The proportion of overweight and obese individuals in the older population is growing worldwide( Reference Ng, Fleming and Robinson 1 ). Due to demographic changes resulting in a continuous increase in the number of older adults, this topic concerns a permanently growing part of the population( 2 ). Obesity in older adults is associated with various diseases, such as type 2 diabetes mellitus and CVD, as well as with restriction of physical function and health-related quality of life (HRQoL), and thus is highly problematic for the public health sector( Reference Mathus-Vliegen 3 , Reference Decaria, Sharp and Petrella 4 ).
Obesity is usually defined by a BMI≥30·0 kg/m2 in both the younger and the older population. According to this definition, in Germany, 33·1 % of men and 34·8 % of women aged 60–69 years, and 31·3 % of men and 41·6 % of women aged 70–79 years, are obese( Reference Mensink, Schienkiewitz and Haftenberger 5 ).
For younger adults, BMI is a useful measure of total body fat. However, the ageing process leads to changes in body composition by loss of height and skeletal muscle mass along with a redistribution of body fat towards more visceral fat. Hence, the associations between BMI≥30·0 kg/m2 and the risk of various health consequences of obesity are attenuated and lose explanatory power( Reference Mathus-Vliegen 3 , Reference Decaria, Sharp and Petrella 4 , Reference Zamboni, Mazzali and Zoico 6 – Reference Villareal, Apovian and Kushner 9 ). Thus, other measures of obesity may be more appropriate for older adults. Despite existing research comparing different measures of obesity with respect to different outcomes in older individuals, it is still unclear which measure and which cut-off point best describe the influence of obesity on health in older adults( Reference Mathus-Vliegen 3 , Reference Decaria, Sharp and Petrella 4 ).
There are methods to accurately measure body fat, for example MRI, but these are generally expensive and complicated and thus impractical in the general medical practice. Multiple other measures have been proposed to operationalize obesity( Reference Mathus-Vliegen 3 , Reference Decaria, Sharp and Petrella 4 , Reference Wagner and Heyward 10 ). In addition to BMI, the usual measure of obesity, there are other anthropometric methods like waist circumference (WC), waist-to-hip ratio (WHR) and waist-to-height ratio (WHtR), which can be assessed simply and inexpensively. Moreover, there are methods to easily measure body fat, such as bioelectrical impedance analysis (BIA). Furthermore, combined measures such as hypertriglyceridaemic waist (HTGW) and sarcopenic obesity (SO) can be determined, which are extensions to the current definitions of obesity by considering additional aspects. HTGW represents the combination of an increased WC and elevated serum TAG levels. Lemieux et al.( Reference Lemieux, Pascot and Couillard 11 ) established this definition in 2000 as a substitute for the metabolic syndrome and Sam et al.( Reference Sam, Haffner and Davidson 12 ) showed that HTGW is a better measure of visceral fat than WC alone. Thus, the assessment of HTGW allows better differentiation between metabolically healthy and ill obese individuals than WC. For SO, there is no standard definition so far, but it is mostly described by the combination of increased body fat and decreased skeletal muscle mass and/or strength, and thus better reflects the changes in body composition in older adults( Reference Kohara 13 ).
Previous studies comparing different measures of obesity in older adults focused mainly on outcomes like mortality, individual diseases and biomarkers( Reference Guasch-Ferré, Bulló and Martínez-González 14 – Reference Sun, van Dam and Spiegelman 21 ). Restrictions of HRQoL and physical constitution, which are important for healthy and successful ageing, have rarely been investigated in this context. To date, mainly one or two measures of obesity have been examined simultaneously in their cross-sectional and/or longitudinal association with HRQoL in older adults, and not much is known about the comparison of several measures in older age groups( Reference López-García, Banegas Banegas and Gutiérrez-Fisac 22 – Reference Kostka and Bogus 25 ). Most studies examined the measures continuously or categorized according to different percentiles to find the best measure per se. In practice, however, the determination of obesity is based on established cut-off points. Thus, the aim of the present study was to compare (cross-sectionally and longitudinally) a variety of measures of obesity in relation to HRQoL and self-rated physical constitution (SRPC) in older adults, using established obesity cut-off points.
Participants and methods
Study population
The population-based Cooperative Health Research in the Region of Augsburg (KORA)-Age cohort study is a follow-up of all participants born before 1944 (i.e. ≥65 years at the baseline examination in 2009) who took part in one of the four MONICA/KORA surveys carried out in Southern Germany( Reference Holle, Happich and Löwel 26 ). The baseline examination included 5991 participants, of whom 4565 returned a postal questionnaire and 4127 took part in a telephone interview. Additionally, a sub-sample of 1079 participants was intensively examined at the study centre. Eight hundred and twenty-two participants were re-examined in 2012. All interviews and examinations were conducted by trained staff( Reference Peters, Döring and Ladwig 27 ). A detailed description of the study population, the assessment and classification of variables, and the statistical analysis can be found in the online supplementary material (section ‘Additional Information on Subjects and Methods’).
Measures of obesity
We compared seven measures of obesity (BMI, WC, WHR, WHtR, fat mass percentage (FMp), HTGW, SO), which were collected at baseline.
Weight was measured with an electronic scale, standing height with a stadiometer. WC was quantified with an inelastic measuring tape at the smallest abdominal girth or, in obese participants, in the middle between the lowest rib and the iliac crest. Hip circumference was measured at the most protruding part of the hips. Fat-free mass was computed by Kyle’s equation( Reference Kyle, Genton and Karsegard 28 , Reference Genton, Karsegard and Kyle 29 ) using body composition parameters assessed by BIA (BIA 2000-S; DATA-INPUT GmbH, Frankfurt, Germany). Fat mass (weight – fat-free mass) and FMp (fat mass/weight) were calculated. TAG levels were determined in non-fasting blood samples (TGL Flex reagent cartridge; Dade Behring, Eschborn, Germany). Mean grip strength from three consecutive measurements was assessed with the JAMAR Dynamometer (Saehan Corp., Masan, Korea).
The measures of obesity were dichotomized at their established, sex-specific if available, obesity cut-off points for the general adult population. Participants were classified as obese at a BMI (weight/height2) of ≥30·0 kg/m2, a WC of ≥102/88 cm or a WHR (WC/hip circumference) of ≥1·00/0·85 for men and women, respectively( 30 ), or a WHtR( Reference Ashwell 31 , Reference Schneider, Friedrich and Klotsche 32 ) (WC/height) of ≥0·6. For consistency and comparability with WC, HTGW was defined as a WC≥102/88 cm for men and women, respectively, and TAG levels of ≥1·7 mmol/l( Reference Saudny, Cao and Egeland 33 , 34 ). The following two groups were established: one group with individuals fulfilling both criteria and another group with individuals fulfilling only one or none of the two criteria. Likewise, SO was defined as a combination of a FMp higher than the sex-specific 60th percentile of the study population( Reference Baumgartner, Wayne and Waters 35 ) (≥30·41/41·10 % for men and women, respectively) and decreased handgrip strength (<30/20 kg for men and women, respectively)( Reference Lauretani, Russo and Bandinelli 36 ). As muscle quality is more important than muscle mass per se, handgrip strength was preferred to muscle mass( Reference Stenholm, Harris and Rantanen 37 ). However, muscle mass was included in the definition of SO in a sensitivity analysis (see online supplementary material, section ‘Sensitivity Analysis for the Definition of SO’). For reasons of consistency and comparability with SO, FMp was also dichotomized at the sex-specific 60th percentile of the study population (Table 2).
Outcomes
HRQoL and SRPC were inquired in the postal questionnaires at baseline and at the 3-year follow-up (2·9 (sd 0·1) years). HRQoL, the primary outcome, was assessed with the Euroqol (EQ)-5D. This generic measure includes five questions concerning mobility, self-care, usual activities, pain/physical discomfort and anxiety/depression( 38 ). The continuous EQ-5D index (range: −0·205 to 0·999; 0·999=no restriction) was calculated using the scoring algorithm for the German population derived by Greiner et al.( Reference Greiner, Claes and Busschbach 39 ).
SRPC, the secondary outcome, was assessed with the question ‘How would you rate your present physical constitution?’, with four response options (‘excellent’, ‘good’, ‘fair’ and ‘poor’). The answers were dichotomized as ‘excellent/good’ and ‘fair/poor’ to obtain adequate group sizes.
Assessment of covariables
Covariables, all collected at baseline, were selected based on theoretical considerations and existing literature investigating this topic( Reference López-García, Banegas Banegas and Gutiérrez-Fisac 22 , Reference Tan, Wee and Lee 23 , Reference Kostka and Bogus 25 ). Age and physical activity were considered continuously, all other variables were grouped into categories.
Sociodemographic variables included age (years), sex (reference=male), marital status (unmarried (=reference), married, divorced, widowed) and years of education (<10 (=reference), Reference Wagner and Heyward10–<Reference Sam, Haffner and Davidson12, ≥Reference Sam, Haffner and Davidson12). Lifestyle variables included physical activity (assessed by the Physical Activity Scale for the Elderly (PASE); score range: 0–365)( Reference Washburn, Smith and Jette 40 ), smoking status (never (=reference), former, current) and alcohol consumption (no (0 g/d; =reference), moderate (>0–<40/20 g/d for men and women, respectively) and high (≥40/20 g/d for men and women, respectively))( Reference Keil, Chambless and Döring 41 ). Additionally, the presence (yes, no (=reference)) of the following diseases was assessed: hip/femoral neck fracture in the last 5 years( Reference Wildner, Döring and Meisinger 42 ), hypertension, diabetes mellitus, lung disease, joint disease, gastrointestinal disease, heart problems, heart attack in the last 3 years, kidney disease, liver disease, cancer occurring in the last 3 years, neurological disease, stroke in the last 3 years and eye disease.
Statistical analysis
As all measures of obesity showed an almost linear relationship with continuous baseline HRQoL and its change over 3 years of follow-up (HRQoL at follow-up – HRQoL at baseline), linear regression was used for HRQoL. Associations with dichotomized baseline SRPC (reference=excellent/good) and its deterioration/improvement over time (reference=no change) were examined with binary logistic regression. Prior to analyses, participants who reported a weight change of >5 kg in the last 6 months (n 65) and those with missing values in at least one of the seven different measures of obesity (additional: n 76), both outcomes (additional: n 21 (cross-sectional)/n 233 (longitudinal)) or the covariables (additional: n 34 (cross-sectional)/n 16 (longitudinal)) were excluded. Thus, the sample size varied between the five analyses (n 883 (baseline HRQoL/baseline SRPC), n 689 (change in HRQoL), n 622/605 (deterioration of SRPC/improvement of SRPC); see online supplementary material, Supplemental Fig. 1). Within each analysis and for each of the seven measures of obesity, three models with different sets of covariables were conducted, respectively: model 1 was adjusted for age and sex; model 2 was additionally adjusted for further sociodemographic and lifestyle variables; model 3 was additionally adjusted for the presence of diseases. In each case, we compared the respective seven models, which differed only in the measure of obesity used, but included the same sample size and covariables, to assess the measure of obesity showing the strongest association with each of the five outcomes. Only measures which were significantly associated with the respective outcome were further examined (linear regression: β estimate (F test); logistic regression: OR and 95 % CI (Wald test)). R 2, Akaike information criterion (AIC) and Schwarz Bayesian information criterion (BIC) were used to find the model with the best goodness-of-fit and thus the best measure of obesity for the prediction of the outcomes. The larger the R 2 and the smaller the AIC and BIC, the better the model adaptation. Additionally, the (changes of the) areas under the receiver-operating characteristic curves ((Δ)AUC) were compared in logistic regression models (see online supplementary material, Supplemental Fig. 2). The AUC ranges between 0·5 (poor discrimination) and 1·0 (optimal discrimination). ΔAUC was applied to verify if the addition of each measure of obesity improved the model. We calculated the ∆AUC by the difference between the AUC of the model with the respective obesity measure and the model without the respective obesity measure containing only the covariables. The ∆AUC were compared between the models with different obesity measures and the higher the AUC/∆AUC, the better the model.
All analyses were repeated using continuous versions of the investigated measures of obesity to demonstrate which measure per se shows the strongest association with the outcomes, independent of the recommended cut-off points (see online supplementary material, section ‘Sensitivity Analysis with Continuous Versions of the Investigated Measures of Obesity’). To provide comparable β estimates, all measures of obesity were Z-transformed (Z=(X–mean)/sd) prior to analysis. P values of <0·05 were considered statistically significant in all analyses, which were performed using the statistical software package SAS version 9.3.
Results
Table 1 displays the characteristics of the study population of the cross-sectional analyses in total and stratified by sex. Participants were 65–93 years old and the proportion of men and women was approximately equal. Men were significantly more physically active (P=0·02), better educated, and consumed more cigarettes and alcohol (all P<0·0001). Marital status (P<0·0001) as well as the prevalence of gastrointestinal diseases (P=0·05), eye diseases (P=0·0001) and cancer (P=0·01) also differed significantly by sex. HRQoL at baseline was high (median=0·887) and 28·2 % of the participants classified their baseline SRPC as fair/poor. Over 3 years of follow-up, the median change of 0 showed no change in HRQoL, but the 25th percentile of –0·112 indicated a slight decrease. Of the study population, 12·2 % reported a deterioration and 9·7 % an improvement of SRPC.
Table 1 Characteristics of the study population used in the cross-sectional analyses
P25, 25th percentile; P75, 75th percentile; PASE, Physical Activity Scale for the Elderly (score range: 0–365); M, male; F, female; HRQoL, health-related quality of life (range: −0·205–0·999); SRPC, self-rated physical constitution.
Data of the KORA-Age study conducted in Southern Germany between 2009 and 2012. Outcomes at follow-up and change of outcomes between baseline and follow-up: n 689 (men, n 350; women, n 339).
Significant results (P<0·05) are highlighted.
* Mann–Whitney–Wilcoxon U test for continuous variables; Pearson’s χ 2 test for categorical variables (Fisher’s exact test if expected frequencies were too low).
† Data are presented as median (P25, P75).
‡ Data are presented as n and column %.
Table 2 displays the prevalence of obesity according to the different obesity measures in total and stratified by HRQoL (no, any restriction) and SRPC (excellent/good, fair/poor). In total, the prevalence of obesity varied considerably between 15·2 % as defined by SO and 60·5 % as defined by WC. Regardless of the measure of obesity, the proportion of obese participants was significantly higher in the group with restrictions or fair/poor health as compared with the group without restrictions or excellent/good health (P=0·01 to P<0·0001).
Table 2 Prevalence of obesity using different obesity measures in the study population used in the cross-sectional analyses
HRQoL, health-related quality of life; SRPC, self-rated physical constitution; WC, waist circumference; WHR, waist-to-hip ratio; WHtR, waist-to-height ratio; FMp, fat mass percentage; HTGW, hypertriglyceridaemic waist; SO, sarcopenic obesity; M, male; F, female.
Data of the KORA-Age study conducted in Southern Germany between 2009 and 2012.
Significant results (P<0·05) are highlighted.
* Restriction in HRQoL (EQ-5D index <0·999) v. no restriction (EQ-5D index=0·999).
† Pearson’s χ 2 test for categorical variables (Fisher’s exact test if expected frequencies were too low).
The results of the linear regression and the model quality criteria of the cross-sectional analysis of HRQoL are shown in Table 3. Obesity, regardless of the measure, was significantly associated with worse HRQoL (P=0·01 to P<0·0001) after adjustment for different sets of covariables. For the fully adjusted model 3, the lowest AIC and BIC were found for the WC model. Regarding R 2, this model explained 17·04 % of the variance of HRQoL and thus more than the BMI model (R 2=16·29 %) with the second best model adaption. The results of the binary logistic regression and the model quality criteria of the cross-sectional analysis of SRPC are presented in Table 4. Again, only the results from the fully adjusted model 3 are reported here. Obesity, defined by all measures except SO, was significantly associated with higher odds for fair/poor SRPC (P=0·01 to P<0·0001), with obesity defined by BMI and HTGW nearly doubling the odds (OR=1·98). The BMI and HTGW models showed the highest R 2 and (Δ)AUC as well as the lowest AIC and BIC. The WC model had the third best model adaption. The results of the sensitivity analysis using continuous measures of obesity are presented in Tables 5 and 6. Except SO, all measures of obesity were significantly associated with both outcomes in the fully adjusted model 3 (P=0·01 to P<0·0001). Regarding HRQoL, the BMI model showed the best model quality criteria R 2, AIC and BIC, followed by the models with WHtR and WC. For every sd increase of Z-standardized BMI, HRQoL decreased by β=–0·033. Regarding SRPC, the WHtR model had the best goodness-of-fit assessed by (Δ)AUC, R 2 , AIC and BIC, followed by the models with WC and BMI. In the longitudinal analyses (see online supplementary material, Supplemental Tables 1–6), neither the dichotomized nor the continuous measures of obesity were significantly associated with change in HRQoL and SRPC over time. Cross-sectionally and longitudinally, SO as defined by extended definitions was not associated with either outcome (Supplemental Tables 7 and 8).
Table 3 Comparison of cross-sectional associations between the categorized measures of obesity and HRQoL
HRQoL, health-related quality of life; AIC, Akaike information criterion; BIC, Schwarz Bayesian information criterion; WC, waist circumference; WHR, waist-to-hip ratio; WHtR, waist-to-height ratio; FMp, fat mass percentage; HTGW, hypertriglyceridaemic waist; SO, sarcopenic obesity; PASE, Physical Activity Scale for the Elderly; M, male; F, female.
Data of the KORA-Age study conducted in Southern Germany between 2009 and 2012.
Linear regression models. Model 1 adjusted for age and sex; model 2 additionally adjusted for further sociodemographic (marital status, years of education) and lifestyle (PASE score, smoking status, alcohol consumption) variables; model 3 additionally adjusted for the presence of diseases (hip/femoral neck fracture in the last 5 years, hypertension, diabetes mellitus, lung disease, joint disease, gastrointestinal disease, heart disease, heart attack in the last 3 years, kidney disease, liver disease, cancer occurring in the last 3 years, neurological disease, stroke in the last 3 years, eye disease).
Results are shown for the following obesity categories: BMI≥30·0 kg/m2; WC≥102/88 cm (M/F); WHR≥1·00/0·85 (M/F); WHtR≥0·6; FMp≥30·41/41·10 % (M/F); HTGW: WC≥102/88 cm (M/F) and TAG≥1·7 mmol/l; SO: FMp≥30·41/41·10 % (M/F) and handgrip strength <30/20 kg (M/F).
Significant results (P<0·05) are highlighted.
Table 4 Comparison of cross-sectional associations between the categorized measures of obesity and SRPC
SRPC, self-rated physical constitution; AUC, area under the receiver-operating characteristic curve; Δ, change; AIC, Akaike information criterion; BIC, Schwarz Bayesian information criterion; WC, waist circumference; WHR, waist-to-hip ratio; WHtR, waist-to-height ratio; FMp, fat mass percentage; HTGW, hypertriglyceridaemic waist; SO, sarcopenic obesity; PASE, Physical Activity Scale for the Elderly; M, male; F, female.
Data of the KORA-Age study conducted in Southern Germany between 2009 and 2012.
Logistic regression models: reference category=excellent/good SRPC. Model 1 adjusted for age and sex; model 2 additionally adjusted for further sociodemographic (marital status, years of education) and lifestyle (PASE score, smoking status, alcohol consumption) variables; model 3 additionally adjusted for the presence of diseases (hip/femoral neck fracture in the last 5 years, hypertension, diabetes mellitus, lung disease, joint disease, gastrointestinal disease, heart disease, heart attack in the last 3 years, kidney disease, liver disease, cancer occurring in the last 3 years, neurological disease, stroke in the last 3 years, eye disease).
Results are shown for the following obesity categories: BMI≥30·0 kg/m2; WC≥102/88 cm (M/F); WHR≥1·00/0·85 (M/F); WHtR≥0·6; FMp≥30·41/41·10 % (M/F); HTGW: WC≥102/88 cm (M/F) and TAG≥1·7 mmol/l; SO: FMp≥30·41/41·10 % (M/F) and handgrip strength <30/20 kg (M/F).
Significant results (P<0·05) are highlighted.
Table 5 Comparison of cross-sectional associations between the continuous measures of obesity and HRQoL
HRQoL, health-related quality of life; AIC, Akaike information criterion; BIC, Schwarz Bayesian information criterion; WC, waist circumference; WHR, waist-to-hip ratio; WHtR, waist-to-height ratio; FMp, fat mass percentage; HTGW, hypertriglyceridaemic waist; SO, sarcopenic obesity; PASE, Physical Activity Scale for the Elderly; M, male; F, female.
Data of the KORA-Age study conducted in Southern Germany between 2009 and 2012.
Linear regression models. Model 1 adjusted for age and sex; model 2 additionally adjusted for further sociodemographic (marital status, years of education) and lifestyle (PASE score, smoking status, alcohol consumption) variables; model 3 additionally adjusted for the presence of diseases (hip/femoral neck fracture in the last 5 years, hypertension, diabetes mellitus, lung disease, joint disease, gastrointestinal disease, heart disease, heart attack in the last 3 years, kidney disease, liver disease, cancer occurring in the last 3 years, neurological disease, stroke in the last 3 years, eye disease).
Results are shown for 1 sd increase in measures of obesity, as they were Z-standardized for direct comparability.
Significant results (P<0·05) are highlighted.
Table 6 Comparison of cross-sectional associations between the continuous measures of obesity and SRPC
SRPC, self-rated physical constitution; AUC, area under the receiver-operating characteristic curve; Δ, change; AIC, Akaike information criterion; BIC, Schwarz Bayesian information criterion; WC, waist circumference; WHR, waist-to-hip ratio; WHtR, waist-to-height ratio; FMp, fat mass percentage; HTGW, hypertriglyceridaemic waist; SO, sarcopenic obesity; PASE, Physical Activity Scale for the Elderly; M, male; F, female.
Data of the KORA-Age study conducted in Southern Germany between 2009 and 2012.
Logistic regression models: reference category=excellent/good SRPC. Model 1 adjusted for age and sex; model 2 additionally adjusted for further sociodemographic (marital status, years of education) and lifestyle (PASE score, smoking status, alcohol consumption) variables; model 3 additionally adjusted for the presence of diseases (hip/femoral neck fracture in the last 5 years, hypertension, diabetes mellitus, lung disease, joint disease, gastrointestinal disease, heart disease, heart attack in the last 3 years, kidney disease, liver disease, cancer occurring in the last 3 years, neurological disease, stroke in the last 3 years, eye disease).
Results are shown for 1 sd increase in measures of obesity, as they were Z-standardized for direct comparability.
Significant results (P<0·05) are highlighted.
Discussion
In this population of older adults, nearly all measures of obesity were significantly inversely associated with baseline HRQoL/SRPC. For the categorized measures, the strongest inverse association with HRQoL as well as the best model adaption was found for WC. Thus, a WC of ≥102 cm for men and of ≥88 cm for women had a slightly better predictive power for HRQoL than a BMI of ≥30·0 kg/m2. However, BMI and HTGW had the best predictive power for SRPC, with WC being almost equally as good. In the longitudinal analysis, none of the obesity measures was significantly associated with change in HRQoL or SRPC over the follow-up period.
These results suggest that with regard to quality of life, simple anthropometric measures are preferable to more complex measures or BIA to determine obesity in older adults. As, in the sensitivity analysis, the continuous BMI remained the best measure per se in connection with HRQoL, the cut-off point of ≥30·0 kg/m2 may not be ideal for older adults. Similarly, WHtR was the best continuous measure concerning SRPC, but performed poorly in the comparison of dichotomized measures. This indicates that the cut-off point of ≥0·6 is not optimal for the older population. Since WC, dichotomized or continuous, was either the best or one of the best measures and easier to assess than HTGW, measuring WC may be a valuable addition to BMI with regard to quality of life.
Overall, the present study confirms the inverse relationship between obesity and HRQoL described in the literature( Reference López-García, Banegas Banegas and Gutiérrez-Fisac 22 – Reference Kostka and Bogus 25 , Reference Hunger, Schunk and Meisinger 43 , Reference Hunger, Thorand and Schunk 44 ). Our results are in line with the only study comparing several measures of obesity with regard to HRQoL in older age groups. Tan et al.( Reference Tan, Wee and Lee 23 ) used anthropometric measures (BMI, WC, waist residuals (regression of WC v. BMI), WHR, WHtR, height) in quintiles. HRQoL was assessed with the Short-Form 36 questionnaire version 2 and was separately analysed for a physical and a mental component. By contrast, the EQ-5D, used in the present study, considers the two parts together, but is dominated by the physical component. Tan et al. examined cross-sectional associations in a slightly younger (mean age: 50·6 (sd 12·2)/49·3 (sd 11·6) years for men/women, respectively) multi-ethnic Asian population (n 4981), stratified by sex. They studied the same model quality criteria R 2, AIC and BIC, which, as in our study, were similar for all measures of obesity. They concluded that BMI, WHtR and WC were the best measures for women in the physical component. For men, there were no significant results. We did not conduct sex-stratified analyses, as no sex-specific differences were found (see online supplementary material, section ‘Analyses to Test for Sex-specific Differences’ and Supplemental Tables 9 and 10). Despite this consistency, the body composition between Asians and Europeans is known to differ greatly( Reference Rush, Freitas and Plank 45 ). Therefore, further studies in Europeans are needed to confirm our results. Our results are also consistent with studies using disability as an outcome, which is important for successful ageing as well. Simple anthropometric measures like BMI and WC showed the strongest associations and complex measures like SO were only rarely associated with disability( Reference Wong, Stevenson and Backholer 46 – Reference Angleman, Harris and Melzer 50 ).
One strength of our study is the investigation of a population-based sample. Our results are thus generalizable to older adults in Germany. Further, in contrast to other studies investigating this topic, a greater variety of measures of obesity (especially by BIA) was available, collected by trained staff using standardized assessment methods. As the outcomes were measured at two points in time, we were able to examine the longitudinal association between obesity at baseline and change in HRQoL/SRPC over the follow-up period. The longitudinal analysis was, however, limited by a short follow-up period, in which HRQoL/SRPC changed only little, as well as by a smaller sample size, compared with other studies in this context. Thus, our longitudinal analysis lacked statistical power, which, together with a possibly weak effect of obesity on future HRQoL/SRPC, might explain the non-significant results. Although the present study is the first comparing various measures of obesity in their longitudinal association with HRQoL/SRPC in older adults, our inability to draw meaningful conclusions on which obesity measure best predicted deterioration of HRQoL or SRPC limits the impact of our study. Further limitations are inaccuracies in the determination of fat and muscle mass by BIA with the equations of Kyle et al.( Reference Kyle, Genton and Karsegard 28 , Reference Genton, Karsegard and Kyle 29 ) and Janssen et al.( Reference Janssen, Heymsfield and Baumgartner 51 ), even though both were validated in older adults. Outcomes and covariables were assessed by self-report and thus misreporting cannot be excluded. But, since all seven models included the same population sample, respectively, these errors had no effect with regard to the comparison of the obesity measures. The large number of covariables in model 3 could have led to over-adjustment. However, all three models showed a similar ranking of measures of obesity.
Conclusion
In conclusion, using established cut-off points of obesity, WC showed the strongest association with HRQoL as well as the best model adaption among the compared measures of obesity. Thus, in older adults, both in health checks and for research concerning HRQoL, WC should be measured in addition to BMI. As WHtR and BMI were appropriate continuous measures, but were less efficient when categorized, future research should focus on special obesity cut-off points for older adults. Our results suggest that the assessment of simple anthropometric measures is sufficient to determine obesity in older adults in medical practice, as more complex measurement techniques such as BIA do not provide additional information with regard to HRQoL.
Acknowledgements
Financial support: The Cooperative Health Research in the Region of Augsburg (KORA) research platform was initiated and financed by the Helmholtz Zentrum München – German Research Center for Environmental Health, which is funded by the German Federal Ministry of Education and Research and by the State of Bavaria. The KORA-Age project was financed by the German Federal Ministry of Education and Research (BMBF) (grant numbers FKZ 01ET0713 and FKZ 01ET1003A) as part of the ‘Health in old age’ programme. S.V. is supported by Kompetenznetz Adipositas (Competence Network Obesity), funded by the German Federal Ministry of Education and Research (grant number FKZ 01GI1121B). The funders had no role in the design, analysis or writing of this article. Conflict of interest: None. Authorship: A.R. conceived and designed the experiments, analysed the data and wrote the paper; S.V. conceived and designed the experiments, supervised the data analysis and revised the paper; R.H. collected the data, provided advice regarding the data analysis and revised the paper; T.d.l.H.G. provided advice regarding the data analysis and revised the paper; M.L. provided advice regarding the data analysis and revised the paper; A.P. collected the data and revised the paper; B.T. conceived and designed the experiments, collected the data, supervised the data analysis and revised the paper. All authors have read and approved the final manuscript. Ethics of human subject participation: The study was approved by the Ethics Committee of the Bavarian Medical Association and all participants provided written informed consent.
Supplementary Material
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S1368980016001270
