Statistical analysis
Data were analyzed using the statistical package of Graphpad Prism version 6.05 (Graphpad software, La Jolla, California, USA). All continuous data were tested for normal distribution using the D’Agostino-Pearson omnibus normality test, and presented as mean (SD) or as median with the IQR, where appropriate. Nominal data were compared using the Chi-square test. Groups were compared using the ANOVA test for unpaired data. A post-hoc analysis was done where ANOVA showed significant differences, using the t-test. For normally distributed data, comparing pre-HUT and post-HUT values, we used paired t test. In data that were not normally distributed the Mann-Whitney test was used. Due to the multiple comparisons, we chose a conservative p-value of <0.01 to be statistically significant.