The data used in this study came from the English Longitudinal Study of Ageing (ELSA), which is an ongoing panel study involving community-dwelling individuals in England aged 50 years or older. ELSA began in 2002 and the sample was composed of participants of the Health Survey for England, which was a study involving a nationally representative sample using random probability stratified in different stages 20. After baseline, follow-up interviews in ELSA occurred every two years and health examinations were performed by a nurse every four years. The first health examination occurred in 2004-2005. A detailed description of the study can be found in a previous publication 21.
The sample of the present study comprised 6,183 individuals aged 60 years or older in 2004, when anthropometric and physical performance data were collected for the first time. Among these individuals, 2,302 were excluded due to missing data on the Short Physical Performance Battery (SPPB), grip strength, waist circumference or other covariates, resulting in a final sample of 3,881 individuals at baseline. These measures were not obtained for individuals who were incapable of 1) performing the walk tests without the use of a gait-assistance device; 2) standing up from a chair a single time without using the arms; 3) perform the standing balance tests; 4) performing the grip strength test; or 5) remaining in the standing position for the measurement of waist circumference. The participants were reevaluated after four (2008) and eight (2012) years.
The SPPB is used to determine the physical performance of older adults through the combined assessment of static balance (feet side by side, semi-tandem and tandem), the 2.4-meter walk test and repeated chair stands 22. The complete battery ranges from 0 to 12 points, with higher scores denoting a better physical performance 5. In the present study, the SPPB score was treated as a discrete quantitative variable 23.
Anthropometric measures and classification of groups
Grip strength was measured using a manual dynamometer (Smedley, range: 0 to 100 kg). During the test, the participant remained standing with the arm alongside the trunk and the elbow flexed at 90 degrees 24. Three maximum strength tests were performed with a one-minute rest period between readings and the highest value was considered for the analysis. Dynapenia was defined as grip strength <26 kg for men and <16 kg for women 25.
Waist circumference was measured using a metric tape at the midpoint between the lowest rib and the upper edge of the iliac crest. The measurement was made twice at the end of the expiratory phase of the respiratory cycle 24. A third measurement was performed if the difference between the first two measurements was greater than 3 cm. Abdominal obesity was defined as waist circumference >102 cm for men and >88 cm for women 26.
Four groups were created considering the absence/presence of abdominal obesity and dynapenia: non-dynapenic/non-abdominal obese (ND/NAO); non-dynapenic/abdominal obese (ND/AO); dynapenic/non-abdominal obese (D/NAO); and dynapenic/abdominal obese (D/AO) 27.
The socioeconomic variables were sex, age group (60 to 69; 70 to 79; 80 years or more), marital status (with or without a conjugal life), level of education (0 to 11; 12 to 13; >13 years) and total household wealth (quintiles).
Smoking was determined by asking the participants whether they were non-smokers, ex-smokers or current smokers. Regarding alcohol intake, the participants were classified as non-drinkers or rare drinkers (up to once per week), frequent drinkers (two to six times per week) or daily drinkers 28. Level of physical activity was determined using an instrument validated by the Health Survey for England 29 that considers the frequency of participation in vigorous, moderate and mild physical activities (more than once per week, once per week, one to three times per months or almost never). Lifestyle was classified as sedentary (no weekly physical activity) or active (mild, moderate or vigorous physical activity at least once per week) 27.
Health status was ascertained by self-reported doctor diagnosis of diabetes, systemic arterial hypertension, stroke, heart disease, lung disease, cancer, joint disease, osteoporosis and falls in the previous 12 months. Cognitive function was evaluated based on the global score of immediate and delayed recall (range: 0 to 20 words) 30. Depressive symptoms were determined using the Center for Epidemiologic Studies Depression Scale (CES-D), considering a cutoff of ≥4 points 31.
The percentage change in weight between waves was analyzed to adjust the models, as weight loss can affect the relation between abdominal obesity and the decline in muscle strength. This variable was categorized as follows: stable weight, weight loss equal to or greater than 5% and weight gain equal to or greater than 5% compared to the previous wave32.
All covariates included in the analyses represent a wide range of risk factors associated with the progression of the decline in physical performance 33. With the exception of age, all variables were treated as time-varying covariates (whenever a variable changes over time for the subjects)34.
Differences in baseline characteristics between (a) included individuals and those excluded due to missing data on the SPPB, grip strength, waist circumference or other covariates and (b) the four groups classified based on the presence/absence abdominal obesity and dynapenia were evaluated using the chi-squared test, analysis of variance (ANOVA) and Tukey’s post hoc test. For all analyses, a p-value <0.05 was considered indicative of statistical significance.
Generalized linear mixed models stratified by sex were used to estimate the trajectories of physical performance using the XTMIXED procedure in Stata 14 SE (Stata Corp, College Station, TX, USA). These models were chosen because they are more appropriate for unbalanced data from studies with repeated measures and enable the statistical modeling of time-dependent changes in the outcome variable (SPPB) and in the magnitude of associations between variables 35,36.
The two models (one for each sex) include the interaction between time (in years of follow-up) and the status of abdominal obesity and dynapenia adjusted by the covariables. Univariate analyses were performed to select covariables to incorporate into the final model per sex. Only covariables that presented associations with a p-value ≤0.20 in the univariate analyses were selected for inclusion in the multiple models using the stepwise forward method.
In the final models, the intercept represents the mean estimated difference in performance on the SPPB at baseline between individuals categorized based on the presence/absence of abdominal obesity and dynapenia considering the ND/NAO group as the reference category. On the slope, time (in years) indicates whether the performance on the SPPB declines independently of the presence of covariables (that is, whether time per se is the determinant of the decline). The interaction between time and each status of abdominal obesity and dynapenia represents the estimated difference in the annual rate of decline in performance on the SPPB (slope) between each of the three groups (ND/AO, D/NAO and D/AO) and the reference group (ND/NAO), enabling the assessment of the annual rate of change in the SPPB in the four groups. The rates of the decline in performance were compared using the ß coefficient and 95% confidence interval (CI).
Sensitivity analysis was performed to investigate whether abdominal obesity (yes/no) and dynapenia (yes/no), when analyzed separately, were able to modify the association found between dynapenic abdominal obesity and the decline in physical performance.
In addition, statistics to estimate average population parameters, such as the marginal average, were used from predictions of a previously adjusted model.