Re‐evaluating diagnostic thresholds for intrahepatic cholestasis of pregnancy: case–control and cohort study

To determine the optimal total serum bile acid (TSBA) threshold and sampling time for accurate intrahepatic cholestasis of pregnancy (ICP) diagnosis.

This article includes Author Insights, a video abstract available at https://vimeo.com/bjog/authorinsights16669 Objective To determine the optimal total serum bile acid (TSBA) threshold and sampling time for accurate intrahepatic cholestasis of pregnancy (ICP) diagnosis.
Design Case-control, retrospective cohort studies.
Setting Antenatal clinics, clinical research facilities.
Population Women with ICP or uncomplicated pregnancies.
Methods Serial TSBA measurements were performed pre-/postprandially in 42 women with ICP or uncomplicated pregnancy. Third-trimester non-fasting TSBA reference ranges were calculated from 561 women of black, south Asian and white ethnicity. Rates of adverse perinatal outcomes for women with ICP but peak non-fasting TSBA below the upper reference range limit were compared with those in healthy populations.
Main outcome measures Sensitivity and specificity of common TSBA thresholds for ICP diagnosis, using fasting and postprandial TSBA. Calculation of normal reference ranges of non-fasting TSBA.
Results Concentrations of TSBA increased markedly postprandially in all groups, with overlap between healthy pregnancy and mild ICP (TSBA <40 lmol/l). The specificity of ICP diagnosis was higher when fasting, but corresponded to <30% sensitivity for diagnosis of mild disease. Using TSBA ≥40 lmol/l to define severe ICP, fasting measurements identified 9% (1/11), whereas non-fasting measurements detected over 91% with severe ICP. The highest upper limit of the non-fasting TSBA reference range was 18.3 µmol/l (95% confidence interval: 15.0-35.6 lmol/l). A re-evaluation of published ICP meta-analysis data demonstrated no increase in spontaneous preterm birth or stillbirth in women with TSBA <19 µmol/l.

Introduction
Intrahepatic cholestasis of pregnancy (ICP) is defined by gestational pruritus and elevated total serum bile acids (TSBA). The incidence of ICP varies between 0.2 and 5.6% of pregnancies, 1,2 which may be accounted for by differences in ethnic populations 3 as well as the diagnostic criteria used. Diagnostic thresholds for ICP range from ≥6 to >15 lmol/l, depending upon hospital and national guidelines, [4][5][6] and there is currently no consensus as to whether TSBA should be measured fasting or postprandially, 7 although it is well documented that TSBA concentrations rise after food. 8,9 The reference interval for TSBA concentration has not been adequately established in pregnant women. Initial studies calculated the TSBA reference intervals based upon Gaussian data distribution, referencing mean AE 29 standard deviations. 10 However, analysis of TSBA levels from four studies during healthy pregnancy demonstrated that normal TSBA concentrations are not normally distributed, 11 with concentrations up to 16.7 lmol/l reported. 12 Intrahepatic cholestasis of pregnancy is associated with an increased risk of adverse pregnancy outcomes, including meconium-stained amniotic fluid, neonatal unit admission, spontaneous preterm birth and stillbirth. [13][14][15] Glantz et al. 15 demonstrated an increased risk of fetal complications in 96 women with fasting TSBA >40 lmol/l, describing this as 'severe ICP'. An individual patient data meta-analysis, comprising fasting or postprandial samples from 5557 women with ICP, determined that only women with peak TSBA levels ≥100 lmol/l have an increased risk of stillbirth; those with TSBA ≥40 lmol/l had an increased risk of spontaneous preterm birth. 14 No pharmacological treatment or monitoring strategy has been found to reduce stillbirth in ICP, 16 such that many women undergo iatrogenic preterm birth with the aim of reducing this risk. 4,14 However, preterm birth is associated with an increased risk for the neonate, including respiratory distress 17 and cognitive impairment in childhood. 18 As a result, correct diagnosis of the severity of ICP is necessary to enable clinicians to identify women with TSBA levels associated with increased risk of stillbirth to facilitate informed decision-making about timing of delivery.
To address concerns regarding accuracy and thresholds for ICP diagnosis, we performed three interlinked studies. In the first, we measured fasting and serial postprandial TSBA levels in response to standardised meals in women with ICP and uncomplicated pregnancies. In the second study, we determined reference intervals for postprandial TSBA from a separate cohort of women in their third trimester of uncomplicated pregnancy. In the third study, data from a published individual patient data meta-analysis 14 were used to compare rates of adverse perinatal outcomes for women with ICP but peak TSBA below the upper reference limit for TSBA identified in Study 2 with matched normal populations, to determine the effect of altering the diagnostic threshold for ICP.

Patient and public involvement
This study was designed in consultation with the charity ICP Support (www.icpsupport.org), whose input was incorporated into the original study protocol, application for ethical approval, patient information leaflets and consent forms. The protocol for Study 1 was modified following feedback from participants to ensure that the breakfast meal was acceptable, and recruitment included women identified through the charity social media and website. A video produced by participants was available for subsequent women, which provided them with information on the timeline for the study (https://m.youtube.com/watch?v=D9Bv8YIUlZI). Similar channels will be used to disseminate the findings, including question and answer sessions, available on the charity's YouTube and Facebook pages, and website. For a study that was quite onerous to the participants, having patient and public involvement from the beginning enabled us to understand how much could be asked of women, and revealed how motivated women affected by ICP, in particular, are to participate in research. Similarly, ongoing feedback received through the study has enabled us to modify protocols for future studies. In future, formal collation of this feedback would be beneficial, particularly where undertaken at multiple sites and by multiple researchers.
Three interlinked studies were performed with the aim of establishing the optimal TSBA threshold and blood sampling time for accurate diagnosis of intrahepatic cholestasis of pregnancy (ICP) (Figure 1).

Study 1: Standardised diet case-control study
Pregnant women between 16 +0 and 39 +2 weeks of gestation with singleton or multifetal pregnancies were recruited prospectively between 2012 and 2017 from Queen Charlotte's and Chelsea Hospital, St Thomas' Hospital and West Middlesex University Hospital, UK, and through the charity ICP Support. Four women with uncomplicated pregnancies, one woman with mild ICP and two women with severe ICP (defined as peak TSBA ≥40 µmol/l) were in their second trimester at the time of study; all other women were in their third trimester at the time of participation. Women with ICP were diagnosed according to each hospital's diagnostic criteria, which included unexplained pruritus and elevated random TSBA greater than the laboratory upper limit of normal, and exclusion of pre-existing liver disorders and pregnancy-specific causes of liver dysfunction such as HELLP (haemolysis, elevated liver enzymes and low platelets) syndrome, pre-eclampsia and acute fatty liver of pregnancy. All women with ICP were diagnosed before study participation, and none with uncomplicated pregnancies later developed ICP.
Serum samples for analysis of TSBA were taken from 42 women given a standardised diet from 18:00 the preceding day until 15:00 on the day of investigation, as previously described. 19 Participants were given fixed meals consisting of an evening meal the preceding day (50 g fat, 100 g carbohydrates, 1000 calories), breakfast (50 g fat, 75 g carbohydrates, 770 calories) and lunch (50 g fat, 100 g carbohydrates, 1000 calories). Blood samples were taken fasting and 3 hours post-lunch. A subset of 23 participants also underwent phlebotomy at 1 and 2 hours post-breakfast, immediately before lunch, and 20 minutes and 1 hour postlunch; sample sizes for each blood sampling time can be found in the Supplementary material (Table S1). Samples where technical difficulties occurred were omitted from the analysis. Participants remained sedentary throughout the study. No adverse events were recorded. Blood samples were analysed blind using the Total Bile Acids Assay Kit (Diazyme, Diazyme Laboratories Inc, Poway, CA, USA) in validated clinical laboratories of the respective hospitals. Participants' maximum TSBA concentration during the study day was used to group women with ICP by disease severity according to previously suggested thresholds: 13,15 seven had mild ICP (all TSBA measurements <40 lmol/l), 11 had 'severe' ICP (one or more TSBA measurements ≥40 lmol/l) and 24 participants had uncomplicated pregnancies. Data were plotted using GRAPHPAD PRISM (version 8.4.1; GraphPad Software Inc., San Diego, CA, USA).
The specificity and sensitivity of ICP diagnosis were calculated using a spread of commonly used diagnostic thresholds (≥6, ≥11 and >15 lmol/l), and ≥40 lmol/l as the threshold for severe ICP. Specificity and sensitivity were calculated for TSBA levels when fasting, and 20 minutes, 1 hour and 3 hours post-lunch.

Study 2: Reference interval serum samples
Non-fasting serum samples were retrospectively analysed from samples collected for a prospective observational cohort study for early prediction of pregnancy complications at King's College Hospital, London, UK. Women attending this visit, held at 30 +0 to 34 +6 weeks of gestation between 2011 and 2014 and at 35 +0 to 37 +6 weeks of gestation between 2014 and 2016, were invited to participate in the study, and serum samples were stored at À80°C from those who provided informed written consent. Maternal characteristics and medical history were recorded. Data on pregnancy outcome were obtained from the maternity computerised records or the general medical practitioners of the women.
Five hundred and sixty-one samples were randomly selected from our database of stored samples with uncomplicated pregnancies resulting in live birth after 38 weeks of gestation of phenotypically normal neonates with birthweight between the 10th and 90th centiles for gestational age. The downloaded database file was sorted at random, and cases were selected according to gestation at sampling to ensure a comparable distribution of samples across the gestational weeks of the third trimester of pregnancy. Samples were analysed retrospectively using Total Bile Acid reagent (Randox) run on a Siemens Advia 1800 by Affinity Biomarker Labs (London, UK). Samples were analysed blind. Five samples produced TSBA values below the limit of detection at <0.3 lmol/l; these data were included at an estimated value of 0.2 lmol/l to retain all data. No adverse events were recorded from performing the venepuncture.
Non-fasting TSBA reference intervals were calculated for the third trimester of pregnancy according to black (n = 160), south Asian (n = 160) or white (n = 241) ancestry, exceeding the minimum suggested 120 samples for determining reference intervals and confidence intervals according to the Clinical Laboratory and Standards Institute. 20 Analyses were performed in STATA software (version 15.1; StataCorp, College Station, TX, USA). TSBA values were log-transformed and differences between ancestral groups were compared using a one-way analysis of variance with post-hoc Bonferroni correction, with multiple linear regression with robust standard errors used to determine the size of the differences in non-fasting TSBA values. 21 Correlations between the log(TSBA) and maternal age, maternal body mass index or gestational age were assessed using Pearson's correlation coefficient. All TSBA values were included for reference interval calculation, according to the International Federation of Clinical Chemistry and Clinical and Laboratory Standards Institute C28-A3 recommendations. Calculation of the lower and upper limits of reference intervals was performed using the nonparametric method 22 with 95% CI.
Study 3: Comparison of adverse perinatal outcome rates for women with ICP with peak bile acids below a suggested diagnostic threshold with background population rates Rates of adverse perinatal outcomes for women with peak TSBA concentrations below or equal to the highest upper limit of the calculated reference interval were obtained from participants in a published individual patient data meta-analysis of women with ICP, 14 selecting women with non-fasted peak TSBA concentrations only. These were compared with published rates of adverse perinatal outcomes in uncomplicated pregnancies or at population levels, matching comparators as closely as possible, and compared using the chi-square test with Yates' correction and WOOLFLOGIT to calculate the 95% confidence interval (CI) in GRAPHPAD PRISM.
De-identified original and summary data, and the study protocol are available upon reasonable request of the authors. There are no core outcome sets relevant to this study.

Results
To assess the diurnal TSBA variation, women with uncomplicated pregnancies and women diagnosed with ICP were given standardised meals with known calorie and fat content, and TSBA was measured fasting and postprandially over the course of the day (Table 1). Women with ICP were separated into those with TSBA <40 lmol/l throughout the day (mild ICP), and those with at least one sample ≥40 lmol/l (severe ICP) (Figure 1).
In all participants, TSBA concentrations increased markedly after a meal, particularly in women with severe ICP (Figure 2A, and see Supplementary material, Table S1). We sought to determine the optimal time after starting the meal to measure TSBA in women with mild or severe ICP, calculating the sensitivity and specificity for fasting and postprandial time-points using different mild (≥6, ≥11 and >15 lmol/l) and severe (≥40 lmol/l) TSBA thresholds currently used 4-6,13 ( Figure 2B-D, and see Supplementary material, Table S2).
Fasting TSBA concentration was 100% specific for the diagnosis of ICP using thresholds ≥11 lmol/l and 91% specific at ≥6 lmol/l ( Figure 2B). For women with mild ICP, all fasting TSBA were <11 lmol/l (0% sensitive), and a threshold of ≥6 lmol/l gave only 29% sensitivity (2/7 women). Optimal sensitivity was observed 20 minutes post-lunch ( Figure 2C), but only a threshold of ≥6 lmol/l achieved >90% sensitivity, which corresponded to just 10% specificity for mild ICP. Hence, no time-point or currently used TSBA threshold achieved both good sensitivity and specificity for the diagnosis of mild ICP. When considering a diagnosis of severe ICP, the fasting TSBA concentrations were only 9% (1/11) sensitive, whereas all postprandial measurements were >90% sensitive ( Figure 2D) with 27% (3/11) of women with severe ICP having postprandial TSBA ≥100 µmol/l. As a result, the majority of women with severe ICP would not be diagnosed as having severe disease using a fasting sample.
Postprandial TSBA concentrations for women with uncomplicated pregnancies were greater for many women than currently used diagnostic thresholds for ICP (Figure 2A, and see Supplementary material, Table S1). We calculated reference intervals for TSBA in uncomplicated pregnancies using non-fasting samples (a sensible time for phlebotomy during pregnancy, when fasting before sampling is particularly unacceptable 24 ). We measured non-fasting TSBA concentrations from 561 women in the third trimester of uncomplicated pregnancies of black, south Asian or white ethnicity (Study 2, Figure 1, Table 2).
The distribution of TSBA concentrations was skewed for all ethnicities, which best transformed to an approximately Gaussian distribution using log transformation (see Supplementary material, Figure S1). Black women had non-fasting TSBA 25.8% higher (95% CI 9.6-44.4%, P = 0.001) than white women, and 24.3% higher (95% CI 5.7-46.1%, P = 0.008) than south Asian women. Non-fasting TSBA levels from south Asian women were similar to those of white women (1.22% higher, 95% CI À12.1 to 16.5%, P = 0.866) (see Supplementary material, Figure S2). No correlation was observed between TSBA and gestational age within the third trimester, maternal age or body mass index in any ethnic group (see Supplementary material, Table S3).   Due to the non-normal distribution of TSBA values, we determined the normal reference range for the different ancestral groups using the nonparametric method. 22 The lower reference limit for TSBA was between 0.7 and 1.0 lmol/l (95% CI 0.2-1.5 lmol/l) for all groups; however, the upper reference limit varied by ethnicity, at 10.4 lmol/l (95% CI 8.7-11.5 lmol/l), 15.5 lmol/l (95% CI 11.5-47.6 lmol/l) and 18.3 lmol/l (95% CI 15.0-35.6 lmol/l) for white, south Asian and black women, respectively ( Table 2).
Applying the highest upper TSBA reference limit of 18.3 lmol/l, two of seven women with mild ICP who participated in the standardised diet study (Study 1) did not reach this threshold (Figure 2A). Applying a TSBA threshold of ≤19 lmol/l, 0.9% (5/561) of women with uncomplicated pregnancy from Study 2 would have been falsely diagnosed, compared with 2.3% (13/561) of women using the >15 lmol/l non-fasting threshold (used by the South Australia Maternal and Neonatal Community of Practice 5 ) or 6.8% (38/561) using a ≥10 lmol/l non-fasting threshold (commonly used in hospitals in the UK 25 ).
We therefore considered the implications for using ≥19 lmol/l as the threshold for the diagnosis of ICP for women who had previously been diagnosed with ICP but whose non-fasting TSBA values remained <19 lmol/l during pregnancy. Using individual patient data previously collected for a meta-analysis of perinatal outcomes in ICP from an international cohort (Study 3, Figure 1), 14 we found that 18.5% (658/3509) of women would not have been diagnosed with ICP using this threshold. To determine whether this would alter the risk of adverse outcomes were this group of women not diagnosed as having ICP, we compared rates of adverse perinatal outcomes of these women with studies in women with uncomplicated pregnancies or at the population level, selecting studies that most closely represented the ICP cohort.
Consistent with previous reports, 14,15 women with ICP whose peak non-fasting TSBA was <19 lmol/l had no increased risk of stillbirth compared with the background population (and see Supplementary material, Table S4). Although these women with ICP had higher rates of preterm birth, this probably resulted from iatrogenic preterm birth as the rate of spontaneous preterm birth was lower than in the background population. It is also probable that the higher preterm birth rate in this cohort of women with ICP contributed to higher rates of neonatal unit admission and low Apgar scores. 26

Main findings
This study has revealed the diurnal variation in TSBA in both cholestatic and uncomplicated pregnancies, demonstrating the difficulties in determining clinically appropriate diagnostic thresholds to identify affected women with acceptable sensitivity and specificity. Considering that the adverse perinatal outcomes of ICP are associated with peak TSBA concentration, measuring non-fasting rather than fasting TSBA has greater clinical relevance for decisions about patient management according to severity of hypercholanaemia. Use of this time-point for diagnosis necessitates definition of the non-fasting normal range for TSBA in uncomplicated pregnancies. The women assessed did not receive a standardised diet; however, they represent a realworld cohort whose TSBA measurements are likely to represent typical levels based upon their normal diets. As a woman's typical TSBA concentrations are probably more relevant to the fetal bile acid exposure and stillbirth risk than TSBA results after a fixed meal stimulant (akin to the oral glucose load in gestational diabetes mellitus), we suggest that this non-fasting reference range is of greater clinical relevance.
Fasting TSBA concentrations of women with mild ICP from our standardised diet study were generally below the current diagnostic thresholds, and most women who had severe ICP (TSBA ≥40 lmol/l 13 ) would not have been diagnosed as such using fasting samples. The European Association for the Study of Liver threshold currently suggests that fasting TSBA ≥11 lmol/l is diagnostic. 6 However, this may miss correct risk stratification, as TSBA levels in women whose fasting TSBA level is <40 lmol/l can rise postprandially to ≥40 lmol/l, when risk of perinatal complications is increased, 13 or even higher to ≥100 lmol/l, when the risk of stillbirth is elevated. 14

Strengths and limitations
A limitation to this study is the relatively small numbers of women included with ICP for the standardised diet study, and that TSBA concentrations were not assessed at all time-points during the study in all women because of technical issues or patient consent. However, with the data stratified into mild and severe disease, differences between fasting and postprandial elevation between these groups were still evident.
Another limitation of the standardised diet study was that most women with severe ICP were treated with ursodeoxycholic acid (UDCA), whereas only one woman with mild ICP took the drug. It is possible that UDCA treatment influenced the TSBA concentration and caused the proportion of UDCA in the TSBA assay to increase, as was reported in a study where most women had TSBA <40 µmol/l. 27 However, a recent study of longitudinal changes in individual bile acids in ICP did not show a significant increase in TSBA concentration in UDCA-treated women. 28 A strength of this study is the large cohort of women analysed for non-fasting TSBA in the third trimester, which demonstrated the variation in the normal range for TSBA between different ethnic groups. Samples were obtained alongside routine clinical appointments, and so reflect a pragmatic time-point upon which to base real-world clinical decisions. Rates of adverse perinatal outcomes assessed using individual patient data provided for meta-analysis were selected from one of the largest international studies of women with ICP yet performed, and comparisons were made with the best matched data obtainable for each comparisonthe selection of which was performed before analysis in order to reduce potential bias.

Interpretation
Our re-evaluation of the reference limits for TSBA based on non-fasting samples holds implications for ICP diagnosis. We demonstrated no definitive evidence of markedly adverse spontaneous perinatal outcomes for women with ICP with non-fasting peak TSBA <19 µmol/l compared with uncomplicated pregnancies, with iatrogenic preterm birth probably responsible for the higher rates of overall preterm birth, neonatal unit admission and low Apgar score, 26 which could potentially be avoided by redefining the diagnostic threshold for ICP. Reassuringly, our findings are similar to previous studies, which reported that fetal complications increased in ICP when TSBA levels were >40 lmol/l, [13][14][15] and another study that did not find any differences in perinatal outcomes between women with ICP when stratified into groups with TSBA <13 lmol/l or between 13 and 40 lmol/l. 29

Conclusion
Based upon the results of this study, we suggest that a non-fasting TSBA value of ≥19 lmol/l is used as a clinical diagnostic threshold for ICP, as an alternative to fasting thresholds. It is important that pregnant woman with otherwise unexplained pruritus whose non-fasting TSBA levels are <19 lmol/l are monitored regularly for changes in TSBA, given that pruritus often predates hypercholanaemia. 30 Use of appropriate diagnostic levels is likely to reduce the number of women diagnosed with ICP, thereby reducing maternal anxiety, avoiding unnecessary antenatal consultations and potential complications associated with iatrogenic preterm birth. Furthermore, the use of postprandial TSBA measurements is more likely to identify women with TSBA ≥100 lmol/l and at increased risk of stillbirth, and will empower clinicians to individualise interventions aimed at reducing the risk of adverse outcomes in this small but important group of women.

Disclosure of interests
CW and CO report personal fees from Mirum Pharmaceuticals, and CW reports personal fees from GSK, outside the submitted work. PTS and CW report grants from King's Health Partners Institute of Women and Children's Health and Tommy's (Registered charity 1060508), and PTS reports grants from ARC South London (NIHR), during the conduct of the study. JRFW reports personal fees from GE Healthcare, grants from Intercept Pharmaceuticals, grants from Novartis Pharmaceuticals, outside the submitted work. Other authors declare no conflict of interests.

Contribution to authorship
CW conceptualised and oversaw the study. ALM, CO and CW wrote the paper. JRFW advised on the measurement times for the standardised diet study. CO, MM, TV, JG, CW and JC were involved with setting up and running the standardised diet study. MM, TV, HMF, CO, JG, JC and ALM processed serum samples from the standardised diet study. Regarding the samples for the reference range of TSBA in uncomplicated pregnancies, KN selected the appropriate samples, AS and KS performed the sample collection and data acquisition. PTS advised on appropriate data and statistical analysis for reference interval calculation and correlations. ALM analysed the TSBA results for the standardised diet, and the reference intervals. CO produced the metaanalysis comparison. All authors edited, revised and contributed to the intellectual content of the manuscript.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Figure S1. Power transformations of non-fasting total serum bile acid values. Figure S2. Total serum bile acid concentrations of women in their third trimester of uncomplicated pregnancy, separated according to ethnicity. Table S1. Total serum bile acid concentrations before and after standardised meals in women in the third trimester of uncomplicated pregnancy, or women diagnosed with intrahepatic cholestasis of pregnancy. Table S2. Sensitivity and specificity of intrahepatic cholestasis of pregnancy diagnosis using different diagnostic thresholds for fasting and postprandial samples. Table S3. Relationships between total serum bile acid values and maternal age, body mass index or gestational age. Table S4. Comparison of adverse perinatal outcomes in an intrahepatic cholestasis of pregnancy cohort with highest recorded total serum bile acids <19 µmol/l compared with uncomplicated pregnancies or background population.