Two-Arm Crossover Randomized Controlled Trial Versus Meta-Analysis of N-of-1 Studies: Comparison of Statistical Efficiency in Determining an Intervention Effect

Abstract

N-of-1 trials are currently receiving broader attention in healthcare research when assessing the effectiveness of interventions. In contrast to the most commonly applied two-arm randomized controlled trial (RCT), in an N-of-1 design, the individual acts as their own control condition in the sense of a multiple crossover trial. N-of-1 trials can lead to a higher quality of patient by examining the effectiveness of an intervention at an individual level. Moreover, when a series of N-of-1 trials are properly aggregated, it becomes possible to detect an intervention effect at a population level. This work investigates whether a meta-analysis of summary data of a series of N-of-1 trials allows us to detect a statistically significant intervention effect with fewer participants than in a traditional, prospectively powered two-arm RCT and crossover design when evaluating a digital health intervention in cardiovascular care. After introducing these different analysis approaches, we compared the empirical properties in a simulation study both under the null hypothesis and with respect to power with different between-subject heterogeneity settings and in the presence of a carry-over effect. We further investigate the performance of a sequential aggregation procedure. In terms of simulated power, the threshold of 80% was achieved earlier for the aggregating procedure, requiring fewer participants.

Keywords: N‐of‐1 trials; comparative effectiveness; crossover Design; meta‐analysis; two‐arm randomized controlled trial.

FIGURE 1

Rejection rates of the three methods for $\delta =0.15$ for increasing sample sizes, with no carry‐over effect, for setting (a) homogeneity ($\tau =0)$, (b) weak heterogeneity $\tau =0.1$, and (c). strong heterogeneity $\tau =0.5$.

FIGURE 2

Rejection rates of the three methods for $\delta =0.15$ for increasing sample sizes, with carry‐over effect, for setting (a) homogeneity ($\tau =0)$, (b) weak heterogeneity $\tau =0.1$, and (c) strong heterogeneity $\tau =0.5$.

FIGURE 3

Rejection rates of the three methods for fixed sample size n = 30 with increasing true treatment effect $\delta$, without and with carry over effect, for setting (a) homogeneity ($\tau =0)$, (b) weak heterogeneity $\tau =0.1$, and (c) strong heterogeneity $\tau =0.5$.

FIGURE 4

Rejection rate of a cumulative meta‐analysis with a Bonferroni correction to detect a true treatment effect $\delta =0.15$, $\tau =0.5,$ and with carry‐over effect.