Synthetic control methods for n-of-1 and parallel-group trials in Alzheimer's disease: A proof-of-concept study using...

With the advent of Alzheimer's disease (AD)-modifying and symptomatic treatments of demonstrated efficacy, enrolling participants as concurrent placebo controls in trials can become increasingly difficult. Synthetic controls have been proposed as a viable alternative to concurrent control groups, but their feasibility and reliability remain untested in AD studies.

I-CONECT trial, which evaluates conversational interactions on cognition, was used to test synthetic control methods. Data from the National Alzheimer's Coordinating Center-Uniform Data Set was used to create synthetic-controls for I-CONECT participants using two methods: 1) case mapping and 2) case modeling. Efficacy estimates were compared between original versus synthetic-controlled trials.

In parallel-group designs, treatment effect sizes for the primary outcome were closely aligned between the original trial (β = 1.67) and synthetic control analyses (β = 1.40-1.65). For n-of-1 designs, the two methods showed high agreement in identifying treatment responders (Kappa = 0.75-0.82).

Synthetic control methods are feasible and reliable to create alternative controls in AD studies.

NCT02871921.

Synthetic control methods are feasible and suitable for evaluating treatment effects in various trial designs such as n-of-1, single-arm, and parallel groups. Synthetic control methods can help replicate early-phase Alzheimer's trials, informing go/no-go decisions for larger-scale studies. The choice of similarity algorithms is critical as it affects the quality of historical case mapping. The National Alzheimer's Coordinating Center-Uniform Data Set (NACC-UDS) provided an ideal pool for identifying historical cases with similar demographic, biological, and social characteristics to participants in trials, enabling the creation of synthetic control groups for Alzheimer's clinical research.