Abstract
Objective
To investigate the association of obesity and all-cause mortality in a sample of middle-aged and elderly population.
Design and Setting
Information of participants was collected in the Dongfeng-Tongji study, a perspective cohort study of Chinese occupational population. The main outcome was risk of death after 8.5 years of follow-up.
Participants and measurements
We examined the association of BMI, waist circumference (WC, and waist–height ratio (WHtR) with all-cause mortality in the Dongfeng-Tongji cohort study (n=26,143). Cox proportional hazard models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CI) for all-cause mortality. Area under the receiver operating characteristic curves and net reclassification improvement (NRI) were used to calculate the power of prediction models.
Results
During a mean of 8.5 years of follow-up, 2,246 deaths were identified. There is a U-shaped association of BMI with all-cause mortality in the middle-aged and elderly Chinese population. Compared with individuals with normal BMI, underweight was positively (HR=2.16, 95% CI: 1.73, 2.69) while overweight (HR=0.75, 95% CI: 0.67, 0.84) and obesity (HR=0.67, 95% CI: 0.56, 0.79) were negatively associated with all-cause mortality after adjustment for potential confounders including WC. In contrast, WC (Q5 vs. Q1, HR=1.55, 95% CI: 1.29, 1.86) and WHtR (Q5 vs.Q1, HR=1.69, 95% CI: 1.40, 2.04) were positively associated with mortality after further adjustment for BMI (P trend < 0.001). Addition of both BMI and WC into the all-cause mortality predictive model significantly increased AUC (P =0.0002) and NRI (NRI = 2.57%, P = 0.0007).
Conclusions
BMI and WC/WHtR were independently associated with all-cause mortality after mutual adjustment. Combination of BMI and WC/WHtR improved the predictive ability of all-cause mortality risk in the middle-aged and elderly population.
Key words: Prospective cohort, body-shape, death
Introduction
Obesity prevalence is increasing in all age groups including the elder population (1). BMI, an indicator of general obesity, was found J-shape, U-shape or null associated with all-cause mortality in elder population (2, 3, 4). Some studies found inverse association of BMI with risk of death (5, 6), indicating an “obesity paradox with the higher BMI related to lower mortality (7, 8, 9). Abdominal obesity, usually assessed by waist circumference (WC), waist-height ratio (WHtR) and waist-hip ratio, was found positively associated with risk of death (10, 11), whereas some studies found no associations between them (12). Researchers proposed that the controversial association between general obesity (BMI) or abdominal obesity (WC or WHtR) and all-cause mortality risk might attribute to the lack of adjustment for each other in the multivariate model (13).
To further explore the independent association of BMI and WC with mortality risk, based on a middle-aged and elderly Chinese population derived from the ongoing Dongfeng-Tongji cohort, we evaluated (1) independent associations of general obesity (BMI) and abdominal obesity (WC or WHtR) with allcause mortality risk; (2) the predictive ability of the BMI and WC/WHtR for all-cause mortality risk.
Methods
Study population
Participants in the present study were derived from the Dongfeng-Tongji cohort. The detailed information of this cohort was described in previous study (14). Briefly, 27,009 subjects (44.6% males) recruited between September 2008 and June 2010 were followed up every 5 years. The first follow-up was conducted from April to October in 2013. After obtaining written informed consent, trained interviewers took questionnaire to collect baseline information by face-to-face interviews. The questionnaires included detailed information on demographics, lifestyle factors, and medical history. Among the 27,009 subjects, 856 individuals with missing data of one or more of the body index (BMI, WC or WHtR) were excluded and 10 subjects with outliers were further excluded. Finally, a total of 26,143 participants with 2,246 deaths (from baseline to November 2016) were included for the further analysis.
The study was approved by the Medical Ethics Committee of the School of Public Health, Tongji Medical College, and Dongfeng General Hospital, DMC. All participants provided written informed consent.
Assessment of endpoint
The cause and date of death were ascertained by means of record linkage with the regional center for disease control and hospital. Death data were coded according to the rules of the 10th international Classification of Diseases (ICD-10).
Assessment of anthropometric data
Weight, height, and waist circumference were measured by trained health technicians following standardized procedures. Height and weight were measured with participants wearing light indoor clothing and no shoes. Waist circumference was measured at the midpoint between the lower ribs and the iliac crest.
Statistical analysis
Multivariate Cox proportional hazard models were used to analyze the association between anthropometric variables and all-cause mortality risk. Individuals were divided into underweight (BMI<18.5 kg/m2), normal weight (18.5 to < 24 kg/m2), overweight (24 to < 28 kg/m2), and obesity ( ≥ 28 kg/m2) groups according to Chinese criteria of weight for adults. The cutoffpoints of abdominal obesity were 90 centimeters (cm) for males and 85 cm for females, respectively (15). Participants were also grouped into sex- specific quintiles according to WC and WHtR. Tests for trend used the quintile medians. Associations between anthropometric index and allcause mortality risk were also evaluated using nonparametric restricted cubic splines, with 3 konts defined at the 5th, 50th, and 90th percentiles of the anthropometric measurements. The covariates adjusted for in the multivariate models were age, sex, smoking status [classified by the smoking index (daily cigars* smoking years): null, reference; < 200, light; 200–400, moderate; >400, heavy] (16), drinking status (never drinker, current drinker, former drinker), physical activity status (yes or no), and family history of cardio-cerebrovascular diseases (yes or no for coronary heart disease [CHD] or stroke). BMI was introduced into the multivariate models to investigate the independent associations of waist circumference and WHtR with all-cause mortality risk, vice versa. Similar analyses were conducted after exclusion of participants with diabetes mellitus (17), CHD, stroke, and cancer at baseline (16,966 subjects remained for analysis). Further sensitivity analysis were run by excluding participants who died during the first year of follow- up (25,963 subjects remained for analysis).
To evaluate the predictive power of the prediction model for all-cause mortality risk, we computed the area under the receiver operating characteristic (ROC) curve, and denoted area under curve (18) based on the predicted risk for each individual obtained from the logistic regression analysis. Furthermore, the contribution of anthropometric index to predict all-cause mortality (or discriminative power of predictive model) risk was assessed by the net reclassification improvement (NRI) method which evaluates the proportion of subjects moving accurately (NRI > 0) or inaccurately (NRI < 0) from one risk category to another after adding variables into the basic model. We divided participants into four risk categories (< 5%, 5% to 10%, 10% to 20%, and ≥ 20%) (19, 20). Basic predictive model including age, sex, smoking status, drinking status, physical activity, and family history of cardio-cerebrovascular diseases was taken as reference. BMI alone, WC alone, WHtR alone, BMI and WC together, BMI and WHtR together were added to the basic model respectively to examine their effects on the predictive ability of all-cause mortality risk.
Analyses were performed with the SPSS version 13.0 (SPSS Inc., Chicago, IL, USA) and SAS version 9.3 (SAS Institute, Cary, North Carolina, USA).Two-sided P < 0•05 was considered as statistically significant.
Results
Baseline characteristics of the participants
Descriptive statistics in different BMI categories are presented in Table 1. Underweight participants smoked heavily. Participants with larger BMI had larger waist circumference and WHtR. Nearly half of overweight participants (44.2%) and majority of general obesity (86.3%) individuals had abdominal obesity. Overweight and obesity participants had higher prevalence of chronic diseases, including DM, CHD, and stroke. There were no significant difference in cancer prevalence among different BMI categories.
Table 1.
Baseline characteristics of the study population (DFTJ cohort, n=26,143)
| Variables | BMI (kg/m2) | P value | |||
|---|---|---|---|---|---|
| <18.5 | 18.5-23.9 | 24-27.9 | > 28 | ||
| N (%) | 687(2.63) | 11152(42.66) | 10551(40.36) | 3753(14.36) | |
| Age, years, (SD) | 64.3(8.6) | 62.9(8.0) | 63.9(7.6) | 64.5(7.6) | <0.0001 |
| Male, n (%) | 322(46.9) | 4692(42.1) | 5086(48.2) | 1552(41.4) | 0.04 |
| Waist circumference, cm, median (Q25, Q75) | 67(63.6,71.5) | 77(73,82) | 86(81.8,90.2) | 94(89,100) | <0.0001 |
| Abdominal obesity, n (%) | 0 | 972(8.7) | 4662(44.2) | 3238(86.3) | <0.0001 |
| Waist-Height Ratio, median (Q25, Q75) | 41.9(39.8,44.4) | 48.1(45.5,50.7) | 53.3(50.9,56.0) | 56.0(53.6,62.1) | <0.0001 |
| Smoking Index, n (%) | 0.017 | ||||
| Never | 450(65.5) | 8006(71.9) | 7327(69.4) | 2761(73.6) | |
| Light | 34(4.9) | 510(4.6) | 577(5.5) | 189(5.0) | |
| Moderate | 46(6.7) | 671(6) | 736(7) | 211(5.6) | |
| Heavy | 157(22.9) | 1965(17.6) | 1911(18.1) | 592(15.8) | |
| Drinking status, n (%) | 0.0004 | ||||
| Never | 522(76.0) | 8315(74.6) | 7526(71.4) | 2834(75.5) | |
| Current | 127(18.5) | 2273(20.4) | 2361(22.4) | 674(18) | |
| Former | 38(5.5) | 562(5) | 660(6.3) | 245(6.5) | |
| Physical activity, n (%) | 596(86.8) | 9956(89.3) | 9426(89.3) | 3241(86.4) | <0.0001 |
| Stroke, n (%) | 21(3.1) | 411(3.7) | 536(5.1) | 202(5.4) | <0.0001 |
| CHD, n (%) | 62(9.0) | 1367(12.3) | 1929(18.3) | 895(23.8) | <0.0001 |
| Cancer, n (%) | 21(3.1) | 397(3.6) | 342(3.2) | 126(3.4) | 0.59 |
| DM, n (%) | 56(8.2) | 1653(14.8) | 2298(21.8) | 1021(27.2) | <0.0001 |
| Family history of CCVD, n (%) | 58(8.3) | 1028(9.6) | 945(9.0) | 306(8.2) | 0.046 |
Smoking index= cigarettes per day* smoking years, CCVD=cardio-cerebrovascular diseases
Associations of BMI with all-cause mortality
During a mean of 8.5 years of follow-up, 2,246 deaths were identified. As showed in Table 2, compared with participants with normal weight, underweight participants consistently had increased risk of death after adjustment for traditional risk factors (HR=1.72;95% CI:1.41to 2.09). Further adjustment for WC (HR=2.16; 95% CI: 1.73 to 2.69) or WHtR (HR=2.34; 95% CI: 1.87 to 2.92) the associations were even stronger. However, overweight and obesity were negatively related to mortality risk. Further controlling for waist circumference, subjects with overweight (HR= 0.75; 95% CI: 0.67 to 0.84) or obesity (HR= 0.67; 95% CI: 0.56 to 0.79) had decreased risk of death. Similar results were obtained after adjustment for WHtR (Figure 1 and Table 2). In the nonparametric restricted cubic splines (Figure 1) we observed similar results. There was an reverse J-shaped relationship between BMI and all-cause mortality. Subjects with 25.6 kg/m2 of BMI had lowest hazard ratio. In sensitivity analysis, we further excluded early deaths during the first year of follow-up and participants with chronic diseases (including CHD, DM, stroke and cancer) at baseline, the results remained unchanged (Supplementary Table 1).
Table 2.
Associations of BMI with all-cause mortality risk (DFTJ cohort, n=26,143)
| BMI (kg/m2) | ||||
|---|---|---|---|---|
| <18.5 | 18.5-23.9 | 24-27.9 | >28 | |
| Participants, n (%) | 687(2.63) | 11152(42.66) | 10551(40.36) | 3753(14.36) |
| Deaths, n (%) | 109(15.87) | 941(8.44) | 860(8.15) | 336(8.95) |
| Age and sex adjusted | 1.78(1.46,2.18) | 1.00 | 0.88(0.80,0.97) | 0.97(0.85,1.09) |
| Model1* | 1.72(1.41,2.09) | 1.00 | 0.89(0.81,0.98) | 0.95(0.84,1.08) |
| Model1*+Waist Circumference | 2.16(1.73,2.69) | 1.00 | 0.75(0.67,0.84) | 0.67(0.56,0.79) |
| Model1*+Waist Height Ratio | 2.34(1.87,2.92) | 1.00 | 0.70(0.63,0.78) | 0.57(0.48,0.67) |
Notes:
Multivariable models with adjustment for age, sex, drinking status (never drinker, former drinker, current drinker), smoking index [cigarettes per day*years (null, light:< 200, moderate: 200-400, heavy: >400)], physical activity (yes or no), family history of cardio-cerebrovascular diseases (yes or no).
Figure 1.
Adjusted hazard ratio (34) and 95% confidence interval (95% CI) for all-cause mortality according to BMI, waist circumference, and waist-to-height ratio.
We conducted stratified analysis by age (<60, 60–70, and >70 years old). Among participants less than 60 years old, the nadir of curve for BMI and mortality was around 24 kg/m2 (Supplementary Figure 1A). However, the curve shifted rightward and became flat with nadir around 26 kg/m2 in elderly groups (60–70 years old and >70 years old) (Supplementary Figure 1B, 1C).We further did the stratified analysis by smoke status. Underweight subjects consistently had increased risk of all-cause mortality among the three subgroups. In never-smokers, overweight or obesity were not associated with mortality risk (Supplementary Figure 2) and the nadir of curve was 25.3 kg/m2 (Supplementary Figure 1D). Among the current-smokers (HR=0.80; 95% CI: 0.66 to 0.96) and ever-smokers (HR=0.79; 95% CI: 0.64 to 0.95), overweight subjects had significant decreased risk of all-cause mortality (Supplementary Figure 2). The BMI with lowest mortality were close to 26 kg/m2 among current smokers and ex-smokers (Supplementary Figure 1E, 1F). The nadir were around 25 kg/m2 when excluding those with pre-existing diseases and/or deaths during the first year of follow-up (Supplementary Figure 1G, 1H, 1I).
After adjustment for traditional risk factors, marginal significant associations were found between waist circumference and all-cause mortality risk (P for trend = 0.08). However, further controlling for BMI, participants in Q2-Q5 of waist circumference had higher all-cause mortality risk with HRs of 1.05, 1.09, 1.29, and 1.58 respectively (P trend < 0.001; Table 3 and Figure 1A.b, Figure 1B.b) when compared with participants in Q1. Similar and even stronger associations were observed for the WHtR with all-cause mortality risk. Further exclusion of early death during the first year of follow up and participants with chronic diseases (CHD, DM, stroke, and cancer) at baseline did not materially change the results (Supplementary Table 3). It's worth noting that waist circumference and WHtR were highly related to BMI among participants, the correlation coefficients were 0.732 and 0.752, respectively (Supplemental Figure 3).
Table 3.
Associations of waist circumference and WHtR with all-cause mortality risk (DFTJ cohort, n=26,143)
| Quintiles of measurements | P Trend | ||||||
|---|---|---|---|---|---|---|---|
| Q1 | Q2 | Q3 | Q4 | Q5 | |||
| Waist Circumference | |||||||
| WC (cm) | Male | < 77.8 | 77.8-83 | 83-88 | 88-93 | > ([0-9]+) | |
| Female | < 74 | 74-78.8 | 78.8-83 | 83-89.4 | > ([0-9]+).4 | ||
| Participants, | 5228 | 5228 | 5229 | 5228 | 5230 | ||
| Deaths, n (%) | 420(8.03) | 406(7.77) | 396(7.57) | 453(8.66) | 571(10.92) | ||
| Age and sex adjusted | 1.00 | 0.94(0.82,1.08) | 0.90(0.78,1.03) | 1.00(0.87,1.14) | 1.12(0.91,1.28) | 0.03 | |
| Model1* | 1.00 | 0.95(0.82,1.08) | 0.90(0.78,1.04) | 1.00(0.87,1.14) | 1.09(0.96,1.24) | 0.08 | |
| Model1*+BMI | 1.00 | 1.05(0.91,1.21) | 1.09(0.93,1.27) | 1.29(1.10,1.51) | 1.58(1.33,1.89) | < 0.001 | |
| Waist-Height ratio | |||||||
| WHtR | Male | <46.8 | 46.8-50 | 50-52.6 | 52.6-55.7 | >55.7 | |
| Female | <47.1 | 47.1-50.3 | 50.3-53.3 | 53.3-57.2 | >57.2 | ||
| Participants | 5228 | 5228 | 5229 | 5228 | 5230 | ||
| Deaths, n (%) | 387(7.40) | 379(7.25) | 398(7.61) | 473(9.05) | 609(11.64) | ||
| Age and sex adjusted | 1.00 | 0.93(0.80,1.07) | 0.93(0.81,1.07) | 1.03(0.90,1.18) | 1.18(1.04,1.34) | 0.0008 | |
| Model1* | 1.00 | 0.93(0.81,1.07) | 0.93(0.81,1.07) | 1.03(0.90,1.18) | 1.15(1.01,1.31) | 0.07 | |
| Model1*+BMI | 1.00 | 1.06(0.92,1.23) | 1.16(0.99,1.35) | 1.41(1.20,1.66) | 1.81(1.51,2.17) | <0.0001 | |
Notes:
Multivariable models with adjustment for age, sex, drinking status (never drinker, former drinker, current drinker), smoking index [cigarettes per day*years (null, light:<200, moderate: 200-400, heavy: >400)], physical activity (yes or no), family history of cardio-cerebrovascular diseases (yes or no).
BMI and waist circumference improved the predictive ability of all-cause mortality risk
The area under the receiver operating characteristic curves for predictive models were shown in Table 4. Compared with the basic predictive model, the AUC increased with statistical significance after addition of both BMI and WC/WHtR into the basic model. For the NRI, after introduction of BMI and WC together into the basic model, the discriminative power increased significantly (NRI = 2.57%; P= 0.0007). Similar findings were obtained for BMI and WHtR (NRI = 3.55%; P < 0.0001), indicating that the correct movement of participants across predefined risk categories. The classification matrix indicates that the net correctly reclassified rate of predicted risk significantly increased when addition of BMI and WC/WHtR simultaneously (Supplemental Table 6 and 7) into basic model, while that increased a little when addition of measurements alone (Supplemental Table 2, 3, and 4)
Table 4.
Area under the receiver operating characteristic curves for the prediction of all-cause mortality in different models and differences between these models(DFTJ cohort, n=26,143)
| AUC (95% CI) | P for difference in AUC | NRI (%) (95% CI) | P for difference in NRI | |
|---|---|---|---|---|
| Basic Model* (reference) | 0.756(0.746,0.766) | |||
| Ref + BMI | 0.757(0.747,0.768) | 0.06 | 0.8 (-0.38, 1.98) | 0.15 |
| Ref + waist circumference | 0.757(0.747, 0.766) | 0.07 | 0.83 (-0.26, 2.01) | 0.14 |
| Ref + Waist-Height Ratio | 0.757(0.747, 0.767) | 0.08 | 0.45 (-0.73, 1.63) | 0.41 |
| Ref + BMI and waist circumference | 0.760(0.750,0.770) | 0.0002 | 2.57 (2.55, 2.58) | 0.0007 |
| Ref + BMI and Waist-Height Ratio | 0.762(0.752,0.772) | <0.0001 | 3.55 (3.53, 3.57) | < 0.0001 |
Notes:
Basic models included age, sex, drinking status (never drinker, former drinker, current drinker), smoking index [cigarettes per day*years (null, light: <200, moderate: 200-400, heavy: >400)], physical activity (yes or no), and family history of cardio-cerebrovascular diseases (yes or no); NRI=Net Reclassification Improvement, computed to indicate the proportion of subjects reclassified correctly (NRI>0) or incorrectly (NRI<0) into various risk categories after adding new variable to the former model, we stratified participants into four risk categories (<5%, 5% to 10%, 10%to 20%, and ≥20%) according to the all-cause mortality.
Discussion
Among the present middle-aged and elderly Chinese population, BMI was reverse J-shaped associated with all-cause mortality risk. Underweight participants had higher risk of death, while overweight and obesity individuals had decreased risk of death after adjustment for waist circumference. WC and WHtR were consistently related to increased risk of death after adjustment for BMI. Combination of BMI and WC/WHtR improved the predictive ability of all-cause mortality risk.
Consistent with the previous studies, (21, 22, 23) in the present study underweight subjects had highest risk of mortality. The underlying diseases of these underweight individuals might explain the higher risk of mortality. However, when we excluded subjects with chronic disease at baseline or those died during the first year of follow-up (deaths are more likely due to the preexisting diseases), the association still remained. Studies indicated that underweight was associated with undernutrition, (4) low grade of inflammation, (24) and sarcopenia, (25) which potentially explained the enhanced risk of death. More studies aims to investigate the underlying mechanism of high risk of mortality in underweight is warranted.
Smoking was an important confounder of causal relationship between BMI and mortality risk. (26,27) A recent meta-analysis indicated that overweight was associated with decreased mortality risk in current smokers but not in never smokers and ex-smokers, and the best BMI was 22–23 kg/m2 among healthy never smokers.(28) Similar results were found in the present study. However, the nadir of BMI shifted to right. In sensitivity analysis, the nadirs of BMI shifted to left- when excluding those with pre-existing diseases and/or deaths during the first year of follow-up as the meta-analysis, but the BMI with lowest mortality was still above the current normal range, which was consistent with other previous studies (2, 29, 30). Shorter duration of follow-up and lower smoking rate among females might result in controversy (8, 28).
For the inconsistent association of BMI with mortality risk, researchers focused on sex, ethnic-origin, (31), and body fat percentage (32), which were associated with smoking status and pre-diagnostic disease. Additionally, the association of BMI and mortality became decreasingly U-shaped with advanced age of 70–95 years in some researches, (33, 34), similar as the present study. In a meta-analysis, poplation-attributable fraction for all-cause mortality due to overweight and obesity was only 5% in East Asia (35). For Asian population, especially middle-aged and advanced age individuals, BMI may not be appreciated to predict mortality independently.
In the present study, overweight and obesity subjects had lower risk of all-cause mortality compared with the normal weight individuals after adjustment for waist circumference or WHtR, which was consistent with the previous studies (5, 6). In the present stratified analysis, subjects equal to or elder than 60 years old with BMI around 26 kg/m2 had the lowest risk of all-cause mortality, consistent with the findings among aged participants from a latest meta-analysis (35). Among those less than 60 years old the nadir of BMI for all-cause mortality was 24.1 kg/m2. This might be due to the decline in height in elders, which results in a false increase in BMI (1). Besides, compared with underweight individuals who were prone to external hazards, overweight or obesity ones might provide a metabolic buffer for chronic diseases in elders. In addition, it is proposed that fat mass plays a role in calorie depot for muscles and brain, and serves as a reserve of energy to be mobilized in the event of depression or an acute illness (17).
Abdominal obesity assessed by WC and WHtR has been repeatedly reported as a risk factor for death in older people (10, 36, 37). In comparison with the subcutaneous adipose tissue, the visceral adipose tissue has stronger effects on the metabolic syndrome and insulin resistance. In addition, WC and WHtR have been identified as better markers of metabolic risk than BMI (10). In the present study before adjustment for BMI in the model, the highest quintile of WC or WHtR was not related to significant risk of death, however, as reported previously (11), controlling for BMI strengthened the positive association of WC and WHtR with all-cause mortality risk, which might be due to that the general obesity offsets part of the effects of abdominal obesity on all-cause mortality (38). Elevated WC or WHtR indicated more fat mass aggregation at abdomen or viscera and less lean mass or subcutaneous adipose in arms and legs among individuals with the same BMI, while peripheral lean mass could be predictive against cardiovascular risk factors (39) and disturb glucose metabolism (40). For the individuals with both lower BMI ,higher WC and WHtR, elevated risk of death might result from the harm effects of both undernutrition and visceral adiposity.
In the present study addition of BMI and WC/WHtR in the traditional prediction model of all-cause mortality significantly increased the prediction power with improved AUC. NRI was considered as more sensitive measurement to assess the contribution of indicator to prediction model when evaluating the proportion of subjects moving accurately or inaccurately from one risk category to another. (19)Addition of both BMI and WC/WHtR into the basic predictive model significantly improved predictive ability of all-cause mortality risk when assessed by NRI. Considering the distribution of body fat, introduction of both abdominal and general obesity into prediction model was more accurate to group participants into higher-risk or lower-risk categories.
In the present study, we obtained the anthropometric data from trained health technicians following standardized procedures rather than self-reported, and we adjusted for confounders much more comprehensively. Nevertheless, limitations should also be considered. Firstly, the duration of the follow-up was relatively short and the risk of general obesity or abdominal obesity might be underestimated. Secondly, in the present study we only examined the effects of BMI and WC/WHtR on all-cause mortality risk but not specific-cause mortality because of the relative small sample size of the specific-cause mortality. Thirdly, although we thoroughly adjusted for traditional risk factors in the multivariate model, the residual confounding may still exist. Finally, our findings were restricted to the middle-aged and elderly Chinese population and might not be generalized to other population.
In conclusion, in the present study BMI and waist circumference were independently associated with all- cause mortality after mutual adjustment, suggesting that it is necessary to measure both BMI and waist circumference to examine the mortality related to obesity in elder population. In addition, combined BMI with waist circumference slightly improved predictive ability of all-cause mortality risk on the basis of traditional risk factors. However, further studies in different population especially in elder population were still warranted to validate our findings.
Ethics approval and consent to participate
The study was approved by the Medical Ethics Committee of the School of Public Health, Tongji Medical College, and Dongfeng General Hospital, DMC. All participants provided written informed consent.
Conflict of interests
None. The authors declare that they have no competing interests
Funding
This work was supported by the grants from the National Natural Science Foundation (grants NSFC-81473051 and 81522040); the National Key Research and Development Program of China (2017YFC0907501) and the Program for HUST Academic Frontier Youth Team.
Authors' contributions
Hua Hu, and Meian He conceived and designed the study. All authors acquired, analyzed, or interpreted data and critically revised the manuscript for important intellectual content. Jing Yuan, Xiaoping Miao checked the data extraction. Hua Hu did the statistical analysis and drafted the manuscript. Meian He obtained funding and supervised the study. Meian He had full access to all of the data and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Electronic supplementary material
Supplementary material is available for this article at https://doi.org/10.1007/s12603-018-1047-z and is accessible for authorized users.
Supplemental Document
References
- 1.Zamboni M, Mazzali G, Zoico E, et al. Health consequences of obesity in the elderly: a review of four unresolved questions. Int J Obes (Lond). 2005;29(9):1011–1029. doi: 10.1038/sj.ijo.0803005. 10.1038/sj.ijo.0803005 [DOI] [PubMed] [Google Scholar]
- 2.Tamakoshi A, Yatsuya H, Lin Y, et al. BMI and all-cause mortality among Japanese older adults: findings from the Japan collaborative cohort study. Obesity (Silver Spring). 2010;18(2):362–369. doi: 10.1038/oby.2009.190. 10.1038/oby.2009.190 [DOI] [PubMed] [Google Scholar]
- 3.Winter JE, MacInnis RJ, Wattanapenpaiboon N, Nowson CA. BMI and all-cause mortality in older adults: a meta-analysis. Am J Clin Nutr. 2014;99(4):875–890. doi: 10.3945/ajcn.113.068122. 10.3945/ajcn.113.068122 PubMed PMID: 24452240. [DOI] [PubMed] [Google Scholar]
- 4.Wirth R, Streicher M, Smoliner C, et al, The impact of weight loss and low BMI on mortality of nursing home residents -Results from the nutritionDay in nursing homes, Clin Nutr., 2015, 1–7 [DOI] [PubMed]
- 5.Price GM, Uauy R, Breeze E, Bulpitt CJ, Fletcher AE. Weight, shape, and mortality risk in older persons: elevated waist-hip ratio, not high body mass index, is associated with a greater risk of death. Am J Clin Nutr. 2006;84(2):449–460. doi: 10.1093/ajcn/84.1.449. 10.1093/ajcn/84.2.449 PubMed PMID: 16895897. [DOI] [PubMed] [Google Scholar]
- 6.Flicker L, McCaul KA, Hankey GJ, et al. Body mass index and survival in men and women aged 70 to 75. J Am Geriatr Soc. 2010;58(2):234–241. doi: 10.1111/j.1532-5415.2009.02677.x. 10.1111/j.1532-5415.2009.02677.x PubMed PMID: 20370857. [DOI] [PubMed] [Google Scholar]
- 7.Janssen I. Morbidity and mortality risk associated with an overweight BMI in older men and women. Obesity (Silver Spring). 2007;15:1827–1840. doi: 10.1038/oby.2007.217. 10.1038/oby.2007.217 [DOI] [PubMed] [Google Scholar]
- 8.Flegal KM, Kit BK, Orpalna H, Graubard BI. Association of all-cause mortality with overweight and obesity using standard body mass index categories. JAMA. 2013;309(1):71–82. doi: 10.1001/jama.2012.113905. 10.1001/jama.2012.113905 PubMed PMID: 23280227, PMCID 4855514. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Zaccardi F, Dhalwani NN, Papamargaritis D, et al. Nonlinear association of BMI with all-cause and cardiovascular mortality in type 2 diabetes mellitus: a systematic review and meta-analysis of 414,587 participants in prospective studies. Diabetologia. 2017;60(2):240–248. doi: 10.1007/s00125-016-4162-6. 10.1007/s00125-016-4162-6 PubMed PMID: 27888288. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Cameron AJ, Magliano DJ, Soderberg S. A systematic review of the impact of including both waist and hip circumference in risk models for cardiovascular diseases, diabetes and mortality. Obes Rev. 2013;14(1):86–94. doi: 10.1111/j.1467-789X.2012.01051.x. 10.1111/j.1467-789X.2012.01051.x PubMed PMID: 23072327. [DOI] [PubMed] [Google Scholar]
- 11.Cerhan JR, Moore SC, Jacobs EJ, et al. A pooled analysis of waist circumference and mortality in 650,000 adults. Mayo Clin Proc. 2014;89(3):335–345. doi: 10.1016/j.mayocp.2013.11.011. 10.1016/j.mayocp.2013.11.011 PubMed PMID: 24582192, PMCID 4104704. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Hollander EL, Bemelmans WJ, Boshuizen HC, et al. The association between waist circumference and risk of mortality considering body mass index in 65-to 74-yearolds: a meta-analysis of 29 cohorts involving more than 58 000 elderly persons. Int J Epidemiol. 2012;41(3):805–817. doi: 10.1093/ije/dys008. 10.1093/ije/dys008 PubMed PMID: 22467292, PMCID 4492417. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Guallar-Castillon P, Balboa-Castillo T, Lopez-Garcia E, et al. BMI, waist circumference, and mortality according to health status in the older adult population of Spain. Obesity (Silver Spring). 2009;17(12):2232–2238. doi: 10.1038/oby.2009.115. 10.1038/oby.2009.115 [DOI] [PubMed] [Google Scholar]
- 14.Wang F, Zhu J, Yao P, et al. Cohort profile: the Dongfeng-Tongji cohort study of retired workers. Int J Epidemiol. 2013;42(3):731–740. doi: 10.1093/ije/dys053. 10.1093/ije/dys053 PubMed PMID: 22531126. [DOI] [PubMed] [Google Scholar]
- 15.Sun DM, Li FF, Zhang Y, Xu XM. Associations of the pre-pregnancy BMI and gestational BMI gain with pregnancy outcomes in Chinese women with gestational diabetes mellitus. Int J Clin Exp Med. 2014;7(12):5784–5789. PubMed PMID: 25664107, PMCID 4307554. [PMC free article] [PubMed] [Google Scholar]
- 16.Ishikawa Y, Furuta R, Miyoshi T, et al. Loss of heterozygosity and the smoking index increase with decrease in differentiation of lung adenocarcinomas:etiologic implications. Cancer Letters. 2002;187:47–51. doi: 10.1016/s0304-3835(02)00383-x. 10.1016/S0304-3835(02)00383-X PubMed PMID: 12359350. [DOI] [PubMed] [Google Scholar]
- 17.Roberson LL, Aneni EC, Maziak W, et al. Beyond BMI: The “Metabolically healthy obese phenotype & its association with clinical/subclinical cardiovascular disease and all-cause mortality–a systematic review. BMC Public Health. 2014;14:14–27. doi: 10.1186/1471-2458-14-14. 10.1186/1471-2458-14-14 PubMed PMID: 24400816, PMCID 3890499. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Rolland Y, Gallini A, Cristini C, et al. Body-composition predictors of mortality in women aged >/= 75 y: data from a large population-based cohort study with a 17-y follow-up. Am J Clin Nutr. 2014;100(5):1352–1360. doi: 10.3945/ajcn.114.086728. 10.3945/ajcn.114.086728 PubMed PMID: 25332333. [DOI] [PubMed] [Google Scholar]
- 19.Tam CH, Ho JS, Wang Y, et al. Use of net reclassification improvement (NRI) method confirms the utility of combined genetic risk score to predict type 2 diabetes. PLoS One. 2013;8(12):e83093. doi: 10.1371/journal.pone.0083093. 10.1371/journal.pone.0083093 PubMed PMID: 24376643, PMCID 3869744. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Petursson H, Sigurdsson JA, Bengtsson C, Nilsen TI, Getz L. Body configuration as a predictor of mortality: comparison of five anthropometric measures in a 12 year follow-up of the Norwegian HUNT 2 study. PLoS One. 2011;6(10):e26621. doi: 10.1371/journal.pone.0026621. 10.1371/journal.pone.0026621 PubMed PMID: 22028926, PMCID 3197688. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Gao S, Jin Y, Unverzagt FW, Cheng Y, Su L, Wang CK. Cognitive function, body mass index and mortality in a rural elderly Chinese cohort. Arch Public Health. 2014;72:9–17. doi: 10.1186/2049-3258-72-9. 10.1186/2049-3258-72-9 PubMed PMID: 24666663, PMCID 3974191. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Pan WH, Yeh WT, Chen HJ, Chuang SY, Chang HY, Chen L. The U-shaped relationship between BMI and all-cause mortality contrasts with a progressive increase in medical expenditure: a prospective cohort study. Asia Pac J Clin Nutr. 2012;21(4):577–587. PubMed PMID: 23017316. [PubMed] [Google Scholar]
- 23.Lin WY, Tsai SL, Albu JB, Lin CC, Li TC, Pi-Sunyer FX. Body mass index and allcause mortality in a large Chinese cohort. CMAJ. 2011;183(6):645–646. doi: 10.1503/cmaj.101303. 10.1503/cmaj.100303 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Nakajima K, Yamaoka H, Morita K, et al. Elderly people with low body weight may have subtle low-grade inflammation. Obesity (Silver Spring). 2009;17(4):803–808. doi: 10.1038/oby.2008.596. 10.1038/oby.2008.596 [DOI] [PubMed] [Google Scholar]
- 25.Szulc P, Duboeuf F, Marchand F, Delmas PD. Hormonal and lifestyle determinants of appendicular skeletal muscle mass in men: the MINOS study. Am J Clin Nutr. 2004;80(2):496–503. doi: 10.1093/ajcn/80.2.496. 10.1093/ajcn/80.2.496 PubMed PMID: 15277176. [DOI] [PubMed] [Google Scholar]
- 28.Aune D, Sen A, Prasad M, et al. BMI and all cause mortality: systematic review and non-linear dose-response meta-analysis of 230 cohort studies with 3.74 million deaths among 30.3 million participants. BMJ. 2016;353:i2156. doi: 10.1136/bmj.i2156. 10.1136/bmj.i2156 PubMed PMID: 27146380, PMCID 4856854. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Chung WS, Ho FM, Cheng NC, Lee MC, Yeh CJ. BMI and all-cause mortality among middle-aged and older adults in Taiwan: a population-based cohort study. Public Health Nutr. 2015;18(10):1839–1846. doi: 10.1017/S136898001400281X. 10.1017/S136898001400281X PubMed PMID: 25482035. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Wang Z, Dong B, Hu J, Adegbija O, Arnold LW. Exploring the non-linear association between BMI and mortality in adults with and without diabetes: the US National Health Interview Survey. Diabetic Medicine. 2016;33(12):1691–1699. doi: 10.1111/dme.13111. 10.1111/dme.13111 PubMed PMID: 26972695. [DOI] [PubMed] [Google Scholar]
- 31.Dankner R, Shanik M, Roth J, Luski A, Lubin F, Chetrit A. Sex and ethnic-origin specific BMI cut points improve prediction of 40-year mortality: the Israel GOH study. Diabetes Metab Res Rev. 2015;31(5):530–536. doi: 10.1002/dmrr.2642. 10.1002/dmrr.2642 PubMed PMID: 25689480. [DOI] [PubMed] [Google Scholar]
- 32.Padwal R, Leslie WD, Lix LM, Majumdar SR. Relationship Among Body Fat Percentage, Body Mass Index, and All-Cause Mortality: A Cohort Study. Ann Intern Med. 2016;164(8):532–541. doi: 10.7326/M15-1181. 10.7326/M15-1181 PubMed PMID: 26954388. [DOI] [PubMed] [Google Scholar]
- 33.Clark DO, Gao S, Lane KA, et al. Obesity and 10-year mortality in very old African Americans and Yoruba-Nigerians: exploring the obesity paradox. J Gerontol A Biol Sci Med Sci. 2014;69(9):1162–1169. doi: 10.1093/gerona/glu035. 10.1093/gerona/glu035 PubMed PMID: 24694355, PMCID 4202260. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Thinggaard M, Jacobsen R, Jeune B, Martinussen T, Christensen K. Is the relationship between BMI and mortality increasingly U-shaped with advancing age? A 10-year follow-up of persons aged 70–95 years. J Gerontol A Biol Sci Med Sci. 2010;65(5):526–531. doi: 10.1093/gerona/glp214. 10.1093/gerona/glp214 PubMed PMID: 20089666. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Angelantonio ED, Bhupathiraju SN, Wormser D, et al. Body-mass index and allcause mortality: individual-participant-data meta-analysis of 239 prospective studies in four continents. The Lancet. 2016;388:776–786. doi: 10.1016/S0140-6736(16)30175-1. 10.1016/S0140-6736(16)30175-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Lu Y, Hajifathalian K M, Ezzati M, Woodward M, Rimm EB, Danaei G. Metabolic mediators of the effects of body-mass index, overweight, and obesity on coronary heart disease and stroke: a pooled analysis of 97 prospective cohorts with 1•8 million participants. The Lancet. 2014;383(9921):970–983. doi: 10.1016/S0140-6736(13)61836-X. 10.1016/S0140-6736(13)61836-X [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Staiano AE, Reeder BA, Elliott S, et al. Body mass index versus waist circumference as predictors of mortality in Canadian adults. Int J Obes (Lond). 2012;36(11):1450–1454. doi: 10.1038/ijo.2011.268. 10.1038/ijo.2011.268 PMCID 4120111. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Janssen I, Katzmarzyk PT, Ross R. Body mass index is inversely related to mortality in older people after adjustment for waist circumference. JAGS. 2005;53(12):2112–2118. doi: 10.1111/j.1532-5415.2005.00505.x. 10.1111/j.1532-5415.2005.00505.x [DOI] [PubMed] [Google Scholar]
- 39.Ferreira I, Snijder MB, Twisk JWR, et al. Central fat mass versus peripheral fat and lean mass: opposite (adverse versus favorable) associations with arterial stiffness? The amsterdam growth and health longitudinal study. J Clin Endocrinol Metab. 2004;89(6):2632–2639. doi: 10.1210/jc.2003-031619. 10.1210/jc.2003-031619 PubMed PMID: 15181034. [DOI] [PubMed] [Google Scholar]
- 40.Snijder MB, Dekker JM, Visser M, et al. Trunk fat and leg fat have independent and opposite associations with fasting and postload glucose levels: The Hoorn Study. Diabetes Care. 2004;27:372–377. doi: 10.2337/diacare.27.2.372. 10.2337/diacare.27.2.372 PubMed PMID: 14747216. [DOI] [PubMed] [Google Scholar]
Uncited references
- 26.Tobias DK, Pan A, Jackson CL, et al. Body-mass index and mortality among adults with incident type 2 diabetes. N Engl J Med. 2014;370(3):233–244. doi: 10.1056/NEJMoa1304501. 10.1056/NEJMoa1304501 PubMed PMID: 24428469, PMCID 3966911. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Gary Whitlock SL. Body-mass index and cause-specific mortality in 900 000 adults: collaborative analyses of 57 prospective studies. Lancet. 2009;373:1083–1096. doi: 10.1016/S0140-6736(09)60318-4. 10.1016/S0140-6736(09)60318-4 PubMed PMID: 19299006. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplemental Document