The details of the identification of relevant evidence on dosing recommendations for patients with liver cirrhosis via a systematic literature review, including the search strategy, screening workflow, and inclusion and exclusion criteria, have recently been published [10]. In brief, structured searches of PubMed and Embase were conducted to identify peer-reviewed publications providing explicit dosing recommendations for patients with liver cirrhosis (Textbox 1). Only publications offering specific and clinically actionable guidance beyond the SmPC were considered eligible. This approach focused on already formulated dosing guidance rather than on isolated primary pharmacokinetic studies to ensure comparability across substances and alignment with real-world prescribing practice.

Reporting quality was evaluated to determine whether the derivation of dosing recommendations was sufficiently transparent and reproducible. Evidence levels were assigned according to the Oxford Centre for Evidence-Based Medicine framework [11]. Reporting checklists (PRISMA for systematic reviews [12], AGREE [Appraisal of Guidelines for Research and Evaluation] for guidelines [13], and CREDES for Delphi procedures [14]) were applied to assess transparency and reproducibility of reporting rather than intrinsic scientific validity or clinical relevance. Publications were excluded only in cases of major reporting deficiencies that prevented understanding or reproduction of how dose recommendations were derived. References with minor or moderate reporting limitations were retained, and their reporting shortcomings were documented with the assigned evidence level [10].

From the included publications, only recommendations allowing for direct clinical application (eg, clear dose adjustments, interval modifications, avoidance statements, or specification of no adjustment required) were considered. Drug classes requiring highly individualized and indication-specific dosing decisions (eg, cytostatics) were excluded as treatment protocols for these agents are typically governed by specialized oncology guidelines and do not permit generalized dose recommendations across patient groups. The extracted recommendations were structured according to active substance and, where available, stratified using the Child-Pugh classification [15,16] to enable direct comparison and consensus validation. Where multiple publications provided differing recommendations for the same active substance, statements were compared to assess consistency. The resulting structured compilation of dosing recommendations served as the basis for the subsequent Delphi consensus process. A detailed description of the comparative methodology is available in the original systematic review [10].

As the systematic literature search was carried out without restrictions on the publication date, the identified active substances will be cross-checked for relevance against the stock list of the pharmacy of University Hospital Bonn, including all drugs and medical devices used by the wards of the Department of Internal Medicine I. Active substances not included in the pharmacy stock list of this hospital of maximum care will be excluded from further processing as their absence indicates that they are not used regularly in clinical practice. However, before the medications are removed, their clinical relevance will be checked again by 2 pharmacists.

For feasibility reasons, and to focus on active substances particularly relevant in cirrhosis, a preselection will be conducted through an internal preassessment procedure to rate (1) the relevance of dose adjustments of each active substance considering the severity of potential ADRs mentioned in the literature, (2) the frequency of prescription particularly in patients with cirrhosis, and (3) the expected benefit of dose adjustment to avoid possible side effects. The expert panel will consist of 2 hepatologists and 1 clinical pharmacologist. Each expert will have specific expertise in the treatment of patients with liver cirrhosis. The following data from the selected literature will be the basis for the evaluation: (1) the authors’ recommendations and their justification; (2) further information from the references on hepatotoxicity and ADRs; (3) information from the corresponding SmPC on precautions, hepatotoxicity, and ADRs; and (4) information from the LiverTox database, assigned to category A, B, C, D, or E* in the Categorization of the Likelihood of Drug-Induced Liver Injury. For analysis of the assessments, the 3 different criteria (clinical relevance of ADRs, frequency of use, and expected benefit of dose adjustment) will be weighted in a ratio of 3:1:2. Considering the point scale from 0 to 3 provided to the experts, a maximum of 18 points per active substance can be achieved. Only the most relevant active substances rated, reaching a predetermined minimum score of 13.5 points, will be included for further processing. The cutoff is pragmatically predefined to focus on substances with the highest expected clinical impact and to improve the feasibility of the consensus evaluation.

The Delphi technique is a well-established method for the evaluation and assessment of controversial issues by experts, with the aim to produce reasonably reliable consensus-validated statements [14,17]. A Delphi procedure will be conducted that meets the three basic characteristics: (1) anonymity (the 2-round consensus process will be conducted anonymously; the list of participants will only be disclosed to the experts after the evaluation has been completed), (2) iteration (2 interlinked sequential rounds will be carried out), and (3) controlled feedback (in the second round, participants will receive both the consensus results from the first round and their own initial responses, allowing them to reconsider and, if necessary, adjust their assessments; in this way, convergence of evaluations from the first to the second round can be achieved). Unlike the classic Delphi approach, additional information will be provided to the experts in the second round, which they can incorporate into their assessments. This approach allows experts to reconcile their initial clinical judgments with available evidence and formal guidance.

A total of 12 experts with particular expertise in drug safety in hospitalized patients with cirrhosis and involved in drug decision-making processes in hospitals were recruited: 6 (50%) hepatologists, 3 (25%) clinical pharmacologists, and 3 (25%) clinical pharmacists. The assessment sheets will undergo pretesting in 2 rounds by a pharmacist, a clinical pharmacologist, and 2 hepatologists to identify any issues related to understanding and processing of the sheet, resulting in the final version.

According to the rules of the Association of the Scientific Medical Societies for structured consensus in Germany, a recommendation will be considered accepted if at least 75% of the voting experts agree; Table 1 provides a concise overview of the agreement thresholds and decision categories applied. In contrast, divergence will be assumed if (1) no majority consensus is reached, (2) a majority is achieved but at least one-third of the votes indicate a conflicting recommendation (eg, standard dose vs avoidance), and (3) the majority rejects the evidence but no consensus on an alternative approach can be reached (this item only applies to the second round).

The first round is intended to capture the usual prescribing practice concerning the choice of an appropriate active substance (drug selection) within a therapeutic class and dose adjustment of a selected drug. In this design, structured external evidence is not provided in the first round to capture routine expert prescribing behavior prior to evidence-informed reassessment in the second round. In this context, the experts will be asked about their common practice with the active substances in an average hospitalized patient with liver cirrhosis who does not have additional comorbidities that are not typically found in patients with cirrhosis. The experts will have the options to select, for each grade of the Child-Pugh classification, whether they usually (1) prescribe the unchanged dose as they do to patients without liver disease (no dose adjustment), (2) make a dose adjustment, or (3) avoid the drug completely. If they select a dose adjustment, the experts will be required to provide the type of dose adjustment (eg, dose reduction, prolongation of the dosing interval, or dose adaptation based on the results of therapeutic drug monitoring [TDM]). A comment field will be provided for each active substance, allowing experts to describe any limitations and specific considerations that appear relevant to them (for the standardized response options used to capture routine prescribing practice, see Multimedia Appendix 1). This assessment yields a valid and unbiased baseline, is itself an objective of the study, and will provide clinically meaningful data on existing prescribing behavior.

After analysis of the first round, the experts will be asked to review the specific dose recommendations from the systematic literature search and select the recommendations they agree with. Additionally, they will be provided with (1) the authors’ justifications for their recommendations; (2) data from the LiverTox database; and (3) information from the corresponding SmPC on precautions, hepatotoxicity, and ADRs. If the experts completely reject the evidence, they will have the opportunity to enter their own alternative suggestions in a designated comment field. The second round is intended to allow for the analysis of differences between baseline routine clinical judgment and a subsequent evidence-informed reassessment in which experts can critically evaluate and either adopt or explicitly reject the published recommendations. Items for which first-round ratings diverge widely are of particular interest in this round as the provision of structured evidence may facilitate convergence. The design also enables meaningful analysis of substances where first-round consensus was already high to assess whether evidence confirms or challenges prevailing practice.

Of note, the provided data have the same content and structure as in the internal preassessment. For certain agents, the literature provides additional information, such as considerations for comorbidities, specific situations requiring further dose modifications, starting doses, maximum doses, or alternative agents appropriate in cirrhosis. Experts will be asked to provide a separate assessment on their agreement or disagreement for each Child-Pugh grade. The planned evaluation scales can be found in Multimedia Appendices 2 and 3, which provide the structured assessment forms applied in the second Delphi round. The final recommendation list will include dose recommendations for which consensus is reached.

Finally, the feasibility and acceptability of the adopted recommendations for drug selection and dosing in liver cirrhosis will be evaluated through a prospective, nonrandomized clinical study involving patients with liver cirrhosis with the following objectives: (1) improvement of drug therapy safety by applying the consensus-validated therapy recommendations in patients with liver cirrhosis and (2) demonstration of clinical usefulness of the consensus-validated therapy recommendations for patients with liver cirrhosis. For this purpose, dose recommendations will be provided to the pharmacists responsible for the wards with hepatologic patients. If discrepancies between the prescribed medication and the adopted dose recommendations are identified in a patient meeting the inclusion criteria, pharmacists will propose a dose adjustment; substitution with an alternative active substance; or another strategy mentioned in the dose recommendations, such as TDM or enhanced monitoring for adverse reactions. The acceptance of the recommendations remains at the discretion of the prescribing physician. The clinical study will be conducted in the hepatology department of University Hospital Bonn.

All adult (≥18 years) hospitalized patients with clinically documented liver cirrhosis of any origin and severity, except for pediatric patients, pregnant and breastfeeding women, and patients after liver transplantation, will be included. Furthermore, patients must be hospitalized for at least 24 hours and must be seen at least once by the pharmacist responsible for the ward.

The primary end point of the study will be the comparison of the rate of deviations from the prescribed medications before and after the implementation of the consensus-validated dose recommendations. Specifically, the rate of deviations over 9 months using the recommendations (cohort B) will be compared with the rate of deviations over 9 months before their implementation (cohort A), first identified at the prescription level and subsequently aggregated at the patient level. A deviation will be defined as any prescription of a drug covered by the recommendations that does not match the corresponding recommendation for drug selection or dose adjustment, assessed at the patient level during the hospital stay. Prescriptions are assessed at the level of individual active substances. In patients with multiple relevant prescriptions, all prescriptions are checked individually for discrepancies. For the primary end point, deviations are assessed at the patient level as a binary outcome, defined as the presence of at least one deviation during the hospital stay (yes or no). Multiple deviations per patient are aggregated, with the primary outcome defined as the presence of at least one deviation per patient. The total number of MRPs per patient, including both clinically relevant (manifest or preventable) and non–clinically relevant MRPs, will be counted and subsequently dichotomized for categorical analyses. MRPs will be identified based on the first complete medication review at initial patient assessment. Combination products will be broken down into individual active substances if these are available separately; otherwise, they will be evaluated as a single entity. Physicians are given a predefined period of 3 days to address identified MRPs. Unresolved issues will be reassessed and counted thereafter. Recurring deviations after prior resolution will be counted as new events. Deviations from recommendations are treated as a process measure of prescribing behavior, whereas secondary end points are intended to characterize feasibility, acceptance, and safety-related process signals associated with implementation of the recommendations and, therefore, are interpreted as exploratory. These may include the acceptance and rejection rates of pharmacists’ recommendations, particularly for drugs with hepatotoxic properties; frequency of typical side effects (classified as “very frequent” or “frequent” according to the SmPC) of the prescribed drugs covered by the recommendations; deviations from recommended serum levels for drugs monitored via TDM; emergency readmission rates within 30 days after discharge; and overall hospital readmission rates. Furthermore, all consensus-validated recommendations rejected by the prescribing physician will be further analyzed with regard to the corresponding reasons.

Currently, there is no literature available regarding inadequate dosing of drugs in hospitalized patients with liver cirrhosis. A random sample of historical data from the clinic showed an inappropriate prescribing rate of 35%. It is expected that the recommendations will reduce this rate to 20%. To detect a significant difference between the 2 cohorts using a chi-square test with the usual .05 significance level (2 sided) and 80% power, 138 patients per group are required, resulting in a total of 276 patients.

The following data will be collected: age, gender, admission type (standard or emergency case), average and actual length of stay, Charlson Comorbidity Index, Child-Pugh score, type of liver cirrhosis, MRPs, severity of MRPs according to the National Coordinating Council for Medication Error Reporting and Prevention [18], ward pharmacists’ recommendations for solution of MRPs, acceptance of recommendations and reasons for rejection of an intervention by physicians, and outcome of the intervention.

The authors will only have access to pseudonymized data, whereas only the ward pharmacists and ward physicians will have access to personal data.

Baseline characteristics will be summarized overall and stratified by cohort. Continuous variables will be presented as means and SDs or, in the case of skewed distributions, as medians and IQRs. Categorical variables will be reported as absolute counts and corresponding percentages. Comparisons between cohorts will be performed using appropriate statistical tests, including the t test (2-sided) or Wilcoxon rank-sum test for continuous variables and the chi-square test or Fisher exact test for categorical variables, as applicable. These analyses are intended to identify potential baseline imbalances between cohorts.

The primary objective is to assess whether implementation of the consensus-validated recommendations is associated with a lower rate of deviations from the recommendations regarding drug selection and dose adjustment. The primary comparison of deviations as defined in the Primary and Secondary End Points section between cohort A (before implementation) and cohort B (after implementation) will be performed using a chi-square test for the dichotomized end point; effect estimates will be reported as odds ratios with corresponding 95% CIs. Additionally, the raw number of deviations (counted as MRPs at the patient level) will be compared between the 2 cohorts using a nonparametric Wilcoxon rank-sum test. If substantial baseline differences between the cohorts are observed, adjusted analyses may be performed using regression models (eg, logistic regression for the binary outcome or Poisson or negative binomial regression for count data) including relevant baseline covariates to account for potential confounding. The main analysis will be conducted at the patient level. In a secondary analysis, the evaluation will be repeated at the prescription level; clustered data structures (multiple prescriptions per patient) will be accounted for using appropriate methods. As the primary objective is to compare prescribing behavior between 2 predefined cohorts within a real-world pretest-posttest design, the primary analysis will be conducted unadjusted for other covariates. To address potential bias, baseline characteristics will be described and compared between cohorts to assess potential imbalances. Furthermore, major organizational changes or parallel quality improvement initiatives related to medication safety will be documented and qualitatively considered when interpreting the results. An exploratory sensitivity analysis using logistic regression adjusting for key clinical variables (eg, age, sex, and Child-Pugh class) will be performed to assess robustness with regard to potential residual confounding. Secondary end points will be analyzed descriptively and, where appropriate, using inferential methods suited to the scale and distribution of the corresponding variables. If distributional assumptions for parametric testing are not met, appropriate nonparametric alternatives will be applied. Missing data will be handled via complete-case analysis without imputation; the extent and pattern of missingness will be described. No formal adjustment for multiple testing will be applied for secondary outcomes, and results will be interpreted accordingly. Analyses will be performed using R (version 4.4.2; R Foundation for Statistical Computing [19]) with RStudio (Posit PBC) [20]. Statistical significance will be set at a P value below .05.