Ursodeoxycholic acid to reduce adverse perinatal outcomes for intrahepatic cholestasis of pregnancy: the PITCHES RCT

Trial design

We carried out a parallel-group, masked, multicentre, randomised, placebo-controlled trial with individual randomisation to UDCA or placebo using a 1 : 1 allocation ratio. There were no substantial changes to the study design or methods after commencement of the trial.

Participants

Women were eligible if the attending clinician considered that they had a diagnosis of ICP (defined as maternal pruritus with a raised, randomly timed, serum bile acid concentration above the upper limit of normal as measured in the local laboratory),12 were between 20⁺⁰ and 40⁺⁶ weeks of pregnancy on day of randomisation, with a singleton or twin pregnancy, had no known lethal fetal anomaly, were aged ≥ 18 years, and were able to give written informed consent. A woman was not included in the trial if a decision had already been made for delivery within the next 48 hours, she had any known allergy to any component of the UDCA or placebo tablets, or if she had a triplet or higher-order multiple pregnancy. We undertook the trial in 33 maternity units in England and Wales. Seventeen units used a threshold of 14 µmol/l as the upper limit of normal, whereas the remaining units used thresholds between 9 and 13 µmol/l, in accordance with local laboratory reference ranges. The trial was approved by the East of England – Essex Research Ethics Committee (number 15/EE/0010).

Interventions

We allocated women to the group taking either UDCA tablets or matched placebo tablets, which were manufactured and supplied by Dr Falk Pharma GmbH (Freiburg im Breisgau, Germany). Each film-coated UDCA tablet contained the active ingredient, 500 mg of UDCA, and the inactive ingredients magnesium stearate, polysorbate 80, povidone K25, microcrystalline cellulose, colloidal anhydrous silica, crospovidone and talc. The matched placebo tablet was identical in colour and shape to the UDCA tablet and contained the same inactive ingredients. The tablets were packaged for oral administration and did not require any special storage conditions.

We recommended that women were started on a dose of two oral tablets per day (equivalent to 500 mg of UDCA twice per day), which was increased by a health-care professional in increments of one tablet per day every 3–14 days if there was no biochemical or symptomatic improvement, to a maximum of four tablets per day. The dose could be reduced to one tablet per day at a clinician’s discretion (e.g. if a woman’s weight was < 50 kg or if gastrointestinal side effects occurred). We advised that doses should be spread evenly throughout the day, but that no specific instructions to take with or without food needed to be given. In addition, we recommended that treatment should be continued from enrolment until the infant’s birth.

Outcomes

Outcomes were recorded on the web-based trial database through case note review by trained researchers after discharge from hospital of the woman and baby.

The primary perinatal outcome was prespecified as a composite of perinatal death (defined as in utero fetal death after randomisation or known neonatal death up to 7 days) or preterm delivery (< 37 weeks’ gestation) or neonatal unit admission for at least 4 hours (from infant delivery until hospital discharge). Each infant was counted once within this composite.

Secondary maternal outcomes (measured on clinical visits between randomisation and delivery) were maternal serum concentration of bile acids, alanine transaminase (or aspartate transaminase), total bilirubin and gamma-glutamyltransferase and maternal itch score (marked by the woman as the worst episode of itch over the past 24 hours in millimetres on a 100-mm visual analogue scale, where 100 mm is the worst possible itch). Additional secondary maternal outcomes (assessed on case note review after maternal hospital discharge) were gestational diabetes mellitus, mode of onset of labour and estimated blood loss after delivery.

Secondary perinatal outcomes (assessed on case note review after infant hospital discharge) included the components of the primary outcome, mode of delivery, birthweight, birthweight centile, gestational age at delivery, presence of meconium, Apgar score at 5 minutes, umbilical arterial pH at birth and total number of nights in the neonatal unit. All other secondary outcomes were descriptive only.

Health resource use post enrolment (i.e. collected at case note review after maternal and infant discharge from hospital) was the total number of nights in hospital (antenatal, intrapartum and postnatal) together with the level of care including adult intensive care unit, mode of delivery and UDCA cost (in the intervention group) for the woman, and total number of nights for the infant in the neonatal unit, together with the level of care (e.g. intensive care) for the infant.

There were no changes to primary or secondary outcomes after the trial started.

Sample size

The sample size was informed by the Cochrane meta-analysis,10 from which the event rate for the primary outcome for infants of untreated women was estimated as 40%. We determined that 550 infants of women with ICP (275 per group) were required to have a 90% chance of detecting (as significant at the two-sided 5% level) a reduction in the primary outcome measure from 40% in the control group to 27% in the treated group, corresponding to an absolute risk reduction of 13% and a risk ratio (RR) of 0.675. This was conservative compared with the effect sizes seen in the Cochrane meta-analysis10 for the three individual end points (RR 0.31, 0.46 and 0.48 for perinatal death, preterm delivery and neonatal unit admission, respectively). We planned to recruit 580 women in total, to allow for the possibility of 5% of infants being lost to follow-up. During the recruitment phase, we amended the protocol to permit continued recruitment of additional participants to allow for women who discontinued the intervention or withdrew from the trial. Interim analyses were undertaken for presentation only to the Data Monitoring Committee, to be reviewed when they met at least annually.

Randomisation

Randomisation was performed using a probabilistic minimisation algorithm13 to ensure approximate balance within the following groups: study centre, gestational age at randomisation (< 34, 34 to < 37, ≥ 37 weeks’ gestation), single versus twin pregnancy, and highest serum bile acid concentration prior to randomisation (< 40 µmol/l, ≥ 40 µmol/l). Randomisation was managed via a secure web-based randomisation program (MedSciNet, Stockholm, Sweden).

Allocation concealment

Packs containing UDCA or placebo were produced by the central manufacturing unit at Guy’s and St Thomas’ NHS Foundation Trust, London, UK, prior to shipping to site pharmacies. Packs were labelled with unique pack identifiers according to a randomly generated sequence known only to the manufacturing unit and the trial programmers. Participants were allocated a pack identifier at randomisation and, if more packs were required, the randomisation program was used to allocate further packs containing the same allocation.

Implementation

The minimisation algorithm was implemented by a MedSciNet database programmer, with balance and predictability checked by an independent National Perinatal Epidemiology Unit Clinical Trials Unit statistician during the trial. Research teams at sites approached women to confirm eligibility and provided verbal and written information. A trained clinician obtained written informed consent. A research team member entered baseline data on a web-based database at study enrolment and then allocated a pack number using the web-based randomisation, which corresponded to a pack for dispensing by that site’s pharmacy.

Clinical teams reviewed participants at routine care clinic visits until delivery. Antenatal care, in particular the timing and mode of delivery, was left to the discretion of the responsible clinician. Research teams undertook standard assessments of safety, with reporting of adverse events and serious adverse events following usual governance procedures.

Masking

Trial participants, clinical care providers, outcome assessors and data analysts were all masked to allocation. The UDCA and placebo tablets appeared identical in size, shape and colour.

Statistical methods

Additional trial information

Participant flow, recruitment and numbers analysed

Between 23 December 2015 and 7 August 2018, 605 women (43%) were recruited out of 1418 women who were found to be eligible, including 37 women with a twin pregnancy (Figure 1), across 33 maternity units (Table 2).

FIGURE 1.

The CONSORT flow diagram. a, Not mutually exclusive.

A total of 305 women were allocated to the UDCA group, with data from 304 women (one woman withdrew, with consent to use all data withdrawn) and 322 infants included in the primary outcome analysis. A total of 300 women were allocated to the placebo group, with all 300 women analysed (one woman withdrew, with consent to use baseline data but not to collect outcome data) and 318 infants included in the primary outcome analysis. Follow-up to maternal and infant discharge from hospital continued until December 2018. As we recruited ahead of schedule, we continued recruitment up to the number of women who discontinued the intervention or withdrew from the trial (with approval of the funder, sponsor and ethics committee), such that our total number of women recruited (n = 605) included the target sample size (n = 550 women), the number who discontinued the intervention (n = 53) and those who withdrew (n = 2). Recruitment ended after 605 women had been enrolled.

Baseline data

Baseline characteristics were similar between the two groups (Table 3). At trial enrolment, the groups were well balanced on minimisation factors (see Table 3).

Outcomes and estimation

There was no evidence of a significant difference between the groups in the incidence of the primary outcome (perinatal death, preterm delivery or neonatal unit admission for at least 4 hours): 74 (23.0%) infants in the UDCA group compared with 85 (26.7%) infants in the placebo group experienced the primary outcome (aRR 0.85, 95% CI 0.62 to 1.15; p = 0.279) (Table 4). Similarly, there was no evidence of a significant difference between the groups in the incidence of the individual components of the primary outcome components (see Table 4). There were three in utero fetal deaths after randomisation, one in the UDCA group and two in the placebo group, with two occurring at 35 weeks’ gestation and one at 37 weeks’ gestation.

There was no evidence of a significant difference in median gestational age at delivery (Table 4); time to delivery is shown in Figure 2. The proportion of women having spontaneous vaginal births or caesarean sections was similar in both groups (Table 4). No evidence of significant differences was seen between the groups in the total number of nights in the neonatal unit or in the main diagnosis for neonatal unit admission (the latter was not formally tested). A full listing of perinatal secondary outcomes is shown in Tables 4 and 5.

FIGURE 2.

Kaplan–Meier survival curve of time from randomisation to delivery estimate.

There was evidence of a significant difference between the groups on post-randomisation maternal itch score, which was lower in the UDCA group than in the placebo group (MD –5.7 mm, 95% CI –9.7 to –1.7 mm; p = 0.005) (Table 6). The actual and estimated mean trajectories are adjusted for the baseline measures and minimisation factors in Figures 3 and 4. Figure 3 shows the mean or geometric mean trajectory for maternal itch score, serum bile acid concentrations and serum alanine transaminase concentrations (with 95% CIs) up to 10 weeks post randomisation; Figures 3a, c and e show the observed trajectories and Figures 3b, d and f show the trajectories adjusted for the baseline value and minimisation factors. For example, Figure 3b shows that, at 1 week after randomisation, the mean itch score is estimated to be 56 mm (95% CI 51 to 61 mm) in the placebo group and 50 mm (95% CI 45 to 55 mm) in the UDCA group, after adjusting for the itch score at randomisation and minimisation factors. Figure 4 shows the same trajectories (actual and estimated) plotted against week of gestation throughout the trial from randomisation onwards. For example, Figure 4c shows that the observed geometric mean of serum bile acid concentration (µmol/l) when women were between 29⁺⁰ and 30⁺⁶ weeks’ gestation was 20 µmol/l (95% CI 17 to 25 µmol/l) in the placebo group and 27 µmol/l (95% CI 22 to 33 µmol/l) in the UDCA group. Serum bile acid concentrations reduced in both groups over time after study enrolment; however, there was evidence of less reduction in serum bile acid concentrations post randomisation in the UDCA group than in the placebo group [adjusted GMR 1.18 µmol/l (95% CI 1.02 to 1.36 µmol/l; p = 0.030)]. In contrast, there was evidence of a reduction in serum alanine transaminase concentrations post randomisation in the UDCA group compared with the placebo group (adjusted GMR 0.74 U/l, 95% CI 0.66 to 0.83 U/l; p < 0.001) (see Table 6, and the actual and estimated trajectories in Figures 3 and 4). Other maternal secondary outcomes are shown in Tables 6 and 7.

FIGURE 3.

Changes in maternal itch score and serum concentrations of bile acid and alanine transaminase between randomisation and delivery over 10 weeks since randomisation with 95% CIs, by allocation. (a) Actual mean trajectory of changes in maternal itch score (mm); (b) estimated mean trajectory of changes in maternal itch score (mm); (c) actual geometric mean trajectory of changes in serum bile acid concentration (µmol/l); (d) estimated geometric mean trajectory of changes in serum bile acid concentration (µmol/l); (e) actual geometric mean trajectory of changes in serum alanine transaminase concentration (U/l); and (f) estimated geometric mean trajectory of changes in serum alanine transaminase concentration (U/l). The estimated geometric mean trajectories (b, d and f) are adjusted for baseline measures and minimisation factors.

FIGURE 4.

Changes in maternal itch score and serum concentrations of bile acid and alanine transaminase between randomisation and delivery by gestational age with 95% CIs, by allocation. (a) Actual mean trajectory of changes in maternal itch score (mm); (b) estimated mean trajectory of changes in maternal itch score (mm); (c) actual geometric mean trajectory of changes in serum bile acid concentration (µmol/l); (d) estimated geometric mean trajectory of changes in serum bile acid concentration (µmol/l); (e) actual geometric mean trajectory of changes in serum alanine transaminase concentration (U/l); and (f) estimated geometric mean trajectory of changes in serum alanine transaminase concentration (U/l). The estimated mean trajectories (b, d and f) are adjusted for baseline measures and minimisation factors.

Around two-thirds of women took a maximum of one tablet twice per day (equivalent to 1000 mg in the UDCA group). Similar numbers of women in both groups discontinued the intervention [24 (7.9%) in the UDCA group compared with 29 (9.7%) in the placebo group], with similar numbers of discontinuations across the groups instigated by clinicians and participants (Table 8).

In a prespecified planned sensitivity analysis excluding infants whose mothers took < 90% of the trial medication, a similar proportion of infants experienced the primary outcome of perinatal death, preterm delivery or neonatal unit admission for at least 4 hours: 49 out of 217 (22.6%) infants in the UDCA group compared with 44 out of 190 (23.2%) in the placebo group (aRR 0.91, 95% CI 0.63 to 1.32; p = 0.627) (Table 9).

Ancillary analyses

In prespecified planned subgroup analyses, we found that there was no evidence of a significant interaction of highest serum bile acid concentration prior to randomisation (stratified as < 40 µmol/l and ≥ 40 µmol/l), gestational age at randomisation (< 34 weeks’ gestation, ≥ 34 weeks’ gestation), or multiplicity (singleton or twin pregnancy) and the incidence of the primary outcome, nor its components, nor important maternal secondary outcomes (itch score and serum bile acid concentration post randomisation) (Figure 5 and Table 10).

FIGURE 5.

Forest plots showing subgroup analysis for the primary outcome and its main components. (a) Primary outcome; (b) preterm delivery; and (c) neonatal unit admission.

In requested post hoc exploratory analyses, the proportion of infants with the primary outcome in mothers with the highest serum bile acid concentrations of ≥ 100 µmol/l at randomisation were similar: 9 out of 23 infants (39.1%) in the UDCA group compared with 7 out of 17 (41.2%) in the placebo group (see Table 7). There was also no evidence of a difference in the proportion of women with spontaneous or iatrogenic preterm birth between the groups (Figure 6).

FIGURE 6.

Forest plots showing post hoc subgroup analysis for (a) spontaneous preterm delivery; and (b) iatrogenic preterm delivery.

Harms

Eight serious adverse events were reported, all of which were considered to be unrelated to the trial intervention. Two occurred in the UDCA group and six occurred in the placebo group, relating to a range of organ systems (Table 11). Seventy-three adverse events were reported: 31 in the UDCA group and 42 in the placebo group. Of note, the same number in each group (n = 10) reported adverse events related to gastrointestinal disturbances.

Economic analysis

There was no evidence of a significant difference in total costs (maternal, infant and cost of UDCA) between the two trial groups: mean £5420 [standard error (SE) £284] in the UDCA group compared with mean £5892 (SE £353) in the placebo group [adjusted difference –£429, 95% CI –£1235 to £377: adjusted p-value 0.297 (Table 12)].

Generalisability

The study was undertaken in 33 maternity units across England and Wales, and included women representative of the wider pregnancy population in terms of demographics and spectrum of disease. Recruitment occurred within time and target, indicating equipoise and willingness to participate from clinicians and pregnant women.

Limitations

Limitations include a primary outcome event rate in the control group that was lower than that estimated for the sample size calculation. At the time of trial inception, we used the best available data (from the PITCH pilot study,11 which had a similar population of women with ICP, and the Cochrane systematic review in this area10) to estimate the event rate for perinatal death, preterm birth and neonatal unit admission. Our subsequent individual patient data analysis2 reported that 13.4% (412/3080) of women with ICP had spontaneous preterm birth, much lower than that reported in the Cochrane systematic review (43.8%; 39/89),10 highlighting the limitations of imprecise numbers to estimate our primary event rate. The trial primary outcome event rate was reviewed by the Data Monitoring Committee, but in the light of the lack of difference between the two groups noted, extending the trial was not considered by them to be necessary. Although it is possible that the trial has insufficient power to show a difference, the lack of effect in both the analysis for women who adhered to the intervention and in subgroup analyses in women at greatest risk of the adverse perinatal outcomes (serum bile acid concentrations ≥ 40 µmol/l at enrolment) suggest that this is unlikely. The difference shown in the study was much smaller (3.7%) than anticipated. It is possible and even likely that this difference may be too small to justify the use of UDCA in pregnancy, unless there was clear evidence that it reduces a robust and important endpoint (such as perinatal death).

In contrast to previous meta-analyses,9,10 we did not find that UDCA reduced maternal serum bile acid concentrations. Enzymic assays used to quantify total serum bile acids detect synthetic UDCA as well as cholic acid and chenodeoxycholic acid, the pathologically elevated bile acids in ICP that are implicated in the pathogenesis of fetal complications. 23 Treatment with UDCA reduces the proportion of cholic acid and chenodeoxycholic acid,23 but may be associated with some degree of increase in total serum bile acid concentration, due to elevated serum concentrations of the drug, making interpretation in clinical practice and in this trial more complex.

Sources of bias

We considered sources of possible bias for the trial. Selection bias is unlikely owing to the randomisation process including robust allocation sequence concealment. Performance bias was reduced by effective masking of the intervention to clinicians, women and data collectors, such that identification of the active treatment was minimal, with the two groups receiving the same antenatal and intrapartum care pathways. Assessment of outcome was also masked, minimising detection bias. Differences in attrition between the groups were minimal, with similar numbers discontinuing the intervention. We have aimed to avoid reporting bias, presenting all prespecified secondary outcomes, including the secondary analyses where effect size measures were calculated, and interpreting secondary outcomes with caution to avoid overinterpretation.

Interpretation

This trial suggests that there is no strong evidence base for routine use of UDCA in women with ICP for clinically useful amelioration of maternal symptoms or for reduction of adverse perinatal outcomes. The results support those from the pilot study,11 which reported only a small reduction in maternal itch symptoms, less than that judged by women and health-care professionals to be clinically useful. In the absence of any discernible significant effect in women with enrolment peak serum bile acid concentrations > 40 µmol/l, or in those presenting prior to 34 weeks’ gestation (on iatrogenic preterm birth) or on outcomes that are not related to clinician behaviour (i.e. iatrogenic preterm birth), a strong biological effect appears unlikely. It is possible that there are selected subgroups not currently identified that may benefit from treatment with UDCA.

As UDCA has been the only treatment consistently proposed in guidelines as a disease-modifying drug, there are no other treatments in current widespread use for prevention of the adverse perinatal outcomes associated with the disease, although a further study is planned evaluating rifampicin. 24 Our recent individual patient data analysis suggests that in women with peak serum bile acid concentrations < 100 µmol/l the risk of stillbirth is similar to that of the general pregnant population, whereas it is significantly higher in women with peak serum bile acid concentrations ≥ 100 µmol/l at any time in the pregnancy. The presence of coexisting pregnancy complications, such as pre-eclampsia or gestational diabetes, may add to the risk of stillbirth. 25 The only intervention that may have an impact on adverse perinatal outcomes is likely to be appropriately planned delivery.

Placing the research into wider context

Unanswered questions and future research

It remains uncertain whether or not there are as yet unidentified groups of women who may respond to UDCA treatment (such as those with or without comorbidities), for reduction of either maternal symptoms or adverse perinatal outcomes. Abnormally high serum bile acid concentrations are associated with stillbirth in ICP,2 and in vitro studies had shown that UDCA may be protective against bile acid-induced cardiac arrhythmias,27 potentially mediated through reduction of specific bile acid species. 23 However, further understanding of the pathophysiology underpinning stillbirth (and, therefore, the target for intervention) is needed. Additional work is also needed to confirm the likely pruritogen in ICP, with progesterone sulfates28 and lysophosphatidic acid29 proposed as candidates, to identify a target for therapeutic treatments to reduce troublesome symptoms of itch. Trials of other unlicensed treatments used for ICP, such as rifampicin, S-adenosylmethionine, dexamethasone, activated charcoal, guar gum and colestyramine, should be considered.