The study cohort comprised of all patients presenting to the Emergency Department (ED) or inpatient units of Weill Cornell Medicine-affiliated Hamad General Hospital (HMC) with suspected adverse drug reactions (ADR) as part of an ongoing prospective study cohort study (Fig. 1). For this sub-study, two reviewers considered patients >65 years from this cohort who were first adjudicated to have “probable” or “highly probable” DILI scores by the updated Roussel Uclaf Causality Assessment method (RUCAM) [10]. This is a universally acceptable, internationally validated structured tool that allocates specific scores to designated patient-drug interaction variables resulting in a quantitative grading of causality of DILI or herb induced liver injury (HILI) adjudication tool. The final score is interpreted thus: Drug-DILI pairs with scores of “0” or <0 indicate that the drug is “excluded” as a cause of DILI or HILI; 1 to 2 indicates that DILI/HILI is “unlikely”; 3 to 5 indicates “possible”; 6 to 8 “probable”; and >8, “highly probable”. We accepted and included DILI-drug pairs that are classed as “probable” or “highly probable” on the updated RUCAM scale. We subsequently utilized the updated scale tailored for the determination of hepatocellular injury for cases with R-ratio (described under case definitions below) of >5; and that specific for the adjudication of cholestatic or mixed injury for cases with R-ratio <2.5, or between 2.5-5. We then utilized the LAAT (Fig. 2), and modified Hallas tools to determine the potential avoidability of these DILI-drug pairs.
We abstracted the following demographic and clinical parameters of each patient (from an online patient data repository to a Microsoft excel database. These include ethnicity, sex, age, current medication, drug allergies, co-morbidities, index DILI-drug pair, and implicated drug. Other parameters include the dates of the following variables: ADR event, drug commenced, ADR drug stopped, and the onset of symptoms/signs. We also abstracted information on the resolution of ADR, the results of any investigations for differential diagnoses of the suspected DILI-drug pair, any documented record of DILI-ADR outcome, as well as if there was any record of any re-challenge. Amongst the range of alternative diagnoses (other than DILI) that we sought clarification include viral/autoimmune hepatitis, non-alcoholic fatty liver disease (NAFLD), liver and biliary ultrasound, as well as if patients have had a liver biopsy.
We excluded cases where there were alternative diagnoses of liver injury (other than the drug) and in cases of deliberate self-harm. We initially piloted the modified Hallas [8] and LAAT [9] tools on 12 randomly selected DILI-dug pairs. This was done for the dual purpose of training the prospective raters, as well as to improve the familiarity of the raters with these tools. We utilized two rating pairs (two clinical pharmacologists, and two general physicians) to rate the modified Hallas and the LAAT tools with the view to ascertain avoidability of DILI-drug phenotype pairs. In the unlikely event where raters require additional information for case ascertainment and causality determination, they were advised to access publicly available portals and databases of therapeutics information {such as Summary of Product Characteristics (SmPC), the European Medicines Agency (EMA); Food and Drug Administration (FDA); Medicines and Healthcare Products Regulatory Agency (MHRA)}. We reported DILI-drug outcomes measures as one of four-point ordinal scale i.e. ‘definitely avoidable’, ‘possibly avoidable’, ‘not avoidable’ and ‘unassessable’. We submitted and obtained ethical approval for this study from the independent review board of the Medical Research Centre (HMC) (Certificate number MRC-01-18-162).
Statistical analyses
Continuous variables were expressed as means (±SD), or median (inter-quartile range) depending on distribution. Categorical variables were presented as numbers (percentages). We estimated the agreement proportions, Fleiss pairwise kappa (k), Krippendorf’s pairwise kappa (as appropriate), and intraclass correlation coefficients between the 38 DILI-ADR pairs.
DILI-ADR avoidability outcomes were classified as categorical variables, with their resulting pairwise interrater agreement proportions, Krippendorf’s kappa statistics with 95% confidence intervals (CI), and intraclass correlation coefficients (ICC). We estimated and compared the exact pairwise agreement and the disagreement between raters in order to determine agreement proportions across multiple raters. All statistical analyses were carried out with GraphPad Prism (version 8.00 for Windows, 2019 GraphPad Software, La Jolla California USA).
Case definitions
-
We define extreme agreement (EA) between a pair of raters where a particular DILI-ADR pair was scored to the same outcome [11]
-
Where DILI-ADR pairs are scored to discordant outcomes by rating pairs; i.e. where one rater scores a DILI-ADR drug pair outcome as “unassessable”, with the other rating pair scoring it as any of the other possible ordinal outcomes (“not avoidable”, “possibly avoidable”, or “definitely avoidable”)
-
We accept Kappa values of ≤ 0.20, 0.21–0.40, 0.41–0.60, 0.61–0.80, and 0.81–1 as representing a slight, fair, moderate, substantial, and almost perfect agreement, respectively [11]
-
DILI was defined as alanine aminotransferase (ALT) or aspartate aminotransferase (AST) levels greater than 5 × the upper limit of normal (ULN) on two consecutive occasions (at least two weeks apart) and/or ALP levels greater than 2 × the ULN on two consecutive (at least two weeks apart) occasions [10]
-
R-ratio: This index assists in the classification of liver injury into potentially hepatocellular, obstructive, and mixed categories. Itis calculated by dividing multiples of the upper limit of the normal ranges of alanine aminotransferase (ALT) by the alkaline phosphatase (ALP). Preferably, values used should be from blood samples that were drawn from the same day (or no more than 48 hours apart). Cases with values of greater than 5, less than 2.5, and between 2.5 to 5 are classified as hepatocellular, obstructive, and mixed Liver injury respectively. It is an initial first step in the estimation of RUCAM scores.
