ROGER EXPLORES… the evolution of clinical trials

by Roger Wilson

The evolution of clinical research is happening rapidly, although the days of the experimental, randomised clinical trial are far from over. We are seeing new trial design ideas with new terms and names which may need explaining. Here is a quick starter explanation for some of them.

Basket trial. This is a trial where one drug (often a targeted therapy) is tested against patients with a range of cancers all of which have the same biomarker. Each patient’s biology would be determined before being accepted on the study. The aim is to identify which cancers respond best to the treatment.

Umbrella trial. An umbrella trial is where a range of targeted drugs, usually from one manufacturer, would be tested on patients with one sub-type of cancer, sometimes with a range of different genetic mutations and biomarkers. The aim is to identify the best treatment(s) for this particular cancer, and the genetic markers which respond best to each drug. An academic-led umbrella study is less common but may use drugs from a range of manufacturers, usually with their support.

Platform study. This is a design which encompasses and evaluates multiple studies in a specific disease or condition under a single master protocol. The rationale for the platform is carefully defined, and each individual patient is validated. There will be laboratory investigations within the platform. Studies may be ‘adaptive’ (see below) by adding or stopping treatment arms (new drugs, new biomarker defined patient groups etc) but the comparator group would usually be standard-of-care for that patient cohort. The aim is a flexible structure for investigating multiple research questions within a unified statistical and operational framework.

Adaptive. This is a study where the design allows for planned adaptation of the protocol during the study. Interim analyses of its data may lead to changes of such elements as size of patient groups, randomisation ratio, or to add new arms based on emerging evidence from other studies. The aim is to pre-specify statistical adaptation within defined rules, but changes must be valuable to answering the research question without introducing bias. There should be a Statistical Analysis Plan which ensures statistical integrity to get to the answer to the research question.

Pragmatic. A pragmatic trial is a study where the comparator group is made up of patients receiving the real-world standard-of-care. Randomisation to a new treatment or approach to care is compared with patients receiving the current standard practice. The aim is to ensure that the experimental treatment will benefit patients in the real-world clinic. Drug studies are increasingly using a pragmatic approach rather than placebo, which patients generally do not like. Pragmatic studies are often cost comparison studies but may also include drug dosing comparisons, use of product/materials in care and the management of clinic visits.

Pairwise comparison. This is an analytical technique which allows the data from a patient receiving an experimental treatment to be compared with a patient having the same disease/condition profile but receiving standard-of-care. Ideally the standard-of-care patient is specifically recruited to the research but in instances of rare cancers retrospective cases may be used. The pairing for comparison may be done on an individual or a group basis.  This requires very special care in selection of comparison patients to ensure that there is no bias introduced. The technique is particularly relevant in platform studies where there are several treatment arms.

The growing use of new trial designs, even in early phases of clinical research, will provide data more quickly and more accurately to inform clinical decision-making. Where a new drug is involved, the regulators should take part in the choice of study design so that they can be more certain of understanding how the results were arrived at. Improving patient involvement at all stages of study design in such trials is valuable for ensuring relevant improvement in the quality of care and the patient experience.

Early phase trials are increasingly using mixed Phase I/II studies whereby dose finding and safety criteria are researched in Phase I, while the Phase II element closes in on efficacy. Small patient groups and short durations make these studies quite demanding to plan, although statistical analysis can be relatively simple and may use well established models. While protocols can be prescriptive, they may allow adaptation if that is rational in the research pathway of the treatment.

Later Phase 3 randomised studies in cancer have become increasingly dominated by surrogate endpoints, justified by the recognition that patients may have further treatments after the study has ended. This distorts the pathway to the gold standard Overall Survival (OS)endpoint. Regulatory bodies recognise the unreliability of surrogate endpoints in traditional Randomized Control Trials (RCT) because the protocol tends to focus on the performance of the treatment rather than the response of the patient. The full analysis of clinical data can be complex and may take a long time to interpret accurately in the real-world clinical context. While the primary endpoint (on which statistical analysis is based) may be a surrogate such as Progression Free Survival (PFS), it is usual to have OS as a secondary endpoint.

Other developments which one can identify include:

• Identifying patient cohorts accurately, e.g. by using biomarkers
• Setting up comparator arms based on ‘real-world’ retrospective data
• Adopting new statistical analysis models (Bayesian probability rather than traditional Frequentist approaches are now common)
• Using Patient-Reported Outcomes (PROs) to complement clinical measurements, sometimes with differing endpoints
• Looking for new patient response appraisal methods, e.g. patient experience (PREMs)[i]

Platform trials like STAMPEDE in prostate cancer, TRACERx in lung cancer, and the RECOVERY trial for Covid-19 have shown that it is possible to test successive new treatments in a randomised manner to deliver results quickly.  These studies were underpinned by accrual to a cohort of relevant patients who committed to that research. They have shown that the platform model can include adaptive and pragmatic studies as well as more traditional RCTs. It provides a flexible base which allows rapid results to be obtained.

Part Two will look at what all this means for sarcoma.

[i] Patient Reported Experience Measures

Credits:  Graphic design by Karl Berger