New Zealand Statistical Association 2024 Conference

Mark Weatherall

University of Otago, Wellington and Medical Research Institute of New Zealand

Validation of clinically important differences for randomised controlled trials in asthma

This is joint work with Richard Beasley, Jon Noble, Allie Eathorne, Pepa Bruce

Introduction
Planning sample sizes for randomised controlled trials (RCTs) includes nominating a minimum clinically important difference (MCID) for an outcome variable. One way of exploring the MCID of an outcome variable is to compare it with other, established, outcome variables.

Methods
The Asthma Control Questionnaire 5 Item (ACQ-5) is a recognised outcome variable in asthma trials with an MCID of 0.5. This outcome variable was measured in two RCTs in asthma, in association with a biomarker of asthma, the Fractional Expired Nitric Oxide (FeNO). It was also measured in a cohort study of a stepped-care approach to asthma medication, in association with a participant reported outcome measure, the Treatment Satisfaction Questionnaire for Medication (TSQM). Scatter plots and linear regression were used to estimate the strength of association between the ACQ-5 and FeNO and TSQM. Logistic regression with Receiver Operating Characteristic Curve plots and c-statistics assessed the discrimination for the MCID of the ACQ-5 for FeNO and TSQM.

Results
In the RCTs the point estimate for association of FeNO with the ACQ-5 MCID was consistent with past suggestions that a 20% relative change in FeNO may be the MICD. The best c-statistic was 0.62 in a sub-group of participants with poor asthma control and high baseline ACQ-5. In the cohort study the c-statistic was 0.67 for TSQM in relation to ACQ-5.

Discussion
Both the outcome variables were associated with the ACQ-5 but with modest discrimination for its MCID. They may still be useful outcome variables capturing different aspects of asthma.