Reappraisal of Cisplatin vs Carboplatin Treatment for Solid Tumors Authors: Pablo Jiménez-Labaig, Oriol Mirallas, Filipa Pereira, Judit Sanz, Enriqueta Felip, Jordi Rodon, Nicoleta Colombo, Matthew D. Galsky, Kevin J. Harrington Published: 2026 ABSTRACT Key PointsQuestionWhat is the certainty of evidence supporting cisplatin vs carboplatin treatment of solid tumors?FindingsIn this umbrella review of 11 meta-analyses, high-certainty evidence was seen only for no difference between agents in ovarian carcinoma and for radiologic response favoring cisplatin in urothelial carcinoma, whereas moderate-certainty evidence was mixed in cervical cancer, germ cell tumors, and non–small cell lung cancer. Cisplatin had greater nonhematologic toxicity and carboplatin had greater hematologic toxicity; for most other tumors and end points, certainty was low or very low.MeaningThese findings suggest that current guideline recommendations for the use of cisplatin or carboplatin, based on meta-analyses, reflect varying certainties of evidence, ranging from high to very low. Introduction Platinum-based chemotherapies heralded a new age of medical oncology following their serendipitous discovery as anticancer agents by Barnett Rosenberg, PhD, in 1965.1 Nowadays, even in the current era of widespread use of immune checkpoint inhibitors across solid tumor oncology, platinum salts continue to serve as a crucial chemotherapy backbone in both the curative and palliative settings for several different tumor types,2 including small cell and non–small cell lung cancers (SCLC and NSCLC) and urothelial, gynecologic, gastrointestinal tract, triple-negative breast, and head and neck cancers, among others.3,4 Among platinum compounds, there has been a long-standing debate about their distinctive relative efficacy and safety profiles.5 Although the debate was resolved for some tumor types, such as colorectal adenocarcinoma, where oxaliplatin showed broad superiority over other platinum drugs,6,7 the controversy frequently continues in clinical practice, specifically between cisplatin and carboplatin.4 Both agents form platinum-DNA adducts that, if unrepaired, trigger apoptosis, but they differ in toxicity and dose adjustment: cisplatin is commonly limited by nephrotoxicity and ototoxicity, whereas carboplatin is generally simpler to administer, is commonly dosed to a target area under the curve, and is associated with better quality of life during treatment while tending to cause more myelosuppression, particularly thrombocytopenia.8,9 Generally, cisplatin is perceived as an agent with overall survival (OS) benefit in distinct tumor types and treatment settings, especially as a radiosensitizer in concomitant regimens in radical or postoperative circumstances.10,11,12,13 However, cisplatin’s potential for permanent toxic effects, along with population aging and an increasing comorbidity burden in patients considered for curative-intent therapy, as well as periodic drug shortages, complicates its use and renews interest in whether carboplatin can be considered noninferior in selected contexts.14 Numerous initiatives have been undertaken to address this question, including Cochrane reviews, the CISCA (Cisplatin vs Carboplatin) and COCIS (Carboplatin or Cisplatin) Meta-analysis groups, and distinct programs from the National Comprehensive Cancer Network (NCCN), European Organisation for Research and Treatment of Cancer, and Japan Clinical Oncology Group organizations, among others.15,16,17,18 However, this debate is still relatively poorly addressed in the literature, without a global meta-analysis that grades the robustness of the evidence toward superiority or noninferiority across all tumor types and treatment settings. We aimed to quantitatively synthesize all available meta-analytical data to assess the certainty of evidence regarding efficacy and safety differences between cisplatin- and carboplatin-based therapies per tumor type. Methods This umbrella meta-analysis was conducted in accordance with the Preferred Reporting Items for Overviews of Reviews (PRIOR) reporting guidelines. The protocol for this study was registered in PROSPERO (CRD420251160357). None of the authors’ institutions required ethical committee approval for this meta-analysis. Search Strategy and Study Selection Criteria Four independent researchers conducted a systematic review of the literature (P.J.L., O.M., F.P., and J.S.). The search utilized the Web of Science database, including its Core Collection, from database inception through January 11, 2026. Inclusion criteria were (1) original meta-analyses, including individual studies comparing cisplatin with carboplatin, either as monotherapy or in combination regimens, and (2) reviews that assessed efficacy and/or safety outcomes. Exclusion criteria, search strategy, data extraction, and synthesis are reported in eTable 1 in Supplement 1. Overlap was assessed at the primary-study level; duplicates were included once, and discrepancies were resolved using the most complete original study data. Individual studies were selected from the eligible meta-analyses and reassessed at the individual-study level to confirm that they met the eligibility criteria, to ensure data accuracy, and to obtain granular data. Eligible studies were pooled for the umbrella analyses, stratified by tumor type, treatment setting, efficacy (measured by overall response rate [ORR], complete response rate [CR], overall survival [OS], progression-free survival [PFS], disease-free survival [DFS], mortality and treatment failure ratios, and 1-, 3-, and 5-year survival rates), and treatment-related adverse events (grade ≥3 cytopenias, emesis, nephrotoxicity, peripheral neurotoxicity, and ototoxicity), alongside their effect size measures (odds ratio, risk ratio, and hazard ratio). The methodologic quality of the included meta-analyses was assessed using the AMSTAR and AMSTAR-2 tools.19 Statistical Analysis Meta-analytical calculations were performed using the metaumbrella package in R, version 2024.12.1 (R Project for Statistical Computing).20 Random-effects models were fitted using the Hartung-Knapp-Sidik-Jonkman method, which provides conservative confidence interval coverage when the number of studies is fewer than 10.21 All results were presented as equivalent odds ratios (eORs), a standardized metric that allows different effect size measures (odds ratios, risk ratios, and hazard ratios) to be converted into a common odds ratio scale, thereby facilitating comparability and synthesis across outcomes. eORs were reported as relative effect sizes, and not as absolute risks or survival probabilities. Values greater than 1 indicated higher outcome rates with cisplatin and values less than 1 indicated higher rates with carboplatin. Certainty of evidence was assessed using an algorithmic GRADE (Grading of Recommendations, Assessment, Development and Evaluation) approach (eFigure 1 in Supplement 1). Each comparison was initially rated as high certainty and downgraded for risk of bias, inconsistency, publication bias, and imprecision. Final ratings were classified as high, moderate, low, or very low according to the number and severity of downgrades. For end points including different study designs, certainty was assigned to the pooled body of evidence using the same rules. Equivalent effect sizes and GRADE procedures followed the method of Gosling et al.20 Overall risk of bias was classified as low if less than 50% of studies were at high risk, high if at least 50% were at high risk, and unclear if no quality assessment was reported. The impact of study design (randomized clinical trials [RCTs] vs observational studies), response criteria (RECIST [Response Evaluation Criteria in Solid Tumors] vs non-RECIST), and treatment frequency (equal vs unequal across arms) on efficacy outcomes was prespecified for examination through sensitivity analyses, restricted to factors with high or moderate GRADE certainty. Statistical significance was set at a 2-sided P < .05. Lack of statistical significance did not establish equivalence or noninferiority unless the evidence was designed and powered for that purpose. Results From 2380 records identified through the systematic review, 225 full texts were assessed for eligibility. Eleven meta-analyses were selected,15,16,22,23,24,25,26,27,28,29,30 involving 52 original studies and 12 683 patients in total (Table 1 and eTable 2 in Supplement 1).31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83 The PRISMA diagram is presented in eFigure 2 in Supplement 1. Several efficacy and safety outcomes were identified across 7 tumor types (gonadal germ cell tumors [GCT], head and neck squamous cell carcinoma [HNSCC], urothelial, cervical, and ovarian carcinomas, NSCLC, and SCLC). Most cancer types were investigated in advanced situations with mainly palliative intention, and only GCT, early-stage cervical carcinoma, and HNSCC were examined with purely curative intent. Table 1. Characteristics of Meta-Analyses Comparing Cisplatin- and Carboplatin-Based Systemic Therapies Across Solid Malignant TumorsSourceTumor type MA design (study types included)Main outcomesAMSTAR/AMSTAR-2 methodologic qualityConclusionsGalsky et al,22 2012Advanced urothelial carcinomaPairwise MA (RCTs)ORR, CRModerate/lowCisplatin regimens had higher ORR and CR vs carboplatinO’Dwyer et al,23 2024Advanced urothelial carcinomaNMA (RCTs and observational studies)ORR, PFS, OS, TRAEsHigh/moderateSplit-dose cisplatin showed ORR, PFS and OS broadly comparable to alternatives in the NMA, supporting split-dose cisplatin as a feasible option in selected patients; comparable safetyXue et al,24 2018Early-stage cervical carcinoma (concomitant CRT)Pairwise MA (observational studies)CR, 3-y PFS/DFS, 3-y OS, TRAEsHigh/lowCarboplatin-RT combination tended to yield lower CR (and nonsuperior survival efficacy) vs cisplatin-RT combination; carboplatin was better toleratedMansoor et al,25 2023Advanced epithelial ovarian cancerPairwise MA (RCTs)OS, PFS, CR, PR, TRAEsHigh/lowOS and PFS were broadly similar between platinum-taxane approaches, with differences mainly in toxicity and quality of life favoring carboplatinGriesinger et al,26 2019Advanced NSCLCPairwise MA (RCTs)OS, 1-y OS, ORR, TRAEsHigh/lowOS was similar across agents; cisplatin showed a small advantage in ORR, with different severe adverse events driving regimen choiceArdizzoni et al,15 2007Advanced NSCLCIndividual patient data MA (RCTs)OS, ORR, mortality ratio, TRAEsModerate/lowModest efficacy advantages for cisplatin (ORR and survival in selected subgroups), endorsing it as reference for fit patients; different toxicity profilesDe Castria et al, 201327Advanced NSCLCPairwise MA (RCTs)OS, 1-y OS, ORR, TRAEsHigh/highOS was broadly similar for cisplatin vs carboplatin, with toxicity trade-offs guiding selectionRossi et al,16 2012Extensive-stage SCLCIndividual patient data MA (RCTs)OS, PFS, ORR, TRAEsModerate/lowOS and PFS were similar across agents; choice mainly depended on toxicity profileChen et al,28 2020Extensive-stage SCLCNMA (RCTs)OS, PFS, ORRHigh/moderateComparable OS, PFS, and ORR, with cisplatin generally more toxic and carboplatin generally better tolerated but more myelosuppressiveZhang et al,29 2025Curative-intent gonadal GCTPairwise MA (RCTs and observational studies)Treatment failure, mortality, TRAEsHigh/moderateCarboplatin showed higher treatment failure, especially in testicular primary tumors; mortality was numerically higher with carboplatin, without statistically significant difference; different toxicity profilesGuan et al,30 2016Locally advanced HNSCCPairwise MA (RCTs and observational studies)5-y OS, 3-y OS, 3-y LRC, TRAEsModerate/lowCisplatin-based therapy was associated with better OS vs carboplatin; LRC generally similar and adverse events differedAbbreviations: AMSTAR, Assessing the Methodological Quality of Systematic Reviews; CR, complete response; DFS, disease-free survival; GCT, germ cell tumor; LRC, locoregional control; NMA, network meta-analysis (MA); NSCLC, non–small-cell lung cancer; ORR, objective response rate; OS, overall survival; PFS, progression-free survival; PR, partial response; RCT, randomized clinical trial; RT, radiotherapy; SCLC, small cell lung cancer; TRAEs, treatment-related adverse events. Characteristics of Meta-Analyses Comparing Cisplatin- and Carboplatin-Based Systemic Therapies Across Solid Malignant Tumors Abbreviations: AMSTAR, Assessing the Methodological Quality of Systematic Reviews; CR, complete response; DFS, disease-free survival; GCT, germ cell tumor; LRC, locoregional control; NMA, network meta-analysis (MA); NSCLC, non–small-cell lung cancer; ORR, objective response rate; OS, overall survival; PFS, progression-free survival; PR, partial response; RCT, randomized clinical trial; RT, radiotherapy; SCLC, small cell lung cancer; TRAEs, treatment-related adverse events. Thirty-four original studies (65.3%) across 36 articles were RCTs34,35,36,37,38,39,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,69,70,71,74,76,77,78,81,82 and 17 (32.7%) were observational studies.31,32,40,41,42,43,44,45,67,71,72,75,79,80,83,84 One study33 was a nonrandomized clinical trial. Forty-three studies (82.7%) in 45 articles evaluated platinum agents in combination regimens,31,32,33,34,35,36,37,38,39,40,41,42,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,69,70,74,75,76,77,78,79,80,81,82,83 although 4 used different partner drugs across arms.33,36,37,51 Five studies in 6 articles (9.6%) were noninferiority trials47,66,69,74,80,81; the remainder were superiority trials. RECIST was the most common response criterion (14 studies in 15 articles [26.9%]),39,40,41,42,43,44,45,48,64,65,66,67,68,70,78 with the remaining studies using pre-RECIST criteria. Nine studies (17.3%) used different dosing schedules between arms.39,40,41,42,50,51,62,64,74 Radiologic Response ORR in advanced urothelial carcinoma was classified as high certainty of evidence according to the GRADE approach. The CR rate in the same tumor type and in early-stage cervical carcinoma showed moderate certainty of evidence. The remaining outcomes showed a low or very low level of certainty of evidence (Table 2 and Figure 1A). Across the meta-analyzed urothelial cancer studies, cisplatin-based regimens were associated with significantly higher ORR (eOR, 1.38 [95% CI, 1.24-1.53]; I2 = 0; P < .001) and CR (eOR, 2.37 [95% CI, 1.64-3.42]; I2 = 0; P = .001) compared with carboplatin-based regimens. Table 2. Efficacy Outcomes Stratified by Decreasing GRADE Class of Evidence and MagnitudeFactorNo. of studiesEffect size, eOR (95% CI)Features used to classify certainty of evidenceGRADE certainty of evidenceNo. of eventsa/No. of patientsP valueI2 statistic, %Egger testMeta-analytical weight, %b Radiologic response Advanced urothelial carcinoma ORR8RR, 1.38 (1.24-1.53)614/1195<.00100.8275.40HighAdvanced urothelial carcinoma CR7RR, 2.37 (1.64-3.42)139/1157.00100.1777.87ModerateEarly-stage cervical carcinoma CR3OR, 2.03 (1.27-3.24)391/520.0200.47100ModerateAdvanced NSCLC ORR15OR, 1.23 (1.06-1.43)1238/4564.00940.1321.69LowAdvanced ovarian carcinoma ORR2RR, 0.89 (0.45-1.75)212/991.270NA100LowAdvanced ovarian carcinoma CR2RR, 0.87 (0.25-3.08)110/991.400NA100LowEarly-stage cervical carcinoma ORR2OR, 1.83 (0-207 055)252/270.6369NA100Very lowExtensive SCLC ORR3RR, 1.07 (0.43-2.66)291/421.79730.3552.26Very lowExtensive SCLC CR4RR, 1.05 (1.00-1.10)53/499.0400.6144.09Very lowAdvanced NSCLC CR13OR, 0.96 (0.69-1.34)43/3160.7900.7837.72Very low OS and DFS Advanced ovarian carcinoma PFS3HR, 0.91 (0.81-1.03)1224/1615.0900.8197.52HighEarly-stage cervical carcinoma 3-y DFS4OR, 1.42 (0.92-2.17)572/824.0800.98100ModerateEarly-stage cervical carcinoma 3-y OS4OR, 1.38 (0.59-3.23)648/824.31360.63100ModerateAdvanced NSCLC 1-y OS10OR, 1.07 (0.89-1.27)1972/4008.43180.8158.71ModerateEarly-stage HNSCC 5-y OS3OR, 1.45 (0.34-6.20)197/349.38600.740Very lowExtensive SCLC OS2HR, 1.43 (0.01-185 095)270/300.5283NA80.33Very lowExtensive SCLC 1-y OS1OR, 1.36 (0.54-3.43)50/78.52NANA0Very lowEarly-stage HNSCC 3-y OS4OR, 0.98 (0.33-2.89)220/483.96680.420Very low Mortality and treatment-failure ratios Advanced NSCLC mortality ratio14RR, 0.99 (0.98-1.01)4043/4646.4000.5250.65ModerateGonadal GCT treatment failure ratio7OR, 0.55 (0.29-1.03)205/1343.06430.20100ModerateGonadal GCT mortality ratio7OR, 0.60 (0.20-1.77)93/968.28620.8993.18LowAbbreviations: CR, complete response; DFS, disease-free survival; eOR, equivalent odds ratio; GCT, germ cell tumors; NA, not applicable in cases with <3 studies; NSCLC, non–small cell lung cancer; OR, odds ratio; ORR, overall objective response rate; OS, overall survival; PFS, progression-free survival; RR, risk ratio. a Indicates across both treatment arms. b Indicates weighted percentage contributed by low-risk-of-bias studies. Efficacy Outcomes Stratified by Decreasing GRADE Class of Evidence and Magnitude Abbreviations: CR, complete response; DFS, disease-free survival; eOR, equivalent odds ratio; GCT, germ cell tumors; NA, not applicable in cases with <3 studies; NSCLC, non–small cell lung cancer; OR, odds ratio; ORR, overall objective response rate; OS, overall survival; PFS, progression-free survival; RR, risk ratio. Indicates across both treatment arms. Indicates weighted percentage contributed by low-risk-of-bias studies. Forest Plots of Efficacy Outcomes Depending on Tumor Type and Treatment Setting Sizes of squares indicate study weights. CR indicates complete response; DFS, disease-free survival; eOR, equivalent odds ratios; GCT, germ cell tumor; NA, not applicable; NSCLC, non–small cell lung cancer; ORR, overall response rate; OS, overall survival; PFS, progression-free survival; SCLC, small cell lung cancer. In early-stage cervical carcinoma, the CR rate was associated with a statistically significant improvement in favor of cisplatin (eOR, 2.03 [95% CI, 1.27-3.24]; I2 = 0; P = .02) supported by moderate certainty of evidence. Similarly, in advanced NSCLC, ORR significantly improved with cisplatin (eOR, 1.23 [95% CI, 1.06-1.43]; I2 = 4%; P = .009), although the certainty of evidence was rated as low. With a very low level of certainty, patients with SCLC had a statistically significant improvement in CR rate with cisplatin-based treatments (eOR, 1.05 [95% CI, 1.00-1.10]; I2 = 0; P = .04). The remaining ORR and CR outcomes were supported by low or very low levels of certainty and did not demonstrate statistically significant differences according to the platinum agent used. In some cases, the pooled estimates were unstable, such as ORR in early-stage cervical carcinoma, likely reflecting sparse data. Survival End Points PFS in advanced ovarian carcinoma demonstrated a high certainty of evidence, showing no statistically significant difference between cisplatin and carboplatin (eOR, 0.91 [95% CI, 0.81-1.03]; I2 = 0; P = .09). Moderate certainty of evidence was presented by the end points assessed in early-stage cervical carcinoma. Cisplatin use showed numerically higher point estimates for DFS (eOR, 1.42 [95% CI, 0.92-2.17]; I2 = 0%; P = .08) and OS (eOR, 1.38 [95% CI, 0.59-3.23]; I2 = 36%; P = .31), although these did not reach statistical significance. The 1-year OS rate in NSCLC achieved a moderate GRADE rating, without differences between agents (eOR, 1.07 [95% CI, 0.89-1.27]; I2 = 18%; P = .43). The remaining factors had very low certainty of evidence, without statistically significant associations (Table 2 and Figure 1B). Treatment Failure and Mortality Ratio The mortality ratio in NSCLC and the rate of relapse in GCT were rated as moderate certainty of evidence, while mortality in GCT was rated as low certainty of evidence (Table 2 and Figure 1C). The NSCLC mortality ratio did not demonstrate a significant difference between the use of cisplatin and carboplatin (eOR, 0.99 [95% CI, 0.98-1.01]; I2 = 0; P = .40). In contrast, in GCTs, pooled estimates showed higher odds of treatment failure (eOR, 0.55 [95% CI, 0.29-1.03]; I2 = 43%; P = .06) and mortality (eOR, 0.60 [95% CI, 0.20-1.77]; I2 = 62%; P = .28) with carboplatin compared with cisplatin, but these differences did not reach statistical significance. Sensitivity Analyses In advanced urothelial carcinoma, ORR remained significantly higher with cisplatin in all sensitivity analyses, although GRADE certainty decreased to very low in RCT-only analyses and to moderate in studies with equal treatment frequency. CR also remained significantly higher with cisplatin, with certainty decreasing from moderate to very low in RCT-only analyses and to low in studies with equal treatment frequency. In GCT, the treatment failure ratio became significant in the RCT-only (eOR, 0.32 [95% CI, 0.15-0.67]; I2 = 0; P = .03) and RECIST-based (eOR, 0.33 [95% CI, 0.25-0.44]; I2 = 0; P =.03) subgroup analyses, suggesting a higher relapse risk with carboplatin, but remained at low certainty and lost significance when restricted to studies with equal treatment frequency (Table 3). Table 3. Sensitivity Analyses for Factors That Achieved High and Moderate GRADE Level of EvidenceFactorNo. of StudiesNo. of events/No. of patientsEffect size, eOR (95% CI)P valueI2 statistic, %Egger testGRADE certainty of evidence Only RCTs Advanced urothelial carcinoma ORR5153/294RR, 1.30 (1.06-1.59).0200.45Very lowAdvanced urothelial carcinoma CR428/256RR, 3.75 (1.38-10.20).0300.10Very lowGonadal GCT treatment failure ratio2123/652OR, 0.32 (0.15-0.67).030NALow Only when RECIST 1.0 and 1.1 were used Advanced urothelial carcinoma ORR4487/943RR, 1.40 (1.18-1.66).00900.79HighAdvanced urothelial carcinoma CR3111/901RR, 2.13 (1.20-3.77).0300.84ModerateGonadal GCT treatment failure ratio4145/767OR, 0.33 (0.25-0.44).00100.04aLow Only when treatment frequency was equal across arms Advanced urothelial carcinoma ORR6441/843RR, 1.36 (1.20-1.54).00200.77ModerateAdvanced urothelial carcinoma CR6113/847RR, 2.70 (1.77-4.13).00200.17LowAdvanced NSCLC 1-y OS91944/3937OR, 1.07 (0.88-1.30).45270.86ModerateAdvanced NSCLC mortality ratio134000/4575RR, 0.99 (0.98-1.01).4200.47ModerateGonadal GCT treatment failure ratio7205/1343OR, 0.55 (0.29-1.03).06430.20ModerateAbbreviations: CR, complete response; eOR, equivalent odds ratio; GCT, germ cell tumors; NA, not applicable in cases with <3 studies; NSCLC, non–small cell lung cancer; OR, odds ratio; ORR, overall objective response rate; OS, overall survival; RCT, randomized clinical trial; RR, risk ratio. a Indicates statistically significant values. Sensitivity Analyses for Factors That Achieved High and Moderate GRADE Level of Evidence Abbreviations: CR, complete response; eOR, equivalent odds ratio; GCT, germ cell tumors; NA, not applicable in cases with <3 studies; NSCLC, non–small cell lung cancer; OR, odds ratio; ORR, overall objective response rate; OS, overall survival; RCT, randomized clinical trial; RR, risk ratio. Indicates statistically significant values. Safety Analyses In the safety analysis comparing carboplatin and cisplatin across all tumor types, the pooled estimate for peripheral neurotoxicity did not demonstrate a significant difference between the 2 agents, with moderate certainty of evidence (eOR, 1.24 [95% CI, 0.86-1.81]; I2 = 30%; P = .24). All other toxicity outcomes reached only very low certainty of evidence, with statistically significant differences (Figure 2 and eTable 3 in Supplement 1). Cisplatin was associated with higher odds of ototoxicity (eOR, 2.86 [95% CI, 1.07-7.64]; I2 = 0; P = .04), nephrotoxicity (eOR, 2.21 [95% CI, 1.63-2.98]; I2 = 0; P < .001), and nausea and vomiting (eOR, 2.17 [95% CI, 1.26-3.71]; I2 = 52%; P = .008) compared with carboplatin. Conversely, carboplatin was associated with higher odds of hematologic toxic effects, including thrombocytopenia (eOR, 0.30 [95% CI, 0.20-0.43]; I2 = 68%; P < .001), anemia (eOR, 0.62 [9 5% CI, 0.43-0.88]; I2 = 63%; P = .01), and neutropenia (eOR, 0.64 [95% CI, 0.44-0.94]; I2 = 89%; P = .03), compared with cisplatin. Safety analyses by tumor type are presented in eTable 4 in Supplement 1. Forest Plot of Odds of Adverse Events With Cisplatin Compared With Carboplatin Across Tumor Types Sizes of squares indicate study weights. eOR indicates equivalent odds ratio. Quality Assessment and Heterogeneity AMSTAR rated 7 meta-analyses as high quality23,24,25,26,27,28,29 and 4 as moderate,15,16,22,30 whereas AMSTAR-2 rated 1 as high27 and 3 as moderate23,28,29 (Table 1 and eTable 5 in Supplement 1). Heterogeneity was generally low, except moderate for ORR in extensive-stage SCLC and OS in early-stage HNSCC, and high for OS in extensive-stage SCLC. Ototoxicity and nephrotoxicity had no heterogeneity; peripheral neurotoxicity was low, whereas nausea and vomiting and hematologic toxic effects showed moderate to high heterogeneity. No publication bias was detected in the analyses (Table 2). Discussion To our knowledge, this is the first umbrella meta-analysis to compile and quantitatively synthesize the evidence across existing meta-analyses assessing efficacy and safety differences between cisplatin and carboplatin. The certainty of evidence was evaluated for each tumor type and treatment setting, providing additional support for clinical practice and guideline-level decision-making beyond the existing evidence base. The review accounted for heterogeneity across studies, including differences in partner drugs and platinum dosing schedules. Several findings merit consideration. First, high-certainty evidence was observed only for radiologic response differences in urothelial carcinoma and for no PFS difference in ovarian carcinoma. In advanced bladder cancer, cisplatin was associated with as much as 38% higher odds of achieving an objective radiologic response compared with carboplatin, as well as a higher probability of CR. Although corresponding PFS and OS data were not available for this comparison, these findings underscore the importance of assessing cisplatin eligibility upfront when selecting a platinum-based therapy, with or without immune checkpoint inhibitor treatments,85 particularly in patients with a high disease burden or significant symptoms. They also highlight the need for caution when extrapolating carboplatin substitution to the neoadjuvant and adjuvant settings.13,86,87 The umbrella review consistently aligns with the strengths of the current guideline recommendations favoring cisplatin over carboplatin for patients who are eligible for either agent in the advanced setting (European Association of Urology [EAU]: strong; European Society for Medical Oncology [ESMO]: recommendation IA; NCCN: carboplatin-based therapy is not a preferred regimen but is useful in certain circumstances, [category 1]),18,88,89 which highlight inferior survival outcomes when carboplatin is substituted. The expanding role of enfortumab vedotin in both curative-intent and palliative settings may reduce the importance of cisplatin eligibility in bladder cancer.90,91,92 However, platinum choice is likely to remain pivotal for patients with disease progression after receiving treatment or who cannot access or are not candidates for the antibody-drug conjugate. On the other hand, in advanced epithelial ovarian carcinoma, cisplatin and carboplatin showed no statistically significant differences in PFS, with a high certainty of evidence. This finding is consistent with clinical practice adopted since 2003, when Ozols et al69 demonstrated the noninferiority of carboplatin vs cisplatin, each combined with paclitaxel, in a phase 3 trial including 792 patients. In that trial, the carboplatin-based regimen was also less toxic and associated with a lower impact on quality of life.93 Overall, these findings reinforce the data from the included 2023 meta-analysis by Mansoor et al25 and align with current guideline recommendations, in which NCCN and ESMO list carboplatin-based therapy as the preferred regimen (category 1 and recommendation IA, respectively).18,94 By contrast, no eligible meta-analysis addressed this question in the context of intraperitoneal chemotherapy, where cisplatin remains the preferred agent because of favorable intraperitoneal pharmacologic properties and reported outcome advantages.95 Regarding early-stage cervical carcinoma, our effect estimates generally favored cisplatin, mostly with moderate certainty, although statistically significant differences were observed only for CR. In the published literature on cervical carcinoma, this uncertainty has been most directly examined in the advanced setting in the Japanese phase 3 JCOG0505 trial,17 where carboplatin plus paclitaxel were noninferior to cisplatin plus paclitaxel for OS, with similar response rates overall, although outcomes appeared less favorable with carboplatin among patients who had not previously received cisplatin. Similarly, in the BEATcc study, subgroup analyses did not show clear differences between platinum agents.96 Conversely, in the early-stage setting, investigational efforts have largely focused on assessing weekly cisplatin as the standard radiosensitizing chemotherapy, consistent with current guideline recommendations from NCCN (preferred regimen, category 1),18 ESMO (recommendation IA),97 and the European Society of Gynecological Oncology, European Society for Radiotherapy and Oncology, and European Society of Pathology, with carboplatin reserved for patients who are ineligible for cisplatin.98 This question was not directly addressed in the current standard-of-care KEYNOTE-A18 regimen99; instead, available evidence for carboplatin with radiotherapy comes from a single-arm phase 1 study100 and a small number of comparative observational cohorts reporting potentially acceptable efficacy and tolerability.43,44,99 For advanced NSCLC, no outcomes with high certainty of evidence were identified. Although this tumor type contributed the largest sample in terms of included studies and patients, the certainty of evidence was downgraded mainly due to a high risk of bias in 8 of the 15 included trials, generally due to inadequate blinding and dose imbalances between treatment arms. Nevertheless, the data suggest no significant differences between agents in OS and mortality ratio, supported by moderate-certainty evidence with relatively narrow confidence intervals. ORR was statistically significantly increased with cisplatin compared with carboplatin, albeit with low certainty of evidence. According to current guideline recommendations from the NCCN,18 American Society of Clinical Oncology,101 and ESMO,102 both carboplatin and cisplatin are generally accepted across NSCLC subtypes in patients eligible for either agent, although carboplatin is acknowledged to have better tolerability and feasibility. These recommendations are consistent with the 2015 Cochrane review,27 which found no statistically significant differences in OS between cisplatin- and carboplatin-based regimens. This appraisal, encompassed in the present umbrella review, includes the phase 3 ECOG 1594 trial, which reported comparable survival and response outcomes across platinum agents in 1207 patients.56 Extensive-stage SCLC showed no statistically significant difference between agents overall, but the certainty of evidence was very low for all evaluated outcomes due to few available studies, wide confidence intervals, and substantial heterogeneity. Only CR in extensive-stage SCLC was found to be significantly more likely with cisplatin. Current ESMO103 and NCCN18 guidelines list cisplatin as the preferred platinum agent in curative-intent settings, while emphasizing carboplatin as an alternative for cisplatin-ineligible patients and as an equivalent option in extensive-stage disease. This is also reinforced by 3677 elderly patients with extensive-stage SCLC in the Surveillance, Epidemiology, and End Results registry for whom similar survival data and fewer hospitalizations were reported with carboplatin use.104 Likewise, in early-stage HNSCC, the data did not demonstrate an efficacy difference between cisplatin and carboplatin; however, the certainty of evidence was very low. In this regard, guideline recommendations strongly favor cisplatin as the standard radiosensitizer in cisplatin-eligible patients.18,105 In gonadal GCT (testicular and ovarian), statistically significant differences were not found in terms of treatment failure and mortality between cisplatin- and carboplatin-based regimens, in contrast with the results reported by the included meta-analyses.29 This discrepancy may be explained by our use of a more conservative statistical approach (Hartung-Knapp-Sidik-Jonkman method). Nonetheless, carboplatin-based regimens showed numerically higher treatment failures with moderate-certainty evidence and a higher mortality ratio with low certainty of evidence. These findings are consistent with current EAU,106 ESMO,107 and NCCN18 guideline recommendations, in which cisplatin-based regimens remain the curative backbone and are preferred to avoid compromising disease control compared with carboplatin. These results should not be extrapolated to stage I seminomas, which were not represented in the focus of the present analysis, for which single-agent adjuvant carboplatin is guideline supported and backed by a positive noninferiority phase 3 RCT vs radiotherapy.18,106,107,108 Last, although efficacy remains central to platinum selection, our findings highlight that treatment choice should not be guided by efficacy alone. Cisplatin-related nephrotoxicity, ototoxicity, and nausea and vomiting may limit feasibility in older patients or those with comorbidities, whereas carboplatin’s hematologic toxicity may be relevant in patients with limited marrow reserve, especially those with a risk of thrombocytopenia. Peripheral neurotoxicity did not differ significantly between agents and was supported by moderate certainty. However, this finding should be interpreted cautiously, as both agents remain neurotoxic. Cisplatin is generally considered more neurotoxic than carboplatin, although carboplatin-related neuropathy may still occur, particularly with cumulative exposure or in combination with other neurotoxic drugs.109,110 Therefore, treatment choice should continue to integrate efficacy goals with tolerability, age, comorbidities (including renal impairment), prior platinum exposure, quality of life, and patient preferences.4 Limitations This study has some limitations. First, the included evidence was methodologically heterogeneous, merging different approaches to deliver platinum therapy across studies with respect to combination partners, dosing schedules, response assessment criteria, and differences in follow-up timings. As a result, pooled estimates should be interpreted as average effects across heterogeneous clinical contexts, rather than regimen-specific conclusions. Low heterogeneity values did not necessarily imply clinical interchangeability, particularly when trials differ in chemotherapy backbone, treatment intent, or response criteria. Sensitivity analyses were intended to assess this interstudy heterogeneity, although it was only possible to assess 3 subgroups. They showed that specific design features influenced the resulting GRADE certainty ratings. For example, the high-certainty signal for ORR in advanced urothelial carcinoma relied largely on observational studies rather than RCTs. Many of these trials were terminated early because of poor accrual. Second, the umbrella meta-analysis highlighted evidence gaps in the published metadata. Several outcomes graded as low or very low certainty of evidence were driven by substantial heterogeneity and/or a high risk of bias, which led to downgrading under the GRADE framework. Notably, NSCLC contributed the largest number of patients and events, yet its certainty of evidence was still limited by the high risk of bias in a subset of included studies, as discussed above. While this constrains inference, it is also an informative finding of the umbrella review, underscoring the need for more complete and higher-quality reporting of metadata to support robust comparative effectiveness assessments.20 Conclusions In this quantitative umbrella review of published meta-analyses comparing cisplatin and carboplatin, the efficacy and safety of these agents were revisited across oncology settings. By quantifying areas of concordance and discordance between published meta-analytic evidence and guideline recommendations, the review identified clinical contexts in which platinum selection was supported by higher-certainty evidence. It also highlighted contexts in which high-certainty evidence remained lacking for a treatment backbone that is still widely used in solid tumor oncology.