Back

Value of the body roundness index vs. body mass index for predicting 9-year mortality: Shizuoka Kokuho Database study

Takeuchi, M.

2025-12-22 public and global health
10.64898/2025.12.19.25342719 medRxiv
Show abstract

BackgroundThe body roundness index (BRI), calculated from height and waist circumference, is a novel obesity metric proposed as an alternative to body mass index (BMI). Although the BRI is reportedly associated with mortality risk, little is known about how the performance of the BRI and BMI compare. MethodsWe used data from the SKDB, a regional healthcare database in Shizuoka, Japan, that includes data from >2 million people from 2012 to 2022. Individuals aged 40 to 74 years who received a Specific Health Checkups in 2013 were included and followed for up to nine years. We investigated the associations of the BRI and BMI with future risk of all-cause death and compared their predictive value using the net reclassification index (NRI) and c-statistics. ResultsAmong 202,133 individuals (median age: 66 years; 64.6% female; median BMI: 22.3 kg/m{superscript 2}), 10,336 deaths (5.2%) occurred over a median follow-up of eight years. Both the BRI and BMI demonstrated J-shaped associations with mortality risk. Incorporating the BRI or BMI into age- and sex-adjusted models improved NRI and c-statistics, with NRI increases of 5.6% (95% confidence interval [CI]: 3.6 to 7.1%) for BRI and 7.8% (95% CI: 6.0 to 9.6%) for BMI. The incremental improvements in the c-statistic were 0.48% (95% CI, 0.47 to 0.48%) for the BRI and 0.65% (95% CI, 0.64 to 0.65%) for BMI, whereas the BRI performed slightly worse than BMI (difference: -0.17%; 95% CI, -0.16 to -0.18%). ConclusionsAlthough BRI was predictive of mortality risk, the BRI did not outperform BMI in our population.

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.