Body composition and mortality in a large cohort with preserved ejection fraction: untangling the obesity paradox

Mayo Clin Proc. 2014 Aug;89(8):1072-9. doi: 10.1016/j.mayocp.2014.04.025. Epub 2014 Jul 16.


Objective: To evaluate the effects of body composition as a function of lean mass index (LMI) and body fat (BF) on the correlation between increasing body mass index (BMI; calculated as the weight in kilograms divided by the height in meters squared) and decreasing mortality, which is known as the obesity paradox.

Patients and methods: We retrospectively assessed 47,866 patients with preserved left ventricular ejection fraction (≥50%). We calculated BF by using the Jackson-Pollock equation and LMI using (1 - BF) × BMI. The population was divided according to the sex-adjusted BMI classification, sex-adjusted LMI classification, and sex-adjusted BF tertiles. The population was analyzed by using multivariate analysis for total mortality over a mean follow-up duration of 3.1 years by using the National Death Index, adjusting for left ventricular ejection fraction, left ventricular mass index, age, sex, and relative wall thickness.

Results: In the entire population, higher BMI was narrowly associated (hazard ratio [HR], 0.99; P<.001) with lower mortality. The higher LMI group was clearly protective (HR, 0.71; P<.001), whereas BF tertile was associated with lower mortality only if no adjustment was made for LMI (HR, 0.87; P<.001 without LMI; HR, 0.97; P=.23 with LMI). In the lean patients, low BMI was clearly associated with higher mortality (HR, 0.92; P<.001) and lower BF tertile was associated with lower mortality only if no adjustment was made for LMI (HR, 0.80; P<.001 without LMI; HR, 1.01; P=.83 with LMI). The underweight patients stratified by BF seemed to have an increased mortality (HR, 1.91; 95% CI, 1.56-2.34) that was independent of LMI. However, in obese patients, both BMI (HR, 1.03; P<.001) and BF (HR, 1.18; P=.003) were associated with higher mortality, even after adjusting for LMI, which remained protective (HR, 0.57; P<.001) independently of BF.

Conclusion: Body composition could explain the inverse J shape of the mortality curve noted with increasing BMI. Body fat seems to be protective in this cohort only if no adjustment was made for LMI, although being underweight stratified by BF seems to be an independent risk factor. Lean mass index seems to remain protective in obese patients even when BMI is not.

MeSH terms

  • Adipose Tissue / physiology
  • Body Composition / physiology*
  • Body Mass Index*
  • Cause of Death
  • Female
  • Humans
  • Louisiana
  • Male
  • Middle Aged
  • Multivariate Analysis
  • Obesity / complications
  • Obesity / mortality*
  • Proportional Hazards Models
  • Retrospective Studies
  • Stroke Volume / physiology*
  • Thinness / complications
  • Thinness / mortality*