R statistical software environment

RNA-sequencing values of transcripts were downloaded in counts units from the University of California Santa Cruz public repository [22 (link)]. The differential expression analysis was performed using limma (limma package version 3.48.3 in R statistical software environment version 4.0.3-The R Foundation for statistical Computing, Vienna, Austria implemented in R studio version 1.4.1717, RStudio, Boston, MA, USA). The statistically significant genes were extracted by filtering by p adjusted value < 0.01.

Continuous variables are expressed as means ± standard deviation, whereas categorical variables are expressed as proportions. We used student's t‐tests and χ² tests to evaluate group differences in continuous and categorical variables, respectively. Kaplan−Meier curves were used to visualize and identify the primary endpoint of patients with or without dementia, and the difference was evaluated by the log‐rank test, also used for the pairwise comparison of subgroups.
The Cox proportional hazards model was constructed to adjust all variables collected for baseline characteristics and MPR. The adjusted covariates were age, sex, hypertension, diabetes, dyslipidemia, chronic kidney disease, end‐stage renal disease, peripheral arterial occlusive disease, chronic obstructive pulmonary disease, liver disease, malignancy, income levels, discharge medications, and MPR of aspirin in the first year. The results were expressed in terms of a hazard ratio (HR) and the corresponding 95% confidence interval (CI). We then performed stepwise regression under Akaike's information criterion to determine the appropriate multivariate model. All reported p‐values were two‐tailed, and p ≤ .050 indicated statistical significance. All statistical analyses were performed using R Statistical Software/environment (version 3.4.3; The R foundation for Statistical Computing).

Continuous variables are expressed as mean ± standard deviation (SD). Categorical variables are expressed as frequencies and percentages. The risk of incident AF associated with DM and proteinuria was compared using Kaplan–Meier analysis. The proportional hazard assumption was tested based on the Schoenfeld residuals. In the sensitivity analyses, we assessed the association between proteinuria, which was entered into the models as a time-varying factor, and each outcome using Cox proportional hazards regression models. The underlying time scale in these models was the observational period. The hazard ratios (HR) of AF for each combination of proteinuria change were calculated in Cox models using the consistently negative proteinuria group as the reference category. Urine dipstick results between the first and third tests were used to define changes in proteinuria. In these analyses, follow-up for AF was initiated from the third health examination (eFigure 1).
All tests were two-tailed, and P-values < 0.05 were considered statistically significant. Balance between different exposure groups was evaluated by standardized differences of all covariates, using a threshold of 0.1 to indicate imbalance. Statistical analyses were conducted using R Statistical Software/Environment (version 3.5.1, The R Project for Statistical Computing, Vienna, Austria).

The sample size of subjects recruited was determined according to availability, willingness and the statistically minimal sample size needed. Statistical analyzes were performed by JM using an R statistical software environment (www.r-project.org, accessed on 11 July 2022). All significance levels were set at α = 0.05. Normality of the data was tested using the Shapiro–Wilk test. If the test rejected normality, the Mann–Whitney test was applied; if not, two-sample t-tests were used.

Sequence pre-processing followed methods previously described [37 (link)], but modified by subsampling at 11,000 sequences per sample. QIIME 1.6.0[38 (link)] was used for initial stages of sequence analysis. Sequences were clustered into OTUs (operational taxonomic units, a proxy for ‘species’) using UCLUST [39 (link)] at 97% sequence similarity. Bacterial diversity was calculated using the following alpha diversity indices: 1) Shannon diversity index; 2) Faith’s phylogenetic distance (PD); and 3) Chao I species estimation; and 4) number of observed OTUs. Relative abundance of bacteria was calculated based on taxonomic classification of sequences using the RDP classifier [40 (link)] at a confidence threshold of 0.8. Microbiome data was analyzed with the R statistical software environment (www.r-project.org). Statistical significance was determined using two-sample Wilcoxon tests and corrected for multiple comparisons by FDR where appropriate.