Efficacy and Safety Evidence Supporting Cancer Drug Approvals in Switzerland (2001-2020): A Meta-Analysis of Pivotal Randomised Controlled Trials.

Research in context Evidence before this study Over the past two decades, the number of cancer drug approvals has increased substantially, reflecting advances in oncology drug development. While overall survival (OS) remains the most robust outcome measure, there is a growing reliance on early or intermediate endpoints such as progression- or recurrence-related time-to-event endpoints (PRTTE), including progression-free survival (PFS) and disease-free survival (DFS), in trials supporting drug approvals. Consequently, there is an ongoing debate on the completeness, robustness, and clinical relevance of the evidence available at the time of approval. We searched PubMed for peer-reviewed, original studies published from database inception to March 1, 2026, without language restrictions, using the combination of search terms related to cancer drug approvals and clinical outcomes. The search strategy included terms such as (“cancer” OR “antineoplastic agents”) AND (“drug approval” OR “regulatory approval”) AND (“overall survival” OR “progression-free survival” OR “disease-free survival”). Our search identified several evaluations of the evidence supporting approvals particularly by the U.S. Food and Drug Administration (FDA). Most evaluations have focused on measures of efficacy without parallel assessment of safety. In addition, a subset of studies acknowledged discrepancies between OS and PRTTE, however, comprehensive assessments comparing the treatment effects between OS and PRTTE across cancer types, drug classes, and trial characteristics remain limited. Added value of this study We systematically evaluated the evidence of clinical efficacy and safety supporting cancer drugs authorised in Switzerland between 2001 and 2020. This meta-analysis included 241 randomised controlled trials (RCTs) supporting approvals of new active substances and indication extensions. PRTTE, mainly PFS, was continuously used more frequently than OS as the primary endpoint in pivotal trials. Approved cancer drugs were associated with a considerable reduction in the risk of death and progression or recurrence, but with modest absolute survival gains and were accompanied by an increase in SAEs, compared with the control arm. However, the differences in treatment effects between OS and PRTTE (PFS) varied across solid cancer types and drug classes in the palliative setting with close alignment in some settings such as head and neck cancer and among immune checkpoint inhibitors, and greater divergence in others, including ovarian cancer and targeted agents. Implications of all the available evidence The magnitude of treatment effects on OS and PRTTE (PFS/DFS) should be interpreted with caution, as concordance between these endpoints may vary by cancer type and drug class. These differences highlight the need to evaluate early or intermediate endpoints within their specific clinical and biological context. From a regulatory perspective, the limited availability of OS as a primary endpoint and the reliance on early or intermediate endpoints highlight the ongoing challenges in drug assessment. Ensuring timely access to meaningful OS data, either at initial approval or through robust post-approval evidence requirements, remains essential for informed decision-making.

Introduction

Despite major improvements in survival across various cancer types in the last two decades, cancer continues to be one of the leading causes of death worldwide. 1 Between 2001 and 2020, the number of cancer drug approvals increased significantly, reflecting scientific innovation and an increased pace of cancer drug development. 2

From a regulatory perspective, a new drug can be approved if it demonstrates a positive benefit-risk profile, meaning its benefits and risks are well characterised and the benefits outweigh the risks for the intended population. 3 Randomised controlled trials (RCTs) are considered the gold standard for primarily assessing the efficacy and safety of new cancer treatments. 4 Overall survival (OS) is considered the most robust and clinically relevant endpoint. 5 However, its use as a primary endpoint is limited by the need for longer follow-up and by factors such as crossover and post progression therapies, which may reduce observable differences in OS between study arms and make the attribution of survival benefit to the initial intervention more challenging to assess. 6 , 7 , 8 As a result, early or intermediate endpoints such as progression- or recurrence-related time to event endpoints (PRTTE), including progression-free survival (PFS) and disease-free survival (DFS), as well as response-based endpoints such as objective response rate (ORR) are often used. 9 , 10 PRTTE endpoints can shorten study duration and costs, thereby facilitating a timely assessment of the efficacy of new therapies, which may lead to faster approvals. 11 However, these measures may be weak predictors of actual OS. 12 With the growing reliance on early or intermediate endpoints, concerns have been raised regarding the completeness, medical relevance, and quality of the clinical evidence supporting oncology drug approvals. Despite systematic evaluations of the evidence supporting cancer drug approvals by the U.S. Food and Drug Administration (FDA) and the European Medicines Agency (EMA), 13 , 14 , 15 , 16 , 17 , 18 , 19 comprehensive assessments of the efficacy and, in particular, safety evidence at approval, as well as differences in treatment effects between OS and PRTTE, remain limited.

We aimed to systematically evaluate clinical efficacy and safety evidence supporting cancer drugs authorised by Swissmedic between 2001 and 2020. Our study included most commonly used and clinically relevant endpoints reported in pivotal trials at the time of approval, and compared treatment effects across cancer types, drug classes, and trial characteristics.

Methods

Search strategy and selection criteria

We retrospectively identified the pivotal clinical trials supporting cancer drug approvals in Switzerland from Jan 1, 2001, to Dec 31, 2020, using the internal Swissmedic database. Pivotal RCTs supporting the approval of new active substances and indication extensions were included. We excluded single-arm studies, trials with non-inferiority design, those supporting non-systemic, biosimilar, and supportive care drugs, and the studies not reporting results for any of the prespecified outcomes of interest (OS, PRTTE, or response rate).

Procedure

Data were extracted from Swissmedic clinical assessment reports and submitted dossiers from the applicants at the time of approval. Data extraction was performed by P.R. using a predefined framework. Data extraction, data quality evaluation and data cleaning were conducted between December 2022 and December 2024.

For each RCT, we collected information on drug class, defined by the mechanism of action, cancer type and approved indication label, trial characteristics (phase, blinding, trial comparator, treatment setting, line of therapy, (co)primary endpoints, type I error-controlled endpoints, crossover), and outcome data.

We classified the outcome data as 1. OS and PRTTE (including PFS, time to progression (TTP), metastasis-free survival (MFS), DFS, relapse-free survival (RFS), and event-free survival (EFS)), 2. Response rate, and 3. Safety outcomes, including serious adverse events (SAEs), defined according to ICH E2A 20 and irrespective of treatment attribution, and adverse events (AEs) grade 3–4 and 5, defined according to the Common Terminology Criteria for Adverse Events (CTC-AE). As the included PRTTE endpoints consisted predominantly of PFS in the palliative setting and DFS in the curative setting, these endpoints are collectively referred to as PFS/DFS. In analyses restricted to the palliative setting, the term PFS is used. Data were extracted according to the approved indication label and the most updated results available at the time of approval. Subgroup-specific results were used when approvals were restricted to subpopulations; otherwise, intention-to-treat data were used.

For time-to-event endpoints, we extracted sample size, number of events per arm, hazard ratios (HRs) with 95% confidence intervals (CIs), and median survival times, when available, with 95% CIs in both arms. Analyses were classified as interim or primary based on whether the prespecified number of events recorded in the statistical analysis plan had been reached at approval.

Data validation and quality control

To ensure data quality, a preplanned, structured validation process was implemented. Systematic review of all entered data from randomly selected applications was performed by independent reviewers who are clinical assessors from Swissmedic. This systematic review was conducted in iterative rounds, with discrepancies resolved by consensus or, when necessary, adjudicated by senior oncologists, improving the consistency of data extraction. After several rounds, covering 30% of indication extensions and 20% of new active substance applications, discrepancies were close to zero; further quality checks were therefore discontinued, and the data were considered sufficiently robust.

Statistical analysis

We descriptively summarised the trial characteristics. Trends in the use of different primary endpoints were summarised across four consecutive five-year intervals.

Odds ratios (ORs) were calculated for response rate and risk ratios (RRs) and risk differences (RDs) for safety outcomes. A continuity correction of 0.5 was applied to account for cases of zero events. To evaluate and compare efficacy and safety outcomes across trials, we identified trials that reported OS, PRTTE, and SAEs data at approval, creating a matched dataset. Safety was primarily assessed using SAEs, as these were most consistently reported, whereas AEs were reported heterogeneously across trials (e.g., grade 3–4 vs. grade 3–4–5).

Effect estimates were pooled using random-effects models for meta-analysis. 21 Statistical heterogeneity was assessed with the I 2 statistics. Meta-analyses were conducted overall and stratified by cancer type (solid tumours vs. haematological neoplasms). Absolute risks for SAEs and AEs were meta-analysed for the experimental and control arms. 22 Absolute median OS and PRTTE gains (treatment minus control) were pooled using Wald approximation-based approach 23 implemented in the R package metamedian. 24

Subgroup analyses were performed for solid tumours by treatment settings. Trials in the palliative setting were further stratified by cancer type, drug class, and trial characteristics. Meta-analyses were conducted for subgroups with at least four trials. In addition, we conducted exploratory analyses based on the line of therapy among solid tumour trials in the palliative setting, where sufficient trial numbers were available.

Univariate meta-regression analyses were used to explore the association between study-level characteristics and treatment effects. Each covariate was examined individually using a mixed-effects meta-regression model with restricted maximum likelihood (REML) to estimate between-study variance. The Knapp-Hartung 25 method was used to adjust the estimated standard errors for the estimated coefficients to account for the uncertainty in the estimate of the amount of residual heterogeneity. The 95% CIs of the regression coefficients were obtained from t statistics and the omnibus test to investigate the association between the covariate and outcome was based on the F test. R packages meta (version 8.2-1) and metafor (version 4.8-0) were used to perform all the meta-analyses. 26 , 27 A two-sided p-value of less than 0.05 was considered statistically significant. Statistical tests were not adjusted for multiple testing. Analyses were conducted using R software (version 4.5.1; R Foundation for Statistical Computing) and RStudio (version 202 5.5.1.513; RStudio, PBC).

Ethics statement

The analysed outcome data did not involve the collection or use of identifiable individual patient data and therefore do not fall under the Swiss Human Research Act; accordingly, approval from an ethics committee was not required.

Reporting

This study was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines, where applicable.

Role of the funding source

Prof. Andreas Wicki as representative of the University of Zurich was involved in the study concept and design, interpretation of the data, writing and editing of the manuscript, and in the decision to submit the paper for publication.

Results

We identified a total of 378 pivotal clinical trials supporting cancer drug approvals in Switzerland between Jan 1, 2001, and Dec 31, 2020. After full review, 241 eligible RCTs fulfilled the inclusion criteria and were included in the final analysis ( Fig. 1 ). These 241 clinical trials supported the approval of new active substances and indication extensions for 102 distinct drugs ( Appendix p 1). Most RCTs (171/241 [71%]) evaluated treatments for solid tumours and 97/241 (40.2%) investigated targeted agents ( Table 1 ). Table 1 summarises the main characteristics of the pivotal trials. Fig. 1 Flowchart of study selection. NAS = new active substance. IE = indication extension. OS = overall survival. PRTTE = progression- or recurrence-related time-to-event endpoints. SAEs = serious adverse events. ∗Applications for antineoplastic agents L01-04. † Six haematology trials were excluded because none reported the outcomes prespecified for this meta-analysis (OS, PRTTE, or response rate). ‡ PRTTE consisting predominantly of progression-free survival and disease-free survival, with limited number of time to progression, relapse-free survival, event-free survival, and metastasis-free survival. Table 1 Characteristics of pivotal randomised controlled trials supporting cancer drug approvals in Switzerland, 2001–2020. Number of trials Total 241 (100%) Cancer type Solid tumours 171 (71%) Breast 34 (19.9%) Lung 28 (16.4%) Colorectal 20 (11.7%) Melanoma 14 (8.2%) Renal 12 (7%) Ovarian 12 (7%) Sarcoma/GIST 8 (4.7%) Gastroesophageal 7 (4.1%) Head and neck 7 (4.1%) Prostate 6 (3.5%) Liver 5 (2.9%) Other a 18 (10.5%) Haematological neoplasms 70 (29%) Leukaemia 31 (44.3%) Multiple myeloma 20 (28.6%) Lymphoma 13 (18.6%) Other b 6 (8.6%) Treatment type Monotherapy 112 (46.5%) Combination 129 (53.5%) Drug class d Targeted agent 97 (40.2%) Monoclonal antibody 58 (24.1%) Immune checkpoint inhibitor 35 (14.5%) Cytotoxic agent 29 (12%) Antibody-drug conjugate 8 (3.3%) Endocrine therapy 5 (2.1%) Other c 9 (3.7%) Indication type New active substance 77 (32%) Indication extension 164 (68%) Treatment setting Curative setting 24 (10%) Palliative maintenance setting 16 (6.6%) Palliative setting 201 (83.4%) Treatment line (palliative setting) First line 103 (51.2%) Second and/or further line 88 (43.8%) Any line 10 (5%) Pivotal trial phase I/II 2 (0.8%) II 16 (6.6%) II/III 2 (0.8%) III 221 (91.7%) Type of blinding Open label 146 (60.6%) Double-blinded 95 (39%) Control arm Active control 185 (76.8%) Placebo control 49 (20.3%) No treatment/observation 7 (2.9%) Primary endpoints e Progression or recurrence related time-to-event 167 (69.2%) Progression-free survival 135 (80.6%) Disease-free survival 14 (8.4%) Event-free survival 3 (1.8%) Recurrence-free survival 4 (2.4%) Time-to-progression 9 (5.4%) Metastasis-free survival 2 (1.2%) Overall survival 73 (30.2%) Response rate 28 (11.6%) Data are n (%). GIST = Gastrointestinal Stromal Tumour. a Thyroid, plural, pancreas, neuroendocrine, brain, tuberous sclerosis complex. b Waldenström macroglobulinaemia, Myelodysplastic Syndrome, Multicentric Castleman’s disease, Polycythaemia Vera, Myelodysplastic/myeloproliferative syndrome. c Immunomodulatory agents, oncolytic virus therapies, macrophage activator, histone deacetylase inhibitor. d Drug classes and the corresponding subclasses are listed in the Appendix p 1. e In some clinical trials, multiple primary endpoints (co-primary) were used, causing the total number to exceed 241 and percentage to exceed 100%.

The proportion of trials with PRTTE as primary endpoint was higher (167/241 [69.2%]) than those with OS as primary endpoint (73/241 [30.2%]). From 2001 to 2020, PRTTE, mainly PFS, were consistently selected more frequently as primary endpoints than OS, reaching 73.8% in 2016–2020 ( Fig. 2 ). OS as primary endpoint remained stable over time, ranging from 26% to 34%. In parallel, the percentage of trials including OS as a key secondary endpoint (controlled under type I error) increased considerably, from 0% in 2001–2005 to 57.3% in 2016–2020. Among trials with OS as the primary endpoint, 32 (43.8%) reported an interim analysis, compared with 82 (49.1%) with PRTTE as the primary endpoint. Fig. 2 Temporal trends in primary endpoints of pivotal randomised controlled trials supporting cancer drug approvals in Switzerland, 2001–2020. RCT = randomised controlled trials. PRTTE = progression- or recurrence-related time-to-event endpoints. OS = overall survival. The figure shows the proportions of primary endpoints and overall survival as a controlled key secondary endpoint over time. In some clinical trials, multiple primary endpoints (co-primary) were used, causing the total percentage to exceed 100%. ∗PRTTE consisting predominantly of progression-free survival and disease-free survival, with limited number of time to progression, relapse-free survival, event-free survival, and metastasis-free survival.

Of 241 trials, 194 reported OS, PRTTE, and SAEs (matched dataset). The new approved cancer drugs reduced the risk of death by 24% compared with the control arm (OS HR 0.76; 95% CI 0.74–0.77; I 2 = 24.5%) ( Table 2 ). The risk of disease progression or recurrence was reduced by 45% (PRTTE (PFS/DFS) HR 0.55; 95% CI 0.52–0.58; I 2 = 86.6%). Table 2 Treatment outcomes for overall survival, progression- or recurrence-related time to event endpoints, and serious adverse events (matched dataset) in randomised controlled trials supporting cancer drug approvals in Switzerland, 2001–2020. Overall Survival Progression- or recurrence-related time-to-event a Serious Adverse Events No. of trials Events/total population experimental arm Events/total population control arm HR (95% CI) P-value b Events/total population experimental arm Events/total population control arm HR (95% CI) P-value b Events/total population c experimental arm Events/total population c control arm RR e (95% CI) P-value b A-ll trials 194 21,203/61,261 20,118/52,523 0.76 (0.74–0.77) 30,596/60,569 30,622/51,879 0.55 (0.52–0.58) 22,004/63,156 15,267/54,503 1.26 (1.21–1.32) Cancer type 0.64 0.078 0.36 Solid tumours 142 18,119/48,808 16,750/40,590 0.76 (0.74–0.78) 25,505/48,214 24,278/40,041 0.57 (0.54–0.60) 16,164/50,711 10,815/42,764 1.28 (1.21–1.36) Haematological neoplasms 52 3084/12,453 3368/11,933 0.75 (0.70–0.79) 5091/12,355 6344/11,838 0.50 (0.44–0.57) 5840/12,445 4452/11,739 1.22 (1.12–1.32) Solid tumours, Treatment setting 0.57 0.037 <0.0001 Curative 13 958/10,159 1134/9890 0.74 (0.68–0.82) 1835/10,159 2424/9890 0.57 (0.46–0.71) 2483/10,335 1706/10,632 1.70 (1.29–2.25) Palliative maintenance 12 1222/3004 1013/1993 0.81 (0.72–0.90) 1825/3004 1600/1993 0.44 (0.35–0.55) 853/3563 357/2715 1.79 (1.39–2.31) Palliative 117 15,939/35,645 14,603/28,707 0.76 (0.74–0.78) 21,845/35,051 20,254/28,158 0.59 (0.55–0.62) 12,828/36,813 8752/29,417 1.20 (1.15–1.26) Solid tumours, Palliative setting Cancer entities <0.0001 0.019 0.026 Lung 24 3178/6550 3017/5520 0.75 (0.71–0.80) 4269/6550 4136/5520 0.62 (0.55–0.70) 2224/6754 1727/5504 1.06 (0-99-1.14) Breast 19 2022/6479 1750/4917 0.81 (0.76–0.87) 3603/6389 3279/4875 0.64 (0.58–0.70) 1586/6602 873/4976 1.41 (1.21–1.64) Colorectal 14 2797/4513 2633/3915 0.80 (0.75–0.86) 3362/4406 2980/3802 0.62 (0.55–0.71) 1849/4751 1422/4092 1.12 (1.01-1.25) Renal 11 1443/3933 1595/3781 0.74 (0.68–0.79) 2143/3866 2369/3714 0.60 (0.52–0.70) 1563/4053 1254/3829 1.18 (1.04–1.35) Melanoma 10 902/2323 836/1919 0.66 (0.58–0.75) 1287/2262 1290/1857 0.53 (0.42–0.66) 907/2336 620/1910 1.25 (1.03–1.51) Sarcoma/GIST 6 296/808 184/477 0.56 (0.43–0.73) 407/808 313/477 0.35 (0.23–0.51) 349/975 185/635 1.21 (0.94–1.56) Gastroesophageal 5 1025/1463 776/1003 0.71 (0.64–0.79) 1244/1529 940/1075 0.61 (0.54–0.68) 628/1522 416/1063 1.03 (0.92–1.14) Liver 5 936/1681 624/994 0.69 (0.62–0.76) 1011/1681 581/994 0.49 (0.43–0.56) 744/1664 417/983 1.10 (0.94–1.29) Prostate 5 1041/3271 843/2067 0.73 (0.66–0.80) 1362/2936 1114/1730 0.55 (0.37–0.82) 904/3251 463/2052 1.23 (0.94–1.59) Ovarian 4 662/1522 697/1524 0.87 (0.78–0.97) 956/1522 1052/1524 0.55 (0.41–0.75) 515/1497 323/1509 1.59 (1.18–2.15) Head and neck 4 556/768 527/641 0.72 (0.64–0.81) 639/768 595/641 0.70 (0.55–0.89) 514/977 400/838 1.08 (0.99–1.19) Drug class 0.02 <0.0001 0.43 Targeted agent 48 4649/13,181 3796/9639 0.74 (0.71–0.78) 6929/12,963 6243/9468 0.48 (0.44–0.52) 4291/13,159 2570/9347 1.25 (1.14–1.37) Immune checkpoint inhibitor 30 4312/8560 4270/7117 0.71 (0.68–0.74) 6099/8560 5632/7117 0.72 (0.66–0.78) 3824/9210 2846/7680 1.13 (1.05–1.21) Monoclonal antibody 25 4139/7452 4122/7136 0.82 (0.79–0.86) 5298/7411 5381/7095 0.64 (0.59–0.69) 2982/7722 2261/7339 1.26 (1.17–1.36) Cytotoxic agent 9 1876/3193 1765/2772 0.77 (0.70–0.85) 2272/2858 1957/2435 0.72 (0.61–0.85) 881/3480 590/3024 1.27 (0.98–1.64) Endocrine therapy 4 781/2764 501/1547 0.80 (0.66–0.96) 982/2764 737/1547 0.52 (0.31–0.86) 774/2752 397/1539 1.07 (0.91–1.27) Submission type 0.33 0.012 0.29 New active substance 45 5770/14,606 4527/10,392 0.75 (0.71–0.78) 8113/14,281 6938/10,108 0.53 (0.48–0.59) 4716/14,542 2736/10,158 1.25 (1.15–1.37) Indication extension 72 10,169/21,039 10,076/18,315 0.76 (0.74–0.79) 13,732/20,770 13,316/18,050 0.62 (0.58–0.66) 8112/22,271 6016/19,259 1.18 (1.11–1.25) Treatment type 0.86 0.0019 0.0013 Monotherapy 58 7061/15,790 6070/12,217 0.75 (0.72–0.78) 9185/15,555 8314/11,975 0.53 (0.48–0.58) 5702/16,235 3950/12,284 1.11 (1.04–1.18) Combination therapy 59 8878/19,855 8533/16,490 0.77 (0.74–0.79) 12,660/19,496 11,940/16,183 0.64 (0.61–0.68) 7126/20,578 4802/17,133 1.30 (1.22–1.39) Treatment line 0.30 0.16 0.13 First line 58 5900/15,546 6256/14,057 0.74 (0.71–0.78) 8431/15,216 9128/13,730 0.59 (0.55–0.64) 5769/16,361 4259/14,639 1.23 (1.15–1.32) Second and/or further line 55 9616/19,104 8101/14,057 0.77 (0.74–0.79) 12,742/18,840 10,634/13,835 0.59 (0.54–0.65) 6648/19,471 4323/14,186 1.16 (1.08–1.24) Any line 4 423/995 246/593 0.77 (0.65–0.93) 672/995 492/593 0.43 (0.34–0.56) 411/981 170/592 1.53 (1.11–2.11) Time of approval 0.043 0.36 0.30 2001–2005 8 1254/2276 1029/1809 0.80 (0.73–0.87) 1018/1874 814/1405 0.60 (0.47–0.76) 677/2274 380/1813 1.36 (1.22–1.51) 2006–2010 18 2519/5540 2513/5133 0.80 (0.75–0.86) 3606/5499 3587/5092 0.63 (0.56–0.71) 2168/5813 1629/5326 1.25 (1.09–1.44) 2011–2015 35 4660/10,131 4213/8349 0.78 (0.75–0.81) 6404/10,070 6090/8287 0.54 (0.48–0.60) 3701/10,396 2495/8378 1.24 (1.13–1.37) 2016–2020 56 7506/17,698 6848/13,416 0.72 (0.69–0.75) 10,817/17,608 9763/13,374 0.60 (0.55–0.65) 6282/18,330 4248/13,900 1.15 (1.08–1.22) Marker-based indication 0.55 0.81 0.34 No 70 11,204/23,401 10,013/18,818 0.76 (0.74–0.79) 14,817/22,999 13,412/18,414 0.59 (0.54–0.64) 8624/23,459 5884/18,613 1.18 (1.11–1.25) Yes 47 4735/12,244 4590/9889 0.75 (0.71–0.79) 7028/12,052 6842/9744 0.58 (0.54–0.63) 4204/13,354 2868/10,804 1.24 (1.15–1.35) Study phase 0.13 0.82 0.66 II 8 572/1202 437/888 0.75 (0.64–0.86) 740/1112 654/846 0.60 (0.50–0.73) 439/1181 281/867 1.25 (1.06–1.47) III 109 15,367/34,443 14,166/27,819 0.76 (0.74–0.78) 21,105/33,939 19,600/27,312 0.58 (0.55–0.62) 12,389/35,632 8471/28,550 1.20 (1.14–1.26) Control arm 0.30 <0.0001 0.94 Active control 90 11,761/25,994 11,915/23,081 0.76 (0.74–0.79) 16,494/25,574 16,560/22,712 0.64 (0.60–0.67) 9370/27,193 7019/23,808 1.20 (1.14–1.27) Placebo control/no treatment 27 4178/9651 2688/5626 0.73 (0.69–0.76) 5351/9477 3694/5446 0.44 (0.39–0.49) 3458/9620 1733/5609 1.21 (1.08–1.35) Blinding type 0.84 0.0009 0.13 Open label 62 8337/16,722 8690/15,601 0.76 (0.73–0.79) 10,902/16,285 11,042/15,161 0.64 (0.60–0.69) 6481/17,947 5054/16,389 1.16 (1.09–1.24) Double blinded 55 7602/18,923 5913/13,106 0.75 (0.72–0.78) 10,943/18,766 9212/12,997 0.53 (0.49–0.57) 6347/18,866 3698/13,028 1.25 (1.17–1.34) Crossover d 0.78 <0.001 0.71 Allowed 23 1698/5118 1604/4091 0.75 (0.69–0.83) 2686/5011 2822/3978 0.46 (0.39–0.54) 1671/5301 1120/4210 1.18 (1.04–1.34) Not allowed 94 14,241/30,527 12,999/24,616 0.76 (0.74–0.78) 19,159/30,040 17,432/24,180 0.62 (0.58–0.65) 11,157/31,512 7632/25,207 1.21 (1.15–1.27) Subgroups with ≥ four trials were included in the analyses. HR = Hazard Ratio. RR = Risk Ratio.CI = confidence interval. GIST = Gastrointestinal Stromal Tumour. a Includes progression-free survival (n = 113), time to progression (n = 2), and metastasis-free survival (n = 2) in the palliative setting, disease-free survival (n = 10), relapse-free (n = 2), and event-free survival (n = 1) in the curative setting, and progression-free survival (n = 12) in maintenance setting. b P-values for moderators were obtained using an F test with Knapp-Hartung adjustment. c Events for serious adverse events refers to the number of patients with at least one serious adverse event. d Protocol allowed crossover from the control arm to the experimental treatment before primary analysis of the trial. e RR was calculated as the risk in the experimental arm divided by the risk in the control arm.

In contrast, treatment was associated with an increased relative risk of SAEs by 26% (RR 1.26; 95% CI 1.21–1.32; I 2 = 82.1%). For AEs grade 5 (death), the pooled RR was 0.98 (95% CI 0.91–1.05) ( Appendix p 3). There was inconsistent reporting of AEs, with 120 trials reporting AEs grade 3–4. The absolute risk of AEs was higher in experimental arm compared to control across all trials (RD 11.54; 95% CI 8.10–14.98) ( Appendix p 4).

Overall, among 93 (38.6%) trials reporting median OS and 146 (60%) reporting median PRTTE (PFS) at approval in both arms, absolute median gains were 2.42 (95% CI 2.11–2.73) months for OS and 3.47 (95% CI 2.89–4.04) months for PRTTE ( Appendix p 5). Looking at response rates, in solid tumours, the odds of achieving a response in the treatment arm were 2.46 times compared with control (OR 2.46; 95% CI 2.20–2.74; I 2 = 80.4%) ( Appendix p 6).

We then stratified trials by cancer type. For OS, the pooled HRs were similar for solid tumours (HR 0.76 [95% CI 0.74–0.78; I 2 = 23.2%]) and haematological neoplasms (HR 0.75 [95% CI 0.70–0.79; I 2 = 28.8%]). The PRTTE (PFS/DFS) benefits were consistently larger in both as compared to OS benefit (solid tumours: PRTTE HR 0.57 [95% CI 0.54–0.60; I 2 = 86.7%]; haematological neoplasms: PRTTE HR 0.50 [95% CI0.44–0.57; I 2 = 86.3%]). The absolute risk of SAEs was higher in experimental than in control arms in both solid and haematological neoplasms (solid: 33.65% [95% CI 31.69–35.61] vs. 27.15% [95% CI 25.06–29.23]; haematological: 45.97% [95% CI 41.15–50.80] vs. 37.76% [95% CI 33.26–42.27]) ( Appendix p 7).

We conducted a subgroup meta-analysis across solid tumour trials in the palliative setting. OS benefit varied significantly across cancer types (p < 0.0001) and drug classes (p = 0.02), with the greatest benefit observed in sarcoma/GIST (HR 0.56 [95% CI 0.43–0.73]) and the least in ovarian cancer (HR 0.87 [95% CI 0.78–0.97]) ( Table 2 ). Across drug classes, immune checkpoint inhibitors showed the greatest OS benefit (HR 0.71 [95% CI 0.68–0.74]), compared with monoclonal antibodies (HR 0.82 [95% CI 0.79–0.86]). PRTTE (PFS) also varied across cancer types (p = 0.019) and drug classes (p < 0.0001), with the greatest benefit observed in sarcoma/GIST (HR 0.35 [95% CI 0.23–0.51]) and with targeted agents (HR 0.48 [95% CI 0.44–0.52]), compared with head and neck cancer (HR 0.70 [95% CI 0.55–0.89]) and cytotoxic agents (HR 0.72 [95% CI 0.61–0.85]). PRTTE effects also differed by comparator type, with greater benefit in placebo-controlled trials (HR 0.44 [95% CI 0.39–0.49]) than in active-controlled trials (HR 0.64 [95% CI 0.60–0.67]; p < 0.0001). The risk of SAEs varied across cancer types (p = 0.026), with the highest RR in ovarian cancer (RR 1.59 [95% CI 1.18–2.15]) and lowest in gastroesophageal cancer (RR 1.03 [95% CI 0.92–1.14]).

Looking at treatment effect differences between OS and PRTTE (PFS), we observed a variation across cancer types and drug classes in the palliative setting ( Fig. 3 ). In some cancer types like ovarian cancer, the difference between pooled HRs of OS and PRTTE was pronounced (OS HR 0.87 vs. PRTTE HR 0.55). In contrast, in cancers like head and neck, pooled HRs for OS and PRTTE were more closely aligned (OS HR 0.72 vs. PRTTE HR 0.70). Across drug classes, while targeted agents showed a larger difference between treatment effect for OS and PRTTE (OS HR 0.74 vs. PRTTE HR 0.48), in immune checkpoint inhibitors the pooled HRs of OS and PRTTE were more comparable (OS HR 0.71 vs. PRTTE HR 0.72). In exploratory analyses, differences between OS and PRTTE varied by line of therapy, differing between first line and later line settings across cancer types in palliative setting ( Appendix p 9). Fig. 3 Treatment outcomes for overall survival, progression- or recurrence-related time-to-event, and serious adverse events of randomised controlled trials supporting solid tumour drug approvals in the palliative setting in Switzerland, 2001–2020. CI = confidence interval. I 2 = the degree of heterogeneity across studies. HR = hazard ratio. RR = risk ratio. GIST = gastrointestinal stromal tumour. Forest plots show pooled treatment effects for OS (blue line), PRTTE (progression-free survival) (green line), and SAEs (orange line). The figure highlights differences in the treatment effects between PRTTE and OS across cancer types and drug classes, illustrating that the degree of alignment between progression-based endpoints and survival varies across these subgroups. (a) Across solid tumour cancer types. (b) Across drug classes. Hazard ratios derived from a random-effects model with the restricted maximum likelihood (REML) method. Risk ratios were derived using a random-effects model with the inverse-variance model. Only subgroups with ≥ four trials were included in the analysis. For serious adverse events, events refer to the number of patients with at least one serious adverse event.

Discussion

Our study systematically evaluated the clinical evidence supporting cancer drug approvals by Swissmedic between 2001 and 2020, focusing on efficacy and safety outcomes reported at the time of regulatory approval. The use of Swissmedic data offers relevant insights, as its regulatory evidence base and decisions are generally well aligned with those of the EMA and FDA. 28

In total, 241 RCTs supported the approvals of 102 distinct cancer drugs. At approval, cancer drugs were associated with reduction in the risk of death (24%) and progression or recurrence (45%). These relative benefits translated into modest median absolute survival gains (OS 2.42 months; PRTTE 3.47 months), based on a limited subset of trials with available median OS and PRTTE (PFS) in both arms. However, long-term survival, an important measure to assess clinical benefit of drug interventions, was generally not available at the time of approval. As a result, HRs and median survival at approval may underestimate the ultimate benefit of a therapy in terms of long-term survival. Our findings are consistent with previous studies of cancer drug approvals, which similarly report modest survival improvements at the time of approval and variability in endpoint selection and clinical benefit. 13 , 15 Michaeli et al. 13 reported pooled HRs of 0.73 for OS and 0.57 for PFS, corresponding to median survival gains of approximately 2–3 months for FDA approvals (2003–2021). Similarly, prior study in Switzerland has shown that 39% of trials supporting approvals meet the ESMO-MCBS threshold for substantial clinical benefit at the time of approval. 29

In terms of safety, we found that newly approved drugs were associated with an increased risk of SAEs (RD 6.91%) and a higher risk of grade 3–4 (RD 11.54%) compared with control arms, consistent with previous studies. 30

OS remains underused as a primary endpoint, being employed in less than a third of approvals. However, OS is increasingly included as a key secondary endpoint, formally controlled for multiplicity within hierarchical or graphical testing strategies that preserve the overall type I error rate. 31 PRTTE endpoints, mostly PFS, remain the most common primary endpoint, reflecting a pragmatic approach to generate earlier efficacy data to support regulatory decisions. However, the validity of these endpoints as surrogates for OS remains uncertain and appears to vary across tumour types, treatment modalities, and lines of therapy. 32

Across trials, we observed consistently larger effects for PRTTE (PFS/DFS) than for OS. This is consistent with previous studies and suggests that early endpoints may overestimate the true magnitude of survival benefit. 13 , 15 , 16 , 33 In our analysis, we described the magnitude of this “shrinking effect” between OS and PRTTE endpoints and showed that it varied by cancer type and drug class in the palliative setting. Some tumour types such as head and neck cancer showed relatively consistent treatment effects for OS and PRTTE (PFS). In contrast, others such as ovarian cancer exhibited substantial discrepancies between these endpoints. These differences may arise from variations in disease biology, mechanism of action, and treatment sequencing. Furthermore, our exploratory analyses suggest that the relationship between OS and PRTTE may vary by treatment line.

Targeted therapies tended to show larger effects on early endpoints than on OS, potentially reflecting acquired resistance and subsequent treatment adapted to the mechanism of resistance, whereas immune checkpoint inhibitors demonstrated more concordant effects, possibly due to durable responses in a subset of patients. However, the OS benefit for targeted agents was very similar to that of immune checkpoint inhibitors. These findings underscore the biological and clinical factors that may influence the relationship between early endpoints and long-term outcomes and highlight important considerations when interpreting early evidence.

Taken together, our findings highlight the challenges faced by regulators and clinicians when relying on early efficacy endpoints. The limited availability of primary analysis and long-term OS data, together with variability in the association between PRTTE and OS, suggests that early endpoints should be interpreted cautiously, particularly when extrapolating to long-term survival benefit. While early endpoints can facilitate timely access to new therapies, their use should be accompanied by robust post-approval evidence generation to confirm survival benefit and more comprehensively characterise safety.

While our study offers new insights, limitations must be acknowledged. First, additional outcomes such as quality of life were not included, as analyses of Swissmedic approvals indicate that these data are heterogeneously reported and insufficiently standardised for pooled meta-analysis. 34 Second, meta-regression analyses were limited to univariate models due to the risk of overfitting given the available data. Third, safety analyses were limited by the lack of individual patient data and primarily relied on SAEs, as reporting of AEs grade 3–4 was available only for a subset of trials. Fourth, substantial heterogeneity was observed across trials for PRTTE, reflecting differences in endpoint definitions, assessment methods, follow-up, and patient populations. Although random-effects models were used, pooled estimates should be interpreted with caution, as they represent average effects across heterogeneous contexts and may not be fully generalisable. In addition, PRTTE and OS endpoints may be influenced by informative censoring, 35 , 36 which could affect treatment-effect estimates, particularly in the absence of individual patient data. We also did not assess the adequacy of comparator arms beyond broad classification, a complex and context-dependent issue that may influence observed effect sizes. Finally, this study includes only trials submitted for regulatory approval and may therefore be subject to publication or submission bias, as studies with unfavourable results are less likely to be submitted. Formal assessment of publication bias was not feasible given the nature of the dataset.

In conclusion, cancer drugs approved by Swissmedic were associated with treatment benefits at the time of approval, but these translated into modest absolute survival gains and were accompanied by increased risk of AEs. Important evidence gaps remain at the time of regulatory approval, including limited availability of median survival gain, long-term OS data, and a frequent reliance on early or intermediate endpoints, whose validity as surrogates varies across clinical settings. These findings highlight the importance of systematic and transparent post-approval monitoring and careful interpretation of early endpoints, taking cancer type and drug class into account to ensure meaningful long-term patient benefit.

Contributors

AW (Wicki), PR, and U-PR led the study concept and design. SDT provided the list of approvals. PR led the collection and analysis of the data. QL provided statistical expertise. All authors had access to the data and analyses. QL, LR, SDT, NM, JB, NB, AG, MI, SJ, KE, BM, EZ, MZ-P, and AW (Wolfer) accessed and verified the data, and conducted data monitoring and quality assurance. AW (Wicki), PR, U-PR, AW (Wolfer), QL, and LR contributed to the interpretation of the data. AW (Wicki), PR, and U-PR drafted and revised the manuscript. All authors contributed to the review and editing of the manuscript and had final responsibility for the decision to submit for publication.

Data sharing statement

Due to legal reasons in Switzerland, the data are not publicly available, nor can they be made available beyond the information given in the publication.

Declaration of interests

AW (Wicki) is the academic supervisor of the Precision Oncology Program, a public-private partnership between the University of Zurich, University Hospital Zurich, and F. Hoffmann-La Roche. AW (Wicki) is also a co-inventor in a patent on the analysis of circulating tumour cells by apheresis, Captaim study (Application PCT/EP2025/052381/WO2025163057A1). U-PR declares travel expenses covered by the organisers for ESMO 2025 as an invited congress speaker and as speaker for The New York Lung Cancer Foundation (NYLCF) annual meeting Lung Cancer: Call-To-Action, March 2026. All other authors declare no competing interests.

References

1 World Health Organization Cancer 2025 https://www.who.int/news-room/fact-sheets/detail/cancer

2 Rahimzadeh P. Li Q. Rudofsky L. Trends in pivotal clinical trial design and biomarker use: a retrospective analysis of oncology drug approval in Switzerland from 2001 to 2020 BMJ Connect Oncol 2 1 2025 e000054

3 US Food and Drug Administration Benefit-risk assessment for new drug and biological products: guidance for industry 2023 https://www.fda.gov/regulatory-information/search-fda-guidance-documents/benefit-risk-assessment-new-drug-and-biological-products

4 Booth C.M. Cescon D.W. Wang L. Tannock I.F. Krzyzanowska M.K. Evolution of the randomized controlled trial in oncology over three decades J Clin Oncol 26 33 2008 5458 5464 18955452 10.1200/JCO.2008.16.5456 PMC2651075

5 Delgado A. Guddati A.K. Clinical endpoints in oncology - a primer Am J Cancer Res 11 4 2021 1121 1131 33948349 PMC8085844

6 Soria J.C. Massard C. Le Chevalier T. Should progression-free survival be the primary measure of efficacy for advanced NSCLC therapy? Ann Oncol 21 12 2010 2324 2332 20497965 10.1093/annonc/mdq204

7 Jönsson L. Sandin R. Ekman M. Analyzing overall survival in randomized controlled trials with crossover and implications for economic evaluation Value Health 17 6 2014 707 713 25236994 10.1016/j.jval.2014.06.006

8 US Food and Drug Administration Approaches to assessment of overall survival in oncology clinical trials: guidance for industry 2025 https://www.fda.gov/regulatory-information/search-fda-guidance-documents/approaches-assessment-overall-survival-oncology-clinical-trials

9 Saad E.D. Buyse M. Statistical controversies in clinical research: end points other than overall survival are vital for regulatory approval of anticancer agents Ann Oncol 27 3 2016 373 378 26578738 10.1093/annonc/mdv562

10 Mittal A. Kim M.S. Dunn S. Wright K. Gyawali B. Frequently asked questions on surrogate endpoints in oncology-opportunities, pitfalls, and the way forward eClinicalMedicine 76 2024 10.1016/j.eclinm.2024.102824 PMC11701476 39764569

11 Sharma R. Gulati A. Chopra K. Era of surrogate endpoints and accelerated approvals: a comprehensive review on applicability, uncertainties, and challenges from regulatory, payer, and patient perspectives Eur J Clin Pharmacol 81 5 2025 605 623 40080138 10.1007/s00228-025-03822-w

12 Haslam A. Hey S.P. Gill J. Prasad V. A systematic review of trial-level meta-analyses measuring the strength of association between surrogate end-points and overall survival in oncology Eur J Cancer 106 2019 196 211 30528804 10.1016/j.ejca.2018.11.012

13 Michaeli D.T. Michaeli T. Overall survival, progression-free survival, and tumor response benefit supporting initial US food and drug administration approval and indication extension of new cancer drugs, 2003-2021 J Clin Oncol 40 35 2022 4095 4106 35921606 10.1200/JCO.22.00535

14 Tibau A. Molto C. Ocana A. Magnitude of clinical benefit of cancer drugs approved by the US food and drug administration J Natl Cancer Inst 110 5 2018 486 492 29244173 10.1093/jnci/djx232

15 Gloy V. Schmitt A.M. Düblin P. The evidence base of US food and Drug Administration approvals of novel cancer therapies from 2000 to 2020 Int J Cancer 152 12 2023 2474 2484 36779785 10.1002/ijc.34473

16 Ladanie A. Schmitt A.M. Speich B. Clinical trial evidence supporting US food and drug administration approval of novel cancer therapies between 2000 and 2016 JAMA Netw Open 3 11 2020 e2024406 10.1001/jamanetworkopen.2020.24406 PMC7656288 33170262

17 Michaeli D.T. Michaeli T. Albers S. Michaeli J.C. Clinical trial design and treatment effects: a meta-analysis of randomised controlled and single-arm trials supporting 437 FDA approvals of cancer drugs and indications BMJ Evid Based Med 29 5 2024 333 341 10.1136/bmjebm-2023-112544 38760158

18 Davis C. Naci H. Gurpinar E. Poplavska E. Pinto A. Aggarwal A. Availability of evidence of benefits on overall survival and quality of life of cancer drugs approved by European Medicines Agency: retrospective cohort study of drug approvals 2009-13 BMJ 359 2017 j4530 10.1136/bmj.j4530 PMC5627352 28978555

19 Chen E.Y. Raghunathan V. Prasad V. An overview of cancer drugs approved by the US Food and Drug administration based on the surrogate end point of response rate JAMA Intern Med 179 7 2019 915 921 31135822 10.1001/jamainternmed.2019.0583 PMC6547222

20 International conference on harmonisation of technical requirements for registration of pharmaceuticals for human use ICH E2A: clinical safety data management: definitions and standards for expedited reporting 1994 https://database.ich.org/sites/default/files/E2A_Guideline.pdf 10.1111/j.1365-2125.1994.tb05705.x PMC1364893 8054244

21 DerSimonian R. Laird N. Meta-analysis in clinical trials Control Clin Trials 7 3 1986 177 188 3802833 10.1016/0197-2456(86)90046-2

22 Wang N. Conducting meta-analyses of proportions in R J Behav Data Sci 3 2 2023 64 126

23 McGrath S. Kimmelman J. Ozturk O. Steele R. Benedetti A. Meta-Analysis of Median Survival Times with Inverse-Variance Weighting 2025 10.1002/sim.70533 41998861

24 McGrath S. Zhao X. Ozturk O. Katzenschlager S. Steele R. Benedetti A. Metamedian: an R package for meta-analyzing studies reporting medians Res Synth Methods 15 2 2024 332 346 38073145 10.1002/jrsm.1686

25 Knapp G. Hartung J. Improved tests for a random effects meta-regression with a single covariate Stat Med 22 17 2003 2693 2710 12939780 10.1002/sim.1482

26 Balduzzi S. Rücker G. Schwarzer G. How to perform a meta-analysis with R: a practical tutorial Evid Based Ment Health 22 4 2019 153 160 31563865 10.1136/ebmental-2019-300117 PMC10231495

27 Viechtbauer W. Conducting meta-analyses in R with the metafor package J Stat Software 36 3 2010 1 48

28 Rohr U.P. Iovino M. Rudofsky L. A decade comparison of regulatory decision patterns for oncology products to all other non-oncology products among Swissmedic, European Medicines Agency, and US Food and Drug Administration Clin Transl Sci 16 9 2023 1569 1581 37408165 10.1111/cts.13567 PMC10499418

29 Adam R. Tibau A. Molto Valiente C. Clinical benefit of cancer drugs approved in Switzerland 2010–2019 PLoS One 17 6 2022 e0268545 10.1371/journal.pone.0268545 PMC9187080 35687539

30 Moreau Bachelard C. Coquan E. du Rusquec P. Paoletti X. Le Tourneau C. Risks and benefits of anticancer drugs in advanced cancer patients: a systematic review and meta-analysis EClinicalMedicine 40 2021 101130 10.1016/j.eclinm.2021.101130 PMC8548931 34746718

31 Bretz F. Maurer W. Brannath W. Posch M. A graphical approach to sequentially rejective multiple test procedures Stat Med 28 4 2009 586 604 19051220 10.1002/sim.3495

32 Chu G. Zhu X. Wu J. Assessing correlation between surrogate endpoints and overall survival for oncology clinical trials Clin Pharmacol Ther 117 6 2025 1706 1717 40013399 10.1002/cpt.3613

33 Locher L. Serra-Burriel M. Trapani D. Nussli E. Vokinger K.N. Why effect sizes are systematically larger for progression-free survival than overall survival in cancer drug trials: prognostic scores as a way forward Eur J Cancer 213 2024 10.1016/j.ejca.2024.115106 39550905

34 Horst A. Löffler-Perez N. Rohr U.P. Wac K. Quality of life in oncology drug approvals: evidence from Swiss decisions in relation to global regulatory trends Qual Life Res 35 3 2026 72 41677984 10.1007/s11136-026-04180-5 PMC12901070

35 Templeton A.J. Amir E. Tannock I.F. Informative censoring — a neglected cause of bias in oncology trials Nat Rev Clin Oncol 17 6 2020 327 328 32273582 10.1038/s41571-020-0368-0

36 Gilboa S. Pras Y. Mataraso A. Bomze D. Markel G. Meirson T. Informative censoring of surrogate end-point data in phase 3 oncology trials Eur J Cancer 153 2021 190 202 34186504 10.1016/j.ejca.2021.04.044

Supplementary data

Supplemetary material PRISMA_checklist

Acknowledgements

This study was funded by the 10.13039/501100006447 University of Zurich . During the preparation of this work, the authors used ChatGPT (OpenAI) to correct grammatical errors and improve readability. The scientific content and the figures were generated without input from AI. All authors reviewed and edited the content and take full responsibility for the final version.

Supplementary data related to this article can be found at https://doi.org/10.1016/j.lanepe.2026.101710 .