Analyzing Daily Weights to Improve Estimates of Weight Loss while Minimizing Impact of Random Day-to-Day Variation in Weight

Introduction and Objective: Accurate estimation of weight loss in short-term studies is challenging because of measurement errors from day-to-day body-weight fluctuations. We compared 2 approaches to optimize estimates of weight loss in individual participants in a 6-week trial of semaglutide in non-diabetic overweight/obese adults (BMI ≥27). Methods: We compared 2 estimates of weight loss in 19 participants: (a) standard 2-point estimate (baseline weight minus weight on Day 42); (b) slope of least-squares regression line for daily weights over time. Results: While the group mean daily weights showed a strong linear relationship with time (r = −0.97), individual-level correlations were substantially weaker (mean r = −0.64), reflecting substantial day-to-day variability. Day-to-day weight variability introduced measurement error in standard 2-point estimates. The slope of the least-squares regression line reduced measurement error by incorporating all 42 days of data for each patient. Bland-Altman analysis of regression slopes derived from odd- vs even-day data was used as an internal benchmark of method precision and demonstrated near-perfect agreement: mean bias: -0.21 ± 0.17 kg/6-week (mean ± SEM). In contrast, mean bias was 2-fold larger for a comparison between the standard 2-point (Day42−Day1) vs. the regression-based method; mean bias: -0.42 ± 0.17 kg/6-week. Similarly, the limits of agreement for the odd-day versus even-day estimates were narrow: -1.72 to 1.26 kg/6week. In contrast, the limits of agreement for the comparison of standard 2-point estimates vs. the regression-based analysis were wider: -2.06 to 9.24 kg/6week. - indicating the confounding impact of random measurement error in the standard 2-point estimate of weight loss. Conclusion: Least-squares regression of daily weight leverages all available data and markedly reduces measurement errors due to random variation. Thus, this approach provides statistically robust estimate of weight loss in individual participants in short-term clinical trials. Disclosure: S. Bagheri: None. H.B. Whitlatch: None. A.L. Beitelshees: None. J. O'Connell: None. S.A. Bargal: None. S.I. Taylor: Other - Royalties; Current; Chiesi USA, Inc. E.A. Streeten: None. Funding: R01DK130238