Associations of Body Mass Index and Waist Circumference with 3-Year All-Cause Mortality Among the Oldest Old: Evidence from a Chinese Community-Based Prospective Cohort Study.

OBJECTIVE

Current international and national guidelines for body mass index (BMI) and waist circumference (WC) have been recommended to all adults. However, whether recommendations applied to the oldest old (aged 80+) is poorly known. The study objective was to investigate the relation of BMI and WC with 3-year all-cause mortality among the oldest old.

DESIGN, SETTING, AND PARTICIPANTS: A total of 4361 Chinese oldest old (mean age 91.8) participated in this community-based prospective cohort study.

MEASUREMENTS

BMI and WC were measured at baseline in 2011 and were used as continuous variables and as categorized variables by recommendations or by tertiles. Adjusted, sex-stratified Cox models with penalized splines and Cox models were constructed to explore the association.

RESULTS

Greater BMI and WC were linearly associated with lower mortality risk in both genders. The mortality risk was the lowest in overweight or obese participants (BMI ≥ 24.0) and was lower in participants with abdominal obesity. Compared to the upper tertile, those in the middle and lower tertile of BMI had a higher risk of mortality for men [hazard ratio (HR): 1.23 (1.02-1.48) and 1.53 (1.28-1.82)] and for women [HR: 1.21 (1.03-1.41) and 1.35 (1.15-1.58)]; it was also found in participants in the middle and lower tertile of WC for men [HR: 1.21 (1.01-1.46) and 1.41 (1.18-1.69)] and for women [HR: 1.35 (1.15-1.58) and 1.55 (1.32-1.81)] (all the P values for trend <.001). These findings were robust in further sensitivity analyses or when using propensity score matching, in subgroup analyses, or in octogenarians, nonagenarians, and centenarians.

CONCLUSIONS

In Chinese oldest old, both higher BMI and higher WC predict better survival in both genders. The finding suggests optimal BMI and WC may be sensitive to age, thus, the current recommendations for the oldest old may need to be revisited.