Adaptive trial design in osteoarthritis research

Purpose (the aim of the study): In 2020, 595 million people were living with osteoarthritis (OA) globally. From 2008 to 2011, the direct medical cost associated with OA in the US alone was around $72 billion. Despite the incidence of OA and economic burden, the development of improved treatments remains behind other rheumatic diseases. To slow down the devastating burden of OA, our field should consider innovative RCTs to provide more effective treatment options, more rapidly.

Adaptive randomised controlled designs are emerging as a more efficient and informative approach to clinical trials. As trial data is collected, preplanned modifications such as dropping treatment arms or revising treatment allocation ratios can be employed to make better use of resources such as: time, funds and participants. For example, a randomised clinical trial exploring three therapies for acute myeloid leukaemia used response-adaptive randomisation to limit patients exposed to inferior treatment arms. Over half of the patients were treated with the best of the three treatments and only five to the least beneficial.

A ratio of primary endpoint length to recruitment length may be a good indicator if a trial can benefit from an adaptive design. A low ratio indicates a large collection of primary data over the recruitment period. Therefore, if any adjustments need to be made, there is time to modify the study whilst recruitment is ongoing.

The aim of this study was to identify the number of adaptive trials in OA research and suggest how many planned or on-going RCTs could benefit from an adaptive design.

Methods: On September 8th 2024, we searched ClinicalTrials.gov for studies containing ‘Osteoarthritis’ in the ‘Condition and disease’ field and ‘Recruiting’ in the ‘Status’ field. Planned and ongoing studies were all included. Data extracted included length of time (in days) to observe primary endpoint after enrolment (primary endpoint length) and total planned recruitment length. A smaller primary endpoint length to recruitment length ratio can indicate the potential benefit of an adaptive design, with a threshold of 0.25 suggested. We calculated the ratio of primary endpoint length to recruitment length for all trials and summarised the proportion of trials that may benefit from an adaptive design.

Results: The search returned 521 titles; studies with non-random allocation, not related OA and insufficient information on study design were removed, with a final sample of 493 trials. No trials referred to an ‘adaptive design’. The primary endpoint length to recruitment length ratio ranged from 0.00 to 10.00, with a mean of 0.39 and a median of 0.21. Approximately 57% (282 trials) of the final sample had a ratio of less than 0.25 and could particularly benefit from an adaptive design (Table 1). Figure 1 depicts a trial that may benefit from adaptive design and one that may not.

Conclusions: The use of an adaptive design in OA studies is limited. The ideal ratio as an indicator of a trial that would benefit from an adaptive design is relatively arbitrary, and several factors require consideration in determining if an adaptive design is appropriate (i.e. platform design). Nevertheless, we suggest a considerable proportion of planned or on-going studies could benefit from an innovative adaptive study design. This would mean future trials become more efficient and provide flexibility to detect clinically relevant findings, if present.