Prism version 7

Fisher’s exact test was used to assess the association between categorical variables. Exact conditional logistic regression analysis was used for the multivariate analysis (16 ). Bootstrapping was performed using random sampling with replacement to create a large number (N = 982) of “phantom samples” known as bootstrap samples. The sample summary is then computed on each of the bootstrap samples. This method can be superior to approaches relying on the asymptotic distribution of the tests that assumes the data come from a normal distribution, allowing the data of the sample study at hand to be utilized as a surrogate for a larger population. Although cross validation remains the preferred approach for validating predictive models, bootstrapping can be used when the sample size is too small to be split into a training and a validation set and/or there is no independent cross-validation cohort (17 (link)) (as is the case in our study).
All tests were 2-sided. P values ≤ 0.05 were considered significant. Fisher’s exact test was performed using Graph-Pad Prism version 7.0 (San Diego, CA, USA). Multivariate exact conditional logistic regression was performed with SAS software, version 9.4 (Cary, NC, USA). Bootstrapping with multiple logistic regression analysis was performed with SPSS version 24.0 (Chicago, IL, USA).