# Methods to Crosswalk Between Cognitive Test Scores Using Data From the Alzheimer’s Disease Neuroimaging Cohort

## TL;DR

This paper introduces new methods to compare cognitive test results from different dementia studies, enabling better analysis of treatment effects.

## Contribution

The paper presents two novel methods for crosswalking cognitive test scores without relying on strong distributional assumptions.

## Key findings

- The methods were applied to crosswalk between CDR-SB, MoCA, and MMSE scores using data from ADNI.
- The paper demonstrates a case where crosswalking is not appropriate, such as with plasma p-tau and cognitive measures.
- The proposed methods offer a flexible and broadly applicable solution for harmonizing cognitive outcomes across studies.

## Abstract

Studies use multiple instruments to measure dementia-related outcomes, making comparisons of intervention or exposure effects challenging. For example, trials of anti-amyloid drugs like TRAILBLAZER-AD 2 (donanemab), CLARITY-AD (lecanemab), and GRADUATE I & II (gantenerumab) report changes in the clinical dementia rating sum of boxes (CDR-SB) as the primary outcome, while earlier trials used measures such as the Mini-Mental State Exam (MMSE) or the Alzheimer’s Disease Assessment Scale-Cognitive Subscale. Comparing or meta-analyzing findings requires integrating effect estimates from different outcome measures. Psychometric methods can harmonize these measures but rely on unverifiable assumptions and do not work with summary statistics. To address this gap, we developed two methods to crosswalk estimated treatment effects on cognitive outcomes that are flexible, broadly applicable, and do not rely on strong distributional assumptions. We present two methods to crosswalk effect estimates using one measure to estimates using another measure, illustrated with global cognitive measures from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Specifically, we develop crosswalks for the following measures and associated change scores over time: the clinical dementia rating scale sum of box (CDR-SB), Montreal cognitive assessment (MoCA), and mini-mental state exam (MMSE) scores. Finally, a setting in which crosswalking is not appropriate is illustrated with plasma phosphorylated tau (p-tau) concentration and global cognitive measures. Given the inconsistent collection and reporting of dementia and cognitive outcomes across studies, these crosswalking methods offer a valuable approach to harmonizing and comparing results reported on different scales.