Appendix S1 Supplementary Material (Methods)
Statistical analysis
The sample size was determined based on an expected in-hospital mortality rate of 35%. A minimum of 200 patients was needed to include a maximum of 7 independent variables in the multivariate models.Assuming an α value of 0.05, the study has 80% power to detect a difference of 20% in in-hospital between patients with S. aureus vancomycin MIC ≥1.5 μg/mL and the rest.
Categorical data are reported as the frequency and percentage. Continuous data are expressed as the mean and standard deviation or the median and interquartile range, as appropriate. The patients’ clinical variables and the S. aureus microbiological variables were compared between survivors at discharge and non-survivors, overall and in the subgroup of patients with MSSA infection. Microbiological variables were compared according to the secondary end-points. Categorical variables were compared using the chi-square test or Fisher exact test, and continuous variables using the two-sample t-test or Kruskal-Wallis test, as appropriate.
To determine the association of clinical and microbiological characteristics with in-hospital death, we constructed a multivariate logistic regression model. All clinical characteristics that differed between groups (significance level, P<0.2) and those considered independent clinical variables had been defined, we introduced the candidate microbiological variables using the same criteria as above, forcing the variable vancomycin MIC ≥1.5 μg/mL into the final model. The following variables were considered as potential confounders of the association between microbiological variables and in-hospital mortality and therefore were included in the first model: female gender, age adjusted Charlson comorbidity index, prosthetic-valve IE, vegetation diameter, congestive heart failure (NYHA class 3 or 4), symptomatic central nervous system complication, acute renal failure, new conduction abnormality, severe sepsis/septic shock, and surgery performed. Unadjusted and adjusted associations are reported using odds ratios and 95% confidence intervals.
As the only primary endpoint was in-hospital mortality and all other clinical endpoints were defined as secondary we did not consider the need for multiplicity adjustment [32]. Therefore all the associations described on secondary outcomes should only be considered exploratory and subject to future research.
Statistical significance was set at P<0.05 and hypothesis tests were two-sided.Statistical analyses were performed with STATA 11.0.