# Investigating between-group effects of a physical activity intervention across the physical activity intensity spectrum using multivariate pattern analysis

## TL;DR

This study explores using multivariate pattern analysis to assess the effects of a physical activity intervention on preschool children across different activity intensities.

## Contribution

The study introduces multivariate pattern analysis as a method to evaluate between-group effects across the physical activity intensity spectrum in intervention trials.

## Key findings

- Higher-resolution activity descriptors improved model fit slightly in multivariate pattern analysis.
- Significant mean differences were detected, with comparable results from multivariate pattern analysis and linear mixed models.
- Multivariate pattern analysis is suitable for evaluating between-group effects but should be used as a secondary approach in cluster trials.

## Abstract

Effects of physical activity (PA) interventions are evaluated in many settings, but novel methods are needed to determine effects of interventions across the PA intensity spectrum. The aim of this study was to test the applicability of multivariate pattern analysis to determine between-group effects of an intervention study across the PA intensity spectrum.

A sample of 698 Norwegian preschool children (mean age 3.8 years) from the Active Learning Norwegian Preschool(er)s cluster randomized controlled trial provided data on PA (ActiGraph GT3X+) before and after short- (7 months) and/or long-term (18 months) follow-ups after a preschool PA intervention. Multivariate pattern analysis and linear mixed models were used to determine effects across the intensity spectrum using PA descriptors of different resolutions (4, 17 and 51 variables).

Higher-resolution PA descriptors led to marginally better model fit than lower-resolution descriptors in multivariate pattern analysis (explained variances 1.27–3.59%). We detected significant standardized mean differences of ± 0.15–0.36, with results for multivariate pattern analysis and linear mixed models being comparable.

We discuss applicability of these approaches for analysis of between-group effects on PA. We conclude that multivariate pattern analysis is a suitable analytic approach to evaluate between-group effect patterns across the PA intensity spectrum of intervention studies, but suggest it being a secondary approach for cluster trials given challenges of appropriately taking clustering into account.

Clinicaltrials.gov, identifier NCT04048967, registered August 7, 2019, (https://clinicaltrials.gov/ct2/show/NCT04048967?term=actnow&rank=1).

The online version contains supplementary material available at 10.1186/s12889-025-25722-5.