Validation and measurement invariance of the Compulsive Exercise Test among Brazilian and American young adults

Abstract

Abstract Purpose To evaluate compulsive exercise, researchers often rely on the widely used Compulsive Exercise Test (CET). However, the measure has shown unstable factor structure in several validation studies and is not available in Portuguese for use in Brazil. We aimed to describe the translation and cultural adaptation of the CET to Brazilian Portuguese, to test several factor structures among Brazilian and US samples of men and women, to test measurement invariance across countries, and to evaluate its internal consistency. Furthermore, we sought to evaluate convergent validity, correlating the CET with a measure of eating disorder symptoms, and to compare compulsive exercise symptoms between countries. Methods Four models of the latent structure of the CET were tested using confirmatory factor analyses (CFAs), three-factor structure with 15 items, three-factor structure with 18 items, four-factor structure with 21 items, and the original five-factor structure with 24 items, in a sample of 1,531 young adults (601 Brazilians and 930 Americans), aged 18–35 years. Results A series of CFAs demonstrated that the three-factor structure with 15 items showed a better fit to the data. This model demonstrated good convergent validity and internal consistency. Results from the CET multigroup CFA showed evidence for the invariance at the configural, metric, and scalar levels across Brazilians and Americans. Furthermore, significant differences were found between Brazilians and Americans, with Brazilians demonstrating higher scores on the Avoidance and rule-driven behavior and Mood improvement subscales, whereas US participants scored higher on the Weight control exercise subscale. Conclusions Results support the three-factor structure with 15 items to be used as a measure of compulsive exercise among Brazilians and Americans, allowing cross-cultural comparisons between these countries. Level of evidence Level V, Cross-sectional, Psychometric study.