The promise of futility trials in neurological diseases


Double-blinded randomized controlled trials (RCTs) have contributed much important evidence to guide treatment decisions in neurology. RCTs are relatively straightforward to conduct, provided that they investigate common diseases, have clearly defined outcome measures, and are of short duration. In neurology, however, many diseases are uncommon, have no consensus outcome measures, and develop over decades. Basic research into neurological diseases continues to identify candidate therapies faster than they can be tested for their clinical utility, leading to a 'translational gap'. Futility trials were initially developed in oncology to efficiently test candidate therapies in phase II trials. As single-arm unblinded studies, futility trials are relatively easy to conduct, and they generally require smaller sample sizes than RCTs. In this article, we discuss futility models, highlighting their advantages as well as challenges to their application in several neurological diseases, including Parkinson disease, stroke and multiple sclerosis.


The best evidence to guide clinical decision-making comes from randomized controlled trials (RCTs), and a general consensus exists on how RCTs should be designed to deliver the most accurate and reliable results. Efforts in many medical specialities by guideline committees and groups such as the Cochrane Collaboration review the quality of RCTs and translate their results into treatment recommendations.

RCTs are easier to conduct—and their results are easier to translate into clinical practice—if they investigate common diseases, include typical patients, have a short duration, and use clear and patient-relevant outcome measures. Many neurological diseases, however, cannot satisfy these criteria: they are uncommon, have a slow and sometimes unpredictable disease course, and lack straightforward and widely used outcome measures. Multiple sclerosis (MS) and Parkinson disease (PD), for example, develop over decades, and RCTs for these diseases generally require hundreds of patients and trial durations of several years.

Basic research into neurological diseases continues to suggest pathophysiological mechanisms and promising candidate treatments. Given the challenges described above, testing all such candidates in RCTs will not be feasible. This situation has been described as a 'translational gap', wherein treatments are available, but no quick and straightforward method is available to check their clinical utility, or to establish proof of concept, in a phase II trial.

Oncology has been faced with a translational gap for longer than other medical specialities. Many chemotherapeutics (and combinations thereof) have been developed, but there was no simple way to test their clinical utility. Since the 1960s, clinical trialists in oncology have developed innovative phase II trial designs to screen candidate therapies, so as to identify which treatments are worthy of further investigation in RCTs, and to reject those that have low probability of success. These alternative trial designs are based on the idea of futility rather than efficacy.

In RCTs, the null hypothesis states that treatment and control are equivalent in terms of efficacy, and the null hypothesis is rejected if outcomes are significantly different between the treatment and control groups. In futility trials, the null hypothesis is that the treatment will increase the number of treatment successes by a minimal clinically important amount. If the treatment does not achieve this effect, the null hypothesis is rejected, and the treatment is deemed futile. If the treatment does achieve this effect, it is declared nonfutile and considered to be worthy of further investigation in, for example, a phase III RCT.

In neurology, some preliminary efforts have been made to use futility models in diseases such as PD and stroke (Table 1), but these models have not been widely adopted as a straightforward and meaningful option for phase II trials. In this Perspectives article, we review the design and use of futility models, and discuss their application in neurological diseases.