text new page (beta)
English (pdf)
Article in xml format
Curriculum ScienTI
Automatic translation
Cited by SciELO 
Cited by Google 
Similars in
SciELO 
Similars in Google 
Permalink
Introduction
Previous studies have shown the relationship between overweight and obesity and the risk of mortality from cardiovascular disease, cancer and all-cause 1,2. Most prospective studies assessing the association between overweight and mortality risk used the body mass index (BMI) to evaluate nutritional status 3. Widely applied to predict the risk of different health events related to overweight or obesity, its use has been questioned, as BMI does not distinguish between the accumulation of lean or fat mass or the distribution of adipose tissue 4, i.e., a high BMI may not necessarily reflect an increase in adiposity. Additionally, BMI validity as an adiposity indicator in older adults is limited 5, as lean mass decreases with aging even without a change in body weight 6.
Waist circumference (WC), an indicator of abdominal adiposity, has been used to complement BMI assessment. High WC is associated with an increase in the prevalence of cardiovascular disease and in the risk of premature death, even when BMI is within the normal range 7. Its use to properly estimate body fat distribution, however, is limited since it disregards that taller individuals have a higher WC 8,9. Additionally, the high correlation between WC and BMI hinders determining the independent role of general and abdominal fat as a risk factor for health outcomes 10,11.
Measures derived from WC, such as waist-to-height ratio (WHtR) and waist-to-hip ratio have also been studied, showing high correlation with BMI 12. Considering these limitations, Krakauer & Krakauer 11 proposed a measure named a body shape index (ABSI), based on WC and adjusted for height and weight. A high ABSI value indicates that WC is greater than expected for a given height and weight, thus corresponding to a more centralized body volume concentration. Thus, ABSI helps quantify the risk associated with abdominal obesity that cannot be attributed to BMI.
ABSI performed better than WC and BMI in predicting the risk of all-cause death in an North American population aged 18 or over, with the authors stressing that their predictive power differed between ethnic groups, with less power for Latinos compared with Caucasians and African Americans 11. In a recent meta-analysis conducted with 24 retrospective cohorts from 15 countries ABSI showed a superior performance in predicting all-cause mortality when compared with WC and BMI 13. Among the included cohorts, no study included only older adults or a Latin American population. To our knowledge, no study have evaluated the association of ABSI with all-cause mortality in the Brazilian population 14, especially in older adults, a rapidly growing group in developing countries such as Brazil 15. As the Brazilian population is highly mixed 16, our study can contribute to assess ABSI applicability and other indirect adiposity measures in different populations.
Moreover, most studies in the literature used only one anthropometric measure, obtained at baseline, to assess the association between ABSI and mortality 13, which is insufficient to reveal its impact on mortality. Thus, it remains unknown to what extent ABSI varies over time and whether this change would affect the risk of mortality. In this context, this study aimed to assess the joint and separate association of abdominal adiposity indicators (ABSI, WC and WHtR) and BMI with mortality among older adults, using anthropometric measures as time dependent variables.
Methodology
Study population and data collection
The Bambuí Cohort Study of Aging is a prospective population-based study conducted in the city of Bambuí, a municipality located in the state of Minas Gerais, Brazil, 215km from Belo Horizonte. The city had about 15,000 inhabitants when the cohort baseline was established in 1997 17. The baseline population consisted of all residents aged 60 and over on January 1, 1997, identified using a complete city census conducted in 1996. Out of 1,742 older adults, 1,606 (92.19%) participated in the interview and 1,454 (83.47%) underwent physical and laboratory tests.
The cohort’s baseline data collection took place from February to May 1997, and from then on, participants were monitored annually by interviews, checking death certificates and physical examinations and collection of biological material, in selected years. Interviews were conducted in the participants’ homes and answered by the older adult or by a close relative, when the interviewee was unable to answer the questions due to cognitive deficit or other health problem. Clinical and anthropometric evaluation, as well as blood collection for biochemical analysis, took place in the field clinic; in the case of a health limiting condition, the evaluations happened at home. All data were collected by interviewers and previously trained and certified technicians, with at least 11 years of schooling. The present study included data collected between the baseline and December 31, 2011, totaling approximately a 15-year follow-up. Of the 1,606 cohort participants, 1,454 underwent anthropometric measurements at the beginning of the study; of these, 88 (5.48%) were excluded due to the lack of information for confounding variables, resulting in a final sample of 1,366 older adults.
The Bambuí project was approved by the Ethics Research Committee of the Oswaldo Cruz Foundation (Fiocruz), in Rio de Janeiro on November 4, 1996. All participants signed an Informed Consent Form (ICF) with information about the research, in addition to the authorization to check death records.
Mortality
The present analysis included all deaths that occurred between the baseline (1997) and December 31, 2011. These were reported by family members during the annual follow-up interviews and confirmed by the Brazilian Ministry of Health’s Mortality Information System (SIM). Death certificates were obtained for 98.90% of the Bambuí cohort participants who had their deaths reported by family members. In the present study, the all-cause time-until-death was the outcome variable.
Anthropometric assessment
Weight and height were obtained using standardized techniques and equipment (CMS Weighing Equipment Ltd., London, United Kingdom), with the older adults wearing light clothes and barefoot. Weight was measured using a portable digital electronic scale, with a capacity of 150kg and precision of 50g. Height was measured in the vertical position, using an aluminium stadiometer capable of measuring up to 2m and with a 0.1cm accuracy. WC was measured using a flexible and inelastic measuring tape at the midpoint between the last rib and the iliac crest, with the interviewee standing up. WHtR was calculated by dividing WC (cm) by height (cm), and BMI by dividing the individual’s weight (kg) by height (m) squared. ABSI was calculated using the formula: WC (m) ÷ (BMI2/3 X height (m)1/2) 11.
All these measures were collected in 1997, 2000, 2002 and 2008, using the same procedures. The anthropometric indicators were calculated for these four years and considered as time dependent variables in the association analyses.
Potential confounding factors
The variables used to adjust our analyses were obtained at baseline and chosen in line with specialized literature, for they are associated with anthropometric indicators and mortality, especially in older populations 18,19.
We considered as sociodemographic variables: age (in years), self-reported skin color (white, brown or black), gender (male or female) and schooling (illiterate, 1-3 years and 4 or more years). Health behaviors included current smoking (those who reported having smoked 100 or more cigarettes during their lives and continued to smoke until the time of the interview), at-risk alcohol consumption (seven or more doses per week in the 12 months prior to the interview) and insufficient physical activity (evaluated using metabolic equivalents-MET, less than 450 MET.minutes/week). Estimation of energy expenditure was based on 23 multiple choice questions and two open questions about physical activities performed in the last 90 days, considering all domains. The questions included the type and frequency of the activity and the average time (in minutes) spent on each activity, which allowed to estimate the metabolic equivalents.
As Bambuí was an endemic area for Chagas disease, the analyses considered positive serology for T. cruzi as a potential confounding variable. Infection was defined by positive serology in three tests: hemagglutination assay (Biolab Mérieux, Rio de Janeiro, Brazil) and two immunoenzymatic tests (Elisa) (Abbott Laboratories Inc., Chicago, USA; and Wiener Laboratories, Rosario, Argentina).
Statistical analysis
The descriptive analysis included measures of central tendency and dispersion for quantitative variables and frequency distribution for qualitative variables. Raw mortality rates were estimated using people-years at risk as the denominator. The association between anthropometric indicators (considering the four measures) and time-until-death was assessed using the Cox proportional hazards model, obtaining hazard ratio (HR) estimates and respective 95% confidence intervals (95%CI). The anthropometric indicators underwent quartile analysis for better comparability between our results and those of existing studies. The proportionality assumption of risks was tested using Schoenfeld residuals.
As no significant interaction was observed between gender and the anthropometric indicators (p > 0.05), the results of the extended Cox regression were presented for the entire sample. All variables considered in the analyses were included in the adjusted model only after verifying their collinearity (variance inflation factor < 5).
All models were estimated without adjustments and, subsequently, including the confounding factors. Thus, the association between anthropometric measures and mortality was evaluated by the following models: Model 1 evaluates the association of each anthropometric indicator adjusted for confounding factors; Model 2 evaluates the association of each abdominal adiposity indicator adjusted for BMI (continuous) and other confounding factors; Model 3 evaluates the association of BMI adjusted for WC (continuous) and other confounding factors; and Model 4 evaluates the association of BMI adjusted for ABSI (continuous) and other confounding factors.
Including individuals with diseases in studies that assess the association between excess weight and mortality has been identified as a potential source of bias 20, as these conditions can lead to weight change and to predisposed higher mortality 21. Thus, to minimize reverse causality, we performed sensitivity analyses with anthropometric indicators classified in quartiles to verify whether the observed associations would be altered by excluding deaths that occurred in the first two years of follow-up or by excluding those that had chronic diseases at the beginning of the study (i.e., cardiovascular disease, diabetes, hypertension, and arthritis or rheumatism).
Smoking has been considered an important confounding factor that can mask any real effect of being overweight 22, as it is associated with weight reduction and an increased risk of death. Thus, we conducted analyses considering the anthropometric indicators in quartiles only among non-smokers, to minimize the residual confusion of this variable.
We conducted additional analyses with cut-off points commonly used in the literature for WC and BMI. WC was classified as increased when ≥ 80cm and < 94cm for women and ≥ 88cm and < 102cm for men, and substantially increased when ≥ 94cm for women and ≥ 102cm for men 23. Underweight was defined as BMI < 18.5kg/m2, normal weight from 18.5 to < 25kg/m2, overweight from 25 to < 30kg/m2 and obesity ≥ 30kg/m2, according to the cut-off points proposed by the World Health Organization (WHO) 24. All analyses were conducted with Stata version 13.0 (https://www.stata.com), except the database preparation, done with R version 3.4.1 (http://www.r-project.org).
Results
A total of 1,366 older adults (61.27% women) participated in this study, aged 60 to 93 years. During a median 14-year follow-up, 672 participants died (287 men and 385 women), leading to 14,259 person-years of observation. The cohort’s overall mortality rate was 47.12 per 1,000 person-years, being higher in men (55.40 per 1,000 person-years) than women (42.40 per 1,000 person-years) (p < 0.001). Cardiovascular diseases (16.62%), cancer (7.32%) and respiratory diseases (5.49%) were the main causes of death in this population.
At baseline, the medians of ABSI, WC, WHtR and BMI were significantly higher among women (0.0856m11/6kg-2/3, 92cm, 0.60 and 25.65kg/m2, respectively) than among men (0.0841m11/6kg-2/3, 89.50cm, 0.54 and 23.66kg/m2, respectively) (p < 0.05). The median of anthropometric indicators measurements was 2.9, ranging from one (16.50%) to four (40%).
Regarding the characteristics of the analyzed sample, most of the older adult participants were white (59.67%), had four years of schooling or more (37.52%), were non-current smokers (83.50%), presented no at-risk alcohol consumption (95.18%), were active (76.39%) and presented negative serology for Trypanosoma cruzi (63.23%). Gender showed a significant association (p < 0.05) with schooling, current smoking, at-risk of alcohol consumption, insufficient physical activity and positive serology for T. cruzi, in unadjusted analyses (Table 1).
Table 1 Characteristics of the studied sample according to gender. Bambuí Cohort Study of Aging, Brazil (1997-2011).
| Variables | Total | Male | Female | p-value * |
|---|---|---|---|---|
| Age in years [median (IQR)] | 67 (63-72) | 66 (62-72) | 67 (63-72) | 0.203 |
| Self-reported skin color | ||||
| White | 59.67 | 62.05 | 58.29 | 0.054 |
| Brown | 34.80 | 32.54 | 36.11 | |
| Black | 5.53 | 5.41 | 5.60 | |
| Schooling (years) | ||||
| Illiterate | 28.85 | 24.62 | 31.32 | < 0.001 |
| 1-3 | 33.63 | 35.71 | 32.41 | |
| 4 or more | 37.52 | 39.67 | 36.27 | |
| Current smoking | ||||
| No | 83.50 | 72.61 | 89.88 | < 0.001 |
| Yes ** | 16.50 | 27.39 | 10.12 | |
| At-risk alcohol consumption *** | ||||
| No | 95.18 | 87.52 | 99.65 | < 0.001 |
| Yes | 4.82 | 12.48 | 0.35 | |
| Insufficient physical activity # | ||||
| No | 76.39 | 81.45 | 71.84 | < 0.001 |
| Yes | 24.61 | 18.55 | 28.16 | |
| Positive serology for Trypanosoma cruzi | ||||
| No | 63.23 | 69.44 | 59.60 | < 0.001 |
| Yes | 36.77 | 30.56 | 40.40 |
IQR: interquartile range.
Note: values expressed in percentages, except when specified.
* p-value of the Mann Whitney test or Pearson’s chi-square test;
** Smoked 100 or more cigarettes during their lives and continued to smoke;
*** Consumed seven or more doses per week in the 12 months prior to the interview;
# Less than 450 MET. minutes/week, considering all domains.
Table 2 shows the results of the association between anthropometric abdominal adiposity indicators and mortality. Older adults in the fourth ABSI quartile had a higher risk of mortality (HR = 1.32, 95%CI: 1.06-1.64) when compared with those in the first quartile, after adjustment for potential confounding factors; this association remained significant even after adjusting for BMI (HR = 1.27, 95%CI: 1.02-1.58). Another abdominal adiposity indicator, WC, was associated with a lower risk of mortality in the third quartile after adjustment for confounders (HR = 0.77, 95%CI: 0.61-0.96), but this association disappeared when adjusted for BMI. When evaluated with cut-off points commonly used in the literature, WC showed no association with mortality after adjustment for selected variables or when adjusted for BMI. WHtR showed no association with mortality after adjustment for potential confounding factors or BMI.
Table 2 Univariate and multivariable analysis of the association between anthropometric abdominal adiposity indicators and mortality. Bambuí Cohort Study of Aging, Brazil (1997-2011).
| Anthropometric indicators * | Unadjusted model | Model 1 | Model 2 (Adjusted for BMI) |
|---|---|---|---|
| HR (95%CI) | HR (95%CI) | HR (95%CI) | |
| A body shape index (m11/6 X kg-2/3) | |||
| Q1 (reference: < 0.0812) | 1.00 | 1.00 | 1.00 |
| Q2 (0.0812-0.0848) | 0.99 (0.78-1.27) | 0.89 (0.69-1.13) | 0.87 (0.68-1.12) |
| Q3 (0.0849-0.0887) | 1.20 (0.96-1.51) | 0.99 (0.79-1.25) | 0.97 (0.77-1.23) |
| Q4 (> 0.0887) | 1.88 (1.52-2.32) | 1.32 (1.06-1.64) | 1.27 (1.02-1.58) |
| Waist circumference (cm) | |||
| Q1 (reference: < 83.10) | 1.00 | 1.00 | 1.00 |
| Q2 (84.10-91.00) | 0.76 (0.62-0.92) | 0.89 (0.73-1.10) | 0.96 (0.77-1.19) |
| Q3 (91.01-98.50) | 0.61 (0.49-0.76) | 0.77 (0.61-0.96) | 0.87 (0.67-1.13) |
| Q4 (> 98.50) | 0.68 (0.55-0.83) | 0.86 (0.69-1.06) | 1.06 (0.76-1.45) |
| Waist circumference (cm): cut-off points | |||
| Normal | 1.00 | 1.00 | 1.00 |
| Increased | 0.71 (0.58-0.86) | 0.84 (0.68-1.03) | 0.90 (0.71-1.12) |
| Substantially increased | 0.64 (0.54-0.76) | 0.82 (0.67-1.01) | 0.95 (0.72-1.24) |
| Waist-to-height ratio | |||
| Q1 (reference: < 0.52) | 1.00 | 1.00 | 1.00 |
| Q2 (0.52-0.57) | 0.80 (0.65-0.98) | 0.93 (0.75-1.14) | 1.00 (0.80-1.25) |
| Q3 (0.58-0.62) | 0.65 (0.52-0.80) | 0.81 (0.65-1.01) | 0.92 (0.70-1.20) |
| Q4 (> 0.62) | 0.68 (0.55-0.84) | 0.84 (0.67-1.05) | 1.03 (0.74-1.43) |
95%CI: 95% confidence interval; BMI: body mass index; HR: hazard ratio.
Model 1: adjusted for age (continuous), gender, self-reported skin color, schooling, current smoking, at-risk alcohol consumption, insufficient physical activity and positive serology for Trypanosoma cruzi, as described in table 1.
Model 2: adjusted for the same variables as model 1 plus BMI (continuous).
* Divided in quartiles or cut-off points commonly used in the literature.
Older adults in the second, third and fourth BMI quartiles had a lower risk of mortality (HR = 0.79, 95%CI: 0.64-0.96; HR = 0.71, 95%CI: 0.57-0.88; HR = 0.69, 95%CI: 0.55-0.86, respectively), association that remained even after adjustment for WC or ABSI (Table 3). For BMI evaluated with common cut-off points, we found that underweight older adults had a higher risk of mortality (HR = 1.42, 95%CI: 1.11-1.80), while being overweight was a protective factor (HR = 0.82, 95%CI: 0.69-0.99), even after adjustment for WC or ABSI (Table 3).
Table 3 Univariate and multivariable analysis of the association between body mass index (BMI) and mortality. Bambuí Cohort Study of Aging Brazil (1997-2011).
| Anthropometric indicator * | Unadjusted model | Model 1 | Model 3 (adjusted for WC) | Model 4 (adjusted for ABSI) |
|---|---|---|---|---|
| HR (95%CI) | HR (95%CI) | HR (95%CI) | HR (95%CI) | |
| BMI (kg/m2) | ||||
| Q1 (reference: < 21.77) | 1.00 | 1.00 | 1.00 | 1.00 |
| Q2 (21.77-24.78) | 0.62 (0.51-0.75) | 0.79 (0.64-0.97) | 0.71 (0.57-0.89) | 0.80 (0.65-0.98) |
| Q3 (24.79-28.00) | 0.52 (0.21-0.64) | 0.71 (0.57-0.88) | 0.59 (0.45-0.78) | 0.71 (0.57-0.88) |
| Q4 (> 28.00) | 0.45 (0.36-0.56) | 0.69 (0.55-0.86) | 0.53 (0.37-0.74) | 0.71 (0.56-0.89) |
| BMI (kg/m2): WHO cut-off points | ||||
| Normal weight | 1.00 | 1.00 | 1.00 | 1.00 |
| Underweight | 1.89 (1.50-2.40) | 1.42 (1.11-1.80) | 1.52 (1.17-1.97) | 1.38 (1.08-1.76) |
| Overweight | 0.70 (0.58-0.83) | 0.82 (0.69-0.99) | 0.77 (0.62-0.95) | 0.82 (0.69-0.99) |
| Obesity | 0.69 (0.53-0.89) | 0.94 (0.72-1.22) | 0.82 (0.59-1.14) | 0.97 (0.74-1.26) |
95%CI: 95% confidence interval; ABSI: a body shape index; HR: hazard ratio; WC: waist circumference; WHO: World Health Organization.
Model 1: adjusted for age (continuous), gender, self-reported skin color, schooling, current smoking, at-risk alcohol consumption, insufficient physical activity and positive serology for Trypanosoma cruzi, as described in Table 1;
Model 3: adjusted for the same variables as model 1 plus WC (continuous);
Model 4: adjusted for the same variables as model 1 plus ABSI (continuous).
* Divided in quartiles or cut-off points commonly used in the literature.
Our analyses adjusted for health conditions such as hypertension, diabetes, cardiovascular diseases or respiratory diseases had no effect on the associations between anthropometric indicators and mortality (data not shown). However, to avoid controlling intermediate conditions in the causal path between obesity and mortality 25 and obtaining a more parsimonious model, we disregarded these variables in our final analyses.
Sensitivity analysis showed that ABSI had no association with mortality in a model adjusted for BMI, after excluding 59 deaths in the first two years of follow-up (30 men and 29 women), restricting analyses to non-smokers (813) or those without chronic illnesses at baseline (343). Excluding deaths in the first two years or restricting the analysis to non-smokers or those without diseases at baseline, resulted in sensitivity analysis for WC or WHtR adjusted for BMI with findings comparable to those of the total study population. For BMI, excluding deaths at the beginning of follow-up or restricting the analysis to non-smokers, did not change the results. Only the third quartile showed an inverse association with mortality in the analyses conducted with those without chronic diseases, in models adjusted for WC or ABSI.
Discussion
This population-based cohort study using anthropometric measures, evaluated four times during follow-up, identified that a higher ABSI was directly associated with mortality independent of BMI; but this association was not found in the sensitivity analyses conducted. After adjusting for WC or ABSI, BMI remained inversely associated with mortality; while WC and WHtR showed no consistent associations with general mortality after adjusting for confounding factors.
International studies conducted in middle-aged and older adults populations found similar results regarding higher mortality risk among older adults with a high ABSI, when evaluating ABSI separately 26,27,28. Conversely, a Chinese cohort study conducted with 780 middle-aged adults showed that ABSI was not associated with mortality 29.
In the present study, the direct association between ABSI and mortality changed only slightly when BMI was added to the model, showing its independence from BMI in the association with mortality 11. Krakauer & Krakauer 11 developed ABSI to be minimally associated with weight, height and BMI, so that it could be used in conjunction with the latter, to quantify the risk associated with abdominal obesity 30. But few authors have assessed the combined contribution of ABSI and BMI to mortality 27. A study conducted with 6,366 Dutch adults over the age of 55 showed that, when evaluated together, the direction and magnitude of the association for both ABSI and BMI remained similar, being comparable to the models that contained each measure separately 27.
The association between ABSI and a higher mortality risk observed among older adults in Bambuí and in other studies evaluating older populations 26,27,28 may result from the fact that a high ABSI corresponds to greater abdominal (visceral) adiposity and a lower amount of lean mass, and the latter has been shown to have a strong negative correlation with mortality 31,32. Adipose tissue, particularly visceral tissue, secretes potential mediators related to the development of chronic diseases 33, which could explain why abdominal fat distribution, assessed by ABSI, was associated with a higher risk of death, regardless of BMI.
Although we found an association between ABSI and mortality for the total population of Bambuí, this association did not recur in the sensitivity analyses performed. Among the studies that evaluated the relationship between ABSI and mortality, only a few conducted analyses to limit the effects of residual confusion or reverse causality 27,28. A direct association between ABSI and mortality among 77,505 postmenopausal women persisted even after restricting analyses to non-smokers 28. A study conducted with older Dutch adults observed similar results, and excluding deaths during the first five years of follow-up or participants who lost more than three kilograms of weight in the last 12 months prior to the study, did not substantially change the association measures when compared with the total study population 27.
A recent meta-analysis verified that ABSI was a better predictor of mortality, when compared with BMI and WC 13. But this study conducted no analyses to assess potential residual confusion due to smoking or reverse causality due to existing diseases, which could affect the validity of the estimates obtained. Due to the limitations presented by this meta-analysis and the loss of significance in the association between ABSI and mortality in our sensitivity analyses, further studies are needed to confirm the hypothesis of association, in different populations, of this index and general mortality.
WC, another abdominal adiposity indicator, was associated with a lower risk of mortality in the third quartile, a result consistent with the assessment of WC with commonly used cut-off points; but this association did not recur after adjustment for BMI. Additionally, WHtR showed no association with mortality when assessed separately or adjusted by BMI.
Despite recommendations endorsing the use of WC and BMI in clinical practice 34, these indicators are strongly correlated, hindering to assess the independent contribution of abdominal adiposity to mortality 10. But a meta-analysis including 58,609 adults aged 65 to 74 years, showed that increased WC was associated with a higher mortality risk, independent of BMI. In this study, the authors reported absence of multicollinearity among the anthropometric indicators assessed as continuous variables, showing the independent role of abdominal adiposity in mortality 35. On the other hand, our results showed that after adjusting for BMI, WC lost its association with mortality, while WC evaluated with cut-off points showed no association with the outcome in any of the adjusted models. In the population of Bambuí, WC and BMI showed a correlation of 0.78, suggesting that WC is not only an indicator of abdominal adiposity, but also of general adiposity, as already reported by other authors 36.
Some studies have shown inconsistent associations between WC and mortality when the analyses are adjusted for BMI 27,37,38 or even in models that evaluated WC separately 36,39,40. A study conducted with older North American adults showed that WC had a negative association with mortality, which became positive after adjusting for BMI 37. Two other studies found that the association between WC and mortality was evident only after adjusting for BMI 27,38. The discrepancies found when adjusting the model by another indicator may result from the collinearity between BMI and WC. However, none of these studies reported having performed analyses to diagnose multicollinearity.
Besides the limitation due to the lack of multicollinearity assessment, there could be other explanations to the conflicting results of studies that assess the relationship between waist circumference and mortality. Among these issues are the adjustment for mediating variables, ethnic differences in body composition, selection of older adults from different age groups and selective survival. Due to these discrepancies, further studies with standardized methods are needed to verify whether WC should be used to assess mortality risk among older adults.
As for the general adiposity indicator, we found that older adults in the second, third and fourth BMI quartiles had a lower risk of mortality, even after adjusting for WC or ABSI, as well as in the analyses using the cut-off points proposed by the WHO. The sensitivity analyses for BMI had results comparable to those obtained for the total population, providing additional evidence that residual confusions due to smoking or weight loss from pre-existing serious illness could not fully explain our results. These findings are consistent with the results of previous systematic reviews and meta- analyses, which showed that overweight is a protective factor for mortality among older adults 3,41. The association between BMI and mortality, regardless of ABSI, is in line with the findings of a Dutch study 27, which showed the importance of total adiposity in all-cause mortality, regardless of abdominal adiposity.
This study has some strong points, such as anthropometric indicators obtained from direct measurements of weight, height and WC, as self-reported data can result in incorrect classification of individuals 42. Additionally, we considered anthropometric measurements obtained at four different times, instead of only at baseline, different from most longitudinal (or cohort) studies investigating this topic. Despite these positive points, this study has some limitations. First, information on pre-existing chronic condition for cancer, which is associated with low body weight or low BMI, was unavailable and could have potentially confused the observed associations. Finally, although we adjusted our study for several confounding factors, due to its observational nature, we must consider the possibility of residual confusion.
Conclusions
Our results showed that ABSI, evaluated at four moments of follow-up, was associated with a higher risk of mortality regardless of BMI, but this association receded after sensitivity analyses, suggesting a possible weakness of this index in predicting mortality. Thus, using ABSI in clinical practice or for assessing risks associated with overweight in epidemiological surveys, in conjunction or replacing BMI and WC, still needs more in-depth studies, including those conducted with older adults. However, given the consistency of associations verified for BMI, our results corroborate the fact that this index remains, at least regarding mortality, as the best indicator of this outcome among older adults.
