FormalPara Key Points

We identifed 18 publications providing advice on drug selection and dose adjustments of drugs in patients with liver cirrhosis by conducting a systematic literature search.

There were 1145 dose recommendations for 481 active substances identified. There were 706 recommendations for 316 substances sufficiently specific for application in clinical practice.

The recommendations for 22 active substances were sufficiently specific and consistent across multiple publications. Recommendations for six active substances were also consistent with the respective Summaries of Product Characteristics.

1 Introduction

A missing dose adjustment or administration of contraindicated drugs is a common drug-related problem [1, 2]. Liver cirrhosis with advanced fibrotic remodelling of the parenchyma as the terminal stage of several chronic liver diseases often leads to an impaired liver function, classified into three grades of severity by the Child-Pugh score considering objective laboratory (bilirubin, albumin, international normalised ratio) as well as subjective parameters (encephalopathy and ascites) [2]. Consequently, doses of drugs metabolised by the liver have to be adjusted to their altered pharmacokinetics [3, 4]. It can be assumed that 33% of patients with liver cirrhosis experience at least one adverse drug reaction during their hospital stay because of missing dose adjustments [5], which results in a considerable prolongation of the hospitalisation in affected patients [5]. Sixty-eight percent of these adverse drug reactions are considered preventable [6]. One known risk factor for the occurrence of adverse drug reactions in cirrhotic patients is polypharmacy [5].

As Summaries of Product Characteristics (SmPCs) usually provide only vague information and standardised recommendations are missing for patients with hepatic impairment [4, 7], dose adjustments remain a major challenge in clinical practice. Moreover, the term “liver cirrhosis” is often not explicitly mentioned in the SmPC of drugs but instead replaced by soft ambiguous terms such as “liver damage”, “liver disease” or “liver dysfunction”, confusing the prescriber as to whether the individual patient actually falls into this category [7, 8]. As the prescription of drugs for patients with liver cirrhosis therefore often depends on the individual experience of the prescribing physician, large differences in drug therapy may occur. There is scientific literature that attempts to assist in the selection and dosage of suitable drugs in this population [9,10,11]. However, it has not resulted in joint recommendations so far.

We therefore conducted a systematic literature search to identify the literature that aimed to provide recommendations on drug selection and dose adjustments of drugs in patients with liver cirrhosis. In this context, it was particularly important to analyse which active substances were covered by the recommendations, whether the recommendations were sufficiently specific and clearly formulated to warrant simple application in clinical practice, and whether the recommendations of the different references were consistent to avoid further confusion.

2 Materials and Methods

2.1 Step 1: Structured Literature Search

A structured literature search was carried out to compile an exhaustive list of recommendations for the selection and dosage of drugs in liver cirrhosis. All publications indexed in the PubMed/Embase database that were available at the cut-off date of October 2020 were screened using a combination of search terms for both liver cirrhosis and drug dose adjustments (see Fig. 1). The search was performed twice with different combinations of search terms (Box 1 and Box 2) and duplicates were removed. Broad search terms were used to identify as many relevant publications as possible.

Fig. 1
figure 1

Original search strings in PubMed as part of the literature search

Full size image

2.2 Step 2: Selection of Literature

All publications whose characteristics met the inclusion and exclusion criteria (see Table 1) were included. Of note, only peer-reviewed reviews, systematic reviews, guideline articles and consensus procedures in English were considered. At first, titles were screened. Then, abstracts belonging to all relevant titles were analysed. If no abstract was available, the full text was reviewed directly. Finally, full texts of all publications that were considered relevant were analysed. Dose recommendations that were based solely on the SmPCs were excluded, as we were searching for recommendations going beyond the formal specifications. Additional literature identified by checking the references of the analysed full texts were also checked for relevance (“forward snowballing”) [12] to render the search as exhaustive as possible. The selected literature was verified by two further reviewers, who were chosen based on their experience in the field of evidence-based medicine and structured literature searches and who checked independently from each other the relevance by screening the title, abstract and full text. If no consensus on inclusion and exclusion of a publication could be reached, the decision was made by a majority vote after a joint group discussion.

Table 1 Inclusion and exclusion criteria
Full size table

2.2.1 Specificity of Dose Recommendations

Only publications that made unambiguous statements on at least one active substance were included after a full-text analysis. The inclusion criterion of a “specific dose recommendation” (Table 1, E3: results) was considered to be fulfilled if (1) a specific dose recommendation was provided (in [%] or in the corresponding unit, e.g. [mg]), (2) a precise indication of the prolongation or shortening of a dose interval was given (in hours [h] or alternating schemes, e.g. “every other day”), (3) it was stated that no dose adjustment was required, (4) it has clearly to be avoided or (5) an alternative active substance was indicated that should be preferred. Additionally, if a publication gave dose recommendations differing for each Child-Pugh class, the criteria had to be fulfilled for each of the three classes. If a publication made two statements within a recommendation (e.g. on dose adjustment and interval extension) or if recommendations were provided for different formulations of an active substance or different route of administration, both statements were checked separately for specificity.

The following scenarios were not categorised as specific: (1) unspecific recommendations such as “precaution” or “start as low as possible” as they could be interpreted in different ways, (2) the general recommendation of therapeutic drug monitoring if no appropriate starting dose was provided and (3) whenever authors were not able to provide recommendations because of a lack of data.

2.3 Step 3: Assessment of Reporting Quality

Because the identified literature showed a high degree of heterogeneity in terms of reporting quality, an assessment of quality criteria was carried out to ensure that the formulation of dose recommendations was based on adequate and transparent criteria. Therefore, all included publications were categorised based on their objectives as (systematic) reviews providing an overview over recommendations (checked against the Preferred Reporting Items for Systematic Reviews and Meta-Analyses [PRISMA] statement [13]), guidance documents or guidelines that adopt clearly formulated recommendations for specific use in clinical practice (checked against the AGREE statement [14]) and reports on Delphi procedures (checked against the CREDES statement [15]). The objective of the publication stated by the authors was the basis for the classification as a guideline or review and the corresponding reporting guideline used for the assessment of the quality of the publication. The objectives were predominantly stated in the introduction, but sometimes also in other sections.

Subsequently, the publications were assessed for transparency and completeness in reporting by reconciliation with all items listed in the applicable statement checklist (PRISMA, AGREE or CREDES). The absence of listed items was assumed to affect the quality of the entire publication if this impaired the reproducibility or the understanding of how the recommendations on dose selection and adjustment were generated. Minor (incomplete but still directly comprehensible reporting on a literature search) or moderate deficiencies (incomplete reporting on a literature search, but indirectly reproducible) led to downgrading of the quality of a reference, serious deficiencies (missing information that prevented the recommendations from being reproduced) led to exclusion.

The categorisation of the references and the quality assessment according to the PRISMA, AGREE or CREDES statements was carried out according to a four-eye principle involving a second reviewer experienced in the field of evidence-based medicine and structured literature searches. Discrepancies in the assessment were jointly discussed. If no consensus could be reached, a third reviewer made the decision after a joint group discussion.

2.4 Step 4: Assessment of the Evidence Level

Additionally, the evidence level of the references was classified on the basis of the literature covered according to the Oxford Centre for Evidence-Based Medicine (CEBM) Levels of Evidence [16]. All evidence levels according to the CEBM were included. Any minor, moderate or serious deficiencies noted under step 3 were added to the level of evidence to get a complete picture of the quality of the individual references. Minor deficiencies were marked by a minus sign after the indicated level of evidence. Moderate deficiencies were marked with two minus signs (see Table 2).

Table 2 Overview of the 20 provisionally included references
Full size table

2.5 Step 5: Extraction of Information for Each Active Substance

All available specific dose recommendations were taken from the finally selected publications. We particularly considered whether a publication provided differentiated recommendations according to Child-Pugh severity grades or simply formulated general recommendations for dosages in liver cirrhosis. Subsequently, a mapping according to the Anatomical Therapeutic Chemical/Defined Daily Dose (ATC/DDD) Index of the World Health Organization (Version 2022) [17] of all active substances was performed to assign them to the associated indications as some active substances can be used in different contexts and dose recommendations are based on the specific indication.

If different specific statements from different publications were available for the same active substance, we subsequently compared the statements for consistency. For active substances with consistent recommendations, an additional comparison with the corresponding SmPC was made. We selected a random sample of five active substances (carvedilol, fentanyl, morphine, paracetamol/acetaminophen and verapamil) with specific but inconsistent recommendations to compare the data basis of the detected divergence. For active substances for which only one recommendation could be identified, no evaluation via a comparison was possible.

3 Results

3.1 Peer-Reviewed Literature Search and Selection

The literature search identified 6031 publications whose titles were reviewed. There were 5881 studies excluded, as they did not meet the inclusion and exclusion criteria. Of the remaining 150 publications, abstract screening resulted in 116 further publications being excluded. After full-text screening of the remaining 34 references, another 16 publications were excluded, resulting in 18 publications being initially classified as relevant. The review of the bibliography of the publications, for which full texts were analysed, led to the inclusion of two additional publications (“forward snowballing”), resulting in a total of 20 references that qualified for further analysis (see flow chart in Fig. 2).

Fig. 2
figure 2

Flow diagram of included publications (according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses [PRISMA] statement). *No abstract was available for three publications. AGREE Appraisal of Guidelines for Research and Evaluation, CREDES Guidance on Conducting and REporting DElphi Studies

Full size image

3.2 Assessment of Reporting Quality

The 20 publications initially included were analysed for their quality according to the specifications provided by the corresponding guidelines for transparent reporting (PRISMA, AGREE and CREDES). Results are shown in Table 2 (detailed results reflecting the checklists of the corresponding guidelines are provided in Tables S1, S2 and S3 of the Electronic Supplementary Material [ESM]) and are also further illustrated in Fig. 3. The assigned evidence level according to the CEBM Levels of Evidence evaluation was based on the elements critical for the quality review. During this whole process, two publications (18, 19) were excluded after joint group discussion, as the provided dose recommendations could not be verified from the data provided. Table 2 illustrates the distribution of the quality of the included publications. Strikingly, the majority of references (n = 8) were assigned to Evidence Level III, which primarily were based on pharmacokinetic studies. Three references were classified as the highest Level of Evidence I, which were all guidelines. Regardless of the age of the publications, the majority of the publications showed minor deficiencies. This was therefore independent from the publication of reporting guidelines over the years such as PRISMA or AGREE. The distribution of deficiencies did not show relevant imbalances between guidelines and reviews. A comparison with Table 3 revealed that five publications [21, 25, 31, 32, 36] with high quality (Evidence Level I or II; no or minor deficiencies) exist. This means a proportion of 28% of all included publications, which were also rather evenly distributed between reviews (n = 2) and guidelines (n = 3). On average, the specificity of these five publications was significantly higher with a proportion of 71% compared to all other publications (62%).

Fig. 3
figure 3

Quality assessment results (according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses [PRISMA] statement)

Full size image
Table 3 Overview of the proportion of specific statements
Full size table

3.3 Characteristics of the Included Publications

The finally identified 18 publications covered a publication period from 1993 to 2019, with a clustering around the year 2014 (Table 2). Eight of the publications were classified as guidelines [9, 10, 20,21,22,23,24,25] and ten were categorised as reviews [11, 26,27,28,29,30,31,32,33,34]. They included a total of 1145 dose recommendations for 481 active substances (see Table 3). An overview of the active substances with the corresponding publication by which they were covered and an assignment of whether the recommendations were specific can be found in Table S4 of the ESM. The most frequently discussed drug classes were psychoanaleptics (N06) and antiepileptics (N03), which both are prominent for use in neuropathic pain. In addition, dose recommendations on analgesics, especially opioids and paracetamol [11, 21, 23, 24, 31, 32], were frequently addressed.

Table 4 shows the classes of active substances for which no dose recommendations were available. This included dermatics and other locally administered agents, drugs used for constipation (ATC level 2 code A06), sex hormones and antiandrogens (G03H), urological drugs (G04), the subgroup of thyreostatics (H03), drugs used for the treatment of osteoporosis (M05), antihelminthics (P02) and finally conventional medicines for common cold (R05). Some of the drugs mentioned in the publications are not listed in the updated 2022 version of the ATC Index [17] anymore, as they have become obsolete in the meantime.

Table 4 Substance groups (ATC Level 2) without coverage by recommendations in the literature
Full size table

3.4 Variations in Dose Recommendations

Of the 1145 dose recommendations for 481 active substances, 706 recommendations (62%) applying to a total of 316 single agents were classified as sufficiently specific to warrant an unambiguous and simple transfer into clinical routine. Thirty-five specific recommendations were given for four entire drug classes [11, 23, 26, 31, 32] and 29 recommendations were embedded in a therapeutic algorithm [23, 31, 34]. In contrast, 439 recommendations applying to 164 active substances were not classified as sufficiently specific (Table 5). As expected, the total number of specific recommendations was higher in the publications classified as guidelines (n = 432) compared with the reviews (n = 274) [see Table 3]. However, considering the total number of recommendations in both types of publication, there were no relevant differences in the proportion of specific recommendations between reviews and guidelines (61% vs 62%) [Table 3].

Table 5 Active substances with purely nonspecific recommendations
Full size table

Table 3 provides a descriptive overview of the number of statements, as well as the number of specific and nonspecific statements per reference. One publication provided recommendations for three active substances that contradicted themselves [21]. Furthermore, authors described a lack of data for 42 substances, so that they were not able to give recommendations on dose adjustments (Table 3, “no data available”). These statements were ultimately defined as nonspecific. In addition, Table S4 of the ESM shows whether the publications provided particular recommendations referring to the presence or the severity of liver cirrhosis or both. Of note, if differences concerning disease severity were considered, all authors applied the Child-Pugh classification. However, only two studies covering ten agents (2% of all 481 active substances) consequently differentiated their recommendations according to the Child-Pugh classification (Table 3), whereas nine references considered the Child-Pugh classification at least for some of their recommendations and seven references did not adjust for disease severity at all.

3.5 Applicability of Dose Recommendations

A direct comparison of different specific dose recommendations available for 98 individual active substances of the different publications revealed that 53 were not consistent or even contradictory. This particularly applied to the class of analgesics. Table 6 shows examples of divergent statements for acetaminophen and morphine, which are also representative for other substances. Analysis of a random sample of different recommendations for the same substance from at least two publications with Evidence Level I or II showed no consistent recommendations, indicating that the divergence was independent from the quality of evidence of the references. Furthermore, analysis of a representative sample of five prominent drugs with specific but inconsistent recommendations revealed different sources of data as basis for the recommendations, explaining the differing conclusions of the authors. For 290 active substances, of which 157 active substances fulfilled the criteria of specificity (33% of all 481 active substances and 50% of the 316 active substances with at least one specific recommendation), there was only a single reference available that provided recommendations precluding a direct comparison.

Table 6 Examples of divergent recommendations from the literature
Full size table

In contrast, for 22 active substances, the available distinct dose recommendations were not only specific but consistent across the references (Table 7). For another 19 active substances, the dose recommendations were consistent with regard to the general direction (i.e. “no dose modification necessary”, ”dose reduction indicated”, “dose interval extension indicated”, “special handling indicated” [e.g. dose adjustment according to laboratory values] or “avoidance of the active substance”) but they differed in the detailed specifications (e.g. in the degree of the dose reduction or in the recommendations for the different Child-Pugh classes) (Table 8).

Table 7 Consensus from the literature and alignment with the SmPCs
Full size table
Table 8 Overview of active substances with consensus on the general direction
Full size table

As a further step, we cross-checked these concordant recommendations with the according SmPCs to find out whether these recommendations were in line with the formal specifications. Table 7 shows that the dose recommendations of six out of 22 active substances were consistent. For eight substances, the recommendations disagreed at least for certain Child-Pugh classes. For four substances, the literature and SmPC did not explicitly agree but either the literature stated that a dose modification was not necessary, which was not explicitly mentioned in the SmPC, or a recommended dose adjustment in the literature was generally confirmed by the SmPC but not substantiated by a specific dose. For two substances, the alignment was not possible as these preparations were not (or no longer) available in any country.

4 Discussion

In this analysis, we were able to identify 18 publications with a certain standard of reporting quality describing dose recommendations of drugs in patients with liver cirrhosis and providing specific recommendations for 22 active substances consistent between different authors, which would allow the direct transfer into clinical practice. Additionally, we were able to show that the dose recommendations of only six of these 22 active substances were consistent with the specifications of the according SmPC.

As only a few suitable publications could be identified based on the defined inclusion and exclusion criteria despite the use of two different search strings, the time period of publication was not restricted. This resulted in literature originating from a time period of 26 years in which the scientific standards have changed significantly. While in the mid-1990s basic pharmacokinetic parameters were primarily considered, new aspects such as the pharmacodynamics or safety data of particular relevant side effects were increasingly added over time. Moreover, although the Child-Pugh score actually has never been validated for the dose adjustment of drugs and lacks variables to quantify drug metabolisation capacity, more recent literature tends to formulate recommendations specifically for each Child-Pugh class while older publications give general recommendations. Consequently, this variable data basis results in different conclusions and differing recommendations for the same drugs, which complicates clinical application. Furthermore, the reporting standards have changed over time. Therefore, we introduced a quality assessment according to the applicable reporting guidelines to warrant comparable quality of the data as the basis for the dose recommendations. Additionally, the evidence grade was analysed according to the CEBM Levels of Evidence.

The 18 publications we finally identified covered 706 specific dose recommendations for 316 individual active substances. We further determined that with 706 of 1145, only 62% of the recommendations were sufficiently specific with unambiguous statements of doses or rules for certain clinical constellations enabling a simple clinical application contributing to the patients’ safety. Interestingly, there were no relevant differences in the specificity of the recommendations between the literature classified as reviews and guidelines, although generally guidelines would be expected to provide more specific clinical guidance than reviews. However, in absolute terms, the guidelines group itself provided far more specific recommendations than the reviews (432 vs 274).

Furthermore, we identified 164 active substances for which recommendations were not sufficiently specific. In addition, after matching the active substances with their corresponding therapeutic subgroup according to level 2 of the World Health Organization ATC/DDD Index (Version 2022) [17], we identified whole groups of drugs, for which no recommendations were available (Table 4). This primarily included dermatics and other locally administered agents, which are not expected to have systemic effects modulated by liver function. However, several active substances that might be relevant for patients with cirrhosis also lack dose recommendations, for instance, the group of antiandrogens for which not only hepatic metabolism is important [35], but also hepatotoxic events are well described.

Overall, our results demonstrate that the selection of drugs and potentially required dose modifications in patients with cirrhosis on the basis of the available evidence remains challenging, as—in contrast to patients with renal impairment—many recommendations for the dosing of drugs are not sufficiently specific or even contradictory. This remains a clinically relevant problem as a correct dose warrants the efficacy and safety of a drug: on the one hand, a drug administered at a too high dose might lead to increased and potentially toxic side effects, on the other hand, a drug administered at a too low dose might lead to insufficient pharmacological effects. Unfortunately, the calculation of the appropriate dose in patients with liver cirrhosis depends on several different patient-related factors, such as increased body fluid, altered electrolytes, decreased serum albumin and other proteins, decreased muscle mass, portosystemic shunts, and impaired metabolisation capacity of the liver with a decreased first-pass effect [36] and drug-related factors such as the proportion of hepatic extraction or bioavailability, metabolisation pathways, protein binding and the route of administration. Except for a few particular substances, it has even not been shown if therapeutic drug monitoring can be used as a surrogate parameter to predict the effects of a drug sufficiently [28].

Possible limitations of this study include: (1) our literature search was based on a robust search strategy and was supplemented by a “snowballing” process. However, relevant articles may not have been included if ambiguous terms, such as “liver disease” or “liver damage” were used instead of the specific term “liver cirrhosis”. Furthermore, the explicit designation of the disease was assumed to be essential for the identification of literature relevant to this particular population. (2) We specifically searched for (systematic) reviews, guideline articles or consensus procedures. Therefore, original articles were only indirectly part of the evidence for this study. Furthermore, it was not the intention of this study to make its own recommendations. (3) As only publications in English were considered, one French publication was not considered. (4) Our search focused on published reviews and guidelines that were indexed in PubMed/Embase. Therefore, the content of the renowned and continuously updated Dutch database was not taken into consideration [9, 25]. However, the active substances listed in the publications and the supplements of the original work were included. (5) Specific therapeutics, such as nutritional supplements (vitamins, electrolytes and minerals) or medical devices were excluded from the search (see inclusion and exclusion criteria) as dose modifications are usually not relevant for their administration. (6) A differentiation according to the severity of cirrhosis was only available for a small proportion of recommendations. Although it seems obvious that the severity of liver cirrhosis might affect the pharmacokinetics and pharmacodynamics of several drugs, this was not reflected by the available data reviewed.

In the future, more robust and methodologically sound clinical studies in patients with liver cirrhosis of all stages are needed. Furthermore, industry sponsors should be obliged to publish the results of trials in patients with liver cirrhosis if required for the authorisation of new drugs. These studies should also help to close the identified recommendation gaps of many relevant substances and also solve identified discrepancies between available scientific data and information provided by the SmPC. As a prerequisite, a standardisation of parameters that should be considered as a basis for a uniform assessment and the formulation of dose recommendations should be established. Furthermore, it should be clarified, for example, by validation studies, whether the application of the Child-Pugh score is an appropriate classification to actually contribute to safer use of drugs in cirrhotic patients.

In the short term, the conduct of a Delphi survey [37] engaging a board of hepatologists and drug experts such as pharmacists and clinical pharmacologists could be an option to evaluate the already available data and agree on standardised recommendations. To our knowledge, such an approach has not been performed yet. However, expert panels have been conducted in a slightly different setting, in which safety levels and subsequent dose recommendations were developed based on the safety and pharmacokinetics of those drugs [9]. A subsequent clinical validation study to analyse the effectiveness of the adopted recommendations for dose adjustment would be desirable.

5 Conclusions

We identified 18 publications that provided recommendations on drug selection and dose adjustment in patients with liver cirrhosis by a systematic literature and subsequent quality review. Only a minority of 22 recommendations were specific and consistent and could be validated clinically. For 164 drugs, no specific recommendations for patients with liver cirrhosis were available. This lack of consistent recommendations could at least partly be overcome by harmonisation through an expert consensus process.