Study population and data analysis set
We used data of the Survey of Health, Ageing and Retirement in Europe (SHARE), a panel study with eight waves carried out between 2004 and 2020 in 28 European countries and Israel mainly in persons aged 50 years or older [19,20,21,22]. We used data from waves 2, 4, 5, 6 and 7 of SHARE which were collected in 2006/2007 (wave 2), 2011 (wave 4), 2013 (wave 5), 2015 (wave 6) and 2017 (wave 7) – these five waves include residents from 19 European countries and Israel. We did not take wave 3 into account because it has a focus on the life history of the participants and, thus, differs from the other waves. The CST-5 was performed only in waves 2 and 5. As we were interested in the results of the CST-5 as the exposure variable, wave 1 was not taken into account either. Participants were interviewed every 2 years. The mean follow-up time was 5.3 years (standard deviation (SD) 2.9 years). For participants with the CST-5 in wave 2, the mean follow-up was 8.3 years (SD 2.6 years), for those with the CST-5 in wave 5, it was 3.5 years (SD 1.0 years). The interviews covered a wide range of topics, including demographics, physical and mental health, cognitive function, health care, lifestyle, social support, housing, employment, pensions, household income, and financial transfers. The study rationale and design have been described elsewhere, and further information on SHARE is available online [19,20,21,22]. SHARE data are available free of charge after registration.
Seventeen thousand forty-nine participants of wave 2, and 29,070 of wave 5 were aged ≥50 years, neither had known diabetes nor took diabetes drugs at the time of the test, had all undergone the CST-5, were followed up in at least one more wave after performing the CST-5 and did not have missing values for education, BMI and physical activity (cf. flow chart in Fig. 1).
Variables
CST-5
In the CST-5, the time to fully rise and sit down again five times in a row is recorded [23]. The equipment necessary to perform the test is very simple (a stop watch and a chair). The CST shows good test-retest reliability, and good construct validity [24]. In the CST-5, participants were asked to fold their arms across their chest and to sit so that their feet were on the floor, and then to stand up keeping their arms folded across their chest. This was demonstrated to the participants by the interviewer, and the participants were asked whether they felt safe to perform the test. If so, time for five sit and stands in a row was measured in seconds. From the results of the CST-5, a five-level categorical variable was built (test not safe; test safe and time in seconds in the lowest, second lowest, second highest, or highest sex-specific quartile). Longer times required to complete the CST-5 indicate a worse performance.
Handgrip strength
A handheld dynamometer (Smedley, S Dynamometer, TTM, Tokyo, 100 kg) was used to measure handgrip strength [25]. The test was performed in either a standing (preferred) or sitting position, with the elbow at a 90 degree angle, the upper arm tight against the trunk, and the wrist in a neutral position. The interviewers told the participants to squeeze the handle as hard as they could. Before the measurements, the participants were asked whether they were willing to have their handgrip measured. If so, they had a practice with one hand. Two alternate measurements were taken from the right and from the left hand. The maximum of the four measurements was used for further analysis. As result of the HGST, a five-level categorical variable was built in an analogous way as for CST-5.
Diabetes and comorbidities
To assess chronic diseases, participants were shown a card with 16 diseases (heart attack including myocardial infarction or coronary thrombosis or any other heart problem including congestive heart failure; high blood pressure or hypertension; high blood cholesterol; stroke or cerebral vascular disease; diabetes or high blood sugar; chronic lung disease such as chronic bronchitis or emphysema; asthma; arthritis; osteoporosis; cancer or malignant tumor; stomach or duodenal ulcer, peptic ulcer; Parkinson disease; cataracts; hip fracture or femoral fracture; other fractures; Alzheimer’s disease, dementia; benign tumor) and asked: “Has a doctor ever told you that you had/Do you currently have any of the conditions on this card? With this we mean that a doctor has told you that you have this condition, and that you are either currently being treated for or bothered by this condition. Please tell me the number or numbers of the conditions.” New-onset diabetes was assessed by two questions: one on whether a doctor had ever told the participants they had diabetes or high blood sugar, and one on the intake of diabetes drugs. We assume that incident diabetes at the age of 50 years or older is very likely to be type 2 diabetes.
Other covariables
The following socio-demographic variables were taken into account: age, sex, educational attainment and country of residence. Age was used as a continuous variable. International Standard Classification of Education codes (ISCED-97) were used which provide the following classification of educational level: pre-primary, primary, lower secondary, upper secondary, post-secondary non-tertiary, first stage of tertiary, second stage of tertiary, other [26]. We classified Austria, Germany, the Netherlands, France, Switzerland, Belgium, and Luxembourg as “Western Europe”; Sweden and Denmark as “Northern Europe”; Spain, Italy, Greece, Portugal, Croatia, and Israel as “Southern Europe”; Czech Republik, Poland, Hungary, Slovenia, and Estonia as “Eastern Europe”.
Body mass index (BMI) was calculated from self-reports of weight and height, and categorized according to WHO recommendations [27]. Furthermore, participants were asked how often they engaged in vigorous physical activity (more than once a week; once a week; one to three times a month; hardly ever or never). All variables were assessed at wave 2 or 5, depending on when the CST-5 was done for the first time.
Statistical analyses
Relative risks with 95% CIs for the association between results of the CST-5 and incident diabetes were estimated from log-linear models with a Poisson working likelihood and robust standard errors. Three models were fitted: model 1: crude, model 2: adjusted for age (as a continuous variable) and sex, model 3: adjusted for age, sex, BMI (< 18.5, 18.5-24.9, 25-29.9, ≥ 30.0 kg/m2), vigorous physical activity, number of chronic diseases, education (ISCED1997), country group (as a categorical variable). An earlier study showed that age, sex, moderate or vigorous exercise, number of morbidity conditions, income, but not smoking, were associated with muscle strength [28]. Therefore, we did not include smoking in the adjustment set.
These regression models were fitted again to examine whether persons with obesity and dynapenia (age related loss of muscle strength) have a larger risk of type 2 diabetes than persons with dynapenia or obesity alone. Two analyses on this question were done:
In a first analysis, an exposure variable with eight categories was built from obesity (> 30 kg/m2, ≤ 30 kg/m2) and time for CST-5 (considered as not safe; little (Q4), medium (Q2, Q3) or much (Q1) time needed). In a second analysis, the definition of the European Working Group on Sarcopenia in Older People (EWGSOP2) of low muscle strength was used (time for the CST-5 > 15 seconds) [29]. The exposure variable for this analysis has four categories: BMI > 30 kg/m2 and time > 15 s; BMI > 30 kg/m2 and time ≤ 15 s; BMI ≤ 30 kg/m2 and time > 15 s; BMI ≤ 30 kg/m2 and time ≤ 15 s.
To assess the ability of the CST-5 to predict diabetes, we used a logistic regression model with strong risk factors of diabetes which can be measured non-invasively (age, sex, BMI, vigorous physical activity, education, number of chronic diseases). We investigated how strongly the area under the receiver operating characteric curve (AROC) improved after adding the results of the CST-5.
To compare the five-level categorical variables for the CST-5 and the HGST, Cramer’s V was calculated. Pearson correlation coefficients were calculated between handgrip strength and time for five stands and sits in the CST-5.
For participants without incident diabetes, follow-up time was calculated as the time between the first wave (either wave 2 or wave 5) and the last wave in which the participant took part. For participants with incident diabetes, follow-up time was calculated as the time between the first wave and the onset time of diabetes. The midpoint between the wave where a participant reported diabetes diagnosis (or intake of diabetes drugs) for the first time and the previous wave in which the participant took part was used as time of diabetes onset.
All statistical analyses were performed using SAS Version 9.4 (SAS Institute, Cary, USA).