Background
Registration trials in oncology are increasingly enrolling patients from many different countries and regions. International trials offer many advantages, including the rapid recruitment of large numbers of participants and the ability to test treatments across different populations with unique attributes, genetics, and ethnicity. When pooling outcomes across regions and countries, studies assume that treatment effects are similar. However, even among high-income countries with well-funded health systems, age-standardized outcomes vary. For example, between 2010 and 2014, the 5-year survival rate for esophageal cancer in Australia was 23.5% compared with 14.7% in Denmark.1 Although there may be differences in coding or timing of diagnosis, real outcome differences driven by regional practice variations, genetic differences, and/or variable treatment responses are likely contributors.
Regional differences in treatment efficacy can generate doubt about the generalizability of results. Although clinical trials strive to standardize on-study treatment to minimize the effects of practice variation on outcome, they cannot control for differences in practice that precede study enrollment or treatments that follow study discontinuation, both of which could influence overall survival (OS). Moreover, the number of enrolled participants for any individual country or region may be small, and outcomes may show considerable variability due to sampling or chance alone.
Despite the growing reliance of cancer trials on international recruitment, no studies quantified how regional differences are reported and how often they are observed. Herein, we examine FDA approvals for solid tumor malignancies based on multi-arm comparative studies, analyze regional differences in outcomes, and provide a question guide to help clinicians determine whether regional differences are likely to be genuine.
Methods
Study Selection and Data Collection
We searched the FDA archives (Drugs@FDA)2 to identify trials supporting drug approvals for solid organ malignancies (excluding lymphoma) between January 2010 and December 2020. For each trial we extracted tumor site, year of approval, number of patients, number of events, and the hazard ratio (HR) for the outcome supporting approval, stratified by region if available. The class of drug was grouped into immunotherapy, chemotherapy, monoclonal antibodies (including antibody drug conjugates), small-molecule targeted agents (including tyrosine kinase inhibitors [TKIs], PARP inhibitors, CDK4/6 inhibitors, mTOR inhibitors), hormone therapy, and other (including combination treatments). We excluded noncomparative studies (single-arm and noncomparative multi-arm studies). Only time-to-event (TTE) endpoints were included. We searched the protocol and/or statistical appendices to determine whether regional analyses were preplanned, and whether country groupings were prespecified.
Data Synthesis and Statistical Analysis
Studies were grouped according to whether a regional subgroup analysis was presented. For parametric data, the means for continuous data were compared using a Student’s t-test, and dichotomous and categorical data were compared using χ2 test or odds ratio (OR). For nonparametric data, the Kruskal-Wallis test was used to compare the median between groups. Changes in the number of participating countries and the reporting of regional variations over time were analyzed using linear regression. For studies presenting regional analyses, we extracted the test for heterogeneity. If not presented, we used Revman 5.4 (The Cochrane Collaboration) to calculate the Cochran’s Q test for heterogeneity using the Deeks method3 and estimated the number of studies where the test for heterogeneity was P<.10. A P value of ≤.1 was chosen by convention, due to the recognized low power of the Deeks test for heterogeneity.3
Based on data regarding country participation extracted from the appendices, regions were grouped into high-income versus mixed-income (including regions with high [HIC], upper-middle [UMIC] and low-middle [LMIC], and low [LIC] income based on World Bank Income Group definitions). We performed a random effects meta-analysis by region followed by a pairwise comparison to determine whether outcomes differed between high-income versus mixed-income regions.
Statistical analyses were performed using STATA, version 12 (StataCorp LP), Revman 5.4, and WINBugs. Statistical significance was defined as P<.05. No corrections were applied for multiple significance testing.
Results
Between January 2010 and December 2020, we identified 229 studies in patients with solid tumors supporting FDA approvals. We excluded 82 studies (63 single-arm, 1 abstract only, 2 pediatric tumors, 4 biosimilar or adjusted-dose studies, 10 no TTE, 2 single-country studies), resulting in an analytic cohort of 147 studies comprising 93,226 patients (supplemental eFigure 1, available with this article at JNCCN.org).
The number of participating countries has increased over time (β = 0.5; P=.04). Among the included studies, 78 (53%) included a regional subgroup analysis and there was a significant increase in the proportion of studies presenting regional analyses over time (β = 0.23; P=.03). Studies with progression-free survival (PFS) as the primary endpoint were less likely to present regional analyses compared with studies with OS as the primary endpoint (OR, 0.36; 95% CI, 0.17–0.76; P=.006). The most common drugs studied in this cohort were small molecules (n=68; 43%). Compared with small-molecule studies, immunotherapy studies were more likely to present regional data (OR, 3.00; 95% CI, 1.25–7.16; P=.009). The most common individual tumor types were lung (n=33; 22%) and breast (n=23; 16%). The mean number of participating patients, sites, and countries were similar between studies that presented regional analyses and those that did not (Table 1).
Baseline Trial Characteristics (N=147)
Among all 147 studies, country participation was available for 109 (74%). The median proportion of participating countries was 81.2% (range, 44%–100%) from HICs, 17% (range, 0%–44%) from UMICs, 0% (range, 0%–22%) from LMICs, and there were no participating LICs. Over time, there have been no significant differences in the proportion of participating HICs (β = 0.0003; P=.94), UMICs (β = 0.001; P=.74), or LMICs (β = −0.001; P=.22).
Among the 78 studies presenting regional analyses, 56 (72%) had a protocol available for review; of these, 38 (68%) specified a regional analysis, whereas 18 (32%) did not despite presenting these results in the primary publication. In 3 studies (5%), the protocol-specified and published regional groupings were inconsistent. Among the 69 studies without a regional analysis, 16 (38%) specified one in their protocol. The endpoints for which regional subgroups were available included OS (n=46), PFS (n=45), and other (1 invasive disease–free survival, 1 metastasis-free survival). No studies provided a subgroup analysis stratified by country-level income.
Among the 46 studies presenting a regional analysis for OS, 6 presented absolute differences in median OS by region, whereas the remaining presented only relative differences (as represented by the HR). One study with only graphical data was excluded from further analysis. Among the remaining 45 studies, 3 provided a test for interaction, 1 presented a P value for regional subgroups from a multivariate cox regression, and the test for heterogeneity was calculated in the remaining 41. In 17.8% (8/45) of studies, the test for heterogeneity P value was ≤.1, suggesting a significant difference in OS HR between regions (Table 2). There were no baseline trial factors that predicted for regional variation in OS (supplemental eTable 1).
Studies With Potential Regional Heterogeneity
Among the 46 studies presenting regional analyses for PFS, 5 presented the absolute differences in median PFS by region, whereas the remaining presented only relative differences. Two studies with only graphical data were excluded from further analysis. Among the remaining 44 studies, 3 presented the test for interaction, whereas this was calculated for 41. In 18% (8/44), the P value for the test for heterogeneity was ≤.10, whereas in 9% (4/44) it was ≤.05, suggesting a significant difference in PFS HR between regions (Table 2). There were no baseline trial factors that predicted for regional variation in PFS (supplemental eTable 1).
There were small variations in the reported and calculated P values for heterogeneity, resulting from differences in calculation methods. There were no instances in which the reported P value was significant and the calculated value was nonsignificant. There was one instance4 in which the calculated P value for heterogeneity was significant (P=.02) while the reported P value for interaction was borderline significant (P=.051).
There was heterogeneity in the regions and country groupings used across studies, limiting pooled analysis. After grouping regions by income level, we found no difference in OS outcomes in high-income regions compared with mixed-income regions (HR, 0.95; 95% CI, 0.84–1.07). However, these data were only available for a minority of the included studies (n=20).
Discussion
Over time, FDA approvals in solid tumor malignancies have been increasingly supported by multicenter and multinational studies. We demonstrate that only 53% of studies presented a regional subgroup analysis, and only 72% of those presenting a regional variability prespecified the subgroup analysis in their protocol. We also found that participation in registration studies in oncology is dominated by HICs, with limited participation from UMICs and LMICs, and no participation from LICs. These findings further support recent work demonstrating that contemporary studies in oncology are predominantly led by researchers in HIC.5
Regional subgroup analyses should be interpreted with caution. With any subgroup analysis, false-positives may occur due to multiple comparisons and false-negatives due to inadequate power.6,7 When subgroups are analyzed without being defined a priori, as was done in 28% of the studies included in this cohort, statistically significant results may be identified by chance. Moreover, there is a risk of maldistribution if studies are not stratified according to prespecified regions. Prior research has demonstrated that if a study has 5 regions and is designed with 80% power to detect a difference between 2 arms, the probability of identifying 1 region that favors the control by chance alone is 50%.8 Despite these limitations, oncology trials are becoming increasingly international and understanding regional variations is of growing importance. The CONSORT diagram specifies that analyses from important subgroups should be prespecified and that such prespecified analyses should be distinguished from exploratory analyses.9 We urge clinicians and journal editors to standardize regional reporting using established regional definitions such as the WHO Regions or World Bank Income Groups, and to prespecify such analyses in their trial protocols.
Even within single-country studies, population-level heterogeneity in outcomes may be detected due to environmental, social, economic, racial, and cultural differences.10 However, the degree of population-level heterogeneity is likely to be even greater between countries where populations do not share common regulatory and health systems. The potential associations between region and cancer outcomes can be considered within the socioecological framework for health, and include the impact of region on the individual, society, environment, community, and cancer health policy.11
The challenge of interpreting regional variation in international trials is not unique to oncology. In cardiology, the international PLATO trial of ticagrelor in acute coronary syndrome12 and the MERIT study of metoprolol for heart failure13 found regional differences in efficacy between US and non-US populations, creating controversy regarding the FDA approval of these drugs. International outcome variations have also been observed in stroke, where an analysis of 3,284 patients across multiple trials and countries found regional variations in index stroke severity, outcome, and mortality that were not explained by case mix alone.14
When regional or country-level differences are detected in oncology, clinicians must evaluate the available data closely to determine whether the differences are genuine or statistical anomalies. This is particularly important in the modern era in which randomized controlled trials are often not repeated, and false-positives will inevitably occur. Several questions to help guide the clinician in this exercise have been proposed previously.15 Here we present an adapted list (Table 3) with examples unique to trials in oncology.
Questions to Guide Clinicians in Determining Whether Regional Variations in Oncology Study Results Are Real
How Have the Regions or Countries Been Grouped, Were They Prespecified, and Does the Grouping Make Biologic and Clinical Sense?
Only 68% of included studies prespecified a regional analysis. On 3 occasions the presented regional grouping in the protocol and publication did not match. Plausible groupings include World Bank Income Groups, geographic proximity, or countries with cultural or ethnic similarities. On the other hand, grouping Australia and New Zealand with LMICs such as Thailand and Vietnam may be hard to justify and interpret. Although it may be tempting to group dissimilar countries together to generate subgroups with similar numbers, such groupings of convenience should be discouraged. Studies should plan to recruit sufficient patients from each region to avoid the potential for data maldistribution, and the numbers of patients in each region should be clearly reported.
Are the Observed Differences in Treatment Effects on a Relative or Absolute Scale?
Absolute differences in treatment efficacy are more likely to demonstrate regional variation,8 because of differences in underlying risk. For example, a screening intervention may decrease the absolute risk of cervical cancer from 15% to 5% in LICs, compared with 3% to 1% in HICs. In both settings, the relative risk reduction is 66%; however, the absolute benefits vary due to the underlying risk (10% absolute benefit in LICs vs 2% in HICs). In oncology, most therapeutic studies are presented in relative terms, and we assume the relative benefits are constant between regions. Although regional differences in the absolute benefits (eg, median OS or PFS) may exist, only a minority of studies presented the absolute differences between regions. Regional differences in absolute benefit could influence decisions on clinical utility and reimbursement, and their reporting should be encouraged. The CONSORT checklist recommends presenting both absolute and relative effect sizes for binary outcomes.9 We would suggest that these guidelines be updated to also recommend presenting absolute and relative effect sizes for continuous outcomes such as median OS and PFS.
What is the Plausibility and Consistency of the Observed Regional Difference?
In 1965, Bradford Hill outlined criteria to help establish causal inference.16 However, the analysis of regional variation is not attempting to establish causality between region and outcome, but rather whether region is an effect modifier in the association between treatment and outcome. Not all of the Bradford Hill criteria are applicable directly to the question of regional variation; however, the questions of plausibility and consistency are of particular interest and should be considered when assessing regional variations.16,17 Examples of plausible biologic explanations for regional variability include genetic variability, such as EGFR mutations in East-Asian patients with non–small cell lung cancer, or ethnic differences in drug metabolism of 5-fluorouracil.18
Consistency of regional variability between similar studies or tumor types should prompt further consideration. The SHARP trial in hepatocellular carcinoma (HCC) identified regional differences in efficacy,19 which was confirmed in real-world studies across 39 countries.20 Regional differences were also seen in the REFLECT21 and RESORCE22 studies in HCC. These regional differences are attributed to differences in the clinical management and underlying etiology of HCC.23 In contrast, the MONARCH 3 study found regional differences in PFS for patients with metastatic breast cancer treated with CDK4/6 inhibitors, whereas Asian patients seemed to derive greater relative PFS benefit (HR, 0.34; 95% CI, 0.2–0.53) compared with North American (HR, 0.76; 95% CI, 0.42–1.38) and European patients (HR, 0.64; 95% CI, 0.45–0.896).24 However, regional differences were not observed for OS outcomes in other studies of first-line CDK4/6 inhibitors, including MONALEESA-7,25 MONALEESA-3,5 or PALOMA-2,26 raising questions about whether this regional variation was real.
What Is the Magnitude of Effect of the Regional Difference and Is It Statistically Significant?
The magnitude of effect and strength of the regional variability should also be considered. The statistical test for heterogeneity can examine whether treatment effect is modified by geographic location, with a P value <.10 suggesting true variation. However, the test for heterogeneity has low statistical power27 and should be interpreted cautiously, especially in the context of multiple subgroup analyses or if the subgroup is not defined a priori. Furthermore, a nonsignificant result should not be used as proof of lack of regional variation. Large differences in HR between regions, even with widely overlapping confidence intervals and a nonsignificant P value, may still warrant further investigation. Only 7 included endpoints (3 PFS and 4 OS endpoints) presented a test for heterogeneity or interaction for the regional subgroup analyses. After calculating the test for heterogeneity for the remaining studies, we found regional differences in approximately 18% of OS and PFS endpoints.
Is There a Clinical Explanation for the Observed Regional Differences?
Although international trials make substantial efforts to standardize populations enrolled and the treatments delivered on study, regional differences may still exist. Regional differences in competing causes of death may result in lower absolute OS in one region compared with another. Differences in competing events are particularly important for adjuvant studies with long follow-up. For example, among patients with breast cancer enrolled in the TAILORx study, almost half of the primary endpoint events were unrelated to breast cancer, and were instead due to second primary cancers or death from other causes.28 The rate of other events could vary significantly between countries with differing health resources and economic status; however, this may not necessarily impact on the relative benefits of a therapy. Transparent reporting of competing events should be mandated, and competing events analysis performed where indicated, particularly in instances where regional variation has been detected.
Differences in the availability of subsequent therapies could also drive regional differences. Unlike PFS, OS is influenced by the investigational treatment and subsequent therapies. Unfortunately, regional differences in subsequent treatments were only reported in one study in our cohort (REFLECT).21 In the lenvatinib arm, 50.5% of patients in the Asia-Pacific region received further anticancer therapy after study compared with only 28% in the Western region. Among those randomly assigned to sorafenib, rates of subsequent anticancer therapy use were 53.9% in the Asia-Pacific compared with 45.2% in the Western region. We urge journals to mandate the reporting of subsequent treatments stratified by region, regardless of whether regional differences in outcome are detected.
Limitations
Given that our study focused on oncology trials leading to FDA drug approval, our findings regarding the reporting and prevalence of regional variation may not be generalized to nonregistration trials. Furthermore, the majority of included studies were phase III. Smaller phase II studies are less likely to be multinational, and generally have smaller samples sizes with fewer participating countries.
Most studies did not present data regarding the number of enrolled participants for each country. Therefore, the proportion of participating countries by income group may not reflect actual patient recruitment. However, it remains the best available estimate regarding participation of LMICs in registration studies in oncology.
There was significant heterogeneity in the definition of regions used across all included studies. Some studies used preexisting regional groupings such as the WHO regions, whereas others generated their own regional groupings. This limited our ability to perform pooled analyses categorized by region. Conversely, as described earlier, results may be difficult to interpret when grouping different racial and ethnic populations together, or countries with varying health resources and income levels, and may warrant additional subgroup analyses.
A small proportion of studies reported the test for interaction or heterogeneity in their analysis. There was minor variability in the reported and calculated P values due to differences in the calculation methods. However, these differences did not impact the number of studies with regional variation identified.
Conclusions
Over time, studies in oncology have become larger, with more participating countries and regions. Regional outcome variations occur in <20% of trials. As studies become more international, the oncology community needs to develop a strategy for reporting and interpreting regional variation. The question guide herein may help clinicians interpret regional variations in trials and determine whether they are likely to be clinically meaningful. Postregistration and real-world studies should be used to confirm benefit, particularly for regions where the primary registration study showed variability.
References
- 1.↑
Arnold M, Rutherford MJ, Bardot A, et al. Progress in cancer survival, mortality, and incidence in seven high-income countries 1995-2014 (ICBP SURVMARK-2): a population-based study. Lancet Oncol 2019;20:1493–1505.
- 2.↑
U.S. Food and Drug Administration. Drugs@FDA Database. Accessed April 15, 2020. Available at: https://www.accessdata.fda.gov/scripts/cder/daf/
- 3.↑
Deeks JJ, Higgins JPT, Altman DG, et al. Analysing data and undertaking meta‐analyses. In: Higgins JPT, Thomas J, Chandler J, et al., eds. Cochrane Handbook for Systematic Reviews of Interventions. Hoboken, NJ: Wiley Blackwell; 2019:241–284.
- 4.↑
Saura C, Oliveira M, Feng YH, et al. Neratinib plus capecitabine versus lapatinib plus capecitabine in HER2-positive metastatic breast cancer previously treated with≥ 2 HER2-directed regimens: phase III NALA trial. J Clin Oncol 2020;38:3138–3149.
- 5.↑
Slamon DJ, Neven P, Chia S, et al. Overall survival with ribociclib plus fulvestrant in advanced breast cancer. N Engl J Med 2020;382:514–524.
- 6.↑
Burke JF, Sussman JB, Kent DM, et al. Three simple rules to ensure reasonably credible subgroup analyses. BMJ 2015;351:h5651.
- 7.↑
Rothwell PM. Treating individuals 2. Subgroup analysis in randomised controlled trials: importance, indications, and interpretation. Lancet 2005;365:176–186.
- 8.↑
Marschner IC. Regional differences in multinational clinical trials: anticipating chance variation. Clin Trials 2010;7:147–156.
- 9.↑
Schulz KF, Altman DG, Moher D. CONSORT 2010 statement: updated guidelines for reporting parallel group randomized trials. Ann Intern Med 2010;152:726–732.
- 10.↑
Moore JX, Royston KJ, Langston ME, et al. Mapping hot spots of breast cancer mortality in the United States: place matters for Blacks and Hispanics. Cancer Causes Control 2018;29:737–750.
- 11.↑
Palafox NA, Reichhardt M, Taitano JR, et al. A socio-ecological framework for cancer control in the Pacific: a community case study of the US affiliated Pacific Island jurisdictions 1997–2017. Front Public Health 2018;6:313.
- 12.↑
Mahaffey KW, Wojdyla DM, Carroll K, et al. Ticagrelor compared with clopidogrel by geographic region in the Platelet Inhibition and Patient Outcomes (PLATO) trial. Circulation 2011;124:544–554.
- 13.↑
Wedel H, Demets D, Deedwania P, et al. Challenges of subgroup analyses in multinational clinical trials: experiences from the MERIT-HF trial. Am Heart J 2001;142:502–511.
- 14.↑
Ali M, Atula S, Bath PM, et al. Stroke outcome in clinical trial patients deriving from different countries. Stroke 2009;40:35–40.
- 15.↑
Yusuf S, Wittes J. Interpreting geographic variations in results of randomized, controlled trials. N Engl J Med 2016;375:2263–2271.
- 16.↑
Hill AB. The environment and disease: association or causation? Proc R Soc Med 1965;58:295–300.
- 17.↑
Fedak KM, Bernal A, Capshaw ZA, et al. Applying the Bradford Hill criteria in the 21st century: how data integration has changed causal inference in molecular epidemiology. Emerg Themes Epidemiol 2015;12:14.
- 18.↑
Loh M, Chua D, Yao Y, et al. Can population differences in chemotherapy outcomes be inferred from differences in pharmacogenetic frequencies? Pharmacogenomics J 2013;13:423–429.
- 19.↑
Cheng AL, Kang YK, Chen Z, et al. Efficacy and safety of sorafenib in patients in the Asia-Pacific region with advanced hepatocellular carcinoma: a phase III randomised, double-blind, placebo-controlled trial. Lancet Oncol 2009;10:25–34.
- 20.↑
Kudo M, Lencioni R, Marrero JA, et al. Regional differences in sorafenib-treated patients with hepatocellular carcinoma: GIDEON observational study. Liver Int 2016;36:1196–1205.
- 21.↑
Kudo M, Finn RS, Qin S, et al. Lenvatinib versus sorafenib in first-line treatment of patients with unresectable hepatocellular carcinoma: a randomised phase 3 non-inferiority trial. Lancet 2018;391:1163–1173.
- 22.↑
Bruix J, Qin S, Merle P, et al. Regorafenib for patients with hepatocellular carcinoma who progressed on sorafenib treatment (RESORCE): a randomised, double-blind, placebo-controlled, phase 3 trial. Lancet 2017;389:56–66.
- 23.↑
Yang JD, Hainaut P, Gores GJ, et al. A global view of hepatocellular carcinoma: trends, risk, prevention and management. Nat Rev Gastroenterol Hepatol 2019;16:589–604.
- 24.↑
Johnston S, Martin M, Di Leo A, et al. MONARCH 3 final PFS: a randomized study of abemaciclib as initial therapy for advanced breast cancer. NPJ Breast Cancer 2019;5:5.
- 25.↑
Im SA, Lu YS, Bardia A, et al. Overall survival with ribociclib plus endocrine therapy in breast cancer. N Engl J Med 2019;381:307–316.
- 26.↑
Finn RS, Martin M, Rugo HS, et al. Palbociclib and letrozole in advanced breast cancer. N Engl J Med 2016;375:1925–1936.
- 27.↑
Hardy RJ, Thompson SG. Detecting and describing heterogeneity in meta-analysis. Stat Med 1998;17:841–856.
- 28.↑
Sparano JA, Gray RJ, Makower DF, et al. Adjuvant chemotherapy guided by a 21-gene expression assay in breast cancer. N Engl J Med 2018;379:111–121.
- 29.
González-Martín A, Pothuri B, Vergote I, et al. Niraparib in patients with newly diagnosed advanced ovarian cancer. New Engl J Med 2019;381:2391–2402.
- 30.
Goetz MP, Toi M, Campone M, et al. MONARCH 3: abemaciclib as initial therapy for advanced breast cancer. J Clin Oncol 2017;35:3638–3646.
- 31.
Fizazi K, Tran NP, Fein L, et al. Abiraterone acetate plus prednisone in patients with newly diagnosed high-risk metastatic castration-sensitive prostate cancer (LATITUDE): final overall survival analysis of a randomised, double-blind, phase 3 trial. Lancet Oncol 2019;20:686–700.
- 32.
Borghaei H, Paz-Ares L, Horn L, et al. Nivolumab versus docetaxel in advanced nonsquamous non–small-cell lung cancer. N Engl J Med 2015;373:1627–1639.
- 33.
Mayer RJ, Van Cutsem E, Falcone A, et al. Randomized trial of TAS-102 for refractory metastatic colorectal cancer. N Engl J Med 2015;372:1909–1919.
- 34.
Rugo HS, Im SA, Cardoso F, et al. Efficacy of margetuximab vs trastuzumab in patients with pretreated ERBB2-positive advanced breast cancer: a phase 3 randomized clinical trial. JAMA Oncol 2021;7:573–584.
- 35.
Kopetz S, Grothey A, Yaeger R, et al. Encorafenib, binimetinib, and cetuximab in BRAF V600E–mutated colorectal cancer. N Engl J Med 2019;381:1632–1643.
- 36.
Burtness B, Harrington KJ, Greil R, et al. Pembrolizumab alone or with chemotherapy versus cetuximab with chemotherapy for recurrent or metastatic squamous cell carcinoma of the head and neck (KEYNOTE-048): a randomised, open-label, phase 3 study. Lancet 2019;394:1915–1928.
- 37.
Wilke H, Muro K, Van Cutsem E, et al. Ramucirumab plus paclitaxel versus placebo plus paclitaxel in patients with previously treated advanced gastric or gastro-oesophageal junction adenocarcinoma (RAINBOW): a double-blind, randomised phase 3 trial. Lancet Oncol 2014;15:1224–1235.
- 38.
Brahmer J, Reckamp KL, Baas P, et al. Nivolumab versus docetaxel in advanced squamous-cell non–small-cell lung cancer. N Engl J Med 2015;373:123–135.
- 39.
Schöffski P, Chawla S, Maki RG, et al. Eribulin versus dacarbazine in previously treated patients with advanced liposarcoma or leiomyosarcoma: a randomised, open-label, multicentre, phase 3 trial. Lancet 2016;387:1629–1637.
- 40.
Pujade-Lauraine E, Ledermann JA, Selle F, et al. Olaparib tablets as maintenance therapy in patients with platinum-sensitive, relapsed ovarian cancer and a BRCA1/2 mutation (SOLO2/ENGOT-Ov21): a double-blind, randomised, placebo-controlled, phase 3 trial. Lancet Oncol 2017;18:1274–1284.
- 41.
Kojima T, Shah MA, Muro K, et al. Randomized phase III KEYNOTE-181 study of pembrolizumab versus chemotherapy in advanced esophageal cancer. J Clin Oncol 2020;38:4138–4148.
