R environment for statistical computing

The Shapiro-Wilk test was used to assess if the data were distributed normally. All values were displayed as the mean ± standard deviation or median with an interquartile range if the data were distributed skewed or nonskewed, respectively. Since each group's values were shown to have a nonskewed distribution, a nonparametric test (the Friedman test) was performed to analyze these data. Per-vessel diameter distributions measured by Mi-croTools were compared between different sublingual fields using linear mixed model analysis [16] . The sublingual field was entered into the model as fixed effects, and individual intercepts for subjects and per-subject random slopes representing the effect on the dependent variables were entered as random effects. p values were calculated using a likelihood ratio test of the full model with the effect in question against a "null model" that lacks the effect in question [17] . p values for individual fixed effects were obtained via the Satterthwaite approximation [18] . A two-sided p value <0.05 was considered as a criteron for significance. All the statistical tests were performed within the R environment for statistical computing, version 3.4.2 (R Foundation for Statistical Computing, Vienna, Austria; http://www.R-project.org/), and the figures were created with the ggplot2 software package, version 2.2.1.

Data were analysed using the R environment for statistical computing (R Foundation for Statistical Computing). Continuous variables are summarised as the mean AE s.e.m., whereas categorical variables are summarised as frequencies and percentages. Clinical specialties were grouped into six categories according to the area of the body and the body system involved (e.g. neurology and neurosurgery were grouped together, as were cardiology, cardiac surgery and vascular). The 'other' group contained another seven specialties grouped together for the purpose of statistical analysis, each having only a small number of patients; these were haematology, hepatology, medical and radiation oncology, interventional radiology and renal and respiratory medicine. One-way analysis of variance (ANOVA) was used to investigate the relationship between continuous BMI and hospital speciality. Multivariate Cox proportional hazards regression was used to analyse the time to leaving hospital, where those who died were censored at their date of death. The initial model included effects of BMI group, age, sex and hospital speciality, as well as all two-way interactions with BMI group. Variables that were significant at the 5% level were retained for the final model. Hazard ratios (HR) and 95% confidence intervals (CI) are provided.

Differential expression was assessed by performing a Student t test on the coefficients of simple linear models created with the ''limma'' R package (version 3.14.4). Correction for multiple hypothesis testing was performed with the Benjamini-Hochberg false discovery rate (FDR). Hierarchical clustering was performed with complete linkage and Euclidean distance metric, using expression values z-normalized to a mean of zero and standard deviation of one across all samples. All microarray analyses were performed using the R environment for statistical computing (version 2.15.1; the R Foundation for Statistical Computing, Vienna, Austria).