Its prospective application enables the CIOMS scale to provide an early causality grading for patients with suspected DILI or HILI; its results can be adapted further to diagnostic and therapeutic measures. This scale is easily used as a bedside tool at a time the disease is developing (Tables 1 and 2). Results do not depend on expert opinion and are quickly available for trained physicians to decide whether DILI and/or HILI should be considered as relevant differential diagnoses due to their clinical experience. Assessment is best started on the day of suspecting DILI or HILI, with a continuous update of the required data and a change in the diagnostic and therapeutic concept if needed. Finally, a complete data set for presentation to regulatory agencies, expert panels and eventually for publication is obtained[24,30-33], including a checklist with additional data helpful in overall case evaluation and causality assessment (Table 3). Therefore, the CIOMS scale should be considered as a standard for causality assessment of DILI and HILI, both for the attending physician and later evaluation stages. Using one single assessment method at all evaluating levels allows comparison of different assessment outcomes.

Liver specificity is a hallmark of the CIOMS scale, in contrast to liver unspecific causality assessment methods or ad hoc approaches[4,24]. The CIOMS items are specially tailored to liver injury and not applicable to liver unrelated adverse drug reactions[24]. All current core elements of hepatotoxicity are considered in the CIOMS scale (Tables 1 and 2): time to onset of increased liver values or symptoms from the beginning and cessation of the drug/herb; course of liver enzymes after cessation; risk factors such as alcohol, age and pregnancy; comedication with other drugs/herbs; search for alternative causes, previously known drug/herb hepatotoxicity; and response to unintentional reexposure[21-25] based on specific criteria (Table 4). The individual items are transparent and facilitate quick and precise answers.

The CIOMS scale is structured and all its items undergo quantitative rather than qualitative assessment and scoring (Tables 1 and 2)[4,5,10,14,21,23,24,29]. Each item is weighted with specific scores based on the answer. The sum of the individual scores gives a final score that may range from -9 to +14 points, allowing for sufficient discrimination. The final score provides causality levels for the individual synthetic drug or herb as highly probable, probable, possible, unlikely or excluded (Tables 1 and 2)[12,24,30-32].

The international CIOMS expert panel defined liver injury in its consensus report as increased alanine aminotransferase (ALT) and/or alkaline phosphatase (ALP) activities of at least 2N, with N as the upper limit of normal[21]. Conversely, the consensus of the international DILI Expert Working Group with participants from Europe, the United States and Japan raised the ALT cut off point to 5N or 3N if total bilirubin values exceeded 2N and considered the 2N of ALP as an appropriate definition criterion[14]. Whereas the DILI Expert Working Group recommendations were based on expert opinion alone[14], those of the CIOMS expert panel were derived from both expert opinion and assessment of reference reexposure DILI cases[21,22].

Raising the ALT cut off to 5N increases the specificity of the hepatotoxicity causality assessment[24], eliminates false positive cases and substantiates hepatotoxicity causality at a higher level of probability[16,24]. The lower threshold of ALT > 2N will include multiple cases with nonspecific enzyme increases and requires more stringent exclusion of causes unrelated to drug(s) and herb(s)[24]. Also for low threshold N values, the inclusion rate of alternative diagnoses must be higher; false positive fulfilment of a hepatotoxicity definition results in high numbers of misattributed cases due to overdiagnosing and overreporting[8,12,17,24,34-48]. This phenomenon is illustrated in a recent HILI study where initial ALT values were available in only 8/22 cases (36%), including 3 cases with a range of 50-69 U/L serum activity[36]. None withstanding, regulatory assessment attributed a possible causality for the incriminated herb to all 22 cases[36,42]. In other spontaneous case collections, initial ALT values were available in 5/24 cases (21%)[35], 19/22 cases (86%)[37], 12/15 cases (80%)[38], and 7/13 cases (54%)[39]. The corresponding figures for ALT in published case reports of HILI were 16/16 cases (100%)[35], 21/21 cases (100%)[32], and 5/8 cases (63%)[33]. ALT values were included in DILI reports for amoxicillin/clavulanic acid, troglitazone, pioglitazone and montelukast in 11% to 88% of the cases, which were not further scored for causality by the Drug Induced Liver Injury Network (DILIN)[6]. ALT underreporting is therefore an issue for both DILI and HILI.

Other arguments merit further considerations. An ALT cut off point of 5N may not be applicable to some types of chronic liver injury like methotrexate liver fibrosis or nodular regenerative hyperplasia; misinterpretation is also possible in some forms of acute liver injury by mitochondrial toxicity in cases of valproate or fialuridine hepatotoxicity[14]. Aspartate aminotransferase (AST) activities may be used instead if ALT activities are unavailable[14,44,45] and other pathologies for AST increases are excluded[14]. ALP increases should be paralleled by γ-glutamyltranspeptidase (γGT) to rule out isolated increases of ALP activities due to bone rather than hepatobiliary disease. However, γGT alone is not an appropriate parameter for liver cell injury[14,36], contrary to published claims[42]. In addition, isolated hyperbilirubinemia is not DILI or HILI specific and may be caused by Gilbert’s syndrome[1,14].

The CIOMS scale takes into account divergent laboratory constellations of the liver injury pattern in the hepatocellular and the cholestatic type of liver injury and therefore provides two different subscales[21,23,24] for the hepatocellular type of injury (Table 1) and for the cholestatic or mixed type (Table 2). These types are differentiated by the ratio R, calculated as the ALT/ALP activity measured at the time liver injury is suspected, with both activities expressed as multiples of N[21,24]. Injury is hepatocellular, if only ALT > 2N, alternatively if R ≥ 5; cholestatic injury is assumed, if only ALP > 2N or R ≤ 2; mixed damage is prevalent, if ALT > 2N and ALP is increased, with R > 2 and R < 5[21,23,24]. Of note, R may vary during the later course of the liver injury independent from the initial attribution of damage type.

Clear challenge criteria are defined with a time frame between beginning of the drug/herb use as the first day of intake and the onset of increased liver enzymes or symptoms at the time of ongoing use, with a high score for 5-90 d and a lower one for < 5 d or > 90 d (Tables 1 and 2). If drug/herb use has been terminated prior to the onset of challenge criteria, then this specific condition must be considered and scored exclusively. Scoring is only possible when the onset occurs within 15 d after cessation for the hepatocellular injury (Table 1) or 30 d for the cholestatic or mixed type (Table 2), a longer interval commonly excludes causality (Tables 1 and 2). An exemption is provided for slowly metabolized chemicals like amiodarone, leflunomide and clavulanate[44,45]; no definitive time frame can be provided in these cases due to varying half lives. The time frame of challenge and latency period were neither specified nor individually scored by other causality assessment methods[4,24], including the DILIN method[49,50].

Precise dechallenge criteria are cornerstones of the CIOMS scale and facilitate causality assessment (Tables 1 and 2). In analogy to the periods mentioned, the physician can easily determine relevant future time points for repeated liver enzyme tests. When dechallenge data are missing in retrospective analyses, the CIOMS scale considers this and provides 0, but not negative points, so that the overall score may still present a probable causality level. Of note, the dechallenge time frame was not specifically considered or scored by the DILIN method[49,50] or by virtually any of the other methods[4,24].

The consensus report of the international CIOMS expert panel considered alcohol and age ≥ 55 years as risk factors each scoring +1 point (Tables 1 and 2)[21], as suggested by DILI cases with positive reexposure[21,22]. The international DILI Expert Working Group specified alcohol intake of > 2 drinks per day (> 14 units/week) in women and > 3 drinks per day (21 units/week) in men as the lower threshold for alcohol intake as a risk factor (Tables 1 and 2)[14]. This limit is in line with the recommendations of NIH LiverTox equalling 1 drink to 10 g ethanol[44,45].

The impact of including alcohol as a risk factor on the overall CIOMS scoring was negligible as only 9/146 patients (6%) of a HILI study cohort were allotted an alcohol related scoring point (Table 5)[12,30,32,36-39,46-48]. In these 9 patients, CIOMS causality grading was changed in only one case (patient 16) and unchanged in the other 8 cases (patients 2, 13, 14, 17, 20, 21, 22, 32) (Table 6). In the single case (patient 16), alcohol as risk factor raised the overall CIOMS scoring from 0 to +01 point, i.e., from excluded to unlikely causality (Table 6). Therefore, alcohol per se as risk factor upgrades the CIOMS causality level in virtually none of the cases in this study cohort.

Age ≥ 55 years was as a risk factor in 44/146 cases (30%) of the analyzed HILI study cohort (Table 5)[12,30,32,36-39,46-48]. In 35/44 patients, the overall CIOMS causality grading remained unchanged whether or not age as a risk factor was included in the CIOMS scale scoring (Table 6). Deletion of age as a risk factor reduced the overall CIOMS grading by one causality level in 9/44 patients, i.e., from unlikely to excluded in 5 cases (patients 1, 3, 15, 42, 43), from highly probable to probable in 1 case (patient 23), and from probable to possible in 3 cases (patients 9, 25, 34). Therefore, within this cohort the risk factor age upgraded the causality levels only marginally, which appears to have no clinical relevance. Overall, age as a risk factor has a limited impact on the final causality gradings by the CIOMS scale.

Risk factors are not considered and/or not scored by various other methods[4,24,49,50], including the DILIN method[49,50]. Conversely, the international DILI Expert Group also accepts the risk factors defined in the CIOMS scale, with modified specifications and limitations, if risk factors for hepatotoxicity are present in addition to those listed in the CIOMS algorithm[14].

Concomitant drugs and herbs are individually assessed for temporal association and hepatotoxic potency (Tables 1 and 2). For reasons of comparison and transparency, each comedicated drug or herb requires a separate analysis by the complete CIOMS scale. This is feasible and easily tabulated (Table 7)[30,47,48]. In patients with multiple drug or herb intakes, causality should be attributed primarily to the product with the highest score.

In this section, the CIOMS scale considers the clinically most relevant alternative causes (Tables 1 and 2). There is no difference in alternative causes made between the two types of liver injury, avoiding the need of subsequent reassessment, if the laboratory based typology changes during the clinical course[45]. Complications of underlying disease(s) are exemplified, such as sepsis, autoimmune hepatitis, chronic hepatitis B or C, primary biliary cirrhosis and sclerosing cholangitis and genetic liver diseases (Tables 1 and 2), in accordance with recent suggestions[45]. Other rare alternative causes are included in a checklist of differential diagnoses as a reminder for the clinician in case of unclear clinical diagnosis (Table 3)[24].

To improve its performance when used as an investigational tool, criteria for competing liver injury causes have been proposed for the CIOMS scale[14,29,44,45,49,50] and were included in the updated CIOMS scale (Tables 1 and 2)[24]. This update ensures correct diagnosis of alternative causes but was limited to details of hepatitis serology and hepatobiliary sonography, as specified by the current knowledge in the field and adapted to actual diagnostic methods[23,24]. The update of the original CIOMS scale substantially improved specificity, i.e., exclusion of alternative causes by hepatitis serology. HBsAg and HBV-DNA quantification were added to distinguish HBV infection from immunization, as was hepatitis C virus (HCV)-RNA to correctly assess HCV infections. Also, clinical and/or biological parameters for cytomegalovirus (CMV), Epstein Barr virus (EBV) or herpes simplex virus (HSV) infection were vague or unknown at the time of compilation[21] but specified in the updated CIOMS scale; also included and specified were infections by hepatitis E virus (HEV) and varicella zoster virus (VZV) (Tables 1 and 2)[24]. Specific diagnostic criteria include polymerase chain reaction 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), if these tests were indicated clinically (Tables 1 and 2). 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[32,33,36-39]. Therefore, the updated CIOMS scale was validated and there is no need for further validation of the updated CIOMS scale versus the original CIOMS scale.

Hepatotoxicity listed in the product information sheet must be checked; in addition, a quick literature search in PubMed will be sufficient to determine whether the observed reaction has been published before. Appropriate information may also be obtained from the NIH LiverTox database[44,45].

To classify an unintentional reexposure test as positive, few criteria are required (Tables 1 and 2), as specified (Table 4)[21,22,24,26,27]. Although reexposure is an important domain, probable causality gradings with the CIOMS scale are achievable even in the absence of a reexposure (Table 7)[12,17,30-32,37].

Each item of the CIOMS scale receives an individual score and the sum of the individual scores provides the final score for the patient (Tables 1 and 2). With +14 down to -9 points, there is a wide range of the final scores, leading to the following causality levels: ≤ 0 points, excluded causality; 1-2, unlikely; 3-5, possible; 6-8, probable; and ≥ 9, highly probable (Tables 1 and 2)[21].

Cases with positive reexposure tests were proposed for validation of the CIOMS scale and used as gold standard[22]. Articles from two databanks were compiled with liver injury confirmed by a positive rechallenge. The mandatory information for inclusion in this series contained the type of liver injury, time interval between administration of the drug and occurrence of the reaction, and results of the positive response to readministration of the drug, in accordance with the conclusions of the International Consensus Meeting on drug induced liver injuries. For the final validation, 49 cases and 28 controls were assessed, as described in detail[22]. Most importantly, the discriminative power of the score was quantified in terms of sensitivity, specificity and predictive values. The cut off point was offset to maximize the combined sensitivity and specificity. Using +5 points as the cut off, sensitivity was 86%, specificity 89%, positive predictive value 93%, and negative predictive value 78% for the CIOMS causality assessment. In another study with 81 cases and 46 controls, sensitivity was 78% and specificity 100% for the CIOMS scale[51], confirming the validation of the CIOMS scale.

The interrater reliability of CIOMS assessment was good by one group[52] but mediocre by the DILIN group[49]. In the latter report, however, 40 cases going back to 1994 were studied. Uncertainties arose from numerous missing, incomplete or outdated medical reports and charts, especially for older cases. In particular, there were high rates (28%) of preexisting liver diseases like chronic hepatitis C virus infection, hemochromatosis and unspecified cirrhosis. Liver sonography was reported in 26/40 cases and found abnormal in 15/26 (58%). These data were nevertheless described as ‘‘best-case scenario’’[49]. Considering these limitations and numerous confounding variables, poor case data quality likely results in mediocre assessment quality, including low interrater reliability[49]. Moreover, problematic data presentation by the principal assessor to external reviewers may have influenced the results as the external reviewers received only a subset of the case report forms and had no access to the original data of the cases[49]. Of interest, no proof has been provided that an expert group opinion improves the CIOMS assessment evaluation, at least according to recent comments and studies[11,49]. In another study comparing the CIOMS scale with the DILIN method, there was considerable interobserver variability in both methods[50].

The CIOMS scale for hepatotoxicity assessment in its original or updated form[4,5,10,14,23,24,28,29] has been extensively used in epidemiological studies, clinical trials, case reports, case series, regulatory analyses and genotyping studies[4]. Additional efforts are still needed to reevaluate causality in most HILI reports for 60 different herbs and herbal products[53]. CIOMS based results were published by the DILIN group[49,50] and by the European Medicines Agency (EMA)[54]. Individual studies[10,16,55,56], the NIH LiverTox[44,45], the international DILI Expert Working Group[14], the Spanish Group for the Study of Drug-Induced Liver Disease[29], and the Hong Kong Herb-Induced Liver Injury Network (HK-HILIN)[57] provided further support for the CIOMS scale.

Among various causality assessment methods, the original and updated CIOMS scales were the preferred tools in cases of DILI[28] and HILI (Table 8)[5], seen for 573 cases from 23 HILI reports evaluating alternative causes[12,32,34,36-39,42,43,47,48,54,57-67].

CIOMS based assessments should be reported or published as an original data set suitable for subsequent and independent assessments, rather than as final scores and corresponding causality levels, to improve data transparency. Scientists, editors and reviewers should strive to obtain appropriate CIOMS based details for all DILI and HILI case reports. This can easily be achieved since the CIOMS scale provides all items in tabulated form for each individual case (Tables 1, 2 and 7). These forms may be communicated as a spontaneous report to regulatory agencies and expert panels or presented for publication to scientific journals as a case report[24,30,31] or case series[12,24,32]. This tabulation is a good basis for further regulatory or scientific assessments and discussions. For regulatory and expert based assessments, there is no need for other causality assessment algorithms to be used subsequently since CIOMS based data are also amenable to regulatory and expert panel evaluations.

The CIOMS scale resulted from intensive expert discussions[21], integrating medical progress and improving the initial qualitative RUCAM[26] and the qualitative CIOMS method[27]. The qualitative RUCAM represented the first objective attempt to assess causality in drug induced liver injury and considered some characteristic features of liver injury, but it had a qualitative rather than a quantitative approach[26]. As an improved version of the qualitative RUCAM[26], the qualitative CIOMS method differentiated the hepatocellular, the cholestatic and the mixed type of liver injury[27]. However, both the qualitative RUCAM[26] and the qualitative CIOMS method[27] were not quantitative, as opposed to the current quantitative CIOMS scale[21] that is now the preferred tool[24].

The scale of Maria and Victorino (MV)[68] was developed to improve upon the CIOMS scale by deleting laboratory items and adding clinical elements, along with simplifying and changing the relative weight of elements in their algorithm[23,44,45]. No data are available for specificity, sensitivity, positive and negative predictive values for the MV scale[68]. Compared to the original CIOMS scale[21], the MV scale[68] showed shortcomings and the results are not equivalent, causing major concern[10,14,23,24,29,44,45,69-71]. This may explain why the MV scale was used in a few DILI studies[1,72,73], but not in 38 other publications of DILI cases[28] or in 23 publications of HILI cases[5]. The MV scale is not commonly recommended for assessing causality in assumed DILI and HILI cases and is certainly no substitute for the CIOMS scale[24].

The TTK scale[25], named for the first three authors Takikawa, Takamori, Kumagi et al[74], is a modification of the CIOMS scale[21] with different evaluations of the chronology, exclusion of comedication, inclusion of the drug lymphocyte stimulation test (DLST) and of eosinophilia in their assessment system[74,75]. The TTK scale is widely used in Japan[74], as recently reviewed[75]. In other countries, this scale is not or rarely considered[5,10,14,24,28,29,44,45,76]. Limited access and lack of standardization have prevented general clinical use of the DLST and consequently TTK scale applications outside Japan[29]; this may be due to methodological difficulties with false positive and false negative cases in the DLST[25,75]. For clinicians, the TTK scale cannot replace the CIOMS scale[25].

The DILIN method provided by the DILIN group requires an expert panel[3,6,11,24,44,45,49,50,77,78], in contrast to the CIOMS scale[21,24]. Consequently, the DILIN method is of limited availability to physicians in need of early results for therapeutic decisions[24]. In particular, the DILIN method is not an appropriate substitute for the CIOMS scale, nor are other expert panel based approaches[24]. This includes the novel Causality Assessment Tool (CAT) specifically designed for herbs and dietary supplements (HDS), which was presented as an abstract[15]. As opposed to CIOMS based results with transparent data presentation (Table 7)[12,30,32,36-39,47,48,71], publications based on the DILIN method lack transparency for individual cases regarding assessed and scored items since only final causality levels are published without details and thereby open for discussions, not allowing valid conclusions[3,6,11,49,50,77,78]. The DILIN method also lacks data on specificity, sensitivity and predictive values, as an expert opinion based method no items can be validated. Individual weighing and scoring of items remain undisclosed and undiscussed, hampering thorough analysis of assessment results by the DILIN method.

In contrast to the liver specific core elements of the original and updated CIOMS scale (Tables 1 and 2)[4,21-24], numerous causality algorithms are liver unspecific[4,24,76,79,80], including the Naranjo scale[81], the World Health Organization (WHO) global introspection method as the WHO method in short[82], and the KL method of Karg and Lasagna[83]. Particularly intensive discussions focused on the Naranjo scale[4,24,25,84-87], the WHO method[4,24,84,87], and the KL method[24,25]. All these methods are obsolete for causality assessment of assumed hepatotoxicity as they lack liver specificity and do not consider hepatotoxicity characteristics.

Retrospective analysis of case data is problematic and may require some assistance evaluating the CIOMS items[14,44,45]; unselected and sometimes undefined, low quality data have to be adapted into a structured algorithm like the CIOMS scale. Therefore, physicians should prospectively use the CIOMS scale, which then may provide complete case data (Table 7)[30].

Missing ALT dechallenge data are factored as 0 points given (Table 1); this condition has been interpreted as a limitation of the CIOMS scale[14]. Retrospective studies commonly lack dechallenge results[6,12,32,34-39] which are included in prospective evaluations (Table 7)[30]. CIOMS performs inaccurately in acute liver failure and liver transplantation if liver values are not available within 30 d after cessation of the incriminated drug or herb. Under these circumstances, 0 but not negative points are credited due to lacking ALT data (Tables 1 and 2).

The CIOMS scale includes only the risk factors of alcohol and age ≥ 55 years[21]. Diabetes, metabolic syndrome, sex, ethnicity and body mass index[14], as well as genetic predisposition[29], are also proposed as potential risk factors; the lack of inclusion in the CIOMS scale has been considered as a limitation[14,29]. However, these factors have not been validated as risk factors; their inclusion into the CIOMS scale requires evidence as independent contributors and a subsequent new validation.

It may be argued that rare alternative causes were not listed in the CIOMS scale (Tables 1 and 2) but this shortcoming was compensated for by the checklist for numerous rare liver diseases as a reminder for the clinician (Table 3)[24].

Safety labels are available for both synthetic and herbal drugs but rarely for other herbal products. This shortcoming of the CIOMS scale may be compensated by a thorough search for prior publications of hepatotoxicity by herbal products; published reports provide an even higher scoring than information obtained only from safety labels.