When You Eat Matters: Two Methods Reveal Distinct Daily Eating Patterns

When we eat may be just as important as what we eat — but measuring meal timing patterns consistently has proven difficult. Different research groups use different statistical tools, making it hard to compare findings across studies. This paper directly tackles that methodological gap.

Researchers from Deakin University and the University of Queensland analyzed food diary data from 672 Australian adults aged 18–65, collected via a smartphone app over one to seven days. They applied two analytical approaches — latent class analysis (LCA) and modified dynamic time warping (MDTW)-based cluster analysis — to identify distinct temporal eating patterns and then tested how well each predicted diet quality and obesity.

Both methods independently identified three eating patterns: a conventional pattern with peaks at breakfast, lunch, and dinner; a late-eating pattern with most intake occurring after 1 PM; and an early, evenly distributed pattern. The conventional pattern was associated with higher diet quality compared to the late-eating pattern. Notably, neither method found a statistically significant association between eating timing and BMI, which challenges some popular assumptions about meal timing and weight.

The two methods agreed on participant classification between 56% and 73% of the time, with only fair statistical agreement (kappa = 0.38). LCA explained slightly more variance in diet quality (6% vs. 4%), while both methods performed nearly identically in predicting obesity risk. The authors suggest LCA is preferable for diet quality research, while MDTW may be more appropriate when integrating multiple health variables simultaneously.

For clinicians and researchers, this study highlights that the choice of analytical method meaningfully shapes findings in chrononutrition research. Standardizing approaches will be essential for building a coherent evidence base around meal timing and metabolic health.