Statistical analysis

The statistical analysis can be reproduced in more detail in the statistical analysis plan (Doc S3). A power analysis, calculated with G*Power [20], yielded a sample size of 76 to 112 patients for an expected small to medium effect size of 0.4–0.5, two measurement time points, four groups, a level of significance of α=0.05, and a power of 0.95. Thus, we included 24 patients per group for a total of 96 patients. Data are reported as the means with 95% confidence intervals (CIs) unless otherwise specified. We performed a univariate analysis of variance (ANOVA) to determine differences in pain reduction on a NRS after surgery. For all other analyses, we used a repeated measures analysis of variance (rANOVA) to determine changes on a NRS or consumption in mg over time. The Greenhouse–Geisser or Huynh–Feldt correction for the F test was used to adjust the degrees of freedom for deviations from sphericity. Hence, the results revealed the interactive effect of time (postoperative days) and group. For significant main and interaction effects, exploratory post hoc analyses for significant results were derived from Fisher’s least significant difference (LSD) tests. Hence, we were interested in comparing all studied groups with each other. For the primary outcome, the multiple imputation method for missing observations was conducted after assessing whether the data were missing at random. For all performed analyses, two-sided P values of P<.05 were considered statistically significant. To account for multiple comparisons, Bonferroni corrections were applied. Analyses were performed with IBM SPSS Statistics software, version 27.0 (IBM Corp., Armonk, NY, USA).