Maternal outcomes:
All analyses in this secondary analysis followed the intention-to-treat principle: all randomly allocated women were analysed according to the group they were allocated to, irrespective of the treatment they received, if any. The analyses required data collected at post-randomisation visits therefore women without post-randomisation visit data were excluded.
For bile acid concentrations, subgroups were defined based on accepted thresholds from the literature related to perinatal risk.2For itch, subgroups were defined based on median itch at baseline in the trial participants.6The effect of baseline bile acid concentration (<40 μmol/L vs ≥40 μmol/L, <100 μmol/L vs ≥100 μmol/L) and baseline itch score (<60 mm vs ≥60 mm) on two maternal outcomes were analysed: i) serum bile acid concentration post-randomisation and ii) itch score post-randomisation. Baseline bile acid trajectory was defined as ‘increasing’ if the first bile acid concentration post-randomisation was greater than or equal to the baseline bile acid concentration and ‘decreasing’ if the first bile acid concentration post-randomisation was less than the baseline bile acid concentration. The effect of this baseline itch trajectory on the two maternal outcomes was also analysed.
As bile acid concentrations demonstrate a lognormal distribution, the geometric mean of all available post-randomisation bile acid concentrations was used to calculate each participant’s mean post-randomisation bile acid concentration, and the trial groups were compared using a geometric mean ratio. An arithmetic mean of all available itch scores post-randomisation was used to calculate each participant’s mean post-randomisation itch score, and the trial groups were compared using a mean difference. An interaction test (likelihood ratio version) was used to test for a difference in treatment effect between the individual subgroups.
The treatment effect in each group was visualised by plotting the average itch score or geometric average bile acid concentration by visit. Visits were seven days apart, plus or minus one day. Only visits which had five or more women contributing results were included. Error bars were added to represent the standard error of each group at each visit. All available pairs of itch scores and bile acid concentrations were plotted for all participants at all time points, and a correlation coefficient between the two was calculated.