Page 92 - 《运动与健康科学》(英文)2024年第2期

TaggedAPTARAEnd214 J. Tarp et al.
14
moderate activity ): no leisure-time PA, below recommenda- models with incremental levels of adjustment. Model 1: age
tion (>07.49 MET-h/week), at recommended level (7.514.9 and sex adjusted. Model 2: Model 1 + sociodemographic varia-
MET-h/week), above recommendation (15 MET-h/week). 5,10 bles, lifestyle factors (smoking, dietary quality indicators,
Quality control was performed by excluding implausible values alcohol intake, PA for transportation and occupation), family
defined as the sum of self-reported behaviors exceeding history of diabetes/cancer/CVD, inclusion method (self-re-
24 h/day. PA for transportation was categorized as passive, ported type 2 diabetes/use of diabetes medication or
walking, or cycling. Occupational PA was classified as seden- biochemistry), poor mental health, and diabetes duration.
Model 3 (main model): Model 2 with further adjustment for
tary, standing, or manual/heavy manual work.TaggedAPTARAEnd
BMI (conceptualizing BMI as a source of confounding 22,23 ).
In an additional Model 4, we further adjusted our main model
TaggedAPTARAH22.5. Outcome ascertainmentTaggedAPTARAEnd
for pharmacological treatment of blood glucose, lipids, and
TaggedAPTARAPAll-cause mortality and CVD, defined as cardiovascular
blood pressure as potential mediating factors.TaggedAPTARAEnd
mortality and major adverse cardiovascular events, were
TaggedAPTARAPThe continuous doseresponse pattern was modeled based
obtained through linkage to national registries. All-cause
on the main model using a restricted cubic spline. Because
mortality was the pre-specified primary outcome, cardiovas-
many participants had 0 MET-h/week, 3 knots were placed at
cular mortality and incidence are secondary outcomes. Partici-
the 10th median and 90th percentiles of the exposure distribu-
pants were followed until death, emigration, loss to follow-up,
tion among participants with non-zero leisure-time PA. Depar-
withdrawal from the study, or end of observation time (Supple-
ture from linearity was assessed by a Wald test of the
mentary Table 2 of Supplementary File 1), whichever occurred
null-hypothesis that the coefficient of the second spline is
first. Cardiovascular mortality and major adverse cardiovas- 24
equal to 0. MET-h/week was winsorized at the 95th percen-
cular events were coded according to the International Classi-
tile in these analyses because the highest values are at the
fication of Diseases, Injuries, and Causes of Death, 10th
greatest risk of reporting error. Absolute risk differences were
Revision. Cardiovascular mortality was any primary cause of
estimated as the standardized 10-year cumulative mortality
death coded as I00 to I99. Major adverse cardiovascular events
using a flexible parametric survival model (not specified in the
included the first incident episode of ischemic heart disease 25
pre-defined analysis plan). TaggedAPTARAEnd
(I20I25) or stroke (I60, I61, I63, and I64) identified from
TaggedAPTARAPEffect modification by sex, whether the patient was identi-
hospital records, in addition to cardiovascular mortality.TaggedAPTARAEnd
fied with type 2 diabetes from self-report or biochemistry, age
(<60 vs. 60 years old), diabetes duration (<5 vs. 5 years),
TaggedAPTARAH22.6. Statistical analysisTaggedAPTARAEnd
and by a history of pre-existing CVD was examined by stratifi-
TaggedAPTARAPA statistical analysis plan was developed and registered at cation and evaluated statistically using the likelihood-ratio
ClinicalTrials.gov (NCT05380232, Supplementary File 2) test. The pre-defined statistical analysis plan specified an
prior to commencing the analysis. To limit the potential influ- analysis repeating model 3 including both individuals with and
ence of major somatic or psychological conditions leading to without pre-existing CVD. Instead, we repeated Model 3 but
reduced PA (i.e., reverse causation), we excluded participants restricted it to individuals with pre-existing CVD only in order
if they had any of the conditions listed in Supplementary to provide estimates directly applicable to this group of
Table 3 of Supplementary File 1 (UK Biobank: n = 6310; high-risk patients. Finally, we performed sensitivity analyses
China Kadoorie Biobank: n = 1679). This list of conditions is by re-analyzing our data (using Model 3) restricted to never
not intended to be complete but to remove individuals with smokers and adjusting for more detailed diet information from
very severe medical conditions based on available data (e.g., 24-h recalls conducted between 2009 and 2012 (UK Biobank,
chronic degenerative neurological problems, renal failure, or subsample only), and restricted to individuals classified as
chronic obstructive pulmonary disease) as well as those indi- “possible type 2 diabetes” who had HbA1c < 48 mmol/mol
viduals with conditions most likely to interfere with engage- (UK Biobank).TaggedAPTARAEnd
ment in PA (e.g., inability to walk or chronic widespread TaggedAPTARAPThe proportional hazards assumption was verified by
pain). In our primary analyses, we also excluded individuals loglog plots and by Schoenfeld residuals plotted against
with a history of pre-existing CVD or cancer at baseline (defi- follow-up time. Cardiovascular mortality and major adverse
nitions provided in Supplementary Table 3 of Supplementary cardiovascular events were modeled using Fine-Gray models 26
File 1, UK Biobank: n = 4163; China Kadoorie Biobank: with death from other causes as a competing event. Statistical
n = 3900). Individuals with pre-existing CVD were included in analyses were performed using Stata Version 16.0 (StataCorp.,
a secondary analysis. Participant flowcharts are presented in College Station, TX, USA). Statistical significance was
Supplementary Fig. 1 of Supplementary File 1.TaggedAPTARAEnd a = 0.05 (two-sided).TaggedAPTARAEnd
TaggedAPTARAPStatistical adjustment was informed by a cohort-specific
directed acyclic graph (Supplementary File 2). Associations
TaggedAPTARAH13. ResultsTaggedAPTARAEnd
are presented as hazard ratios (HRs) with 95% confidence
intervals (95%CIs) from Cox proportional hazards regression TaggedAPTARAPWe identified 29,236 (5.8%) and 30,155 (5.9%) adults with
models. Age was modeled as the timescale, and follow-up type 2 diabetes in the UK Biobank and China Kadoorie
started 3 years after the baseline assessment (delayed entry). biobank, respectively. From these, a total of 14,913 and
Cohorts were analyzed separately based on 4 regression 17,457 participants with type 2 diabetes and no history of