From: Clinical trial design in the era of precision medicine
Trial design | Features | Advantages | Disadvantages | Challenges |
---|---|---|---|---|
Basket | One molecular alteration, multiple histologies | -Rare molecular alterations -Test treatment in diverse tumor types in parallel | -Alterations are not driver in every tumor type -Different mechanisms of resistance based on tumor type -Lack of comparative arm | Recruiting rare subsets across multiple disease types |
Umbrella | One histology, multiple molecular alterations | -Biomarker assessment -Improved enrollment rates when biomarker prevalence is low -Parallel evaluation of multiple treatment agents -Flexibility of dropping failing drugs | -Inadequate sample size -Multiple treatments matching molecular alterations -Suboptimal selection of treatment targets | Intra-patient heterogeneity of molecular findings, making it difficult to categorize patients |
Platform | Combines umbrella and basket features to create broad-based trial | -Allows the addition or exclusion of new investigational arms during the trial -Enables evaluation of multiple hypotheses in a single protocol -Shortens time -Lowers costs | -Complicated design -Administrative and logistical complexity -Long-term nature -High execution costs | Complexity of statistical analysis and of monitoring of extremely heterogeneous patient groups |
Adaptive | During the course of the study, the trial is changed as data are collected and analyzed | -Drops ineffective arms early -Modifies patient randomization to more effective treatments -Improves biomarker selection -Requires fewer participants -Requires shorter follow-up time | -Complicated design -Administrative and logistical complexity -Miss important secondary outcome data due to early elimination of treatment arms | Dependent on intense statistical monitoring; constant need to adapt the design may make the interpretation of the results difficult |
Telescope (seamless) | Seamless transition from phase I to II and sometimes to III | -Combines learning and confirmatory stages -Shortens duration of drug development and approval -Reduces administrative costs -Reduces effort -Focuses on promising treatments to be used in later trial stages -Drops failing treatments early -Focuses on responding subpopulations in later trial stages | -Design complexity | Designing all phases of the trial (I, II and III) without taking into consideration preliminary data from phase I trial. The long time period required to complete the study, which may be associated with change in practice and experimental drugs that gain regulatory approval in the interim, therefore making the interpretation of the results challenging |
N-of-1 | Personalized combination therapy; Patient-centric trial where each patient gets a customized therapy. The efficacy of the matching strategy rather than the individual therapies is evaluated | -Based on unique patient characteristics and tumor profile -Addresses molecular complexity and heterogeneity -Customized treatment | -Lack of comparator -Heterogeneity of treatments -Complexity of analysis/statistical algorithms | Difficult fit between individualized therapy and the way clinical oncology is practiced wherein physicians often specialize in specific types of cancer Rarity of patient characteristics Need to analyze the robustness of the strategy (algorithm) for matching, rather than drug combinations, since the latter differ from patient to patient |
Exceptional responders | In-depth understanding of unusual patients | -Biomarker identification -Highlights molecular pathways associated with response to treatments | -Rare cases -Requires validation | Lack of uniformity in available biomarker data and correlation with patient characteristics and clinical outcomes |
Registry protocols | Structured real-world data | -Collection of data in parallel: cancer incidence, patient demographics, treatment patterns, molecular profiling data, and clinical outcomes -Enables correlations -Lower costs | -Complex analysis | Need to analyze the robustness of the strategy (algorithm) for matching, rather than drug combinations, since the latter differ from patient to patient |
Real-world data | Data derived from electronic medical records and insurance data, as examples | -Collection of data in parallel: cancer incidence, patient demographics, treatment patterns, molecular profiling data, and clinical outcomes -Enables correlations -Lower costs -Results are representative of population -Safety data on vulnerable subpopulations -Accelerates drug approval | -Inaccurate reporting -Data discrepancies -Subjective assessment of benefit -Machine learning may improve information synthesis | Lack of structuring of data Inaccuracies propagated in the medical records, due to cloning of notes and lack of routine secondary checks Difficulty harmonizing records to draw conclusions |
Patient-reported outcomes | Patients report outcomes, often via digital devices | -Improves symptom control -Improves quality of life -Minimizes emergency department visits/hospitalizations -Improves patient survival -Improves physician-patient communication -Increased access to communication with treating team in case of limited access to hospital (rural areas) -Increased access to communication with treating team during COVID-19 pandemic | -Cost of applications -Difficulty in using the technology -Under-reporting of symptom severity -Underestimated symptom severity | Lack of medical knowledge on the part of patients may influence their interpretation of clinical events |
Home-based trials | Patients stay at home—the trial comes to them rather than having them travel to the trial | -Increased access to innovative treatments in case of limited access to site-based clinical trials -Increased access to innovative treatments during COVID-19 pandemic -Facilitates patients with difficulties in traveling | -Patient recruitment -Monitoring issues -Patient safety | Difficulty in recruiting patients with rare subsets and challenges in proactively engaging their physicians |