Continuous outcome estimation in N-of-1 trials for accelerated decision-making

Abstract

OBJECTIVE: N-of-1 trials aim to determine the therapeutic effect for a single individual. This individualized approach necessitates collecting multiple data points over time through repeated alternating periods of active treatment and a comparator or control condition. The extended duration of the treatment periods may increase patient burden, prolong placebo exposure, and increase the likelihood of study discontinuation. In theory, treatment responders (or non-responders) can be identified early during the trial if the therapeutic effect is strong (or completely lacking). There are no theoretical constraints to evaluate treatment efficacy more regularly-instead of only after a predetermined number of treatment periods. Regularly updating estimates on treatment effects allows clinicians to accelerate clinical decision-making regarding N-of-1 study termination. This study examined the value of continuous treatment effect estimation using Bayesian hypothesis testing in N-of-1 trials to accelerate and nuance clinical decision-making. METHODS: An N-of-1 trial with severe epilepsy was simulated and three N-of-1 trials in neurological conditions were (re-)analyzed continuously with consecutive data points using Bayesian hypothesis testing and/or a minimally clinically important threshold (30% seizure frequency reduction). Trial duration based on Bayesian testing with strong evidence for treatment effects was compared to original trial duration. RESULTS: Original trial duration could be reduced between 9.5% and 35% of the trial length by using continuous outcome estimation in two of the analyzed trial examples. The moment that strong evidence supporting beneficial treatment effects using Bayesian hypothesis testing and a significant probability of minimally clinically important differences are achieved during the trial may differ. Obtaining additional data points and alternating interventions over time improve certainty of the estimates of treatment effects. SIGNIFICANCE: Treatment efficacy decisions can be expedited when outcome estimation is performed continuously rather than delayed until the end of the trial. Clinical significance of N-of-1 trial outcome can be improved combining both Bayesian hypothesis testing and a minimally clinically important threshold.