Systematic review of basket trials, umbrella trials, and platform trials: a landscape analysis of master protocols

Systematic searches were conducted on July 8, 2019, in MEDLINE, EMBASE, and the Cochrane Central Register of Controlled Trials. As no validated literature search strategy has been published, our strategies were developed on the basis of a review of key papers, including the Draft Guidance of the US Food and Drug Administration (FDA) [3‐6, 15]. We complemented the search terms of “master protocols”, “basket trials”, “umbrella trials”, and “platform trials” with several search terms specific to “adaptive trial designs” to improve the sensitivity of our search. The search strategies for each database are presented in Additional file 1: Tables S1–S3. We supplemented our database searches with a review of bibliographies from included publications. In addition, we searched trial registries (ClinicalTrials.gov and ISRCTN registry) for registered master protocols. Search terms used for ClinicalTrials.gov are reported in Additional file 1: Table S4. The list of published reviews related to master protocols that we reviewed is provided in Additional file 1: Table S5.

Complete study eligibility is described in Table 1. In brief, we included peer-reviewed publications, conference abstracts, and clinical registry records reporting on master protocols (basket trials, umbrella trials, and platform trials) that have been proposed, are ongoing, or have already been conducted. We defined “basket trials” as any prospective clinical trials that investigated the utility (e.g., effectiveness, dosage, and safety) of intervention(s) in a study population of multiple diseases with common predictive biomarkers or other common predictive patient characteristics that can be used to predict whether a patient will respond to a specific intervention (or both) as the unifying eligibility criteria. We defined “umbrella trials” as any prospective clinical trials that investigated the utility of targeted interventions based on predictive biomarkers or other patient characteristics or both. In umbrella trials, the single disease population (e.g., single histology cancer) is stratified into multiple subgroups on predictive biomarkers or other characteristics or both. We defined “platform trials” as any clinical trials that allowed for the intervention arm(s) to be dropped and the flexibility of introducing new intervention(s) during the trial. Graphical displays of basket trials, umbrella trials, and platform trials are provided in Fig. 1. We excluded non-English language studies.

Two reviewers (JJHP and MJZ) independently reviewed all abstracts and proceedings identified in the literature searches. The full-text publications of potentially relevant abstracts were then retrieved and assessed for eligibility. Two reviewers also screened the bibliographies of published literature reviews on master protocols (JJHP and ES) and trial registries (JJHP and LD). Discrepancies in study selection were resolved by discussion or, when necessary, by a third investigator (KT or EJM).

Study design elements, patient characteristics, and outcomes were extracted independently by two investigators (JJHP and ES) using a standardized, piloted data extraction form. We recorded information on trial registry, trial recruitment status, phase, randomization, masking, number of clinical centers, sample size, trial duration, interventions and control, disease area, age of population, number of conventional diseases recruited, key eligibility for stratification, number of subgroups defined, and geographic location of the master protocols. Discrepancies were resolved by discussion.

A meta-analysis was not conducted for this study, and we present the findings of this landscape analysis descriptively. We report on temporal trends of master protocols, geographical representation, and trial and disease characteristics of each of the three master protocols (basket trials, umbrella trials, and platform trials).

This study was not funded.

The study selection process is presented in Additional file 1: Figure S1. We identified 5869 abstracts from our database searches, and 140 more records were identified through hand searches of bibliographies and trial registries. Of these, 639 records were selected for full-text review. In total, 214 publications describing 83 trials met our inclusion criteria. Thirty-four trials were available only through trial registries, and three trials were in the pre-recruitment phase (NCT03339843, NCT03915678, and NCT03872427). A complete list of trials and the corresponding citations is provided in Additional file 1: Tables S6–S8. In summary, we identified 49 basket trials, 18 umbrella trials, and 16 platform trials.

There has been a rapid increase in the number of master protocols published in the last five years (Fig. 2). From our literature search, we identified nine completed and published master protocol trials, including results. The first master protocol conducted was a basket trial called the Imatinib Target Exploration Consortium Study B2225 [16, 17], which started in 2001. This was followed by the platform trial STAMPEDE, which was first proposed in 2005 [8, 9, 18‐28]. We identified 68 ongoing master protocols (39 basket trials, 17 umbrella trials, and 12 platform trials) recruiting patients; of these, 11 basket trials [29‐38], eight umbrella trials [39‐46], and four platform trials [24, 47‐50] have published results (Additional file 1: Table S10).

At the time of writing (August 1, 2019), one platform trial (LEAP; NCT03092674) is suspended for an unscheduled safety data review [51]. EBOLA (NCT02380 625), a platform trial supported by the Bill & Melinda Gates Foundation in response to the 2014 West Africa Ebola outbreak, has been terminated, as it could not be launched in time in response to the outbreak [52].

Trial characteristics of the master protocols are presented in Additional file 1: Table S9, and the sample size distribution of these master protocols displayed as box plots is provided in Fig. 3.

The majority of master protocols were basket designs, and 49 are identified in the current review. Among basket trials, all but one involved a drug investigation (n = 48/49); NCT03003195 was the exception as a proposed vaccine basket trial. The majority of basket trials were exploratory (i.e., phase I or II; n = 47/49) and were open-label (n = 46/49); more than half of the included basket trials investigated only a single intervention arm (n = 28/48; one did not report information on the number of interventions), and the majority did not involve a control group or randomization (n = 44/49). The median sample size of basket trials was 205 participants (interquartile range, Q3-Q1 [IQR]: 500–90 = 410), and the median study duration was 22.3 (IQR: 74.1–42.9 = 31.1) months. ALCHEMIST (NCT02193282, NCT02595944, and NCT02201992) and CLUSTER (NCT02059291) [53‐55] were the only phase III basket trials, which were comprised of three interventions arms and were of an open-label design.

Our review returned 16 platform trials. All of the platform trials involved investigation of experimental drugs. The median sample size was 892 (IQR: 1835–255 = 1580), and the median study duration was 58.9 (IQR: 101.3–36.9 = 64.4) months. Nearly all platform trials were of open-label design (n = 12/14; two trials did not report information on blinding), similar to basket and umbrella trials. However, phase III investigation was more common among platform trials (n = 7/15; one did not report information on phase) in contrast to basket and umbrella trials; four of these seven platform trials were seamless II/III trials. In the majority of platform trials, patients were assigned by randomization (n = 15/16). PRISM (NCT03527147) was the only non-randomized platform trial, although this is currently a phase I study. However, the trial registry of PRISM indicates that future arms may be added. In STAMPEDE [8, 9, 18‐28] and I-SPY2 [49, 50, 56‐61], several agents have graduated from the phase II evaluation with seamless transitions into phase III evaluations. The phase III evaluation for the I-SPY program is called I-SPY3.

The patient and disease characteristics of master protocols are provided in Additional file 1: Table S10. Most studies were in adult populations (n = 69/83), and nearly all were in the field of oncology (n = 76/83). No umbrella trials were conducted outside of oncology. Notably, two basket trials were conducted for other clinical indications, namely hereditary periodic fevers (CLUSTER; NCT02059291) [53‐55] and complement-mediated disorders (TNT0009 Basket trial). Additionally, five platform trials have been designed for influenza (ALIC4E; ISRCTN27908921) [62], Ebola (EBOLA) [52], pneumonia (REMAP-CAP; NCT02735707), pre-operative surgery (UPMC REMAP; NCT03861767), and Alzheimer’s disease (The DIAN-TU platform; NCT01760005) [48].

The information on the geographical representation of the current master protocols is provided in Additional file 1: Table S11. The majority of current master protocols have taken place in the US (n = 44/83) (Fig. 4). Other high-income countries such as the UK (n = 25), France (n = 23), Spain (n = 17), and Canada (n = 13) were the next most common countries. There were no master protocols observed from low-income countries, although the EBOLA (NCT02380625) trial had been proposed for Guinea, Sierra Leone, and Liberia [52]. Two upper-middle-income countries, Brazil and Mexico, were involved in the DIAN-TU platform trial (NCT01760005), but these countries accounted for only three of 36 study sites [48]. China, an upper-middle-income country, has centers participating in FUTURE (NCT03805399), GBM AGILE [47, 56, 63], TRUMP (NCT03574402), and VE-BASKET (NCT01524978) [64, 65] trials, but it should be noted that China accounts for only a minority of study sites in GBM AGILE and VE-BASKET.