Article Text

PDF

Letter
Two heads are better than one: benefits of joint models for ALS trials
  1. Ruben P A van Eijk1,2,
  2. James Rooney3,
  3. Orla Hardiman3,4,
  4. Leonard H van den Berg1
  1. 1Neurology, Brain Center Rudolf Magnus, University Medical Center Utrecht, Utrecht, The Netherlands
  2. 2Biostatistics & Research Support, Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, The Netherlands
  3. 3Academic Unit of Neurology, Trinity Biomedical Sciences Institute, Trinity College Dublin, Dublin, Ireland
  4. 4Neurology, Beaumont Hospital, Dublin, Ireland5
  1. Correspondence to Dr Ruben P A van Eijk, Neurology, Brain Center Rudolf Magnus, University Medical Center Utrecht, Utrecht 3584, The Netherlands; R.P.A.vanEijk-2{at}umcutrecht.nl

Statistics from Altmetric.com

Kaji et al recently conducted a phase II/III clinical trial to evaluate the safety and efficacy of methylcobalamin in 343 patients with amyotrophic lateral sclerosis (ALS).1 Both coprimary endpoints (survival and ALS Functional Rating Scale [ALSFRS-R]) failed to show significant benefit (p=0.19 and p=0.13, respectively). The choice for two coprimary endpoints is interesting and may have circumvented important pitfalls encountered by previous ALS trials.2 As ALS significantly reduces the patient’s life expectancy, evaluating a drug’s therapeutic potential to improve survival is fundamental. Survival time, however, may be influenced by life-extending interventions (eg, gastrostomy or tracheostomy) and provides little information about the patient’s functioning during life. The ALSFRS-R is a welcome alternative that is clinically relevant, easily obtained and highly predictive of survival time. Unfortunately, positive phase II results on the ALSFRS-R have until now translated poorly into phase III survival endpoints.2 Moreover, treatments for ALS may affect both survival and function simultaneously. Any trial based on a single endpoint might, therefore, only partially capture the full treatment effect. Evaluating both endpoints simultaneously could circumvent these pitfalls and provide essential treatment clues.

In the trial of Kaji et al,1 the authors analysed the ALSFRS-R and survival time as two separate entities. Nevertheless, both endpoints are highly predictive of the other endpoint; that is, a fast ALSFRS-R progression rate predicts a short survival time and vice versa. By adjusting the ALSFRS-R for survival, and survival for the ALSFRS-R, a large part of the uncertainty in either endpoint can be explained. There are several methods to combine survival time and the ALSFRS-R.3 4 The Combined Assessment of Functional and Survival (CAFS) is a well-known composite endpoint and is currently being used as either primary or key secondary endpoint in several placebo-controlled trials (eg, arimoclomol [NCT03491462] or deferiprone [NCT03293069]).5 Nevertheless, the CAFS makes suboptimal use of the available information, and our recently proposed joint modelling framework could further improve the sensitivity of ALS trials.3 To illustrate the merits of a combined analysis, we performed a (simplified) simulation study based on the hypothesised treatment effect and sample size as reported by Kaji et al;1 the details of the simulation are described elsewhere.3 We assumed a reduction in the ALSFRS-R progression rate of 30.0%, an HR reduction of 50.0% (HR 0.5) as survival benefit and a sample size of 240 patients (120 per arm). For each simulated trial, we fitted the joint modelling framework and extracted the p values for survival (ie, Cox proportional hazard model), ALSFRS-R (ie, linear mixed effects [LME] model) and the combined test (ie, joint model of the Cox and LME models). In this simulation, the Cox model has 70.0% power, the LME model 83.6% and the joint model 93.1%. In terms of sample size,4 enrolment of 160 patients would have been sufficient for the joint model to detect the hypothesised treatment effect with 80% power.

These results underscore the large efficiency gains that can be achieved when combining the analysis of two related endpoints. In this case, the sample size could have been reduced by approximately a third (or detect a smaller effect with an identical sample size). Moreover, the joint modelling framework incorporates all collected data, adjusts for informative missing (ie, missing data in ALSFRS-R due to death) and could be easily extended with other endpoints such as muscle strength or lung function. Optimising the use of information in clinical trials could significantly improve their efficiency. In the end, this may speed up the development of effective drugs against this debilitating disease.

References

View Abstract

Footnotes

  • Contributors RPAvE: study concept, design, analysis, interpretation of data and drafting the manuscript. JR: study concept, design, analysis, interpretation of data and drafting the manuscript. OH: study concept and critical revision of the manuscript for intellectual content. LHvdB: study concept and critical revision of the manuscript for intellectual content.

  • Funding This study was funded by the Netherlands ALS Foundation (grant number: TRICALS).

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Provenance and peer review Not commissioned; internally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.