Data were obtained from the SHARE, which was biennial longitudinal research aimed at assessing the population aged 50 or older across European countries by using probability-based sampling. Details about sampling methodology could be found in the published official article19. The information of this survey concerned health, socio-economic status, and social and family networks. Six waves (1, 2, 4, 5, 6, 7) and one retrospective life history wave (3) were conducted ever since 2004. Since the second wave, SHARE has successively incorporated more European countries such as Belgium, Czech Republic, Poland, Ireland and so on based on the first wave. Thus, we analyzed data from five-panel waves of the SHARE but excluding the first and third waves in the present study. SHARE was reviewed and approved by the ethics committee of the University of Mannheim and the ethics council of the Max Planck Society. We confirm that all methods were performed in accordance with the relevant guidelines and regulations, and informed consent was obtained from all participants and their legal guardians.
A total of 37152 people participated in the second wave which carried out in 2006, The exclusion criteria are as follows: (1) excluding those who participated less than three waves during the following years (n = 17021); (2) excluding those with cancer, Parkinson’s disease, Alzheimer’s disease and stroke at any survey may affect handgrip strength measurement (n = 1353); (3) excluding participants with missing data of alcohol intake, smoking status, family economic level, physical activity, European depression scale, grip strength and cognitive function (n = 1087); (4) excluding individuals with depression at baseline (wave 2) so that avoiding causing reverse causality (n = 3728). Ultimately, 13936 individuals were left in this study. The selection process of the study population was shown in Fig. 1.
Depression was evaluated using the European depression (EURO-D) scale. The scale is a 12-item binary scale including the following symptoms: depression, pessimism, suicidality, guilty, sleep, irritability, fatigue, appetite, interests, enjoyment, concentration and tearfulness, which has been validated by European Depression Concerted Action Project20. THE total EURO-D scale ranged from 0 to 12 with a score above three representing clinically depression.
Physical activity was assessed by asking two questions about how often to engage in moderate and vigorous-intensity physical activity in daily life, respectively. The following response options were given: more than once a week, once a week, one to three times a month and hardly ever or never. Participants who reported less than ‘one to three times a month’ for either moderate or vigorous-intensity physical activity were defined as physical inactivity21.
Grip strength was measured using a Smedley handheld dynamometer (100 kg)22. Participants were requested to sit or stand while keeping the upper arms tight against the trunk with their elbows at a 90° angle and then squeeze the handles as hard as possible for 5 s. Two alternate measures from their right and left hands were performed, the highest value of four measures in each survey wave was used in the present study. Due to the lack of a standard grouping method for grip strength until now, we used GBTM to explore a suitable group according to genders based on the panel data. The specific grouping method could be found in the supplementary file of S-figure 1 and 2.
Covariates in this study were acquired from the questionnaire, including age, gender (male/female), European region (central, northern, southern and western Europe), married status (married, living with a spouse/other married status such as divorced, widow), education (primary/secondary/tertiary education), employment status (employed/retired/unemployment), family economic level, smoking status (never, ever and current smoker), alcohol intake (whether drinking exceed two glasses), heart attack (yes/no), hypertension (yes/no), hyperlipidemia (yes/no), diabetes (yes/no), mobility limitation (yes/no), body mass index (BMI) and cognitive function. Education was categorized based on the International Standard Classification of Education23. The family economic level was determined by one question: “Is your household able to make ends meet?” The answers included easily, fairly easily, with some difficulty and with great difficulty. BMI was calculated as weight in kilograms divided by height in meters squared. Cognitive function was assessed from four domains: time orientation, memory, verbal fluency and numeracy24,25. The scores of each domain ranged from 0 to 5, 0 to 20, 0 to 100 and 0 to 5, respectively. To avoid the proportion of memory and fluency accounted too high, we convert these two indexes into decimal systems. The sum of the above scores was used to assess cognitive function.
All statistical analyses were conducted using STATA version 16.0 (Stata Corp, College Station, Texas). GBTM was used to identify low, moderate and high grip strength groups of individuals following similar patterns of grip strength according to genders. Data was given in the form of means and standard deviations (SDs) for continuous variables or as percentages for categorical variables. One-way analysis of variance (ANOVA) was used to examine the difference of means for continuous variables with normal distribution; otherwise, the Kruskal-Wallis test was used. Pearson’s χ2 test was performed to compare the distribution of the categorical variables between the three grip strength groups. Multiplicative interaction was assessed through the grip strength-physical inactivity interaction term in the generalized estimated equation (GEE) model. After determining the joint effect of grip strength and exercise on depression, time-variable was interacted with physical inactivity according to grip strength groups to identify the independent effect of exposure on the change of depression over time. The depression was treated as a continuous variable in all GEE models. The independent working correlation structure was chosen in the GEE analysis.
All of the GEE models were adjusted for the potential confounders including age, gender, European region, married status, education, employment status, family economic level, smoking status, alcohol intake, heart attack, hypertension, hyperlipidemia, diabetes, mobility limitation, BMI and cognitive function. A two-tailed P value less than 0.05 was recognized as statistically significant.