Statistical analysis

Baseline characteristics will be summarised by treatment group to assess comparability. We are not planning to adjust the analyses for baseline variables.

Data analysis will be performed according to the intention-to-treat principle. The primary outcome will be analysed by a comparison of scores using an independent t test, with 95% confidence intervals and two-tailed p values reported. Significance levels will be set at 0.05.

The 24-h opiate usage and hospital length of stay will by analyses using independent t tests with 95% confidence intervals and two-tailed p values reported.

Nausea and vomiting will be recorded as categorical variables and analysed using the Pearson chi-squared test.

For the analysis of pain scores across the four periods a linear regression analysis will be used with the dependent variable being the pain score and the independent variables being ‘period’, ‘group’ and ‘period*group’ interaction. This analysis will be conducted within the framework of Generalised Estimating Equations, which corrects the bias in the estimates of the SE caused by having multiple observations per person and thus correlation in the data. The output generated would be the difference between the groups across all periods (95% CI and p value) and the difference between the groups in each period (95% CI and p value).