Potential conflict of interest: Nothing to report.
Intrahepatic cholestasis of pregnancy (ICP) is characterized by pruritus, elevated bile acids, and, specifically, elevated disulphated progesterone metabolites. We aimed to study changes in these parameters during treatment with dexamethasone or ursodeoxycholic acid (UDCA) in 40 out of 130 women included in the Swedish ICP intervention trial (26 randomized to placebo or UDCA, 14 randomized to dexamethasone). Serum bile acid profiles and urinary steroid hormone metabolites were analyzed using isotope-dilution gas chromatography–mass spectrometry and electrospray–mass spectrometry. We found that all patients displayed ICP-typical serum bile acid profiles with >50% cholic acid at baseline but almost 80% UDCA upon treatment with this bile acid. In UDCA-treated patients, relative amounts of disulphated progesterone metabolites in urine decreased by 34%, 48% (P < 0.05), and 55% (P < 0.05) after 1, 2, and 3 weeks of treatment, respectively, which was significantly correlated to improvements of pruritus scores but not to serum bile acid levels. In contrast, in patients randomized to dexamethasone or placebo, no changes in steroid metabolites or pruritus scores were observed. Conclusion: UDCA treatment in ICP decreased urinary excretion of disulphated progesterone metabolites, suggesting that amelioration of pruritus is connected to stimulation of hepatobiliary excretion of progesterone disulphates. (HEPATOLOGY 2008.)
Intrahepatic cholestasis of pregnancy (ICP) is a liver disease of as yet undefined etiology and pathogenesis. ICP is characterized by pruritus and elevated serum bile acids (≥10 μmol/L) with onset in the second half of pregnancy and persisting until delivery.1 In the observational study of the Swedish ICP intervention trial,2 the prevalence of pruritus in pregnancy was 2.1%, and that of ICP was 1.5%. Fetal complication rates were related to maternal serum bile acid levels and increased when bile acid levels exceeded 40 μmol/L.2 In the double-blind, placebo-controlled intervention study of the Swedish ICP trial,3 effects of treatment with dexamethasone, ursodeoxycholic acid (UDCA), or placebo were investigated in 130 patients. UDCA but not dexamethasone significantly reduced alanine aminotransferase and bilirubin in the entire study group and improved pruritus and serum bile acid levels in women presenting with the severe form of ICP with bile acids ≥40 μmol/L.3
Compared with other liver diseases during pregnancy and normal pregnancies, women with ICP have a predominance of cholic acid (CA) among serum bile acids, and specifically, increased levels of steroid monosulphates and disulphates (predominantly progesterone metabolites) in serum and urine (reviewed by Reyes and Sjovall4). Meng and coworkers showed that treatment with UDCA not only lowered plasma bile acid levels but also the levels of sulphated progesterone metabolites, possibly by increasing their hepatobiliary excretion.5, 6 To validate these findings in the Swedish ICP intervention trial population, we analyzed the conjugation patterns of estrogen and progesterone metabolites in urine using electrospray–mass spectrometry (ESMS) and the serum bile acid composition using isotope-dilution gas chromatography–mass spectrometry (GCMS).
CA, cholic acid; CDCA, chenodeoxycholic acid; DCA, deoxycholic acid; ESMS, electrospray–mass spectrometry; GCMS, gas chromatography–mass spectrometry; GlcA, glucuronide; GlcNAc, N-acetylglucosamine; ICP, intrahepatic cholestasis of pregnancy; UDCA, ursodeoxycholic acid.
Patients and Methods
Samples of serum (2-5 mL) and urine (50-100 mL) for bile acid and steroid analysis were collected in a fasting state from 40 out of the 130 women included in the Swedish ICP intervention trial.3 These 40 patients comprised all patients enrolled at Sahlgrenska University Hospital/East in Gothenburg, which was by far the largest center of the Swedish ICP study. Pruritus was estimated on a visual analogue scale with the endpoints “no pruritus at all” at 0 mm and “worst possible pruritus” at 100 mm. Twenty-six women had been randomized to placebo or UDCA, and 14 had been randomized to dexamethasone. A maximum of 4 samples were available from each patient. The first sample was obtained at randomization, and the next 3 were taken weekly during the study period, in which patients were given either UDCA (1 g/day as a single dose for 3 weeks), dexamethasone (12 mg/day as a single dose for 1 week and placebo during the second and third weeks), or placebo for 3 weeks. Four samples were available from 7 women randomized to placebo or UDCA and from 9 women randomized to dexamethasone, respectively. Not all samples were available for analysis from the remaining patients, because these women gave birth during the study period. All samples were frozen immediately and stored at −20°C until analyzed.
The study was approved by the Swedish Medical Products Agency and the Ethics Committee of the Faculty of Medicine at the University of Gothenburg. All patients gave written informed consent according to the Declaration of Helsinki.
Analysis of Bile Acids and Steroids.
Deuterium-labeled bile acids were obtained from QMX Laboratories Ltd. (Thaxted, U.K.). The sources of other materials were the same as previously described.7 Total serum bile acids were analyzed with an enzymatic, colorimetric method.3 Relative amounts of individual serum bile acids were analyzed via isotope-dilution GCMS as described by Ewerth et al.8 Serum samples (n = 135) were added with known amounts of 2,2,4,4-D2–labeled CA, deoxycholic acid (DCA), chenodeoxycholic acid (CDCA), and UDCA, hydrolyzed with cholylglycine-hydrolase, extracted with octadecylsilane-bonded silanol, purified via anion exchange–chromatography, and analyzed on a type 6890N GCMS system equipped with a type 5975 mass selective detector (Agilent Technologies, Inc., Santa Clara, CA). Urine samples (n = 135) were extracted with octadecylsilane-bonded silanol, blow-dried under a stream of nitrogen, and analyzed via ESMS on Micromass Quattro 1 and Quattro Micro (Micromass, Manchester, U.K.) mass spectrometers. Samples in 70% aqueous ethanol were injected into the ion source in a stream of the same solvents at a flow of 10 μL/minute. The settings were as follows: capillary, 4.2 kV; extractor, 9 V; cone voltage, 30-35 V; RF lens, 0.2 V. Spectra of negative ions (deprotonated molecules) were recorded in the range m/z 200-800, which includes ions derived from doubly charged sulphated progesterone metabolites and bile acid conjugates. Changes in the relative excretion rates of estrogen and progesterone metabolites were semiquantitatively estimated from individual ion peak intensities in relation to the ion intensity of pregnanediol glucuronide at m/z 495.3 (100%).
In 5 women randomized to dexamethasone or UDCA, and in 4 women randomized to placebo, a comprehensive work-up of all serum and urine samples, including separation into different groups of bile acid and steroid hormone conjugates on the lipophilic anion-exchanger Lipidex-DEAP and characterization by ESMS and GCMS,7 was performed in order to rule out significant amounts of serum bile acids differing from those estimated by isotope-dilution GCMS, and to verify the validity of the identification and interpretation of ions observed in ESMS analysis of urine.
Comparison of serum bile acid peaks recorded via conventional GCMS and isotope-dilution GCMS did not reveal any significant peak (>0.05 μmol/L) other than CA, DCA, CDCA, and UDCA, except isoUDCA in women allocated to UDCA treatment. IsoUDCA is a major metabolite of orally administered UDCA and represented approximately 5% of all serum bile acids during UDCA treatment, as previously shown in healthy humans and in patients with other cholestatic liver diseases.7, 9 Group separation of urinary steroid conjugates confirmed that the ion at m/z 495.3 was exclusively derived from pregnanediol glucuronides and not from single-charged pregnanetriol disulphates that have the same molecular weight (data not shown).
Means, standard deviations, medians, and 25%-75% interquartile ranges were calculated for descriptive purposes. After testing for normal distribution, individual bile acids and relative ion intensities were evaluated by paired, 2-tailed Student t test. Total serum bile acids and pruritus scores were not normally distributed and thus were first analyzed for trend changes over time by Friedman test and then for differences during treatment, compared with randomization, via Wilcoxon signed rank test. Correlations between changes in serum bile acids, pruritus scores, and relative metabolite ion intensities were calculated by linear regression analysis. All calculations were performed with SAS/STAT software (SAS Institute Inc., Cary, NC).
Women randomized to placebo (n = 13), dexamethasone (n = 14), or UDCA (n = 13) did not differ in maternal age (28.5 ± 3.3, 27.2 ± 2.6, and 27.5 ± 6.3 years, respectively), gestational age (33.8 ± 1.7, 32.5 ± 2.8, and 33.5 ± 2.9 weeks, respectively), alanine aminotransferase levels (186 ± 210, 180 ±162, 222 ± 96 U/L, respectively), and bilirubin levels (13.4 ± 11.7, 14.4 ± 12.5, and 15.2 ± 7.2 μmol/L, respectively). There was 1 twin pregnancy in the placebo group and 1 in the UDCA group. These data, in addition to pruritus scores and serum bile acid levels at inclusion (Table 1), did not differ significantly from those of the total number of 130 patients included in the intervention trial of the Swedish ICP study.3 Therefore, this single-center group of 40 patients was considered representative of the entire study population.
Table 1. Pruritus Scores and Bile Acid Profiles
Pruritus Scores, Visual Analogue Scale (mm)
Serum Bile Acids
Relative Amounts, % (Mean ± Standard Deviation)
25%–75% Interquartile Range
25%–75% Interquartile Range
P < 0.05 versus week 0 (Wilcoxon signed rank test).
P < 0.01–0.001 versus week 0 (paired, 2-tailed Student t test).
Table 1 summarizes pruritus score data and serum bile acid levels. Pruritus scores at randomization did not differ between patients treated with dexamethasone, UDCA, or placebo. There was no significant reduction in pruritus scores during treatment with dexamethasone, whereas UDCA significantly improved pruritus during the 3-week treatment period.
Serum Bile Acid Profiles.
At randomization, patients treated with dexamethasone tended to have lower serum bile acid levels (median, 19 μmol/L) compared with patients randomized to UDCA (median, 33 μmol/L). The 7 women randomized to UDCA that finished the study per protocol also presented with higher bile acid levels (mean ± standard deviation, 51.3 ± 29.8 μmol/L) than the entire UDCA group from the Swedish ICP intervention trial3 [39.7 ± 47.7 μmol/L (n = 47)]. However, these differences were not statistically significant.
Bile acid profiles in all patients showed a pattern typical for ICP: an approximately 50% predominance of CA (Table 1). This pattern did not change in patients randomized to placebo nor in patients randomized to dexamethasone despite a significant reduction of total bile acids in this treatment group after 2 and 3 weeks. No correlation (r < 0.5) could be detected between the severity of ICP as measured by total serum bile acids and relative amounts of CA. In patients randomized to UDCA, total serum bile acid levels were significantly reduced after 2 weeks of treatment. UDCA became enriched in serum by almost 80%, whereas the primary bile acids CDCA and CA were significantly reduced. These changes already occurred after the first week of treatment and remained at the same levels during the entire treatment period (Table 1).
Steroid Hormone Metabolites in Urine.
Screening of urine samples for progesterone and bile acid metabolites via ESMS revealed essentially the same pattern in all women at enrollment, with no significant differences between treatment groups. The spectra were dominated by an anion at m/z 239.1, representing doubly charged disulphates of pregnanediols. Ions at m/z 247.1, 493.3, 495.3, and 509.4 were the next most common, representing doubly charged pregnanetriols and deprotonated glucuronides of pregnanolones, pregnanediols, and pregnanetriols, respectively. Ions indicating monosulphates of pregnanolones, pregnanediols, and pregnanetriols were recorded at m/z 399.3, 413.3, and 415.4, respectively, and double conjugates of pregnanediol with sulphate and N-acetylglucosamine (GlcNAc) and with glucuronide (GlcA) and GlcNAc10 were recorded at m/z 602.5 and 698.4, respectively. The ions at m/z 463.4 and 479.3 represented estriol and estetrol glucuronides, respectively (Fig. 1A). Typical signs of cholestasis were anions at m/z 263.7 and 288.7, indicating doubly charged ions of sulphated glycine-amidated or taurine-amidated dihydroxy bile acids (Fig. 1A). In all samples from patients randomized to UDCA, anions at m/z 528.4, 624.4, and 651.6 emerged during treatment (Fig. 1B) indicating sulphate, GlcA, and GlcNAc conjugates of glycine-amidated UDCA.11 No specific ion indicating treatment with dexamethasone was detected.
By comparing relative intensities of steroid hormone metabolite ions, no changes were observed for any compound in patients randomized to placebo or dexamethasone. However, in the group of patients randomized to UDCA, the relative intensities of the ion at m/z 239.1 decreased significantly (P < 0.05) after 2 (n = 8) and 3 weeks (n = 7) of treatment, whereas the relative intensities of other ions did not change. Figure 2 presents the data for relative ion intensities for patients randomized to placebo (n = 7), dexamethasone (n = 9), and UDCA (n = 7) who finished the study per protocol.
Pruritus Scores Correlate with Urinary Pregnanediol Disulphate but not Serum Bile Acid Levels.
A linear regression analysis of possible associations between changes in serum bile acid levels and pruritus scores did not reveal significant correlations in any treatment group. A linear regression analysis for urinary pregnanediol disulphate levels at time point 0 versus pruritus scores did not reveal any statistically significant correlation for the entire study population (P = 0.29) or the UDCA group (P = 0.27). However, in patients randomized to UDCA, improvements of pruritus scores and reductions in urinary pregnanediol disulphate excretions correlated with each other significantly [r = 0.48, 1.0, and 0.60 after 3 weeks of treatment, respectively; P < 0.05 for trend over all time points (each patient was tested 4 times)] (Fig. 3). There was no significant correlation between pruritus scores and individual CDCA and CA levels in UDCA-treated patients. Additionally, reduction in total serum bile acid and urinary pregnanediol disulphate levels did not correlate with each other.
Over time, various combinations of serum bile acids, aminotransferases and bilirubin have been used in the diagnosis of ICP,1 making the comparison of results from different studies difficult.12 Fasting serum bile acids ≥10 μmol/L is the most commonly used variable for diagnosing ICP1 and was the diagnostic criterion in the Swedish ICP trial.2 Postprandial elevated bile acids could have resulted in incorrect inclusion. However, in the present study, we show that the diagnosis of ICP was correct in all cases, because all women presenting with enzymatically, colorimetrically measured serum bile acids ≥10 μmol/L had the ICP-typical serum bile acid composition, consisting of approximately 50% CA, whereas in normal pregnancies, the ratio of CA/CDCA/DCA is approximately 1.1:1.6:1.0.4 We, however, did not observe extreme ratios, such as the 32:8:1 ratio observed in women with ICP in a Chilean study.5, 6 Instead, the relative amounts of CA, DCA, and CDCA in our trial at inclusion were very similar to those recently described in the Lithuanian ICP study13 in which 42 patients were allocated to treatment with either cholestyramine or UDCA. In the latter group, an enrichment of serum UDCA of about 50% was found after 2 weeks, which is lower than the 78.5 ± 7.4% enrichment that we observed after this duration of treatment of the patients in our study. This discrepancy is most likely explained by the higher dose of UDCA administered in our study. Our patients were given 1 g/day of UDCA irrespective of body weight, while patients in the Lithuanian study were treated with a dosage of 8-10 mg/kg/day (i.e., 560-700 mg/day).13
Elevated serum bile acids, and high relative amounts of CA, are typical diagnostic criteria for ICP; however, these parameters are not specific for ICP for which the etiology is still unknown. Rather, compared with normal pregnancies, ICP is biochemically characterized by altered profiles of sulphated progesterone metabolites in serum and urine,4 in particular increased plasma levels of 5α-pregnane-3α,20α-diol and 5β-pregnane-3α,20α-diol disulphates and increased urinary excretion of sulphated metabolites with a 3α,5α/β configuration. These changes may precede the clinical appearance of cholestasis and the rise of serum bile acid levels.14 The importance of progesterone metabolites is further supported by the finding that 34 out of 50 consecutive patients with ICP studied by Bacq et al.15 had been treated with oral progesterone. It should be noted that no statistically significant correlations were found between aberrances of steroid sulphate profiles, increase of serum bile acid concentrations, and clinical cholestasis.5, 6, 15 These findings suggest that the changes in steroid sulphate profiles and levels in plasma are not secondary to cholestasis in general or a defect in bile acid–dependent bile excretion.4
We have now identified a significant correlation between the improvement of the most important clinical feature of ICP, i.e., pruritus and the diminished urinary excretion of ICP-specific progesterone metabolites during UDCA treatment. This result is based on 4 repeated measurements in the each patient with clearly defined inclusion criteria and is strengthened by a double-blind, placebo-controlled design in comparison to treatment with dexamethasone.
The changes in progesterone metabolite patterns in ICP may be due either to a change in the metabolic pathways of progesterone or the biliary excretion of specific isomers and/or conjugates of the metabolites. A hypothetical change of metabolic pathways could affect either the reductive pathways of progesterone metabolism or the pathways of conjugation. Both processes could result in changed proportions of conjugated isomeric pregnanolones and pregnanediols in plasma. The disulphate of 5α-pregnane-3α,20α-diol is a major steroid in the bile of pregnant women. UDCA or its major metabolite in humans, isoUDCA,7, 9 may interfere with the metabolism of progesterone by their affinity for human liver 3α-16 or 3β-hydroxysteroid dehydrogenases17 so that a lesser amount of steroids with a 3α-hydroxy-5α(H) structure are formed. This is unlikely, however, because the amounts of glucuronidated pregnanediols are virtually the same with and without UDCA treatment (Meng and colleagues5, 6 and this study). A possible negative effect of UDCA on the conjugation of pregnanediols with sulphate is also unlikely, because UDCA does not seem to affect messenger RNA expression of SULT2A1 in human liver.18 Finally, an enhanced sulphate deconjugation during UDCA treatment can be excluded from in vivo studies performed in healthy pregnant women using steroid sulphates labeled with stable isotopes. These studies showed that sulphated 3α/β-hydroxysteroids cannot be isomerized and that they are irreversibly sulphated.19, 20 Thus, it appears unlikely that UDCA changes the pattern of sulphated progesterone metabolites toward normal and ameliorates cholestasis (Meng and colleagues5, 6 and this study) due to an effect on the metabolic pathways of progesterone.
Rather, the changes in progesterone metabolite patterns in ICP are most likely due to impaired hepatobiliary excretion of specific isomeric metabolites, which is improved by oral administration of UDCA. Final evidence for this assumption would require analysis of bile samples; however, it is considered unethical to perform this type of analysis in pregnant women. Blood samples were also limited, so we were unable to analyze progesterone metabolite patterns in serum; however, changes in serum progesterone metabolite patterns are highly likely to be the same as seen in urine.6
The beneficial effect of oral UDCA favors the hypothesis of a specific abnormality in the hepatobiliary secretion of sulphated progesterone metabolites. UDCA reduces the increased urinary excretion of steroid sulphates, an indirect sign of an increased hepatobiliary excretion of these compounds. The nature of these transport proteins remains to be determined. We recently showed that both UDCA and rifampicin increase membrane transport protein expression in human liver; UDCA increases bile salt export pump (ABCB11), MDR3 (ABCB4), and MRP4 (ABCC4) expression and rifampicin MRP2 (ABCC2) expression.18 We therefore speculate that the effect of UDCA is indirect, possibly by facilitating bile acid export via the bile salt export pump. In addition, induction of MDR3 by UDCA may be of importance in the minority of ICP patients, for which we and others have defined ICP-related mutations or haplotypes of the ABCB4 gene.21–23 In agreement with the familial background of ICP, it is tempting to speculate about genetic polymorphisms that affect the structure and/or regulation of proteins that transport the large amounts (several 100 mg)4 of sulphated progesterone metabolites that are produced during pregnancy into bile.
UDCA is commonly used for the treatment of cholestatic liver diseases. In primary biliary cirrhosis, it has become the standard treatment. Combined analysis of the 3 largest randomized clinical trials revealed that this treatment increased survival.24 However, pruritus, the most annoying symptom in primary biliary cirrhosis or primary sclerosing cholangitis,25, 26 is not improved; therefore, UDCA is not recommended for the treatment of pruritus in most cholestatic liver diseases where rifampicin and opioids are effective.27 ICP appears to be the only exception, because it is a cholestatic liver disease in which UDCA seems to be more effective in alleviating pruritus.27, 28 The effect of UDCA on pruritus in ICP was recently confirmed to be better than that of cholestyramine in the Lithuanian study,13 which might indicate that reduction of serum bile acids alone is not the only major mechanism behind improvement of pruritus in ICP. We assume that pruritus, the most typical symptom of ICP, results from decreased hepatobiliary secretion of pruritogens transported by the same proteins as the sulphated progesterone metabolites—as long as these metabolites are not pruritogens themselves, a notion that remains to be investigated.
In a previous study, we reported a significant improvement of pruritus in UDCA-treated patients presenting with a severe form of ICP only (total serum bile acids ≥40 μmol/L).3 For the entire study population, the apparent improvement of pruritus by UDCA (cf. Fig. 1D in Glantz et al.3) did not reach statistical significance. The disparate outcome for UDCA treatment on pruritus in the group of 40 patients described in the present study as compared with the larger group of 130 patients in Glantz et al.3 can be at least partially explained by higher—though not significantly different statistically—serum bile acid levels at inclusion. One might also speculate whether the degree of compliance was different between our most carefully followed single-center cohort and the other patients. However, blood samples from the latter were not available for UDCA enrichment analysis.
In conclusion, we found that UDCA decreased urinary progesterone disulphate excretion in women with ICP, a decrease that parallels improvement of pruritus. Enhanced hepatobiliary excretion of progesterone disulphates, presumably due to induction of membrane transport proteins by UDCA, is the most likely explanation.
We wish to thank Professor Emeritus Jan Sjövall for the helpful discussions and suggestions. We also extend special thanks to the study nurse, Ann Christiansson.