Herbal hepatotoxicity in traditional and modern medicine: actual key issues and new encouraging steps (original) (raw)

Key issues

Herbal products in traditional and modern medicine are commonly perceived by the general population as well tolerated and devoid of major adverse reactions. One of the most important goals in clinical practice is to offer patients an efficient therapy for their ailment(s), without harming their health. However, efficacy and safety by the use of herbs in traditional and modern medicine are features that may apply to some herbs and patients but certainly not to others. Similar shortcomings are known from conventional chemical drugs, which also are not effective in all patients. The risk of rare adverse reactions occurring in various organs including the liver relates to both, herbs (NIH, 2014a) and synthetic drugs (NIH, 2014b).

Hepatotoxicity case reports

General aspects

Basic knowledge of hepatotoxicity by drugs and numerous other chemicals was summarized by Hyman Zimmerman in his pioneering book 25 years ago, briefly mentioning already some herbs as culprits and referencing a few case reports of herbal hepatotoxicity (Zimmerman, 1999). Since then, many more HILI cases emerged, which previously were reported (Pittler and Ernst, 2003b) and recently analyzed in publications on herbal TCM preparations (Teschke, 2014; Teschke et al., 2014c, 2015b), other commonly used herbal products (Teschke et al., 2013f), and herbal and dietary supplements (HDS) (Halegoua-De Marzio et al., 2013; Teschke et al., 2013d; Navarro et al., 2014; Robles-Diaz et al., 2015).

Epidemiology

Epidemiology data of hepatotoxicity cases in connection with herbal use are crucial to assess, both in traditional and modern medicine. Actually, the true prevalence of herbal hepatotoxicity is the total number of HILI cases in the population at a given time (Teschke et al., 2013f). It represents an estimate of how common herbal hepatotoxicity is within a population and at a fixed time. Conversely, the incidence of herbal hepatotoxicity is expressed as the total number of new HILI cases during a certain period of time, divided by the number of individuals in the population initially at risk. Therefore, incidence differs from prevalence measuring new HILI cases; for chronic liver injury, these values may change.

Incidence commonly provides information about the risk of acquiring HILI, whereas prevalence signifies how widespread HILI is. The true prevalence and incidence of HILI (Navarro, 2009) and HDS (Navarro et al., 2014; Robles-Diaz et al., 2015) is unknown. Global epidemiology considerations of prevalence and incidence refer to all herbs contained in herbal drugs and herbal supplements, whereas specific epidemiology is restricted to one single herb. Global epidemiology data therefore may be used for health economy assessment whereas specific epidemiology data pertain to herbal product safety. For this purpose, true global prevalence and incidence of HILI still has to be determined through cohort studies or case control-studies, a difficult approach. For an appropriate assessment of the risks from a specific herbal product, there is lack of quantitative data for consumption of herbal products, number of HILI patients, and the population at risk. In addition, herbal product authentication is missing in most cases of suspected HILI and impedes causality assessment for the incriminated herb (Teschke et al., 2013f). Case underreporting and overdiagnosing also prevent determination of the true incidence; future studies will have to address these issues in order to provide firm data of prevalence and incidence in HILI.

For DILI by synthetic drugs, respective data are available: the estimated annual incidence rate of DILI at a coordinating center in Spain was 34.2 ± 10.7 cases per 106 inhabitants (Andrade et al., 2005), and in a French study it was 13.9 ± 9 per 106 inhabitants per year (Sgro et al., 2002).

Compilation of hepatotoxicity cases

For herbal TCM with potential liver injury, we identified 44 different TCM herbs and 21 herbal TCM mixtures, published in case reports and case series as provided by appropriate references (Table 1). These referenced reports present clinical case details, summarized in part also earlier (Teschke, 2014). The 12 most common Chinese herbal medicines with hepatotoxicity detailed in a recent review (Ma et al., 2014) are also included in the present compilation (Table 1).

Compilation of reported cases with suspected hepatotoxicity by herbal traditional Chinese medicine (TCM).

Data represent an update of cases retrieved from a selective literature search for published cases of herbal TCM associated with suspected hepatotoxicity (Teschke, 2014; Teschke et al., 2014c, 2015b) and of a recent compilation most common Chinese herbal medicines with hepatotoxicity (Ma et al., 2014). In some cases, causality for individual herbs and herbal mixtures was established using the CIOMS (Council for International Organizations of Medical Sciences) scale or its modifications, and by positive reexposure test results. For other cases, information was fragmentary and did not necessarily allow a firm causal attribution.

Other herbs and herbal products unrelated to TCM showed reported potential hepatotoxicity for 111 items (Table 2), presented as an update of an earlier compilation (Teschke et al., 2012h). Most of the actual 111 items identified single herbs, rarely mixtures with HDS as examples with various ingredients (Table 2). Numerous other HDS with assumed potential hepatotoxicity are listed in compilations of other reports published just recently (Bunchorntavakul and Reddy, 2013; Navarro et al., 2014; Robles-Diaz et al., 2015) and hence were not included in the present compilation (Table 2).

Compitalion of commonly used herbs and herbal products with reported hepatotoxicity.

Data are retrieved from a selective literature search for selective reports of herbs and herbal products with hepatotoxicity and actualized from a previous report (Teschke et al., 2012h). In numerous cases, causality was proposed, but not necessarily established and open for discussion.

In the past, some review articles focused exclusively on HILI by TCM herbs and herbal preparations (Ma et al., 2014; Teschke, 2014; Teschke et al., 2014c, 2015b) as a primarily neglected topic, which was otherwise considered as part of overall assessments on herbal hepatotoxicity in few publications (Zimmerman, 1999; Abdualmjid and Sergi, 2013; Bunchorntavakul and Reddy, 2013), including an official and well updated NIH statement (NIH, 2014a).

Symptomatology

Clinical symptoms of herbal hepatotoxicity in traditional and modern medicine are variable and described in published case reports, case series, and regulatory presented spontaneous reports as referenced (Tables 1, 2). Symptoms are mostly unspecific and sometimes difficult to direct to the liver, which delays early recognition of the unfolding liver injury (Teschke et al., 2013f, 2014c; Ma et al., 2014). Clinical signs may emerge alone or in combination with other features, while jaundice is the symptom initially best recognized by the patient, facilitating the search for advice by the primary care physician. In detail, patients with herbal TCM hepatotoxicity experience fatigue (67.3%), jaundice (60.3%), anorexia (58.0%), nausea (35.9%), and fever (19.3%), but signs such as rash, pruritus, and pale stools have also been reported (Ma et al., 2014). In another study of 16 cases of Greater Celandine (GC) with established HILI, symptoms were present in 15 cases (Teschke et al., 2012a). Single or multiple symptoms were anorexia (n = 3), fatigue (n = 5), nausea (n = 6), vomiting (n = 2), dyspepsia (n = 1), bloating (n = 1), abdominal discomfort (n = 1), right upper quadrant pains (n = 1), epigastric pains (n = 1), unspecified abdominal pains (n = 1), fever (n = 1), dark urine (n = 3), pale stool (n = 1), pruritus (n = 3), and jaundice (n = 15) (Teschke et al., 2012a). For reasons of transparency, narrative case details and clinical data of patients with assumed HILI should be provided in tabular form, as done previously (Teschke et al., 2008a, 2011a, 2012a,b,d,e; Teschke, 2010a) and shown for GC hepatotoxicity as example (Table 3). Detailed information also allows characterization of HILI by a single herb such as GC (Table 4).

Compilation of narrative case details and clinical data of patients with HILI by Greater Celandine (GC) and established causality.

Narrative compilation with details of relevant clinical data of 16 patients with liver injury by the use of the herb Greater Celandine (GC), data are derived from a review article (Teschke et al., 2012a) and based on previous reports (Teschke et al., 2011a, 2012e). In these studies, highly probable and probable causality levels for GC were established in all patients presented as final diagnoses, using the updated CIOMS scale for the individual causality assessment. Half of the patients (cases 01–08) were derived from published case reports, the other half (cases 09–16) from spontaneous reports of the German regulatory agency (BfArM, 2005). Outcome was favorable in all cases. Abbreviations: AIH, autoimmune hepatitis; ALP, alkaline phosphatase; ALT, alanine aminotransferase; AMA, antimitochondrial antibodies; ANA, antinuclear antibodies; AST, aspartate aminotransferase; BfArM, Bundesinstitut für Arzneimittel und Medizinprodukte, Bonn; CD, comedicated drug(s); CIOMS, Council for International Organizations of Medical Sciences; CMV, cytomegalovirus; EBV, Epstein Barr virus; ERCP, endoscopic retrograde cholangiopancreaticography; GC, Greater Celandine; HAV, hepatitis A virus; HBV, hepatitis B virus; HCV, hepatitis C virus; HSV, herpes simplex virus; LKM, liver kidney microsomal antibodies; MRCP, magnetic resonance cholangiopancreaticography; MRT, magnetic resonance tomography; PCR, polymerase chain reaction; SMA, smooth muscle antibodies; VZV, varicella zoster virus.

Preferred documentation as example: clinical characteristics of GC hepatotoxicity.

Preferred documentation: The data are based on the cases of 16 patients with GC hepatotoxicity and highly probable or probable causality levels for GC and derived from a previous report (Teschke et al., 2012a). Abbreviations: ALT, alanine aminotransferase; ALP, alkaline phosphatase; GC, Greater Celandine.

Although, clinical features are quite similar in HILI cases by traditional and modern medicine (Chau et al., 2011; Teschke et al., 2013f; Ma et al., 2014), there is one exception that relates to the hepatic sinusoidal obstruction syndrome (HSOS), formerly hepatic veno-occlusive disease (HVOD); this special liver injury is caused by pyrrolizidine alkaloids (PAs) contained in various TCM herbs, with its major diagnostic features of abdominal distension and pain, ascites, malaise, hepatomegaly, and body weight gain due to ascites and edema caused by fluid accumulation (Wang and Gao, 2014). Jaundice was most frequent with 84.8% in 100/118 cases of PA induced HSOS by Tu San Qi (Gynura segetum), ascites with 99.2% (121/122 cases, and hepatomegaly with 92.0% (104/113 cases) (Lin et al., 2011; Gao et al., 2012).

In a typical HILI case unrelated to PAs, the chronology of symptoms may follow a particular stepwise pattern, as described for HILI caused by Indian Ayurvedic herbs through an excellent observation by a patient under treatment for her vitiligo (Teschke and Bahre, 2009). Her symptoms started with pruritus, followed by loss of appetite, fatigue, nausea, vomiting, dark urine, light stool, until finally jaundice was recognized by her family physician; this sequence of symptoms stretched over almost 4 months under continued herbal medication.

Patients with HILI may be asymptomatic with increased values observed by chance, monosymptomatic, or polysymptomatic. Latency period describes the interval between initiation of herb use and time of onset, evidenced by emerging symptoms or increased liver values. Liver injury by herbal TCM develops slowly with clinical symptoms appearing between 1 week and 1 month (Ma et al., 2014), or up to 150 days (Chau et al., 2011); with a longer latency period of 5–260 weeks for green tea extracts (GTE) (Mazzanti et al., 2009; Teschke et al., 2014a); or 1 week–24 months for other herbs such as kava (Teschke et al., 2008a); and 28–134 days for Greater Celandine (GC) (Teschke et al., 2011a). Finally, published HILI symptoms (Chau et al., 2011; Teschke et al., 2013a; Ma et al., 2014) are similar to those of DILI (Andrade et al., 2004; Liss and Lewis, 2009).

Clinical course

The clinical course of HILI is variable with details provided in most publications as referenced (Tables 1, 2). For HILI cases, some details of treatment modalities by herbal products of traditional and modern medicine are provided, with focus on daily and cumulative dose, treatment duration, latency period, and reexposure duration (Table 5). With cessation of herbal use, clinical signs usually vanish along with improvements or normalization of initially increased liver values, as illustrated by few examples (Verucchi et al., 2002; Vanderperren et al., 2005; Teschke and Bahre, 2009; Furukawa et al., 2010; Valente et al., 2010; Yang et al., 2010). A well described dechallenge of liver values in suspected HILI is one of the key items to suspect causality for a particular herb. Patients with HILI caused by herbal TCM or modern herbal medicine commonly experience an acute type of liver injury, which is self-limited upon withdrawal of the offending herb with an overall good prognosis. Whether herbs may cause chronic forms of HILI has not yet been evaluated in detail (García-Cortés et al., 2008). However, persistence of increased liver values raises the question whether these are due to a preexisting liver disease present prior to herbal use rather than to HILI (Picciotti et al., 1998). The acute type of HILI rarely may progress to acute liver failure (Stadlbauer et al., 2005; Fong et al., 2010). This is a serious condition that may require a liver transplant and eventually leads to death (Perharic-Walton and Murray, 1992; Yoshida et al., 1996; Haller et al., 2002; Adachi et al., 2003; Estes et al., 2003; Yuen et al., 2006; Sohn et al., 2008; Fong et al., 2010). Between 1992 and 2008; in Seoul (Korea) alone, 24 patients underwent liver transplantation due to toxic hepatitis caused by herbal TCM (Sohn et al., 2008), causing concern in view of poorly documented efficacy of herbal TCM (Manheimer et al., 2009; Teschke, 2014).

Some chracteristics of daily and cumulative doses, treatment duration, latency period, and reexposure period of cases with hepatotoxicity by herbs of traditional and modern medicine.

In all 22 hepatotoxicity cases, causality for the respective herb or herbal mixture was ascertained by positive reexposure test results based on established criteria. Unclear: ?, Adapted data from a previous report (Teschke et al., 2014b).

Cessation of herbal use is the only therapeutic approach for HILI patients. Other options including evidence based therapy for treating patients with HILI are lacking, but on a case basis treatment was reported with glycyrrhizin (Inoue et al., 2011), ursodesoxycholic acid (Jorge and Jorge, 2005; Inoue et al., 2011), or corticosteroids (Weinstein et al, 2012).

Hepatotoxicity criteria

HILI case assessment mandates clear hepatotoxicity criteria for disease characterization including causality assignment (Teschke et al., 2013f, 2014b). Laboratory-based criteria of HILI are best defined by alanine aminotransferase (ALT) and/or alkaline phosphatise (ALP) values, expressed as N in multiples of the upper limit of their normal range (Figure 1). For ALT, recommendations initially were at >2N (Bénichou et al., 1993; Danan and Bénichou, 1993) and currently are at >5N (Björnsson et al., 2012; Teschke et al., 2014c,d) or at 3N if total bilirubin values exceed 2N (Aithal et al., 2011); for ALP, values of >2N are considered diagnostic (Bénichou et al., 1993; Danan and Bénichou, 1993; Aithal et al., 2011). Restricting the ALT criteria to >5N will eliminate unspecific ALT increases and substantiate causality at a high level of probability (Björnsson et al., 2012). Considering patients with ALT values of >2N will initially also include numerous cases with nonspecific increases, which then require thorough assessment and stringent exclusion of causes unrelated to the used herb(s). For low threshold values, the rate of alternative diagnoses is high (Teschke et al., 2013g), findings that are plausible and not unexpected (Teschke et al., 2014e). Other values such as aspartate aminotransferase (AST) are not required, unless to be used as substitute for ALT if not available.

Concern emerges whenever hepatotoxicity is assumed even if liver values were only marginally increased, not reported, or not assessed. These problems are not uncommon for cases of assumed HILI, presented for instance by the US Pharmacopeia (USP) (Mahady et al., 2008) relating to both black cohosh (BC) (Teschke, 2010c; Teschke et al., 2011d,e; Teschke and Schulze, 2012) and green tea extracts (Sarma et al., 2008; Liss and Lewis, 2009); by the WHO, relating to kava (WHO, 2007a; Teschke and Wolff, 2009); the German regulatory agency BfArM (Bundesinstitut für Arzneimittel und Medizinprodukte) (BfArM, 2002, 2005) relating to kava (Schmidt et al., 2005; Teschke et al., 2008a); or the Drug Commission of the German Medical Association (DCGMA, 2011) relating to Pelargonium sidoides (PS) (Teschke et al., 2012b,e). In published case reports receiving the benefit of appropriate peer reviews, the presented HILI cases commonly provide high values of aminotransferases and/or ALP and basic data support of potential hepatotoxicity, as shown also for some cases with a positive reexposure test results (Teschke et al., 2014b). In other HILI case series, however, criteria were not or incompletely documented; neglecting these aspects in effect invalidates the causality assessment.

In spontaneous reports of regulatory agencies, a clear hepatotoxicity definition was provided by EMA (2007) but not by the U.S. Pharmacopeia (USP) (Mahady et al., 2008; Sarma et al., 2008), the German BfArM (2002), or the WHO (2007a). For instance, EMA mentions cases with assumed HILI by BC but clarifies that a causal attribution cannot be made with the required certainty in face of missing liver values (EMA, 2007). Consequently, missing regulatory hepatotoxicity definitions represent confounding variables and result in false high signal cases due to regulatory case overreporting and overdiagnosing (Teschke et al., 2013f).

For reasons of transparency and assessment of case data quality, each HILI case series should provide tabulated information of available or missing case details, as done in various reports (Teschke et al., 2011a, 2012b,e) and shown as example (Table 6) (Teschke, 2010c).

Preferred documentation as example: overview of known information regarding all 69 patients with primarily suspected but not established HILI by black cohosh (BC).

PreferreThe group of the 69 cases consisted of 11 case reports (cases 1–11), 13 TGA cases from Australia (cases 12–24), 2 CADRMP cases from Canada (cases 25 and 26), 7 MedWatch/ FDA cases from the United States (cases 27–33), and 33 EMA cases from the European Union (cases 34–69). Details and references of the 69 cases were published earlier (Teschke et al., 2010). Abbreviations: ALP, alkaline phosphatase; ALT alanine aminotransferase; BC, black cohosh; CMV, cytomegalovirus; EBV, Epstein Barr virus; HAV, hepatitis A virus, HBV, hepatitis B virus, HCV, hepatitis C virus; HSV, herpes simplex virus; VZV, varicella zoster virus.

HILI case characteristics

Hepatotoxicity classification is mandatory in cases of assumed HILI to facilitate further evaluation of reexposure results and CIOMS assessments (Teschke et al., 2013f). Based on specific laboratory constellations, differentiation of the hepatocellular, cholestatic or mixed form of hepatotoxicity is feasible by comparing serum activities of ALT and ALP at the time HILI diagnosis is first suspected (Bénichou et al., 1993; Danan and Bénichou, 1993; Teschke et al., 2014d). Enzyme activity is expressed as a multiple of the upper limit of the normal range (N), and the ratio (R) of ALT/ALP is calculated. Liver injury is classified as hepatocellular, if ALT > 2N alone or R = 5; cholestatic, when there is an increase of ALP > 2N alone or when R = 2; of the mixed type if ALT > 2N, ALP is increased, and 2 < R < 5 (Figure 1). In a HILI case series of herbal TCM consisting of 27 patients, the pattern of liver injury was hepatocellular in 82% of the cases, cholestatic in 11%, and mixed in 7% (Chau et al., 2011).

Liver histology

Liver biopsy in HILI and DILI cases requires special attention in any clinical hepatology setting, balancing benefits and risks for the patient (Teschke and Frenzel, 2014). Published and spontaneous HILI reports often contain detailed histological descriptions of liver biopsy findings, mostly associated with pictures obtained by microscopy. This erroneously implies that liver biopsy is an essential part of routine case assessments (BfArM, 2002; Teschke et al., 2008a, 2011a, 2012a,b,f,g; Teschke, 2010c). Histology data were also presented by narrative HILI case reports lacking even any causality for a particular herb (Teschke et al., 2012f,g). This raises the question whether a liver biopsy is justified, considering also that there were no histological findings recognized as specific for all hepatotoxicity cases (Ramachandran and Kakar, 2009). Liver biopsy in chronic hepatotoxicity cases to define prognosis in the absence of an expected specific therapy option remains debatable (Teschke and Frenzel, 2014).

To evaluate liver histology findings, a retrospective case analysis of pathological changes in HILI selectively caused by one single herb with established causality appears the best approach. For instance, HILI cases of kava and GC have such an established causality track. In 12 GC HILI patients with a probable or highly probable causality grading for GC, prevailing histological features included hepatitis, single or confluent liver cell necrosis, inflammation, rarely fibrosis, and cholestasis (Teschke et al., 2011a, 2012a,e). In eight HILI patients with a highly probable, probable or possible causality for kava, liver histology showed hepatitis, liver cell necrosis, and rarely bile duct proliferation and intrahepatic cholestasis (Teschke et al., 2008a). Therefore, at least for the two herbs GC and kava, the histological features are quite uniform and restricted to two major features, hepatitis and liver cell necroses. These histological characteristics, however, are also found in most other liver diseases unrelated to herbs, obviating liver biopsy in suspected HILI cases due to unspecific results.

Additional insights are provided by the analysis of cases with positive reexposure tests, done unintentionally with the incriminated herb or herbal mixture, and available liver histology results. For instance, in HILI by a herbal mixture of TCM, total liver necrosis prevailed (Perharic-Walton and Murray, 1992); germander (Teucrium chamaedrys) caused hepatocyte necrosis with lobular inflammatory infiltration mainly by mononuclear cells, associated with slightly fibrous portal tracts containing inflammatory cells (Larrey et al., 1992); senna use resulted in liver cell necrosis around the central veins as well as portal and lobular infiltration by lymphocytes, histiocytes, and rare plasma cells (Beuers et al., 1991); chaparral intake was associated with hepatocellular necrosis combined with inflammation, portal tract expansion, mild cholestasis and fibrous septation (Batchelor et al., 1995); and the herbal TCM Chinese skullcap (Scutellaria baicalensis) was considered to cause acidophil bodies, ballooned hepatocytes, lobular inflammatory cell infiltrates including eosinophils, and portal tracts containing mononuclear cells and eosinophils (Yang et al., 2012a). Based on causality established by positive reexposure results, these few examples may provide some insight in morphological liver changes due to herbal use.

Histological features usually are not clinically relevant, but some clinicians still consider a liver biopsy an important part of the diagnostic work-up in suspected hepatotoxicity cases. The question is whether histological results changed the initial diagnosis or benefited the individual patient. In two cases of initially suspected HILI, however, histological findings of giant cell hepatitis were reported and completely ignored (Dunbar and Solga, 2007; Schoepfer et al., 2007), while the clinical course and this particular histological pattern best fitted with an existing severe virus infection with hepatic involvement rather than herbal hepatotoxicity (Teschke and Schwarzenboeck, 2009).

Clearly, the pathologist is not helpful offering diagnoses such as HILI or liver injury compatible with or suggestive for herbal use. Overall, liver histology as a supporting routine method for assessing HILI cases is not recommended, it commonly adds little new specific diagnostic clues as information to the case without benefit for the patient; as an invasive procedure, rare but potentially life threatening complications may occur (Teschke and Frenzel, 2014).

Alternative causes

Unrecognized alternative diseases are a real clinical problem when caring for patients with initially assumed but later not confirmed HILI. Several hundred liver diseases have to be considered as diagnoses alternative to HILI, to be ruled out under clinical aspects and with specific diagnostic tools. As a reminder for clinicians, a checklist with details for these alternative diagnoses is available (Table 7) (Teschke et al., 2014d). Numerous missed diagnoses were found upon reevaluation of initially assumed HILI cases, with similar problems for DILI (Figure 2) (Teschke et al., 2013g, 2014a). Exclusion of hepatitis E and infections by cytomegalovirus (CMV), Epstein Barr virus (EBV), herpes simplex virus (HSV), and varicella zoster virus (VZV) should be obligatory rather than facultative (Table 7).

Differential diagnoses of HILI.

This tabular compilation represents an update of a previous version (Teschke et al., 2013f). AAA, Anti-actin antibodies; AMA, antimitochondrial antibodies; ANA, antinuclear antibodies; ASGPR, Asialo-glycoprotein-receptor; BMI, body mass index; CIOMS, Council for International Organizations of Medical Sciences; CT, computer tomography; CYP, cytochrome P450; DPH, pyruvate dehydrogenase; HAV, hepatitis A virus; HBc, Hepatitis B core; HBV, hepatitis B virus; HCV, hepatitis C virus; HEV, hepatitis E virus; HILI, herb induced liver injury; HIV, human immunodeficiency virus; LKM, liver kidney microsomes; LP, liver-pancreas antigen; LSP, liver specific protein; MRC, magnetic resonance cholangiography; MRT, magnetic resonance tomography; p-ANCA, perinuclear antineutrophil cytoplasmatic antibodies; PCR, polymerase chain reaction; SLA, soluble liver antigen; SMA, smooth muscle antibodies; TSH, thyroid stimulating hormone; TTG, tissue transglutaminase.

Causality evaluation

Reexposure

Establishing the diagnosis of HILI may be cumbersome in the usual clinical setting, with experts not available in place and in time. Since, convincing biomarkers for all HILI cases are lacking, the gold standard for the diagnosis of hepatotoxicity still is a positive unintentional reexposure test result, if available (Bénichou et al., 1993; Danan and Bénichou, 1993; García-Cortés et al., 2008; Chalasani and Björnsson, 2010). Details of essential criteria are based on the conclusions of an International Consensus Meeting, as referred previously (Bénichou et al., 1993). Accordingly, required data are baseline ALT levels before reexposure, designed ALTb, and reexposure ALT levels, designed ALTr. The reexposure test is positive, if ALTr = 2ALTb and ALTb is below 5N, with N as the upper limit of the normal value (Table 8). Other variations lead to uninterpretable results. Some HILI reports mentioned a positive reaction upon reexposure, and these cases were further analyzed. Not all reexposure tests could be confirmed, partially due to lack of any details; however, for numerous herbs and herbal preparations of TCM and modern medicine, valid reexposure test results confirmed causality in the assessed HILI cases (Table 9) (Teschke et al., 2014b,d); using these cases, some characteristic features of daily and cumulative doses, divergences between treatment duration and latency period, and reexposure duration are evident (Table 5).

Conditions of unintentional reexposure tests in suspected HILI cases.

Conditions and criteria for an unintentional reexposure test, adapted from a previous report (Teschke et al., 2014b). Accordingly, required data for the hepatocellular type of liver injury are the ALT level, just before reexposure, designed as baseline ALT or ALTb, and the ALT levels during reexposure, designed as ALTr. Response to reexposure is positive, if both criteria are met: first, ALTb is below 5N with N as the upper limit of the normal value, and second ALTr ≥2ALTb. Other variations lead to negative or uninterpretable results. For the cholestatic (±hepatocellular) type of liver injury, corresponding values of ALP are to be used rather than of ALT. Abbreviations: ALP, Alkaline phosphatase; ALT, Alanine aminotransferase; n.a., not available.

Analysis of reported positive reexposure test results in cases of suspected herbal Traditional Chinese Medicine (TCM) induced liver injury.

Compilation of some clinical details and laboratory values for assessment of reported positive reexposure test results in 25 cases with suspected herbal hepatotoxicity by TCM products. Data are derived from previous reports, which may provide additional details (Teschke, 2014; Teschke et al., 2014b). Unless otherwise stated, reexposure was commonly unintentional. Criteria for a positive reexposure test result were used as described in Table 4, restricted to criteria provided for the hepatocellular type of liver injury. Accordingly, essential data are the ALT levels at baseline before reexposure (ALTb), and the ALT levels during reexposure (ALTr). Response to reexposure is positive if ALTr = 2ALTb and ALTb < 5N, with N as the upper limit of the normal value. Other combinations lead to negative or uninterpretable results. Serum enzyme activities were provided in U/L or multiples of N. Details for calculation of the R value are presented before (Figure 1). Abbreviation: ALT, alanine aminotransferase; AST, aspartate aminotranferase; N, upper limit of normal; R, ratio; TCM, Traditional Chinese Medicine.

CIOMS/RUCAM

Physicians at armlength from the patient with HILI are well advised to consider a pragmatic thorough clinical evaluation in connection with a prospective structured approach assessing causality, providing the diagnosis in time while the disease is unfolding and without delay due to waiting periods for expert rounds' conclusions months thereafter; this is a crucial issue worldwide. We clearly prefer the CIOMS scale (Council for International Organizations of Medical Sciences), also called RUCAM (Roussel Uclaf Causality Assessment Method), in its original form (Bénichou et al., 1993; Danan and Bénichou, 1993) or better its update (Tables 10, 11) (Teschke et al., 2013a, 2014d). Discussions focused on strengths and weaknesses of CIOMS, a learning system and not immutable (Andrade et al., 2004; Rochon et al., 2008; Aithal et al., 2011; García-Cortés et al., 2011; Teschke and Wolff, 2011; Teschke and Schulze, 2012; Teschke et al., 2008c, 2013e, 2014d; Lewis, 2014; NIH, 2014b). Outlined suggestions for improvement and refinement are incorporated in the updated CIOMS scale (Tables 10, 11) (Teschke et al., 2013e, 2014d). Included is now the search for additional competing causes such as sepsis; or autoimmune hepatitis, chronic hepatitis B and C, primary biliary cholangitis and sclerosing cholangitis, and genetic liver diseases. HBsAg and HBV-DNA quantification were added to distinguish HBV infection from immunization, as was HCV-RNA to correctly assess HCV infections. Specific diagnostic criteria now include PCR detection and titer changes of the respective antibodies (IgM, IgG) for CMV, EBV, HEV, HSV, and VZV infections. Hepatobiliary sonography was supplemented by color Doppler sonography including assessments of the liver vessels, endosonography, computed tomography (CT), and magnetic resonance cholangiography (MRC. Alcohol as risk factor is now specified by an intake of >2 drinks per day (>14 units/week) in woman and >3 drinks per day (21 units/week) in men, whereby one drink corresponds to 10 g ethanol. For comparison and method validation, causality has been evaluated in 101 hepatotoxicity cases by both the original and updated CIOMS scales, with identical causality results published in 6 studies (Teschke et al., 2014d). Therefore, the updated CIOMS scale was validated, and there is no need for further validation of the updated CIOMS scale vs. the original CIOMS scale.

![

            CIOMS scale for the hepatocellular type of injury in HILI cases.
        ](https://www.frontiersin.org/files/Articles/134515/xml-images/fphar-06-00072-i0001.webp)

CIOMS scale for the hepatocellular type of injury in HILI cases.

The updated CIOMS scale is derived and modified from a previous report (Teschke et al., 2014c). The above items specifically refer to the hepatocellular type of injury rather than to the cholestatic or mixed type (shown in Table 11). Abbreviations: ALT, Alanine aminotransferase; AST, Aspartate aminotransferase; CIOMS, Council for International Organizations of Medical Sciences; CMV, Cytomegalovirus; CT, Computer tomography; DILI, Drug induced liver injury; EBV, Epstein Barr virus; HAV, Hepatitis A virus; HBc, Hepatitis B core; HBsAg, Hepatitis B antigen; HBV, Hepatitis B virus; HCV, Hepatitis C virus; HEV, Hepatitis E virus; HILI, Herb induced liver injury; HSV, Herpes simplex virus; MRC, Magnetic resonance cholangiography; N, upper limit of the normal range; VZV, Varicella zoster virus. Total score and resulting causality grading: = 0, excluded; 1–2, unlikely; 3–5, possible; 6–8, probable; ≥9, highly probable.

![

            CIOMS scale for the cholestatic or mixed type of injury in HILI cases.
        ](https://www.frontiersin.org/files/Articles/134515/xml-images/fphar-06-00072-i0002.webp)

CIOMS scale for the cholestatic or mixed type of injury in HILI cases.

For details see legend to Table 10.

A selective compilation shows that numerous international registries and regulatory agencies as well as associated groups actually apply CIOMS for HILI and DILI cases (Table 12), with its advantages over other causality assessing approaches (Teschke et al., 2013a) including the method of DILIN (Drug-Induced Liver Injury Network) (Table 13). As opposed to the DILIN method, CIOMS fascinates by its stringent scoring system (Table 13), with its generation of a quantitative assessment (score) to address probability, which is more precise than a simple yes or no (Gunawan and Kaplowitz, 2004). Support for CIOMS was actually provided by Navarro as DILIN member and senior author of a HDS case series by applying CIOMS for causality assessment (Halegoua-De Marzio et al., 2013), whereas CIOMS was not in favor of Lewis, another DILIN member (Lewis, 2014). Although, connected with an actual commentary on a single HILI case report that did not undergo any formal causality assessment, Navarro also correctly acknowledged that CIOMS/RUCAM is the most frequently referenced scoring system (Fenkel and Navarro, 2011). This assumption supports earlier systematic analyses of 2008 (Tajiri and Shimizu, 2008). Of 61 DILI reports that were reviewed in the PubMed database over the last decade, showing that CIOMS was the most used scale. In a recent confirmative study of 573 HILI cases (Table 14), CIOMS again was the most used method applied in 275 cases (48.0%) (Teschke et al., 2013g), in line with mainstream opinion (Tables 12, 14) (Wai, 2006; Aithal et al., 2011; NIH, 2014a,b; Björnsson et al., 2012, 2013. The CIOMS scale was widely used for hepatotoxicity assessments in epidemiological studies, clinical trials, case reports, case series, regulatory analyses, and genotyping studies, as referenced in detail recently (Teschke et al., 2013f).

Selective compilation of international registries and regulatory agencies, and associated groups applying the CIOMS scale for causality evaluation in suspected HILI and DILI cases.

Core elements of the updated CIOMS scale vs. DILIN method.

Adapted and derived from a previous report (Teschke et al., 2013a). Latency period indicates time from herb start to symptoms, alternatively to abnormal liver tests. The symbol + shows that this item is present and the symbol 0 indicates lack of this item. Abbreviations: ALT, Alanine aminotransferase; ALP, Alkaline phosphatase; CIOMS, Council for International Organizations of Medical Sciences; CMV, cytomegalovirus; DILIN, Drug-Induced Liver Injury Network;EBV, Epstein Barr virus; HAV, Hepatitis A virus; HBV, Hepatitis B virus; HCV, Hepatitis C virus; HEV, Hepatitis E virus; HSV, Herpes simplex virus; VZV, Varicella zoster virus.

Compilation of causality assessment methods used in suspected HILI cases.

The data are derived from a study evaluating alternative causes in suspected HILI cases (n = 573) (Teschke et al., 2013g). For the 275 CIOMS cases, causality assessment was performed with the updated CIOMS scale, the original CIOMS scale, or early CIOMS version. Abbreviations: Ad hoc, ad hoc approach; CIOMS, Council for International Organizations of Medical Sciences scale; DILIN, Drug Induced Liver Injury Network method; KL, Karch & Lasagna method; Naranjo, Naranjo scale; WHO, World Health Organization method.

CIOMS is structured, quantitative, and specific and validated for hepatotoxicity, and considers all its core elements (Tables 10, 11) (Teschke et al., 2014d). It was developed by an international expert panel and validated by cases with positive reexposure tests as gold standard, showing good sensitivity (86%), specificity (89%), positive predictive value (93%), and negative predictive value (78%) (Bénichou et al., 1993). Of note, the scales for the hepatocellular and the cholestatic (± hepatocellular) type of injury differ slightly (Tables 10, 11).

The CIOMS scale was conceptualized and developed in consensus meetings organized at the request of the Council for International Organizations of Medial Sciences (CIOMS) (Bénichou et al., 1993; Danan and Bénichou, 1993), aiming to overcome experts' previous problems with unstructured and unquantified evaluations lacking defined and scored items,resulting in debated causality assignments. This CIOMS scale represented a breakthrough in DILI and HILI causality assessment methods and extended, specified, and quantified preceding versions (Danan, 1988; Bénichou, 1990). The basis for CIOMS was provided by eight experts in hepatology from 6 countries and included J. P. Benhamou (France), J. Bircher (Germany), G. Danan (France), W. C. Maddrey (USA), J. Neuberger (UK), F. Orlandi (Italy), N. Tygstrup (Denmark), and H. J. Zimmerman (USA) (Bénichou et al., 1993; Danan and Bénichou, 1993). These experts in the field evaluated DILI cases for case characteristics, hepatotoxicity criteria, liver injury pattern, and reexposure criteria; they standardized DILI case assessment with specific, quantitative items and validated their method with established positive reexposure DILI case results (Bénichou et al., 1993; Danan and Bénichou, 1993). CIOMS was developed for assessment of a single drug containing a synthetic product and may be used for a single herb containing multiple chemical constituents, but does not allow causality attribution to a single constituent.

CIOMS provides a differentiating range of causality grades for the responsible agent(s) and clearly delineates liver specific criteria for challenge, dechallenge, exclusion of unrelated diseases, and comedication (Bénichou et al., 1993; Danan and Bénichou, 1993). It even takes into account atypical chronology with +1 point for challenge periods of <5 days or > 90 days, whereas the period of 5–90 days renders +2 points (Table 10). It is well adapted for cases with missing data. Physicians suspecting herbal hepatotoxicity can easily use CIOMS, results are readily available within a few minutes.

To facilitate valid actual assessment and possible external reevaluation, CIOMS scale data should be provided individually point by point for each patient (Tables 10, 11), along with the list of alternative diagnoses to be excluded (Table 7). In publications of HILI cases and for submission to regulatory agencies as spontaneous reports, the completed CIOMS scale with all items including individual and final scores should be supplied to ensure data transparency, as published before (Teschke et al., 2008a, 2011a,d,e, 2012b,d,h; Teschke and Bahre, 2009; Teschke, 2010a,c) and presented as example, using the CIOMS scale of the hepatocellular type of injury (Table 15).

Preferred documentation as example: Tabulated causality assessment of 15 patients with primarily suspected HILI by Pelargonium sidoides (PS).

Preferred documentation of causality assessment using the scale of the updated CIOMS (Council for International Organizations of Medical Sciences), considering the items for the hepatocellular type of liver injury and the data of a previous report of 15 patients with primarily suspected HILI by Pelargonium sidoides (PS) (Teschke et al., 2012b). In the above section 6 of concomitant herb(s)/drug(s), the following products were considered: synthetic drugs, dietary supplements including herbal ones, and polyherbal products. In the section 7 (search for non-herb/drug causes), the symbol of—denotes that the obtained result was negative and that of + was positive, whereas lack of a symbol indicates that assessment was not performed. ALT denotes alanine aminotransferase, AST aspartate aminotransferase, HAV hepatitis A virus, HBc hepatitis B core, HBV hepatitis B virus, HCV hepatitis C virus, CMV cytomegalovirus, EBV Epstein Barr virus, HSV herpes simplex virus, PCR Polymerase Chain Reaction, VZV varicella zoster virus. Total points/causality: ≤0/excluded; 1–2/unlikely; 3–5/possible; 6–8/probable; ≥9/highly probable.

Thus, we strongly recommend for HILI case assessment a sequential approach, starting with thorough clinical evaluations and concomitant prospective causality evaluation by the updated CIOMS scale (Tables 10, 11), followed by optional expert opinion based on scored CIOMS items, if uncertainty remains, and finally for reasons of transparency, appropriate documentation of case details (Tables 3, 6, 7, 15).

DILIN method

With use restricted to its own country, the United States DILIN causality method (Chalasani et al., 2008) does not operate within the international HILI and DILI mainstream domains, as opposed to CIOMS (Table 12). The DILIN method may create problems even in its homeland when physicians are waiting for conclusions of expert circles at times HILI is unfolding. Representing a post-clinical, postponed evaluation rather than a rapid assessment of HILI as a critical disease, the DILIN method will not gain the same international popularity as its counterpart CIOMS (Tables 9, 10, 12), also due to other major shortcomings (Table 13).

Although, various CIOMS criteria (Danan and Bénichou, 1993) have been incorporated in the DILIN method (Rochon et al., 2008), the DILIN group missed the chance to establish the fundaments of a newly conceptualized causality method for DILI and HILI, considering the well known pros ad cons of CIOMS (García-Cortés et al., 2011; Teschke et al., 2013a, 2014d), rather than commenting on few shortcomings of CIOMS at the expense on its own DILIN method (Lewis, 2014). As opposed to the transparent CIOMS results shown by tables with individual scored items (Tables 10, 11) and the applied scale with actual 15 assumed HILI cases (Table 15), the DILIN method lacks both transparency and individual item scorings (Teschke et al., 2013a, 2014d). Results presented as percentage ranges only and not given as clearly defined, individually scored items before are irrelevant. Proving a moderate reliability of their DILIN causality approach is far from concordance (Rochon et al., 2008; Teschke et al., 2014d); expert opinion validation therefore seems to be irrelevant. DILIN should close the present evaluating gap by recommending physicians in clinical practice to use CIOMS a priori to improve causality assessment already at origin of clinical data, while DILIN may later use these CIOMS-based itemized data for own assessment.

Naranjo scale

Causality assessment of hepatotoxicity cases by the Naranjo scale (Naranjo et al., 1981) with its known shortcomings is problematic (Table 13), although favored by the United States Pharmacopeia (USP) (Mahady et al., 2008) but rejected by the United States DILIN group (Lewis, 2014) and other groups (García-Cortés et al., 2011). This scale relates toxic drug reactions to general pharmacological drug actions rather than specifically to idiosyncratic reactions like hepatotoxicity; it contains drug concentrations and monitoring, dose relationship including decreasing dose, placebo response, cross-reactivity, and confirmation of the ADR using unidentified objective evidence, which are irrelevant for HILI (Naranjo et al., 1981; Teschke and Wolff, 2011; Teschke and Schulze, 2012). The hepatotoxicity unspecific feature of the Naranjo scale is unacceptable in suspected HILI cases, its results are heavily disputed (Liss and Lewis, 2009; Mahady et al., 2008; Sarma et al., 2008; Teschke, 2010c; Teschke and Wolff, 2011; Teschke et al., 2011d,e; Teschke and Schulze, 2012); this also pertains to a shortened version with only 5 of the original 10 items (Teschke and Schulze, 2012). Lack of hepatotoxicity specificity of the Naranjo algorithm was associated with missing definition of liver ADR, unclear time frame and latency period, undefined time frame for dechallenge, lacking risk factor definition, insufficient evaluation of alternative diagnoses, inappropriate assessment of comedication; and missing definition of a positive rechallenge test (Naranjo et al., 1981; Teschke and Schulze, 2012). This scale also was considered insensitive, allowing a possible causality even in the absence of essential data by virtue of the patient simply having taken the suspected agent (Liss and Lewis, 2009; Sarma et al., 2008). The Naranjo scale as modified by USP (Mahady et al., 2008) did not exclude alternative causes such as idiopathic autoimmune hepatitis, alcoholic or cardiac hepatopathy, other preexisting liver diseases, DILI, and drug-induced rhabdomyolysis (Sarma et al., 2008; Teschke, 2014; Teschke et al., 2011d,e). It therefore appears that the USP approach (Mahady et al., 2008) is an invalid tool for causality assessment in suspected HILI, leading to the conclusion that quality of causality assessment is more important than quantity of counted cases, not vice versa (Teschke et al., 2012g). Use of this method has raised concern about judgment validity by the USP regarding cases of hepatotoxicty by green tea (Liss and Lewis, 2009; Teschke and Schulze, 2012).

WHO method

The WHO method in short consists of both the WHO scale and the global introspection by experts (WHO, 2000b) and was applied for assessing causality in cases of kava hepatotoxicity by the WHO (2007a) and of PS hepatotoxicity as erroneously assumed by the German regulatory agency BfArM and the Drug Commission of the German Medical Association (DCGMA, 2011), but the value of this hepatotoxicity unspecific method was heavily debated (Stammschulte and Gundert-Remy, 2012; Teschke et al., 2012b,c,d) and judged obsolete before (Teschke and Wolff, 2011), considering its known shortcomings (Table 13). In general, global introspection represents a strategy in evaluating the likelihood of drug causality for adverse reactions (Kramer, 1986). Surprisingly, this method also has never been validated for any ADRs (Teschke et al., 2012c); as early as 1986, global introspection by experts has been shown to be neither reproducible nor valid (Kramer, 1986). Both the questions and the answers are ambiguous (Teschke and Wolff, 2011). Specifically, the assessor considers factors that might causally link one or more drugs to an observed ADR, lists all factors, weighs their importance, and decides the probability of drug causation (Kramer, 1986). No specific check list or level of strength is given.

The WHO scale was not validated by a gold standard, is not quantitative, not specific for hepatotoxicity (WHO, 2000b; Teschke and Wolff, 2011; Teschke et al., 2012b,c,d, 2013b,e). Reliability, sensitivity, specificity, positive and negative predictive values are unknown. Its scope is also limited since it cannot discriminate between a positive and a negative correlation, thereby stimulating overdiagnosing and overreporting (Teschke et al., 2013b). The WHO method ignores uncertainties in daily dose, temporal association, start, duration, and end of herbal use, time to onset of ADR, and course of liver values after herb discontinuation. Insufficiently considered or ignored are comedications, pre-existing liver diseases, numerous alternative explanations, and exclusion of virus infections by hepatitis A - C, CMV, EBV, HSV, and VZV (Teschke et al., 2012b,d). Similarly, case duplications and retracted cases remained undetected by the WHO method (Teschke et al., 2012a). Despite these flaws, the WHO method was used for causality assessment in herbal hepatotoxicity cases (Elinav et al., 2007; Schoepfer et al., 2007; DCGMA, 2011; Stammschulte and Gundert-Remy, 2012; Teschke et al., 2012b,d); claimed causality for PS was not confirmed after reevaluation in two studies (Teschke et al., 2012b,d).

Other approaches

Other attempts to evaluate causality in assumed HILI cases exist (Hung et al., 2011), also the ad-hoc assessment (Kaplowitz, 2001), which was preferentially used for kava cases by the German regulatory agency (BfArM, 2002) and in detail disputed subsequently (Teschke and Wolff, 2011), and the Karch & Lasagna method applied in some HILI cases of Herbalife® and considered obsolete recently (Teschke et al., 2013d), due to known shortcomings.

Questionable and lacking causality

Although causality was firmly established for various herbal TCM preparations as well as other herbs and herbal products in reported HILI cases (Tables 1, 2) (Teschke et al., 2011a, 2012a,e, 2014a), causality problems emerged with a few herbs and herbal preparations as evidenced by some full length published reports with detailed analyses. Among these are black cohosh with a possible causality grading in one single HILI case (Teschke, 2010c) and lacking causality in another study (Naser et al., 2011), kava with a highly probable causality level in one HILI case confirmed by a positive reexposure test result (Teschke et al., 2008a), and a confirmed causality grading assessed by a positive reexposure test result for a Herbalife® product in a single HILI case (Teschke et al., 2013d); however, CIOMS/RUCAM based causality for some Herbalife® products was highly probable in one patient and probable in six patients, as preliminarily reported in abstract form without any case details including case data quality (Halegoua-De Marzio et al., 2013), which was described as poor and scattered before (Teschke et al., 2013c).

With Ba Jiao Lian (Dysosma pleianthum), this TCM herb was not further considered as hepatotoxic (Teschke, 2014; Teschke et al., 2014c, 2015b), since not all diagnostic criteria were fulfilled for cases of hepatotoxicity by this herb (NIH, 2014c; Teschke, 2014). In detail, after herbal use at recommended doses, the patients manifested abnormal liver function tests associated with nausea, vomiting, diarrhea, abdominal pain, thrombocytopenia, leucopenia, sensory ataxia, altered consciousness and persistent peripheral tingling or numbness. However, the increase of the aminotransferases was marginal, with preference of AST rather than ALT. The AST increase could reflect isolated damage of the mitochondria around the hepatic central vein or muscular damage, because of the associated increase of creatine phosphokinase, findings not in support for a clinically relevant toxic liver disease (Teschke, 2014). Evidence against a hepatotoxic potential was also provided for Jing Tian San Qi (Sedum aizoon) as another herbal TCM (Teschke, 2014), based on the results of recent studies showing that in patients with HSOS, the hepatotoxic PAs in the herbal TCM Tu San Qi (Gynura segetum) were responsible rather than the misidentified Sedum aizoon lacking these alkaloids (Dai et al., 2006; Gao et al., 2006, 2012; Wu et al., 2008; Lin et al., 2011; Wang and Gao, 2014).

Pathogenetic aspects of HILI

Any HILI case report should describe details to ensure a pathogenetic case classification, using appropriate criteria that characterize two major forms of HILI (Figure 1). One of these is named idiosyncratic, the other one intrinsic (Zimmerman, 1999; Teschke et al., 2008a). The idiosyncratic form of injury is unpredictable and independent of the dose; its metabolic and immunologic subtypes require special attention in clinical practice (Figure 1). Conversely, the intrinsic form of liver injury is predictable and dose dependent (Figure 1). Although, valid data are lacking, it appears that most HILI cases are of the idiosyncratic rather than the intrinsic form.

Idiosyncratic form

As an example, clinical assessment characterized kava hepatotoxicity as an idiosyncratic liver injury linked to a metabolic aberration in unusually susceptible humans, providing an overall low incidence of kava hepatotoxicity in the normal population (Teschke et al., 2008a). This rarity of kava hepatotoxicity was also considered in the recent kava trial and evaluated as a positive risk/benefit constellation (Court, 2014), opposing previous regulatory assumptions to the contrary (BfArM, 2002). In accordance with other HILI cases of the idiosyncratic form of injury, human kava hepatotoxicity is not reproducible in experimental animals. Therefore, results of preclinical assessments with kava in experimental studies showing lack of liver toxicity are not transferrable to humans with another susceptibility setting and are not suitable to ensure safe use in humans. Since experimental reproducibility is missing, the lack of an experimental model prevents analytical evaluations directed to a proposed molecular mechanism of kava hepatotoxicity. Regarding human kava hepatotoxicity, characteristics of the metabolic subtype of the idiosyncratic form of injury prevail, based on the variable duration of exposure of 1week up to 12 months, associated with a weak dose dependency (Teschke et al., 2008a). Overall, most plants are fairly well tolerated by humans, whether used as normal food, herbal drugs, or HDS.

The pathophysiology of idiosyncratic HILI in humans is difficult to assess due to lack of experimental reproducibility and hence missing existence of an experimental animal model of HILI. There are abundant studies related to effects of herbs on animals or in vitro cell systems, but uncertainty exists whether these experimental results are transferable to human idiosyncratic HILI conditions. However, pathogenetic aspects are well assessable for HILI cases of the intrinsic form, due to available animal models with experimental hepatotoxicity and the possibility of transferring their results to human conditions.

Intrinsic form

Germander (Teucrium chamaedrys) hepatotoxicity is a typical liver injury of the intrinsic form, since it is dose dependent and reproducible in mice (Larrey and Faure, 2011). Due to its experimental reproducibility in animals, the molecular pathogenesis of Germander hepatotoxicity can easily be studied in experimental hepatotoxicity and transferred to human Germander hepatotoxicity. Germander components are neoclerodane diterpenoids that are oxidized by the cytochrome P450 3A isoform into reactive metabolites. These deplete hepatic stores of glutathione and cytoskeleton associated protein thiols, form plasma membrane blebs, and cause apoptosis contributing to liver cell necrosis (Larrey et al., 1992; Larrey and Faure, 2011).

PAs are other good examples for the intrinsic form of liver injury, which again is clearly dose dependent, thereby predictable, and hence preventable. For herbs containing PAs, each consumer of these herbs is at a dose dependent risk developing HSOS as a specific entity of liver disease (Smith and Desmond, 1993; Sperl et al., 1995; Stillman et al., 1977; Fu et al., 2004). PA containing plants are probably the most common poisonous plants affecting not only humans but also livestock and wildlife, with more than 6.000 plants containing PAs and about 3% of the world's flowering plants containing PAs (Fu et al., 2004). Some of these plants have caused toxic liver disease, recognized as epidemics and sometimes primarily assigned to viral hepatitis and not necessarily to toxic plants (Tandon et al., 1976a,b, 2008). Human embryotoxicity caused by PAs has been described in a newborn whose mother drank one cup of a tea containing PAs per day throughout pregnancy (Roulet et al., 1988; Fu et al., 2004). Some PA containing plants such as Crotalaria species (Bush tea, Rattlebox), Ilex paraguarensis (Mate tea), Symphytum species (Comfrey), Senecio species (Groundsel), Heliotropium species, and Compositae species (Indian herbs) that caused HILI are tabulated (Table 2). These herbs also injure cattle and house animals (Fu et al., 2004) and cause experimental hepatotoxicity in animals (Lin et al., 2011). PAs can be quantified in the serum of patients with HSOS (Lin et al., 2011; Larrey and Faure, 2011). The pathogenesis of PA hepatotoxicity has been elucidated in experimental studies, which showed the involvement of hepatic microsomal cytochrome P450 in the activation of PAs (Larrey and Faure, 2011).

Finally, herbal TCM products containing more than 19 g dose of Radix bupleuri may increase the hepatotoxicity risk (Lee et al., 2011); this dose dependency was confirmed in experimental animals and provided insights into some pathogenetic processes (Liu et al., 2014).

Juristical considerations

Black cohosh

Legal aspects of HILI case assessment rarely provide particular juridical and clinical challenges. Two court decisions merit attention, one relates to BC and the other one to kava. In 2005, a report was published representing a case of a 50 year old woman with fulminant liver failure and liver transplantation in assumed connection with the use of BC (Levitsky et al., 2005). A product liability action was filed by the patient after her recovery (Nebraska, 2006).

The decision of the judge answered the question whether in the specific case under discussion sufficient evidence establishes the herb as a generally or individually specific cause for the observed liver disease; for black cohosh, both aspects of causation were denied. General causation refers to the previously established hepatotoxicity by the same herb, but this was denied because of lack of convincing data. Specific causation refers to the case under discussion; this was also refuted on grounds of conflicting case data, poor case data quality, and numerous confounding variables (Nebraska, 2006). Our clinical diagnosis in this case was herpetic hepatitis and liver disease unrelated to BC or comedicated drugs, and CIOMS based assessment led to an excluded causality for both BC and comedicated drugs (Teschke and Schwarzenboeck, 2009): For this case of BC overdose, conclusions for an update of 22.12.2006 were provided (EMA, 2007): Worst case causality scoring would be possible, if comments of the expert would not be taken into account; because of the clinical experience of the expert and the requested obligatory causality in front of an American court, preference is given keeping the causality at a probable level (EMA, 2007). The conclusions of EMA are cloudy, difficult to reconcile, and ignore presented details, since the judge actually excluded both involved experts from expert testimony as to causation according to Daubert, rule 702 (Nebraska, 2006). As explained in detail, this rule requires that an expert be qualified to render a testimony on the subject, and that his testimony be reliable and relevant. None of the experts obviously met these and other required qualifications. Following the trial, USP reduced the causality for BC in this case from a probable to a possible level (Mahady et al., 2008). An erratum clarified the case conditions (Levitsky et al., 2008). Our clinical diagnosis in this case was herpetic hepatitis and liver disease unrelated to BC or comedicated drugs; CIOMS based assessment led to an excluded causality for both BC and comedicated drugs (Teschke and Schwarzenboeck, 2009). In retrospect, this court case calls for a thorough transparent documentation of HILI cases, associated with an unbiased expert opinion.

Kava

Kava was in the focus of another trial in connection with its marketing withdrawal by the German regulatory agency BfArM (Schmidt, 2014). Almost 12 years after the German regulatory agency BfArM issued an intermediate withdrawal of marketing authorization for products containing extracts of kava (Piper methysticum, Piperaceae) root and/or rhizoma (BfArM, 2002), the case has been reviewed by the German administrative court in Cologne (Court, 2014). According to the court's ruling on June 11, 2014, there was no justification for the ban of kava medicinal products issued by the German BfArM (Schmidt, 2014). The court ruled that based on available evidence, the benefit/risk ratio of kava medicinal products was confirmed as positive and must be considered as positive (Court, 2014), with credit given for previous reports (Schmidt, 2014) assessing causality in cases of assumed kava hepatotoxicity with the CIOMS scale (Teschke et al., 2008a, 2010; Teschke and Wolff, 2009; Teschke, 2010a). Credit was also given (Court, 2014) to the work of others (Sarris et al., 2013). Their double blind, randomized, placebo controlled trial was performed with a well defined noble kava drug that is on the market in Australia, showing both efficacy of kava in patients treated for their generalized anxiety disorders and lack of overt adverse reactions (Sarris et al., 2013), confirming kava efficacy based on a previous Cochrane study (Pittler and Ernst, 2003a). As a consequence of the court's ruling, German kava products have been formally restored to their market status on June 2002. As an update, BfArM appealed the court's ruling on June 30, 2014, but justification for the appeal has not yet been published (Schmidt, 2014). Clearly, the ruling is a major breakthrough, as it strengthens the legal certainty and predictability of regulatory decisions for herbal medicinal product manufacturers in general. It is also a call for the BfArM and other regulatory agencies to present transparent and appropriate clinical documentations of future HILI cases to be evaluated by clinically well trained regulatory assessors and external experts in the field, to be more self-critically, not to dismiss expert views to the contrary a priori, and providing rather than refuting original HILI case data in anonymous form to interested requesting scientists to assist in case evaluations.

Essentials of herbal product quality

Herb authentication and product identification

Good quality of herbal drugs and other herbal products is prerequisite for safe human use (Table 16) (Teschke et al., 2013c). However, shortcomings of herbal products are well documented, both in herbal TCM and herbal modern medicine. Herbal authentication was an issue for BC (Sarma et al., 2008; Health Canada, 2010) and various TCM herbs (Haller et al., 2002; Teschke, 2014; Teschke et al., 2014b). There also was considerable debate whether kava products used by patients with kava hepatotoxicity might have been of poor quality including inappropriate herb authentication. This led to a comprehensive assessment and a proposal for a Kava Quality Standardization Code (Teschke and Lebot, 2011). Actually, several guidelines exist already for Good Agricultural Practices (GAP) and Good Manufacturing Practices (GMP), applicable to medicinal plants and herbal medicines to ensure their product quality (WHO, 2000a, 2003, 2006, 2007b). Despite these official precautionary recommendations for quality improvements, batch and product variability is not unusual (Lebot, 2006; Schmidt, 2007; Teschke and Lebot, 2011; Teschke et al., 2013c). Violation of GAP or GMP rules also will result in herbal products that lack efficacy, safety, or both.

Proposal for international harmonization: requirements for regulatory approved herbal drugs.

Data are derived and adapted from a previous report (Teschke et al., 2013c).

When plants are considered for human use as ingredients of a herbal drug and dietary supplement, a clear definition and identification of plant family, subfamily, species, subspecies, and variety is mandatory, best done by a professional classical botanical description for any herb. Neglect may cause variation in plant family and species, contributing to the overall batch and product variability. Appropriate information should be provided by the manufacturers in the consumer's leaflet, thereby being available to the physician who suspects liver injury induced by a herbal product. The leaflet requires the name of the herbal product and the manufacturer's address who will provide additional information upon request. Therefore, all essential data of herb identification and the herbal product should be available before reporting HILI case details as spontaneous reports to the regulatory agencies or as case report publication. However, pitfalls are evident already at this stage of case evaluation.

In kava drug hepatotoxicity as an example, herb identification problems were evident. The manufacturers did not provide details on kava variety identification, so this specific information was missing in all spontaneous reports and case report publications (BfArM, 2002; Schmidt, 2007; WHO, 2007a; Teschke et al., 2008a; Teschke, 2010a,b, 2011). In the South Pacific region of origin, several hundred kava varieties exist—also called kava cultivars—and are grouped into noble, medicinal, and Two-Day varieties (Lebot, 2006; Schmidt, 2007; Teschke and Lebot, 2011; Teschke et al., 2011b). They differ in their kavalactone composition and their pleasant and unwanted, possibly toxic effects. In cases of suspected hepatotoxicity, it remained unclear which kava variety had to be incriminated. Interestingly, regulatory approval of kava drugs neither considered different kava varieties nor required respective labeling (BfArM, 2002). Thus, kava hepatotoxicity remains unexplained.

Other problems of herb or product identification have been described in detail in various cases of initially suspected herbal hepatotoxicity (EMA, 2007; WHO, 2007b; Mahady et al., 2008; Teschke et al., 2011a,b,d,e). Incomplete herb description complicates accurate association of herbs with liver injury and allows only general assumptions (Teschke, 2010c; Teschke et al., 2013e,f). Besides overall herb descriptions, the brand name of the herbal product has been provided in only a few case reports, and data for manufacturer, plant part, and extraction solvent normally was fragmentary (Teschke et al., 2011e, 2013c). For instance, the rate of undetermined herbal products was 10/16 cases (63%) among published case reports (Teschke et al., 2011e). This high rate questions the validity of any causality attribution. Additional problems arise from herbal mixtures, in which individual ingredients are not specified (Teschke, 2010c; Teschke et al., 2011d,e). Again, case reports have been published as HILI even if the patients were not sure whether they used a herbal product at all (Teschke et al., 2011e).

For GC, case reports assumed causality for hepatotoxicity in all 21 cases, but details on the GC product were fragmentary (Teschke et al., 2012e). Out of these 21 cases, seven patients used a GC monopreparation and four patients a GC polyherbal product, brand names and manufacturers were known in only nine patients. Fears of liability may contribute to the restriction of detailed product specifications by the authors. On the other hand, the regulatory agency did not hesitate to provide all relevant data of GC products from spontaneous cases of GC hepatotoxicity (Teschke et al., 2011a).

Therefore, unless complete data for herbal identification, ingredients, and name of the herbal product are provided in each HILI case, a valid causality assignment is not realistic. Identification problems are evident also in most HDS providing little specific information (Navarro et al., 2014; Robles-Diaz et al., 2015).

Plant part specification

Reports on HILI rarely provide details of the plant part used, ignoring specific toxic properties attributable to different parts of a plant. The regulatory recommendation for kava drugs was to use its peeled rhizome (Teschke and Lebot, 2011). In various assumed HILI cases by kava it remained unclear, whether also unpeeled rhizomes, peeled and unpeeled roots, and/or stem peelings were used, hampering evaluation of the causative agent of kava hepatotoxicity (WHO, 2007a; Teschke, 2011). For the U.S. FDA, peeled kava rhizomes were recommended for kava supplements (Teschke and Schulze, 2010), and according to the Australian Therapeutic Goods Administration, the commonly used medicinal kava products are derived from peeled rhizomes (Sarris et al., 2009). Plant part specification can be a major regulatory, agricultural, manufactural, pharmaceutical, and clinical issue (Teschke and Lebot, 2011).

Solvents and solubilizers

Herbal drugs and supplements are specified as extracts that are either water based or prepared from organic solvents like ethanol or acetone, but regulatory advice is often lacking (Teschke and Lebot, 2011). Thus, herbal extracts will substantially differ in their composition depending on the solvent. In addition, numerous solubilizers like macrogol, craspovidon, mentha oil, methyl acrylic acid polymer and polysorbate polyols may be included in herbal products to facilitate gastrointestinal uptake (Teschke, 2010b). Therefore, solvents and solubilizers may influence the composition of chemicals in the herbal product and selectively affect the bioavailability for the liver as the target organ. These variations hamper causality attribution in suspected HILI cases, leading to the recommendation that kava drugs and supplements should be water based extracts lacking any solvents or solubilizers (WHO, 2007a; Teschke and Schulze, 2010; Teschke et al., 2011b).

Misidentifications, impurities, and adulterants

For herbal product quality, not only plant misidentification but also contaminants, impurities and adulterants still remain key problems (Kang-Yum and Oransky, 1992; Espinoza et al., 1995; Gertner et al., 1995; Huang et al., 1997; Ko, 1998; Ernst, 2002; Estes et al., 2003; Lebot, 2006; Schmidt, 2007; Seeff, 2007; WHO, 2007b; Mahady et al., 2008; Navarro, 2009; Teschke et al., 2009, 2011c,e; Health Canada, 2010; Larrey and Faure, 2011; Teschke and Lebot, 2011). Adulterants are not uncommon in herbal TCM mixtures; they usually consist of synthetic drugs to provide or fortify product efficacy. Although rarely addressed by analytical approaches in patients with actually reported HILI, Health Canada was the only regulatory agency with recently reported interest in the analytical assessment of herbal products to evaluate quality, providing evidence for misidentification of herbs in some products and presenting results that the accused herb was not present in the herbal products used by the affected patients (Health Canada, 2010).

It remains to be established to what extent misidentifications, impurities, and adulterants are responsible for individual HILI cases. For instance, possible causality for hepatotoxicity cases of the herbal TCM mixtures Chaso and Onshido was ascribed to N-nitroso-fenfluramine, found as adulterant in these slimming aid products that had been produced in China and sold in Japan (Adachi et al., 2003). However, there is only little clinical or experimental evidence for a potential hepatotoxicity by this adulterant (Kanda et al., 2003a,b; Lau et al., 2004). It rather appears that green tea as ingredient was the causative agent if supplied as extract (Teschke, 2014).

Misidentification may create major clinical challenges and harm dramatically the health of consumers, shown for the following cases (Teschke, 2014). Until 2008, overall 41 cases from China with HSOS, the former HVOD, were reported and causally attributed to the herbal TCM Jing Tian San Qi (Sedum aizoon, syn. Stonecrop) (Wu et al., 2008), but causal attribution to Sedum aizoon was obviously incorrect. Sedum aizoon lacks PAs, and when applied to experimental animals, HSOS did not eme.g., Lin et al., 2011, suggesting that a herb containing PAs likely is reponsible for the reported cases (Wu et al., 2008). In line with this is another hepatotoxicity case from Hong Kong with HSOS that initially also was ascribed to Sedum aizoon, but it turned out to have been caused by the herbal TCM Tu San Qi (Gynura segetum) (Lin et al., 2011). The name and appearance of Sedum aizoon is similar to the one of Gynura segetum, but botanical differentiation was considered possible for the eye of experts (Lin et al., 2011). Comparative studies with both herbs provided clear supportive evidence for Gynura segetum as culprit for additional cases of HSOS as compared to Sedum aizoon. Respective studies in mice showed that Gynura segetum as the PA containing herb but not Sedum aizoon lacking PAs causes experimental HSOS as assessed by liver histology results (Lin et al., 2011). In an earlier experimental study, a model of the hepatic veno-occlusive disease was established by PAs derived from a herb described erronously as Sedum aizoon (Gao et al., 2006), which again does not contain PAs (Lin et al., 2011; Gao et al., 2012; Wang and Gao, 2014). This suggests that the described experimental model (Gao et al., 2006) was due to the action of a herb containing PAs, most likely Gynura segetum (Lin et al., 2011; Gao et al., 2012; Wang and Gao, 2014), rather than to Sedum aizoon lacking PAs (Gao et al., 2012). Based on these well founded considerations, evidence for a hepatotoxic potential of Jing Tian San Qi is lacking. The herbal TCM Sedum aizoon should therefore not be tabulated any more as hepatotoxic herb, as done until recently (Teschke et al., 2012h).

Gynura segetum was involved in other cases of herbal misidentification. In two Chinese women, HSOS emerged, which was induced by PAs of the herbal TCM Gynura segetum (syn. Ju Shan Qi, Ju Ye San Qi, Shan Chi, San Qi Cao, Shan Chi, Shan Chi) (Dai et al., 2006). Additional six cases were earlier suspected (Kumana et al., 1983, 1985); in at least four cases, the culprit was the PA containing herb Heliotropium lasiocarpum rather than Gynura segetum (Culvenor et al., 1986).