Data source and study population
In our single-center study performed at MJ Healthcare Center in Beijing retrospectively, we enrolled participants aged ≥ 20 years undergoing at least two regular health check-ups from 2009 to 2019. These participants attended one health check-up from 2009 to 2012 and another one or two check-ups from 2013 to 2019. Each health check-up included physical examination, biochemical tests, a pulmonary function test (without bronchodilator) and a questionnaire. Information in the questionnaire included past medical histories, medication use, current health situation, family history, and exercise frequency. Participants were excluded at baseline if they had (1) cardiovascular diseases, (2) chest surgery, (3) previous or current chronic pulmonary diseases (e.g., lung cancer, asthma, pulmonary tuberculosis), and (4) medication history for cardiovascular disease or asthma.
Definitions of diabetic status
Diagnosis of diabetic status was made according to either of the following three conditions: (1) according to the International Statistical Classification of Diseases and Related Health Problems, 11th revision (ICD-11), codes 5A11, subjects with fasting plasma glucose (FPG) levels ≥ 126 mg/dL (7.0 mmol/L) at the first health check-up; (2) participants who mentioned diabetes in their past medical history in the questionnaire; (3) participants who stated that they took medication for diabetes in the questionnaire.
For both prediabetic and normal individuals, there should be no mention of diabetes or use of medication for diabetes in the questionnaire. Diagnosis of prediabetic status was made based on the FPG level at the first health check-up, with FPG level ≥ 110 mg/dL (6.1 mmol/L) but < 126 mg/dL (7.0 mmol/L) . Diagnosis of normal status was also made based on FPG level, with FPG < 110 mg/dL (6.1 mmol/L).
Definitions of different lung function status
We grouped participants into Normal, Preserved Ratio Impaired Spirometry (PRISm), COPD groups according to their baseline lung function. PRISm: FEV1/FVC ≥ 0.7 and FEV1% < 80%. COPD: FEV1/FVC < 0.7. These definitions were made based on spirometry without bronchodilator.
Pulmonary function test performance
In this study, FEV1 and FVC were measured using a portable spirometer (CHEST HI-101, Japan), which was calibrated daily with a 1-L syringe. Pulmonary function tests were performed by experienced technicians at MJ Healthcare Center. Each participant completed two spirometry attempts without bronchodilator while seated. If the two readings differed significantly, a third measurement was required . Lung function data were reported in absolute values, a percentage of predict value (% pred)  and 1-s rate (FEV1/FVC). Spirometry was performed based on recommendations from the Epidemiology Standardization Project  and the American Thoracic Society . The quality control and reproducibility were coordinated by a group formed by the Meinian Institute of Health and China-Japan Friendship Hospital. We calculated the ratios of observed to predicted FEV1 based on Chinese population references .
Collection and classification of other variables
To stratify our population by age, we divided participants into the following six age groups: 20 ~ 30, 30 ~ 40, 40 ~ 50, 50 ~ 60, 60 ~ 70, and > 70 years. For stratification analysis of body mass index (BMI), according to the Chinese criteria, we divided participants into three groups: BMI < 24 kg/m2, 24 ≤ BMI < 28 kg/m2 (overweight), and BMI ≥ 28 kg/m2 (obese). For stratification analysis of physical activity frequency, the participants were divided into three groups based on their self-reported details in the questionnaire: < 1 time/week, 1–6 times/week, and ≥ 7 times/week. For smoking status, we stratified the participants into the never smoking group, former smoker group, and current smoker group according to their self-reported details in the questionnaire. Hypertension was diagnosed based on previous medical history or anti-hypertension medicine history self-reported by the participants in the questionnaire. Diabetes medication history was also obtained from the questionnaire. Total cholesterol, triglycerides, FPG, and high-density lipoprotein (HDL) data were obtained from biochemical tests.
Differences across groups were compared using ANOVA for continuous variables and χ2 test for categorical variables. Multivariable linear regression analysis was performed to measure the association of FEV1, FVC, FEV1%, FVC%, FEV1/FVC with both FPG level at baseline from 2009 to 2012 and at the time when lung function parameters were measured again, adjusting for age, sex, BMI, smoking status, physical activity frequency, total cholesterol, triglycerides, HDL, hypertension status, and medication use. Participants were excluded if any variable was not available at baseline.
For the longitudinal study, to incorporate correlations for all observations arising from the same person, we employed linear mixed model with correlated measurement errors to analyze the association of diabetic status with lung function parameters. An Unstructured covariance matrix was specified and maximum likelihood was used to estimate the unknown covariance parameters. Only fixed effects were investigated. The subgroup analysis based on the baseline lung function (PRISm, COPD, Normal) was also performed for the association of FPG and lung function through linear regression models and mixed models.
In the trajectory analysis, we defined the outcome of interest and the time variable as FPG and the follow-up time at three time points. Bayesian Information Criterion (BIC) values were used to select best models and we finally fitted 3 groups for the trajectory of FPG with the order of each equation that described the changes over time in each group as quadratic, cubic, and cubic. The linear regression model was applied to investigate the association between FPG trajectories and lung function with the change rate of pulmonary parameters as the response variable. The change rate of pulmonary parameter was calculated as: (pulmonary parameter at the last measurement time – pulmonary parameter at baseline)/ follow-up time.
All analyses were two-tailed, performed using SAS 9.4 (SAS Institute Inc., SAS Campus Drive, Cary, North Carolina 27,513, USA), with P-value < 0.05 considered to be statistically significant. The linear mixed model was performed using the Proc Mixed procedure with Repeated statement in the SAS system. The trajectory analysis was performed using the SAS Trajectory Procedure (Proc Traj) .