Oral herbal medicine for women with intrahepatic cholestasis in pregnancy: a systematic review of randomized controlled trials

Background Intrahepatic cholestasis of pregnancy (ICP) is a pregnancy complication whose range has been calculated to be between 0.01 and 15.6% all around the world. We wanted to systematically evaluate the effect and safety of oral herbal medicine on treatment for ICP. Methods Details of the methods could be found in the registered protocol on PROSPERO (CRD42018096013). Trials assessing the effectiveness of herbal medicine for ICP were searched from seven electronic databases from inception to 28th February 2020. RevMan 5.3 software was used to perform all statistical analysis. Meta-analysis, additional analysis, Trial Sequential Analysis (TSA) and Grading of Recommendations Assessment, Development and Evaluation (GRADE) were conducted if data permitted. Results Totally 43 randomized controlled trials with 3556 patients were included. Meta-analysis showed potential good adjunctive effect of herbal medicine on decreasing the pruritus scores (MD -0.58, 95% CI − 0.79 to − 0.36), the serum TBA scores (MD − 3.99 μmol/L, 95% CI − 4.24 to − 3.74) on the basis with Ursodesoxycholic acid. Compared to the medicine alone, significantly lower incidence of fetal distress (RR 0.41, 95% CI 0.32 to 0.51), asphyxia neonatorum (RR 0.35, 95%CI 0.25 to 0.49), cesarean section (RR 0.73, 95% CI 0.63 to 0.85), postpartum hemorrhage (RR 0.45, 95% CI 0.28 to 0.72) were observed in the combination group. But the comparison between herbal medicine and medicine showed inconsistent results among trials. Insufficient information could be used to evaluate the safety of herbal medicine for ICP. Conclusion This review found the current evidence may support the effectiveness of combination of herbal medicine and conventional medicine for decreasing the maternal pruritus scores, the serum TBA, and the number of fetal distress, or asphyxia neonatorum events related to this condition (which was supported by TSA results). Since there were obvious statistical and clinical heterogeneity among trials, and the methodological quality of the included studies was poor, the level of the evidence could only be defined as “very low” according to the GRADE criteria. Further high quality studies are still needed to testify the effectiveness and safety of herbal medicine for ICP.


Background
Intrahepatic cholestasis of pregnancy (ICP) is a pregnancy complication with prevalence ranging between 0.01 and 15.6% [1][2][3][4]. ICP presents as persistent pruritus, typically on the skin of palms and soles, with elevated bile acid levels, and resolves with delivery [5]. Generally, the onset of ICP occurs in the second and third trimesters and disappears spontaneously after delivery, usually within weeks [5][6][7]. The clinical importance of ICP lies in the potential fetal risks, including spontaneous preterm birth, iatrogenic preterm birth and fetal death [8]. While the pathophysiology of ICP is still poorly understood, gene, hormone, and environment play roles. Bile acids need to enter hepatocytes or bile ducts through the action of some transporters which genetic mutations can cause cholestasis [9,10]. Estrogen can decrease the expression of ABCB11 / BSEP gene to inhibit the function of bile salt delivery pump, or decrease the activity of Na+ / K+ ATPase to inhibit the uptake of bile acid by hepatocytes that eventually leads to intrahepatic cholestasis. Sulfated progesterone metabolites can also have an adverse effect on bile acid excretion by inhibiting the function of BSEP [11]. Although the mechanism is unknown, studies have shown that the plasma and serum selenium concentrations and glutathione peroxidase activities in ICP patients are lower than those in healthy pregnant women, and the incidence rate is higher in winter. Some environmental factors such as pesticide pollutants, erucic acid in rape and selenium deficiency in food may lead to ICP [12]. Most often the disease affects women over the age of 35 years, [13] with personal history of cholestasis associated with the use of oral contraceptives, personal or family history of cholestasis of pregnancy, [14] biliary disease, [1] or liver disease, in multiple gestation pregnancy, [15] or in vitro fertilisation pregnancies [16]. Likewise, seasonal variations, [17] low selenium intake, erucic acid, increased gut absorption of bacterial endotoxins, pollutants, infections, and medicine are factors suspected as causing the disease [15][16][17][18][19][20].
When treating the ICP, doctors always focused on reducing maternal symptoms, improving results of liver tests, and reducing total bile acid (TBA) concentration. They commonly used Ursodeoxycholic acid (UDCA), Sadenosylmethionine (SAMe), dexamethasone, or cholestyramine as well as vitamin K (preventing postpartum bleeding) as therapies [20]. Whlie there was insufficient evidence to recommend early-term delivery or to support therapies above according to one Cochrane review [21]. However, the review found that UDCA seemed to improve the maternal symptom of pruritus, which agreed with the result of a meta-analysis by Bacq [22]. Bacq strongly suggested that UDCA was also beneficial for the fetal outcome, including total prematurity, fetal distress and neonatal respiratory distress syndrome; however, the Cochrane Review did not agree [21].
With the possible underlying mechanism of herbal medicinal for this condition, we conducted this systematic review to explore the potential effectiveness and safety of oral administration of herbal medicine in treating ICP.

Eligibility criteria
Published and unpublished studies, inany language, were included where thefollowing PICOS (patient, intervention, comparator, outcome, study type) criteria were met: i) Patient: Patients, of any age, with intrahepatic cholestasis in pregnancy. Intrahepatic cholestasis in pregnancy (ICP or OC), diagnosed in accordance with recognized criteria (e.g. Guidelines for diagnosis and treatment of intrahepatic cholestasis in pregnancy (2015) by Chinese Medical Association). ii) Intervention: Herbal medicine through oral administration. iii) Comparators: Controls include placebo and conventional medicine (such as UDCA, SAMe, dexamethasone, vitamin K etc.). Chinese medicine combined with conventional medicine compared to the conventional medicine alone are also included. The controls in protocol were selected as Guidelines for diagnosis and treatment of intrahepatic cholestasis in pregnancy (2015) by Chinese Medical Association. After analyzing the data, most of the controls were not reported in the study, so there are some differences. iv) Outcomes: Primary outcomes included the changes in Ribalta score, maternal serum TBA values, and incidence of adverse birth events (e.g. newborn deaths, fetal distress and suffocation). Secondary outcomes included changes in bile acid content, incidence of cesarean sections, postpartum hemorrhage, adverse effects of medicine, changes in ALT values, and changes in AST values for the maternal, as well as the averages gestational age at birth and the incidence of premature births. v) Study type: Randomized controlled trials.
Literatures that unable to obtain the analysable data, as well as the piratical documents would be excluded.

Data extraction (selection and coding)
Trials retrieved using the search strategy and those from additional sources were screened independently by two review authors to identify trials that potentially meet the inclusion criteria outlined above. The full texts of these trials were retrieved and independently assessed for eligibility by other two reviewers. Any disagreement was resolved through discussion with a third reviewer (Huijuan Cao). A predesigned form was used to extract data from the included trials for assessment of trial quality.
Extracted information included: i) General information: Including document number, title, first author, year(s) conducted, location (city, country), source, etc.; ii) Methodological related information: type of design, grouping number, random allocation method, Random concealment method, method of blinding, participants blinded, loss of follow up, report of selective outcome, calculation of sample size, baseline comparability; iii) Participants information: diagnostic criteria, inclusion criteria, exclusion criteria, source, sample size, age, gender, disease course. iv) Intervention information: Types of interventions, intervention performer, treatment duration; v) Outcome measures: Treatment outcomes: changes in Ribalta score, changes in TBA values, number of neonatal deaths, number of fetal distress or asphyxiation, changes in ALT values, changes in AST values, number of caesarean sections, postpartum hemorrhage, adverse medicine effects, average gestational age at birth, number of premature deliveries.

Risk of bias assessment
We assessed the methodological quality of the included trials using the risk of bias tool recommended by the Cochrane Collaboration (Higgins and Green, 2009). Seven elements were assessed: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data (according to record the missing data and the method to deal with it), selective reporting (determined by the consistency of the predefined and reported outcomes) and other bias (assessed according to sample size calculation, inclusion/exclusion criteria for participant recruitment, comparability of baseline data, funding sources).

Strategy for data analysis
All statistical analyses were performed using RevMan 5.3 (The Cochrane Collaboration) software. The results of each single trial were described. We presented results as risk ratio (RR) with its 95% confidence intervals (CI) for dichotomous outcomes, and mean difference (MD) with 95% CI for continuous outcomes. Statistical heterogeneity between the included trials were assessed using the I 2 value, and a meta-analysis were conducted if it had proved to be no significant clinical (relating to the participants, interventions, controls, and outcomes) and statistical heterogeneity (I 2 values are less than 75%) among the included trials. If the I 2 value was less than 25%, a fixed-effect model (FEM) was used to synthesize the data, and if the I 2 value was between 25 and 75%, we explored the sources of the heterogeneity. If the results of sensitive analysis or subgroup analysis explained the statistical heterogeneity successfully (i.e., I 2 value is less than 25%), we used the FEM to synthesize the data as well, otherwise, random effect model (REM) was used to combine the data. Data would not be synthesized if there was a significant level of statistical heterogeneity (i.e., I 2 value is greater than 75%) which was could not be explain or to handle (by subgroup analysis).

Additional analysis
If one meta-analysis included more than 10 trials, we explored the possibility of publication bias by using a funnel plot. If one meta-analysis contained more than 8 included studies, we performed a Trial Sequential Analysis (TSA) was performed if there were more than 8 included studies in the meta-analysis. We calculated the required sample size of each meta-analysis and tested the robustness of the result with the TSA versioned 0.9.5.10 (Copenhagen: The Copenhagen Trial Unit, Center for Clinical Intervention Research, 2017). The information size required by the diversity-adjustment that we used was estimated from a control event proportion of the included trials and a priori intervention effect of 5%, and the diversity we used was estimated in the included trials.

Evidence quality assessment
The Grading of Recommendations Assessment, Development and Evaluation criteria (GRADE) was conducted to assess the certainty of the evidence for each primary outcomes with meta-analysis. Considering the following aspects, such as study design, risk of bias, outcome consistency of trials, directness and accuracy of evidence and possibility of publication bias, we judged whether to downgrade the evidence. Finally, the evidence was rated at four levels: high, moderate, low or very low.

Selection results
A total of 1604 trials were retrieved according to the search strategy, and the literature was screened according to the predefined criteria. After reading the title and abstract, 1262 trials were screened out, and further another 187 trials were excluded after reading the full text. Forty-three randomized controlled trials were finally included. All the included trials were published in Chinese. The specific literature screening process is shown in Fig. 1.
Basic characteristics of trials (see Table 1) A total of 3556 women with ICP were included in the 43 trials . The sample size ranged from 30 to 188 (an average of 41 patients in each group). The age distribution ranged from 18 to 42 years old, and the range of gestational age was 18-40 weeks. All cases were recruited from the outpatient/inpatient Department of Obstetrics and Gynaecology. Thirty-eight trials reported baseline comparisons in terms of demographic characteristics, and the other five trials only mentioned randomization, which did not explain whether the baseline characteristic of the participants was comparable.

Quality evaluation of inclusion studies
None studies used the appropriate random number table method for sequence generation [32,34,38,49,57,60,68,69,71], one study used coinage method to generate random numbers [58]. It is unclear whether the remaining studies had used adequate randomization allocation methods. None of the 43 studies reported the allocation concealment or blinding of participants or investigators. One study reported and explained the dropouts and losses to follow-up [40], and none of the others reported dropouts. Two study has selective reporting outcomes [34,51], because one of them intended to use liver function data as outcome measures but was not reported in result and another one reported the number of adverse birth events but was not reported in protocol. All studies did not report sample size estimates and financial support, also the baseline comparability was unclear. Thus, all the included trials were considered to be at high risk of other bias. So all the other bias risks of the 43 studies are unclear. In summary, all of the included trials were assessed as having high risk of bias due to the undefined methods of randomization, absence of blinding and the potential inappropriate method on dealing with missing data (See Fig. 2).
Four studies reported the number of deliveries by cesarean section [31,[33][34][35], the results showed that there was no difference between herbal medicine and conventional medicine (RR 0.97, 95% CI 0.80 to 1.18, I 2 = 0%, P = 0.78, 4 trials, 338 patients). One trial reported the number of postpartum hemorrhage [35] and the results showed that there was no significant difference in this outcome between herbal medicine and conventional medicine (RR 0.47, 95% CI 0.09 to 2.36, P = 0.36, 1 trial, 60 patients). One trial reported changes in maternal ALT (unknown as whole blood or serum) and the results showed that there was no significant difference in this outcome between herbal medicine and conventional medicine (MD -12.80 U/L, 95% CI − 25.81 to 0.21, P = 0.05, 1 trial, 98 patients) [31]. Four studies reported changes in maternal serum ALT and the results       showed that there was no significant difference in this outcome between herbal medicine and conventional medicine (MD 0.90, 95% CI − 5.10 to 6.90, I 2 = 67%, P = 0.77, 4 trials, 300 patients) [32][33][34][35]. Two studies reported changes in maternal serum AST and the results showed that there was no significant difference in this outcome between herbal medicine and conventional medicine (MD 11.20 U/L, 95% CI 7.86 to 14.54, P < 0.00001, 1 trial, 120 patients) [33,35]. Three studies reported the gestational age of the newborn (MD and 95% CI were 1.80 weeks, 1.04 to 2.56; − 0.50 weeks, − 1.31 to 0.31; 0.70 weeks, − 0.35 to 1.75; respectively) [31,33,34], meta-analysis of all the above studies could not be conducted due to the obvious statistical heterogeneity. Another trials reported the numbers of the preterm births, which showed no differences between herbal medicine and conventional medicine (RR 0.28, 95% CI 0.06 to 1.27, I 2 = 0%, P = 0.10, 12 trials, 120 patients) [32,35].

Combination of herbal medicine and conventional medicine vs. conventional medicine alone
Thrity-nine trials contributed data to this comparison [33,.
The average gestational age of the newborn One trial reported the average gestational age of the newborn which indicated that there was no difference between thecombination group and conventional medicine group (MD 0.49 weeks, 95% CI 0.01 to 0.97, 3 trials, 263 women) [33,66,69].

Adverse events
Five trials reported the number of adverse events [48,53,55,68,70]. One trial reported six cases of nausea, headache and flushin in treatment group [48]. One trial reported one case of loose stool, disappeared in the following course, two cases of nausea [53]. One trial reported no adverse events during the treatment [55]. One trial reported one case of headache and one case of tachycardia in treatment group, and one case of tachycardia and one case of constipation in control group [68]. Another trial reported two cases of nausea, two cases of acid reflux, one case of bloating in treatment group, and one case of acid reflux, two cases of bloating, one case of nausea, one case of rash in control group [70].

Funnel plot
A funnel plot analysis of 24 trials was performed to examine outcome for the serum TBA value of patients. The result showed potential asymmetry (Fig. 3). The potential asymmetry may be caused by small study effects, different methodological quality, or even Heterogeneity in intervention effects.

Trial sequential analysis (TSA)
We conducted TSA with the data from the two metaanalyses above. For the outcome of numbers of fetal distress in comparing herbal medicine combined conventional medicine to conventional medicine alone, TSA illustrated that the cumulative Z-curve across the traditional boundary of 5% significance (horizontal line) as well as the monitoring boundaries (inward sloping curves) (see Fig. 4). After the sixth study, the significance testing had been performed each time a new trial was added to the meta-analysis, which means the sample size achieved the required 349 participants and we had enough power to confirm the evidence (that with adjunction of herbal medicine, the therapy may decrease 12% more fetal distress) controlling for the risk of random error.
For the outcome of numbers of asphyxia neonatorum in the same comparison, the result was similar with the two crossings. TSA also illustrated that the cumulative Z-curve across the horizontal line and the inward sloping curves (see Fig. 5), which means the sample size achieved the required 376 participants and we had enough power to confirm the evidence (that the combination therapy may decrease 10% more cases of asphyxia neonatorum).

Summary of main findings
Totally 43 trials with 3556 women were included, all of them were assessed as having high risk of bias. Results from these trials showed potential better effect of herbal medicine when combined with conventional medicine on decreasing pruritus scores, reducing adverse birth events (which was supported by the TSA results) and improving the liver functions according to the serum TBA, ALT, AST values compared to conventional medicine alone. However, due to the poor methodology of the included studies and the obvious statistical heterogeneity among trials, quality of the evidence for all these outcomes were "low" and "very low" according to the GRADE assessment (see Tables 3 and 4). When compared to conventional medicine, the single application of herbal medicine showed no better/worse effect for the condition, we could not draw any firm conclusion on this type of comparison due to the small sample size of involved trials. Furthermore, few data were reported to indicate the adverse events in herbal medicine group, which also made the safety of the herbal medicine was unclear.
Compare to the previous studies The results of a meta-analysis preliminarily showed that traditional Chinese medicine combined with conventional medicine (or physical therapy) had a certain therapeutic effect on neonatal jaundice in treatment of recurrent spontaneous abortion [74]. Another metaanalysis showed that Yinzhihuang Oral Liquid was more effective in improving itching symptoms of pregnant women, reducing serum total bilirubin, total bilirubin, glycocholic acid and increasing neonatal weight of newborns than the control group in treatment of intrahepatic cholestasis of pregnancy [75]. The third meta-analysis showed that Yinchenhao Decoction had better effect on treating neonatal jaundice, shortening recovery time of serum total bilirubin and jaundice subsidence time than that of conventional medicine alone [76]. All these findings indicated that herbal medicine may have effect on improving the liver function of pregnancy women, which was consistent with our findings. Since we did not find any other review focused on this disease through the literature searching, this is the probably the first study assessed the herbal medicine for ICP and the current evidence may support the effectiveness of combination of herbal medicine and conventional medicine for decreasing the maternal pruritus scores, the serum TBA, and the number of fetal distress, asphyxial events or asphyxia neonatorum events related to this condition.

Implications for practice
Though we only got "very low" quality evidence to support the adjunctive effect of herbal medicine in treating ICP. The results showed a statistically meaningful advantage of herbal medicine on improving the symptoms or reducing adverse birth events when combined with conventional medicine. Considering the uncertain safety of the herbal product, we suggest the application of herbal medicine would be recommended in consideration of the expertise and experience of the clinician. Treatment duration would be 10 to 14 days according to this review. The most frequently used prescription in our review is Yinchenhao Decoction, in which Yinchen and Zhizi are the core component. Bian found that Yinchenhao decotion can induce liver fibrosis by dimethylnitrosamine, reduce hydroxyproline and improve liver function and hepatic histology after 2 weeks of treatment in rats [77]. Mentimes, a systematic review indicated that Yinchenhao decotion can significantly improve cholestasis by reducing elevated serum markers [78]. Practitioners may choose to use modified Yinchenhao Decoction in treating this condition.

Implications for future studies
There are some deficiencies in this study that should be improved in future studies. In terms of scheme design and method application, most studies did not give a clear introduction to the method of randomization and blinding. Although the blind method may not be easy to operate for patients due to the difficulty in the operation of herbal medicine placebo, a blind method should be implemented for the outcome assessors or statisticians to avoid bias. High quality studies are needed to verify the adjuvant efficacy and safety of Chinese herbal medicine. Meanwhile, cost effect analysis may be considered to be conducted in the future.

Conclusions
This review found "very low" quality evidence which support the effectiveness of combination of herbal medicine and conventional medicine for decreasing the maternal pruritus scores, the serum TBA, and the number of adverse birth events. TSA analysis showed the results of benefit of combination of herbal medicine and conventional medicine for decreasing the number of fetal The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI)  The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI)