reza Sari Motlagh , Fahad Quhal , Marco Moschini , Francesco Soria , Jeremy Y.C. Teoh , David D’Andrea , Mohammad Abufaraj , Simone Albisinni , Wojciech Krajewski , Shin Egawa , Pierre I. Karakiewicz , Michael Rink , Shahrokh F. Shariat *, European Association of Urology– Young Academic Urologists Urothelial Carcinoma Working Group (EAU-YAU)


Introduction
Survival outcomes for patients with advanced or metastatic urothelial carcinoma (AMUC) are extremely poor with the overall 5-yr survival rates of only approximately 5% [1,2]. Moreover, given that only approximately 50% of all AMUC patients receive any oncological treatment and only approximately 15-20% of all patients receive a second therapy line, the choice of an optimal first-line therapy is of utmost importance [3]. Platinum-based regimens are recommended as a first-line treatment for patients with previously untreated AMUC and deemed fit for chemotherapy [2]. Despite being the standard of care for decades, the available platinum agents remain insufficiently compared regarding their efficacy and safety. The only phase 3 trial comparing cisplatin and carboplatin in the literature was unable to detect any significant difference in survival benefits; however, the study was underpowered, failing to reach its accrual goal [4]. To date, small, single-center, phase 2 studies have demonstrated the superiority of cisplatin over carboplatin [5][6][7][8]. Contrarily, a phase 2 trial by Dogliotti et al [9] comparing the efficacy of gemcitabine/cisplatin with that of gemcitabine/carboplatin in AMUC reported no clinically significant difference in objective response rates (ORRs; 65.9% and 56.4%, respectively), median survival (12.8 and 9.8 mo, respectively), and/or time to disease progression (8.3 and 7.7 months, respectively), despite not being designed with sufficient power to detect significant differences between the study arms in terms of efficacy. Therefore, a thorough assessment of carboplatin versus cisplatin with respect to efficacy appears to be necessary, particularly given that approximately one-third of all patients deemed eligible for cisplatin actually receive carboplatin [10].
Recently, the DANUBE, IMvigor 130, and KEYNOTE 361 trials investigated the efficacy of immune checkpoint inhibitors (ICIs) and/or chemotherapy in the first-line setting for AMUC [10][11][12]. Interestingly, data from all three studies suggest that carboplatin might not be inferior to cisplatin and that carboplatin-based chemotherapy is more effective in contemporary series than in historic series [10][11][12][13][14]. In fact, based on the results from the Keynote-361 trial, the Food and Drug Administration has revised the indication for pembrolizumab, which was previously indicated as a treatment for all patients ineligible for cisplatincontaining chemotherapy [12]. Pembrolizumab is now only approved as a treatment of patients who are not eligible for any platinum-containing chemotherapy, thus emphasizing the importance of carboplatin in this treatment setting. Therefore, the aim of the current study was to re-evaluate the efficacy of carboplatin as a first-line treatment for AMUC using the recently reported data.

Evidence acquisition
The protocol has been registered in the International Prospective Register of Systematic Reviews database (PROS-PERO: CRD42021272996).

Search strategy
A systematic review, a meta-analysis (MA), and a network meta-analysis (NMA) were conducted on phase 3 randomized controlled trials (RCTs) in AMUC patients treated with first-line ICIs or chemotherapy according to the Preferred Reporting Items for Systematic Reviews and Metaanalyses (PRISMA) statement [15]. A completed PRISMA 2009 checklist was used to describe the methodology of our study (c). PubMed, Web of Science, and Scopus were searched to identify reports published up to August 2021 that investigated first-line systemic therapy for AMUC.
The following keywords were used in our search strategy: (urothelial carcinoma OR bladder cancer OR bladder carcinoma OR urothelial cancer) AND (metastatic OR advanced) AND (randomized). Furthermore, we reviewed relevant abstracts presented at major conferences, such as the American Society of Clinical Oncology and the European Society for Medical Oncology. The primary outcome of interest was overall survival (OS) and complete response (CR), and the secondary outcome was ORR. Initial screening was per-formed independently by two investigators based on the titles and abstracts of the articles to identify ineligible reports. The reasons for exclusions were noted. Potentially relevant reports were subjected to full-text reviews, and the relevance of the reports was confirmed after the data extraction process. Disagreements were resolved via consensus with a separate committee of investigators.

Inclusion and exclusion criteria
Studies were included if they investigated AMUC patients (Patients) who had undergone carboplatin-or cisplatinbased chemotherapy (Intervention) compared with those treated with immunotherapy (Comparison) as a first-line treatment to assess their differential effects on OS, CR, and ORR (Outcome) in phase 3 randomized studies only. We also included RCTs comparing carboplatin-and cisplatinbased chemotherapies. We excluded observational studies, reviews, letters, editorials, replies from authors, case reports, and articles not published in English. Moreover, we excluded phase 2 trials. The references of all papers were scanned for additional studies of interest.

Data extraction
Two investigators independently extracted the following information from the included articles: study name, publication year, number of patients, treatment compound, age, sex, performance status (PS), primary tumor site, disease status, programmed death ligand-1 (PD-L1) status, cisplatin eligibility, subsequent therapy, oncological outcomes, and follow-up. Subsequently, the hazard ratios (HRs) and 95% confidence intervals (CIs) associated with OS were retrieved. All discrepancies regarding data extraction were resolved by consensus with the committee of investigators.

Risk of bias assessment
The ''risk of bias'' (RoB) evaluation of each study was performed using the Cochrane Collaboration's tool for assessing RoB [16]. This tool assesses selection (random sequence generation and allocation concealment), performance, detection, attrition, reporting, and other sources of bias ( Supplementary Fig. 1). The RoB of each study was assessed independently by two authors. Disagreements were resolved by consultation with the coauthors.

Meta-analysis
ORR was defined as the proportion of enrolled and randomly assigned patients who achieved the best response of CR or partial response based on investigator assessment. First, forest plots were used to assess the HRs and to describe the relationships between treatment and survival outcomes (ICI therapy vs carboplatin-based chemotherapy and ICI therapy vs cisplatin-based chemotherapy). Second, forest plots were used to summarize the variables for dichotomous outcomes and to describe the relationships between treatment and CR/ORR (ICI therapy vs carboplatin-based chemotherapy and ICI therapy vs cisplatin-based chemotherapy). Dichotomous variables were presented as proportions and compared using odds ratios (ORs) and 95% CIs. The outcomes of the studies included in this MA were evaluated for heterogeneity using Cochrane's Q test and I 2 statistics. Significant heterogeneity was indicated by p 0.05 in Cochrane's Q tests and a ratio of 50% in I 2 statistics. We used fixed-effect models to calculate nonheterogeneous results. Random-effect models were used in cases of heterogeneity [17][18][19].

Network meta-analysis
An NMA was conducted with random-and fixed-effect models using a frequentist approach for the direct and indirect comparisons of the treatments evaluated, with immunotherapy as the common comparator arm (carboplatin-vs cisplatin-based chemotherapy) [20,21]. In OS assessment, contrast-based analyses were applied with estimated differences in the log HR and the standard error calculated from the published HR and CI [22]. The relative treatment effects were presented as HR and 95% credible interval (CrI) [20]. In the assessment of CR/ORRs, arm-based analyses were performed to estimate ORs and 95% CrIs from raw data presented in the selected manuscripts [20]. Network plots were utilized to illustrate the connectivity of the treatment networks in terms of OS/CR/ORR. All statistical analyses were performed using R 3.6.3 and Review manager 5.3; statistical significance was set at p < 0.05.

Study selection and characteristics
Our initial search identified 2582 publications, and after the elimination of duplicates, a total of 1872 publications remained. A further 1840 articles were excluded after screening the titles and abstracts, and full-text reviews were performed for the remaining 32 articles (Supplementary Fig. 2). In accordance with the selection criteria, four articles comprising 3340 patients were identified for inclusion. Three studies, published between 2020 and 2021, comprised an assessment of first-line therapy and compared ICI therapy with chemotherapy (including carboplatin and cisplatin) [10][11][12][13][14]. The data extracted from these three studies are outlined in The forest plot in Figure 1A showed that cisplatin-based chemotherapy was not significantly different from ICI    Figure 1B showed that cisplatin-based chemotherapy was not significantly different from ICI monotherapy in terms of CR benefits (pooled HR, 1.16; 95% CI, 0.70-1.92; p = 0.57). The Cochrane's Q test (p = 0.20) and I 2 test (I 2 = 39%) revealed no significant heterogeneity. The forest plot in Figure 1C indicated that cisplatin-based chemotherapy was associated with a significantly better ORR than ICI monotherapy (pooled OR, 0.54; 95% CI, 0.40-0.74; p < 0.001). The Cochrane's Q test (p = 0.93) and I 2 test (I 2 = 0%) revealed no significant heterogeneity.

ICI therapy versus carboplatin-based chemotherapy
The forest plot in Figure 2A showed that carboplatin-based chemotherapy was not significantly different from ICI monotherapy in terms of OS benefits (pooled HR, 0.90; 95% CI, 0.78-1.04; p = 0.16). The Cochrane's Q test (p = 0.70) and I 2 test (I 2 = 0%) revealed no significant heterogeneity. The forest plot in Figure 2B showed that cisplatin-based chemotherapy was not significantly different from ICI monotherapy in terms of CR benefits (pooled HR, 0.89; 95% CI, 0.52-1.53; p = 0.67). The Cochrane's Q test (p = 0.84) and I 2 test (I 2 = 0%) revealed no significant heterogeneity. The forest plot in Figure 2C indicated that carboplatinbased chemotherapy was associated with a significantly better ORR than ICI monotherapy (pooled OR, 0.58; 95% CI, 0.42-0.80; p < 0.001). The Cochrane's Q test (p = 0.55) and I 2 test (I 2 = 0%) revealed no significant heterogeneity.

Network meta-analysis
An NMA of the three treatments was performed with regard to OS, CR, and ORR. The networks of eligible comparisons were graphically represented in network plots in terms of OS ( Supplementary Fig. 3A) and CR/ORR ( Supplementary  Fig. 3B). We compared cisplatin-and carboplatin-based chemotherapies with ICI therapy as a common comparator arm.

Discussion
We conducted a systematic review and MA to assess the efficacy of carboplatin and cisplatin as first-line therapies in patients with AMUC. We also performed an NMA to indirectly compare carboplatin and cisplatin; this approach revealed several findings of interest.  [2]. In patients with AMUC, guideline recommendations are tailored according to eligibility for cisplatinbased treatment [2]. Cisplatin-based chemotherapy is preferred for patients who have adequate renal function, good PS, and absence of comorbidities [2,23]. However, approximately 50% of patients are unfit to receive cisplatincontaining regimens, and treatment options for these patients include carboplatin-based regimens [2,24]. Of the standard regimens available, gemcitabine plus cisplatin have attained improved OS compared with gemcitabine plus carboplatin, although significant statistical differences were not recorded [9]. Moreover, patients unfit for cisplatin typically have prognostic factors such as poor PS that are associated with poor survival outcomes [2]. Thus, while several phase 2 trials of carboplatin versus cisplatin combination chemotherapy have shown a lower CR rate and shorter OS for the carboplatin arms, it remains unclear whether carboplatin-based chemotherapy may offer a prognosis comparable with that of cisplatin-based chemotherapy [25].
Carboplatin shares a common mechanism of action with cisplatin but exhibits different pharmacokinetics [26]. One mechanism of action common to all platinum agents is that following cellular uptake, these agents bind covalently to DNA nucleobases to form a variety of DNA adducts, and induce apoptosis through the inhibition of tumor cell apoptosis and other mechanisms [27]. However, cisplatin and carboplatin differ in terms of the extent of DNA adduct formation, which has been hypothesized to account for differences in their efficacy [27]. Further, cisplatin exhibits greater mutagenicity than carboplatin and is more likely to damage the DNA [28]. Carboplatin is less likely to cause renal damage due to its structural configuration; the structure is unlikely to form a substrate for organic cation transporter 2, the transporter involved in cisplatin uptake, thus making its uptake by proximal tubular cells unlikely [29]. ICIs have antitumor activity in urothelial carcinoma (UC) and a more favorable safety profile than chemotherapy; however, trials of first-line ICI monotherapy have not yet shown improved OS when compared with chemotherapy alone [30]. In our current MA, it was shown that cisplatinand carboplatin-based chemotherapies elicit a greater ORR than and offer similar OS/CR benefits to ICI monotherapy. Switch-maintenance therapy could potentially serve as a novel treatment strategy intended to enhance antitumor activity in UC through the use of agents with different mechanisms of action [31]. Maintenance treatment can target tumor cell populations surviving after first-line chemotherapy, thus increasing the depth of responses and/or prolonging treatment effects, while avoiding cumulative toxicity, potential cross-resistance, and increased treatment cost [32]. It is well known that chemotherapy exerts not only direct cytotoxic effects on tumor cells, but also induces antitumor immune responses by promoting the release and presentation of tumor antigens, and by reducing immunoinhibitory cells [33]. Several chemotherapeutic agents and platinum-based combinations induce immunogenic cell death, stimulating immune responses against tumors through the release of signals from dying cells [34,35]. Additionally, chemotherapy may also upregulate the expression of PD-L1, a key immune checkpoint molecule [36][37][38]. The expression of damage-associated molecular patterns, including ATP and HMGB1, using non- Pooled HR (95% credible interval) was derived from network metaanalysis. Bold indicates statistically significant comparison.  small cell lung cancer (NSCLC) cells, has been recorded with cisplatin and carboplatin to a certain extent [39,40]. Furthermore, it has been suggested that both cisplatin and carboplatin have a role in promoting antitumor immune responses by reducing the number of myeloid-derived suppressor cells [41,42]. Of particular note, it is also suggested that cisplatin not only enhances T-cell activity, but also induces tumor cell PD-L1 upregulation, thus possibly accounting in part for the additive antitumor activity between cisplatin-based chemotherapy and PD-L1/ programmed death protein-1 (PD-1) inhibition [37,[43][44][45]. Indeed, the expression of PD-L1 in NSCLC cells is shown to be upregulated following preoperative cisplatin-based chemotherapy in NSCLC patients (before vs after chemotherapy, 11% vs 26%; p = 0.017) [36]. Overall, there is a clear rationale for exploring ICIs as a first-line maintenance therapy for AMUC, given the immunogenic nature of UC, antitumor activity and favorable safety profile of ICIs, and cytotoxic and immunogenic effects of chemotherapy [31]. In the JAVELIN Bladder 100 phase 3 trial, avelumab as first-line maintenance therapy led to significant prolongation of OS, compared with the best supportive care (BSC), in patients with AMUC not experiencing disease progression on first-line platinum-containing chemotherapy [46]. While the trial design permitted the inclusion of patients who had received first-line combination chemotherapy with cisplatin plus gemcitabine or carboplatin plus gemcitabine [46], carboplatin-treated patients tended to be less fit than cisplatin-treated patients, as reflected by a higher proportion of Eastern Cooperative Oncology Group PS 1 (49% vs 32%), median age (71 vs 66 yr), and rate of renal impairment (63% vs 36%) [47]. The improvement in OS with avelumab versus BSC was similar irrespective of the first-line chemotherapy, with the HRs being 0.69 (95% CI, 0.51-0.94) and 0.66 (95% CI, 0.47-0.91) in the cisplatin plus gemcitabine and carboplatin plus gemcitabine subgroups, respectively, and the median postchemotherapy OS being 25.3 and 19.9 mo, respectively, with avelumab maintenance therapy [46,47]. Similarly, in our current MA/NMA, we found that while cisplatin was slightly more efficacious than carboplatin, the two agents were not significantly different in terms of efficacy. Further, the JAVELIN Bladder 100 phase 3 trial demonstrated that those achieving CR with first-line chemotherapy had an HR for OS of 0.81 (0.47-1.38) on avelumab and BSC, with their median OS remaining unreached and faring better than those achieving partial response with first-line chemotherapy [46]. Moreover, another study demonstrated that those achieving CR with preoperative chemotherapy had significantly prolonged cancer-specific and recurrence-free survival [48]. While these data point to the importance of achieving CR with chemotherapy, this study demonstrated no difference in CR rate with cisplatin-based versus carboplatin-based chemotherapy. However, it must be considered that while the DANUBE trial did not comprise patients with PS 2, the KEYNOTE361 and IMvigor130 trials included several patients with poor PS in their carboplatin arms [10][11][12][13][14]. Carboplatin was shown to be comparable with cisplatin with respect to efficacy in our study, and there is a possibility that the efficacy of carboplatin might have been underestimated owing to its use in patients with worse PS, which is known to be associated with worse survival; therefore, carboplatin needs to be reassessed for its efficacy in a similar patient cohort.
Despite being comprehensive in nature, this systematic review has some limitations. First, the patient characteristics differed at the time of study enrollment among the DANUBE, IMvigor130, and KEYNOTE361 trials, despite having similar study designs, treatment lines, and target diseases. Indeed, a much larger proportion (48%) of patients undergoing chemotherapy received subsequent ICI therapy in the KEYNOTE361 trial than in the other trials, likely contributing to the favorable OS outcomes in its chemotherapy arm as well as to the underestimation of the efficacy of pembrolizumab in patients receiving ICI therapy [12]. Moreover, it must be noted that, despite being uniformly categorized as immunotherapy, ICIs included both PD-1 and PD-L1 inhibitors, which differ in their mechanisms of action and possibly efficacy [49]. Second, while the NMA involved an indirect approach to compare outcomes from the RCTs, this approach falls short of a head-to-head comparison. Moreover, the only available phase 3 direct comparison between cisplatin and carboplatin was underpowered. Thus, the findings reported herein need to be validated in well-designed comparative trials. Third, given the lack of data for OS and ORR with cisplatin-based chemotherapy in the KEYNOTE361 trial, this MA evaluated the available data on all chemotherapeutic regimens from the RCTs, including carboplatin-based chemotherapy. Moreover, it must be noted that despite being categorized uniformly as cisplatin-based chemotherapy, this included both patients who were cisplatin eligible and those treated with cisplatin, which differed strictly. Fourth, while the method of Guyot et al [50] represents a better method for analysis of survival over time, it was not available for use in this study, given the paucity of survival curve data with cisplatin versus carboplatin from all the RCTs included. Finally, the OS data from the IMvigor130 trial remained immature at the time of this review, and the study outcomes might vary considerably in their final analyses. Furthermore, as the CheckMate 901 and NILE trials are still underway, the value of carboplatin-based chemotherapy in patients with AMUC could vary depending on the results of these trials.

Conclusions
Our analyses suggest that, in AMUC patients, there is no OS/CR difference between cisplatin-and carboplatinbased chemotherapy compared with ICI monotherapy; however, both chemotherapies offer a more favorable ORR than ICI monotherapy. Moreover, our MA/NMA reveals that there is no difference in OS, CR, and ORR between cisplatin-and carboplatin-based chemotherapy. This suggests a need for a reappraisal of the efficacy and role of carboplatin in the era of ICIs. Carboplatinbased chemotherapy seems to be more effective in contemporary series than in historical controls; moreover, it offers additive effects to ICI therapy, is associated with fewer adverse effects than cisplatin, and is preferentially used for patients with poor PS. These findings might be of value in determining personalized treatment strategies for AMUC patients.