Analysis of Changes in Weight, Waist Circumference, or Both, and All-Cause Mortality in Chinese Adults

Abstract

Importance: Although numerous studies have separately investigated the associations of changes in weight or waist circumference with mortality risk, few studies have examined the associations of concurrent changes in these 2 anthropometric parameters with all-cause mortality.

Objective: To assess the associations of changes in body weight, waist circumference, or both, combined with all-cause mortality.

Design, setting, and participants: This cohort study used data from 2 longitudinal cohort studies in Dongfeng-Tongji and Kailuan, China. Participants included 58 132 adults (aged 40 years and older) with measures of weight and waist circumference at baseline and follow-up visit. Statistical analysis was performed from June 2020 to September 2021.

Exposures: Changes in weight and waist circumference between 2 visits (2008-2010 to 2013 in the Dongfeng-Tongji cohort, and 2006-2007 to 2010-2011 in the Kailuan study). Stable weight was defined as change in weight within 2.5 kg between the 2 visits and stable waist circumference was defined as changes within 3.0 cm. Changes were categorized as loss, stable, or gain for weight and waist circumference separately, and created a 9-category variable to represent the joint changes.

Main outcomes and measures: All-cause mortality from follow-up visit (2013 in Dongfeng-Tongji cohort and 2010-2011 in Kailuan study) until December 31, 2018. Cox proportional hazard regression models were used to estimate the associations with adjustment for potential confounders. Results were obtained in the 2 cohorts separately and pooled via fixed-effect meta-analysis.

Results: A total of 10 951 participants in the Dongfeng-Tongji cohort (median [IQR] age, 62 [56-66] years; 4203 [38.4%] men) and 47 181 participants in the Kailuan study (median [IQR] age, 51 [46-58] years; 36 663 [77.7%] men) were included in the analysis. During 426 072 person-years of follow-up, 4028 deaths (523 in the Dongfeng-Tongji cohort and 3505 in the Kailuan study) were documented. When changes in weight and waist circumference were examined separately, U-shape associations were found: both gain and loss in weight (weight loss: pooled hazard ratio [HR], 1.33; 95% CI, 1.23-1.43; weight gain: HR, 1.10; 95% CI, 1.02-1.19) or waist circumference (waist circumference loss: HR, 1.14; 95% CI, 1.05-1.24; waist circumference gain: HR, 1.11; 95% CI, 1.03-1.21) were associated with higher mortality risk compared with stable weight or waist group. When changes in weight and waist circumference were jointly assessed, compared with participants with stable weight and waist circumference (16.9% of the total population [9828 of 58 132] with 508 deaths), participants with different combinations of weight and waist circumference change all had higher mortality risks except for those with stable weight but significant loss in waist. Notably, those who lost weight but gained waist circumference (6.4% of the total population [3698 of 58 132] with 308 deaths) had the highest risk of all-cause mortality (HR, 1.69; 95% CI, 1.46-1.96; absolute rate difference per 100 000 person-years in the Dongfeng-Tongji cohort: 414; 95% CI, 116-819; and in the Kailuan study: 333; 95% CI, 195-492) among the joint subgroups.

Conclusions and relevance: In this cohort study, weight loss with concurrent waist circumference gain was associated with a higher mortality risk in middle-aged and older Chinese adults. This study's findings suggest the importance of evaluating the changes in both body weight and waist circumference when assessing their associations with mortality.

Figure.. Adjusted Hazard Ratios for All-Cause Mortality Based on the Joint Changes in Weight and Waist Circumference

The multivariable-adjusted model included the joint categories of weight and waist circumference changes, weight, height, and waist circumference at cohort recruitment, smoking status, alcohol intake status, dietary pattern, educational attainment, physical activity, hypertension, and diabetes; and stratified by age at risk (5-year interval) and sex. Cohort-specific results were pooled together using fixed-effect meta-analyses. Separate results for the DFTJ cohort and the Kailuan study are shown in eFigure 2 in the Supplement.