GlobeCitizen
知名会员
- 注册
- 2012-07-14
- 消息
- 1,133
- 荣誉分数
- 88
- 声望点数
- 158
Below is an article to be published in one of the prestigious medical journals regarding using external data (real-life data) in the design and analyzing of clinical trial data, which is directed by DI who is the vice president of PR for the company when asked by an investor for direction (as if it's something new, instead it's not):
https://www.dr-bala.net/NWBO/Lancet/Rahman_Lancet_102021.pdf
Cancer Trials and Design Principles 4Leveraging external data in the design and analysis of clinical trials in neuro-oncology
Rifaquat Rahman, Steffen Ventz, Jon McDunn, Bill Louv, Irmarie Reyes-Rivera, Mei-Yin C Polley, Fahar Merchant, Lauren E Abrey, Joshua E Allen, Laura K Aguilar, Estuardo Aguilar-Cordova, David Arons, Kirk Tanner, Stephen Bagley, Mustafa Khasraw, Timothy Cloughesy, Patrick Y Wen, Brian M Alexander*, Lorenzo Trippa*
Integration of external control data, with patient-level information, in clinical trials has the potential to accelerate the development of new treatments in neuro-oncology by contextualising single-arm studies and improving decision making (eg, early stopping decisions). Based on a series of presentations at the 2020 Clinical Trials Think Tank hosted by the Society of Neuro-Oncology, we provide an overview on the use of external control data representative of the standard of care in the design and analysis of clinical trials. High-quality patient-level records, rigorous methods, and validation analyses are necessary to effectively leverage external data. We review study designs, statistical methods, risks, and potential distortions in using external data from completed trials and real-world data, as well as data sources, data sharing models, ongoing work, and applications in glioblastoma.
IntroductionDrug development is associated with inefficiency, long timelines, and poor success rates in oncology, in which less than 10% of drug candidates are ultimately approved by the US Food and Drug Administration (FDA).1,2 As new unproven therapies emerge at an accelerated pace, there has been an increasing interest in novel approaches to clinical trial design that could improve efficiency.3,4Within neuro-oncology, the use of trial designs with potential for increased efficiency is of interest, particularly in the study of glioblastoma, a disease with an unmet need for better therapies because of its poor prognosis.5 There are several distinctive challenges in drug development for glioblastoma, such as the inability to completely resect tumours, difficulty in crossing the
blood–brain barrier, tumour heterogeneity, imaging challenges for monitoring disease course, and a unique immune environment.6,7 With few treatment advances over the past two decades, the clinical trial landscape in glioblastoma has been characterised by long development times, low patient participation, problematic surrogate outcomes, and poor decision making.8,9 Indeed, poor early phase decision making has been repeatedly high-lighted as a major problem in the development of therapeutics10 and continues to increase interest in novel clinical trial designs.Randomised controlled trials (RCTs) are the gold standard for clinical experimentation and evaluation of therapies. RCTs control for systematic bias from known and unknown confounders by randomly assigning
patients to receive an experimental therapy or standard of care, which allows for the evaluation of treatment effects. However, RCTs can be difficult to conduct in some neuro-oncology settings. A small percentage of patients with cancer participate in clinical trials,11 and RCTs can have slow patient accrual because of patient reluctance to enrol in studies with a control group, which is a pronounced problem in settings with ineffective standards of care, such as recurrent glioblastoma.12–14 Precision medicine further complicates this issue by focusing trials on biomarker-defined subgroups of patients who might benefit from targeted therapies.15 These subgroups are often small, resulting in substantial challenges to conducting RCTs with adequate sample sizes to detect treatment effects.15,16The design and implementation of clinical trials that leverage external datasets, with patient-level information on pretreatment clinical profiles and outcomes to support testing of experimental therapies and study decision
making, has attracted interest in neuro-oncology.17,18 A phase 2b trial in recurrent glioblastoma19 used a prespecified, eligibility-matched, external control group (including individual pretreatment profiles and outcomes), with data from patients with glioblastoma from major neurosurgery centres as a comparator group to evaluate MDNA55. After implementation of this trial design, investigators reported improved survival in patients receiving MDNA55 compared with the external control group.19 Several ongoing neuro-oncology trials are actively exploring similar approaches to use external data in the design and analysis of clinical studies.The Society of Neuro-Oncology hosted the 2020 Clinical Trials Think Tank on Nov 6, 2020, with a virtual session dedicated to trial designs leveraging external data. Experts in the field of neuro-oncology were paired with experts in data science and biostatistics, and representatives from industry, patient advocacy, and the FDA. The interdisciplinary session focused on challenges in drug development, data sharing and access, regulatory considerations of novel trial designs, and emerging methodological approaches for using external data. Although participation was broad in terms of expertise, most participants were from the USA and provided a US perspective on the topic. The discussion from the Clinical
e457 www.thelancet.com/oncology Vol 22 October 2021SeriesCalifornia Los Angeles, Los Angeles, CA, USA (Prof T Cloughesy MD); Foundation Medicine, Cambridge, MA, USA
(B M Alexander)Correspondence to: Dr Rifaquat Rahman, Department of Radiation Oncology, Dana-Farber/Brigham and Women’s Cancer Center, Harvard Medical School, Boston, MA 02215, USA
rrahman@bwh.harvard.edu
Trials Think Tank serves as a framework for this Series paper, which focuses on the use of external data to design, conduct, and analyse clinical trials with an emphasis on possible applications in neuro-oncology. We review trial designs, methodologies, approaches, regulatory considerations, and current barriers in data sharing and access.
Early phase trial designs
Early phase trials are typically designed to obtain preliminary estimates of treatment efficacy and toxic effects, which will inform the decision to pursue a phase 3 trial or stop drug development. Often in neuro-oncology, these early phase studies are single-arm trials that test the superiority of an experimental therapeutic compared with an established benchmark for the current standard of care (eg, median overall survival or other point estimates).20 Importantly, there can be substantial differences between populations or standards to assess outcomes across trials,21 which can lead to inappropriate comparisons and inadequate evaluations of the experimental therapy. Another major challenge with single-arm trials is the choice of a primary endpoint. Response rate might not correlate with overall survival and is difficult to interpret in glioblastoma,22 and single-arm studies are suboptimal for reliable inference on improvements in time-to-event endpoints, such as survival. On the basis of these known limitations, single-arm designs have been posited as a possible reason for poor decision making and unsuccessful phase 3 trials in glioblastoma.23,24The risk of biased conclusions from single-arm trials has been examined extensively and frameworks have been developed to help guide the choice between RCTs versus single-arm designs for glioblastoma.10,25 Despite well documented limitations, single-arm trials remain the most common early phase trial design in glioblastoma.25 Alternative trial designs have been proposed, including the incorporation of randomisation, seamless phase 2/3 study designs,26 and Bayesian outcome-adaptive trials to overcome limitations and
improve the evaluation of therapeutic candidates in the early phase of development.27–29
Overview of trial designs that include external data
Trial designs that leverage external data can generate valuable inferences in settings in which single-arm trials are suboptimal and RCTs are infeasible.30 External data can play a role in supporting key decisions in the drug development process, including regulatory approvals and decision making in early phase trials. The use of external patient-level datasets has the potential to improve accuracy of trial findings and inform decision making (eg, determining the sample size of a subsequent confirmatory phase 3 trial or selecting the phase 3 patient population). External data can also be incorporated into RCTs31—eg, within interim analyses18—although these designs remain largely unexplored.Externally augmented clinical trial design refers to the broad class of designs that leverage external data for decision making during a study or in the final analysis. These designs rely on access to well curated patient-level data for the standard of care from one or more relevant data sources to allow for adjustment of differences in pretreatment, covariates between enrolled patients and external data, and to derive treatment effect estimates. Given the need for statistical adjustments, the external dataset ideally includes a comprehensive set of potential confounders.32 Pretreatment covariates have been thoroughly studied for adult primary brain tumours.33,34An example of an externally augmented clinical trial design consists of a single-arm study combined with an external control group (ie, an external dataset with patient-level outcomes and pretreatment profiles), which is used as a comparator to evaluate the experimental treatment. This design (figure 1A), an externally controlled single-arm trial, infers the treatment effect by using adjustment methods to account for differences in patient pretreatment profiles between the external control group and the experimental group.35 In this design, the external control group is used to contextualise the outcome data from a single-arm study. Instead of using benchmark estimates (eg, median survival) for the standard of care, data analyses and treatment effect estimates are based on patient-level data from an external dataset.Hybrid randomised trial designs constitute another type of externally augmented clinical trial. With adequate external data and statistical planning,36 these designs have the potential of reducing sample size while maintaining the benefits of randomisation. An example of a two-stage hybrid design is shown in figure 1B. The study has an initial 1:1 randomisation to an internal control group and an experimental group. If the interim analysis does not identify differences between adjusted primary outcome distributions in the internal randomised control group and the external control group, then different randomisation ratios (eg, 2:5) can be used in the second stage of the study. If there is evidence of inconsistencies between the external and internal control groups (eg, unmeasured confounders or different measurement standards of outcomes and prognostic variables), the trial can continue with 1:1 randomisation. In the outlined example, the potential increase of randomisation probability for the experimental group can be attractive and might accelerate trial accrual. Indeed, patients with a brain tumour and an inadequate standard of care might be more likely to enrol in a trial if the probability of receiving the experimental therapy is higher.25
Externally augmented clinical trial designsAlong with high quality and complete data, a statistically rigorous study design is the most important element of an externally augmented clinical trial. As with any clinical trial, the design, including the sample size, a detailed
www.thelancet.com/oncology Vol 22 October 2021 e458Series
plan for interim decisions, and statistical methods for data analyses should be prespecified. Additionally, a plan for how missing data in the trial and external data sources will be handled is important. Potential distortion mechanisms that can bias treatment effect estimates and undermine the scientific validity of externally augmented clinical trial findings have been carefully examined and include unmeasured or misclassified confounders and data quality issues, such as the use of different standards to capture or measure outcomes.37–39Risks of introducing bias (table) and compromising the control of false positives and false negatives when using external patient-level data can differ substantially across externally augmented clinical trial designs, which span from single-arm studies (figure 1A) to hybrid randomised studies (figure 1B). Quantitative analyses of these and other risks (eg, exposure of patients to inferior treatments) are necessary before trial initiation. The decision to use external data should take account of several factors in addition to the study population and available patient-level datasets, including the stage of drug development (eg, early phase 2 vs confirmatory trials); the specific decisions (eg, early stopping of a phase 2 study for futility40 or sample size re-estimation during the study41) that will be supported by external data; resources (including maximum sample size); and the potential trial designs and statistical methods for data analysis.Candidate externally augmented clinical trial designs and statistical methods for data analysis can present markedly different trade-offs between potential efficiencies (eg, discontinuing early randomised trials of ineffective treatments by leveraging external data) and risks of poor operating characteristics (eg, bias and poor control of false positive results). The value of integrating external data is context specific and strictly dependent on the trial design and methods selected for data analysis and decision making.We describe three examples of externally augmented clinical trials with markedly different risks of poor operating characteristics. The purpose of these examples is to illustrate how external information can be used for making different decisions during or at completion of a trial.The first example is a single-arm trial with an external control group. We consider either binary primary out-comes (eg, tumour response) or time-to-event outcomes with censoring (eg, overall survival). This design uses procedures developed for observational studies,42 such as matching, propensity score methods,43,44 or inverse
A
External control
dataset
External control
dataset
Enrolled patients Enrolled patients
Analyses and treatment
effect estimate
Adjustment methods
(eg, matching)
B
Randomisation during
first phase of the trial (1:1)
Interim analysis Randomisation during
second phase of the trial
External control
dataset
Control group
Experimental
group
Do early control
data support the
use of external
data?
No
Yes
Altered
randomisation
(eg, 2:5)
Randomisation
(1:1)
C
Control group
External control
dataset
External
control
dataset
Experimental
group
Adjustment
methods
(eg, matching)
Interim and
futility
analysis
Final analysis
and treatment
estimate
Figure 1: Clinical trial designs
(A) A clinical study with patient enrolment to a single experimental group and an external control group (externally controlled single-arm trial). (B) A two-stage hybrid randomised trial design. (C) A randomised trial design that uses external data for interim futility analyses to support the decision to continue or discontinue the study. If the trial is not discontinued, the final analysis does not use external data.
e459 www.thelancet.com/oncology Vol 22 October 2021Series
probability weighting,45 which are applicable to the comparison of single-arm trial data (experimental treatment) with external patient-level data representative of the standard of care treatment (external control; figure 1A). These procedures have been developed to estimate treatment effects in non-randomised studies and have generated numerous contributions to the statistical literature.46 They compare outcome data Y in the experi-mental and control treatment groups with adjustments that account for confounders X.The literature on adjustment methods applicable to externally augmented clinical trial designs (eg, matching)43,44 is largely based on assumptions that are difficult to demonstrate,42 including the absence of unmeasured confounders.47 In the context of externally controlled single-arm trial designs, in which these assumptions might be violated, the investigator can attempt to evaluate the risk of bias and other statistical properties of treatment effect estimates (eg, the coverage of confidence intervals) computed using adjustment methods. Patient-level data from a library of completed RCTs in a specific clinical setting—eg, newly diagnosed glioblastoma patients—facilitate the comparison of externally controlled single-arm trial, RCT, and single-arm trial designs. For example, a treatment effect estimate computed using only data from a previously completed RCT can be compared with a second treatment effect estimate, computed using the experi-
mental group of the same RCT and external data (figure 2).48 This method can be repeated considering different RCTs, adjustment methods, and external datasets. These comparisons describe the consistency between RCT results and hypothetical results obtained from a smaller externally controlled single-arm trial (ie, the experimental group of the RCT, or part of it) leveraging external data; similar evaluation frameworks have been discussed elsewhere.49,50The leave-one-out algorithm is an alternative evaluation approach that requires a collection of completed RCTs with the same control treatment.17 The algorithm has been used to compare the application of candidate causal inference methods in externally controlled single-arm trials. This approach has been applied to a collection of studies in newly diagnosed glioblastoma and requires pretreatment profiles and outcomes from patients receiving standard of care (radiotherapy and temozolomide). This aspect is relevant because data access barriers can be substantially different for the control and experimental groups. The algorithm17 iterates the following three operations for each RCT in the data collection: (1) randomly selects n (the sample size of a hypothetical externally controlled single-arm trial) patients (without replacement) from the radiotherapy and temozolomide group (control) and uses patient pretreatment profiles (X) and outcomes (Y) of these
Randomised controlled trial
Control
group
Experimental
group
Experimental
group
External
data
Adjustment method
Treatment effect estimates
Single-arm trial plus external data
Randomised
controlled
trial
Single-arm
trial plus
external
data
Figure 2: Validation schema
A treatment effect estimate computed using only data from a previously completed randomised controlled trial is compared with a second treatment effect estimate, computed using the experimental group from the same randomised controlled trial and external control data.
Description Example Methods to avoid or reduce bias
Measured confounders The distributions of pretreatment patient characteristics that correlate with outcomes in the trial population and in the external control group are differentThe external control group has a higher Karnofsky performance status or age on average than the trial populationMatching; inverse probability weighting; marginal structural modelsUnmeasured confoundersThe distributions of unmeasured pretreatment patient characteristics that correlate with outcomes in the trial population and in the external control group are differentSupportive care (not captured in the datasets) differs between patients in the clinical trial and in the external control groupValidation analyses can indicate the risk of bias before the trial beginsDifferences in defining prognostic variables or outcomesThe definition of clinical measurements might vary between datasets, leading to differences in the definitions of outcomes or prognostic variables between the clinical trial and the external control groupMeasurement of tumour response with different response criteria or at different intervals in external control armsData dictionaries and validation analyses can reveal these discrepancies
before the trial beginsImmortal time bias In the external dataset the time-to-event outcome cannot occur during a time window because of the study design or other causesIn glioblastoma, different real-world datasets capture patient survival from diagnosis or from other timepoints, such as surgeryExplicit and detailed definitions of the time to event outcomes for the trial and external dataset can reveal the risk of bias
Table: Potential causes of bias in clinical trials with an external control group
www.thelancet.com/oncology Vol 22 October 2021 e460Series
patients to define a fictitious single-arm study (the experimental group); (2) the data on patients receiving radiotherapy and temozolomide in the remaining studies are used as external data, these datasets are combined into a single data matrix (the external control group); and (3) a treatment effect estimate is computed by comparing the fictitious externally controlled single-arm trial (step 1) and the external data (step 2), using a candidate adjustment method, which is also used to test the null hypothesis (H0—ie, the treatment does not improve the primary outcome).For each study in the data collection, steps 1–3 (similar to cross-validation) can be repeated to evaluate bias, variability of the treatment effect estimate, and the risk of false positive results. By design, the treatment effect in this fictitious comparison is null, because patients receiving radiotherapy and temozolomide are being compared with those receiving radiotherapy and temozolomide from different studies. This facilitates interpretability and produces bias summaries for the statistical plan of an externally controlled single-arm trial. An analysis using this leave-one-out algorithm approach in patients with newly diagnosed glioblastoma showed high false positive error rates in standard single-arm clinical trials (higher than the α level),10 which can be considerably reduced (up to 30%) by using external control data from previously completed trials.17The first validation approach (figure 2) attempts to replicate results of a completed RCT, whereas the leave-one-out algorithm approach is based on subsampling a control group. Both approaches are valuable strategies that can detect potential distortion mechanisms (eg, unmeasured confounders or inconsistent definitions of primary outcomes), which undermine the scientific validity of externally controlled single-arm trials. These approaches require patient-level data from several RCTs with adequate sample sizes to produce reliable analyses of
the risk of bias and false positive results in future externally controlled single-arm trials. Importantly, positive findings from retrospective analyses should not be overinterpreted using either approach because relevant changes in available treatments and technologies, or other factors, can rapidly make the entire data collection obsolete and inadequate.51The second type of externally augmented clinical trials are hybrid randomised trial designs that combine external and randomised control data to estimate potential treatment effects.52 Figure 1B represents a two-stage hybrid design. In the first stage, patients are randomly assigned to the experimental group and the internal control arm. Interim analysis can be used for early stopping for futility and to determine sample sizes for the experimental and control groups in the second stage. These decisions are based on a similarity measure comparing estimates of conditional outcome distributions (Pr[Y|X]) of the external and internal control groups and preliminary treatment effect estimates. The proportion of patients randomised to the internal control group during the second stage can be reduced or increased on the basis of a prespecified interim analysis, which involves sum-maries in support of or against the integration of external data to estimate the effects of the experimental treatment.On the basis of results from a phase 2 study of MDNA55,19 investigators are planning an open-label, phase 3 registration study with implementation of a hybrid randomised design in patients with recurrent IDH-wild-type glioblastoma. The study team is considering a 3:1 (experimental to control) randomisation ratio with a final comparison of overall survival between patients receiving MDNA55 and the control groups (external and internal).53The third type of externally augmented clinical trials incorporates external data to support futility analyses. External data can be incorporated into RCTs for use in interim decisions.18,31 In such a design (figure 1C), interim analyses use predictions based on early data from the RCT combined with external data. These predictions express the probability that the trial will generate significant evidence of positive treatment effects. The trial is discontinued by design if the predictive probability becomes smaller than a fixed threshold. After completing the enrolment and follow-up phases, the final analysis does not use external data. The primary endpoint of the trial is positive (indicating improved outcomes with the experimental treatment) if the p value, computed using only RCT data (excluding external information), has a value lower than the targeted control of false positive results α.In ideal settings, without unmeasured confounders and other distortion mechanisms, leveraging external data for interim futility analyses can reduce the expected sample size when the experimental treatment is ineffective and can reduce early stopping probability
when the experimental therapy is superior, thus increasing the power.18 Additionally, the outlined design maintains a rigorous control of type I error probability, even in the presence of unmeasured confounders, because the external data are excluded from the final data analyses. The efficiency gains and risks associated with integration of external data into interim decisions have been quantified for trials of patients with newly diagnosed glioblastoma using evaluation analyses that built on a collection of datasets from completed RCTs, the leave-one-out algorithm, and other similar procedures.18As externally augmented clinical trials require numerous context-specific considerations, from relevant aspects of the external datasets to the feasibility of alternative designs, a discussion with regulatory agencies in early stages of trial planning is strongly recommended.
External data sourcesThe use of external controls to evaluate new treatments is
dependent on the availability of high-quality external data. Selecting appropriate datasets is crucial and checklists have been developed to provide guidance on data quality.54
e461 www.thelancet.com/oncology Vol 22 October 2021Series
Data considerations for external controls include appropriate capture of patient-level data,55 consistent definition of covariates and endpoints, and adequate temporality of the data, as small temporal lags can substantially affect the trial analysis.49 Investigators should consider potential biases that occur if the endpoint definitions are inconsistent across studies. For example, survival can be measured from the date of diagnosis, randomisation, or adjuvant treatment start. The definition of the outcome should be explicit and consistent during the trial and in the external datasets.56 Missing data are another important consideration in analyses with external data.57,58 Although there are methods to address missing data (eg, multiple imputation and likelihood-based methods), their use within externally augmented clinical trial designs has not been extensively studied. In general, the external control population and trial population should be similar to reduce the risk of bias. Potential unmeasured confounders, inconsistencies in definitions, and different measurement standards of covariates and outcomes across datasets need to be scrutinised using data dictionaries and study protocols. Contemporaneous controls are ideal, but historical controls with patient-level data might be helpful in appropriate contexts. For example, disease settings without a recent change in the standard of care (eg, glioblastoma) or a long track record of time-stable outcomes might have more flexibility in the temporality of data, but this benefit should be weighed against the possibility of unmeasured aspects of care such as advances in imaging, radiotherapy, surgical techniques, and supportive care that might change over time.59 Additionally, the data should be traceable (ie, an audit trail should be available to document data management processes) to support a marketing application.54The two most relevant sources of external data are previously completed clinical trials and real-world data derived from clinical practice. The use of data from clinical trials can be advantageous, given that the data are typically collected in a rigorous environment with vetting procedures. Clinical trials are often done in specialised institutions and enrol clearly defined patient populations. Data from patients previously enrolled on RCTs and given standard of care might be more likely than real-world data cohorts to contain pretreatment profiles similar to those of patients who will be enrolled in future trials. The use of data collection forms, intensive monitoring, and specialised personnel enable adherence to clear protocols to produce standardised data.15 Data from previous trial participants exist, as there have been several negative phase 3 randomised trials in glioblastoma, sponsored by cooperative groups and industry, with many patients receiving the current standard of care (radiotherapy and temozolomide).60–63 Data access, however, can be challenging because of impediments to data sharing64 and contemporary trial data might not be made available by trial sponsors.Real-world data is derived from registries, claims and billing data, personal devices and applications, or electronic health records. Because real-world data is generally not collected for research purposes, there can be concerns about data organisation and quality, confounding, selection mechanisms, and ultimately, bias.65–67 Advances in the quality of electronic health records data have created opportunities, and new curated datasets can be linked to molecular or radiological data with high fidelity. Efforts to harmonise real-world data from disparate data sources and novel methods to incorporate such data into clinical studies provide an avenue to inform trial design68 and regulatory decision making,69 but further work is required to validate these approaches.Although differences between real-world data and data from clinical studies have been reported,70 methodological work on the use of joint models and analyses to compensate for the scarcity of trial data and the potential distortions of real-world data (eg, measurement errors or unknown selection mechanisms) is in early stages.Methodological work has primarily focused on overall survival,17,48 which is more likely to be adequately captured in external datasets relative to other outcomes. Radiological endpoints, such as progression-free survival, require caution because of the risk of inconsistent assessments across datasets. In real-world data, radiological outcomes might not be determined by standardised response assessment criteria and central radiological review would probably not be routinely implemented. Although brain tumour trials71,72 often use consensus guidelines by the Response Assessment in Neuro-Oncology working group, these criteria include subjective components,73 and datasets of previously completed trials can still include misclassification errors. Quality of life, neurological function, and neurocognitive outcomes are increasingly incorporated into clinical trials.72,74 These non-survival outcomes can provide meaningful measures of the clinical benefit of a therapy and serve as valuable endpoints in neuro-oncology trials.75 Nevertheless, missing measurements of these outcomes are common across datasets from completed trials and real-world data. Exploration of the use of external data to analyse radiographical outcomes, patient-centred outcomes, and safety outcomes in neuro-oncology trials remains scarce.
External control groups beyond neuro-oncologyCarrigan and colleagues48 leveraged a curated real-world data dataset of 48 856 patients (electronic health records from a Flatiron database) to reanalyse 11 completed trials in advanced non-small-cell lung cancer. In this study, the external control groups were defined with matching methods. The external control groups were able to recapitulate treatment effect estimates (hazard ratios) for ten of 11 RCTs. This finding suggests that real-world data can potentially be used as external controls partly because of the large number of patients
www.thelancet.com/oncology Vol 22 October 2021 e462Series
in these datasets, as many patient records can meet the RCT inclusion and exclusion criteria and be used as comparators. Notably, the external control groups derived from electronic health records did not recapitulate the results for one RCT. On inspection, the authors of the study felt this discordance was associated with a biomarker subgroup population that was not sufficiently represented in the electronic health record dataset. These findings underscore the need to account for biomarkers and well represented sub populations in external control groups.A study by Ray and colleagues76 provides an example of the use of external data to contextualise a single-arm trial. In an FDA led retrospective analysis, the outcome of invasive disease-free survival from a single-arm study77 of adjuvant paclitaxel and trastu zumab in patients with HER2-positive early breast cancer was analysed using an external control group derived from clinical trials (control therapies included anthracycline, cyclophosphamide, taxane, and trastuzumab; or taxane, carboplatin, and trastuzumab).78 The de-escalated regimen that combines adjuvant paclitaxel and trastuzumab had been adopted in clinical practice on the basis of initial single-arm trial results.77 This retrospective analysis used propensity score matching to adjust for differences in pretreatment patient profiles in the single-arm trial and the external control dataset. The analysis estimated similar outcome distributions for adjuvant paclitaxel and trastuzumab and the control, which supported the use of a de-escalated regimen, particularly in light of higher toxic effects with the control regimens.
Considerations and implications for regulatory decision makingIn the USA, the 21st Century Cures Act directed the FDA to develop guidance for evaluation and use of real-world data, and to consider the potential role of real-world data in drug development and regulatory decision making. For example, real-world data could be used to support approvals for new indications or be integrated into existing monitoring requirements after approval.79 Accordingly, the FDA launched a real-world data programme to lay the foundation for rigorous use of such data in regulatory decisions.80 Several ongoing initiatives are providing guidance on data quality, data standards, and study designs that incorporate real-world data.81 Also, other regulatory institutions, such as the European Medicines Agency and Health Canada, have shown an openness towards better understanding and potentially leveraging real-world data for drug development.82,83Within a regulatory scope, the use of external control data can support expedited approval, extend on-label use of a therapy to a new indication or subgroup, and more generally support regulatory decision making.84 For example, in terms of rare diseases, the FDA approved blinatumomab for adults with relapsed or refractory acute lymphocytic leukaemia on the basis of a study that used data from a previous clinical trial as a comparator.85In a regulatory context, there is a high burden of proof for investigators to show the scientific rigor of study designs and analyses that leverage external control data with an appropriate risk level. Comparative analyses with standard RCT designs are fundamental to evaluate robustness and efficiencies of externally augmented clinical trial designs. The external data, study design, and analysis should be tailored to each specific clinical context and intended regulatory use. Each element should be carefully considered and scrutinised when evaluating the risks of biased treatment effect estimates and inadequate control of false positive findings.
Data sharing modelsDespite the appeal of patient-level data from previous clinical trials, data access is a barrier to studying and implementing externally augmented clinical trial designs in early phase and late phase trials. Data sharing efforts from industry funded RCTs are increasing, but substantial room for improvement remains.86,87 Beyond implications for externally augmented clinical trials, clinical trial data sharing allows investigators to carry out analyses that generate new knowledge and which have been deemed “essential for expedited translation of research results into knowledge, products, and procedures to improve human health” by the National Institutes of Health.88 Substantial challenges and appropriate concerns about data sharing remain,89 including the need to ensure patient privacy and academic credit; the use of adequate standards for combining data from different sources; and the allotment of resources required to deidentify patient records and provide infrastructures for data sharing. The patient perspective serves as an important counterpoint, assuming that privacy is protected; studies indicate that patients are in favour of data sharing for purposes that can help to advance clinical outcomes.90Advances towards simpler data access could transform the ability to do secondary analyses91 and leverage external data in future clinical studies. For many data sharing platforms, a gatekeeper model is used, often with long approval processes, restrictive criteria for data access, and limitations on data use. These requirements act as a mechanism of passive resistance and delay access to data from completed trials. An increasing number of data sharing platforms such as Vivli,92 YODA,93 and Project Data Sphere,94 are aligned with more open sharing models for clinical trial datasets.95 Nonetheless, most data from previously completed neuro-oncology trials remain difficult to access.New policies might be necessary for data sharing and to accelerate the study of new therapeutics. An important consideration is the modification of incentives for data sharing.96 A systematic effort from cooperative groups, industry, academia, and other stakeholders could help to achieve this goal. Regulatory requirements that ensure
e463 www.thelancet.com/oncology Vol 22 October 2021Series
timely data sharing and patient advocacy groups could play key roles in hastening this process. Additionally, initiatives and agreements to prospectively share patient-level data from the control groups of multiple cooperative RCTs could be beneficial to participating studies and create opportunities to extend data sharing.
ConclusionAt the end of our Clinical Trials Think Tank session, there was a strong interest and desire to continue to collaborate and investigate, validate, and implement externally augmented clinical trial designs in glio-blastoma and, on a broader level, in neuro-oncology. Efforts to form industry cooperative group partnerships and selection of datasets and statistical methods were set as goals to continue towards an advancement in understanding the role of externally augmented clinical trials for drug development in neuro-oncology.The use of external data to design and analyse clinical studies has the potential to accelerate drug development and can contribute to rigorous evaluation of new
treatments. RCTs will remain the indisputable gold standard for evaluation of treatments, but external datasets can supplement information gathered from RCTs and single-arm studies. Further methodological work can help to identify the appropriate clinical contexts, data, and statistical designs for externally augmented clinical trials that generate inference on treatment effects of experimental therapies, with well controlled risks regarding their their accuracy and scientific validity.There is a continuum of approaches for leveraging external data, and use of externally augmented clinical trial designs should be tailored to the disease context. An emphasis on high-quality patient-level data, rigorous methods, and biostatistical expertise are crucial in the successful implementation of externally augmented clinical trials. Data access to previously completed clinical
trials and real-world data is improving, but new policies and initiatives for data sharing could further unlock the value of external data. Continued collab orations between stakeholders, including indus try, academia, biostatisticians, clinicians, regulatory agencies, and patient advocates, are crucial to understand the appropriate use of externally augmented clinical trial designs in neuro-oncology.
Search strategy and selection criteria
We searched PubMed for articles using the search terms “external control arms”, “synthetic control arms”, “neuro-oncology trial design”, “glioblastoma trial design”, published from Jan 1, 2000, to May 1, 2021. Articles were also identified through searches of the authors’ own files. Only papers published in English were reviewed. The final reference list was generated on the basis of relevance to the scope of this Series paper.
Contributors
RR, PYW, BMA, and LT conceptualised this Series paper and wrote the first draft. All authors contributed to the interpretation of findings and editing. All authors approved the final manuscript.
Declaration of interests
RR received research support from the Project Data Sphere, outside of submitted work. IR-R reports employment and owns stocks of Roche and Genentech. FM reports employment at Medicenna Therapeutics. LEA reports employment and owns stocks of Novartis. JEA reports employment and owns stocks of Chimerix. LKA and EA-C report employment at Candel Therapuetics. SB reports grants and personal fees from Novocure; grants from Incyte, GSK, and Eli Lilly; and personal fees from Bayer and Sumitomo Dainippon. MK reports personal fees from Ipsen, Pfizer, Roche, and Jackson Laboratory for Genomic Medicine and research funding paid to his institution from Specialised Therapeutics. TC reports personal fees from Roche, Trizel, Medscape, Bayer, Amgen, Odonate Therapeutics, Pascal Biosciences, DelMar Pharmaceuticals, Tocagen, Karyopharm, GW Pharmaceuticals, Kiyatec, AbbVie, Boehinger Ingelheim, VBI Vaccines, Dicephera, VBL Therapeutics, Agios, Merck, Genocea, Puma, Lilly, Bristol Myers Squibb, Cortice, Wellcome Trust; and stock options from Notable Labs. TC has a patent (62/819,322) with royalties paid to Katmai and is a board member for the 501c3 Global Coalition for Adaptive Research. PYW reports personal fees from Abbvie, Agios, AstraZeneca, Blue Earth Diagnostics, Eli Lilly, Genentech, Roche, Immunomic Therapeutics, Kadmon, Kiyatec, Merck, Puma, Vascular Biogenics, Taiho, Tocagen, Deciphera, and VBI Vaccines; and research support from Agios, AstraZeneca, Beigene, Eli Lily, Genentech, Roche, Karyopharm, Kazia, MediciNova, Merck, Novartis, Oncoceutics, Sanofi-Aventis, and VBI Vaccines. BMA reports employment at Foundation Medicine; personal fees from AbbVie, Bristol Myers Squibb, Precision Health Economics, and Schlesinger Associates; and research support from Puma, Eli Lilly, Celgene. SV, JM, BL, M-YCP, DA, KT, and LT declare no competing interests.
Acknowledgments
We thank Amy Barone, Pallavi Mishra-Kalyani, and the Food and Drug Administration for contributing and helping with preparation of this Series paper. We also thank the Society for Neuro-Oncology and their staff for arrangements and coordination of the 2020 Clinical Trials Think Tank meeting. We thank Johnathan Rine for help with preparation of the figures. RR was supported by the Joint Center for Radiation Therapy Foundation grant. LT and SV were supported by the National Institutes of Health (grant 1R01LM013352–01A1).
References
1 Hwang TJ, Carpenter D, Lauffenburger JC, Wang B, Franklin JM, Kesselheim AS. Failure of investigational drugs in late-stage clinical development and publication of trial results. JAMA Intern Med 2016; 176: 1826–33.2 Wong CH, Siah KW, Lo AW. Estimation of clinical trial success rates and related parameters. Biostatistics 2019; 20: 273–86.3 Hu M, Yang M, Liu Y. Statistical adaptation to oncology drug development evolution. Contemp Clin Trials 2020; 99: 106180.4 Mandrekar SJ, Sargent DJ. Clinical trial designs for predictive biomarker validation: theoretical considerations and practical challenges. J Clin Oncol 2009; 27: 4027–34.5 Alexander BM, Cloughesy TF. Adult glioblastoma. J Clin Oncol 2017; 35: 2402–09.6 Wen PY, Weller M, Lee EQ, et al. Glioblastoma in adults: a Society for Neuro-Oncology (SNO) and European Society of Neuro-Oncology (EANO) consensus review on current management and future directions. Neuro Oncol 2020; 22: 1073–113.7 Barbaro M, Fine HA, Magge RS. Scientific and clinical challenges within neuro-oncology. World Neurosurg 2021; published online Feb 18. https://doi.org/10.1016/j.wneu.2021.01.151.8 Vanderbeek AM, Rahman R, Fell G, et al. The clinical trials landscape for glioblastoma: is it adequate to develop new treatments?
Neuro Oncol 2018; 20: 1034–43.9 Lee EQ, Chukwueke UN, Hervey-Jumper SL, et al. Barriers to accrual and enrollment in brain tumor trials. Neuro Oncol 2019;
21: 1100–17.
www.thelancet.com/oncology Vol 22 October 2021 e464Series
10 Vanderbeek AM, Ventz S, Rahman R, et al. To randomize, or not to randomize, that is the question: using data from prior clinical trials to guide future designs. Neuro Oncol 2019; 21: 1239–49.11 Ostrom QT, Patil N, Cioffi G, Waite K, Kruchko C, Barnholtz-Sloan JS. CBTRUS statistical report: primary brain and other central nervous system tumors diagnosed in the United States in 2013–2017. Neuro Oncol 2020; 22 (suppl 2): iv1–96.12 Eborall HC, Stewart MCW, Cunningham-Burley S, Price JF, Fowkes FGR. Accrual and drop out in a primary prevention randomised controlled trial: qualitative study. Trials 2011; 12: 7.13 Featherstone K, Donovan JL. “Why don’t they just tell me straight, why allocate it?” The struggle to make sense of participating in a randomised controlled trial. Soc Sci Med 2002; 55: 709–19.14 Feinberg BA, Gajra A, Zettler ME, Phillips TD, Phillips EG, Kish JK. Use of real-world evidence to support FDA approval of oncology drugs. Value Health 2020; 23: 1358–65.15 Eichler H-G, Pignatti F, Schwarzer-Daum B, et al. Randomized controlled trials versus real world evidence: neither magic nor myth. Clin Pharmacol Ther 2021; 109: 1212–18.16 Berry DA. The brave new world of clinical cancer research: adaptive biomarker-driven trials integrating clinical practice with clinical research. Mol Oncol 2015; 9: 951–59.17 Ventz S, Lai A, Cloughesy TF, Wen PY, Trippa L, Alexander BM. Design and evaluation of an external control arm using prior clinical trials and real-world data. Clin Cancer Res 2019;
25: 4993–5001.18 Ventz S, Comment L, Louv B, et al. The use of external control data for predictions and futility interim analyses in clinical trials.
Neuro Oncol 2021; published online June 9. https://doi.org/10.1093/neuonc/noab141.19 Sampson JH, Achrol A, Aghi MK, et al. MDNA55 survival in recurrent glioblastoma (rGBM) patients expressing the interleukin-4 receptor (IL4R) as compared to a matched synthetic control. Proc Am Soc Clin Oncol 2020; 38 (suppl): 2513 (abstr).20 Thall PF, Simon R. Incorporating historical control data in planning phase II clinical trials. Stat Med 1990; 9: 215–28.21 Neuenschwander B, Capkun-Niggli G, Branson M, Spiegelhalter DJ. Summarizing historical information on controls in clinical trials. Clin Trials 2010; 7: 5–18.22 Reardon DA, Galanis E, DeGroot JF, et al. Clinical trial end points for high-grade glioma: the evolving landscape. Neuro Oncol 2011;
13: 353–61.23 Sharma MR, Karrison TG, Jin Y, et al. Resampling phase III data to assess phase II trial designs and endpoints. Clin Cancer Res 2012;
18: 2309–15.24 Tang H, Foster NR, Grothey A, Ansell SM, Goldberg RM, Sargent DJ. Comparison of error rates in single-arm versus randomized phase II cancer clinical trials. J Clin Oncol 2010;
28: 1936–41.25 Grossman SA, Schreck KC, Ballman K, Alexander B. Point/counterpoint: randomized versus single-arm phase II clinical trials for patients with newly diagnosed glioblastoma. Neuro Oncol 2017; 19: 469–74.26 Stallard N, Todd S. Seamless phase II/III designs.
Stat Methods Med Res 2011; 20: 623–34.27 Alexander BM, Trippa L, Gaffey S, et al. Individualized screening trial of innovative glioblastoma therapy (INSIGhT): a Bayesian adaptive platform trial to develop precision medicines for patients with glioblastoma. JCO Precis Oncol 2019; 3: 1–13.28 Alexander BM, Ba S, Berger MS, et al. Adaptive global innovative learning environment for glioblastoma: GBM AGILE.
Clin Cancer Res 2018; 24: 737–43.29 Buxton MB, Alexander BM, Berry DA, et al. GBM AGILE: a global, phase II/III adaptive platform trial to evaluate multiple regimens in newly diagnosed and recurrent glioblastoma.
Proc Am Soc Clin Oncol 2020; 38 (suppl): TPS2579 (abstr).30 Thorlund K, Dron L, Park JJH, Mills EJ. Synthetic and external controls in clinical trials—a primer for researchers. Clin Epidemiol 2020; 12: 457–67.31 Viele K, Berry S, Neuenschwander B, et al. Use of historical control data for assessing treatment effects in clinical trials. Pharm Stat 2014; 13: 41–54.32 VanderWeele TJ, Shpitser I. On the definition of a confounder.
Ann Stat 2013; 41: 196–220.33 Pignatti F, van den Bent M, Curran D, et al. Prognostic factors for survival in adult patients with cerebral low-grade glioma.
J Clin Oncol 2002; 20: 2076–84.34 Gittleman H, Lim D, Kattan MW, et al. An independently validated nomogram for individualized estimation of survival among patients with newly diagnosed glioblastoma: NRG Oncology RTOG 0525 and 0825. Neuro Oncol 2017; 19: 669–77.35 Davi R, Mahendraratnam N, Chatterjee A, Dawson CJ, Sherman R. Informing single-arm clinical trials with external controls.
Nat Rev Drug Discov 2020; 19: 821–22.36 Normington J, Zhu J, Mattiello F, Sarkar S, Carlin B. An efficient Bayesian platform trial design for borrowing adaptively from historical control data in lymphoma. Contemp Clin Trials 2020;
89: 105890.37 Webster-Clark M, Jonsson Funk M, Stürmer T. Single-arm trials with external comparators and confounder misclassification: how adjustment can fail. Med Care 2020; 58: 1116–21.38 Thompson D. Replication of randomized, controlled trials using real-world data: what could go wrong? Value Health 2021; 24: 112–15.39 Seeger JD, Davis KJ, Iannacone MR, et al. Methods for external control groups for single arm trials or long-term uncontrolled extensions to randomized clinical trials.
Pharmacoepidemiol Drug Saf 2020; 29: 1382–92.40 Snapinn S, Chen M-G, Jiang Q, Koutsoukos T. Assessment of futility in clinical trials. Pharm Stat 2006; 5: 273–81.41 Gould AL. Sample size re-estimation: recent developments and practical considerations. Stat Med 2001; 20: 2625–43.42 Imbens GW, Rubin DB. Causal inference for statistics, social, and biomedical sciences: an introduction, 1st edn. New York, NY, USA: Cambridge University Press, 2015.43 Rubin DB. The use of matched sampling and regression adjustment to remove bias in observational studies. Biometrics 1973; 29: 185–203.44 Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika 1983; 70: 41–55.45 Li L, Greene T. A weighting analogue to pair matching in propensity score analysis. Int J Biostat 2013; 9: 215–34.46 Pearl J. An introduction to causal inference. Int J Biostat 2010; 6: 7.
47 Lin W-J, Chen JJ. Biomarker classifiers for identifying susceptible subpopulations for treatment decisions. Pharmacogenomics 2012;
13: 147–57.48 Carrigan G, Whipple S, Capra WB, et al. Using electronic health records to derive control arms for early phase single-arm lung cancer trials: proof-of-concept in randomized controlled trials.
Clin Pharmacol Ther 2020; 107: 369–77.49 Abrahami D, Pradhan R, Yin H, Honig P, Baumfeld Andre E, Azoulay L. Use of real-world data to emulate a clinical trial and support regulatory decision making: assessing the impact of temporality, comparator choice, and method of adjustment.
Clin Pharmacol Ther 2021; 109: 452–61.50 Franklin JM, Patorno E, Desai RJ, et al. Emulating randomized clinical trials with nonrandomized real-world evidence studies: first results from the RCT DUPLICATE initiative. Circulation 2021;
143: 1002–13.51 Beaulieu-Jones BK, Finlayson SG, Yuan W, et al. Examining the use of real-world evidence in the regulatory process.
Clin Pharmacol Ther 2020; 107: 843–52.52 Hobbs BP, Carlin BP, Sargent DJ. Adaptive adjustment of the randomization ratio using historical control data. Clin Trials 2013;
10: 430–40.53 Medicenna. Medicenna provides MDNA55 rGBM clinical program update following positive end of phase 2 meeting with the U.S. Food and Drug Administration (FDA). 2020. Medicenna Provides MDNA55 rGBM Clinical Program Update Following Positive End of Phase 2 Meeting with the U.S. Food and Drug Administration (FDA) - Medicenna Therapeutics (accessed June 4, 2021).54 Miksad RA, Abernethy AP. Harnessing the power of Real-World Evidence (RWE): a checklist to ensure regulatory-grade data quality.
Clin Pharmacol Ther 2018; 103: 202–05.55 Ghadessi M, Tang R, Zhou J, et al. A roadmap to using historical controls in clinical trials—by Drug Information Association Adaptive Design Scientific Working Group (DIA-ADSWG).
Orphanet J Rare Dis 2020; 15: 69.56 Backenroth D. How to choose a time zero for patients in external control arms. Pharm Stat 2021; 20: 783–92.
57 Kilburn LS, Aresu M, Banerji J, Barrett-Lee P, Ellis P, Bliss JM. Can routine data be used to support cancer clinical trials? A historical baseline on which to build: retrospective linkage of data from the TACT (CRUK 01/001) breast cancer trial and the National Cancer Data Repository. Trials 2017; 18: 561.58 Little RJ, D’Agostino R, Cohen ML, et al. The prevention and treatment of missing data in clinical trials. N Engl J Med 2012;
367: 1355–60.59 Basch E, Deal AM, Dueck AC, et al. Overall survival results of a trial assessing patient-reported outcomes for symptom monitoring during routine cancer treatment. JAMA 2017; 318: 197–98.60 Gilbert MR, Dignam J, Won M, et al. RTOG 0825: phase III double-blind placebo-controlled trial evaluating bevacizumab (Bev) in patients (Pts) with newly diagnosed glioblastoma (GBM).
Proc Am Soc Clin Oncol 2013; 31 (suppl): 1 (abstr).61 Gilbert MR, Dignam JJ, Armstrong TS, et al. A randomized trial of bevacizumab for newly diagnosed glioblastoma. N Engl J Med 2014;
370: 699–708.62 Chinot OL, Wick W, Mason W, et al. Bevacizumab plus radiotherapy-temozolomide for newly diagnosed glioblastoma.
N Engl J Med 2014; 370: 709–22.63 Weller M, Butowski N, Tran DD, et al. Rindopepimut with temozolomide for patients with newly diagnosed, EGFRvIII-expressing glioblastoma (ACT IV): a randomised, double-blind, international phase 3 trial. Lancet Oncol 2017; 18: 1373–85.64 Mbuagbaw L, Foster G, Cheng J, Thabane L. Challenges to complete and useful data sharing. Trials 2017; 18: 71.65 Pearl J. Causality: models, reasoning and inference, 2nd edn. New York, NY, USA: Cambridge University Press, 2009.66 Collins R, Bowman L, Landray M, Peto R. The magic of randomization versus the myth of real-world evidence. N Engl J Med 2020; 382: 674–78.67 Larrouquere L, Giai J, Cracowski J-L, Bailly S, Roustit M. Externally controlled trials: are we there yet? Clin Pharmacol Ther 2020;
108: 918–19.68 Liu R, Rizzo S, Whipple S, et al. Evaluating eligibility criteria of oncology trials using real-world data and AI. Nature 2021;
592: 629–33.69 Corrigan-Curay J, Sacks L, Woodcock J. Real-world evidence and real-world data for evaluating drug safety and effectiveness. JAMA 2018; 320: 867–68.70 Unger JM, Barlow WE, Martin DP, et al. Comparison of survival outcomes among cancer patients treated in and out of clinical trials.
J Natl Cancer Inst 2014; 106: dju002.71 Chukwueke UN, Wen PY. Use of the Response Assessment in Neuro-Oncology (RANO) criteria in clinical trials and clinical practice. CNS Oncol 2019; 8: CNS28.72 Wen PY, Chang SM, Van den Bent MJ, Vogelbaum MA, Macdonald DR, Lee EQ. Response assessment in neuro-oncology clinical trials. J Clin Oncol 2017; 35: 2439–49.73 Huang RY, Neagu MR, Reardon DA, Wen PY. Pitfalls in the neuroimaging of glioblastoma in the era of antiangiogenic and immuno/targeted therapy—detecting illusive disease, defining response. Front Neurol 2015; 6: 33.74 Gilbert MR, Rubinstein L, Lesser G. Creating clinical trial designs that incorporate clinical outcome assessments. Neuro Oncol 2016;
18 (suppl 2): ii21–25.75 Blakeley JO, Coons SJ, Corboy JR, Kline Leidy N, Mendoza TR, Wefel JS. Clinical outcome assessment in malignant glioma trials: measuring signs, symptoms, and functional limitations.
Neuro Oncol 2016; 18 (suppl 2): ii13–20.76 Ray EM, Carey LA, Reeder-Hayes KE. Leveraging existing data to contextualize phase II clinical trial findings in oncology. Ann Oncol 2020; 31: 1591–93.77 Tolaney SM, Barry WT, Dang CT, et al. Adjuvant paclitaxel and trastuzumab for node-negative, HER2-positive breast cancer.
N Engl J Med 2015; 372: 134–41.78 Amiri-Kordestani L, Xie D, Tolaney SM, et al. A Food and Drug Administration analysis of survival outcomes comparing the adjuvant paclitaxel and trastuzumab trial with an external control from historical clinical trials. Ann Oncol 2020; 31: 1704–08.79 Franklin JM, Glynn RJ, Martin D, Schneeweiss S. Evaluating the use of nonrandomized real-world data analyses for regulatory decision making. Clin Pharmacol Ther 2019; 105: 867–77.80 US Food and Drug Administration. Framework for FDA’s real-world evidence program. 2018. https://www.fda.gov/media/120060/download (accessed June 2, 2021).81 US Food and Drug Administration. Complex innovative trial design pilot meeting program. 2021. Complex Innovative Trial Design Pilot Meeting Program (accessed June 26, 2021).82 European Medicines Agency. EMA regulatory science to 2025. 2020. https://www.ema.europa.eu/en/docume...tory-science-2025-strategic-reflection_en.pdf (accessed June 3, 2021).83 Government of Canada. Strengthening the use of real world evidence for drugs. 2018. Regulatory review of drugs and devices: Strengthening the use of real world evidence for drugs - Canada.ca (accessed June 3, 2021).84 Burcu M, Dreyer NA, Franklin JM, et al. Real-world evidence to support regulatory decision-making for medicines: considerations for external control arms. Pharmacoepidemiol Drug Saf 2020; 29: 1228–35.85 Gökbuget N, Kelsh M, Chia V, et al. Blinatumomab vs historical standard therapy of adult relapsed/refractory acute lymphoblastic leukemia. Blood Cancer J 2016; 6: e473.86 Boutron I, Dechartres A, Baron G, Li J, Ravaud P. Sharing of data from industry-funded registered clinical trials. JAMA 2016;
315: 2729–30.87 Miller J, Ross JS, Wilenzick M, Mello MM. Sharing of clinical trial data and results reporting practices among large pharmaceutical companies: cross sectional descriptive study and pilot of a tool to improve company practices. BMJ 2019; 366: l4217.88 National Institutes of Health. Final NIH statement on sharing research data. 2003. NIH Guide: FINAL NIH STATEMENT ON SHARING RESEARCH DATA (accessed June 15, 2021).89 Longo DL, Drazen JM. Data sharing. N Engl J Med 2016;
374: 276–77.90 Mello MM, Lieou V, Goodman SN. Clinical trial participants’ views of the risks and benefits of data sharing. N Engl J Med 2018;
378: 2202–11.91 Arfè A, Ventz S, Trippa L. Shared and usable data from phase 1 oncology trials-an unmet need. JAMA Oncol 2020; 6: 980–81.92 Bierer BE, Li R, Barnes M, Sim I. A global, neutral platform for sharing trial data. N Engl J Med 2016; 374: 2411–13.93 Krumholz HM, Waldstreicher J. The Yale Open Data Access (YODA) project—a mechanism for data sharing. N Engl J Med 2016;
375: 403–05.94 Bertagnolli MM, Sartor O, Chabner BA, et al. Advantages of a truly open-access data-sharing model. N Engl J Med 2017; 376: 1178–81.95 Pisani E, Aaby P, Breugelmans JG, et al. Beyond open data: realising the health benefits of sharing data. BMJ 2016; 355: i5295.96 Lo B, DeMets DL. Incentives for clinical trialists to share data.
N Engl J Med 2016; 375: 1112–15.
https://www.dr-bala.net/NWBO/Lancet/Rahman_Lancet_102021.pdf
Cancer Trials and Design Principles 4Leveraging external data in the design and analysis of clinical trials in neuro-oncology
Rifaquat Rahman, Steffen Ventz, Jon McDunn, Bill Louv, Irmarie Reyes-Rivera, Mei-Yin C Polley, Fahar Merchant, Lauren E Abrey, Joshua E Allen, Laura K Aguilar, Estuardo Aguilar-Cordova, David Arons, Kirk Tanner, Stephen Bagley, Mustafa Khasraw, Timothy Cloughesy, Patrick Y Wen, Brian M Alexander*, Lorenzo Trippa*
Integration of external control data, with patient-level information, in clinical trials has the potential to accelerate the development of new treatments in neuro-oncology by contextualising single-arm studies and improving decision making (eg, early stopping decisions). Based on a series of presentations at the 2020 Clinical Trials Think Tank hosted by the Society of Neuro-Oncology, we provide an overview on the use of external control data representative of the standard of care in the design and analysis of clinical trials. High-quality patient-level records, rigorous methods, and validation analyses are necessary to effectively leverage external data. We review study designs, statistical methods, risks, and potential distortions in using external data from completed trials and real-world data, as well as data sources, data sharing models, ongoing work, and applications in glioblastoma.
IntroductionDrug development is associated with inefficiency, long timelines, and poor success rates in oncology, in which less than 10% of drug candidates are ultimately approved by the US Food and Drug Administration (FDA).1,2 As new unproven therapies emerge at an accelerated pace, there has been an increasing interest in novel approaches to clinical trial design that could improve efficiency.3,4Within neuro-oncology, the use of trial designs with potential for increased efficiency is of interest, particularly in the study of glioblastoma, a disease with an unmet need for better therapies because of its poor prognosis.5 There are several distinctive challenges in drug development for glioblastoma, such as the inability to completely resect tumours, difficulty in crossing the
blood–brain barrier, tumour heterogeneity, imaging challenges for monitoring disease course, and a unique immune environment.6,7 With few treatment advances over the past two decades, the clinical trial landscape in glioblastoma has been characterised by long development times, low patient participation, problematic surrogate outcomes, and poor decision making.8,9 Indeed, poor early phase decision making has been repeatedly high-lighted as a major problem in the development of therapeutics10 and continues to increase interest in novel clinical trial designs.Randomised controlled trials (RCTs) are the gold standard for clinical experimentation and evaluation of therapies. RCTs control for systematic bias from known and unknown confounders by randomly assigning
patients to receive an experimental therapy or standard of care, which allows for the evaluation of treatment effects. However, RCTs can be difficult to conduct in some neuro-oncology settings. A small percentage of patients with cancer participate in clinical trials,11 and RCTs can have slow patient accrual because of patient reluctance to enrol in studies with a control group, which is a pronounced problem in settings with ineffective standards of care, such as recurrent glioblastoma.12–14 Precision medicine further complicates this issue by focusing trials on biomarker-defined subgroups of patients who might benefit from targeted therapies.15 These subgroups are often small, resulting in substantial challenges to conducting RCTs with adequate sample sizes to detect treatment effects.15,16The design and implementation of clinical trials that leverage external datasets, with patient-level information on pretreatment clinical profiles and outcomes to support testing of experimental therapies and study decision
making, has attracted interest in neuro-oncology.17,18 A phase 2b trial in recurrent glioblastoma19 used a prespecified, eligibility-matched, external control group (including individual pretreatment profiles and outcomes), with data from patients with glioblastoma from major neurosurgery centres as a comparator group to evaluate MDNA55. After implementation of this trial design, investigators reported improved survival in patients receiving MDNA55 compared with the external control group.19 Several ongoing neuro-oncology trials are actively exploring similar approaches to use external data in the design and analysis of clinical studies.The Society of Neuro-Oncology hosted the 2020 Clinical Trials Think Tank on Nov 6, 2020, with a virtual session dedicated to trial designs leveraging external data. Experts in the field of neuro-oncology were paired with experts in data science and biostatistics, and representatives from industry, patient advocacy, and the FDA. The interdisciplinary session focused on challenges in drug development, data sharing and access, regulatory considerations of novel trial designs, and emerging methodological approaches for using external data. Although participation was broad in terms of expertise, most participants were from the USA and provided a US perspective on the topic. The discussion from the Clinical
e457 www.thelancet.com/oncology Vol 22 October 2021SeriesCalifornia Los Angeles, Los Angeles, CA, USA (Prof T Cloughesy MD); Foundation Medicine, Cambridge, MA, USA
(B M Alexander)Correspondence to: Dr Rifaquat Rahman, Department of Radiation Oncology, Dana-Farber/Brigham and Women’s Cancer Center, Harvard Medical School, Boston, MA 02215, USA
rrahman@bwh.harvard.edu
Trials Think Tank serves as a framework for this Series paper, which focuses on the use of external data to design, conduct, and analyse clinical trials with an emphasis on possible applications in neuro-oncology. We review trial designs, methodologies, approaches, regulatory considerations, and current barriers in data sharing and access.
Early phase trial designs
Early phase trials are typically designed to obtain preliminary estimates of treatment efficacy and toxic effects, which will inform the decision to pursue a phase 3 trial or stop drug development. Often in neuro-oncology, these early phase studies are single-arm trials that test the superiority of an experimental therapeutic compared with an established benchmark for the current standard of care (eg, median overall survival or other point estimates).20 Importantly, there can be substantial differences between populations or standards to assess outcomes across trials,21 which can lead to inappropriate comparisons and inadequate evaluations of the experimental therapy. Another major challenge with single-arm trials is the choice of a primary endpoint. Response rate might not correlate with overall survival and is difficult to interpret in glioblastoma,22 and single-arm studies are suboptimal for reliable inference on improvements in time-to-event endpoints, such as survival. On the basis of these known limitations, single-arm designs have been posited as a possible reason for poor decision making and unsuccessful phase 3 trials in glioblastoma.23,24The risk of biased conclusions from single-arm trials has been examined extensively and frameworks have been developed to help guide the choice between RCTs versus single-arm designs for glioblastoma.10,25 Despite well documented limitations, single-arm trials remain the most common early phase trial design in glioblastoma.25 Alternative trial designs have been proposed, including the incorporation of randomisation, seamless phase 2/3 study designs,26 and Bayesian outcome-adaptive trials to overcome limitations and
improve the evaluation of therapeutic candidates in the early phase of development.27–29
Overview of trial designs that include external data
Trial designs that leverage external data can generate valuable inferences in settings in which single-arm trials are suboptimal and RCTs are infeasible.30 External data can play a role in supporting key decisions in the drug development process, including regulatory approvals and decision making in early phase trials. The use of external patient-level datasets has the potential to improve accuracy of trial findings and inform decision making (eg, determining the sample size of a subsequent confirmatory phase 3 trial or selecting the phase 3 patient population). External data can also be incorporated into RCTs31—eg, within interim analyses18—although these designs remain largely unexplored.Externally augmented clinical trial design refers to the broad class of designs that leverage external data for decision making during a study or in the final analysis. These designs rely on access to well curated patient-level data for the standard of care from one or more relevant data sources to allow for adjustment of differences in pretreatment, covariates between enrolled patients and external data, and to derive treatment effect estimates. Given the need for statistical adjustments, the external dataset ideally includes a comprehensive set of potential confounders.32 Pretreatment covariates have been thoroughly studied for adult primary brain tumours.33,34An example of an externally augmented clinical trial design consists of a single-arm study combined with an external control group (ie, an external dataset with patient-level outcomes and pretreatment profiles), which is used as a comparator to evaluate the experimental treatment. This design (figure 1A), an externally controlled single-arm trial, infers the treatment effect by using adjustment methods to account for differences in patient pretreatment profiles between the external control group and the experimental group.35 In this design, the external control group is used to contextualise the outcome data from a single-arm study. Instead of using benchmark estimates (eg, median survival) for the standard of care, data analyses and treatment effect estimates are based on patient-level data from an external dataset.Hybrid randomised trial designs constitute another type of externally augmented clinical trial. With adequate external data and statistical planning,36 these designs have the potential of reducing sample size while maintaining the benefits of randomisation. An example of a two-stage hybrid design is shown in figure 1B. The study has an initial 1:1 randomisation to an internal control group and an experimental group. If the interim analysis does not identify differences between adjusted primary outcome distributions in the internal randomised control group and the external control group, then different randomisation ratios (eg, 2:5) can be used in the second stage of the study. If there is evidence of inconsistencies between the external and internal control groups (eg, unmeasured confounders or different measurement standards of outcomes and prognostic variables), the trial can continue with 1:1 randomisation. In the outlined example, the potential increase of randomisation probability for the experimental group can be attractive and might accelerate trial accrual. Indeed, patients with a brain tumour and an inadequate standard of care might be more likely to enrol in a trial if the probability of receiving the experimental therapy is higher.25
Externally augmented clinical trial designsAlong with high quality and complete data, a statistically rigorous study design is the most important element of an externally augmented clinical trial. As with any clinical trial, the design, including the sample size, a detailed
www.thelancet.com/oncology Vol 22 October 2021 e458Series
plan for interim decisions, and statistical methods for data analyses should be prespecified. Additionally, a plan for how missing data in the trial and external data sources will be handled is important. Potential distortion mechanisms that can bias treatment effect estimates and undermine the scientific validity of externally augmented clinical trial findings have been carefully examined and include unmeasured or misclassified confounders and data quality issues, such as the use of different standards to capture or measure outcomes.37–39Risks of introducing bias (table) and compromising the control of false positives and false negatives when using external patient-level data can differ substantially across externally augmented clinical trial designs, which span from single-arm studies (figure 1A) to hybrid randomised studies (figure 1B). Quantitative analyses of these and other risks (eg, exposure of patients to inferior treatments) are necessary before trial initiation. The decision to use external data should take account of several factors in addition to the study population and available patient-level datasets, including the stage of drug development (eg, early phase 2 vs confirmatory trials); the specific decisions (eg, early stopping of a phase 2 study for futility40 or sample size re-estimation during the study41) that will be supported by external data; resources (including maximum sample size); and the potential trial designs and statistical methods for data analysis.Candidate externally augmented clinical trial designs and statistical methods for data analysis can present markedly different trade-offs between potential efficiencies (eg, discontinuing early randomised trials of ineffective treatments by leveraging external data) and risks of poor operating characteristics (eg, bias and poor control of false positive results). The value of integrating external data is context specific and strictly dependent on the trial design and methods selected for data analysis and decision making.We describe three examples of externally augmented clinical trials with markedly different risks of poor operating characteristics. The purpose of these examples is to illustrate how external information can be used for making different decisions during or at completion of a trial.The first example is a single-arm trial with an external control group. We consider either binary primary out-comes (eg, tumour response) or time-to-event outcomes with censoring (eg, overall survival). This design uses procedures developed for observational studies,42 such as matching, propensity score methods,43,44 or inverse
A
External control
dataset
External control
dataset
Enrolled patients Enrolled patients
Analyses and treatment
effect estimate
Adjustment methods
(eg, matching)
B
Randomisation during
first phase of the trial (1:1)
Interim analysis Randomisation during
second phase of the trial
External control
dataset
Control group
Experimental
group
Do early control
data support the
use of external
data?
No
Yes
Altered
randomisation
(eg, 2:5)
Randomisation
(1:1)
C
Control group
External control
dataset
External
control
dataset
Experimental
group
Adjustment
methods
(eg, matching)
Interim and
futility
analysis
Final analysis
and treatment
estimate
Figure 1: Clinical trial designs
(A) A clinical study with patient enrolment to a single experimental group and an external control group (externally controlled single-arm trial). (B) A two-stage hybrid randomised trial design. (C) A randomised trial design that uses external data for interim futility analyses to support the decision to continue or discontinue the study. If the trial is not discontinued, the final analysis does not use external data.
e459 www.thelancet.com/oncology Vol 22 October 2021Series
probability weighting,45 which are applicable to the comparison of single-arm trial data (experimental treatment) with external patient-level data representative of the standard of care treatment (external control; figure 1A). These procedures have been developed to estimate treatment effects in non-randomised studies and have generated numerous contributions to the statistical literature.46 They compare outcome data Y in the experi-mental and control treatment groups with adjustments that account for confounders X.The literature on adjustment methods applicable to externally augmented clinical trial designs (eg, matching)43,44 is largely based on assumptions that are difficult to demonstrate,42 including the absence of unmeasured confounders.47 In the context of externally controlled single-arm trial designs, in which these assumptions might be violated, the investigator can attempt to evaluate the risk of bias and other statistical properties of treatment effect estimates (eg, the coverage of confidence intervals) computed using adjustment methods. Patient-level data from a library of completed RCTs in a specific clinical setting—eg, newly diagnosed glioblastoma patients—facilitate the comparison of externally controlled single-arm trial, RCT, and single-arm trial designs. For example, a treatment effect estimate computed using only data from a previously completed RCT can be compared with a second treatment effect estimate, computed using the experi-
mental group of the same RCT and external data (figure 2).48 This method can be repeated considering different RCTs, adjustment methods, and external datasets. These comparisons describe the consistency between RCT results and hypothetical results obtained from a smaller externally controlled single-arm trial (ie, the experimental group of the RCT, or part of it) leveraging external data; similar evaluation frameworks have been discussed elsewhere.49,50The leave-one-out algorithm is an alternative evaluation approach that requires a collection of completed RCTs with the same control treatment.17 The algorithm has been used to compare the application of candidate causal inference methods in externally controlled single-arm trials. This approach has been applied to a collection of studies in newly diagnosed glioblastoma and requires pretreatment profiles and outcomes from patients receiving standard of care (radiotherapy and temozolomide). This aspect is relevant because data access barriers can be substantially different for the control and experimental groups. The algorithm17 iterates the following three operations for each RCT in the data collection: (1) randomly selects n (the sample size of a hypothetical externally controlled single-arm trial) patients (without replacement) from the radiotherapy and temozolomide group (control) and uses patient pretreatment profiles (X) and outcomes (Y) of these
Randomised controlled trial
Control
group
Experimental
group
Experimental
group
External
data
Adjustment method
Treatment effect estimates
Single-arm trial plus external data
Randomised
controlled
trial
Single-arm
trial plus
external
data
Figure 2: Validation schema
A treatment effect estimate computed using only data from a previously completed randomised controlled trial is compared with a second treatment effect estimate, computed using the experimental group from the same randomised controlled trial and external control data.
Description Example Methods to avoid or reduce bias
Measured confounders The distributions of pretreatment patient characteristics that correlate with outcomes in the trial population and in the external control group are differentThe external control group has a higher Karnofsky performance status or age on average than the trial populationMatching; inverse probability weighting; marginal structural modelsUnmeasured confoundersThe distributions of unmeasured pretreatment patient characteristics that correlate with outcomes in the trial population and in the external control group are differentSupportive care (not captured in the datasets) differs between patients in the clinical trial and in the external control groupValidation analyses can indicate the risk of bias before the trial beginsDifferences in defining prognostic variables or outcomesThe definition of clinical measurements might vary between datasets, leading to differences in the definitions of outcomes or prognostic variables between the clinical trial and the external control groupMeasurement of tumour response with different response criteria or at different intervals in external control armsData dictionaries and validation analyses can reveal these discrepancies
before the trial beginsImmortal time bias In the external dataset the time-to-event outcome cannot occur during a time window because of the study design or other causesIn glioblastoma, different real-world datasets capture patient survival from diagnosis or from other timepoints, such as surgeryExplicit and detailed definitions of the time to event outcomes for the trial and external dataset can reveal the risk of bias
Table: Potential causes of bias in clinical trials with an external control group
www.thelancet.com/oncology Vol 22 October 2021 e460Series
patients to define a fictitious single-arm study (the experimental group); (2) the data on patients receiving radiotherapy and temozolomide in the remaining studies are used as external data, these datasets are combined into a single data matrix (the external control group); and (3) a treatment effect estimate is computed by comparing the fictitious externally controlled single-arm trial (step 1) and the external data (step 2), using a candidate adjustment method, which is also used to test the null hypothesis (H0—ie, the treatment does not improve the primary outcome).For each study in the data collection, steps 1–3 (similar to cross-validation) can be repeated to evaluate bias, variability of the treatment effect estimate, and the risk of false positive results. By design, the treatment effect in this fictitious comparison is null, because patients receiving radiotherapy and temozolomide are being compared with those receiving radiotherapy and temozolomide from different studies. This facilitates interpretability and produces bias summaries for the statistical plan of an externally controlled single-arm trial. An analysis using this leave-one-out algorithm approach in patients with newly diagnosed glioblastoma showed high false positive error rates in standard single-arm clinical trials (higher than the α level),10 which can be considerably reduced (up to 30%) by using external control data from previously completed trials.17The first validation approach (figure 2) attempts to replicate results of a completed RCT, whereas the leave-one-out algorithm approach is based on subsampling a control group. Both approaches are valuable strategies that can detect potential distortion mechanisms (eg, unmeasured confounders or inconsistent definitions of primary outcomes), which undermine the scientific validity of externally controlled single-arm trials. These approaches require patient-level data from several RCTs with adequate sample sizes to produce reliable analyses of
the risk of bias and false positive results in future externally controlled single-arm trials. Importantly, positive findings from retrospective analyses should not be overinterpreted using either approach because relevant changes in available treatments and technologies, or other factors, can rapidly make the entire data collection obsolete and inadequate.51The second type of externally augmented clinical trials are hybrid randomised trial designs that combine external and randomised control data to estimate potential treatment effects.52 Figure 1B represents a two-stage hybrid design. In the first stage, patients are randomly assigned to the experimental group and the internal control arm. Interim analysis can be used for early stopping for futility and to determine sample sizes for the experimental and control groups in the second stage. These decisions are based on a similarity measure comparing estimates of conditional outcome distributions (Pr[Y|X]) of the external and internal control groups and preliminary treatment effect estimates. The proportion of patients randomised to the internal control group during the second stage can be reduced or increased on the basis of a prespecified interim analysis, which involves sum-maries in support of or against the integration of external data to estimate the effects of the experimental treatment.On the basis of results from a phase 2 study of MDNA55,19 investigators are planning an open-label, phase 3 registration study with implementation of a hybrid randomised design in patients with recurrent IDH-wild-type glioblastoma. The study team is considering a 3:1 (experimental to control) randomisation ratio with a final comparison of overall survival between patients receiving MDNA55 and the control groups (external and internal).53The third type of externally augmented clinical trials incorporates external data to support futility analyses. External data can be incorporated into RCTs for use in interim decisions.18,31 In such a design (figure 1C), interim analyses use predictions based on early data from the RCT combined with external data. These predictions express the probability that the trial will generate significant evidence of positive treatment effects. The trial is discontinued by design if the predictive probability becomes smaller than a fixed threshold. After completing the enrolment and follow-up phases, the final analysis does not use external data. The primary endpoint of the trial is positive (indicating improved outcomes with the experimental treatment) if the p value, computed using only RCT data (excluding external information), has a value lower than the targeted control of false positive results α.In ideal settings, without unmeasured confounders and other distortion mechanisms, leveraging external data for interim futility analyses can reduce the expected sample size when the experimental treatment is ineffective and can reduce early stopping probability
when the experimental therapy is superior, thus increasing the power.18 Additionally, the outlined design maintains a rigorous control of type I error probability, even in the presence of unmeasured confounders, because the external data are excluded from the final data analyses. The efficiency gains and risks associated with integration of external data into interim decisions have been quantified for trials of patients with newly diagnosed glioblastoma using evaluation analyses that built on a collection of datasets from completed RCTs, the leave-one-out algorithm, and other similar procedures.18As externally augmented clinical trials require numerous context-specific considerations, from relevant aspects of the external datasets to the feasibility of alternative designs, a discussion with regulatory agencies in early stages of trial planning is strongly recommended.
External data sourcesThe use of external controls to evaluate new treatments is
dependent on the availability of high-quality external data. Selecting appropriate datasets is crucial and checklists have been developed to provide guidance on data quality.54
e461 www.thelancet.com/oncology Vol 22 October 2021Series
Data considerations for external controls include appropriate capture of patient-level data,55 consistent definition of covariates and endpoints, and adequate temporality of the data, as small temporal lags can substantially affect the trial analysis.49 Investigators should consider potential biases that occur if the endpoint definitions are inconsistent across studies. For example, survival can be measured from the date of diagnosis, randomisation, or adjuvant treatment start. The definition of the outcome should be explicit and consistent during the trial and in the external datasets.56 Missing data are another important consideration in analyses with external data.57,58 Although there are methods to address missing data (eg, multiple imputation and likelihood-based methods), their use within externally augmented clinical trial designs has not been extensively studied. In general, the external control population and trial population should be similar to reduce the risk of bias. Potential unmeasured confounders, inconsistencies in definitions, and different measurement standards of covariates and outcomes across datasets need to be scrutinised using data dictionaries and study protocols. Contemporaneous controls are ideal, but historical controls with patient-level data might be helpful in appropriate contexts. For example, disease settings without a recent change in the standard of care (eg, glioblastoma) or a long track record of time-stable outcomes might have more flexibility in the temporality of data, but this benefit should be weighed against the possibility of unmeasured aspects of care such as advances in imaging, radiotherapy, surgical techniques, and supportive care that might change over time.59 Additionally, the data should be traceable (ie, an audit trail should be available to document data management processes) to support a marketing application.54The two most relevant sources of external data are previously completed clinical trials and real-world data derived from clinical practice. The use of data from clinical trials can be advantageous, given that the data are typically collected in a rigorous environment with vetting procedures. Clinical trials are often done in specialised institutions and enrol clearly defined patient populations. Data from patients previously enrolled on RCTs and given standard of care might be more likely than real-world data cohorts to contain pretreatment profiles similar to those of patients who will be enrolled in future trials. The use of data collection forms, intensive monitoring, and specialised personnel enable adherence to clear protocols to produce standardised data.15 Data from previous trial participants exist, as there have been several negative phase 3 randomised trials in glioblastoma, sponsored by cooperative groups and industry, with many patients receiving the current standard of care (radiotherapy and temozolomide).60–63 Data access, however, can be challenging because of impediments to data sharing64 and contemporary trial data might not be made available by trial sponsors.Real-world data is derived from registries, claims and billing data, personal devices and applications, or electronic health records. Because real-world data is generally not collected for research purposes, there can be concerns about data organisation and quality, confounding, selection mechanisms, and ultimately, bias.65–67 Advances in the quality of electronic health records data have created opportunities, and new curated datasets can be linked to molecular or radiological data with high fidelity. Efforts to harmonise real-world data from disparate data sources and novel methods to incorporate such data into clinical studies provide an avenue to inform trial design68 and regulatory decision making,69 but further work is required to validate these approaches.Although differences between real-world data and data from clinical studies have been reported,70 methodological work on the use of joint models and analyses to compensate for the scarcity of trial data and the potential distortions of real-world data (eg, measurement errors or unknown selection mechanisms) is in early stages.Methodological work has primarily focused on overall survival,17,48 which is more likely to be adequately captured in external datasets relative to other outcomes. Radiological endpoints, such as progression-free survival, require caution because of the risk of inconsistent assessments across datasets. In real-world data, radiological outcomes might not be determined by standardised response assessment criteria and central radiological review would probably not be routinely implemented. Although brain tumour trials71,72 often use consensus guidelines by the Response Assessment in Neuro-Oncology working group, these criteria include subjective components,73 and datasets of previously completed trials can still include misclassification errors. Quality of life, neurological function, and neurocognitive outcomes are increasingly incorporated into clinical trials.72,74 These non-survival outcomes can provide meaningful measures of the clinical benefit of a therapy and serve as valuable endpoints in neuro-oncology trials.75 Nevertheless, missing measurements of these outcomes are common across datasets from completed trials and real-world data. Exploration of the use of external data to analyse radiographical outcomes, patient-centred outcomes, and safety outcomes in neuro-oncology trials remains scarce.
External control groups beyond neuro-oncologyCarrigan and colleagues48 leveraged a curated real-world data dataset of 48 856 patients (electronic health records from a Flatiron database) to reanalyse 11 completed trials in advanced non-small-cell lung cancer. In this study, the external control groups were defined with matching methods. The external control groups were able to recapitulate treatment effect estimates (hazard ratios) for ten of 11 RCTs. This finding suggests that real-world data can potentially be used as external controls partly because of the large number of patients
www.thelancet.com/oncology Vol 22 October 2021 e462Series
in these datasets, as many patient records can meet the RCT inclusion and exclusion criteria and be used as comparators. Notably, the external control groups derived from electronic health records did not recapitulate the results for one RCT. On inspection, the authors of the study felt this discordance was associated with a biomarker subgroup population that was not sufficiently represented in the electronic health record dataset. These findings underscore the need to account for biomarkers and well represented sub populations in external control groups.A study by Ray and colleagues76 provides an example of the use of external data to contextualise a single-arm trial. In an FDA led retrospective analysis, the outcome of invasive disease-free survival from a single-arm study77 of adjuvant paclitaxel and trastu zumab in patients with HER2-positive early breast cancer was analysed using an external control group derived from clinical trials (control therapies included anthracycline, cyclophosphamide, taxane, and trastuzumab; or taxane, carboplatin, and trastuzumab).78 The de-escalated regimen that combines adjuvant paclitaxel and trastuzumab had been adopted in clinical practice on the basis of initial single-arm trial results.77 This retrospective analysis used propensity score matching to adjust for differences in pretreatment patient profiles in the single-arm trial and the external control dataset. The analysis estimated similar outcome distributions for adjuvant paclitaxel and trastuzumab and the control, which supported the use of a de-escalated regimen, particularly in light of higher toxic effects with the control regimens.
Considerations and implications for regulatory decision makingIn the USA, the 21st Century Cures Act directed the FDA to develop guidance for evaluation and use of real-world data, and to consider the potential role of real-world data in drug development and regulatory decision making. For example, real-world data could be used to support approvals for new indications or be integrated into existing monitoring requirements after approval.79 Accordingly, the FDA launched a real-world data programme to lay the foundation for rigorous use of such data in regulatory decisions.80 Several ongoing initiatives are providing guidance on data quality, data standards, and study designs that incorporate real-world data.81 Also, other regulatory institutions, such as the European Medicines Agency and Health Canada, have shown an openness towards better understanding and potentially leveraging real-world data for drug development.82,83Within a regulatory scope, the use of external control data can support expedited approval, extend on-label use of a therapy to a new indication or subgroup, and more generally support regulatory decision making.84 For example, in terms of rare diseases, the FDA approved blinatumomab for adults with relapsed or refractory acute lymphocytic leukaemia on the basis of a study that used data from a previous clinical trial as a comparator.85In a regulatory context, there is a high burden of proof for investigators to show the scientific rigor of study designs and analyses that leverage external control data with an appropriate risk level. Comparative analyses with standard RCT designs are fundamental to evaluate robustness and efficiencies of externally augmented clinical trial designs. The external data, study design, and analysis should be tailored to each specific clinical context and intended regulatory use. Each element should be carefully considered and scrutinised when evaluating the risks of biased treatment effect estimates and inadequate control of false positive findings.
Data sharing modelsDespite the appeal of patient-level data from previous clinical trials, data access is a barrier to studying and implementing externally augmented clinical trial designs in early phase and late phase trials. Data sharing efforts from industry funded RCTs are increasing, but substantial room for improvement remains.86,87 Beyond implications for externally augmented clinical trials, clinical trial data sharing allows investigators to carry out analyses that generate new knowledge and which have been deemed “essential for expedited translation of research results into knowledge, products, and procedures to improve human health” by the National Institutes of Health.88 Substantial challenges and appropriate concerns about data sharing remain,89 including the need to ensure patient privacy and academic credit; the use of adequate standards for combining data from different sources; and the allotment of resources required to deidentify patient records and provide infrastructures for data sharing. The patient perspective serves as an important counterpoint, assuming that privacy is protected; studies indicate that patients are in favour of data sharing for purposes that can help to advance clinical outcomes.90Advances towards simpler data access could transform the ability to do secondary analyses91 and leverage external data in future clinical studies. For many data sharing platforms, a gatekeeper model is used, often with long approval processes, restrictive criteria for data access, and limitations on data use. These requirements act as a mechanism of passive resistance and delay access to data from completed trials. An increasing number of data sharing platforms such as Vivli,92 YODA,93 and Project Data Sphere,94 are aligned with more open sharing models for clinical trial datasets.95 Nonetheless, most data from previously completed neuro-oncology trials remain difficult to access.New policies might be necessary for data sharing and to accelerate the study of new therapeutics. An important consideration is the modification of incentives for data sharing.96 A systematic effort from cooperative groups, industry, academia, and other stakeholders could help to achieve this goal. Regulatory requirements that ensure
e463 www.thelancet.com/oncology Vol 22 October 2021Series
timely data sharing and patient advocacy groups could play key roles in hastening this process. Additionally, initiatives and agreements to prospectively share patient-level data from the control groups of multiple cooperative RCTs could be beneficial to participating studies and create opportunities to extend data sharing.
ConclusionAt the end of our Clinical Trials Think Tank session, there was a strong interest and desire to continue to collaborate and investigate, validate, and implement externally augmented clinical trial designs in glio-blastoma and, on a broader level, in neuro-oncology. Efforts to form industry cooperative group partnerships and selection of datasets and statistical methods were set as goals to continue towards an advancement in understanding the role of externally augmented clinical trials for drug development in neuro-oncology.The use of external data to design and analyse clinical studies has the potential to accelerate drug development and can contribute to rigorous evaluation of new
treatments. RCTs will remain the indisputable gold standard for evaluation of treatments, but external datasets can supplement information gathered from RCTs and single-arm studies. Further methodological work can help to identify the appropriate clinical contexts, data, and statistical designs for externally augmented clinical trials that generate inference on treatment effects of experimental therapies, with well controlled risks regarding their their accuracy and scientific validity.There is a continuum of approaches for leveraging external data, and use of externally augmented clinical trial designs should be tailored to the disease context. An emphasis on high-quality patient-level data, rigorous methods, and biostatistical expertise are crucial in the successful implementation of externally augmented clinical trials. Data access to previously completed clinical
trials and real-world data is improving, but new policies and initiatives for data sharing could further unlock the value of external data. Continued collab orations between stakeholders, including indus try, academia, biostatisticians, clinicians, regulatory agencies, and patient advocates, are crucial to understand the appropriate use of externally augmented clinical trial designs in neuro-oncology.
Search strategy and selection criteria
We searched PubMed for articles using the search terms “external control arms”, “synthetic control arms”, “neuro-oncology trial design”, “glioblastoma trial design”, published from Jan 1, 2000, to May 1, 2021. Articles were also identified through searches of the authors’ own files. Only papers published in English were reviewed. The final reference list was generated on the basis of relevance to the scope of this Series paper.
Contributors
RR, PYW, BMA, and LT conceptualised this Series paper and wrote the first draft. All authors contributed to the interpretation of findings and editing. All authors approved the final manuscript.
Declaration of interests
RR received research support from the Project Data Sphere, outside of submitted work. IR-R reports employment and owns stocks of Roche and Genentech. FM reports employment at Medicenna Therapeutics. LEA reports employment and owns stocks of Novartis. JEA reports employment and owns stocks of Chimerix. LKA and EA-C report employment at Candel Therapuetics. SB reports grants and personal fees from Novocure; grants from Incyte, GSK, and Eli Lilly; and personal fees from Bayer and Sumitomo Dainippon. MK reports personal fees from Ipsen, Pfizer, Roche, and Jackson Laboratory for Genomic Medicine and research funding paid to his institution from Specialised Therapeutics. TC reports personal fees from Roche, Trizel, Medscape, Bayer, Amgen, Odonate Therapeutics, Pascal Biosciences, DelMar Pharmaceuticals, Tocagen, Karyopharm, GW Pharmaceuticals, Kiyatec, AbbVie, Boehinger Ingelheim, VBI Vaccines, Dicephera, VBL Therapeutics, Agios, Merck, Genocea, Puma, Lilly, Bristol Myers Squibb, Cortice, Wellcome Trust; and stock options from Notable Labs. TC has a patent (62/819,322) with royalties paid to Katmai and is a board member for the 501c3 Global Coalition for Adaptive Research. PYW reports personal fees from Abbvie, Agios, AstraZeneca, Blue Earth Diagnostics, Eli Lilly, Genentech, Roche, Immunomic Therapeutics, Kadmon, Kiyatec, Merck, Puma, Vascular Biogenics, Taiho, Tocagen, Deciphera, and VBI Vaccines; and research support from Agios, AstraZeneca, Beigene, Eli Lily, Genentech, Roche, Karyopharm, Kazia, MediciNova, Merck, Novartis, Oncoceutics, Sanofi-Aventis, and VBI Vaccines. BMA reports employment at Foundation Medicine; personal fees from AbbVie, Bristol Myers Squibb, Precision Health Economics, and Schlesinger Associates; and research support from Puma, Eli Lilly, Celgene. SV, JM, BL, M-YCP, DA, KT, and LT declare no competing interests.
Acknowledgments
We thank Amy Barone, Pallavi Mishra-Kalyani, and the Food and Drug Administration for contributing and helping with preparation of this Series paper. We also thank the Society for Neuro-Oncology and their staff for arrangements and coordination of the 2020 Clinical Trials Think Tank meeting. We thank Johnathan Rine for help with preparation of the figures. RR was supported by the Joint Center for Radiation Therapy Foundation grant. LT and SV were supported by the National Institutes of Health (grant 1R01LM013352–01A1).
References
1 Hwang TJ, Carpenter D, Lauffenburger JC, Wang B, Franklin JM, Kesselheim AS. Failure of investigational drugs in late-stage clinical development and publication of trial results. JAMA Intern Med 2016; 176: 1826–33.2 Wong CH, Siah KW, Lo AW. Estimation of clinical trial success rates and related parameters. Biostatistics 2019; 20: 273–86.3 Hu M, Yang M, Liu Y. Statistical adaptation to oncology drug development evolution. Contemp Clin Trials 2020; 99: 106180.4 Mandrekar SJ, Sargent DJ. Clinical trial designs for predictive biomarker validation: theoretical considerations and practical challenges. J Clin Oncol 2009; 27: 4027–34.5 Alexander BM, Cloughesy TF. Adult glioblastoma. J Clin Oncol 2017; 35: 2402–09.6 Wen PY, Weller M, Lee EQ, et al. Glioblastoma in adults: a Society for Neuro-Oncology (SNO) and European Society of Neuro-Oncology (EANO) consensus review on current management and future directions. Neuro Oncol 2020; 22: 1073–113.7 Barbaro M, Fine HA, Magge RS. Scientific and clinical challenges within neuro-oncology. World Neurosurg 2021; published online Feb 18. https://doi.org/10.1016/j.wneu.2021.01.151.8 Vanderbeek AM, Rahman R, Fell G, et al. The clinical trials landscape for glioblastoma: is it adequate to develop new treatments?
Neuro Oncol 2018; 20: 1034–43.9 Lee EQ, Chukwueke UN, Hervey-Jumper SL, et al. Barriers to accrual and enrollment in brain tumor trials. Neuro Oncol 2019;
21: 1100–17.
www.thelancet.com/oncology Vol 22 October 2021 e464Series
10 Vanderbeek AM, Ventz S, Rahman R, et al. To randomize, or not to randomize, that is the question: using data from prior clinical trials to guide future designs. Neuro Oncol 2019; 21: 1239–49.11 Ostrom QT, Patil N, Cioffi G, Waite K, Kruchko C, Barnholtz-Sloan JS. CBTRUS statistical report: primary brain and other central nervous system tumors diagnosed in the United States in 2013–2017. Neuro Oncol 2020; 22 (suppl 2): iv1–96.12 Eborall HC, Stewart MCW, Cunningham-Burley S, Price JF, Fowkes FGR. Accrual and drop out in a primary prevention randomised controlled trial: qualitative study. Trials 2011; 12: 7.13 Featherstone K, Donovan JL. “Why don’t they just tell me straight, why allocate it?” The struggle to make sense of participating in a randomised controlled trial. Soc Sci Med 2002; 55: 709–19.14 Feinberg BA, Gajra A, Zettler ME, Phillips TD, Phillips EG, Kish JK. Use of real-world evidence to support FDA approval of oncology drugs. Value Health 2020; 23: 1358–65.15 Eichler H-G, Pignatti F, Schwarzer-Daum B, et al. Randomized controlled trials versus real world evidence: neither magic nor myth. Clin Pharmacol Ther 2021; 109: 1212–18.16 Berry DA. The brave new world of clinical cancer research: adaptive biomarker-driven trials integrating clinical practice with clinical research. Mol Oncol 2015; 9: 951–59.17 Ventz S, Lai A, Cloughesy TF, Wen PY, Trippa L, Alexander BM. Design and evaluation of an external control arm using prior clinical trials and real-world data. Clin Cancer Res 2019;
25: 4993–5001.18 Ventz S, Comment L, Louv B, et al. The use of external control data for predictions and futility interim analyses in clinical trials.
Neuro Oncol 2021; published online June 9. https://doi.org/10.1093/neuonc/noab141.19 Sampson JH, Achrol A, Aghi MK, et al. MDNA55 survival in recurrent glioblastoma (rGBM) patients expressing the interleukin-4 receptor (IL4R) as compared to a matched synthetic control. Proc Am Soc Clin Oncol 2020; 38 (suppl): 2513 (abstr).20 Thall PF, Simon R. Incorporating historical control data in planning phase II clinical trials. Stat Med 1990; 9: 215–28.21 Neuenschwander B, Capkun-Niggli G, Branson M, Spiegelhalter DJ. Summarizing historical information on controls in clinical trials. Clin Trials 2010; 7: 5–18.22 Reardon DA, Galanis E, DeGroot JF, et al. Clinical trial end points for high-grade glioma: the evolving landscape. Neuro Oncol 2011;
13: 353–61.23 Sharma MR, Karrison TG, Jin Y, et al. Resampling phase III data to assess phase II trial designs and endpoints. Clin Cancer Res 2012;
18: 2309–15.24 Tang H, Foster NR, Grothey A, Ansell SM, Goldberg RM, Sargent DJ. Comparison of error rates in single-arm versus randomized phase II cancer clinical trials. J Clin Oncol 2010;
28: 1936–41.25 Grossman SA, Schreck KC, Ballman K, Alexander B. Point/counterpoint: randomized versus single-arm phase II clinical trials for patients with newly diagnosed glioblastoma. Neuro Oncol 2017; 19: 469–74.26 Stallard N, Todd S. Seamless phase II/III designs.
Stat Methods Med Res 2011; 20: 623–34.27 Alexander BM, Trippa L, Gaffey S, et al. Individualized screening trial of innovative glioblastoma therapy (INSIGhT): a Bayesian adaptive platform trial to develop precision medicines for patients with glioblastoma. JCO Precis Oncol 2019; 3: 1–13.28 Alexander BM, Ba S, Berger MS, et al. Adaptive global innovative learning environment for glioblastoma: GBM AGILE.
Clin Cancer Res 2018; 24: 737–43.29 Buxton MB, Alexander BM, Berry DA, et al. GBM AGILE: a global, phase II/III adaptive platform trial to evaluate multiple regimens in newly diagnosed and recurrent glioblastoma.
Proc Am Soc Clin Oncol 2020; 38 (suppl): TPS2579 (abstr).30 Thorlund K, Dron L, Park JJH, Mills EJ. Synthetic and external controls in clinical trials—a primer for researchers. Clin Epidemiol 2020; 12: 457–67.31 Viele K, Berry S, Neuenschwander B, et al. Use of historical control data for assessing treatment effects in clinical trials. Pharm Stat 2014; 13: 41–54.32 VanderWeele TJ, Shpitser I. On the definition of a confounder.
Ann Stat 2013; 41: 196–220.33 Pignatti F, van den Bent M, Curran D, et al. Prognostic factors for survival in adult patients with cerebral low-grade glioma.
J Clin Oncol 2002; 20: 2076–84.34 Gittleman H, Lim D, Kattan MW, et al. An independently validated nomogram for individualized estimation of survival among patients with newly diagnosed glioblastoma: NRG Oncology RTOG 0525 and 0825. Neuro Oncol 2017; 19: 669–77.35 Davi R, Mahendraratnam N, Chatterjee A, Dawson CJ, Sherman R. Informing single-arm clinical trials with external controls.
Nat Rev Drug Discov 2020; 19: 821–22.36 Normington J, Zhu J, Mattiello F, Sarkar S, Carlin B. An efficient Bayesian platform trial design for borrowing adaptively from historical control data in lymphoma. Contemp Clin Trials 2020;
89: 105890.37 Webster-Clark M, Jonsson Funk M, Stürmer T. Single-arm trials with external comparators and confounder misclassification: how adjustment can fail. Med Care 2020; 58: 1116–21.38 Thompson D. Replication of randomized, controlled trials using real-world data: what could go wrong? Value Health 2021; 24: 112–15.39 Seeger JD, Davis KJ, Iannacone MR, et al. Methods for external control groups for single arm trials or long-term uncontrolled extensions to randomized clinical trials.
Pharmacoepidemiol Drug Saf 2020; 29: 1382–92.40 Snapinn S, Chen M-G, Jiang Q, Koutsoukos T. Assessment of futility in clinical trials. Pharm Stat 2006; 5: 273–81.41 Gould AL. Sample size re-estimation: recent developments and practical considerations. Stat Med 2001; 20: 2625–43.42 Imbens GW, Rubin DB. Causal inference for statistics, social, and biomedical sciences: an introduction, 1st edn. New York, NY, USA: Cambridge University Press, 2015.43 Rubin DB. The use of matched sampling and regression adjustment to remove bias in observational studies. Biometrics 1973; 29: 185–203.44 Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika 1983; 70: 41–55.45 Li L, Greene T. A weighting analogue to pair matching in propensity score analysis. Int J Biostat 2013; 9: 215–34.46 Pearl J. An introduction to causal inference. Int J Biostat 2010; 6: 7.
47 Lin W-J, Chen JJ. Biomarker classifiers for identifying susceptible subpopulations for treatment decisions. Pharmacogenomics 2012;
13: 147–57.48 Carrigan G, Whipple S, Capra WB, et al. Using electronic health records to derive control arms for early phase single-arm lung cancer trials: proof-of-concept in randomized controlled trials.
Clin Pharmacol Ther 2020; 107: 369–77.49 Abrahami D, Pradhan R, Yin H, Honig P, Baumfeld Andre E, Azoulay L. Use of real-world data to emulate a clinical trial and support regulatory decision making: assessing the impact of temporality, comparator choice, and method of adjustment.
Clin Pharmacol Ther 2021; 109: 452–61.50 Franklin JM, Patorno E, Desai RJ, et al. Emulating randomized clinical trials with nonrandomized real-world evidence studies: first results from the RCT DUPLICATE initiative. Circulation 2021;
143: 1002–13.51 Beaulieu-Jones BK, Finlayson SG, Yuan W, et al. Examining the use of real-world evidence in the regulatory process.
Clin Pharmacol Ther 2020; 107: 843–52.52 Hobbs BP, Carlin BP, Sargent DJ. Adaptive adjustment of the randomization ratio using historical control data. Clin Trials 2013;
10: 430–40.53 Medicenna. Medicenna provides MDNA55 rGBM clinical program update following positive end of phase 2 meeting with the U.S. Food and Drug Administration (FDA). 2020. Medicenna Provides MDNA55 rGBM Clinical Program Update Following Positive End of Phase 2 Meeting with the U.S. Food and Drug Administration (FDA) - Medicenna Therapeutics (accessed June 4, 2021).54 Miksad RA, Abernethy AP. Harnessing the power of Real-World Evidence (RWE): a checklist to ensure regulatory-grade data quality.
Clin Pharmacol Ther 2018; 103: 202–05.55 Ghadessi M, Tang R, Zhou J, et al. A roadmap to using historical controls in clinical trials—by Drug Information Association Adaptive Design Scientific Working Group (DIA-ADSWG).
Orphanet J Rare Dis 2020; 15: 69.56 Backenroth D. How to choose a time zero for patients in external control arms. Pharm Stat 2021; 20: 783–92.
57 Kilburn LS, Aresu M, Banerji J, Barrett-Lee P, Ellis P, Bliss JM. Can routine data be used to support cancer clinical trials? A historical baseline on which to build: retrospective linkage of data from the TACT (CRUK 01/001) breast cancer trial and the National Cancer Data Repository. Trials 2017; 18: 561.58 Little RJ, D’Agostino R, Cohen ML, et al. The prevention and treatment of missing data in clinical trials. N Engl J Med 2012;
367: 1355–60.59 Basch E, Deal AM, Dueck AC, et al. Overall survival results of a trial assessing patient-reported outcomes for symptom monitoring during routine cancer treatment. JAMA 2017; 318: 197–98.60 Gilbert MR, Dignam J, Won M, et al. RTOG 0825: phase III double-blind placebo-controlled trial evaluating bevacizumab (Bev) in patients (Pts) with newly diagnosed glioblastoma (GBM).
Proc Am Soc Clin Oncol 2013; 31 (suppl): 1 (abstr).61 Gilbert MR, Dignam JJ, Armstrong TS, et al. A randomized trial of bevacizumab for newly diagnosed glioblastoma. N Engl J Med 2014;
370: 699–708.62 Chinot OL, Wick W, Mason W, et al. Bevacizumab plus radiotherapy-temozolomide for newly diagnosed glioblastoma.
N Engl J Med 2014; 370: 709–22.63 Weller M, Butowski N, Tran DD, et al. Rindopepimut with temozolomide for patients with newly diagnosed, EGFRvIII-expressing glioblastoma (ACT IV): a randomised, double-blind, international phase 3 trial. Lancet Oncol 2017; 18: 1373–85.64 Mbuagbaw L, Foster G, Cheng J, Thabane L. Challenges to complete and useful data sharing. Trials 2017; 18: 71.65 Pearl J. Causality: models, reasoning and inference, 2nd edn. New York, NY, USA: Cambridge University Press, 2009.66 Collins R, Bowman L, Landray M, Peto R. The magic of randomization versus the myth of real-world evidence. N Engl J Med 2020; 382: 674–78.67 Larrouquere L, Giai J, Cracowski J-L, Bailly S, Roustit M. Externally controlled trials: are we there yet? Clin Pharmacol Ther 2020;
108: 918–19.68 Liu R, Rizzo S, Whipple S, et al. Evaluating eligibility criteria of oncology trials using real-world data and AI. Nature 2021;
592: 629–33.69 Corrigan-Curay J, Sacks L, Woodcock J. Real-world evidence and real-world data for evaluating drug safety and effectiveness. JAMA 2018; 320: 867–68.70 Unger JM, Barlow WE, Martin DP, et al. Comparison of survival outcomes among cancer patients treated in and out of clinical trials.
J Natl Cancer Inst 2014; 106: dju002.71 Chukwueke UN, Wen PY. Use of the Response Assessment in Neuro-Oncology (RANO) criteria in clinical trials and clinical practice. CNS Oncol 2019; 8: CNS28.72 Wen PY, Chang SM, Van den Bent MJ, Vogelbaum MA, Macdonald DR, Lee EQ. Response assessment in neuro-oncology clinical trials. J Clin Oncol 2017; 35: 2439–49.73 Huang RY, Neagu MR, Reardon DA, Wen PY. Pitfalls in the neuroimaging of glioblastoma in the era of antiangiogenic and immuno/targeted therapy—detecting illusive disease, defining response. Front Neurol 2015; 6: 33.74 Gilbert MR, Rubinstein L, Lesser G. Creating clinical trial designs that incorporate clinical outcome assessments. Neuro Oncol 2016;
18 (suppl 2): ii21–25.75 Blakeley JO, Coons SJ, Corboy JR, Kline Leidy N, Mendoza TR, Wefel JS. Clinical outcome assessment in malignant glioma trials: measuring signs, symptoms, and functional limitations.
Neuro Oncol 2016; 18 (suppl 2): ii13–20.76 Ray EM, Carey LA, Reeder-Hayes KE. Leveraging existing data to contextualize phase II clinical trial findings in oncology. Ann Oncol 2020; 31: 1591–93.77 Tolaney SM, Barry WT, Dang CT, et al. Adjuvant paclitaxel and trastuzumab for node-negative, HER2-positive breast cancer.
N Engl J Med 2015; 372: 134–41.78 Amiri-Kordestani L, Xie D, Tolaney SM, et al. A Food and Drug Administration analysis of survival outcomes comparing the adjuvant paclitaxel and trastuzumab trial with an external control from historical clinical trials. Ann Oncol 2020; 31: 1704–08.79 Franklin JM, Glynn RJ, Martin D, Schneeweiss S. Evaluating the use of nonrandomized real-world data analyses for regulatory decision making. Clin Pharmacol Ther 2019; 105: 867–77.80 US Food and Drug Administration. Framework for FDA’s real-world evidence program. 2018. https://www.fda.gov/media/120060/download (accessed June 2, 2021).81 US Food and Drug Administration. Complex innovative trial design pilot meeting program. 2021. Complex Innovative Trial Design Pilot Meeting Program (accessed June 26, 2021).82 European Medicines Agency. EMA regulatory science to 2025. 2020. https://www.ema.europa.eu/en/docume...tory-science-2025-strategic-reflection_en.pdf (accessed June 3, 2021).83 Government of Canada. Strengthening the use of real world evidence for drugs. 2018. Regulatory review of drugs and devices: Strengthening the use of real world evidence for drugs - Canada.ca (accessed June 3, 2021).84 Burcu M, Dreyer NA, Franklin JM, et al. Real-world evidence to support regulatory decision-making for medicines: considerations for external control arms. Pharmacoepidemiol Drug Saf 2020; 29: 1228–35.85 Gökbuget N, Kelsh M, Chia V, et al. Blinatumomab vs historical standard therapy of adult relapsed/refractory acute lymphoblastic leukemia. Blood Cancer J 2016; 6: e473.86 Boutron I, Dechartres A, Baron G, Li J, Ravaud P. Sharing of data from industry-funded registered clinical trials. JAMA 2016;
315: 2729–30.87 Miller J, Ross JS, Wilenzick M, Mello MM. Sharing of clinical trial data and results reporting practices among large pharmaceutical companies: cross sectional descriptive study and pilot of a tool to improve company practices. BMJ 2019; 366: l4217.88 National Institutes of Health. Final NIH statement on sharing research data. 2003. NIH Guide: FINAL NIH STATEMENT ON SHARING RESEARCH DATA (accessed June 15, 2021).89 Longo DL, Drazen JM. Data sharing. N Engl J Med 2016;
374: 276–77.90 Mello MM, Lieou V, Goodman SN. Clinical trial participants’ views of the risks and benefits of data sharing. N Engl J Med 2018;
378: 2202–11.91 Arfè A, Ventz S, Trippa L. Shared and usable data from phase 1 oncology trials-an unmet need. JAMA Oncol 2020; 6: 980–81.92 Bierer BE, Li R, Barnes M, Sim I. A global, neutral platform for sharing trial data. N Engl J Med 2016; 374: 2411–13.93 Krumholz HM, Waldstreicher J. The Yale Open Data Access (YODA) project—a mechanism for data sharing. N Engl J Med 2016;
375: 403–05.94 Bertagnolli MM, Sartor O, Chabner BA, et al. Advantages of a truly open-access data-sharing model. N Engl J Med 2017; 376: 1178–81.95 Pisani E, Aaby P, Breugelmans JG, et al. Beyond open data: realising the health benefits of sharing data. BMJ 2016; 355: i5295.96 Lo B, DeMets DL. Incentives for clinical trialists to share data.
N Engl J Med 2016; 375: 1112–15.