{
  "abstract": "Objective Low-middle-income countries (LMICs) have increased their participation in international oncology trials. However, considerable disparities in treatment standards across countries have raised ethical concerns regarding the use of control arms that may not align with the established standards of care in high-income countries. This trial aims to describe the control arms of randomised oncology trials recruiting in Brazil, an LMIC where the majority of patients receive care through the public health system, the Unified Health System (SUS), which provides limited access to cancer treatments.Methods and analysis This cross-sectional study included randomised clinical trials recruiting in Brazil on 4 December 2023 (ClinicalTrials.gov). Abstracted data included sample size, sponsor, tumour site, study phase and control arm. Two independent investigators classified control arms as superior, equal or inferior based on National Comprehensive Cancer Network (NCCN), Brazilian private insurance and SUS standards. Data were summarised in means, medians and proportions. Fisher’s exact test compared categories. A p<0.05 was considered statistically significant.Results A total of 98 studies were included. The median intended sample size was 555 patients (54–6000). Most studies were phase 3 (84.7%) and pharma-sponsored (97%). Lung (29.6%) and breast (24.4%) were the most commonly studied tumour sites. Regarding treatment setting, 23 studies (23.5%) were (neo)adjuvant trials, 48 (49.0%) first-line and 27 (27.5%) second-line or later. Overall, 80 (81.7%), 82 (83.7%) and 58 studies (59.1%) employed control arms considered equivalent to the standards of NCCN, private insurance and SUS, respectively. 18 studies (18.3%) had a suboptimal control arm according to NCCN guidelines, while 16 studies (16.3%) according to Brazilian private insurance. No studies used control arms inferior to SUS standards. Of the 18 control arms inferior to NCCN, 3 were superior and 15 were equal to standard of care offered by SUS. No studies had their control arms superior to NCCN or private insurance; whereas, 40 (40.9%) were superior to SUS.Conclusion A significant number of studies employed control arms inferior to NCCN guidelines; however, these were considered superior or equal to the standards offered by SUS. Such discrepancies may hinder the appropriate interpretation of study findings.",
  "authors": [
    {
      "affiliations": [
        "Division of Medical Oncology, Hospital de Base, São José do Rio Preto, SP, Brazil"
      ],
      "name": "Ana Elisa Boracini Sanches"
    },
    {
      "affiliations": [
        "Division of Medical Oncology, Hospital de Base, São José do Rio Preto, SP, Brazil"
      ],
      "name": "Luiza Aleixo Barros Leite Fadul"
    },
    {
      "affiliations": [
        "Department of Molecular Biology, Faculdade de Medicina de São José do Rio Preto, São José do Rio Preto, SP, Brazil"
      ],
      "name": "Debora Aparecida Pires de Campos Zuccari"
    },
    {
      "affiliations": [
        "Division of Medical Oncology, Hospital de Base, São José do Rio Preto, SP, Brazil"
      ],
      "name": "Beatriz de Menezes Dobbert"
    },
    {
      "affiliations": [
        "Division of Medical Oncology, Hospital de Base, São José do Rio Preto, SP, Brazil"
      ],
      "name": "Lorena Forner"
    },
    {
      "affiliations": [
        "Division of Medical Oncology, Hospital de Base, São José do Rio Preto, SP, Brazil"
      ],
      "name": "Júlia Belone Lopes"
    },
    {
      "affiliations": [
        "Division of Hematology and Medical Oncology, University of Florida College of Medicine, Gainesville, Florida, USA"
      ],
      "name": "Daniel Vilarim Araujo"
    }
  ],
  "full_text": "WHAT IS ALREADY KNOWN ON THIS TOPIC The recent decentralisation of global clinical research, marked by the growing participation of low-income and middle-income countries, has contributed to advancements in local research infrastructure and expanded access to innovative treatments for patients enrolled in clinical trials. However, the same phenomenon raises concerns about the selection of control arms and its implications for the interpretation of study results.WHAT THIS STUDY ADDS We observed a significant number of ongoing multicentre trials in Brazil with suboptimal control arms compared with global standards, despite being appropriate according to national guidelines.HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE, OR POLICY Greater regulatory oversight is necessary to define appropriate control arms, ensuring that study outcomes effectively address global scientific questions.Introduction Cancer incidence and mortality continue to rise, positioning the disease as one of the leading causes of death worldwide. In 2020, cancer was responsible for nearly 10 million deaths globally, with approximately 7 million of these occurring in low-income and middle-income countries (LMICs). 1 2 While cancer incidence is relatively similar across regions, LMICs experience disproportionately higher mortality rates.3 This disparity is largely attributed to socioeconomic factors limiting health promotion, early detection and access to modern oncologic treatments.4Over the past 30 years, the global clinical research landscape has undergone significant decentralisation, marked by the increasing participation of LMICs in clinical trials. Between 2007 and 2016, the number of new trials conducted in high-income countries (HICs) rose by 25% (from 3399 to 4266); whereas, LMICs experienced a 400% increase in the same period (from 363 to 1389).5–7 This globalisation of research facilitates patient recruitment and potentially enhances the generalisability of trial results across diverse populations and healthcare settings.8However, this expansion also underscores the critical importance of control arm selection. Control arms—typically reflecting the local standard of care—serve as the benchmark for evaluating experimental interventions. Yet, standards of care vary significantly across countries, influenced by disparities in healthcare infrastructure, access to health technologies and regulatory approval processes.9 In LMICs, limited access to innovative treatments often results in control arms that, while aligned with local standards, fall short of international benchmarks such as those established by the National Comprehensive Cancer Network (NCCN).This variability raises ethical and scientific concerns regarding the use of ‘suboptimal control arms,’ wherein the comparator arm provides treatment inferior to the global standard. Such practices can create a false perception of benefit for the experimental intervention, compromising the principle of equipoise and potentially leading to the approval of therapies without adequately addressing their optimal sequencing or true clinical value.10 11 Consequently, the interpretation and applicability of trial results become problematic.Several oncology randomised clinical trials (RCTs) illustrate this issue. For instance, the EMPOWER-Lung 1 and CheckMate 9LA trials, both initiated in 2017, evaluated immune checkpoint inhibitors (ICIs) versus chemotherapy in the first-line treatment of non-small cell lung cancer (NSCLC).12 13 However, the superiority of pembrolizumab over chemotherapy in this setting had already been demonstrated by the Keynote-024 and Keynote-042 trials in 2016.14 15 Besides these, as of December 2021, there were an additional 15 active NSCLC trials that still used chemotherapy as the control arm, with 6 of them not allowing the use of immunotherapy, even in subsequent treatment lines.16 Such practices hinder the interpretation of study results and the appropriate use of the tested medications.17 18In Brazil, classified by the World Bank as a middle-income country, participation in global clinical trials has provided important opportunities for patients to access innovative treatments otherwise unavailable in the public health system. However, this benefit is limited to a small fraction of the population.19 Approximately 74% of Brazilians depend exclusively on the Unified Health System (Sistema Único de Saúde – SUS), which faces chronic underfunding, restricted access to high-cost therapies and significant delays in incorporating new technologies. Consequently, the standard of care offered in SUS often lags behind international guidelines, raising ethical and methodological concerns when trials conducted in Brazil use local standards as control arms. In such scenarios, control arms that are suboptimal by international benchmarks may be justified by local practice but risk producing results that overestimate the benefit of experimental treatments, with limited relevance to global clinical decision-making.20This study aims to characterise the control arms of randomised oncology trials conducted in Brazil, assessing their alignment with treatment standards established by SUS, Brazilian private health insurance and NCCN guidelines. Additionally, we will describe the demographic and clinical characteristics of these trials and identify factors associated with the use of suboptimal control arms.Materials and Methods Study design and research strategy This is a cross-sectional study that analysed RCTs recruiting patients in Brazil for the treatment of solid tumours. The search was conducted on 4 December 2023, using the ClinicalTrials database ( https://clinicaltrials.gov) with the following prompt: Cancer or Cancers, Neoplasm or Neoplasms, malignancy, melanoma or melanomas, or sarcoma or sarcomas.Since the study was based exclusively on publicly available data and did not involve direct research with humans, approval by the institutional ethics committee or informed consent was not required.Patient and Public Involvement Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.Inclusion and exclusion criteria Regarding inclusion criteria, only randomised clinical trials, studies with interventions aimed at the treatment of solid tumours and those exclusively recruiting adult patients (18 years or older) were included. The trials needed to be actively recruiting in Brazil as of the search date and include a control group that allowed for comparative analysis of treatment standards.Studies were excluded if they did not employ a randomisation process, such as single-arm trials or those lacking a clearly defined control group; targeted haematological malignancies or paediatric populations; involved local therapies (eg, radiotherapy or surgery) as part of the intervention; or focused on non-cancer-modifying treatments, including symptom management and supportive care.Data abstraction For each trial, the following data were extracted: ClinicalTrials.gov identifier, study phase, title, sponsor, estimated sample size and whether the trial was conducted as a multicentre or single-centre study. Additionally, we recorded whether other LMICs were included in the recruitment process, details of the intervention and control arm treatments, tumour type, International Classification of Diseases (ICD-10) code, primary outcomes and treatment line (neoadjuvant/adjuvant, first-line, second-line or subsequent lines).Characterization of control arms We extracted data of the control arm of each study and categorised its adequacy according to the treatment standards of SUS, Brazilian private health insurance and NCCN. For SUS, classification was guided by both the Diagnostic and Therapeutic Guidelines (Diretrizes Diagnósticas e Terapêuticas, DDT) from the Brazilian Ministry of Health 21 and the Oncology Conduct Manual (third edition) from the Cancer Institute of the State of São Paulo (ICESP).22 In cases of discrepancy between these two sources, we adopted a best-case approach, selecting the guideline recommending the most effective available treatment. For private healthcare standards, we referred to the official package inserts for injectable drugs and the National Supplementary Health Agency’s (ANS) List of Procedures and Health Events for oral therapies.23 24 NCCN guidelines were accessed through their official website, with the January 2024 versions used as the reference. The NCCN was selected as a reference because of its global recognition in oncology care and its regularly updated recommendations.25Each control arm was categorised as superior, equivalent or inferior relative to the respective standard. To ensure data reliability, two independent researchers conducted data extraction and classification, with discrepancies resolved by a third reviewer.Statistical analysis Demographic data were summarised in means, medians and proportions. χ² and Fisher exact tests were used to assess differences between groups. P<0.05 was considered statistically significant. Analyses were performed using jamovi cloud.Results Study Characteristics A total of 277 studies were initially identified through the ClinicalTrials.gov platform. Following a detailed screening process, 98 studies met the inclusion criteria and were included in the final analysis. Excluded studies comprised 67 (24%) non-randomised trials, 36 (12%) haematologic malignancy trials, 32 (11%) studies evaluating non-disease-modifying treatments, 18 (6%) trials involving radiotherapy and 14 (5%) with other local therapy ( figure 1).Figure 1Flowchart of clinical trials selection.The planned sample sizes of included studies varied widely, ranging from 54 to 6000 participants, with a median of 555 patients per study. Most studies (n=94; 95.9%) had a multicentre design, while only four (4.1%) were single-centre trials. All multicentre studies also recruited participants from other LMICs.Regarding study phase, the majority were phase 3 trials (84.7%), followed by phase 2 (12.2%) and phase 1 (3.1%) studies. Industry sponsorship predominated, accounting for 97% of studies, with only 3% funded by academic or governmental sources.In terms of treatment setting, 24.2% of trials were conducted in the neoadjuvant or adjuvant context, 48.5% in the first-line metastatic setting and 27.3% in second-line or later lines of treatment. Lung cancer (29.6%) and breast cancer (24.4%) were the most frequently studied tumour types, followed by colorectal cancer, gastric cancer and melanoma (each representing 6.1%). Other tumour types collectively accounted for 27.7% of the included studies (table 1).Table 1Characteristics of clinical trialsTotalN=98Number of patients (range)555 (54–6000)Phase (n/%) 13 (3.1%) 212 (12.2%) 383 (84.7%)Tumour type Lung29 (29.6%) Breast24 (24.4%) Colorectal6 (6.1%) Gastric6 (6.1%) Melanoma6 (6.1%) Others28 (27.7%)Line of treatment (Neo)adjuvant23 (23.5%) First line48 (49%) Second or higher27 (27.5%)Sponsor Pharma95 (97%) Other3 (3%)Multicentric Yes94 (96%) No4 (4%)Control Arms Characteristics The control arms of the included studies were evaluated against three reference standards: SUS, private health insurance in Brazil and NCCN guidelines. When compared with the NCCN, 81.7% (80 studies) presented control arms considered equivalent to the established standard, while 18.3% (18 studies) were classified as inferior. No study had their control arm classified as superior to the NCCN standard. A similar pattern was observed when comparing to the standards of Brazilian private health insurance, with 83.7% (82 studies) considered equivalent to the standard, and 16.3% (16 studies) classified as inferior, with no study evaluated as superior to this standard. In contrast, comparison with SUS standards revealed greater variability. 58 studies (59.1%) had control arms considered equivalent to the SUS standard (considering the best-case scenario), while the remaining 40 studies (40.9%) were classified as superior. Notably, no study was categorised as inferior to the SUS standard ( table 2).Table 2Characteristics of control armsDoes the control arm match the best possible treatment?Is the control arm?PYesNoSuperiorEqualInferiorNCCN80 (81.7%)18 (18.3%)080 (81.7%)18 (18.3%)Private insurance82 (83.7%)16 (16.3%)082 (83.7%)16 (16.3%)SUS (n=98/98) (best case)49 (50%)49 (50%)40 (40.9%)58 (59.1%)0%<0.001SUS (n=85/98) (DDT—Health ministry)45 (53%)40 (47%)39 (45.9%)46 (54.1%)0%<0.001SUS (n=95/98) (ICESP)43 (45.2%)52 (54.8%)43 (45.2%)52 (54.8%)0%<0.001P value refers to comparisons in the “is the control arm?” column—three separate comparisons, each using a different SUS version (best case, DDT, ICESP), were performed and are reported.DDT, Diretrizes Diagnósticas e Terapêuticas; ICESP, Cancer Institute of the State of São Paulo; NCCN, National Comprehensive Cancer Network; SUS, Unified Health System.Inferior Control Arms Compared to NCCN Standards Among the 18 studies with control arms classified as inferior to NCCN standards, one-third (n=6; 33.3%) involved patients with lung cancer. The primary reasons for this classification included the absence of targeted therapy in four studies and the absence of immunotherapy in two studies. Regarding treatment lines, half of the studies (n=9; 50%) were conducted in the first-line setting, followed by three studies (16.6%) in the second-line setting, other three studies (16.6%) in neoadjuvant/adjuvant setting and two studies (11.1%) in subsequent treatment lines. One of the studies (5%) included treatment from first-line to subsequent lines ( online supplemental table S1).SP110.1136/bmjonc-2025-000808.supp1Supplementary dataIn terms of sponsorship, the vast majority of these trials (n=17; 94.4%) were funded by pharmaceutical companies, while only one study (5.5%) was sponsored by a private hospital (online supplemental table S1).An important observation is that, despite their classification as inferior to NCCN standards, three of these studies were considered to offer control arms superior to the SUS standard. In two of these cases, this was attributed to the unavailability of immunotherapy for lung and bladder cancer within the Brazilian public health system. The remaining 15 studies were considered equivalent to SUS standards, suggesting that while these control arms were suboptimal compared with global benchmarks, they remained aligned with the standards of care available in the Brazilian public sector (figure 2).Figure 2Inferior control arms compared with Brazil health system.DISCUSSION In our study, the majority of the 98 evaluated trials were phase 3 studies (84.7%). This aligns with reported global trends showing an increase in the number of phase 3 trials being conducted in LMICs. A key factor driving this trend is the need for large patient populations to efficiently conduct phase 3 trials, which recruitment in LMICs helps fulfil. Additionally, the lower cost per participant in these regions compared with HICs further incentivises sponsors to conduct studies in LMICs. 26 Conversely, although phase 1 and 2 trials are also increasing in LMICs, the trend is less pronounced.26 This may be due to the greater complexity of early-phase trials, which require advanced infrastructure and expertise that may not yet be widely available in LMICs.Another important finding of our study was the high proportion of industry-sponsored clinical trials (97%), which is also consistent with global data.26 While industry sponsorship plays a crucial role in advancing clinical research, it also raises concerns about potential financial incentives and corporate interests. Pharmaceutical companies have strong financial incentives to obtain drug approvals, which can influence study design, choice of control arms and patient selection to maximise the likelihood of positive outcomes.27–29Notably, 18.3% of trials employed control arms deemed suboptimal by NCCN guidelines. Although frequent updates to oncology guidelines pose challenges—since trials are often designed years before initiation—this alone does not fully explain the continued use of suboptimal controls.30 It is possible that some trial designs strategically select less rigorous control arms to enhance positive results, thereby expediting drug approvals.31 These practices raise ethical concerns, as patients assigned to such control groups may receive treatment below the current standard of care, which can compromise both patient outcomes and the external validity of trial results.However, when contextualised within Brazil’s healthcare system, this picture shifts. Among the trials classified as suboptimal by NCCN standards, none were inferior to the standard of care provided by Brazil’s SUS. In fact, 59.1% of all analysed trials had control arms equivalent to SUS standards, while the other 40.9% offered superior options. Among the 18 studies with suboptimal NCCN control arms, three were superior and 15 were equivalent to SUS standards. This underscores a crucial point: for patients reliant on the public health system, clinical trial participation often provides access to treatments that are at least as good as, if not better than, standard care. Nonetheless, disparities within Brazil’s dual healthcare system persist. When comparing control arms to those available in the private sector, 16.3% of trials were considered inferior. This highlights the internal inequities in treatment access and emphasises the ethical complexity of conducting global trials in LMICs. While participation in such trials may be justified locally, their findings require careful contextualisation within the broader literature.Decentralising research to LMICs brings several important benefits. Most notably, although only a limited number of eligible patients can participate in trials, this often represents their only opportunity to access modern, potentially life-saving therapies that would otherwise be unavailable. Beyond patient-level advantages, clinical research also delivers essential financial and logistical support to local research centres, fostering investments in infrastructure and enhancing the training and expertise of healthcare professionals.32 33 These improvements contribute to strengthening the overall healthcare and research environment in LMICs.An additional concern is that nearly half of the trials (48.5%) evaluated first-line treatments. In LMICs, where access to subsequent lines of therapy is often limited, this raises important ethical and practical issues. Trials that introduce agents already proven effective in later treatment lines into earlier settings may place patients in the control arms at a disadvantage, especially when effective post-progression therapies are unavailable.34 35 While this dynamic may expedite drug approvals, it complicates the generalisability of results to high-income countries, where patients typically have access to comprehensive post-protocol care. Although our analysis focused exclusively on control arms, we acknowledge that limited access to subsequent therapies is a critical issue that requires further study.Despite criticisms regarding the ethics and external validity of industry-sponsored trials in LMICs, our findings indicate that most trials conducted in Brazil maintain control arms aligned with global standards. Even among those categorised as suboptimal by NCCN criteria, none compromised the standard of care provided by SUS. Given the significant benefits of trial participation—ranging from expanded treatment access to infrastructure development—we believe that opening such trials in LMICs remains justified and beneficial for patients, investigators and healthcare systems alike.However, it is important to acknowledge that these trials often fail to answer questions directly relevant to LMIC populations, as the innovative treatments under investigation may never be broadly accessible within public health systems. Conversely, their inclusion of LMIC patients with limited access to optimal post-protocol care can artificially enhance perceived treatment efficacy, complicating extrapolation of results to HIC settings.35 Therefore, regulatory agencies and funding bodies must play a central role in overseeing trial design, ensuring the adequacy of control arms and addressing access to appropriate post-progression therapies.Our study has limitations. Defining the standard of care for each trial involved some subjectivity, though this was mitigated by independent evaluations and consensus resolution. Additionally, our analysis was based on NCCN guidelines current as of January 2024; given the rapid evolution of oncology standards, these findings should be interpreted with caution. However, the consistency of our results when compared with Brazil’s private sector standards lends credibility to our conclusions. Another limitation is the lack of critical appraisal in determining whether the control arms were superior, equivalent or inferior to the standard of care, as we relied solely on guidelines. We acknowledge that guidelines, despite their widespread use, also have inherent limitations. Regarding SUS control arms, we also relied on protocols from the ICESP and the Ministry of Health, though regional variations in treatment access may not be fully captured. Additionally, since this was a cross-sectional study, we were unable to determine whether the trials were intended for regulatory registration or other purposes. We acknowledge that this distinction could have important implications for study design and interpretation. Finally, our study did not assess post-progression therapy and crossover strategies, as many of the analysed trials remain ongoing.Conclusion In conclusion, while many trials in Brazil use control arms that do not meet NCCN guideline standards, these treatments are generally equal or superior to those available through the public health system (SUS). This discrepancy reflects differences between global and local standards of care and complicates the interpretation and generalisability of study results. These findings highlight the need for enhanced regulatory oversight to ensure ethical trial design and reliable evidence to guide clinical practice.",
  "title": "Analysis of control arms in oncology randomised trials in Brazil: a cross-sectional study",
  "uid": "82111e9a-eda8-5988-b03a-7f3151bde508"
}