Replacement of sedentary behavior with various physical activities and the risk of all-cause and cause-specific mortality | BMC Medicine

Study design and population

The UK Biobank [15] is the largest repository of genetic and environmental factors related to disease pathogenesis or prevention in the United Kingdom to date (www.ukbiobank.ac.uk/). Invitations were mailed to 9.2 million individuals aged 37 to 73 years who were registered with the National Health Service (NHS) in the UK and resided within a short travel distance of one of 22 dedicated assessment centers (typically approximately 25 miles). Between 2006 and 2010, the UK Biobank recruited 502,000 participants (5.5% of those invited); collected their genetic information, blood samples, lifestyle, and environmental exposure data; and subsequently tracked their health and medical records for several decades. The UK Biobank program was approved by the Northwest Multicenter Research Ethics Committee (16/NW/0274). Informed consent was obtained from all the participants. This study was conducted based on data under application number 90060.

The National Health and Nutrition Examination Survey (NHANES) [16] is an ongoing health and nutritional survey program in the U.S. for adults and children conducted by the National Center for Health Statistics (NCHS) in the United States since 1999 (www.cdc.gov/nchs/nhanes/about_nhanes.htm/). The survey annually reviews a nationally representative sample of approximately 5000 individuals, which are located across counties nationwide. The NHANES interview component covers demographic, socioeconomic, dietary, and health-related questions, and the examination component includes physiological measurements, laboratory tests, etc. The NHANES employs a complex, multistage probability sampling design to select participants representing the civilian, noninstitutionalized U.S. population. Oversampling of certain demographic subgroups is conducted to increase the reliability and precision of health indicator estimates for these specific subgroups. The NHANES data are released biennially, and we utilized survey data from six survey cycles, namely, 2007–2008, 2009–2010, 2011–2012, 2013–2014, 2015–2016, and 2017–2018, supplemented by public-use linked mortality files (LMFs) updated to 2019. Because the public-use LMFs do not provide data on respiratory mortality after 2015, we used data from the first four cycles when studying respiratory mortality.

Assessment of SB and PA

The SB time from the UK Biobank was defined as the total time the participants spent watching TV, using a computer, or driving. They were asked how many hours they spent on these activities in a typical day. For participants whose sedentary time varied greatly in the last 4 weeks, they were required to provide the average amount of time. In the NHANES cohort, the SB time was based on the participants’ self-assessment [17]; they were asked how much time they usually spent sitting or reclining in a typical day, except for time spent sleeping. The answer range was limited to 0–24 h, and answers > 18 h required confirmation. We then categorized sedentary time into three levels according to the relevant literature: < 5 h/day, 5–8 h/day, and > 8 h/day [13, 18].

In the UK Biobank, participants reported 5 types of PA: walking for pleasure (not as a means of transport), light DIY (e.g., pruning, watering the lawn), heavy DIY (e.g., weeding, lawn mowing), strenuous sports (defined as inducing sweating and hard breathing), and other exercises (e.g., swimming, cycling). In the NHANES cohort, PA was classified as work activity (defined as paid or unpaid work, household chores, and yard work) and recreational activity, which, according to the degree of increase in breathing or heart rate, was divided into vigorous and moderate activity. Walking or bicycling (for transportation) was also included in the NHANES questionnaire. The participants reported the frequency and average duration of engagement in each activity per week. Using this information, we calculated the average daily duration of each activity.

Assessment of outcomes

The death data for the UK Biobank cohort was obtained from the NHS Information Centre and NHS Central Register. The NHANES linked data was collected from several NCHS population surveys with death certificate records from the National Death Index. The duration of follow-up was determined based on the earliest occurrence among the following endpoints: death, loss to follow-up, or May 25, 2022, for the UK Biobank, and December 31, 2019, for the NHANES. According to the 10th revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-10), the following specific causes of death were defined: cancer (C00–D48 for the UK Biobank; C00–C97 for the NHANES), CVD (I00–I79 for the UK Biobank; I00–I09, I11, I13, I20–I51 for the NHANES), respiratory diseases (J09–J18, J40–J47 for both the UK Biobank and the NHANES), and digestive diseases (K20–K93 for the UK Biobank).

Assessment of covariates

For both the UK Biobank and NHANES cohorts, we adjusted for critical covariates as follows: age, sex, race, socioeconomic status, education level, employment status, body mass index (BMI), smoking status, alcohol consumption frequency, dietary habits, overall health rating, and sleep duration. Owing to differences in questionnaire formats, methods of classifying covariates were slightly different between the two cohorts, and the details are shown in Additional file 1: Table S1–S2. In terms of dietary habit covariates, vegetable and fruit intake, and processed meat intake were adjusted for in the UK Biobank cohort, whereas in the NHANES, dietary quality from 24-h dietary recalls was used to determine healthy eating index (HEI) scores [19].

Statistical analysis

According to the sedentary time categories, we conducted descriptive statistics of the population characteristics. For categorical variables, systematic missing values and responses of “do not know” or “prefer not to answer” were consolidated into the “missing” category. Percentages were employed for descriptive purposes, and the chi-square test was used to examine the differences between groups. For continuous variables, absent values were imputed with the median. Variables conforming to a normal distribution were described using the mean and standard deviation (SD), while those deviating from normality were characterized by the median and interquartile range (IQR). The Kruskal–Wallis rank sum test was employed to assess variations between groups.

Our study included two retrospective analyses of longitudinal cohorts from the UK Biobank and NHANES. A Cox proportional hazards model was used to estimate the relationship between SB and the risk of all-cause and cause-specific mortality, with hazard ratios (HRs) and 95% confidence intervals (CIs) used to describe the results. All variables met the proportional hazards assumption via the Schoenfeld residual method (Additional file 1: Fig. S1). Restricted cubic spline models were employed to evaluate potential nonlinear relationships between the daily sedentary time and the risk of all-cause and cause-specific mortality. This analysis excluded individuals whose sedentary time fell outside the 0.5th–99.5th percentile.

On the basis of the assumption that the total daily discretionary time remains unchanged, we subsequently used ISM to estimate the effect of replacing SB with a certain type of PA on mortality. In the UK Biobank cohort, the model was as follows: h(t) = h₀(t) exp (β₁ walking for pleasure + β₂ light DIY + β₃ heavy DIY + β₄ strenuous exercise + β₅ other exercises + β₆ total discretionary time + β₇ covariates). For the NHANES cohort, the model was follows: h(t) = h₀(t) exp (β₁ walk or bicycle + β₂ moderate work activity + β₃ vigorous work activity + β₄ moderate recreational activity + β₅ vigorous recreational activity + β₆ total discretionary time + β₇ covariates). The total discretionary time was the sum of the sedentary time and total PA time. h(t) represents the hazard function of the Cox model, where h₀(t) denotes the baseline hazard. Coefficients β₁ to β₆ represent the effects of different types of PA replacing 30 or 60 min of SB. Furthermore, subgroup analyses in this study were based on sex, age, BMI, smoking status, and sleep duration.

We also conducted several sensitivity analyses to confirm the reliability of the main findings as follows: (1) excluding death cases within the first 2 years of follow-up, (2) using chained equations with five imputations for critical missing values, and (3) using the Fine–Grey competitive risk model to re-estimate the cause-specific mortality risk. In competing risk models, for cause-specific mortality, competing events refer to deaths from other specified causes, excluding the primary cause under consideration.

The statistical analyses in this study were performed using the R-4.3.0 software (R Foundation for Statistical Computing, Vienna, Austria), IBM SPSS Statistics 26 (IBM Corporation, Armonk, NY, USA), and Stata MP 17 (StataCorp LP, College Station, USA). A two-sided P value < 0.05 was considered statistically significant.